Computer Simulations Reveal Substrate Specificity ... - ACS Publications

Mar 17, 2016 - design targets early in the 1970s.2−4 Potent inhibitors of PNP may be useful ..... vicinity of N9 and in some cases the ether oxygen ...
0 downloads 0 Views 2MB Size
Download PDF
Article pubs.acs.org/biochemistry

Computer Simulations Reveal Substrate Specificity of Glycosidic Bond Cleavage in Native and Mutant Human Purine Nucleoside Phosphorylase Geir Villy Isaksen,† Kathrin Helen Hopmann,† Johan Åqvist,‡ and Bjørn Olav Brandsdal*,† †

Centre for Theoretical and Computational Chemistry, Department of Chemistry, Faculty of Science and Technology, University of Tromsø, N9037 Tromsø, Norway ‡ Department of Cell & Molecular Biology, Uppsala University, SE-75124 Uppsala, Sweden S Supporting Information *

ABSTRACT: Purine nucleoside phosphorylase (PNP) catalyzes the reversible phosphorolysis of purine ribonucleosides and 2′-deoxyribonucleosides, yielding the purine base and (2′deoxy)ribose 1-phosphate as products. While this enzyme has been extensively studied, several questions with respect to the catalytic mechanism have remained largely unanswered. The role of the phosphate and key amino acid residues in the catalytic reaction as well as the purine ring protonation state is elucidated using density functional theory calculations and extensive empirical valence bond (EVB) simulations. Free energy surfaces for adenosine, inosine, and guanosine are fitted to ab initio data and yield quantitative agreement with experimental data when the surfaces are used to model the corresponding enzymatic reactions. The cognate substrates 6aminopurines (inosine and guanosine) interact with PNP through extensive hydrogen bonding, but the substrate specificity is found to be a direct result of the electrostatic preorganization energy along the reaction coordinate. Asn243 has previously been identified as a key residue providing substrate specificity. Mutation of Asn243 to Asp has dramatic effects on the substrate specificity, making 6-amino- and 6-oxopurines equally good as substrates. The principal effect of this particular mutation is the change in the electrostatic preorganization energy between the native enzyme and the Asn243Asp mutant, clearly favoring adenosine over inosine and guanosine. Thus, the EVB simulations show that this particular mutation affects the electrostatic preorganization of the active site, which in turn can explain the substrate specificity.

P

nucleoside covering it from accessible solvent.7 The catalytic site residues comprise Tyr88, Phe200, Glu201, Asn243, and His257, which interact with the nucleoside and residues Ser33, His64, Arg84, His86, Ala116, Ser220, and two conserved water molecules hydrogen bonding to the phosphate group.7 From available crystal structures, it is found that the phosphate loop (residues 57−65) can be in the open or closed conformation, resulting in His64 being oriented outward from or inward to the active site, respectively.18,19 Molecular dynamics analysis has indicated that correlated motions between the phosphate loop and the Phe159* loop (residues 155−160) exist in the homotrimer and that these in the closed conformation could be important for stabilizing the catalytic residues.20,21 Despite being studied for decades, the catalytic mechanism of low-mm PNPs is still not adequately elucidated, and a number of key questions remain unanswered. First, the phosphate

urine nucleoside phosphorylase (PNP, EC 2.4.2.1) catalyzes the reversible phosphorolysis of ribonucleoside and 2′-deoxyribonucleoside derivatives of hypoxanthine, guanine, and many of their analogues.1 PNPs have been studied since the 1950s and were recognized as important drug design targets early in the 1970s.2−4 Potent inhibitors of PNP may be useful for targeting undesirable cell proliferation in Tcell cancers and in autoimmune diseases and as immunosuppressive agents, e.g., to prevent tissue transplant rejection.5 Crystal structures suggest that two main classes of PNP exist: the trimeric “low-molecular mass” (low-mm) class that is mainly found in mammals,6−10 but also in some microorganisms,11,12 and the hexameric “high-molecular mass” (highmm) class found in bacteria.13,14 Low-mm PNPs are highly specific for 6-oxopurine nucleosides and exhibit negligible activity for 6-aminopurine nucleosides, whereas high-mm PNPs have a broader specificity, including both 6-amino- and 6oxopurine nucleosides.15−17 Human PNP is a homotrimer with the catalytic site located near the subunit−subunit interface. From one adjacent subunit, Phe159* is oriented in the active site in contact with the © XXXX American Chemical Society

