Origins of the Different Metal Preferences of Escherichia coli Peptide

Jul 24, 2008 - In PDF, the substrate carbonyl is activated by the chemical step itself, and becomes the fifth coordination partner of zinc only in a l...
0 downloads 0 Views 232KB Size
Download PDF
10280

J. Phys. Chem. B 2008, 112, 10280–10290

Origins of the Different Metal Preferences of Escherichia coli Peptide Deformylase and Bacillus thermoproteolyticus Thermolysin: A Comparative Quantum Mechanical/Molecular Mechanical Study Minghui Dong and Haiyan Liu* Hefei National Laboratory for Physical Sciences at the Microscale, and School of Life Sciences, UniVersity of Science and Technology of China (USTC), Hefei, Anhui, 230027, China ReceiVed: NoVember 26, 2007; ReVised Manuscript ReceiVed: May 15, 2008

The Escherichia coli peptide deformylase (PDF) and Bacillus thermoproteolyticus thermolysin (TLN) are two representative metal-requiring peptidases having remarkably similar active centers but distinctively different metal preferences. Zinc is a competent catalytic cofactor for TLN but not for PDF. Reaction pathways and the associated energetics for both enzymes were determined using combined semiempirical and ab initio quantum mechanical/molecular mechanical modeling, without presuming reaction coordinates. The results confirmed that both enzymes catalyze via the same chemical steps, and reproduced their different preferences for zinc or iron as competent cofactors. Further analyses indicated that different feasibility of the nucleophilic attack step leads to different metal preferences of the two enzymes. In TLN, the substrate is strongly activated and can serve as the fifth coordination ligand of zinc prior to the chemical steps. In PDF, the substrate carbonyl is activated by the chemical step itself, and becomes the fifth coordination partner of zinc only in a later stage of the nucleophilic attack. These leads to a much more difficult nucleophilic attack in PDF than in TLN. Different from some earlier suggestions, zinc has no difficulty in accepting an activated substrate as the fifth ligand to switch from tetra- to penta-coordination in either PDF or TLN. When iron replaces zinc, its stronger interaction with the hydroxide ligand may lead to higher activation barrier in TLN. In PDF, the stronger interactions of iron with ligands allow iron-substrate coordination to take place either before or at a very early stage of the chemical step, leading to effective catalysis. Our calculations also show combined semiempirical and ab initio quantum mechanical modeling can be efficient approaches to explore complicated reaction pathways in enzyme systems. I. Introduction Numerous metalloenzymes catalyze a large number of important biochemical reactions.1 Many of them prefer or disfavor certain divalent metals as their catalytic cofactors. Such metal preferences and the associated mechanisms have been extensively studied using structural analyses and other experimental techniques.2–5 The selective preferences between zinc and other divalent cations (for example, iron) in hydrolases are especially interesting. Zinc has been found to be involved in more than 400 enzymes.6–8 It has fully occupied d orbitals, is not redox active, and can act as a strong Lewis acid.9 Its variable coordination numbers, flexible coordination geometries, and rapid ligand exchange rates allow it to accommodate necessary structural rearrangements in many chemical reactions.1,10,11 These and other distinctive characteristics of zinc ions may play important roles during catalysis. Despite of the wide preferences for zinc, exceptional metallohydrolases have been identified that prefer other divalent cations rather than zinc as their competent cofactors. Among them, the Escherichia coli peptide deformylase (PDF) and a number of other PDFs appear as cases of special interest. These enzymes catalyze the deformylation of nascent peptides in bacteria.10,11 Despite the low activity of their zinc forms, they contain a classical HEXXH zinc-binding motif,6,8 which is also * To whom correspondence should be addressed. Telephone: (+86)5513607451. E-mail: [email protected].

found in many other zinc metalloproteases. The structures of PDFs in complex with various inhibitors11–14 strongly implicated that, as many zinc proteases, these enzymes adopt a mechanism of using a metal-activated water molecule as a nucleophile (Figure 1). However, iron instead of zinc is believed to be the physiological metal cofactor for them.2,5 It has been found that the Zn2+-bound form of E. coli PDF has much lower activity than the Fe2+-bound form, while the Ni2+, Co2+ forms retained the same catalytic activity as the native enzyme.2,5,10,15,16 Different explanations for such an unusual metal preference have been proposed based on a range of experimental and theoretical studies.17–19 In several structural studies,11 only small differences at the active site of the crystal structures of Zn2+-, Fe2+-, and Ni2+-PDFs have been noticeable, including a slightly smaller tetrahedral volume at the zinc metal center.19 This has been interpreted as a tighter binding of Zn2+, which might hinder the metal’s ability to alternate between a tetrahedral groundstate and a hypothetical penta-coordinated transition state.19 Recent studies reported that, for some members of other subfamilies of PDFs, the zinc-bound forms are almost as active as the iron-bound forms.20 While making the situation more complicated, these observations also provided some new clues for a clearer understanding of the mechanisms underlying the different reactivity of the same metal ion in identical chemical reactions but different enzymes. Computational studies can provide insight into enzyme catalysis at the level of electronic structures that are difficult to obtain by experimental means. In a recent work,21 Leopoldini

10.1021/jp711209j CCC: $40.75  2008 American Chemical Society Published on Web 07/24/2008

Metal Preferences of PDF and TLN

J. Phys. Chem. B, Vol. 112, No. 33, 2008 10281

Figure 1. Chemical steps of PDF (or TLN)-catalyzed peptide hydrolysis (M ) Zn2+ or Fe2+).

et al. reported a density functional theory (DFT) study to address the metal ion dependence of PDF activity. They used a cluster model of the active site of the protein, including up to the firstshell metal ligands, and the cluster solvated in a homogeneous continuum. Their results did not show dependence of the hydrolysis activity on different metal ions (Zn2+, Ni2+, and Fe2+), presumably because specific interactions with environments have not been considered. Recently, Xiao and Zhang17 and Wu et al.18 reported independent ab initio quantum mechanical/molecular mechanical (QM/MM) studies on PDF including the protein environments in atomic details. Both works showed that Zn-PDF but not Fe-PDF prefers a tetracoordinated metal center as the ground state, and coordination between the substrate and Fe in the penta-coordinated state significantly reduces the barrier for the nucleophilic attack step as compared with the tetra-coordinated zinc form. These previous studies have focused on the reactivity of different metal ions in one enzyme that does not prefer zinc. For a more complete understanding, appropriate comparisons between the reactivity of the same metal ion in different enzymes utilizing the same chemical mechanism are desirable. The zincpreferring PDFs (for example, LiPDF from Leptospira interrogans22 or AtPDF1A from Arabidopsis thaliana69) would have been the first choices for such comparative studies. However, the information contained in the available structures for these enzymes does not allow the construction of a reliable model of

the active enzyme-substrate (ES) complex, mostly because both LiPDF and AtPDF1A contain a long inserted CD loop. At least in LiPDF, this loop has been observed to be in at least three different conformations: open, closed, and half-open.22 In both the half-open and closed forms, the CD-loop folds into the substrate binding site, indicating high flexibility of the substrate binding site. To build complete ES models from currently available structures of LiPDF or AtPDF1A, we would need to remodel this highly flexible CD loop, whose conformation may be closely coupled with substrate binding, more or less blindly. A much better characterized example of zinc-preferring metallopeptidases is thermolysin (TLN) from Bacillus thermoproteolyticus. TLN catalyzes the hydrolysis of peptide bonds specifically on the amino side of large hydrophobic residues.23–25 It is different from E. coli PDF for its high activity in zincbound form. Other metal ions are less preferred: it has been reported that when replacing Zn2+ by Fe2+, TLN retains an activity of about 60%,3 while the replaced forms of some other transition metals show no catalytic activity even at high concentrations of metal ions.2,3 Despite their different metal preferences, it has been well documented that E. coli PDF and TLN have extraordinary similar active sites, including positioning and relative orientations of the zinc-binding motif and the catalytic generic acid, and share the same catalytic mechanism11,26 for peptide hydrolysis proposed long ago.27 The structural comparisons of the

