A fast and simple method to calculate protonation states in proteins

Prediction of protein pK a with representation learning

The behavior of proteins is closely related to the protonation states of the residues. Therefore, prediction and measurement of pK a are essential to understand the basic functions of proteins. In this work, we develop a new empirical scheme for protein pK a prediction that is based on deep representation learning. It combines machine learning with atomic environment vector (AEV) and learned quantum mechanical representation from ANI-2x neural network potential (J. Chem. Theory Comput. 2020, 16, 4192). The scheme requires only the coordinate information of a protein as the input and separately estimates the pK a for all five titratable amino acid types. The accuracy of the approach was analyzed with both cross-validation and an external test set of proteins. Obtained results were compared with the widely used empirical approach PROPKA. The new empirical model provides accuracy with MAEs below 0.5 for all amino acid types. It surpasses the accuracy of PROPKA and performs significantly better than the null model. Our model is also sensitive to the local conformational changes and molecular interactions.

Using polarizable POSSIM force field and fuzzy-border continuum solvent model to calculate pK(a) shifts of protein residues

Our Fuzzy-Border (FB) continuum solvent model has been extended and modified to produce hydration parameters for small molecules using POlarizable Simulations Second-order Interaction Model (POSSIM) framework with an average error of 0.136 kcal/mol. It was then used to compute pKa shifts for carboxylic and basic residues of the turkey ovomucoid third domain (OMTKY3) protein. The average unsigned errors in the acid and base pKa values were 0.37 and 0.4 pH units, respectively, versus 0.58 and 0.7 pH units as calculated with a previous version of polarizable protein force field and Poisson Boltzmann continuum solvent. This POSSIM/FB result is produced with explicit refitting of the hydration parameters to the pKa values of the carboxylic and basic residues of the OMTKY3 protein; thus, the values of the acidity constants can be viewed as additional fitting target data. In addition to calculating pKa shifts for the OMTKY3 residues, we have studied aspartic acid residues of Rnase Sa. This was done without any further refitting of the parameters and agreement with the experimental pKa values is within an average unsigned error of 0.65 pH units. This result included the Asp79 residue that is buried and thus has a high experimental pKa value of 7.37 units. Thus, the presented model is capable or reproducing pKa results for residues in an environment that is significantly different from the solvated protein surface used in the fitting. Therefore, the POSSIM force field and the FB continuum solvent parameters have been demonstrated to be sufficiently robust and transferable. © 2016 Wiley Periodicals, Inc.

pK(A) in proteins solving the Poisson-Boltzmann equation with finite elements

Knowledge on pK(A) values is an eminent factor to understand the function of proteins in living systems. We present a novel approach demonstrating that the finite element (FE) method of solving the linearized Poisson-Boltzmann equation (lPBE) can successfully be used to compute pK(A) values in proteins with high accuracy as a possible replacement to finite difference (FD) method. For this purpose, we implemented the software molecular Finite Element Solver (mFES) in the framework of the Karlsberg+ program to compute pK(A) values. This work focuses on a comparison between pK(A) computations obtained with the well-established FD method and with the new developed FE method mFES, solving the lPBE using protein crystal structures without conformational changes. Accurate and coarse model systems are set up with mFES using a similar number of unknowns compared with the FD method. Our FE method delivers results for computations of pK(A) values and interaction energies of titratable groups, which are comparable in accuracy. We introduce different thermodynamic cycles to evaluate pK(A) values and we show for the FE method how different parameters influence the accuracy of computed pK(A) values.

Uncovering the determinants of a highly perturbed tyrosine pKa in the active site of ketosteroid isomerase