Received: December 16, 2015 Revised: March 11, 2016

A

DOI: 10.1021/acs.biochem.5b01347 Biochemistry XXXX, XXX, XXX−XXX

Article

Biochemistry catalytic role and in particular its protonation state are unclear, although the most relevant crystal structure indicates that it is doubly negatively charged.19 This is also verified from previous vibrational spectroscopic studies indicating that the phosphate group is dianionic in the active site of PNP.22 In spite of this, existing proposed reaction mechanisms are inconsistent in the sense that they employ a singly, doubly, or even triply protonated phosphate group.12,23,24 Second, the amino acids stabilizing the transition state and their role have not been completely elucidated. Important clues and insights regarding key residues based on crystal structures of transition-state inhibitors exist.25−27 Finally, the purine ring protonation state for the glycosidic bond cleavage is highly controversial. A positively charged ribonucleoside protonated at N7 has previously been proposed for the low-mm calf spleen enzyme.27 It has, however, been experimentally shown by fluorescence studies that the corresponding purine base binds to PNP in the anionic form in the reverse nucleoside synthesis reaction.28 Thus, in phosphorolysis, this implies that the ribonucleoside N7 is not protonated, resulting in an anionic base after glycosidic bond cleavage. Studies on human erythrocyte24 and calf spleen6 enzymes suggest that the negative charge is delocalized onto N7 and N9 of the anionic base. Another issue is that the rate-limiting step of some PNPs is known to be the release of purine.21,29,30 This complicates the matter as experimentally measured kcat values and corresponding thermodynamic parameters cannot be used directly to evaluate the glycosidic bond cleavage step of the PNP reaction mechanism, but only as an upper threshold for the activation free energies. Available pre-steady-state kinetics, however, provides valuable information regarding the activation free energies and demonstrates that the rate-limiting step occurs after glycosidic bond cleavage.21,29,30 In this study, we apply extensive empirical valence bond (EVB) simulations to investigate the glycosidic bond cleavage step of phosphorolysis in human PNP. We here present EVB free energy surfaces for the reference reactions fitted to density functional theory (DFT) data for inosine (INO), guanosine (GUO), and adenosine (ADO). Transferring the fitted EVB parameters to the enzyme model system reproduces the experimental activation free energy trends for INO, GUO, and ADO in both native and Asn243Asp mutated HsPNP and provides valuable insight into both the reaction mechanism and substrate specificity.

energies (kilocalories per mole) at a 298 K and 1 atm standard state. The DFT models contain the nucleophile (H2PO4− or HPO42−) and the full nucleoside (inosine, adenosine, or guanosine). It was assumed that the nucleoside is neutral in the initial state. Different conformers of all structures were evaluated, with the lowest conformers exhibiting a hydrogen bond between a ribose OH and a nitrogen of the purine base. This conformation is in line with previous DFT calculations by Schramm and co-workers on the reaction of inosine with HPO42−.38 EVB Reference Reaction. The phosphorolytic cleavage of the glycosidic bond driven by a single protonated phosphate as the nucleophile for inosine (INO), guanosine (GUO), and adenosine (ADO) was modeled with the EVB method.39,40 The reactant and the product state were described as an interconversion between two different resonance structures, ϕi, corresponding to each valence bond state (Figure 2, ϕ1 → ϕ2). The energetics associated with each resonance form was determined by a specific potential energy function, εi, describing the bonding arrangement and the charge distribution for each structure. The difference in εi between the valence bond states (Δε = ε1 − ε2) was utilized as the reaction coordinate. The diagonal elements of the effective EVB Hamiltonian describe the energies for each diabatic resonance structure, and the mixing between these states is reflected by the off-diagonal terms. Each diagonal energy function is given by an analytical force field of the form i i i Hii = εi = Ubnd + Unb,rr + Unb,rs + Ussi + α i

(1)