10282 J. Phys. Chem. B, Vol. 112, No. 33, 2008 active sites of PDF and TLN28 and a previous theoretical study on TLN29,30 may have provided some clues to understand the different reactivity of zinc in the two enzymes. For example, in TLN, the residue His231 at the active site could be positively charged and polarize the substrate carbonyl oxygen. Similar charged group-substrate interactions have been observed in other zinc-preferring metalloenzymes,9,31 while necessary constituents for such interactions seem to be absent in E. coli PDF. Theoretical studies can contribute critically to elucidate whether these or other factors (e.g., differences between the coordination environments of the metal center) are the inherent origins of the different reactivity associated with the same metal ion in different enzymes. Considering the complexity and necessary approximations in any theoretical models of enzyme systems, comparative modeling of different enzymes at the same levels of theory and with the same computational techniques is necessary for robust and conclusive analyses of results for different systems. In this work, we use QM/MM models32–39 to perform such a comparative study of E. coli PDF and B. thermoproteolyticus TLN. The main focus is on the origin of different reactivity of the same metal ion in different enzymes, provided that the enzymes use identical chemical steps. QM/MM allows for accurate modeling of the chemical reactions at the enzyme active sites, with the effects of protein environments properly included.32,33,37–39 To avoid presuming the specific chemical steps in the reactions, we first use the nudged elastic band (NEB) method40,41 adapted for enzymatic systems42 and semiempirical QM/MM to determine minimum energy paths (MEPs), then refine the paths using ab initio QM/MM. This is different from most previous ab initio QM/MM studies on enzyme systems, which relied on intuitively defined reaction coordinates to drive the optimizations. By leaving out the specification of chemical steps from input, our results can testify to what extent PDF and TLN would use identical chemical reaction steps for catalysis, as indicated by prior structural comparisons. Methodologically, these two enzymes also provide good systems to investigate whether the above combined usage of semiempirical and ab initio QM can provide an efficient alternate route for exploring complicated enzymatic reaction paths. For each enzyme system, we construct models of the reactant and product, and first determine an MEP using the NEB method. On the basis of the NEB results, we define reaction coordinates comprising key interatomic distances, and determine ab initio QM/MM potential energy profiles43–45 along the reaction coordinates using restrained optimization. This procedure is applied to the zinc forms of both enzymes. Similar ab initio QM/MM potential energy profiles for the iron forms are obtained by replacing the metal ion from zinc to iron in the starting structures of the restrained optimizations and reoptimizing the paths. II. Materials and Methods 1. Starting Structures of the Enzyme-Substrate-Solvent Systems. Each model contains three parts: the enzyme, the substrate molecule, and the solvating water molecules. For E. coli PDF, the starting model of the ES complex has been constructed from a 2.9 Å resolution crystal structure (PDB ID: 1BSK) of E. coli PDF in complex with an inhibitor (S)-2-O(H-phosphonoxy)-L-caproyl-L-leucyl-p-nitroanilide.14 We replaced the inhibitor by the substrate formyl-Met-Gly-Gly-CH3 molecule and a water molecule using Insight II46 by deleting and replacing heavy atoms of the inhibitor. For TLN, we started from a 1.7 Å resolution crystal structure (PDB ID: 4TMN) of

Dong and Liu TLN in complex with a phosphonamidate.47 We similarly replaced the inhibitor by a substrate formyl-Phe-Gly-Gly-CH3 molecule using Insight II.46 Hydrogen atoms were added to the structures with the TINKER package.48 The ES systems were solvated with a solvent water sphere, which is centered on the active site (the metal ion) with the solvent boundary extending at least 10 Å away from any solute atoms. Water molecules overlapped with solute heavy atoms (water oxygen-solute atom distance < 2.3 Å) have been deleted. 2. QM/MM Models. The enzymatic systems have been described by combined QM/MM methods.32,33,36,38,49 Our QM subsystem for PDF consisted of the substrate molecule, the metal ion and its ligands (side chains of Cys90, His132, and His136), Glu133 side chain, Gln50 side chain, and a water molecule, with the Ca-Cβ bonds of the QM protein residues treated as pseudobonds.43,50,51 The resulting QM subsystem has 84 atoms. For TLN, the QM subsystem contained the substrate, the metal ion and its ligands (side chains of His142, His146, and Glu166), Glu143 side chain, and the nucleophilic water molecule, in total 87 atoms connected to the MM subsystem by pseudobonds. The rest of the systems including explicit solvent molecules form the respective MM parts. In our semiempirical QM/MM model, the QM subsystems were described by the AM152 model, and the MM subsystems by the AMBER force field.53 In our ab initio QM/MM model, the QM parts were described by the B3LYP/6-31G* model,54–56 and the MM parts were again described by the AMBER force field.53 3. Equilibration of the Initial Structures. The initial structures have been optimized and equilibrated by restrained optimizations followed by molecular dynamics (MD) simulations under the semiempirical QM/MM model. Initial relaxation of atoms introduced by model building was achieved by minimization until the root-mean-square (rms) energy gradient fell below 1.0 kcal mol-1 Å-1, with heavy atoms present in the crystal structures restrained with a position restraining force constant of 60.0 kcal mol-1 Å-2. Two hydrogen bonds (the hydrogen bonds between the carboxyl oxygen atom of the substrate and the backbone NH of Leu91, and between the carbonyl oxygen of the substrate and the Gln50) important to the correct orientations of the substrate in the active site were also restrained by harmonic restraining potentials on the NHLeu-Osub and NHGln-Osub distances with minimum values in the initial crystal structure-derived model (1.901 Å and 2.180 Å, respectively). With the same restraints, including the hydrogen-bond restraints, further relaxation was achieved by 30 ps QM/MM MD simulations. Using the SHAKE algorithm57 to constrain all bond lengths, all MD simulations in this study were carried out with a time step of 2 fs. The temperature was maintained at 300 K using the weak coupling method,58 with a coupling time of 0.1 ps. Then the shell of solvating water molecules was equilibrated by a 200 ps MD simulation, in which the protein and substrate atoms were fixed. After that the water molecules were optimized until the rms gradient was below 1.0 kcal mol-1 Å-1. The partial charges of charged residues at the surface of the protein are then down-scaled according to a set of Poisson-Boltzmann (PB) calculations to mimic the screening effects of bulk solvent.59,60 The computations of the scaling factors were carried out on the solvent-equilibrated system, including a shell of explicit solvent molecules in our calculations, and only those charged residues at least 18 Å away from the geometric center of the QM part were scaled. In later productive calculations, these charge scaling factors for surface charged residues were applied; atoms within a sphere of 18 Å around the geometric center of the QM part were considered as