Within the idiosyncratic enzyme active-site environment, side chain and ligand pKa values can be profoundly perturbed relative to their values in aqueous solution. Whereas structural inspection of systems has often attributed perturbed pKa values to dominant contributions from placement near charged groups or within hydrophobic pockets, Tyr57 of a Pseudomonas putida ketosteroid isomerase (KSI) mutant, suggested to have a pKa perturbed by nearly 4 units to 6.3, is situated within a solvent-exposed active site devoid of cationic side chains, metal ions, or cofactors. Extensive comparisons among 45 variants with mutations in and around the KSI active site, along with protein semisynthesis, (13)C NMR spectroscopy, absorbance spectroscopy, and X-ray crystallography, was used to unravel the basis for this perturbed Tyr pKa. The results suggest that the origin of large energetic perturbations are more complex than suggested by visual inspection. For example, the introduction of positively charged residues near Tyr57 raises its pKa rather than lowers it; this effect, and part of the increase in the Tyr pKa from the introduction of nearby anionic groups, arises from accompanying active-site structural rearrangements. Other mutations with large effects also cause structural perturbations or appear to displace a structured water molecule that is part of a stabilizing hydrogen-bond network. Our results lead to a model in which three hydrogen bonds are donated to the stabilized ionized Tyr, with these hydrogen-bond donors, two Tyr side chains, and a water molecule positioned by other side chains and by a water-mediated hydrogen-bond network. These results support the notion that large energetic effects are often the consequence of multiple stabilizing interactions rather than a single dominant interaction. Most generally, this work provides a case study for how extensive and comprehensive comparisons via site-directed mutagenesis in a tight feedback loop with structural analysis can greatly facilitate our understanding of enzyme active-site energetics. The extensive data set provided may also be a valuable resource for those wishing to extensively test computational approaches for determining enzymatic pKa values and energetic effects.

Assignment of protonation states in proteins and ligands: combining pKa prediction with hydrogen bonding network optimization

Among the many applications of molecular modeling, drug design is probably the one with the highest demands on the accuracy of the underlying structures. During lead optimization, the position of every atom in the binding site should ideally be known with high precision to identify those chemical modifications that are most likely to increase drug affinity. Unfortunately, X-ray crystallography at common resolution yields an electron density map that is too coarse, since the chemical elements and their protonation states cannot be fully resolved.This chapter describes the steps required to fill in the missing knowledge, by devising an algorithm that can detect and resolve the ambiguities. First, the pK (a) values of acidic and basic groups are predicted. Second, their potential protonation states are determined, including all permutations (considering for example protons that can jump between the oxygens of a phosphate group). Third, those groups of atoms are identified that can adopt alternative but indistinguishable conformations with essentially the same electron density. Fourth, potential hydrogen bond donors and acceptors are located. Finally, all these data are combined in a single "configuration energy function," whose global minimum is found with the SCWRL algorithm, which employs dead-end elimination and graph theory. As a result, one obtains a complete model of the protein and its bound ligand, with ambiguous groups rotated to the best orientation and with protonation states assigned considering the current pH and the H-bonding network. An implementation of the algorithm has been available since 2008 as part of the YASARA modeling & simulation program.

Electrostatic solvation energy for two oppositely charged ions in a solvated protein system: salt bridges can stabilize proteins

Born-type electrostatic continuum methods have been an indispensable ingredient in a variety of implicit-solvent methods that reduce computational effort by orders of magnitude compared to explicit-solvent MD simulations and thus enable treatment using larger systems and/or longer times. An analysis of the limitations and failures of the Born approaches serves as a guide for fundamental improvements without diminishing the importance of prior works. One of the major limitations of the Born theory is the lack of a liquidlike description of the response of solvent dipoles to the electrostatic field of the solute and the changes therein, a feature contained in the continuum Langevin-Debye (LD) model applied here to investigate how Coulombic interactions depend on the location of charges relative to the protein/water boundary. This physically more realistic LD model is applied to study the stability of salt bridges. When compared head to head using the same (independently measurable) physical parameters (radii, dielectric constants, etc.), the LD model is in good agreement with observations, whereas the Born model is grossly in error. Our calculations also suggest that a salt bridge on the protein's surface can be stabilizing when the charge separation is < or =4 A.

