J. Phys. Chem. B 2010, 114, 15673–15686
15673
Do Halide Motifs Stabilize Protein Architecture? Peng Zhou,† Feifei Tian,‡,§ Jianwei Zou,| Yanrong Ren,⊥ Xiuhong Liu,† and Zhicai Shang†,* Department of Chemistry, Zhejiang UniVersity, Hangzhou 310027, China, College of Bioengineering, Chongqing UniVersity, Chongqing 400044, China, Key Laboratory for Molecular Design and Nutrition Engineering, Ningbo Institute of Technology, Zhejiang UniVersity, Ningbo 315100, China, Department of Biological and Chemical Engineering, Chongqing Education College, Chongqing 400067, China, and Center for Heterocyclic Compounds, Department of Chemistry, UniVersity of Florida, GainesVille, Florida 32611, United States ReceiVed: June 8, 2010; ReVised Manuscript ReceiVed: September 15, 2010
Halide anions are traditionally recognized as the structure maker and breaker of bulk water to indirectly influence the physicochemical and biological properties of biomacromolecules immersed in electrolyte solution, but here we are more interested in whether they can be structured in the protein interior, forming that we named “halide motifs”, to stabilize the protein architecture through direct noncovalent interactions with their context. In the current work, we present a systematical investigation on the energy components in 782 highquality protein halide motifs retrieved from the Protein Data Bank (PDB), by means of the continuum electrostatic analysis coupled with nonelectrostatic considerations, as well as hybrid quantum mechanical/ molecular mechanical (QM/MM) examination. We find that most halide motifs (91.6%) in our data set are substantially stabilizing, and their average stabilization energy is significantly larger than that previously obtained for sophisticated protein salt bridges (-15.16 vs -3.66 kcal/mol). Strikingly, nonelectrostatic factors, especially the dispersion potential, rather than the electrostatic aspect, dominate the energetic profile of the pronouncedly charged halide motifs, since the expensive cost for electrostatic desolvation penalty requires being paid off using the income receiving from the favorable Coulomb interactions during the motif formation. In addition, all the energy terms involved in halide motifs, regardless of their electrostatic or nonelectrostatic nature, highly depend on the degree of the motif’s burial in the protein, and the buried halide motifs are generally associated with a high stability. The results presented herein should be of valuable use in establishing a knowledge framework toward understanding the functional implications underlying anion structured in a biological molecule. 1. Introduction Inorganic ions were found as early as a century ago to exert a variety of biological effects on protein.1 Originally, it was thought that an ion’s influence on protein properties was caused at least in part by “making” or “breaking” bulk water structure.2 Recent time-resolved and thermodynamic studies of water molecules in salt solution, however, suggested that, instead of remodeling water structure through ions, direct ion-protein interactions as well as the ionic interactions with water molecules that are bound to the proteins seem to be also responsible for these effects.3 Previously, the interaction of metal cations with protein moieties by coordinate bonding has been well-characterized in the structural biology community.4,5 In contrast, the ubiquitous anions in the biological world, especially the small, hard halogen ion series, i.e., F-, Cl-, Br-, and I-, which were identified as a good hydrogen-bond acceptor and a friendly ion-pair partner in small molecular adducts,6,7 are traditionally recognized as counterions or co-ions and their direct interactions with proteins, as well as the effect of these * To whom correspondence should be addressed. E-mail: shangzc@ zju.edu.cn. Phone: +86-0571-87952379. Fax: +86-0571-87951895. † Zhejiang University. ‡ Chongqing University. § Ningbo Institute of Technology. | Chongqing Education College. ⊥ University of Florida.