Metal Preferences of PDF and TLN active, and the positions of atoms outside of this sphere were fixed. This approach may reduce the computational cost and leave out possible unrealistic fluctuations of boundary atoms due to finite size effects.42 The active sphere was further optimized (until the rms energy gradient below 1.0 kcal mol-1 Å-1) and equilibrated with the position restraining force constant on nonhydrogen atoms present in the initial structures gradually reduced from 60.0 to 40.0, 20.0, and 0.0 kcal mol-1 Å-2 in a series of 30 ps MD simulations. Finally, free optimizations were preformed until the maximum gradient on any active atom fell below 0.1 kcal mol-1 Å-1 without any position or hydrogen bond restraints. To obtain models of the hydrolyzed products, the covalent structure of the QM parts of the above reactant structures was initially modified using the Gaussian 03 package:61 the bond lengths of newly formed (Ow-Csub, Hw-Nsub) and broken bonds (Csub-Nsub) adjusted to target values(from 2.549 to 1.324 Å, from 3.172 to 1.061 Å, and from 1.341 and 2.600 Å, respectively). Throughout our discussions, we will use Ow to refer to the oxygen, Hw for the transferred proton of the nucleophilic water, M2+ for the metal ion, and Csub, Osub, and Nsub for the respective atoms of the scissile peptide bond of the substrate. The bond length modification was performed using the redundant internal coordinate62 modification utility of the Gaussian program, which minimizes displacements in a redundant internal coordinate space while modifying the adjusted distances. Initial structures thus obtained were optimized with gradually weakening position restraints on all unmodified atoms, in which the restraining force constant was decreased from 1000.0 to 500.0, 250.0, 125.0, 60.0, 30.0, 15.0, 10.0, and 5.0 kcal mol-1 Å-2. Finally, free optimizations were preformed until the maximum gradient on any active atom fell below 0.1 kcal mol-1 Å-1. 4. Computing MEPs Using NEB and Semiempirical QM/ MM. For PDF, the NEB path involved 60 images. For TLN, the NEB path involved 55 images. Initial structures of the images have been constructed using simple linear interpolations between the optimized Cartesian coordinates of the reactants and products. In the NEB framework, the separations between adjacent images have been computed as the rms displacements of the Cartesian coordinates of only those atoms involved in bond forming/breaking42 (these bonds are Ow-Csub, Hw-Nsub, Ow-Hw, and Csub-Nsub). The paths were optimized with the projected velocity Verlet algorithm in successive stages, in which the positions of environment atoms have been restrained to the respective starting positions or optimized positions of the previous stages using gradually decreasing force constants (from 1000.0 to 500.0, 250.0, 125.0, 60.0, 30.0, 15.0, 10.0, and 5.0 kcal mol-1 Å-2). Finally, restraint-free optimizations were performed until the maximum residual NEB gradients on any active atom fell below 0.1 kcal mol-1 Å-1. 5. QM Subsystem Optimization Using the Ab Initio QM/ MM Model. Ab initio QM/MM calculations have been carried out using Gaussian 0361 and the TINKER programs.48 On the basis of the semiempirical NEB results, we chose as reaction coordinates the following quantities for the four successive reaction steps: RC1 ) R(Ow-Hw), RC2 ) R(Ow-Csub), RC3 ) R(Hw-Nsub), and RC4 ) R(Csub-Nsub). For each NEB image, one of these four distances was chosen as the reaction coordinate to be constrained in the ab initio QM/MM optimization based on the location of the image on the potential energy profiles. For example, if an image was between INT1 and INT2 on the potential energy profile, RC2 would be chosen as the reaction coordinate to be constrained. Starting from each of the NEB

J. Phys. Chem. B, Vol. 112, No. 33, 2008 10283

Figure 2. Semiempirical QM/MM potential energy profiles obtained from NEB calculations on Zn2+-PDF (a) and Zn2+-TLN (b). See Figure 1 for definitions of different states.

optimized images, the QM part has been reoptimized under the ab initio QM/MM Hamiltonian, with the corresponding reaction coordinate fixed to the NEB-optimized value. As the MM-MM interactions as well as the QM/MM van der Waals interactions were the same in the ab initio and semiempirical QM/MM models, and to save computational efforts, the MM parts have been fixed at positions obtained from the NEB QM/MM calculations, and iterative optimization of the ab initio QM and MM parts has not been attempted. For the iron-bound systems, we replaced zinc by iron and reoptimized the QM parts by the ab initio QM/MM model with the same reaction coordinate constraints. On the basis of previous ab initio calculations,21 the high-spin state is selected for Fe2+ in these calculations. III. Results and Discussions 1. Reaction Steps. Pelmenschikov et al.29 proposed a deformylation reaction pathway including the following successive steps: a proton transfer from the water molecule to a glutamate, the nucleophilic attack of the carbonyl carbon of the substrate formyl group by the metal-coordinated hydroxide, the protonation of the peptide nitrogen of the substrate, and the breakage of the peptide C-N bond. Our NEB results are consistent with this pathway, which is depicted in Figure 1. Figure 1 also defines notations used in this paper for the transition states and intermediate species along the pathway. 2. Semiempirical MEPs. In Figure 2, the potential energy profiles for the PDF and the TLN reactions obtained from NEB

10284 J. Phys. Chem. B, Vol. 112, No. 33, 2008

Dong and Liu

Figure 4. Ab initio QM/MM energy profiles for (a) Zn2+- and Fe2+PDFs and (b) Zn2+- and Fe2+-TLNs. Figure 3. The interatomic distances along the semiempirical QM/MM MEPs obtained by NEB for Zn2+-PDF (a) and Zn2+-TLN (b). Positions of different intermediate and transition states defined in Figure 1 are indicated.

optimizations are presented. The changes of key distances along the MEP are shown in Figure 3. These semiempirical QM/MM results show that the two enzymes share the same reaction mechanism as shown in Figure 1, although no reaction coordinate suggested from Figure 1 has been prespecified in the NEB calculations. The transition barriers computed for Zn2+PDF, especially the TS2 and TS3 barriers, are significantly higher than the corresponding barriers for Zn2+-TLN, qualitatively reflecting the inversed preferences for zinc of the two enzymes. Despite this, at the AM1/MM level, relatively high activation barriers were obtained for both Zn2+-PDF and Zn2+TLN, especially for TS3. The intrinsic low accuracy of the semiempirical model, however, does not allow for further relevant discussions of these energy profiles. As the energy profiles in Figure 2 and distance profiles in Figure 3 supported the proposed mechanism, we decided to choose four successively forming or breaking bonds (Ow-Hw, Ow-Csub, Hw-Nsub, and Csub-Nsub for the four reaction steps, respectively) along the MEP to define reaction coordinates to be constrained in subsequent optimizations using the ab initio QM/MM model. Figure 3 shows that these distances change in a sequential order during the successive reaction steps. 3. Ab initio QM/MM Energy Profiles. Figure 4 presents the potential energy profiles by reaction coordinate-constrained ab initio QM/MM optimizations. As the local minima and maxima on the semiempirical and ab initio QM/MM profiles do not exactly coincide, the reaction coordinate values at respective intermediate states have been redefined based on the

corresponding ab initio QM/M energy profiles. For PDFs, INT1 corresponds to the structure optimized from the NEB image with an Ow-Hw distance of 2.000 Å, INT2 corresponds to the image with an Ow-Csub distance of 1.468 Å, and INT3 corresponds to the image with a Hw-Nsub distance of 1.027 Å. For TLNs, INT1 corresponds to the structure optimized from the NEB image with an Ow-Hw distance of 1.984 Å, INT2 corresponds to the image with an Ow-Csub distance of 1.489 Å, and INT3 corresponds to the image with a Hw-Nsub distance of 1.051 Å. In Table 1, the key interatomic distances of different species along the pathways for the two enzymes in different metal forms are compared. Compared with the semiempirical results for the zinc form enzymes, the overall reaction mechanism suggested by the ab initio QM/MM calculations is the same, with the computed barriers for TS2 being significantly higher and those for TS3 being significantly lower. On the ab initio QM/MM surfaces, the TS2 barrier for Zn2+-PDF is ca. 8 kcal/mol higher than that for Fe2+-PDF. For TLN, the TS2 barrier computed for the zinc form is, however, 2 kcal/mol lower than that for the iron form. For other intermediate and transition states, Zn2+-PDF is also always associated with higher energies compared with Fe2+PDF, while Zn2+-TLN is associated with lower energies compared with Fe2+-TLN. Thus the ab initio QM/MM computations not only support that the two enzymes catalyze peptide hydrolysis via the same reaction steps, but also successfully reproduced their different preferences for zinc or iron. 4. Individual Species and Chemical Steps. To obtain insight into the origins of such differences, we discuss in more details the results for different reaction steps in the following paragraphs. (1) Ground States. In ES structures, the catalytic glutamate residues have been assumed to be deprotonated. The metal ions