Continuum electrostatic calculations of the pKa of ionizable residues in an ion channel: dynamic vs. static input structure

We have computed the pK(a)'s of the ionizable residues of a protein ion channel, the Staphylococcus aureus toxin alpha-hemolysin, by using two types of input structures, namely the crystal structure of the heptameric alpha-hemolysin and a set of over four hundred snapshots from a 4.38 ns Molecular Dynamics simulation of the protein inserted in a phospholipid planar bilayer. The comparison of the dynamic picture provided by the Molecular Simulation with the static one based on the X-ray crystal structure of the protein embedded in a lipid membrane allows analyzing the influence of the fluctuations in the protein structure on its ionization properties. We find that the use of the dynamic structure provides interesting information about the sensitivity of the computed pK(a) of a given residue to small changes in the local structure. The calculated pK(a) are consistent with previous indirect estimations obtained from single-channel conductance and selectivity measurements.

Predicting pKa

One of the most important physicochemical properties of small molecules and macromolecules are the dissociation constants for any weakly acidic or basic groups, generally expressed as the pK(a) of each group. This is a major factor in the pharmacokinetics of drugs and in the interactions of proteins with other molecules. For both the protein and small molecule cases, we survey the sources of experimental pK(a) values and then focus on current methods for predicting them. Of particular concern is an analysis of the scope, statistical validity, and predictive power of methods as well as their accuracy.

MCCE2: improving protein pKa calculations with extensive side chain rotamer sampling

Multiconformation continuum electrostatics (MCCE) explores different conformational degrees of freedom in Monte Carlo calculations of protein residue and ligand pK(a)s. Explicit changes in side chain conformations throughout a titration create a position dependent, heterogeneous dielectric response giving a more accurate picture of coupled ionization and position changes. The MCCE2 methods for choosing a group of input heavy atom and proton positions are described. The pK(a)s calculated with different isosteric conformers, heavy atom rotamers and proton positions, with different degrees of optimization are tested against a curated group of 305 experimental pK(a)s in 33 proteins. QUICK calculations, with rotation around Asn and Gln termini, sampling His tautomers and torsion minimum hydroxyls yield an RMSD of 1.34 with 84% of the errors being <1.5 pH units. FULL calculations adding heavy atom rotamers and side chain optimization yield an RMSD of 0.90 with 90% of the errors <1.5 pH unit. Good results are also found for pK(a)s in the membrane protein bacteriorhodopsin. The inclusion of extra side chain positions distorts the dielectric boundary and also biases the calculated pK(a)s by creating more neutral than ionized conformers. Methods for correcting these errors are introduced. Calculations are compared with multiple X-ray and NMR derived structures in 36 soluble proteins. Calculations with X-ray structures give significantly better pK(a)s. Results with the default protein dielectric constant of 4 are as good as those using a value of 8. The MCCE2 program can be downloaded from http://www.sci.ccny.cuny.edu/~mcce.

Analysis of the electrochemistry of hemes with E(m)s spanning 800 mV

The free energy of heme reduction in different proteins is found to vary over more than 18 kcal/mol. It is a challenge to determine how proteins manage to achieve this enormous range of E(m)s with a single type of redox cofactor. Proteins containing 141 unique hemes of a-, b-, and c-type, with bis-His, His-Met, and aquo-His ligation were calculated using Multi-Conformation Continuum Electrostatics (MCCE). The experimental E(m)s range over 800 mV from -350 mV in cytochrome c(3) to 450 mV in cytochrome c peroxidase (vs. SHE). The quantitative analysis of the factors that modulate heme electrochemistry includes the interactions of the heme with its ligands, the solvent, the protein backbone, and sidechains. MCCE calculated E(m)s are in good agreement with measured values. Using no free parameters the slope of the line comparing calculated and experimental E(m)s is 0.73 (R(2) = 0.90), showing the method accounts for 73% of the observed E(m) range. Adding a +160 mV correction to the His-Met c-type hemes yields a slope of 0.97 (R(2) = 0.93). With the correction 65% of the hemes have an absolute error smaller than 60 mV and 92% are within 120 mV. The overview of heme proteins with known structures and E(m)s shows both the lowest and highest potential hemes are c-type, whereas the b-type hemes are found in the middle E(m) range. In solution, bis-His ligation lowers the E(m) by approximately 205 mV relative to hemes with His-Met ligands. The bis-His, aquo-His, and His-Met ligated b-type hemes all cluster about E(m)s which are approximately 200 mV more positive in protein than in water. In contrast, the low potential bis-His c-type hemes are shifted little from in solution, whereas the high potential His-Met c-type hemes are raised by approximately 300 mV from solution. The analysis shows that no single type of interaction can be identified as the most important in setting heme electrochemistry in proteins. For example, the loss of solvation (reaction field) energy, which raises the E(m), has been suggested to be a major factor in tuning in situ E(m)s. However, the calculated solvation energy vs. experimental E(m) shows a slope of 0.2 and R(2) of 0.5 thus correlates weakly with E(m)s. All other individual interactions show even less correlation with E(m). However the sum of these terms does reproduce the range of observed E(m)s. Therefore, different proteins use different aspects of their structures to modulate the in situ heme electrochemistry. This study also shows that the calculated E(m)s are relatively insensitive to different heme partial charges and to the protein dielectric constant used in the simulation.