Here the subscripts bnd and nb are abbreviations for bonded and nonbonded interactions, respectively, whereas r and s denote the reacting fragments and the surroundings, respectively. The bond energies for the reacting atoms are described by the Morse bond potentials. Charges for the reacting atoms were assigned in accordance with the atomic electrostatic surface potential (ESP) charges obtained by singlepoint B3LYP/6-31G**+ calculations using Jaguar.41 The DFT calculations where performed using the standard Poisson− Boltzmann continuum solvation model (PBF) with water as the solvent (dielectric constant of 80.37, density of 0.99823 g/mL, and probe radius of 1.4 Å) and with the optimized gas-phase structure as the gas-phase reference energy. All reactant- and product-state geometries were optimized with the Polak−Ribier conjugate gradient minimization scheme prior to DFT calculations with MacroModel.42 The OPLS-2005 force field43,44 with water as the solvent was used for the minimization, and the convergence threshold, converging on the gradient, was set to 0.05. Remaining atom parameters and charges for the potential energy function, εi, were assigned in accordance with the standard force field (OPLS-AA). The last term of the Hamiltonian, αi, represents the intrinsic gas-phase energy of the given resonance structure with all fragments at infinite separation.39,40 The off-diagonal matrix elements, Hij, representing the adiabatic mixing of the VB states are typically described by a constant or an exponential function. Here we have applied a constant for describing the off-diagonal elements (Aij). The EVB potential energy surfaces for the reference reactions were fitted to the ab initio free energy surfaces by adjusting the off-diagonal term as well as the gas-phase shift αi. MD/EVB Simulation Details. Atomic coordinates for human purine nucleoside phosphorylase (HsPNP) in complex

■

METHODS DFT Method. All calculations were performed with the hybrid functional B3LYP31,32 as implemented in the Gaussian 09 package. 3 3 The polarizable continuum method IEFPCM34−36(solvent=water) and the Grimme empirical dispersion correction D237 were included in all calculations. Geometry optimizations were performed at the B3LYP-D2/6311G(d,p)/IEFPCM level of theory, followed by single-point calculations at the B3LYP-D2/6-311+G(2d,2p)/IEFPCM level to obtain improved electronic energies. Transition states were located through linear transit calculations, followed by unconstrained geometry optimizations. Thermodynamic parameters were obtained from frequency calculations at the same level of theory as optimizations. The latter were also employed to confirm the nature of the stationary states, with only positive eigenvalues for minima and one imaginary frequency for transition states. All reported DFT energies are Gibbs free B

DOI: 10.1021/acs.biochem.5b01347 Biochemistry XXXX, XXX, XXX−XXX

Article

Biochemistry with a transition-state analogue, Immucillin (ImmH), and phosphate were obtained from the crystallographic structure of Protein Data Bank (PDB) entry 1RR6.19 Coordinates for the phosphate loop (residues 57−65) oriented inward to the active site were obtained from PDB entry 4EAR.18 Enzyme−substrate complexes for inosine (INO), guanosine (GUO), and adenosine (ADO) were created by modifying ImmH. All MD/EVB simulations were conducted with the molecular dynamics package Q,45 where all input file generations and simulation analysis were performed with Qgui.46 All simulations were performed with spherical boundary conditions where the system was immersed in a spherical droplet of water molecules centered at the phosphate anion with a 25 Å radius for the reference and enzyme reaction. Water molecules were described using the TIP3P potential.47 The nonbonded potential was truncated at 10 Å for solvent−solvent interactions. Long-range electrostatics was treated using a multipole expansion method.48 Interactions between the reacting fragment and surroundings were not truncated, and they were thus allowed to interact with the entire system. All systems were heated from 1 to 298 K over 31 ps, using a stepwise scheme, followed by an equilibration period of 100 ps. SHAKE49 was used to constrain bonds and angles on solvent molecules. A time step of 1 fs was used for the production phase, and the temperature was maintained at 298 K using a weak coupling to an external bath. Each enzyme and water reaction free energy profile involved 500 ps and comprised 51 discrete FEP steps. All simulations were averaged over 10 and 100 independent simulations, resulting in a total of 5 and 50 ns of simulation time for the reactions in water and enzyme, respectively. Reorganization Energies. The EVB diabatic free energy profiles of the reactant and product state represent microscopic equivalents of the Marcus parabolas in electron-transfer theory50 and can be used to obtain the microscopic reorganization energy (λ). There are in principle two ways of estimating the total reorganization energy (see Figure S7). The first method is simply the energy of the diabatic free energy functional for state 1, Δg1, at the minimum (product state) of state 2, Δg2. The second method involves shifting the minima of the functional to the same heights and taking the resulting height of the intersection, the intrinsic barrier, as λ/4 (Figure S7). However, the free energy profiles of the diabatic states do not correspond to perfect parabolas, and the curvature between the states can be different. To compute the reorganization energies, we fitted the diabatic free energy profiles up to the intersection point to perfect parabolas by polynomial regression as described in ref 51. The reported reorganization energies in this work are taken as an average of the two methods described above (Figure S7).