Metal Preferences of PDF and TLN

J. Phys. Chem. B, Vol. 112, No. 33, 2008 10285

TABLE 1: Interatomic Distances (Angstroms) at Different States in Structures Optimized by Ab Initio QM/MM Zn2+ -PDF

Fe2+ -PDF

Zn2+ -TLN

Fe2+ -TLN

Zn2+ -PDF

Fe2+ -PDF

Zn2+ -TLN

Fe2+ -TLN

Zn2+ -PDF

Fe2+ -PDF

Zn2+ -TLN

Fe2+ -TLN

Ow-M2+ Osub-M2+ Ow-Csub Osub-Csub Csub-Nsub

1.995 4.029 2.831 1.239 1.341

ES 1.983 3.665 2.767 1.240 1.341

2.025 2.448 2.666 1.249 1.336

2.206 2.381 2.440 1.249 1.336

1.972 3.992 2.856 1.240 1.341

TS1 1.933 3.615 2.795 1.239 1.342

1.983 2.470 2.648 1.247 1.337

2.007 2.343 2.415 1.250 1.337

1.921 3.702 2.594 1.240 1.344

INT1 1.923 3.631 2.597 1.240 1.343

1.940 2.490 2.618 1.245 1.340

1.966 2.319 2.390 1.253 1.338

Ow-M2+ Osub-M2+ Ow-Csub Osub-Csub Csub-Nsub

2.194 2.048 1.709 1.327 1.379

TS2 2.237 2.031 1.709 1.333 1.375

2.473 1.972 1.497 1.350 1.425

2.507 1.988 1.497 1.354 1.422

2.632 1.946 1.465 1.362 1.451

INT2 2.435 1.967 1.485 1.365 1.438

2.179 2.035 1.485 1.339 1.485

2.227 2.030 1.478 1.347 1.472

2.232 2.018 1.426 1.362 1.555

TS3 2.231 2.000 1.454 1.357 1.501

2.183 2.056 1.454 1.333 1.523

2.227 2.052 1.449 1.340 1.523

Ow-M2+ Osub-M2+ Ow-Csub Osub-Csub Csub-Nsub

2.274 2.031 1.387 1.321 1.670

INT3 2.201 2.069 1.399 1.327 1.669

2.177 2.096 1.422 1.316 1.626

2.206 2.098 1.417 1.322 1.626

2.380 2.018 1.356 1.286 1.975

TS4 2.351 2.045 1.350 1.290 1.975

2.176 2.109 1.376 1.296 1.833

2.190 2.111 1.382 1.306 1.813

2.466 1.997 1.272 1.276 2.620

EP 2.209 2.040 1.286 1.272 2.614

2.209 2.125 1.286 1.272 2.464

2.251 2.129 1.281 1.272 2.463

are coordinated with protein residues and water. There are some major differences between the ES states of PDF and TLN. For PDF, the oxygen of the substrate carbonyl group was found to not be coordinated with the metal ion, with Osub-M2+ distances of 4.029 Å and 3.665 Å in Zn2+-PDF and Fe2+-PDF, respectively. The carbonyl group of the substrate forms hydrogen bonds with the neutral backbone NH of Leu91 and Gln50. The metal-coordinated water forms hydrogen bonds with Glu133 and Gln50. For TLN, the carbonyl oxygen of the substrate not only coordinates with the metal center with Osub-M2+ distances of 2.448 Å and 2.381 Å in Zn2+-TLN and Fe2+-TLN, respectively, but also forms hydrogen bonds with the charged side chain of His231. The CdO bond of the substrate is strongly polarized. (2) Proton Abstraction from the Nuleophilic Water by a Glutamate. Although on the semiempirical QM/MM surfaces the TS1 barriers are 9 kcal/mol for Zn2+-PDF and 6 kcal/mol for Zn2+-TLN, on the ab initio QM/MM surfaces the TS1 barriers are very low (below 1 kcal/mol). Given the similar pKa values of the metal-bound water (∼6.5) and the carboxyl group of glutamate (∼5), the proton should be readily transferable from the water molecule to the carboxylate in both enzymes. The ab initio results are chemically sound, and the higher TS1 barriers from the semiempirical calculations likely reflect the underestimated stability of the small hydroxide anion by the QM model. Besides the proton transfer, no other significant changes have been observed from ES to INT1. (3) The Nucleophilic Attack Step. The potential energies associated with TS2 imply that Zn2+-PDF should have much lower activity than Fe2+-PDF, and Zn2+-TLN should have similar or slightly higher activity than Fe2+-TLN. To obtain insight into why only in PDF but not in TLN is the zinccatalyzed nucleophilic attack much more difficult, we compared the courses of correlated changes of some key interatomic distances, starting from INT1 to INT2. These included the distances R(Ow-M2+), R(Ow-Csub), R(Osub-M2+), and also the bond length of the carbonyl group R(Osub-Csub). Using R(Ow-Csub) as a reaction coordinate, Figure 5a-c respectively shows how R(Osub-M2+), R(Ow-M2+), and R(Osub-Csub) change during the reaction course. For PDFs, we observe very sharp drops in R(Osub-M2+) when R(Osub-Csub) decreased to around 2.4 Å in Fe2+-PDF and around 2.0 Å in Zn2+-PDF (Figure 5a), accompanied with little changes in the

reaction coordinate R(Ow-Csub). Extra calculations with R(Ow-Csub) fixed at these respective values and R(Osub-M2+) gradually changed from the starting to ending values of the drops were performed. These calculations showed that the potential energy surfaces are indeed almost flat for changing R(Osub-M2+) at the respective R(Ow-Csub) distances. Figure 6 shows how the total QM/MM energies relative to INT1 changes with R(Osub-M2+). Figures 5a and 6 indicates that in Zn2+PDF the coordination between Osub and the metal ion does not take place until the nucleophile is sufficiently close to the carbonyl carbon. There is large energy increase during this early phase of the nucleophilic attack (R(Ow-Csub) from 2.6 to 2.0 Å). In Fe2+-PDF, the coordination between the metal ion and Osub takes place at an earlier stage (at R(Ow-Csub) around 2.4 Å), associated with a significantly lower barrier for the nucleophilic attack. In TLN, the metal ion, and the substrate oxygen are already close to each other (2.448 Å in Zn2+-TLN and 2.381 Å in Fe2+TLN) in INT1 (Figure 5a), and the distances are only slightly further shortened along with the proceeding of the nucleophilic attack. Comparing the covalent bond lengths involving Csub in INT1 of the PDFs and of the TLNs indicate weaker polarization of the substrate in PDFs. In PDFs, the Osub-Csub bond lengths are shorter (1.240 Å) than in TLNs (1.245 Å in Zn2+-TLN and 1.253 Å in Fe2+-TLN (see Table 1). In both Zn2+- and Fe2+-PDFs, abrupt lengthening of this bond takes place exactly at respective Ow-Csub distances at which the Osub-M2+ coordination is formed (Figure 5b). The Csub-Nsub bond lengths in INT1 of Zn2+- and Fe2+-PDFs are longer (1.344 and 1.343 Å, respectively) (see Table 1) as compared with the lengths of 1.340 Å in Zn2+-TLN and of 1.338 Å in Fe2+-TLN, also consistent with weaker substrate polarization in PDFs. We emphasize that we are comparing covalent bond lengths and the differences cannot be large. If we were comparing each bond length separately, not too much meaning should be put into such small differences. The above discussions are meaningful only because consistent changes in different bonds have been observed and can be correlated with changes along the reaction paths. One interesting difference between PDF and TLN is in the locations of TS2. In both Zn2+-PDF and Fe2+-PDF, TS2 is located at an Ow-Csub distance of 1.709 Å, while in TLNs, TS2 is located at an Ow-Csub distance of 1.497 Å (see Table 1).