High-throughput pKa screening and prediction amenable for ADME profiling

Recent technological advances have made it possible for several new pK(a) assays to be used in drug screening. In this review, a critical overview is provided of current new methodologies for high-throughput screening and prediction of pK(a). Typical applications of using pK(a )constants and charge state for absorption, distribution, metabolism and excretion (ADME) profiling and quantitative structure-activity relationship modelling complements the methodological comparisons and discussions. The experimental methods discussed include high-throughput screening of pK(a) by multiplexed capillary with ultraviolet absorbance detection on a 96-capillary format instrument, capillary electrophoresis and mass spectrometry (CEMS) based on sample pooling, determination of pK(a) by pH gradient high-performance liquid chromatography, and measurement of pK(a) by a mixed-buffer liner pH gradient system. Comparisons of the different experimental assays are made with emphasis on the newly developed CEMS method. The current status and recent progress in computational approaches to pK(a) prediction are also discussed. In particular, the accuracy limits of simple fragment-based approaches as well as quantum mechanical methods are addressed. Examples of pK(a) prediction from in-house drug candidates as well as commercially available drug molecules are shown and an outline is provided for how drug discovery companies can integrate experiments with computational approaches for increased applications for ADME profiling.

Reproducing basic pKa values for turkey ovomucoid third domain using a polarizable force field

We have extended our previous studies of calculating acidity constants for the acidic residues found in the turkey ovomucoid third domain protein (OMTKY3) by determining the relative pKa values for the basic residues (Lys13, Arg21, Lys29, Lys34, His52, and Lys55). A polarizable force field (PFF) was employed. The values of the pKa were found by direct comparison of energies of solvated protonated and deprotonated forms of the protein. Poisson-Boltzmann (PBF) and surface generalized Born (SGB) continuum solvation models represent the hydration, and a nonpolarizable fixed-charge OPLS-AA force field was used for comparison. Our results indicate that (i) the pKa values of the basic residues can be found in close agreement with the experimental values when a PFF is used in conjunction with the PBF solvation model, (ii) it is sufficient to take into the account only the residues which are in close proximity (hydrogen bonded) to the residue in question, and (iii) the PBF solvation model is superior to the SGB solvation model for these pKa calculations. The average error with the PBF/PFF model is only 0.7 pH unit, compared with 2.2 and 6.1 units for the PBF/OPLS and SGB/OPLS, respectively. The maximum deviation of the PBF/PFF results from the experimental values is 1.7 pH units compared with 6.0 pH units for the PBF/OPLS. Moreover, the best results were obtained while using an advanced nonpolar energy calculation scheme. The overall conclusion is that this methodology and force field are suitable for the accurate assessment of pKa shifts for both acidic and basic protein residues.