interactions on protein functions, have long been underappreciated in the field of biology. It is well-known that the electrolyte solutions of halide salts have a fundamental impact on the various properties of the immersing proteins, ranging from physicochemical aspects such as solubility, crystallizability, isoelectric point, and surface tension, to biological facet such as stability, allosteric propensity, and catalysis activity.8,9 In recent years, a number of experimental and theoretical works have been addressed to elucidate the molecular mechanism of the specific interactions of proteins with dissolved halide anions.10 It was found that, for example, the small anions such as F- are prone to pair with charged groups, whereas larger ones such as I- are more likely to be bound on hydrophobic patches of the protein surface,11,12 and many ion-specific phenomena like this could be readily explained by incorporating ionic dispersion potentials into classical double-layer theory.13-15 Although these studies have provided an atomistic insight into the global effects of solvent ions imposing on the apparent properties of protein, the details about the local direct interactions of individual halide ions with protein moieties, in particular with those of strong halophilicity such as polar hydrogen atoms and basic groups, still remained largely unexplored. Previously, Ko and Jo employed the CHARMM force field to dynamically simulate halide conduction in the ion channel of Escherichia coli ClC protein and demonstrated that the chloride ion can be located metastably at the dehydrating pore of ClC protein without water coordination, and this
10.1021/jp105259d 2010 American Chemical Society Published on Web 11/04/2010
15674
J. Phys. Chem. B, Vol. 114, No. 47, 2010
Zhou et al.
Figure 1. Crystallographic evidence showing that the role of water molecules in mediating the hydrogen-bond network in biological molecules can be functionally replaced by halide ions. (a) Usually, a conserved water molecule (Ow) is located in the active pocket of HIV-1 protease to mediate the hydrogen-bond network with its cognate substrates and noncognate inhibitors (PDB ID: 2cej). (b) However, there has an exception that a chloride ion (Cl-) is observed at the water’s position in the complex of HIV-1 protease with its nonpeptide inhibitor UCSF8 (PDB ID: 1aid). This figure was modified from our recent publication.18
stabilization energy mainly stems from the interactions of the ion with surrounding polar and nonpolar hydrogen atoms protruding from the protein residues.16 Very recently, by exhaustively surveying all the protein crystal structures deposited in the current Protein Data Bank (PDB),17 we have found a striking magnitude (>10 000) of halides located in the interior or attached at the surface of proteins, from which we identified more than 6000 that we named halogen-ionic bridges that show a potential role in conferring stability and specificity for the structure of proteins and their complexes with small ligands and nucleic acids.18 In addition, there was also various experimental evidence indicating that some proteins, many of which are membrane proteins or ion channels, possess specific halidebinding sites.19-22 For example, a crystallographic study revealed that a Cl- is functionally bound between the residues Ile50/ Ile50′ of HIV-1 protease and the protonated tertiary amine of its nonpeptide inhibitor UCSF8, a haloperidol derivative which strongly inhibits both wild type and mutant HIV-1 proteases (Figure 1b).23 It is known that, however, this Cl- position is commonly occupied by a conserved water molecule (Figure 1a).24 If considering that halide cofactors are ubiquitous in the protein world and that their marked electrostatic characteristics must incur intensive interactions from the protein context, then we are interested in whether the structured halide ions generally stabilize protein architecture and how much free energy is required to move a halide ion from the solvent (free state) to the surface, interface, or interior of a protein and to bind it there (structured state). In fact, the analogous questions have also arisen previously for protein salt bridges,25 and both experimental and computational estimates of the free energy contribution of the salt bridges range from being stabilizing,26-28 to being insignificant,29,30 to being destabilizing.31,32 However, to the best of our knowledge, no such studies have been systematically performed on the structured motifs involving halide anions in proteins, albeit halide motifs share the very similar physicochemical features with pairing salt bridges; for instance, both of them are charged entities and their formations would incur