10286 J. Phys. Chem. B, Vol. 112, No. 33, 2008

Dong and Liu

Figure 6. Relative total ab initio QM/MM energies as functions of the Osub-M2+ distances in different systems during the nucleophilic attack step.

Figure 5. The (a) Osub-M2+, (b) Osub-Csub, and (c) Ow-M2+ distances as functions of the Ow-Csub distance in different systems during the nucleophilic attack step. The QM parts have been optimized using the ab initio QM/MM model.

Changing the metal ion between zinc and iron in either PDF or TLN has at most minor effects on the positions of TS2 in the respective enzymes, although different metal ions results in different barrier heights, especially in PDF. In Fe2+-PDF, the energies of INT2 relative to INT1 are almost lowered by the same amount as TS2 as compared with Zn2+-PDF. Should the Hammond postulate63 apply, lower INT2 in Fe2+-PDF would result in a TS2 with a shorter Ow-Csub distance. A possible explanation for the unchanged TS2 position is that when Fe2+PDF is compared with Zn2+-PDF, besides stronger stabilization of INT2, the overall slope of the potential energy surface, or the rate at which the total energy increases with the Ow-Csub distance changes, from INT1 to TS2 is also reduced because of the earlier coordination of the metal with the substrate (which results in earlier polarization of the substrate). The much later

TS2 in TLNs compared with PDFs may also be explained by the reduced slope from INT1 to TS2 because of the stronger polarization of the substrate in TLN. In all four systems, the coordination between the metal and the nucleophilic water changes in very similar ways as functions of the Ow-Csub distance (Figure 5c), and the metal centers always pass through a penta-coordinated form along the path. In both Zn2+-PDF and Fe2+-PDF, when the metal-Osub coordination is formed at the respective Ow-Csub distances around 2.0 and 2.4 Å, there are moderate increases (0.1 to 0.2 Å) in the Ow-M2+ bond length (Figure 5c), but the Ow-M2+ coordination remains strong. Because the TS2 in the PDFs are earlier, in TS2 states of Zn2+-PDF and Fe2+-PDF the metal ions are still in the penta-coordinated state. In the TLNs the Ow-M2+ distances are significantly longer (>2.4 Å) at TS2 (Table 1), but this is only because that the TS2 in TLNs are later and the Ow-Csub bond has almost completely formed. At this Ow-Csub bond length the Ow-M2+ distances in PDFs are of similar lengths (Figure 5c). After INT2 the Ow-M2+ distances are shortened again, probably because of the further shifting of the abstracted proton (Hw) away from Ow and the accompanying reorganization of the MM environments which have been fixed to the NEB-optimized structures in the ab initio QM/MM optimizations. (4) Proton Transfer from Glutamate to Substrate Nitrogen. On the ab initio QM/MM surfaces obtained by us, this step is associated with a 5-7 kcal/mol barriers in Zn2+- and Fe2+-PDFs and 2-3 kcal/mol barriers in Zn2+- and Fe2+-TLNs (TS3 relative to INT2). That the computed TS3 states are 1.4 kcal/mol (in Fe2+-TLN) to 4.6 kcal/mol (in Fe2+-PDF) higher than TS2 is probably because of that we have started from the pathway optimized by semiempirical QM/MM and frozen the MM part, and the actual TS3 energies relative to TS2 may be lower in these systems. Reasons for this include that the ab initio QM/MM barriers from INT2 to TS3 are significantly lower than the corresponding barriers on the semiempirical QM/MM surfaces (ca. 21 and 17 kcal/mol for Zn2+-PDF and Zn2+-TLN, respectively), and that previously reported ab initio QM/MM calculations on only PDF indicated lower TS3 energies relative to TS2.17,18 In the calculations of Xiao and Zhang,17 reorganization of the hydrogen bonds involving the glutamate has been explicitly considered as a reaction coordinate before TS3. We have not tried to introduce such a reaction coordinate or to perform completely relaxed ab initio QM/MM calculations on the entire pathway. One reason is that on our surfaces computed

Metal Preferences of PDF and TLN with constrained MM the differences between TS2 and TS3 are already quite small. Most importantly, our major purpose is to obtain insights into origins of the different reactivity of Zn2+ and Fe2+ in different enzymes, not to reproduce the ab initio QM/MM results on Zn2+- and Fe2+- PDF reported before. (6) Breaking of the Csub-Nsub Bond and Formation of the Final Products. The carbon-nitrogen bond of the substrate in INT3 is already quite ionic, with Csub-Nsub distances (1.670 Å in PDFs and 1.626 Å in TLNs) much longer than an ordinary covalent Csub-Nsub bond. The Csub-Nsub bonds are completely broken and change from 1.670 Å (1.626 Å in TLNs) in INT3 to 2.620 Å (2.463 Å in TLNs) at transition states TS4, with computed activation barriers between 1.7 kcal/mol (Fe2+-TLN) to 4.2 kcal/mol (Fe2+-PDF) by ab initio QM/MM in different systems, comparable to the barriers computed by semiempirical model for the zinc forms. With the proton transferred to the substrate amide, TS4 is lower in energy than both TS2 and TS3. In addition, the remaining proton on Ow is automatically transferred to the now deprotonated glutamate accompanying the broken of the Csub-Nsub bond. In the final EP complex, the hydrolyzed formate is coordinated to the metal cation almost in a bidentate manner, except that the Ow-M2+ distances in Zn2+-PDF and Fe2+-PDF are longer than in TLN (Table 1). 5. Comparisons of Results on the PDF Systems with Previously Reported Ab Initio QM/MM Studies. The PDF part of our study differed from earlier ones in numerous details about constructing the starting models and approximating the QM/MM interactions. For example, we have constructed the initial model from PDB structure 1BSK and equilibrated the starting system with a water and a deprotonated Glu133, instead of from the structure 1BS6 and equilibrated with a hydroxide and a protonated Glu133. We have included Gln50 in the QM part instead of the MM part. We have used the pseudobond approach to treat the QM/MM boundary as Xiao and Zhang17 while Wu et al.18 have instead employed the link atom and ONIOM approach. In our ab initio QM/MM modeling we have constrained the MM part to the reaction path obtained using the NEB approach and a semiempirical QM/MM model, and so on. Given all these different approximate treatments of such complex molecular systems, we should not expect exact numerical agreements between results from different studies. Despite these, our results on PDF are in remarkable agreement with the two recently reported studies on PDF in the following key aspects, suggesting the robustness of the QM/MM methodology. First, in all studies the iron form of E. coli PDF has been found to be associated with lower barriers or more active than the zinc form, in agreement with biochemical experiments. Second, all calculations show that in Zn2+-PDF, the metal does not coordinate with Osub before the nucleophilic attack step and the metal center is in a tetra-coordinated instead of pentacoordinated geometry. Third, in Fe2+-PDF, the tetra- and pentacoordination forms can be very close in energy, thus may exist either in INT1 or form at very early stage of the nucleophilic attack step. The calculations of Wu et al.18 showed that the penta-coordinated form could be marginally more stable (by 0.32 kcal/mol) relative to the tetra-coordinated form. The calculation by Xiao and Zhang17 indicated a Fe2+-Osub distance of 2.38 Å before the nucleophilic attack, much longer than the distance for a strong coordination. Although our calculations indicate that Fe2+ is not yet coordinated with the substrate in INT1, the coordination formed at a very early stage of the nucleophilic attack (Figure 5a). Finally, the positions of the transition states for the nucleophilic attack step obtained from different calculations are also in qualitative agreement. Wu et