activation parameters. However, before we can examine the details of the enzymatic reaction, a proper reference reaction in solution must be identified. Reference Reaction in Solution. Several alternative reaction mechanisms for the reversible phosphorolysis of purine nucleoside by PNP have been proposed over the past few decades. The reaction mechanism proposed by Erion et al. involves a proton transfer from H2PO4− to His86 prior to the nucleophilic attack, leaving a singly protonated phosphate nucleophile.24 In the QM/MM studies by Saen-Oon et al., a singly protonated phosphate was employed, but with N7 of GUO protonated prior to the chemical step (the protonation state of His86 is not mentioned and therefore thought to be neutral).52 H2PO4− is proposed to act as the nucleophile with a proton transfer to the glycosidic N(9) subsequent to ribose 1phosphate formation in the mechanism suggested by Tebbe et al.12 Another important point is the pKa value for N7 of guanine of 3.2, which makes it unlikely that protonation of this site occurs prior to phosphorolysis.53 Studies of the binding of guanine (the product formed upon GUO phosphorolysis) also show that it binds to PNP in the anionic state, further indicating that protonation (at N7 or N9) occurs after phosphorolysis.28 The reaction profiles for phosphate reacting with the nucleosides INO, GUO, and ADO at the DFT level [B3LYP/6-311G(d,p), including empirical dispersion corrections and IEFPCM solvation] have been calculated. In the calculations, the nucleosides are in the expected neutral form. The DFT energy profile for purine phosphorolysis has not been reported previously; however, Schramm and co-workers have reported DFT-optimized transition-state structures for the reaction between HPO42− and INO.38 The lowest-lying conformers obtained here for this reaction involve a hydrogen bond from the ribose 5′-OH to purine N3 (Figure 1) and are

RESULTS AND DISCUSSION The main goal of the work presented here is to generate an EVB description of the free energy surface for the PNPcatalyzed glycosidic bond cleavage that allows us to reliably address issues related to the catalytic mechanism and substrate specificity. This class of enzymes has been studied for more than 50 years, but the reaction mechanism is still highly controversial. An EVB description of the PNP reaction is consequently of considerable interest. The calibrated free energy surfaces can be used for quantitative investigations of the PNP enzyme mechanism in greater detail and on a larger scale by computing mutational effects as well as thermodynamic

thus similar to the previously reported structures.38 The ring puckering in the reactants can be described as C2′-endo. For the reaction with the three nucleosides, INO, GUO, and ADO, the optimized scissile bond lengths are very similar, with both breaking and forming bond lengths of ∼2.4 Å in the transition state (Figure 1). These distances are significantly different from those obtained in earlier DFT studies on the reaction of INO with HPO42− (2.74 and 2.08 Å for the forming O−C bond and the breaking C−N bond, respectively).38 This might be due to differences in the computational methods (earlier calculations did not include solvation or empirical dispersion corrections).38 The C1′−O bond distance decreases from 1.42 Å at the

Figure 1. Optimized transition-state geometries [B3LYP-D2/ IEFPCM/6-311+G(d,p)] for purine phosphorolysis with HPO42−: (A) inosine, (B) guanosine, and (C) adenosine.