a noticeable desolvation penalty. In the current work, we aim to preliminarily address the issues relating to the stability and energy contribution of halide motifs casting to protein architecture. First, we describe a thermodynamic cycle based method by solving a nonlinearized Poisson-Boltzmann equation to separately calculate different electrostatic terms (including Coulomb interaction and desolvation penalty) associated with the formation of a halide motif in a protein matrix. Further, this method is coupled with the London formula,33 MahantyNinham theory,34 and empirical surface area model35 to account for not only the electrostatic effect but also nonpolar (nonelectrostatic) contributions from dispersion potential, dispersion selfenergy, and cavity energy. Subsequently, we use a procedure that we named energy decomposition analysis (EDA) to systematically investigate the energy components of 782 highquality halide motifs recruited from the current PDB-deposited protein crystal structures. Yet as a complementary approach of the empirical EDA method, the rigorous ONIOM-based QM/ MM scheme36 is employed to examine the structural feature and energetic property for the binding of halide ions to haloalkane dehalogenase within the whole protein framework. We also pick several specific interesting motifs to give relevant discussions. This study would help to clarify how halide ions participate in shaping protein architecture and how they contribute to the stability and specificity of protein folding and binding. 2. Materials and Methods 2.1. Definition of a Halide Motif. Crystallographic studies using X-ray and neutron diffractions have provided a large quantity of biomolecular models. These models include a number of atomic sites, reflected in the electron density maps as accumulative peaks. The necessary condition for presenting a clear peak in the density map is the repeated occurrences of a corresponding atom at the periodic crystal lattice. Therefore, the explicit providence of a halide ion in the crystal structure file (or PDB file) implies that this ion is conserved at the binding site in its parent protein. Here, we call the structural entity of
Do Halide Motifs Stabilize Protein Architecture?
J. Phys. Chem. B, Vol. 114, No. 47, 2010 15675
Figure 2. Some examples of halide motifs in proteins and at the interfaces of protein complexes with small ligands and nucleic acids. (a) A bromide motif in the protein interior (PDB ID: 1doc). (b) An iodide motif at the protein-protein interface (PDB ID: 2vgz). (c) A chloride motif at the protein-rRNA interface (PDB ID: 1jj2). (d) A chloride motif at the protein-small ligand interface (PDB ID: 2j90). Halide ion is shown as a ball; the biomolecular atoms, groups, and residues that directly interact with the halide ion are shown in stick style; the direct interactions (such as hydrogen bonding) are represented by green broken lines.
a halide anion and all the atoms, groups, and residues that directly or indirectly interact with the halide ion within a complete biological unit recorded in the protein PDB file as a “halide motif” (Figure 2). In a halide motif, the central ion is structured as compared to those being free in solution and acts as a mediator or adaptor to modulate the local configuration of protein structure. 2.2. Selection of High-Quality Halide Motifs from the PDB. Recently, we have performed an exhaustive survey of all the crystal structures deposited in the PDB and identified more than 10 000 halide ions that are bound in biological molecules through noncovalent chemical forces.18 From those we herein only collected those present in monomeric proteins or protein complexes and with high-quality (resolution e 1.8 Å), reliable validation (R-factor e 2.0), low uncertainty (B-factor e 30 Å2), and sufficient presence (occupancy ) 100%). To further reduce the redundancy of selected samples, for each protein only one halide ion with the least B-factor value was considered. As a result, 782 high-quality halide motifs were extracted from the PDB to define a distinct data set for subsequent analysis (provided in Supporting Information, Table S1). The protein structures of these selected halide motifs were treated with
following procedure: remove water molecules, metal ions, and other cofactors, except the halide ions under consideration; repair missing side chains of protein residues with the SCWRL program;37 add hydrogen atoms to protein residues using the REDUCE38 (for uncharged moieties) and PROPKA39 (for titratable groups, at pH 7.0) programs. REDUCE was tested in our previous work to have a high precision in reproducing the hydrogen atom’s positions determined by neutron diffraction.40 2.3. Energy Decomposition Analysis (EDA). Formation of a halide motif in protein can be regarded as a process during which the halide ion-protein interaction is established (Figure 3b) instead of the original interaction of halide ion/protein with water (Figure 3a). In this process, two kinds of physicochemical forces, i.e., electrostatics and dispersion, are thought to contribute significantly to the change in systematic free energy (∆GTOT). For a long time, the electrostatic effect has been used solely to describe the thermodynamic behavior of halide ions conducting in colloidal and biological systems.41-43 In recent years, however, Ninham and Bostro¨m et al. demonstrated that dispersion potential also plays an important role in many ionrelated chemical and biological processes, especially for those involving the anions with a large, diffuse, and polarizable
15676
J. Phys. Chem. B, Vol. 114, No. 47, 2010
Zhou et al.
