J. Phys. Chem. B 2007, 111, 4547-4552
4547
The Mechanism of Cyclic Nucleotide Hydrolysis in the Phosphodiesterase Catalytic Site E. Alan Salter and Andrzej Wierzbicki* Department of Chemistry, UniVersity of South Alabama, Mobile, Alabama 36688 ReceiVed: October 6, 2006; In Final Form: February 23, 2007
The cyclic nucleotide phosphodiesterase superfamily of enzymes (PDEs) catalyzes the stereospecific hydrolysis of the second messengers adenosine and guanosine 3′,5′- cyclic monophosphate (cAMP, cGMP) to produce 5′-AMP and 5′-GMP, respectively. The PDEs are targets of high-throughput screening to determine selective inhibitors for a variety of therapeutic purposes. The catalytic pocket where the hydrolysis takes place is a highly conserved region and has several residues which are absolutely conserved across the PDE families. In this study, we consider a model cyclic substrate in which the adenine/guanine base has been replaced with a hydrogen atom, and we present results of a quantum computational investigation of the hydrolysis reaction as it occurs within the PDE catalytic site using the ONIOM hybrid (B3LYP/6-31g(d):PM3) method. We characterize the bound substrate, the bound hydrolyzed product, and the transition state which connects them for our model cyclic substrate placed in a truncated model of the PDE4D2 catalytic site. We address the role that the conserved histidine proximal to the bimetal system of the catalytic site, along with its conserved glutamine partner, plays in the generation of the hydroxide nucleophile. Our study provides computational evidence for several key features of the cAMP/cGMP hydrolysis mechanism as it occurs within the protein environment across the PDE superfamily.
Introduction The cyclic nucleotide phosphodiesterase superfamily of enzymes (PDEs) degrades the second messengers adenosine and guanosine 3′,5′-cyclic monophosphate (cAMP, cGMP) by a stereospecific hydrolysis of the P-O3′ bond to produce 5′-AMP and 5′-GMP, respectively. The PDEs are targets of highthroughput screening to find selective inhibitors for a variety of therapeutic purposes. Among these are to counter erectile dysfunction1 (PDE5), asthma, and chronic obstructive pulmonary disease (COPD) via relaxation of smooth muscle cells2-4 (PDE3, PDE4), to reduce inflammation associated with asthma and COPD2-4 (PDE3, PDE4), to perform as cardiotonic agents to remedy congestive heart failure5 (PDE3), and to act as cancer drugs by inducing apoptosis in neoplastic cells6 (PDE2, PDE5). The human genome encodes 21 PDE genes categorized into 11 families, and each gene family has differing regulatory regions, substrate specificity, and binding affinities. However, the catalytic pocket where the hydrolysis takes place is a highly conserved region and has several residues which are absolutely conserved across the PDE families.7,8 PDE hydrolysis has long been known to proceed with inversion of the phosphorus configuration, implying a trigonal bipyramidal transition state or intermediate.10,11 In the mechanism proposed by Xu et al.,7 on the basis of insights gained from their crystal structure of the catalytic domain of unliganded PDE4B and their molecular modeling of cAMP binding orientations, the phosphate oxygens of the substrate are assumed to bind to the bimetal system (Zn2+ and Mg2+) within the catalytic site. Coordinated to both metal ions is a bridge hydroxide ion or water molecule which can carry out a nucleophilic attack on the phosphorus atom, while an aspartate (Asp 392, PDE4B labeling of ref 7 is used throughout this report) holds the nucleophile in place via a hydrogen bond,7 Figure 1. As the reaction proceeds, the phosphorus center presumably inverts as the antipodal phosphate ester bond (P-
O3′) breaks and the ring opens, and the process can be imagined to pass through a pentavalent intermediate or transition state. Whether the protonation of O3′ follows or is concurrent with the nucleophilic attack is unknown. A nearby histidine (His 234) possibly provides a proton to the leaving alkoxide at O3′ to complete the hydrolysis, and an associated glutamine (Glu 413) can serve to modify the pKa of N of His 234 to facilitate the proton donation.7 The crystal structures of PDE4/AMP8,9 and PDE5/GMP12 seem to support the mechanism as outlined above, and the former structures are highly suggestive that the bridge species is indeed the nucleophile, but this hypothesis has yet not been proven. Of course, bound orientations of cAMP and cGMP cannot be determined from X-ray analysis because the hydrolysis reaction ensues. Substrate preference in the PDE family for cAMP versus cGMP is attributed primarily to the residues at the distal region of the catalytic site, away from the metal ions, where interactions occur with the respective adenine or guanine bases.7-9,12-14 Accordingly, the general aspects of the PDE hydrolysis mechanism should be independent of the nucleotide type and should be concerned