Influence of nonlinear electrostatics on transfer energies between liquid phases: charge burial is far less expensive than Born model

The widely used Born model describes the electrostatic response of continuous media using static dielectric constants. However, when applied to a liquid environment, a comparison of Born model predictions with experimental values (e.g., transfer free energies and pK(a) shifts) found that agreement is only achieved by using physically unrealistic dielectric constants for proteins, lipids, etc., and/or equally unrealistic atomic radii. This leads to questions concerning the physical origins for this failure of the Born model. We partially resolve this question by applying the Langevin-Debye (LD) model of a continuous distribution of point, polarizable dipoles, a model that contains an added dependence of the electrostatic response on the solvent's optical dielectric constant and both gas- and liquid-phase dipole moments, features absent in the Born model to which the LD model reduces for weak fields. The LD model is applied to simple representations of three biologically relevant systems: (i) globular proteins, (ii) lipid bilayers, and (iii) membrane proteins. The linear Born treatment greatly overestimates both the self-energy and the transfer free energy from water to hydrophobic environments (e.g., a protein interior). By using the experimental dielectric constant, the energy cost of charge burial in either globular or membrane proteins of the Born model is reduced by almost 50% with the nonlinear theory as is the pK(a) shift, and the shifts agree well with experimental trends.

Prediction and rationalization of protein pKa values using QM and QM/MM methods

We describe the development and application of a computational method for the prediction and rationalization of pKa values of ionizable residues in proteins, based on ab initio quantum mechanics (QM) and the effective fragment potential (EFPs) method (a hybrid QM/MM method). The theoretical developments include (1) a covalent boundary method based on frozen localized orbitals, (2) divide-and-conquer methods for the ab initio computation of protein EFPs consisting of multipoles up to octupoles and dipole polarizability tensors, (3) a method for computing vibrational free energies for a localized molecular region, and (4) solutions of the polarized continuum model of bulk solvation equations for protein-sized systems. The QM-based pKa prediction method is one of the most accurate methods currently available and can be used in cases where other pKa prediction methods fail. Preliminary analysis of the computed results indicate that many pKa values (1) are primarily determined by hydrogen bonds rather than long-range charge-charge interactions and (2) are relatively insensitive to large-scale dynamical fluctuations of the protein structure.

Electrostatic polarization is crucial for reproducing pKa shifts of carboxylic residues in Turkey ovomucoid third domain

We have computed pKa shifts for carboxylic residues of the serine protease inhibitor turkey ovomucoid third domain (residues Asp7, Glu10, Glu19, Asp27, and Glu43). Both polarizable and nonpolarizable empirical force fields were employed. Hydration was represented by the surface generalized Born and Poisson-Boltzmann continuum model. The calculations were carried out in the most physically straightforward fashion, by directly comparing energies of the protonated and deprotonated protein forms, without any additional parameter fitting or adjustment. Our studies have demonstrated that (i) the Poisson-Boltzmann solvation model is more than adequate in reproducing pKa shifts, most likely due to its intrinsically many-body formalism; (ii) explicit treatment of electrostatic polarization included in our polarizable force field (PFF) calculations appears to be crucial in reproducing the acidity constant shifts. The average error of the PFF results was found to be as low as 0.58 pKa units, with the best fixed-charges average deviation being 3.28 units. Therefore, the pKa shifts phenomena and the governing electrostatics are clearly many-body controlled in their intrinsic nature; (iii) our results confirm previously reported conclusions that pKa shifts for protein residues are controlled by the immediate environment of the residues in question, as opposed to long-range interactions in proteins. We are confident that our confirmation of the importance of explicit inclusion of polarization in empirical force fields for protein studies will be useful far beyond the immediate goal of accurate calculation of acidity constants.

Calculations of pH-dependent binding of proteins to biological membranes

Binding of proteins to membranes is often accompanied by titration of ionizable residues and is, therefore, dependent on pH. We present a theoretical treatment and computational approach for predicting absolute, pH-dependent membrane binding free energies. The standard free energy of binding, DeltaG, is defined as -RTln(P(b)/P(f)), where P(b) and P(f) are the amounts of bound and free protein. The apparent pK(a) of binding is the pH value at which P(b) and P(f) are equal. Proteins bind to the membrane in the pH range where DeltaG is negative. The components of the binding free energy are (a) the free energy cost of ionization state changes (DeltaG(ion)), (b) the effective energy of transfer from solvent to the membrane surface, (c) the translational/rotational entropy cost of binding, and (d) an ideal entropy term that depends on the relative volume of the bound and free state and therefore depends on lipid concentration. Calculation of the first term requires determination of pK(a) values in solvent and on the membrane surface. All energies required by the method are obtained from molecular dynamics trajectories on an implicit membrane (IMM1-GC). The method is tested on pentalysine and the helical peptide VEEKS, derived from the membrane-binding domain of phosphocholine cytidylyltransferase. The agreement between the measured and the calculated free energies of binding of pentalysine is good. The extent of membrane binding of VEEKS is, however, underestimated compared to experiment. Calculations of the interaction energy between two VEEKS helices on the membrane suggest that the discrepancy is mainly due to the neglect of protein-protein interactions on the membrane surface.