Figure 3. Schematic representation of the halide motif formed in protein. Initially, the Cl- (green) is in dissolved state and can move freely in solvent (cyan) (a). After the formation of halide motif, this Cl- is bound at the halide-binding site of protein (yellow) and becomes structured (b). During this process, the water molecules around the Cl- and infused in the halide-binding site are drained, which would incur significant desolvation penalties due to the intrinsic electrostatic and dispersion interactions between the Cl-/protein and water.
Figure 4. A organization tree showing the subordinate relation of the energy components involved in the formation of a halide motif. The change in systematic free energy (∆GTOT) is composed of electrostatic (ele∆Gtot) and nonelectrostatic (nes∆Gtot) contributions. Both electrostatic and nonelectrostatic contributions can be further attributed to two aspects as direct nonbonded interaction and indirect desolvation effect. In the leaves of this tree the words in red and blue represent, in general, favorable (∆G < 0) and unfavorable (∆G > 0) contributions to systematic free energy during the binding of a halide ion to protein, respectively.
electron cloud system.13-15 In addition to electrostatic and dispersion forces, here we also considered the cavity energy contribution to the ∆GTOT, which accounts for the energy demand for creating a void cavity in solvent to accommodate a solute.44 In this study, conformation changes associated with the halide motif formation were ignored; we assumed that the protein structure is identical before and after binding by the halide ion. This is because conformation changes accompanying the binding are very complicated and thus unpredictable and also because our QM/MM analysis manifested that the impact of halide ion on protein configuration is considerably modest (vide post). Here, we decompose the ∆GTOT into the different energy terms that can be calculated directly using available methods. As shown in Figure 4, ∆GTOT, the change in systematic free energy upon the formation of a halide motif in protein, is composed of electrostatic (ele∆Gtot) and nonelectrostatic (nes∆Gtot) contributions, and both of them can be further attributed to two aspects such as direct nonbonded interaction (ele∆Gint and nes ∆Gint) and indirect desolvation effect (ele∆Gdslv and nes∆Gdslv).
For the electrostatic facet, direct nonbonded interaction gives rise to, in general, favorable free energy contribution (ele∆Gintcoul) due to the Coulomb interaction of the negatively charged halide ion with the charges in protein, whereas the indirect desolvation effect incurs an unfavorable free energy penalty (ele∆Gdslvplt) due to the charged halide ion, as well as the polar/charged protein moieties comprising the halide-binding site, being buried in the low dielectric protein interior from a high dielectric solvent (water) during the binding. For the nonelectrostatic aspect, direct nonbonded interaction arises from the favorable dispersion potential (ele∆Gintdisp) between the polarizable halide ion and its neighboring protein atoms in binding state, and an indirect desolvation effect is made up of favorable cavity energy contribution (nes∆Gdslvcav) due to the reduction in solventaccessible surface areas (ASAs) of halide ion and protein during the binding and favorable contribution from the dispersion selfenergy (nes∆Gdslvself) of the halide ion due to its move from a water medium to the protein contextsthe latter can provide a lower dispersion potential for the halide ion than the former. In the following parts, we will describe the methods used for
Do Halide Motifs Stabilize Protein Architecture?
