DFT Investigation on the Mechanism of the Deacetylation Reaction

Feb 27, 2008 - WenJuan Huang , Jorge Llano and James W. Gauld ... Marcy Hernick , Samuel G. Gattis , James E. Penner-Hahn and Carol A. Fierke...
0 downloads 0 Views 459KB Size
Download PDF
3462

J. Phys. Chem. B 2008, 112, 3462-3469

DFT Investigation on the Mechanism of the Deacetylation Reaction Catalyzed by LpxC Jesse J. Robinet and James W. Gauld* Department of Chemistry and Biochemistry, UniVersity of Windsor, Windsor, Ontario N9B 3P4, Canada ReceiVed: July 11, 2007; In Final Form: December 12, 2007

LpxC is a key enzyme in the biochemical synthesis of Lipid A, an important outer cell-membrane component found in a number of pathogenic bacteria. Using DFT, we have investigated the binding of the substrate within its active site as well as the deacetylation mechanism it catalyzes. The substrate is found to preferentially coordinate to the active site Zn2+ via its carbonyl oxygen between a Zn2+-bound H2O and an adjacent threonine (Thr191). Furthermore, upon substrate binding a nearby Glu78 residue is found to readily deprotonate the remaining Zn2+-bound H2O. Unlike several related metallopeptidases, the mechanism of LpxC is found to proceed via four steps; (i) initial hydroxylation of the substrates’ carbonyl carbon to give a gem-diolate intermediate, (ii) protonation of the amide nitrogen by the histidine His265-H+, (iii) a barrier-less change in the active site-intermediate hydrogen-bond network and finally, (iv) C-N bond cleavage. Notably, the ratedetermining step of the mechanism of LpxC is found to be the initial hydroxylation while the final C-N bond cleavage occurs with an overall barrier of 23.6 kJ mol-1. Furthermore, LpxC uses a general acid/base pair mechanism as indicated by the fact that both His265-H+ and Glu78 are accordingly involved.

1. Introduction UDP-3-O-(R-3-hydroxymyristoyl)-N-acetylglucosaminedeacetylase (LpxC) has attracted increasing attention in recent years as an antibacterial target.1-8 This is due in large part to the fact that it catalyzes the first committed step in the production of Lipid A,9 which anchors the lipopolysaccharides that form the outer membrane of all gram-negative bacteria.10 Significantly, Lipid A is responsible for the pathogenesis of gram-negative bacteria and, furthermore, is also thought to be involved in their resistance to some antibiotics.4 Thus, information on the catalytic mechanism of LpxC may provide insights for creating more effective antibiotics against gram-negative bacteria. Specifically, LpxC catalyzes the second step in the biosynthesis of Lipid A, deacetylation of UDP-3-O-(R-3-hydroxymyristoyl)-N-acetylglucosamine to give UDP-3-O-(R-3-hydroxymyristoyl)glucosamine and acetate (Scheme 1).10 LpxC is a metalloenzyme that has been found to employ a zinc(II) metal ion in its active site.11 Through mutagenesis,12 EXAFS studies13 and X-ray crystallographic efforts,14 it is now known that the Zn2+ ion directly coordinates to three amino acid residues; two histidines (His79 and His238) and an aspartate (Asp242). The fourth ligand is a water molecule, a typical key component of zinc hydrolases.15,16 Recently, Gennadios et al.17 crystallized several LpxC-bound ligand complexes and examined other available crystal structures of LpxC. They observed that the active site Zn2+ ion exhibits both square pyramidal and tetrahedral geometries owing to the flexibility of its coordination sphere. Interestingly, from the structure of the LpxC-imidazole complex, believed to show the Zn2+ in its native state, the Zn2+ ion is square pyramidal with two water molecules coordinated. Therefore, they17 proposed that binding of the substrate displaces one of these water molecules in order to accommodate a substrate‚‚‚Zn interaction, while the second water molecule is retained. This suggested coordination geometry for the substrate * Author to whom correspondence should be addressed. E-mail: gauld@ uwindsor.ca.

was further supported by the X-ray crystal structure of LpxC cocrystallized with the substrate analogue TU-514.17 It is generally accepted that LpxC proceeds via a general acid/ base mechanism common to many Zn2+-dependent enzymes,16,18 namely, a general base deprotonates a Zn2+-bound H2O. The resulting hydroxide then nucleophilically attacks the substrate’s carbonyl carbon to form a tetrahedral (gem-diolate) intermediate. The amino nitrogen of the gem-diolate then accepts a proton from a general acid, thus cleaving the amide bond and forming the desired products. Although, the general mechanism is agreed upon, the exact nature of the general acid and base are unclear. Initially, Whittington et al.14 proposed a mechanism (Scheme 2) for LpxC in which an active site glutamate (Glu78) acts as the general base. They further proposed that the resulting Glu78-OH then acts as the general acid, deeming it to be more suitably positioned than a protonated active site histidine (His265-H+) which was proposed to instead simply help stabilize the transition state.14 Shortly thereafter, however, on the basis of the structure of LpxC complexed with TU-514 and mutagenesis studies, Coggins et al.19 proposed that His265 is in fact more appropriately suited to serve as both the mechanisms general acid and the base, with Glu78 serving only a secondary role. More recently, however, Coggins et al.20 determined the pKa of His265 to be 7.6. This suggests that His265 is likely protonated within the LpxC active site and thus unable to accept a proton from the Zn2+-bound water. Recently, on the basis of kinetic studies of wild type and mutant enzymes, both McClerren et al.21 and Hernick et al.22 proposed that Glu78 acts as the general base in the mechanism. This proposal was further supported by structural features observed in the LpxCcacodylate and -palmitate complexes obtained by Hernick et al.22 However, these structures also suggest that His265-H+ might be in fact more suitably positioned than Glu78-OH to protonate the amino nitrogen of the intermediate. On the basis of these findings, they proposed a mechanism in which His265H+ and Glu78 work together as a general acid/base pair (Scheme 2).22 In particular, they proposed that Glu78 accepts a proton