J. Phys. Chem. B, Vol. 112, No. 33, 2008 10287 al. reported Ow-Csub distances of 1.63 Å for Zn2+-PDF and 1.75 Å for Fe2+-PDF. Xiao and Zhang reported 1.70 Å and 1.80 Å, respectively, for the same distances. We obtained 1.71 Å for both systems (Table 1). The differences between the results of different calculations are mostly not qualitative and have to do with the relative energies between different species occurred in the same chemical steps. The only exception is that for Zn2+-PDF, Wu et al. reported automatic proton transfer from Glu133 to the substrate nitrogen accompanying the nucleophilic attack. For Fe2+-PDF they obtained a barrier of ca. 3.5 kcal/mol for this proton transfer after the nuleophilic attack. Using reaction coordinate driven Xiao and Zhang17 obtained two successive steps for the same proton transfer in both Zn2+ and Fe2+-PDFs, associated with barriers of 1-2 kcal/mol. The first step is the reorganization of the hydrogen bonds and the second the proton transfer. We obtained higher barriers for this proton transfer as well as for the subsequent Csub-Nsub bond breaking step, probably because the MM part has been constrained to the semiempirical QM/ MM NEB results. As stated before we did not try to confirm this by completely relaxing the MM parts, because our main purposes are to obtain insights into the origins of the different ion-selectivity between TLN and PDF, not the exact energetic of all the reaction steps for PDF. The differences between the Zn2+- and Fe2+-PDFs consistently produced by our calculations and previously reported ones strongly suggest that for these purposes we should focus on comparing the nucleophilic attack step in both enzymes. For these comparisons our computational results should be of sufficient accuracy. 6. Insights into Origins of the Different Metal Preferences of E. coli PDF and TLN. We have applied the same levels of theoretical models and computational procedures to the PDF and TLN systems in both the zinc- and iron-bound forms. The computed reaction pathways and associated energetics are not only consistent with previous experimental suggestions that both enzymes catalyze peptide hydrolysis via the same chemical steps, but also reproduced their different preferences of zinc or iron as catalytic cofactors. The above analyses also indicate that such different preferences may mainly correspond to major differences in the energetic of the nucleophilic attack step (from INT1 to INT2 via TS2). Such energetic differences can be further interpreted in terms of detailed structural changes along the reaction pathways. Our results about the differences between Fe2+-PDF and Zn2+-PDF largely reproduced the prior ab inito studies. Thus we will focus on the comparisons between PDF and TLN. Before the nucleophilic attack in PDFs, the tetra-coordinated state is more stable for zinc, and either the penta-coordinated or the tetra-coordinated state is only marginally more stable for iron. In Zn2+-TLN and Fe2+-TLN, the penta-coordinated states should be much more stable than the tetra-coordinated states. In PDF, the metal centers start from tetra-coordination and enter penta-coordination during the nucleophilic attack step (very early for Fe2+-PDF and quite late for Zn2+-PDF), with newly forming Osub-M2+ coordination. In TLN, the metal centers started from penta-coordination, with the Osub-M2+ bond present already in ES and INT1. The dual roles of the metal ion to coordinate and activate the water nucleophile and to activate the substrate carbonyl have been suggested for several extensively studied zinc hydrolases (e.g., carboxypeptidase A).29,64 Our results also suggested that coordination between the metal and Osub may be of key importance for reducing the slope of the potential energy surface for the nucleophilic attack. First of all, it may significantly

10288 J. Phys. Chem. B, Vol. 112, No. 33, 2008

Dong and Liu

TABLE 2: Wiberg Bond Indices (WBI) at INT1, TS2, and INT2 States Zn2+-PDF

Fe2+-PDF

WBIOw-M2+ WBIOsub-M2+ WBIOw-Csub WBIOsub-Csub WBICsub-Nsub

0.1995 0.0037 0.0306 1.5868 1.2409

0.2354 0.0048 0.0294 1.5806 1.2500

WBIOw-M2+ WBIOsub-M2+ WBIOw-Csub WBIOsub-Csub WBICsub-Nsub

0.0941 0.1479 0.5229 1.1827 1.1373

0.0904 0.1766 0.5286 1.1589 1.1496

WBIOw-M2+ WBIOsub-M2+ WBIOw-Csub WBIOsub-Csub WBICsub-Nsub

0.0419 0.2038 0.8042 1.0702 0.9745

0.0572 0.2144 0.7760 1.0717 0.9978

Zn2+-TLN

Fe2+-TLN

INT1 0.1814 0.0604 0.0515 1.5501 1.2212

0.1986 0.0774 0.0510 1.5252 1.2331

0.0477 0.2049 0.7383 1.1283 0.9994

0.0544 0.2293 0.7328 1.1197 1.0126

0.0843 0.1713 0.7803 1.1662 0.9092

0.0848 0.1956 0.7933 1.1397 0.9162

TS2

INT2

group in E. coli PDFs at INT1. In the different systems, the bond orders of Osub-Csub, Osub-M2+ and Ow-M2+ vary in a somewhat coordinated way. The INT1 and TS2 of PDFs are associated with higher Osub-Csub and Ow-M2+ bond orders and lower Osub-M2+ bond orders relative to TLNs. In INT2, these relations are reversed. PDFs are associated with lower Osub-Csub and Ow-M2+ bond orders and higher Osub-M2+ bond orders. As has already pointed out in prior studies, the interactions of Zn2+ with the substrate carbonyl and nucleophilic hydroxide are always weaker than those of Fe2+. For iron, the difference binding environments of the substrate between PDF and TLN may have much smaller effects on the substrate-metal coordinates than for zinc. Thus Fe2+-PDF remains active as substrate-metal coordination can still take place before the nucleophilic attack. On the other hand, the stronger interactions between the iron and the hydroxide in Fe2+-TLN relative to Zn2+-TLN, however, may reduce the nucleophilicity of the hydroxide. Thus Fe2+-TLN is computed to be less active than Zn2+-TLN. IV. Conclusions

stabilize the developing negative charge on the carbonyl oxygen. Another effect that can be as important is that it can weaken the metal-nucleophile coordination, and thus increase the nucleophilicity of the attacking hydroxide. Figure 5c shows that, during the nucleophilic attack step, the Ow-M2+ distances increase gradually in all systems, suggesting that weakening the metal-hydroxide bond may help to reduce the difficulty of the nucleophilic attack. Figure 5c also shows that, in both Zn2+and Fe2+-PDFs, obvious lengthening of the Ow-M2+ bonds is observed accompanying the formation of the Osub-M2+ coordination. Thus the nucleophilic attack step in TLNs and in Fe2+PDF is much easier than in Zn2+-PDF because of the initially or earlier-established Osub-M2+ coordination. The different Osub-M2+ coordination in the two enzymes most likely arose indirectly from the different environments of the substrate rather than of the metal ions. We performed ab initio gas-phase cluster calculations on the zinc-catalyzed nucleophilic reactions (results not shown), switching one of the metal ligands from a deprotonated thiol (mimicking PDF) to a carboxylate (mimicking TLN), and this had no significant influence on the activation barriers of the nucleophilic attack step. Although, relative to iron, zinc prefers tetra- over pentacoordination25 in TLN and in many other zinc-enzymes,29,65 and zinc can still be penta-coordinated at ground states. In E. coli PDF, the substrate carbonyl forms hydrogen bonds with the neutral amides of Leu91 backbone and Gln50 side chain. In TLNs, the carbonyl is hydrogen bonded with the positively charged side chain of His231, which may more strongly activate the substrate than the neutral groups in PDF before the nucleophilic attack takes place and better stabilize the intermediate during the nucleophilic attack. The importance of His231 in TLN has been reported before, and a role as a generalacid-base catalyst to activate the substrate has even been suggested for it.66,67 The replacement of His231 by an alanine resulted in a dramatic reduction in catalytic efficiency.28 Polarized significantly already by His231, the substrate carbonyl in TLN can serve as the fifth zinc ligand. In E. coli PDF this carbonyl may have not been sufficiently activated by the neutral environment alone and may not serve as the fifth zinc ligand until substantial activation takes place during the attack of the hydroxide. The Wiberg Bond indices68 of key bonds in different species listed in Table 2 supporting the weaker activation of the carbonyl