J. Phys. Chem. B, Vol. 114, No. 47, 2010 15677
calculating these energy components (i.e., the five energy terms at the leaves of the organization tree shown in Figure 4). As aforementioned, the contribution to the change in systematic free energy (∆GTOT) can be broken down into five terms, i.e., Coulomb interaction between halide ion and protein in binding state (ele∆Gintcoul), electrostatic desolvation penalty on halide ion and protein during the binding (ele∆Gdslvplt), dispersion interaction between halide ion and protein in binding state (nes∆Gintdisp), dispersion self-energy change of halide ion during the binding (nes∆Gdslvself), and cavity energy change of halide ion and protein during the binding (nes∆Gdslvcav):
∆GTOT )
∆Gintcoul +
ele
∆Gdslvplt +
ele
nes
nes
∆Gintdisp +
∆Gdslvself +
∆Gdslvcav (1)
nes
Here, the methods used for calculating these five energy terms are discussed in detail as follows. Coulomb Interaction between Halide Ion and Protein (ele∆Gintcoul). The contribution of Coulomb interaction between the structured halide ion and the charges in protein was calculated by solving the nonlinear Poisson-Boltzmann equation with the charges of protein atoms and halide ion, respectively, set to actual partial charges and 0. The potential (Φ) was then determined at the position of the halide ion, and ele ∆Gintcoul was computed as coul
ele
∆Gint
) Φq
(2)
where q is the formal charge -1 of the halide ion. Electrostatic DesolWation Penalty on Halide Ion and Protein (ele∆GdslWplt). The electrostatic desolvation contribution was separated into two steps for calculation. In one calculation, the radii of all atoms/ion in the protein-halide complex system were set to actual values, but only the halide ion was set with the actual charge -1 and all protein atoms with 0. Then the reaction field energy was computed (eleGrct1hal). Subsequently, the protein was removed from this system (only the halide ion was left) and the reaction field energy was recomputed (eleGrct2hal). Similarly, in a separate calculation the partial charges were placed only on the protein atoms (and the charge of halide ion was set to 0), and the reaction field energies eleGrct1prt and eleGrct2prt of this system, respectively, with and without the presence of halide ion were computed. In this manner, ele∆Gdslvplt can be obtained by
∆Gdslvplt ) (eleGrct1hal -
ele
TABLE 1: Charges and Different Radii Sets for Halide Ions radii (Å) ion
formal charge
crystala
modified crystalb
adjustedc
van der Waalsd
FClBrI-
-1 -1 -1 -1
1.36 1.81 1.95 2.16
1.423 1.937 2.087 2.343
1.51 2.13 2.25 2.51e
1.909 2.252 2.298 2.548
a
From ref 50. b From ref 51. It was obtained by increasing 7% from crystal radii. c From ref 52. It was obtained by optimizing agreement between theoretical and experimental free energies of solvation for nine different salts when both Born and dispersion self-energy are taken into account. d From ref 53. e This value was obtained by extrapolation.
TABLE 2: Static Polarizability Volumes (r) and Ionization Energies (I) for Protein Atoms and Halide Ions atom/ion
R (Å3)
I (kcal/mol)
H C N O S FClBrI-
0.67a 1.76a 1.10a 0.80a 2.90a 1.91b 4.86b 6.49b 9.65b
313.62c 259.69c 335.20c 314.07c 238.93c 78.41d 83.29d 77.55d 70.48d
a From ref 54. b From ref 55. It was obtained on the basis of rigorous ab initio quantum mechanical calculations. c From ref 56. d From ref 57. The anions’ ionization energies are obtained from their electron affinities.