10.1021/jp075415m CCC: $40.75 © 2008 American Chemical Society Published on Web 02/27/2008

DFT Study of LpxC Mechanism

J. Phys. Chem. B, Vol. 112, No. 11, 2008 3463

SCHEME 1: LpxC Catalyzes the Conversion of UDP-3-O-(R-3-Hydroxymyristoyl)-N-acetylglucosamine to UDP-3-O-(R3-Hydroxymyristoyl)glucosamine and Acetate10

SCHEME 2: Proposed Mechanisms for the Deacetylation Reaction Catalyzed by LpxC in Which: (a) Glu78 Acts as Both the General Acid and Base14,22 and (b) Glu78 and His265-H+ Act as the General Base and Acid, Respectively22

from the Zn2+-bound H2O, while His265-H+ protonates the intermediate’s nitrogen, thus cleaving the amide bond. In this present study, we have used density functional theory in order to investigate the roles of the active site amino acids as well as to elucidate the overall catalytic mechanism of LpxC.

SCHEME 3: Schematic Illustration of the LpxC Active Site Model Used for This Study Where Atoms Held Fixed Throughout Optimizations Are Bold and Colored Red

2. Computational Methods All calculations were performed using the density functional theory (DFT) method B3LYP, a combination of Becke’s threeparameter hybrid exchange functional23 with the Lee-YangParr correlation functional24 as implemented in Jaguar 5.5.25 The LACVP** basis set was employed for all optimizations. Effects due to the polar environment around the active site were modeled using the Poisson-Boltzmann polarizable continuum method (PB-PCM) as implemented in Jaguar (i.e., PB-PCMB3LYP/LACVP**).25 A dielectric constant of 4.0 was chosen as it has been previously suggested for modeling the environment of proteins.26,27 All structures were characterized as minima or transition structures via gas-phase frequency calculations in Gaussian 03.28 All relative energies given herein were obtained by performing single-point calculations at the B3LYP/ LACV3P+** level of theory in combination with the PB-PCM solvation method ( ) 4.0) using the above optimized structures (i.e., PB-PCM-B3LYP/LACV3P+**//PB-PCM-B3LYP/ LACVP**). The chemical model employed in this study (Scheme 3) was obtained from the LpxC-cacodylate crystal structure PDB: 1YHC.22 In our system, the substrate UDP-3-O-(R-3-hydroxymyristoyl)-N-acetylglucosamine was modeled by N-methylacetamide; that is, the sugar has been replaced by a -CH3. In addition, we have included the Zn2+ ion as well as its coordinated ligands (His79, His238, Asp242, and water). We

have also modeled the amino acid residues that are proposed to be catalytically important, namely, Glu78, His265-H+, and Thr191. These residues are modeled as follows: histidines (His) as imidazoles, glutamate (Glu) and aspartate (Asp) as acetate, and threonine (Thr) as methanol. Furthermore, each of these model amino acids have had their terminal carbon atom fixed at the crystal structure position in order to maintain the integrity of the active site (fixed atoms are shown in red bold in Scheme 3). We note that in the crystal structure one of the oxygens of

3464 J. Phys. Chem. B, Vol. 112, No. 11, 2008

Figure 1. Potential energy surfaces for nonenzymatic hydrolysis of the amide bond of N-methylacetamide via concerted (red dashed) and stepwise (blue solid) mechanisms.

Asp242 is more tightly coordinated to the Zn2+ ion than the other with O‚‚‚Zn2+ distances of 2.12 and 2.70 Å. In the chemical model used in this present study, the O‚‚‚Zn2+ distances are in closer agreement with each other and generally lie between the two values observed in the crystal structure. For example, in the bound active site structure 1A, they are 2.23 and 2.29 Å respectively, with one oxygen being more tightly coordinated than the other (see Table S2). Thus, while throughout this study, Asp242, more specifically the acetate used to model it, is shown as interacting via both its carboxylic oxygens with the Zn2+ ion; it should be remembered that, in all structures, one of its oxygens is in fact preferentially coordinated in agreement with the crystal structure. 3. Results and Discussion 3.1. Nonenzymatic Deacetylation Mechanism. In order to provide greater insights into the LpxC catalyzed deacetylation mechanism, we first investigated the water-assisted or nonenzymatic hydrolysis of the N-methylacetamide (our substrate model) amide bond via both concerted (red dashed) and stepwise (blue solid) mechanisms. The resulting potential energy surfaces (PESs) are given in Figure 1. It should be noted that these and related reactions have also been previously studied elsewhere.29-36 However, the present results provide structures and barrier heights that can be directly compared to those obtained for the LpxC mechanism detailed herein. Thus, only the most important features of relevance to LpxC are briefly discussed. For nonenzymatic hydrolysis of the amide bond in Nmethylacetamide, the initial reactant complex (1aq) lies 29.2 kJ mol-1 lower in energy than the isolated reactants as a result of the hydrogen-bond chain formed by the waters to the carbonyl oxygen of the methylacetamide. The concerted mechanism then proceeds in one step via the six-membered ring transition structure TSCaq in which the oxygen (OA) of one of the waters (H2OA) attacks at the carbonyl carbon (CS) of the substrate. Simultaneously, however, the second water (H2OB) accepts a proton from H2OA while concomitantly donating one of its own to the substrates amide nitrogen, cleaving the amide bond (Figure 1). In TSCaq, the CS-N bond has lengthened markedly by 0.19 Å to 1.54 Å relative to that observed in 1aq. Notably, the barrier for this mechanism is quite high at 175.4 kJ mol-1 relative to 1aq, while the resulting hydrogen-bonded product complex 3aq lies 11.9 kJ mol-1 higher in energy than 1aq. Similar to the concerted pathway, the alternate stepwise mechanism also begins by attack of the oxygen (OA) of one of