Molecular Dynamics Simulations of Rhodopsin in Different One-Component Lipid Bilayers

Four 20 ns molecular dynamic simulations of rhodopsin embedded in different one-component lipid bilayers have been carried out to ascertain the importance of membrane lipids on the protein structure. Specifically, dimyristoyl phosphatidylcholine (DMPC), dipalmitoyl phosphatidylcholine (DPPC), palmitoyl oleoyl phosphatidylcholine (POPC), and palmitoyl linoleyl phosphatidylcholine (PLPC) lipid bilayers have been considered for the present work. The results reported here provide information on the hydrophobic matching between the protein and the bilayer and about the differential effects of the protein on the thickness of the different membranes. Furthermore, a careful analysis of the individual protein-lipid interactions permits the identification of residues that exhibit permanent interactions with atoms of the lipid environment that may putatively act as hooks of the protein to the membrane. The analysis of the trajectories also provides information about the effect of the bilayer on the protein structure, including secondary structural elements, salt bridges, and rigid-body motions.

Assignment of polar states for protein amino acid residues using an interaction cluster decomposition algorithm and its application to high resolution protein structure modeling

We have developed a new method (Independent Cluster Decomposition Algorithm, ICDA) for creating all-atom models of proteins given the heavy-atom coordinates, provided by X-ray crystallography, and the pH. In our method the ionization states of titratable residues, the crystallographic mis-assignment of amide orientations in Asn/Gln, and the orientations of OH/SH groups are addressed under the unified framework of polar states assignment. To address the large number of combinatorial possibilities for the polar hydrogen states of the protein, we have devised a novel algorithm to decompose the system into independent interacting clusters, based on the observation of the crucial interdependence between the short range hydrogen bonding network and polar residue states, thus significantly reducing the computational complexity of the problem and making our algorithm tractable using relatively modest computational resources. We utilize an all atom protein force field (OPLS) and a Generalized Born continuum solvation model, in contrast to the various empirical force fields adopted in most previous studies. We have compared our prediction results with a few well-documented methods in the literature (WHATIF, REDUCE). In addition, as a preliminary attempt to couple our polar state assignment method with real structure predictions, we further validate our method using single side chain prediction, which has been demonstrated to be an effective way of validating structure prediction methods without incurring sampling problems. Comparisons of single side chain prediction results after the application of our polar state prediction method with previous results with default polar state assignments indicate a significant improvement in the single side chain predictions for polar residues.

Effect of different treatments of long-range interactions and sampling conditions in molecular dynamic simulations of rhodopsin embedded in a dipalmitoyl phosphatidylcholine bilayer

The present study analyzes the effect of the simulation conditions on the results of molecular dynamics simulations of G-protein coupled receptors (GPCRs) performed with an explicit lipid bilayer. Accordingly, the present work reports the analysis of different simulations of bovine rhodopsin embedded in a dipalmitoyl phosphatidylcholine (DPPC) lipid bilayer using two different sampling conditions and two different approaches for the treatment of long-range electrostatic interactions. Specifically, sampling was carried out either by using the statistical ensembles NVT or NPT (constant number of atoms, a pressure of 1 atm in all directions and fixed temperature), and the electrostatic interactions were treated either by using a twin-cutoff, or the particle mesh Ewald summation method (PME). The results of the present study suggest that the use of the NPT ensemble in combination with the PME method provide more realistic simulations. The use of NPT during the equilibration avoids the need of an a priori estimation of the box dimensions, giving the correct area per lipid. However, once the system is equilibrated, the simulations are irrespective of the sampling conditions used. The use of an electrostatic cutoff induces artifacts on both lipid thickness and the ion distribution, but has no direct effect on the protein and water molecules.