added to the surface of protein-halide complex to approximately account for solvent ion size.46 For each run, potentials were iterated to a convergence of less than 10-4 kT/e. In this study, for protein atoms the sophisticated PARSE parameter set47 was used to define their charges and sizes; for halide anions, the formal charge of -148 was assigned. Because the continuum electrostatic model is susceptible to ionic size,49 four kinds of ionic radii that have been employed in calculating the hydration energy of halide anions, i.e., crystal radii,50 modified crystal radii,51 adjusted radii,52 and van der Waals radii,53 were examined (Table 1). Dispersion Interaction between Halide Ion and Protein (nes∆Gintdisp). We computed the nonretarded dispersion potential between halide ion and protein in a complex system using a pair summation approximation on the basis of the quantum mechanical London formula33
Grct2hal) +
ele
(eleGrct1plt -
Grct2plt)
ele
(3)
∆Gintdisp )
nes
3RXRiIXIi
∑ - 2(I i
All of the above procedures were calculated by finite difference solutions to the nonlinearized Poisson-Boltzmann equation, as implemented in the DELPHI program,45 with ionic strength 0.145 M and dielectric constants of 4 for solute and 80 for solvent. A grid spacing of 1.2 grid per angstrom, in which the longest linear dimension of the solute occupied 60% of the lattice, was used to determine the size of the cubic lattice, and the boundary potentials were set to the sum of the Debye-Hu¨ckel values. The center of the cubic lattice was always placed at the original position of the halide ion in the crystal structure, no matter if this ion was present or absent in the calculation. A probe radii of 1.4 Å was adopted to define the dielectric boundary and a 2.0 Å ion exclusion radius (Stern layer) was
X
+ Ii)rXi6
(4)
where subscripts X and i represent the halide ion and any protein atom, r denotes the distance between them, and R and I are the corresponding polarizability volume54,55 and ionization energy,56,57 respectively (Table 2). The summation is taken over all protein atoms in the studied system. It is worth noting here that this London formula was originally derived for vacuum condition but was thought to be a reasonably good approximation to describe dispersion interactions of structured halide ion with protein atoms in the low dielectric protein context. Dispersion Self-Energy Change of Halide Ion (nes∆GdslWself). The concept of dispersion self-energy in molecular systems was developed by Mahanty and Ninham.34 It is a quantum electro-
15678
J. Phys. Chem. B, Vol. 114, No. 47, 2010
Zhou et al.
dynamic dipolar polarization analogue of the Born electrostatic energy. Previously, Bostro¨m and co-workers demonstrated that dispersion self-energy plays an important role in addition to the sophisticated Born self-energy in contributing to ionic solvation energy.52 They also gave an analytical expression for the effect of dispersion self-energy on the hydrated free energy of isolated ions.52 Unfortunately, structured halide anions concerned here are immersed in the complicated “protein medium” that cannot be treated analytically. Therefore, we herein can only compute the magnitude of the dispersion selfenergy change associated with a halide ion moving from water to protein using an empirical approach. Briefly, in ideal conditions the change in dispersion self-energy accompanied with transferring a halide ion from bulk water to an infinitely extended protein medium can be accurately quantified as58
self ∆Gideal(waterfprotein) )
4kBT
∞
∑′ √πr3 n)0
[
Rprotein*(iωn) εprotein(iωn) Rwater*(iωn) εwater(iωn)
]
(5)
where kB is Boltzmann’s constant, r is ionic radius, T is the temperature (298 K is considered here), and R*(iωn) denotes the excess ionic polarizability in a specific medium with dielectric function ε(iωn) and can be obtained using the inverted Clausius-Mosotti/Lorenz-Lorentz equation.59 The summation is taken over discrete frequencies (ωn ) 2πkBTn/p), and the prime on the summation indicates that the n ) 0 term carries a weight of 1/2. Using this strategy, we calculated the ideal dispersion self-energies of transfer for F-, Cl-, Br-, and I- as -3.07, -3.83, -3.59, and -3.44 kcal/mol, respectively. A more detailed description of this calculation procedure is found in the Appendix. In the actual cases, however, structured halide ions are partially or completely packed in a finite protein matrix. This means that both protein and water are responsible for the dispersion self-energy of structured halide ions. Considering that dispersion force is considerably short-range relative to, for example, long-range electrostatic and dipole-dipole interactions, the effect of medium (protein/water) on a structured halide ion principally arises from a limited region close to the surface of the halide ion. Therefore, nes∆Gdslvself can be estimated according to the ratio of halide ion’s solvent-accessible surface area (ASA) in structured state (S2) to that in free state (S1):
(
∆Gdslvself ) 1 -
nes
)
S2 self ∆Gideal(waterfprotein) S1