■

C

DOI: 10.1021/acs.biochem.5b01347 Biochemistry XXXX, XXX, XXX−XXX

Article

Biochemistry

reactions were taken as an average of the parameters obtained from these 10 simulations. HsPNP-Catalyzed Glycosidic Bond Cleavage. The enzyme-catalyzed reactions have been modeled using trimeric HsPNP. It can be noted that the possibility that the enzyme also displays activity in the monomeric form exists.54 However, a recent study shows that PNP monomers are unstable in solution, and consequently, packing of monomers into trimers is necessary for enzyme stability.55 The phosphate loop was here modeled in the closed conformation, resulting in His64 being organized in the active site within interacting distance of Ser33 and His86. All EVB simulations were repeated 100 times, resulting in a total simulation time of 51 ns for each substrate. Guanosine. Steady-state kinetics shows that wild-type HsPNP catalyzes phosphorolysis of guanosine at a rate of 28 ± 2 s−1, corresponding to an activation free energy of ∼15.5 kcal/mol at 298 K.15 However, as guanine release is indicated to be rate-limiting, these values can serve as only an upper threshold for the activation free energy of the chemical step.21,29,30 According to pre-steady-state kinetics with GUO as the substrate, the rate for the chemical step involving C1′− N9 bond cleavage is 154 s−1, corresponding to an activation free energy of ∼14 kcal/mol.21 The average activation and reaction free energies of 12.8 and −7.1 kcal/mol, respectively, obtained from the EVB simulations are thus fully compatible with the experimental rate constants (Table 1). HsPNP is able to lower the activation and reaction free energies for GUO by an astonishing 17.8 and 16.3 kcal/mol, respectively, compared to those of the reference reaction in water (Figure 2). The major contribution to catalysis originates from a significant reduction in the electrostatic EVB potential activation energy in HsPNP compared to that of the reference reaction in water. It is, however, well established that the enzyme’s catalytic power originates from optimized electrostatic preorganization along the reaction pathway with respect to the corresponding reaction in water.56 This electrostatic stabilization is a result of smaller reorganization energies, which enzymes have paid for during the folding process.40,56 Analysis of the EVB transition-state structures further reveals an extensive H-bonding network in the active site stabilizing GUO and the phosphate group (Figure 3). Some of the interactions, in particular those stabilizing the phosphate group, are found to be critical for lowering the reaction free energy. Simulations randomly disrupting these interactions (see below) typically increase the activation and reaction free energies by several kilocalories per mole. In particular, we found the interaction of the phosphate oxygen with the backbone NH groups of Ser33 and Ala116 to be crucial for the stability (Figure 3). In our initial simulations necessary for optimizing the EVB starting structure, these interactions were in some runs randomly broken, leading to a domino effect in which the phosphate group tumbles and disrupts many of the stabilizing interactions illustrated in Figure

reactant to 1.28 Å at the transition state (TSinosine), in agreement with a postulated oxocarbenium character of the transition state.24 Overall, the computed reaction type with HPO42− as the nucleophile can be described as SN2 (ANDN). The computed barriers [B3LYP-D2/6-311+G(2d,2p)/ IEFPCM] for the reaction of the three nucleosides with HPO42− are 28.5 kcal/mol for INO and 30.6 kcal/mol for GUA and ADO (Table S1). These values are functional-dependent, and for example, PBE predicts a lower value of 23.7 kcal/mol for INO. Changing the empirical dispersion correction from D2 to D3 has a minor effect, with a computed barrier of 27.0 kcal/ mol for INO. Note that the formed product in all reactions is the phosphorylated ribose and the anion of the purine base. The endergonic reaction energies (Table S1) thus reflect only formation of the purine anion, not the final neutral purine. As discussed above, there is some ambiguity regarding the protonation state of the phosphate nucleophile. If H2PO4− is employed as a nucleophile for the reaction with INO, the B3LYP-D2 barrier increases by 2.6 kcal/mol, from 28.5 to 31.1 kcal/mol (see Methods for details of the DFT calculations). Also, the scissile bond distances change substantially, with 2.28 and 2.68 Å for the forming O−C bond and the breaking C−N bond, respectively, giving the transition state with H2PO4− some SN1 character. This reflects the weaker nucleophilicity of H2PO4− compared to that of HPO42−. The reaction energy is substantially more endergonic with H2PO4− than with HPO42− (Table S1). On the basis of the DFT calculations described above, EVB reference reactions were run for the three purine bases together with a singly protonated phosphate in water (Figure 2). The