Robinet and Gauld the waters (H2OA) at the substrates CS center while the second water (H2OB) again concomitantly accepts one of its protons. Now, however, H2OB donates one of its own protons to the substrates carbonyl oxygen (OS) to form the tetrahedral gemdiol intermediate 2aq. This reaction step proceeds via the sixmembered transition structure TS1aq at a cost of 172.5 kJ mol-1. This is slightly lower than the barrier for the concerted mechanism. In contrast to TSCaq, in TS1aq, the CSsN bond has elongated by only 0.02 Å to 1.37 Å, while the carbonyl CSdOS distance has increased from 1.25 Å to 1.31 Å. The resulting intermediate 2aq lies 90.2 kJ mol-1 higher in energy than the initial reactant complex 1aq. The remaining water H2OB is now hydrogen bonded to both of the gem-diols hydroxyl groups with OB‚‚‚HOS- and OB‚‚‚HOA- distances of 2.08 and 2.00 Å respectively, and to the substrates amide nitrogen with an -OBH‚‚‚N distance of 1.86 Å. In 2aq the CS-N and both CS-O bonds have lengths of 1.49 and 1.41 Å respectively. These distances are similar to those for the single C-N (1.47 Å) and C-O (1.42 Å) bonds in CH3NH2 and CH3OH respectively, as optimized at the same level of theory (see Computational Methods; Table S2, Supporting Information). In the second step, H2OB abstracts a proton from one of the gem-diols C-OH groups while simultaneously donating one of its own to the amino nitrogen. This reaction proceeds via the sixmembered transition structure TS2aq at a cost of 49.8 kJ mol-1 relative to 2aq, 140.0 kJ mol-1 relative to 1aq. It is noted that the breaking CS-N bond in TS2aq is significantly longer (1.61 Å) than that observed for TSCaq in the concerted mechanism (see above). In summary, the stepwise mechanism is preferred, as it has a slightly lower overall barrier than the concerted mechanism. Furthermore, initial hydroxylation of the substrates carbonyl carbon is the rate-determining step as it corresponds to the largest barrier in the stepwise mechanism. 3.2. Active Site-Bound Substrate Complex. We began our investigation of the actual LpxC catalyzed deacetylation mechanism by considering substrate binding within the active site, specifically its energetically preferred position. When bound, the substrate’s uridine 5′-diphosphate interacts with a region of basic amino acids37 while its 3-O-(R-3-hydroxymyristoyl) moiety is believed to interact with a hydrophobic region.14 More mechanistically important, however, the substrate is believed to displace one of the two water molecules coordinated to the Zn2+ ion upon binding.17 Specifically, it displaces either the Zn2+-coordinated H2O that is also hydrogen bonded to the nearby threonine (Thr191) residue (1A) or that which is hydrogen bonded to the active site histidine (His265) and glutamate (Glu78) residues (1B). We have considered both possibilities with the resulting lowest energy structure of each shown in Figure 2. The lowest energy structure was found to correspond to 1A, that is, in which the substrate replaces the Zn2+-bound H2O that is also hydrogen bonded to Thr191. Indeed, the carbonyl oxygen (OS) of the substrate is now coordinated to the Zn2+ at a distance of 2.16 Å. Concomitantly, OS also forms a strong hydrogen bond (1.79 Å) with the -OH group of Thr191 (Figure 2). Interestingly, in 1A the remaining Zn2+-bound H2O has transferred a proton to the glutamate Glu78. Thus, the initial substrate complex now in fact contains a “protonated” Glu78 (Glu78OH) and a Zn2+‚‚‚OH- moiety. Furthermore, the oxygen of the resulting hydroxide (O1) forms two almost equidistant strong hydrogen bonds with Glu78-OH and the nearby protonated histidine His265-H+ with O1‚‚‚HOGlu78 and O1‚‚‚H-NHis265 bond lengths of 1.47 and 1.48 Å respectively (see Figure 2). The strength of these hydrogen bonds is also illustrated by

DFT Study of LpxC Mechanism

Figure 2. Lowest energy active site-bound substrate complexes: (1A) substrate bound between a Zn2+-bound H2O and Thr191 residue, (1B) substrate bound between His265-H+ and a Zn2+ bound H2O (see text). Select optimized distances (angstroms) are shown. Ligands only involved in binding the Zn2+ are shown as tubes while all others are shown as balls and sticks. [Zn (purple), C (gray), N (blue), O (red), H (white)].