(6)
where S2/S1 is also called solvent-accessible rate (ASA%), which characterizes the degree of a structured halide ion exposing to solvent. CaWity Energy Change of Halide Ion and Protein (nes∆GdslWcaW). The cavity energy accounts for the free energy demand for creating a cavity in solvent to accommodate a solute. Both calorimetric experiments and theoretical analyses revealed that cavity energy basically linearly relates to the ASA of solute molecule, and based upon this finding, the empirical surface area (SA) model has been proposed to analyze the cavity energy contrition to biomolecular folding and binding.35 In this study, the SA model was applied to estimate the change in cavity energy upon the formation of a halide motif in protein:
∆Gdslvcav ) γ(∆Sprotein + ∆Shalide)
nes
(7)
where γ ) 0.00542 kcal/mol Å2, as recommended by Kollman et al.,35 and ∆Sprotein and ∆Shalide are the changes in ASAs of, respectively, protein and halide ion due to the protein-halide binding. We computed the solute’s surface area using Sanner’s algorithm implemented in the MSMS program,60 with a solvent probe radius of 1.4 Å and the PARSE radii parameters for protein atoms.47 For the halide ions, analysis of electron density distributions in ionic crystals revealed that the electron density due to positive ions begins to become significant at a distance of about the ionic radius from the center of the anion, indicating that ionic radius could provide a reasonable measure of the cavity radius formed by an anion.51 In this respect, we here adopted ionic crystal radii (see Table 1) to perform this computation. 2.4. ONIOM-Based QM/MM Scheme. As a comparison with the empirical EDA approach, we also implemented the hybrid quantum mechanical/molecular mechanical (QM/MM) calculations in the Gaussian 03 suite of programs61 to examine the structural and energetic properties for several well-characterized halide motifs formed in haloalkane dehalogenase within the whole framework of protein-halide complexes. QM/MM was carried out using the ONIOM method.36 It enables different levels of theory to be applied to different parts of a molecular system and combined to produce a consistent energy expression. The objective is to perform a high-level calculation on just a small part of the system and to include the effects of the remainder at lower level of theory, with the end result being of similar accuracy to a high-level calculation on the full system.62 Previously, we have successfully applied the ONIOM-based QM/MM scheme to investigate protein-ligand interactions including halogen atoms;63,64 hence, this hybrid technique is expected to study the noncovalent interactions involving halide ions in proteins with reasonable results. Briefly, the investigated systems were partitioned into two layers: the structured halide ion and the protein residues that are directly bound to the halide ion were included in the QM layer and treated with a high level of density functional theory (HCTH407/Lanl2DZ for I- or HCTH407/6-31+G* for other atoms/ions), while the rest of the protein atoms in the MM layer were modeled by a low level of molecular force field (AMBER parm96). Water molecules were described by the TIP3P model,65 and the force field parameters developed by Joung et al. were used for halide ions.66 It is known that, owing to the localcorrected functional (LCF) used, density functional theory (DFT) normally underappreciates the dispersion force that has been recognized as a non-negligible factor affecting the physicochemical behavior of the diffuse, polarizable anions,13-15 but this theory is much “cheaper” than other electron-correlation methods such as perturbation theory and coupled cluster. By a comprehensive comparison of 25 DFT methods, Johnson et al. recently concluded that HCTH407 functional can reproduce the intermolecular dispersion potentials best for a series of dimer systems.67 Therefore, we herein adopted the HCTH407 to perform exhaustive calculations in the QM layer. The structures of protein-halide complex systems were fully optimized on the basis of the two-layer ONIOM-based QM/ MM methodology described above without any constraints. The electrostatic interactions between the two layers were treated in terms of a mechanical embedding scheme to save computational cost. After the optimization procedure, the QM subsystem was stripped from the optimized model structure to perform further single-point energy analysis by using rigorous Møller-
Do Halide Motifs Stabilize Protein Architecture?
J. Phys. Chem. B, Vol. 114, No. 47, 2010 15679
Figure 5. (a) EDA-calculated versus experimentally determined free energies for the binding of halide ions to DhlA based on different sets of ionic radii. The experimentally determined free energies for binding halide ions to DhlA were derived from corresponding dissociation constants, which were reported in ref 74. (b) Halide motif in DhlA (PDB ID: 2eda). The halide ion X- is hydrogen-bound to Trp125 and Trp175, while it is electrostatically repulsed by Asp124.