primarily with how the phosphate diester moiety and the ribose ring interact with the bimetal system and the highly conserved proximal residues of the catalytic site. A computational argument in favor of hydroxide as the bridge species (and therefore the likely nucleophile in the hydrolysis rather than water) has been provided by Zhan and Zheng.15 Their gas-phase models comprised the two metal ions, simple representations of only the four residues directly bound to the metals, four waters of hydration, plus a bridge hydroxide/water, as in the unliganded PDE4B structure of Xu et al.7 Their study showed that whereas a hydroxide ion remains in a stable bridge position upon optimization, a bridge water molecule drifts away from the bridge position to result in a weakened interaction with the Zn2+ ion. In the case of their better model system (with the four residues represented as imidazoles and acetate ions), the
10.1021/jp066582+ CCC: $37.00 © 2007 American Chemical Society Published on Web 04/11/2007
4548 J. Phys. Chem. B, Vol. 111, No. 17, 2007
Salter and Wierzbicki
Figure 1. Computational model of the PDE catalytic site, on the basis of the PDE4D2 crystal structure.9 A two-layer ONIOM method was employed: B3LYP/6-31g(d) for the high-level region (ball and stick) and PM3 for the low-level region (tube). Most truncated amino acid residues are terminated at the alpha carbons as methyls. Our model cyclic substrate, intended to represent cAMP/cGMP, is shown bound to the metals (Zn2+ and Mg2+) via the phosphate oxygens Oa and Ob. Arrows indicate the proton transfers and attack of the P center by the bridge hydroxide which occur during hydrolysis of the P-O3′ bond. One water of hydration (W1) has been displaced by the substrate and is not shown. For the unliganded model of the PDE catalytic site, waters of hydration W1 and W2 are placed near the locations marked by the phosphate oxygens Oa and Ob, respectively. PDE4B labeling is used.
bridge water drifted out of range of binding to the Zn2+ ion entirely. On the basis of these qualitative results and an estimation of the pKa of their small bimetal system, they concluded that the bridge ligand in the catalytic site is a hydroxide ion.15 The generation of hydroxide was not considered in their study, and the His 234/Glu 413 pair of residues were not included in their model. Very recently, Xiong et al.16 revisited this question by carrying out molecular dynamics simulations and ONIOM hybrid (B3LYP/6-31 g(d):Amber) calculations on the PDE4 and PDE5 catalytic sites. Both sets of calculations held that a bridge water drifts away from the Zn2+ ion, in support of the original conclusion of Zhan and Zheng that the bridge species is a hydroxide ion.16 The quantum mechanical layer of Xiong et al.’s ONIOM hybrid calculations16
comprised exactly the same atoms of the best gas-phase model of Zhan and Zheng.15 Although the generation of the bridge hydroxide was not a subject of Xiong et al.’s study, they reported that during a certain partial optimization test of the PDE4 model, a bridge water splits to protonate the carboxylate group of Asp 392 and to form a bridge hydroxide.16 To date, no quantum mechanical studies of cAMP/cGMP binding to the PDE catalytic site nor any other aspects of the PDE hydrolysis mechanism have been reported. We have placed a model cyclic substrate in a truncated model of the PDE4D2 catalytic site,9 Figure 1 (pdb entry 1PTW), to carry out a quantum computational study of the hydrolysis reaction as it occurs inside the PDE catalytic site. Our computational model includes the model substrate, the bimetal
Mechanism of Cyclic Nucleotide Hydrolysis
J. Phys. Chem. B, Vol. 111, No. 17, 2007 4549
Figure 2. Hydroxide generation within the substrate-free PDE catalytic site. Optimized ONIOM structures are shown. The His 234/Glu 413 tandem is explicitly shown; for the purpose of clarity, the other 10 residues of the computational model (see Figure 1) are hidden from view. (a) Bridge water arrangement. A water molecule occupies the bridge position with r(O-Mg) ) 2.382 Å and r(O-Zn) ) 2.280 Å. W1 and W2 are coordinated to Mg2+ and Zn2+, respectively. W1, initially a water molecule, upon optimization has dissociated, yielding a hydroxide (r(OW1-Mg) ) 2.028 Å) and a proton which is bound to N of His 234; Glu 413 is protonated at the expense of Nδ. Subsequent proton shifts or ligand rearrangement can place the hydroxide at the bridge position. (b) Bridge hydroxide arrangement. A hydroxide ion occupies the bridge position with r(O-Mg) ) 2.298 Å and r(O-Zn) ) 2.090 Å. For comparison, the experimental distances from the unliganded PDE4B (chain B, pdb entry 1FOJ7) are 2.58 Å and 1.94 Å, respectively. W2, initially bound to Zn2+, upon optimization has drifted away to engage in a hydrogen bond with His 234. The energy of structure (b) is essentially equivalent to (a). W1 and W2 are displaced by the phosphate moiety when the substrate binds as in Figure 1.