the correspondingly lengthened OGlu78-H and NHis265-H bonds which are 1.04 and 1.12 Å respectively. As a result, the Zn2+‚‚‚O1 distance (2.18 Å) is in fact slightly longer than the Zn2+‚‚‚OS coordination bond. Furthermore, we note that the Zn2+‚‚‚O1 distance in 1A is also longer than the Zn2+‚‚‚OH2 bond (2.15 Å) in 1B (see below). This suggests that the hydroxide ion in 1A does not in fact become more tightly bound to the Zn2+ ion once formed and thus perhaps can still readily act as a nucleophile. The alternate structure 1B lies significantly higher in energy than 1A by 89.7 kJ mol-1. The substrate again coordinates to the Zn2+ via its carbonyl oxygen, though now the distance is notably longer at 2.36 Å. The remaining H2O remains coordinated to the Zn2+ ion at a distance of 2.15 Å and, furthermore, forms a strong hydrogen bond (1.65 Å) with the Thr191 hydroxyl group. Importantly, however, it has not donated a proton to Glu78 because of the fact that it lies 5.14 Å from the nearest carboxylic oxygen of Glu78. We note that we also examined a 1B-type structure in which the remaining H2O had donated a proton to Glu78 (see Table S2). However, the resulting structure was found to still lie higher in energy than 1A by over 30 kJ mol-1. Furthermore, the resulting Zn2+‚‚‚OHdistance was considerably shorter at 1.91 Å (Table S2), further suggesting that the OH- formed in 1A will be the better nucleophile. It is noted that several experimental17,22,38 studies based on, for example, binding of cacodylate within the active site17,22 and effects on the kinetics38 of the reaction upon mutating Thr191 to an Alanine have suggested that the substrates carbonyl oxygen forms a hydrogen bond to the -OH group of Thr191, at least during the course of the reaction. These results

J. Phys. Chem. B, Vol. 112, No. 11, 2008 3465 appear to provide further support for 1A as the preferred binding mode; thus, it was used as the bound active site model for further mechanistic studies. 3.3. LpxC Catalyzed Deacetylation Reaction. The catalytic mechanism of LpxC was then investigated. The potential energy surface obtained for the overall mechanism is given in Figure 3. Analogous to the nonenzymatic mechanism, the first step corresponds to nucleophilic attack of the Zn2+-bound hydroxide’s oxygen (O1) at the substrate’s carbonyl carbon (CS) to form the gem-diolate like intermediate (2). This step proceeds via TS1 (Figure 4) with a barrier of 87.7 kJ mol-1, which is approximately only half that of the corresponding barrier for nucleophilic attack in the nonenzymatic mechanism (cf. Figure 1). Furthermore, the resulting tetrahedral intermediate 2 now lies just 66.9 kJ mol-1 higher in energy than the initial reactant complex 1A. Thus, the significantly lower barrier in LpxC is likely indicative both of the greater nucleophilicity of the oxygen of the OH- ion compared with that in H2O, as well as stabilization of the resulting gem-diolate intermediate by the active site. We note that the O1‚‚‚CS distance in TS1 (1.62 Å) is significantly shorter than observed for TS1aq (1.95 Å) suggesting that the transition structure occurs later in the reaction, more closely resembling the intermediate in accordance with Hammond’s postulate.39 It also possibly reflects in part structural constraints due to the active site. Interestingly, in TS1, the Zn2+‚‚‚O1H- bond has essentially broken, having a distance of 3.41 Å. In contrast, while the O1‚‚‚HOGlu78 and O1‚‚‚HNHis265 hydrogen bonds have also lengthened markedly to 1.81 and 1.63 Å, respectively, they have not been broken. It should also be noted that in 2 the CSsNS bond has also lengthened significantly by 0.19 Å to 1.52 Å and now is in fact slightly longer than observed in the nonenzymatic gem-diolate intermediate 2aq (1.49 Å, see above). In addition, the newly formed CS-O1 bond resembles a long C-O single bond with a length of 1.48 Å. This may be due in part to the fact that, while the CS-OS bond has elongated from 1.27 Å in 1A to 1.33 Å, reflecting the increased oxyanion nature of OS in 2, it appears to still have some double bond character as indicated by its shorter length compared with, for example, r(C-O) in CH3OH (1.42 Å). Furthermore, the -O1H oxygen forms a strong hydrogen bond (1.73 Å) with the -OH group of Glu78. Because of the increased oxyanion nature of the substrate’s carbonyl oxygen OS, its interaction with the metal ion has strengthened with the OS‚‚‚Zn2+ distance shortening considerably from 2.16 Å (1A) to 1.97 Å. In contrast, upon transfer from the Zn2+ ion to the substrate, O1 is no longer coordinated to the Zn2+ as indicated by the O1‚‚‚Zn2+ distance of 3.19 Å. This also illustrates the flexibility of the coordination sphere of Zn2+. It should also be noted that the Zn2+ coordination environment of 2 is similar to that observed in the cacodylate-LpxC crystal structure: the oxyanion and not the hydroxyl group of the substrate is coordinated to the Zn.22 Other changes in active site interactions upon hydroxylation of the substrate are also apparent. For instance, upon transfer the interaction between the -O1H oxygen and Glu78-OH proton has lengthened considerably from 1.47 Å to 1.73 Å. Furthermore, the acidic proton of His265-H+ now forms a strong hydrogen bond with the nitrogen center (NS) of intermediate 2 (1.62 Å). Remarkably, despite OS having increased negative charge, its hydrogen bond with the hydroxyl group of Thr191 is now in fact markedly longer by 0.32 Å at 2.11 Å than observed in 1A. Again, similar to the nonenzymatic mechanism (cf. Figure 1) the second step is found to correspond to protonation of

3466 J. Phys. Chem. B, Vol. 112, No. 11, 2008

Robinet and Gauld

Figure 3. Potential energy surface for the overall reaction mechanism of LpxC. Key residues of minima are schematically shown in order to highlight changes in the system during the reaction. For clarity, the Thr191 and Glu78-OH residues have been omitted.