TABLE 3: Thermodynamic Data for Binding Halide Ions to DhlA calculated by EDA (kcal/mol)a PDB ID
ion
ASA%
1edb 1cij 2eda
ClBrI-
0.00 0.00 0.00
ele
coul
∆Gint 5.27 4.87 4.38
ele
∆Gdslv 13.11 13.68 13.38
plt
nes
disp
∆Gint
-13.00 -14.93 -15.91
nes
∆Gdslv
self
-3.83 -3.59 -3.44
experimentb nes
∆Gdslvcav
∆GTOT
Kd (mM)
∆G (kcal/mol)c
-0.78 -0.83 -0.95
0.77 -0.80 -2.54
75 10 5
-1.56 -2.77 -3.19
a Calculated on the basis of the modified crystal radii of Rashin and Honig.51 b From ref 74. Determined by stopped-flow fluorescence experiments. c Free energies, ∆G, were derived from corresponding dissociation constants, Kd, at the experimental conditions.
Plesset second-order perturbation theory in conjunction with the modified effective core potential (ECP) basis set [MP2/ Lanl2DZ+(df)68] for I- or Dunning’s augmented correlation consistent basis set (MP2/aug-cc-pVDZ) for other atoms/ions; the interaction potential ∆EintMP2 of halide ion with protein moieties in the stripped portion was assessed under the supermolecular approach,69 with the associated basis set superposition error (BSSE) being eliminated by means of the standard counterpoise method of Boys and Bernardi.70 Moreover, dispersion potential was estimated approximately as the difference between the ∆EintMP2 obtained at the MP2 level and the ∆EintHF derived from the Hartree-Fock (HF) level with the same basis sets. 3. Results and Discussion 3.1. EDA and QM/MM Analyses of Halide Ions Binding to Haloalkane Dehalogenase. Haloalkane dehalogenase (DhlA) is one of best-studied members in the dehalogenase family, which is capable of hydrolyzing short-chain haloalkanes to produce the corresponding alcohols and halide ions.71 The conversion of reactant to products in the active site of DhlA requires a series of complicated reactions, including an SN2 reaction yielding an alkyl enzyme intermediate and a hydrolysis reaction hydrolyzing the intermediate.72 This process appears to follow an ordered step by releasing the producing halide ion, which is noncovalently bound nearby the catalytic pocket of DhlA.73 The kinetic mechanism and dissociation constants for the release of different halide ions from DhlA has been studied by Schanstra et al. using the stopped-flow fluorescence experiments on the basis of its reverse process, i.e. binding of halide ions to DhlA.74 Here, with the experimental data we performed a computational study of the halide motifs in DhlA to shed some
light on the structural and energetic basis of these motifs and, more importantly, to test our method with this well-characterized paradigm. Energy Decomposition Analysis (EDA). The crystal structures of the bound complexes of DhlA with Cl-, Br-, and Iwere extracted from the PDB entries 1edb, 1cij, and 2eda, respectively. These structures were treated with the procedure described in section 2.2 and were analyzed using the EDA approach. Because the continuum electrostatic model used in EDA is susceptible to the ionic radii adopted,49 we herein gave a comparison of the performance of the four ionic radii sets (as listed in Table 1) in reproducing experimentally measured affinities74 of halide ions binding to DhlA. As can be seen from Figure 5a, all radii sets considered here underestimated, but not by much, the stability of halide motifs in DhlA; this is not unanticipated, since some factors such as allosteric effect and entropic cost were overlooked by the EDA method. Both of the root-mean-square (rms) errors and the correlations (r) between EDA-calculated and experimentally determined free energies for the binding of different halide ions to DhlA increase in the order crystal radii < modified crystal radii < adjusted radii < van der Waals radii, in which the crystal radii and modified crystal radii seem to yield a better result, with rms