Fast empirical pKa prediction by Ewald summation

pK(a) calculations for macromolecules are normally performed by solving the Poisson-Boltzmann equation, accounting for the different dielectric constants of solvent and solute, as well as the ionic strength. Despite the large number of successful applications, there are some situations where the current algorithms are not suitable: (1) large scale, high-throughput analysis which requires calculations to be completed within a fraction of a second, e.g. when permanently monitoring pK(a) shifts during a molecular dynamics simulation; (2) prediction of pK(a)s in periodic boundaries, e.g. when reconstructing entire protein crystal unit cells from PDB files, including the correct protonation patterns at experimental pH. Such in silico crystals are needed by 'self-parameterizing' molecular dynamics force fields like YASARA YAMBER, that optimize their parameters while energy-minimizing high-resolution protein crystals. To address both problems, we define an empirical equation that expresses the pK(a) as a function of electrostatic potential, hydrogen bonds and accessible surface area. The electrostatic potential is evaluated by Ewald summation, which captures periodic crystal environments and the uncertainty in atom positions using Gaussian charge densities. The empirical proportionality constants are derived from 217 experimentally determined pK(a)s, and despite its simplicity, this pK(a) calculation method reaches a high overall jack-knifed accuracy, and is fast enough to be used during a molecular dynamics simulation. A reliable null-model to judge pK(a) prediction accuracies is also presented.

Factors influencing the energetics of electron and proton transfers in proteins. What can be learned from calculations

A protein structure should provide the information needed to understand its observed properties. Significant progress has been made in developing accurate calculations of acid/base and oxidation/reduction reactions in proteins. Current methods and their strengths and weaknesses are discussed. The distribution and calculated ionization states in a survey of proteins is described, showing that a significant minority of acidic and basic residues are buried in the protein and that most of these remain ionized. The electrochemistry of heme and quinones are considered. Proton transfers in bacteriorhodopsin and coupled electron and proton transfers in photosynthetic reaction centers, 5-coordinate heme binding proteins and cytochrome c oxidase are highlighted as systems where calculations have provided insight into the reaction mechanism.

Inferring ideal amino acid interaction forms from statistical protein contact potentials

We have analyzed 29 different published matrices of protein pairwise contact potentials (CPs) between amino acids derived from different sets of proteins, either crystallographic structures taken from the Protein Data Bank (PDB) or computer-generated decoys. Each of the CPs is similar to 1 of the 2 matrices derived in the work of Miyazawa and Jernigan (Proteins 1999;34:49-68). The CP matrices of the first class can be approximated with a correlation of order 0.9 by the formula e(ij) = h(i) + h(j), 1 <or= i, j <or= 20, where the residue-type dependent factor h is highly correlated with the frequency of occurrence of a given amino acid type inside proteins. Electrostatic interactions for the potentials of this class are almost negligible. In the potentials belonging to this class, the major contribution to the potentials is the one-body transfer energy of the amino acid from water to the protein environment. Potentials belonging to the second class can be approximated with a correlation of 0.9 by the formula e(ij) = c(0) - h(i)h(j) + q(i)q(j), where c(0) is a constant, h is highly correlated with the Kyte-Doolittle hydrophobicity scale, and a new, less dominant, residue-type dependent factor q is correlated ( approximately 0.9) with amino acid isoelectric points pI. Including electrostatic interactions significantly improves the approximation for this class of potentials. While, the high correlation between potentials of the first class and the hydrophobic transfer energies is well known, the fact that this approximation can work well also for the second class of potentials is a new finding. We interpret potentials of this class as representing energies of contact of amino acid pairs within an average protein environment.