Figure 4. Optimized geometries of transition structure TS1 and gemdiolate intermediate 2 arising from the addition of the hydroxide ion at the substrate’s carbonyl carbon. Select optimized distances (angstroms) are shown. Ligands only involved in binding the Zn2+ are shown as tubes while all others are shown as balls and sticks. [Zn (purple), C (gray), N (blue), O (red), H (white)].

intermediate 2’s nitrogen (NS) center. If this reaction were to proceed via a single bifunctional acid/base amino acid residue

as has been previously proposed,14 Glu78-OH would now act as an acid and donate its proton derived from the initial Zn2+‚‚‚OH2 moiety. However, on the basis of the geometry of 2, this seems unlikely as the distance between glutamate’s proton and the amino nitrogen is 4.07 Å and furthermore, the intermediate’s hydroxyl group lies between them. Thus, the optimized structure of 2 appears to support the proposal of Hernick et al. that His265-H+ is more aptly positioned to provide the proton in the alternate general acid/base pair mechanism.22 Indeed, the transfer of a proton from His265-H+ to NS proceeds via TS2 with a barrier of only 2.5 kJ mol-1 to give the protonated gem-diolate intermediate 3 which lies just 37.8 kJ mol-1 higher in energy than 1A. We note that the optimized structure of TS2 is indicative of such a low, early reaction barrier with the transferring H+ lying much closer to the appropriate nitrogen center of His265 than that of the intermediate; r(NHis265‚‚‚H+) ) 1.19 Å while r(NS‚‚‚H+) ) 1.45 Å (Figure 5). In the resulting intermediate 3, the newly formed NS-H group hydrogen bonds (1.82 Å) to the basic N center of His265. However, while protonation of the NS center in 3 has caused the CS-NS bond to lengthen significantly to 1.62 Å, it has not resulted in its cleavage. This can also be seen by considering the dihedral angle ∠C-CS-OS-O1, that is, a measure of the pyramidality of the CS center. While in 3 it has indeed increased to 127.2° from that observed in 2 (120.3°), it is still considerably less than expected if the CS-NS bond had broken; in acetate it is 180.0°. This is unlike that observed in the nonenzymatic mechanism where protonation of the amide nitrogen occurred simultaneously with C-N bond cleavage. Interestingly, the stability of 3 appears to be due in part to its coordination to the Zn2+ ion via its formal oxyanion oxygen OS; see Figure 3. For related models, when the Zn2+‚‚‚OS interaction was broken, the CS-NS bond instantly cleaved (not shown). We note that in 3 the CS-O1 and CS-OS distances have both shortened slightly from that observed in 2 (Table S2, Supporting Information) by 0.04 and 0.02 Å, respectively, to 1.44 and 1.31 Å. All other active site interactions observed in 2 are retained in 3 although

DFT Study of LpxC Mechanism

J. Phys. Chem. B, Vol. 112, No. 11, 2008 3467

Figure 5. Optimized geometries of the transition structures and intermediates arising from proton transfer from His265-H+ to the amino nitrogen of the intermediate (TS2 and 3), as well as reorientation of His265 within the active site (TS3 and 4). Select optimized distances (angstroms) are also shown. Ligands only involved in binding the Zn2+ are shown as tubes while all others are shown as balls and sticks. [Zn (purple), C (gray), N (blue), O (red), H (white)].

the Glu78-OH‚‚‚O1 hydrogen bond has lengthened slightly to 1.83 Å, while that between the Thr191-OH group and the oxyanion of the intermediate has shortened by 0.09 Å to 2.02 Å compared with that observed in 2. It should be noted that related doubly protonated intermediates have been found to be occur in the mechanisms of other previous theoretically investigated metallopeptidases, for example, thermolysin,40 peptide deformylase,41-43 matrix metalloproteinases,44 and methionine aminopeptidaes.45 However, the above results suggest LpxC forms this intermediate in two steps, whereas in each of the examples40-45 listed above, the analogous intermediate is formed in one step via concomitant attack of an hydroxide at the carbonyl carbon and a proton transfer to the amide nitrogen. From previous computational studies on the aforementioned metallopeptidase enzymes,40-45 cleavage of the CS-NS bond is typically achieved by the transfer of the remaining CS-O1H proton from the intermediate to some nearby base in the active site. However, at this stage of the reaction, there appears to be no amino acid residue immediately capable of accepting this proton in the LpxC active site in 3. The Glu78-OH residue still has the proton from the initial Zn2+-bound H2O while the basic nitrogen of His265 is hydrogen bonded to a proton of the intermediate’s amino group. However, if the histidine were to break this hydrogen bond and instead form one with the CSO1H group of the intermediate, such a proton transfer might then be possible. Indeed, such a change in the hydrogen-bonding network is found to proceed via TS3 with essentially no barrier to give intermediate 4 lying just 11.3 kJ mol-1 higher in energy than 1A. That is, 3 is not thermodynamically or kinetically stable