We have investigated the catalytic mechanisms of E. coli PDF and B. thermoproteolyticus TLN in both their zinc and iron-bound forms. The two enzymes have been modeled using exactly the same computational procedure so that meaningful comparative analyses could be carried out. Our calculations not only confirmed that the four systems carry out catalysis via the same chemical steps, but also reproduced the different metal ion preferences of the two enzymes. Comparing our results with two previously reported ab initio QM/MM calculations on Zn2+- and Fe2+-PDF indicates that many of the key qualitative features concerning the critical nucleophilic attack steps have been reproduced despite the different details of the computational models and the complexity of the systems. Comparing the results for PDFs and TLNs suggests that their different ion preferences are mainly associated with differences between the two enzymes during the nucleophilic attack step. Considering both a zinc-preferring and an iron-preferring enzyme allowed us to focus on how the same metal ion is associated with the different reactivities of different enzymes that use identical chemical steps. In agreement with prior studies, we also found that direct coordination between the metal and the substrate carbonyl is critical for reducing the difficulty of the nucleophilic attack. Our study indicated that zinc-substrate coordination takes place before the nucleophilic attack in Zn2+-TLN, but at a relatively late stage during the nucleophilic attack in Zn2+-PDF. The substrate carbonyl needs to be activated first to serve as the fifth coordinated ligand for zinc. Substrate activation may have been achieved through polarizing interactions with environment in TLN, but is achieved only after the hydroxide has sufficiently approached the substrate carbonyl in E. coli PDF. Once the substrate has been sufficiently activated, we found little energy cost for the zinc center to switch from the tetra- to the penta-coordinated state. The different substrate activation in TLN and E. coli PDF have much smaller effects on iron, as iron interacts more strongly with its ligand and can coordinate with the substrate at a very early stage, even in E. coli PDF. The above results highlight the important interplay between nonbonded substrate-environment interactions and chemical mechanisms in enzyme catalysis. Given the same chemical steps in PDF and TLN catalysis, it is quite remarkable that the nonbonded

Metal Preferences of PDF and TLN substrate-environment interactions, which would just indirectly tune the finer structure and energetics of the chemical steps, can bring about such dramatically different metal preferences. Recently, PDFs whose zinc forms are almost as active as other zinc hydrolases have been identified. These include A. thaliana PDF1A (AtPDF1A)20,69 and L. interrogans PDF (LiPDF).70–72 AtPDF1A,20 LiPDF,72 and E. coli PDF share the conserved PDF fold. Comparisons of the metal coordination of available structures reveal a perfectly conserved catalytic core, supporting that all PDFs adopt the same catalytic mechanism. When more structural data about the ES complex of the zinc-preferring PDFs are available, it would be interesting to investigate whether the same insight obtained here applies to the comparisons between E. coli PDF and these zincpreferring PDFs. By using the NEB approach, we were able to explore the MEPs without presuming specific chemical steps. The consistent results for two enzymes suggest that the combined semiempirical and ab initio QM/MM approach used here can be an efficient way to explore reaction paths of complicated reaction systems. Acknowledgment. We thank Dr. Yingkai Zhang for helpful discussions and suggestions. Financial support from the Chinese Natural Science Foundation (Grant Numbers 90403120 and 30670485) and from the Chinese Ministry of Science and Technology (Grant Number 2006AA02Z303) are acknowledged. Coordinates of the optimized QM/MM structures are available upon request. References and Notes (1) Frausto da Silva, J. J. R.; Williams, R. J. P. The Biological Chemistry of the Elements. The Inorganic Chemistry of Life, 1st ed.; Oxford University Press: New York, 1991. (2) Ragusa, S.; Blanquet, S.; Meinnel, T. J. Mol. Biol. 1998, 280, 515– 523. (3) Holmquist, B.; Vallee, B. L. J. Biol. Chem. 1974, 249, 4601–4607. (4) Holmquist, B.; Vallee, B. L. Biochemistry 1976, 15, 101–107. (5) Rajagopalan, P. T. R.; Datta, A.; Pei, D. H. Biochemistry 1997, 36, 13910–13918. (6) Vallee, B. L.; Auld, D. S. Biochemistry 1990, 29, 5647–5659. (7) Vallee, B. L.; Auld, D. S. Proc. Natl. Acad. Sci. U.S.A. 1990, 87, 220–224. (8) Lipscomb, W. N.; Strater, N. Chem. ReV. 1996, 96, 2375–2433. (9) Hernick, M.; Fierke, C. A. Arch. Biochem. Biophys. 2005, 433, 71–84. (10) Rajagopalan, P. T. R.; Yu, X. C.; Pei, D. H. J. Am. Chem. Soc. 1997, 119, 12418–12419. (11) Becker, A.; Schlichting, L.; Kabsch, W.; Groche, D.; Schultz, S.; Wagner, A. F. V. Nat. Struct. Biol. 1998, 5, 1053–1058. (12) Durand, D. J.; Green, B. G.; O’Connell, J. F.; Grant, S. K. Arch. Biochem. Biophys. 1999, 367, 297–302. (13) Huntington, K. M.; Yi, T.; Wei, Y. M.; Pei, D. H. Biochemistry 2000, 39, 4543–4551. (14) Hao, B.; Gong, W. M.; Rajagopalan, P. T. R.; Zhou, Y.; Pei, D. H.; Chan, M. K. Biochemistry 1999, 38, 4712–4719. (15) Groche, D.; Becker, A.; Schlichting, I.; Kabasch, W.; Schultz, S.; Wagner, A. F. V. Biochem. Biophys. Res. Commun. 1998, 246, 342–346. (16) Meinnel, T.; Lazennec, C.; Blamquet, S. J. Mol. Biol. 1995, 254, 175–183. (17) Xiao, C.; Zhang, Y. K. J. Phys. Chem. B 2007, 111, 6229–6235. (18) Wu, X. H.; Quan, J. M.; Wu, Y. D. J. Phys. Chem. B 2007, 111, 6236–6244. (19) Jain, R. K.; Hao, B.; Liu, R. P.; Chan, M. K. J. Am. Chem. Soc. 2005, 127, 4558–4559. (20) Serero, A.; Giglione, C.; Meinnel, T. J. Mol. Biol. 2001, 314, 695– 708. (21) Leopoldini, M.; Russo, N.; Toscano, M. J.Phys.Chem. B 2006, 110, 1063–1072. (22) Zhou, Z. C.; Song, X. M.; Gong, W. M. J. Mol. Biol. 2005, 280, 42391–42396. (23) Endo, S. J. Ferment. Technol. 1962, 40, 346–353. (24) Matsubara, H.; Sasaki, R.; Singer, A.; Jukes, T. H. A. Biochem. Biophys 1966, 115, 324. (25) Morihara, K.; Tsuzuki, H. Eur. J. Biochem. 1970, 15, 374.