Figure 2. Calculated reaction free energy profiles at 298 K for the glycosidic bond cleavage step in water and in HsPNP with adenosine, guanosine, and inosine as substrates.

reaction free energy surfaces were fitted to ΔG⧧ values of 30.6, 30.6, and 28.5 kcal/mol and ΔG0 values of 9.4, 9.2, and 7.2 kcal/mol for ADO, GUO, and INO, respectively. A total of 10 reference reaction simulations were utilized for each substrate, and the final EVB parameters applied in the enzyme-catalyzed

Table 1. Average Activation and Reaction Free Energies (kilocalories per mole) Calculated at 298 K for the Chemical Step (ϕ1 → ϕ2) for the Reaction in Native and Asn243Asp HsPNP for Inosine (INO), Guanosine (GUO), and Adenosine (ADO) INO GUO ADO a

⟨ΔG⧧⟩

⟨ΔG0⟩

ΔG⧧expta

⟨ΔG⧧⟩N243D

⟨ΔG0⟩N243D

ΔG⧧expt,N243L

13.0 ± 0.2 12.8 ± 0.1 20.2 ± 0.2

−7.0 ± 0.3 −7.1 ± 0.2 0.9 ± 0.5

15.1 ± 0.1 15.5 ± 0.1 21.0 ± 0.1

15.9 ± 0.2 16.8 ± 0.4 14.7 ± 0.2

−2.3 ± 0.3 −4.3 ± 0.5 −6.9 ± 0.3

17.0 ± 0.1 16.8 ± 0.2 16.0 ± 0.3

ΔG⧧expt computed from kcat (s−1) values reported by Stoeckler et al.,15 where substrate release is thought to be rate-limiting. D

DOI: 10.1021/acs.biochem.5b01347 Biochemistry XXXX, XXX, XXX−XXX

Article

Biochemistry

pathway but also the backbone interactions of Ser33 and Ala116 stabilizing the phosphate group’s optimal orientation. In addition, the phosphate O3 atom is stabilized by the backbone carbonyl oxygen of Ala116 via a water molecule [W1 (Figure 3)] and by an H-bond from the Ser220 hydroxyl. When the interactions of the ribose 2′- and 3′-hydroxyls with the phosphate O3 and O4 atoms, respectively, are counted, a total of nine H-bonds are involved in stabilizing the phosphate group transition state (Figures 3 and 4). The free energies are very sensitive with respect to this extensive network of interactions. Average bond lengths of 2.4 and 2.9 Å are observed for the HO3PO−C1′ and C1′−N9 bonds, respectively, at the transition state for GUO in HsPNP. This indicates that the glycosidic bond (C1′−N9) is almost fully dissociated prior to the nucleophilic attack. In the ground state, the predominant ribose ring conformation in the enzyme is C4′-endo, whereas the transition-state configuration also shows development of a C2′-endo conformation, which previously has been reported by Saen-Oon et al.52 They computed the transition state with GUO as the substrate, although with N7 protonated, and found that the leaving group departs before participation from the phosphate. The reported transition-state bond distances are 2.76 ± 0.01 and 2.49−2.63 Å for the O4−C1′ and C1′−N9 bonds, respectively. However, these results cannot be expected to be directly comparable to our calculations due to the GUO charge difference with respect to the protonated N7. It must be emphasized that fluorescence experiments have indicated that guanine binds in the anionic form in the reverse reaction, and consequently, any protonation must take place after the glycosidic bond cleavage.28 The ribose ring is at the transition state stabilized by Hbonds between the 5′-OH and His257, Tyr88, and 4′-OH and between the main chain HN group of Met119 and 3′-OH (Figure 3). However, in many of the EVB simulations, the 5′OH−His257 interaction is lost some time after the transition state toward the product state, where the 5′-OH moves into the vicinity of N9 and in some cases the ether oxygen of the ribose ring, as well. This observation is not unexpected considering the high negative charge that develops on guanine N9 and N7 in the intermediate state (ϕ2). The base is observed with two perfectly oriented H-bonds from N1 and N2 to Glu201, as illustrated in Figure 3. Glu201 is also linked to O6 through a water molecule (W2 in Figure 3). This H-bond pattern has previously been reported for binary and ternary complexes of bovine PNP with hypoxanthine.6,57 Moreover, mutational studies have demonstrated that Glu201 plays a crucial role in the catalytic activity.15 In the case of guanosine, the mutation of Glu201 to Ala decreases the catalytic efficiency by 6 orders of magnitude. Solution studies

Figure 3. Snapshot of the transition state for guanosine in HsPNP as observed from the EVB simulations. The red dotted lines represent the reaction coordinate (O4−C1′−N9).