with respect to rearrangement to the energetically lower lying 4. It should be noted that this “hydrogen-bond shift” is concomitant with a slight rotation of the substrates amino group (see below). A related shift in an active site-substrate hydrogenbond network has been previously observed by Leopoldini et al.45 in an alternate high-energy pathway for methionine aminopeptidase. However, in that enzyme, it was found to occur concomitantly with cleavage of the C-N bond and with an overall barrier of 87.4 kJ mol-1. In 4, His265 now forms a strong hydrogen bond (1.61 Å) with the substrate’s -O1H proton (Figure 5). In addition, it no longer interacts with the NS-H amino proton (3.25 Å) which instead now forms a hydrogen bond with an oxygen of the Zn2+ coordinating Asp242 residue with a distance of 1.80 Å. The other proton on NS lies 2.63 Å away from the basic nitrogen of His265. The hydrogen bonds from Thr191 and Glu78-OH to OS and O1 respectively have both shortened to 1.96 Å and 1.75 Å, respectively. In addition, the Zn2+‚‚‚O1 distance has also shortened significantly to 2.55 Å (Table S2), while the Zn‚‚‚OS interaction has lengthened slightly to 2.04 Å. The CSNS bond of 4 has only lengthened marginally by 0.01 Å to 1.63 Å with a concomitant marginal increase in the ∠C-CS-OSO1 dihedral angle to 128.3° (Figure 5 and Table S2). The CSOS and CS-O1 bonds have slightly lengthened and shortened respectively to 1.32 and 1.42 Å. The next and final step in the mechanism of LpxC is cleavage of the CS-NS bond. This is achieved by transfer of the CSO1H proton of the substrate to the nitrogen of His265 with concomitant cleavage of the C-N bond. This reaction proceeds

3468 J. Phys. Chem. B, Vol. 112, No. 11, 2008

Figure 6. Optimized geometries of the transition structure and the resulting products complex associated with the transfer of the substrate’s hydroxyl proton to His265. Select optimized distances (angstroms) are also shown. Ligands only involved in binding the Zn2+ are shown as tubes while all others are shown as balls and sticks. [Zn (purple), C (gray), N (blue), O (red), H (white)].

via TS4 at a cost of only 15.0 kJ mol-1 relative to 4, 26.3 kJ mol-1 overall. In the resulting final product complex 5, the desired ultimate products have been formed but remain bound in the active site. Furthermore, it lies 26.0 kJ mol-1 lower in energy than 1A (Figure 3). Thus, importantly, LpxC differs from some related metalloenzymes by having a barrier for this final C-N bond cleavage. For example, in thermolysin,40 peptide deformylase,41-43 and methionine aminopeptidase, 45 the analogous mechanistic step occurs without a barrier. In 5, the acetate derived from the initial substrate is coordinates to the Zn2+ ion through both of its oxygens, though not equally, as indicated by the Zn2+‚‚‚O1 and Zn2+‚‚‚OS optimized distances of 2.40 and 2.15 Å, respectively (Figure 6). As expected for an acetate anion the CS-O1 and CS-OS bonds are both much shorter, though again differing slightly at 1.29 and 1.27 Å respectively. The difference in C-O bond lengths and Zn2+‚‚‚O coordination distances reflects the different hydrogen-bonding interactions of the oxygens of the product acetates. The O1 oxygen, originally derived from the Zn2+-bound H2O, forms two reasonably strong hydrogen bonds with the now re-protonated histidine [r(O1‚‚‚+H-NHis265) ) 1.73 Å] and the glutamate (Glu78) residue [r(O1‚‚‚H-OGlu78) ) 1.75 Å]; see Figure 6. That is, the O1 oxygen essentially sits in the position originally occupied by the Zn2+-bound H2O. In contrast, the acetates OS oxygen forms just one reasonably strong hydrogen

Robinet and Gauld bond with the hydroxyl group of the adjacent Thr191 residue, r(OS‚‚‚H-OThr191) ) 1.86 Å. The product methylamine is found to also retain its hydrogen bond with an oxygen of Asp242, albeit it is now slightly longer at 2.18 Å. The strong Glu78OH‚‚‚O1 hydrogen bond suggests that release of the product acetate may also result in regeneration of the initial Glu78 anion. We note that the structure of 5 is similar to that reported for the product analogue palmitate bound to the Zn2+ ion in the active site. Indeed, the experimentally determined interatomic distances for the Zn2+‚‚‚O1, Zn2+‚‚‚OS, Glu78(O)‚‚‚O1, His265(N)‚‚‚O1, and Thr191(O)‚‚‚OS interactions are 2.3, 2.1, 3.0, 2.9, and 3.1 Å, respectively.22 Comparing these values to those obtained in our model, 2.4, 2.2, 2.7, 2.8, and 2.8 Å respectively, we see that our predicted interaction distances are generally in good agreement with those of the crystal structure. 3.4. Overall Mechanism. It can be seen from Figure 3 that the highest overall barrier for the LpxC mechanism, and hence the rate-determining barrier of the mechanism, is nucleophilic attack of the hydroxide at the substrates carbonyl carbon with a barrier of 87.7 kJ mol-1. This is similar to the stepwise nonenzymatic hydrolysis of the amide bond (Figure 1) where initial hydroxylation corresponds to the highest overall barrier. The calculated barrier for LpxC is slightly higher than that obtained for some related zinc metalloezymes. For example, in the dinuclear zinc methionine aminopeptidase, the barrier for initial hydroxylation of the substrate has been calculated to be approximately 65.7 kJ mol-1.45 Importantly, however, as previously noted by Leopoldini et al.,45 in thermolysin,40 peptide deformylase,41-43 and methionine aminopeptidase,45 initial hydroxylation occurs with concomitant protonation of the amide nitrogen. In contrast, the above results suggest that in LpxC these instead occur as two separate steps. After the initial hydroxylation of the substrates carbonyl carbon, all subsequent intermediates and transition structures are lower in energy than the one preceding it. The fourth and final step, deprotonation of the intermediates C-OH group with concomitant cleavage of its C-N bond, is found to occur with a barrier of 15.0 kJ mol-1. This again distinguishes LpxC from the related metalloenzymes40-43,45 noted above in which the second and final step was found to be cleavage of the C-N bond without a barrier. Intriguingly, the mechanism detailed herein or LpxC shows some similarities to an alternative higher-energy mechanism considered by Leopoldini et al.45 for the dinuclear zincmetalloenzyme methionine minopeptidase. In that mechanism, they found that initial hydroxylation of the substrates carbonyl carbon occurred with a much higher barrier of 93.2 kJ mol-1 and, furthermore, did not occur with concomitant protonation of the amide nitrogen. Rather, this latter protonation occurred as a distinct second step. In addition, they found that the third and final step of their higher-energy mechanism corresponded to a concomitant hydrogen-bond network shift and cleavage of the C-N bond with an overall barrier of 87.4 kJ mol-1. In LpxC, the hydrogen-bond shift occurs without a barrier and precedes C-N bond cleavage, which itself occurs with a low overall barrier of just 26.3 kJ mol-1. 4. Conclusions From the present DFT investigation into the hydrolysis of an amide bond via a nonenzymatic mechanism and the LpxC enzyme, several conclusions can be drawn: (1) Binding of the substrate to the LpxC active site preferentially occurs with the substrate’s carbonyl oxygen coordinating to the Zn2+ ion between the Zn-bound water and the threonine residue. Interestingly, upon coordination of the substrate, the