J. Phys. Chem. B, Vol. 112, No. 33, 2008 10289 (26) Meinnel, T.; Blamquet, S; Dardel, F. J. Mol. Biol. 1996, 262, 375– 386. (27) Matthews, B. W. Chem. ReV 1968, 21, 491. (28) Ragusa, S.; Mouchet, P.; Lazennec, C.; Dive, V.; Meinnel, T. J. Mol. Biol. 1999, 289, 1445–1457. (29) Pelmenschikov, V.; Blomberg, M. R. A.; Siegbahn, P. E. J. Bio. Inorg. Chem. 2002, 7, 284–298. (30) Antonczak, S.; Monard, G.; Ruiz Lopez, M.; Rivail, J. L. J. Mol. Model. 2000, 6, 527–538. (31) Mazel, D.; Pochet, S.; Marliere, P. EMBO J. 1994, 13, 914–923. (32) Warshel, A.; Levitt, M. J. Mol. Biol. 1976, 103, 227–249. (33) Field, M. J.; Bash, P. A.; Karplus, M. J. Comput. Chem. 1990, 11, 700–733. (34) Gao, J. In ReView in Computational Chenistry; VCH: New York, 1995; Vol. 7, pp 119-185. (35) Cui, Q.; Karpulus, M. In Protein Simulation: AdVances in Protein Chemistry; Daggett, V., Ed.; Elsevier: London, 2003; Vol. 66, pp 315-372. (36) Friesner, R. A.; Guallar, V. Annu. ReV. Phys. Chem. 2005, 56, 389– 427. (37) Zhang, Y. K. Theo. Chem. Acc. 2006, 116, 43–50. (38) Gao, J.; Truhlar, D. G. ReV. Phys. Chem. 2002, 53, 467–505. (39) Liu, H. Y.; Müller-Plathe, F.; van Gunsteren, W. F. J. Phys. Chem. 1995, 102, 1722–1730. (40) Henkelman, G.; Jonsson, H. J. Chem. Phys. 2000, 113, 9978–9985. (41) Maragakis, P.; Andreev, S. A.; Brumer, Y.; Reichman, D. R.; Kaxiras, E. J. Chem. Phys. 2002, 117, 4651–4658. (42) Xie, L.; Liu, H. Y.; Yang, W. T. J. Chem. Phys. 2004, 120, 8039– 8052. (43) Zhang, Y. K.; Lee, T. S.; Yang, W. T. J. Chem. Phys. 1999, 110, 46–54. (44) Zhang, Y. K.; Liu, H. Y.; Yang, W. T. J. Chem. Phys. 2000, 112, 3483–3492. (45) Zhang, Y. K. J. Chem. Phys. 2005, 122, 024114. (46) Insight II; Accelrys Software, Inc.: San Diego, CA, 2000. (47) Holden, H. M.; Tronrud, D. E.; Monzingo, A. F.; Weaver, L. H.; Matthews, B. W. Biochemistry 1987, 26, 8542–8553. (48) Ponder, J. W. TINKER: Software Tools for Molecular Design, version 3.6. The most updated version for the TINKER program can be obtained via the Internet at http://dasher.wustl.edu/tinker. Site maintained by J. W. Ponder (accessed February 1998). (49) Singh, U. C.; Kollman, P. A. J. Comput. Chem. 1986, 7, 718–730. (50) Zhang, Y. K.; Liu, H. Y.; Yang, W. T. In Computational Methods for Macromolecules: Challenges and Applications (Lecture Notes in Computational Science & Engineering, Vol. 24); Schlick, T., Gan, H. H., Eds.; Springer-Verlag: New York, 2002; pp 332-354. (51) Antes, I.; Thiel, W. J. Phys. Chem. A 1999, 103, 9290–9295. (52) Dewar, M. J. S.; Zoebisch, E. G.; Healy, E. F.; Stewart, J. J. P. J. Am. Chem. Soc. 1985, 107, 3902–3909. (53) Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Gould, I. R.; Merz, K. M.; Ferguson, D. M.; Spellmeyer, D. C.; Fox, T.; Caldwell, J. W.; Kollman, P. A. J. Am. Chem. Soc. 1995, 117, 5179–5197. (54) Becke, A. D. J. Chem. Phys. 1993, 98, 5648–5652. (55) Petersson, G. A.; AI-Laham, M. A. J. Chem. Phys. 1991, 94, 6081– 6090. (56) Petersson, G. A.; Bennett, A.; Tensfeldt, T. G.; Al-Laham, M. A.; Shirley, W. A.; Mantzaris, J. J. Chem. Phys. 1988, 89, 2193–2218. (57) Ryckaert, J. P.; Ciccotti, G.; Berendsen, H. J. C. J. Comput. Phys. 1977, 23, 327–341. (58) Berendsen, H. J. C.; Postma, J. P. M.; Gunstereno, W. F. V.; DiNola, A.; Haak, J. R. J. Chem. Phys. 1984, 81, 3684–3690. (59) Simonson, T.; Archontis, G.; Karplus, M. J. Chem. Phys. 1997, 101, 8349–8362. (60) Parmaneck, M. S.; Li, G.; Zhang, X.; Cui, Q. J. Theor. Comput. Chem. 2002, 1. (61) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Montgomery, J. A., Jr.; Vreven, T.; Kudin, K. N.; Burant, J. C.; Millam, J. M.; Iyengar, S. S.; Tomasi, J.; Barone, V.; Mennucci, B.; Cossi, M.; Scalmani, G.; Rega, N.; Petersson, G. A.; Nakatsuji, H.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.; Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Klene, M.; Li, X.; Knox, J. E.; Hratchian, H. P.; Cross, J. B.; Bakken, V.; Adamo, C.; Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.; Austin, A. J.; Cammi, R.; Pomelli, C.; Ochterski, J. W.; Ayala, P. Y.; Morokuma, K.; Voth, G. A.; Salvador, P.; Dannenberg, J. J.; Zakrzewski, V. G.; Dapprich, S.; Daniels, A. D.; Strain, M. C.; Farkas, O.; Malick, D. K.; Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Ortiz, J. V.; Cui, Q.; Baboul, A. G.; Clifford, S.; Cioslowski, J.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.; Peng, C. Y.; Nanayakkara, A.; Challacombe, M.; Gill, P. M. W.; Johnson, B.; Chen, W.; Wong, M. W.; Gonzalez, C.; Pople, J. A. Gaussian 03, revision A.5; Gaussian, Inc.: Pittsburgh, PA, 2003. (62) Baker, J. J. Comput. Chem. 1993, 14, 1085–1100. (63) Hammond, G. S. J. Am. Chem. Soc. 1955, 77, 334–338.

10290 J. Phys. Chem. B, Vol. 112, No. 33, 2008 (64) Cross, J. B.; Vreven, T.; Meroueh, S. O.; Mobashery, S.; Schlegel, H. B. J. Phys. Chem. B 2005, 109, 4761–4769. (65) Pelmenschikov, V.; Siegbahn, P. E. M. Inorg. Chem. 2002, 41, 5659–5666. (66) Bachovchin, W. W. Magn. Reson. Chem. 2001, 39, S199–S213. (67) Hedsatrom, L. Chem. ReV. 2002, 102, 4501–4524. (68) Wiberg, K. B. Tetrahedron 1968, 24, 1083. (69) Fieulaine, S.; Juillan Binard, C.; Serero, A.; Dardel, F.; Giglione, C.; Meinnel, T.; Ferrer, J. J. Biol. Chem. 2005, 280, 42315–42324.

Dong and Liu (70) Li, Y. K.; Ren, S. X.; Gong, W. M. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2002, 58, 846–848. (71) Li, Y. K.; Chen, Z. F.; Gong, W. M. Biochem. Biophys. Res. Commun. 2002, 295, 884–889. (72) Zhou, Z. C.; Song, X. M.; Li, Y. K.; Gong, W. M. J. Mol. Biol. 2004, 339, 207–215. (73) Shindyalov, I. N.; Bourne, P. E. Protein Eng. 1998, 11, 739–747.

JP711209J