3. The activation free energy for these runs typically ranges from 18 to 22 kcal/mol compared to ∼13 kcal/mol for the simulations where the interaction to Ser33 and Ala116 is maintained. It should be emphasized that this rarely occurs with the optimized starting structures utilized for the final simulations. The EVB simulations indicate that Ser33 together with His64 and His86 play critical roles in orienting the phosphate group for the glycosidic bond cleavage step. In the reactant state, there are H-bonds from Ser33 and His64 to the phosphate oxygens and between His86 and the phosphate proton (Figure 4). The interaction of phosphate proton with His86 stays intact throughout the entire simulation, but as the reaction proceeds toward the transition state, the His64 H-bond is shifted from the phosphate to the Ser33 hydroxyl as illustrated in Figures 3 and 4. This shift stabilizes the interaction of Ser33 with the nucleophilic phosphate oxygen (O4) as the phosphoester bond forms, allowing Ser33 to move in parallel to the phosphate along the reaction coordinate. Toward the product state, the Ser33 hydroxyl moves within interacting distance of both the phosphate and the ribose ether oxygen (Figure 4). Interestingly, a recent study has identified Ser33-His64-His86 as the catalytic triad and proposed this to stabilize the phosphate in the active site.18 The initial EVB simulations optimizing the EVB starting structure show that these interactions are disrupted if the interactions of the phosphate oxygen with the backbone NH groups of Ser33 and Ala116 are broken, leading to significantly higher free energies, as mentioned above. This demonstrates the importance of not only the stabilization of the phosphate along the reaction

Figure 4. Snapshot of the S33−H64−H86 triad interactions with the phosphate group around the reactant state (RS), transition state (TS), and product state (PS) as observed in the EVB simulations. E

DOI: 10.1021/acs.biochem.5b01347 Biochemistry XXXX, XXX, XXX−XXX

Article

Biochemistry

IEFPCM and D2 (see the computational details)] and found that the imino tautomer is ∼11 kcal/mol higher in energy, which in principle rules out the possibility for this tautomer. The computed activation and reaction free energies for ADO are 20.2 and 0.9 kcal/mol, respectively, and in excellent agreement with the experimental kcat of 0.0024 ± 0.0002 s−1, corresponding to a ΔG⧧ of ∼21.0 kcal/mol15 (Table 1). It is rather remarkable that our EVB simulations are able to capture the fact that HsPNP displays negligible activity for 6aminopurine and not 6-oxopurine nucleosides. Furthermore, the EVB simulations indicate that the low activity originates from the chemical step where the ΔG⧧ for ADO is ∼7 kcal/mol higher than those of GUO and INO (Figure 2 and Table 1). It is therefore likely that the rate-limiting step with ADO as the substrate is the glycosidic bond cleavage and not adenine release. Nevertheless, HsPNP is still able to lower the activation and reaction free energies by 10.4 and 8.7 kcal/mol, respectively, with respect to that of the reference reaction in water. This clearly demonstrates that HsPNP catalyzes the chemical step with ADO as the substrate, but at rates much lower than those with GUO and INO. The transition-state structure shares many of the same interactions as described above for GUO and INO, but with three distinct differences. First, the Ser33 hydroxyl is not interacting with the phosphate oxygen (O4) but is quickly shifted in the direction of His257 and comes within interacting distance of the ribose ether oxygen (Figure S2). This movement is also observed in some of the simulations with GUO and INO, but rather toward the product state, and is likely a result of the phospho−ester bond distance (∼1.3 Å) potentially bringing the Ser33 hydroxyl and 4′-OH too close together. Interestingly, the HO3PO−C1′ and C1′−N9 bond distances are 2.3 and 2.9 Å, respectively, indicating a phospho− ester bond more established than those with GUO and INO at the transition state. Second, the base N1 atom lacks a proton, which consequently leaves no H-bonding options to Glu201. Instead, W2 is observed as an H-bond donor to N1, and Glu201 is shifted away from the base (Figure S2). Finally, Asn243 has flipped compared to GUO and INO and makes Hbond interactions with both N7 and the N6 amino group. Even with the flipped Asn243, W3 stays within interacting distance of N7, as observed with INO and GUO. The two H-bonds between N243 and ADO (N6 and N7) are not optimal as the two amino groups come quite close to each other (Figure S2). Moreover, the zigzag pattern described above is not observed for ADO, and the low activity could potentially be a direct result of this. It is also noteworthy that the computed average reorganization energies listed in Table 2 clearly demonstrate that HsPNP is less preorganized with respect to the enzyme dipoles for the chemical step with ADO as the substrate. Reversing the 6-Oxopurine and 6-Aminopurine Activity. It is rather remarkable that the presented EVB