DFT Study of LpxC Mechanism active site Glu78 readily deprotonates the Zn-bound H2O, resulting in a Zn2+‚‚‚OH- moiety and now neutral Glu78-OH residue. (2) Similarly, in both the step-wise nonenzymatic and the LpxC mechanisms, the initial step where the carbonyl carbon of the substrate is hydroxylated is the rate-determining step. However, the barrier for the LpxC mechanism (87.7 kJ mol-1) is approximately half that of the nonenzymatic system (172.5 kJ mol-1). This is due in part to the increased nucleophilicity of the Zn2+‚‚‚OH- oxygen compared with the H2O oxygen as well as stabilization of the transition structure and initial gemdiolate intermediate by the enzymes active site. (3) Analogous to the nonenzymatic mechanism the proceeding step in the LpxC catalyzed reaction is a proton transfer to the amino nitrogen of the intermediate. LpxC employs the protonated histidine residue (His265-H+) in order to do so. However, unlike the nonenzymatic system and several related metallopeptidases, the amide bond is not cleaved at this second step. Rather, it appears that the resulting protonated intermediate is stabilized by coordination to the Zn2+ ion via its C-O- oxyanion. (4) The third step in the LpxC catalyzed reaction is a change in the hydrogen-bonding network such that the basic nitrogen of His265 changes from hydrogen bonding to an N1-H group of the intermediate and instead hydrogen bonds with the intermediates CS-O1H hydroxyl group. The fourth and final step in the reaction is transfer of the CS-O1H proton onto His265 to regenerate the initial His265-H+ with concomitant cleavage of the C-N bond with an overall barrier of 26.3 kJ mol-1. (5) Overall, LpxC utilizes His265-H+ and Glu78 in an acid and base pair mechanism. Energetically, after the initial ratedetermining step, all subsequent minima and transition states are lower in energy than the corresponding stationary point preceding it. Acknowledgment. We thank the Natural Sciences and Engineering Research Council of Canada (NSERC), Canadian Foundation for Innovation (CFI), and Ontario Innovation Trust (OIT) for funding and Sharcnet and University of Waterloo for additional computational resources. J.R. also thanks Ontario Graduate Scholarship (OGS) program for financial support. Supporting Information Available: Full citation for reference 28, optimized Cartesian coordinates, and calculated energy values for all species in this present study. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Onishi, H. R.; Pelak, B. A.; Gerckens, L. S.; Silver, L. L.; Kahan, F. M.; Chen, M.-H.; Patchett, A. A.; Galloway, S. M.; Hyland, S. A.; Anderson, M. S.; Raetz, C. R. H. Science 1996, 274, 980. (2) Jackman, J. E.; Fierke, C. A.; Tumeyi, L. N.; Pirrungi, M.; Uchiyama, T.; Tahir, S. H.; Hindsgaul, O.; Raetz, C. R. H. J. Biol. Chem. 2000, 275, 11002. (3) Clements, J. M.; Coignard, F.; Johnson, I.; Chandler, S.; Palan, S.; Waller, A.; Wijkmans, J.; Hunter, M. G. Antimicrob. Agents Chemother. 2002, 46, 1793. (4) White, R. J.; Margolis, P. S.; Trias, J.; Yuan, Z. Curr. Opin. Pharmacol. 2003, 3, 502.