system, the bridge hydroxide, 12 proximal residues, and waters of hydration, all treated quantum mechanically, to simulate the protein environment. In this study, we have characterized the bound substrate, the bound hydrolyzed product, and the transition state which connects them. Also of interest to us are the generation of the hydroxide at the bridge position and the role that the key proximal His 234/Glu 413 tandem, absent in the model of Zhan and Zheng15 and only treated by molecular mechanics in the model of Xiong et al.,16 plays in it. Computational Methods To study the catalytic mechanism of cyclic nucleotide hydrolysis, we have carried out calculations on an SGI Altix 350 Supercomputer using the Gaussian0317 implementation of the ONIOM hybrid method.18 The B3LYP/6-31g(d) model was applied to the high-level region, and the semiempirical PM3 model was applied to the low-level region, as indicated in Figure 1. Our system comprises a model cyclic substrate (charge ) -1) and a truncated model (charge ) 0) of the PDE4D2 catalytic site9 (pdb entry 1PTW, 2.30 Å res.). In our model substrate, the adenine/guanine base has been replaced with a hydrogen atom, Figure 1. Of the 12 residues explicitly represented in our model, Glu 413, Asp 392, Thr 345, Glu 304, Asp 275, His 274, His 238, and His 234 are absolutely conserved across the PDEs.7,8 The high-level region contains the metals, the substrate, all waters of hydration, the bridge hydroxide, and parts of seven residues: the imidazole rings of His 274, His 238, and His 234 and the carboxylates of Glu 413, Asp 392, Glu 304, and Asp 275. In our model catalytic site, most of the truncated amino acid residues are terminated at the alpha carbons as methyl groups. Partial optimizations were carried out with the metal ions frozen along with most residues of the site, while
the model cyclic substrate and the side chains of His 234, Glu 413, and Asp 392 were free to move, as were two waters of hydration on the near side of the Mg2+ ion, Figure 1. In the final stages of optimization of the transition state, however, it was necessary to freeze these two waters to achieve full convergence. The default density functional grid and default convergence criteria were used. The putative stable structures and transition state were confirmed by computing analytical frequencies. Estimates of ∆Gq and ∆G for the hydrolysis reaction were determined by vibrational frequency analysis at the ONIOM(B3LYP/6-31g(d):PM3) level and follow-up singlepoint ONIOM(B3LYP/6-311g(d,p):PM3) energies. To investigate the generation of hydroxide as the bridge ligand, we have also carried out ONIOM(B3LYP/6-31g(d): PM3) calculations for our model system of the substrate-free catalytic site with different sets of waters of hydration included (charge ) 0). Two important cases are illustrated in Figure 2, wherein all but the key His 234/Glu 413 pair of the 12 residues in the model are hidden from view. The cases we have considered are based primarily upon the crystal structure of unliganded PDE5A (pdb entry 2H40) as determined very recently by Wang et al.,19 which indicates that there are four strongly bound waters of hydration of Mg2+ and one water of hydration of Zn2+, in addition to the bridge hydroxide/water. Two waters, denoted in our models as W1 (bound to Mg2+) and W2 (bound to Zn2+), would be eventually displaced by the phosphate oxygens Oa and Ob of the substrate when it binds, see Figures 1 and 2. We have also considered cases on the basis of the original unliganded PDE4B structure (pdb entry 1FOJ) which, in contrast, indicates that W1 is present and that W2 is not.7 Although W1 is very weakly bound in chain A of the dimeric structure 1FOJ, it is strongly bound in chain B, and Xu
4550 J. Phys. Chem. B, Vol. 111, No. 17, 2007
Salter and Wierzbicki