have also shown that there is a lack of activity for substrates that are not protonated at N1, indicating that Glu201 plays a key role in substrate recognition.16,58,59 In contrast, Tebbe et al. proposed that it is rather Asn243 that is responsible for substrate recognition and that the main role of Glu201 is stabilization of the purine intermediate in the transition state.12 This differs from the finding of Erion et al., who suggested that the main task of Asn243 is in stabilizing the transition-state negative charge delocalized between N7 and N9.24 Interestingly, Asn243 is in our simulations found to donate an H-bond to O6 and not to N7, which accepts an H-bond from a water molecule instead (W3 in Figure 3). This observation is similar to the Cellulomonas PNP ternary complex with 8-iodoguanine in which Asn246 (Asn243 in HsPNP) has no direct contact with the base but is bridged via a water molecule to O6.12 This observation is in particular interesting because the proposed role of Asn243 in stabilization of a negatively charge purine intermediate is not fully consistent with experimental data, which, for example, does not explain why ADO has negligible substrate activity.15 The EVB simulations show a water molecule (W3) stabilizing the N7 negative charge, whereas Asn243 is stabilizing O6 and part of a zigzag pattern of hydrogen bonds (N243−O6−W2−E201−N1−N2) similar to that reported by Tebbe et al.,12 which may be the key for substrate recognition. Inosine. To our knowledge, there are no experimental data available for the glycosidic bond cleavage step with INO as the substrate. However, the experimentally determined kcat of 57 ± 5 s−1, corresponding to a ΔG⧧ of ∼15.1 kcal/mol, for INO is very similar to that of GUO.15 The calculated activation and reaction free energies of 13.0 and −7.0 kcal/mol (Table 1) for the chemical step are, however, in excellent agreement with the experimental rates, assuming that substrate release is ratelimiting. The enzyme is thus able to lower the activation and reaction free energies by 15.5 and 14.2 kcal/mol, respectively, with respect to that of the reference reaction in water, which is comparable to the results for GUO. The EVB simulations further demonstrate that INO shares the exact same interactions at the transition state as GUO (Figure S1), with the exception of the N2−Glu201 interaction that is not an option (see the discussion above). In particular, the zigzag H-bond pattern (see the discussion above) involving Asn243, O6, W2, N1, and Glu201, which is potentially responsible for the substrate specificity, is perfectly conserved in all the EVB simulations with INO (Figure S1). Given that both INO and GUO are natural substrates of HsPNP (6oxopurines), it is not surprising that they share the same interaction patterns and have similar activation and reaction free energies. The computed bond distances at the transition state are 2.4 and 2.8 Å for the HO3PO−C1′ and C1′−N9 bonds, respectively, which is comparable to that of GUO. However, the glycosidic bond is ∼0.1 Å shorter at the transition state, indicating that the nucleophilic attack potentially happens earlier with INO as a substrate than it does with GUO. Adenosine. Solution studies have demonstrated that lowmm PNPs display low activity for substrates that are not protonated at N1.16,58,59 Because native HsPNP shows detectable activity for ADO, it could be suggested that it adopts the imino tautomer with a proton at N1 in the active form. However, as the population of the imino tautomer is