J. Phys. Chem. B, Vol. 112, No. 11, 2008 3469 (5) McClerren, A. L.; Endsley, S.; Bowman, J. L.; Andersen, N. H.; Guan, Z.; Rudolph, J.; Raetz, C. R. H. Biochemistry 2005, 44, 16574. (6) Mdluli, K. E.; Witte, P. R.; Kline, T.; Barb, A. W.; Erwin, A. L.; Mansfield, B. E.; McClerren, A. L.; Pirrung, M. C.; Tumey, L. N.; Warrener, P.; Raetz, C. R. H.; Stover, C. K. Antimicrob. Agents Chemother. 2006, 50, 2178. (7) Barb, A. W.; McClerren, A. L.; Snehelatha, K.; Reynolds, C. M.; Zhou, P.; Raetz, C. R. H. Biochemistry 2007, 46, 3793. (8) Kadam, R. U.; Chavan, A.; Roy, N. Bioorg. Med. Chem. Lett. 2007, 17, 861. (9) Andersen, M. S.; Bull, H. G.; Galloway, S. M.; Kelly, T. M.; Mohan, S.; Radika, K.; Raetz, C. R. H. J. Biol. Chem. 1993, 268, 19858. (10) Raetz, C. R. H.; Whitfield, C. Annu. ReV. Biochem. 2002, 71, 635. (11) Jackman, J. E.; Raetz, C. R. H.; Fierke, C. A. Biochemistry 1999, 38, 1902. (12) Jackman, J. E.; Raetz, C. R. H.; Fierke, C. A. Biochemistry 2001, 40, 514. (13) McClure, C. P.; Rusche, K. M.; Peariso, K.; Jackman, J. E.; Fierke, C. A.; Penner-Hahn, J. E. J. Inorg. Biochem. 2003, 94, 78. (14) Whittington, D. A.; Rusche, K. M.; Shin, H.; Fierke, C. A.; Christianson, D. W. Proc. Nat. Acad. Sci. U.S.A. 2003, 100, 8146. (15) Vallee, B. L.; Auld, D. S. Biochemistry 1990, 29, 5647. (16) Hernick, M.; Fierke, C. A. Arch. Biochem. Biophys. 2005, 433, 71. (17) Gennadios, H. A.; Whittington, D. A.; Li, X.; Fierke, C. A.; Christianson, D. W. Biochemistry 2006, 45, 7940. (18) Lipscomb, W. N.; Stra¨ter, N. Chem. ReV. 1996, 96, 2375. (19) Coggins, B. E.; Li, X.; McClerren, A. L.; Hindsgaul, O.; Raetz, C. R. H.; Zhou, P. Nat. Struct. Biol. 2003, 10, 645. (20) Coggins, B. E.; McClerren, A. L.; Jiang, L.; Li, H.; Rudolph, J.; Hindsgaul, O.; Raetz, C. R. H.; Zhou, P. Biochemistry 2005, 44, 1114. (21) McClerren, A. L.; Zhou, P.; Guan, Z.; Raetz, C. R. H.; Rudolph, J. Biochemistry 2005, 44, 1106. (22) Hernick, M.; Gennadios, H. A.; Whittington, D. A.; Rusche, K. M.; Christianson, D. W.; Fierke, C. A. J. Biol. Chem. 2005, 280, 16969. (23) Becke, A. D. J. Chem. Phys. 1993, 98, 1372. (24) Lee, C.; Yang, W.; Parr, R. G. Phys. ReV. B 1988, 37, 785. (25) Schro¨dinger, L. L. C. Jaguar 5.5 Portland, OR, 1991-2003. (26) Siegbahn, P. E. M.; Blomberg, M. R. A. Chem. ReV. 2000, 100, 421. (27) Noodleman, L.; Lovell, T.; Han, W.-G.; Li, J.; Himo, F. Chem. ReV. 2004, 104, 459. (28) Frisch, M. J. et al. Gaussian 03; Gaussian, Inc.: Pittsburgh, PA, 2004. (29) Oie, T.; Loew, G. H.; Burt, S. K.; Binkley, J. S.; MacElroy, R. D. J. Am. Chem. Soc. 1982, 104, 6169. (30) Krug, J. P.; Popelier, P. L. A.; Bader, R. F. W. J. Phys. Chem. 1992, 96, 7604. (31) Jenssen, J. H.; Baldridge, K. K.; Gordon, M. S. J. Phys. Chem. 1992, 96, 8340. (32) Antonczak, S.; Ruiz-Lopez, M. F.; Rivail, J.-L. J. Am. Chem. Soc. 1994, 116, 3912. (33) Antonczak, S.; Ruiz-Lopez, M. F.; Rivail, J.-L. J. Mol. Model. 1997, 3, 434. (34) Kallies, B.; Mitzner, R. J. Mol. Model. 1998, 4, 183. (35) Zahn, D. Eur. J. Org. Chem. 2004, 4020. (36) Gorb, L.; Asensio, A.; Tun˜o´n, I.; Ruiz-Lopez, M. F. Chem. Eur. J. 2005, 11, 6743. (37) Gennadios, H. A.; Christianson, D. W. Biochemistry 2006, 45, 15216. (38) Hernick, M.; Fierke, C. A. Biochemistry 2006, 45, 15240. (39) Hammond, G. S. J. Am. Chem. Soc. 1955, 77, 334. (40) Pelmenschikov, V.; Blomberg, M. R. A.; Siegbahn, P. E. M. J. Biol. Inorg. Chem. 2002, 7, 284. (41) Wu, X.-H.; Quan, J.-M.; Wu, Y.-D. J. Chem. Phys. B 2007, 111, 6236. (42) Xiao, C.; Zhang, Y. J. Phys. Chem. B 2007, 111, 6229. (43) Leopoldini, M.; Russo, N.; Toscano, M. J. Phys. Chem. B 2006, 110, 1063. (44) Pelmenschikov, V.; Siegbahn, P. E. M. Inorg. Chem. 2002, 41, 5959. (45) Leopoldini, M.; Russo, N.; Toscano, M. J. Am. Chem. Soc. 2007, 129, 7776.