Figure 3. Hydrolysis within the PDE catalytic site. The Zn2+ (dark blue) and Mg2+ (green) ions are shown as balls. Optimized ONIOM structures are shown. For the purpose of clarity, most of the computational model shown in Figure 1 is hidden from view. (a) Bound cyclic substrate. The electron-poor phosphorus atom is the center of a trigonal bipyramidal complex with d(O5′-Oa-Ob-P) ) -4.6°, r(O3′-P) ) 1.837 Å, and r(POhyd) ) 1.948 Å. (For comparison, computed B3LYP/6-31g(d) gas-phase values for the free cyclic substrate are d(O5′-Oa-Ob-P) ) -34.4°, r(O3′-P) ) 1.725 Å.) The phosphate oxygen coordination distances are r(Ob-Zn) ) 2.089 Å and r(Oa-Mg) ) 1.978 Å. Glu 413 is protonated at the expense of Nδ of His 234. The structure is the same as that of Figure 1. (b) Transition state. The geometry about the phosphorus atom resembles the hydrolysis product with regard to inversion, d(O5′-Oa-Ob-P) ) 18.6°. Ring opening is essentially complete with the P-O3′ bond broken. A stronger bond to the bridge hydroxide is present, r(P-Ohyd) ) 1.728 Å. A proton is midway between N of His 234 and O3′. The single imaginary vibrational mode evinces the concerted transfer of this proton while Nδ accepts a proton from Glu 413 and while the phosphorus center completes its umbrella inversion. The phosphate oxygen coordination distances are r(Ob-Zn) ) 2.076 Å and r(Oa-Mg) ) 1.972 Å. ∆Gq ≈ 3.5 kcal/mol with respect to a. (c) Bound product. The new tetrahedral center is established with d(O5′-Oa-Ob-P) ) 28.8° and r(P-Ohyd) ) 1.659 Å. O3′ and Nδ of His 234 are protonated. The phosphate coordination distances are r(Ob-Zn) ) 2.072 Å and r(Oa-Mg) ) 1.977 Å. For the process bound cyclic substrate f bound product, ∆G ≈ -3.4 kcal/mol.
et al. identified W1 as part of a distorted octahedral complex for Mg2+.7 A cocrystallized arsenate ion with variable occupancy near the bimetal system (see Figure 3 of ref 7) may be responsible for the apparently variable binding of W1 and the total displacement of W2 in their structure. We have carried out optimizations in which all waters (and hydroxide, if present) are free to move, in addition to the side chains of His 234, Glu 413, and Asp 392. Optimized structures were confirmed as stable minima by computing analytical vibrational frequencies. Results and Discussion We considered three possible scenarios for the generation of hydroxide at the bridge position in the substrate-free catalytic site. First, in the absence of waters of hydration W1 and W2 and with a water molecule placed in the bridge position, the bridge water readily dissociates as N of His 234 extracts a proton during energy optimization, leaving a bridge hydroxide. For reference, the resulting arrangement is as shown in Figure 1, if the substrate is removed. Overall, the free energy of the system drops by about 37 kcal/mol for the dissociation process, according to our ONIOM(B3LYP/6-31g(d):PM3) calculations; only a small barrier must be overcome as Glu 413 initially extracts a proton from Nδ of His 234. Second, we considered a case analogous to the simple model system of Zhan and Zheng15 in which W1 is present as in Figure 2a but W2 is absent. The presence of W1 is indicated by an oxygen atom found in this general position near Mg2+ in the crystal structures of unliganded PDE4B and unliganded PDE5A. During an energy optimization for this system, we found that W1 dissociates to produce a hydroxide as N of His 234 extracts a proton, while the bridge water remains in place, unlike the previously
mentioned results for the models of Zhan and Zheng15 and Xiong et al.16 The model system of Zhan and Zheng15 does not include the His 234/Glu 413 tandem, and naturally, this hydroxide generation cannot be seen in their model. The hybrid (QM/MM) model systems of Xiong et al.,16 although comprise the entire catalytic domains and have no frozen atoms, only include the His/Glu tandem as part of the molecular mechanics region of the hybrid calculation and consequently cannot realize this process either. Third, we considered a case in which W1 and W2 are both present, with W2 coordinated to Zn2+, as indicated by the very recent crystal structure of unliganded PDE5A.19 Again, dissociation of W1 occurs without a barrier during energy minimization, with a free-energy drop of about 21 kcal/mol; the optimized structure is shown in Figure 2a. We conclude that analogous to the action in a “catalytic triad”, N of His 234 extracts a proton from W1 (rather than from a serine hydroxyl) to yield a hydroxide ion, while Glu 413 extracts a proton from Nδ. We have not taken into account the full effects of the free energy of solvation in the true protein environment for this process, but our expectation is that partial solvation of the protonated histidine should enhance the free-energy drop. The hydroxide can move to the bridge position, see Figure 2b, via ligand rearrangement or by proton shift with an estimated net free-energy change of