The Influence of a Transmembrane pH Gradient on Protonation Probabilities of Bacteriorhodopsin: The Structural Basis of the Back-Pressure Effect

Bacteriorhodopsin pumps protons across a membrane using the energy of light. The proton pumping is inhibited when the transmembrane proton gradient that the protein generates becomes larger than four pH units. This phenomenon is known as the back-pressure effect. Here, we investigate the structural basis of this effect by predicting the influence of a transmembrane pH gradient on the titration behavior of bacteriorhodopsin. For this purpose we introduce a method that accounts for a pH gradient in protonation probability calculations. The method considers that in a transmembrane protein, which is exposed to two different aqueous phases, each titratable residue is accessible for protons from one side of the membrane depending on its hydrogen-bond pattern. This method is applied to several ground-state structures of bacteriorhodopsin, which residues already present complicated titration behaviors in the absence of a proton gradient. Our calculations show that a pH gradient across the membrane influences in a non-trivial manner the protonation probabilities of six titratable residues which are known to participate in the proton transfer: D85, D96, D115, E194, E204, and the Schiff base. The residues connected to one side of the membrane are influenced by the pH on the other side because of their long-range electrostatic interactions within the protein. In particular, D115 senses the pH at the cytoplasmic side of the membrane and transmits this information to D85 and the Schiff base. We propose that the strong electrostatic interactions found between D85, D115, and the Schiff base as well as the interplay of their respective protonation states under the influence of a transmembrane pH gradient are responsible for the back-pressure effect on bacteriorhodopsin.

The determinants of carboxyl pKa values in turkey ovomucoid third domain

A computational methodology for protein pK(a) predictions, based on ab initio quantum mechanical treatment of part of the protein and linear Poisson-Boltzmann equation treatment of the bulk solvent, is presented. The method is used to predict and interpret the pK(a) values of the five carboxyl residues (Asp7, Glu10, Glu19, Asp27, and Glu43) in the serine protease inhibitor turkey ovomucoid third domain. All the predicted pK(a) values are within 0.5 pH units of experiment, with a root-mean-square deviation of 0.31 pH units. We show that the decreased pK(a) values observed for some of the residues are primarily due to hydrogen bonds to the carboxyl oxygens. Hydrogen bonds involving amide protons are shown to be particularly important, and the effect of hydrogen bonding is shown to be nonadditive. Hydrophobic effects are also shown to be important in raising the pK(a). Interactions with charged residues are shown to have relatively little effect on the carboxyl pK(a) values in this protein, in general agreement with experiment.

Calculating pKa values in enzyme active sites

The ionization properties of the active-site residues in enzymes are of considerable interest in the study of the catalytic mechanisms of enzymes. Knowledge of these ionization constants (pKa values) often allows the researcher to identify the proton donor and the catalytic nucleophile in the reaction mechanism of the enzyme. Estimates of protein residue pKa values can be obtained by applying pKa calculation algorithms to protein X-ray structures. We show that pKa values accurate enough for identifying the proton donor in an enzyme active site can be calculated by considering in detail only the active-site residues and their immediate electrostatic interaction partners, thus allowing for a large decrease in calculation time. More specifically we omit the calculation of site-site interaction energies, and the calculation of desolvation and background interaction energies for a large number of pairs of titratable groups. The method presented here is well suited to be applied on a genomic scale, and can be implemented in most pKa calculation algorithms to give significant reductions in calculation time with little or no impact on the accuracy of the results. The work presented here has implications for the understanding of enzymes in general and for the design of novel biocatalysts.

On the evaluation and optimization of protein X-ray structures for pKa calculations

The calculation of the physical properties of a protein from its X-ray structure is of importance in virtually every aspect of modern biology. Although computational algorithms have been developed for calculating everything from the dynamics of a protein to its binding specificity, only limited information is available on the ability of these methods to give accurate results when used with a particular X-ray structure. We examine the ability of a pKa calculation algorithm to predict the proton-donating residue in the catalytic mechanism of hen egg white lysozyme. We examine the correlation between the ability of the pKa calculation method to obtain the correct result and the overall characteristics of 41 X-ray structures such as crystallization conditions, resolution, and the output of structure validation software. We furthermore examine the ability of energy minimizations (EM), molecular dynamics (MD) simulations, and structure-perturbation methods to optimize the X-ray structures such that these give correct results with the pKa calculation algorithm. We propose a set of criteria for identifying the proton donor in a catalytic mechanism, and demonstrate that the application of these criteria give highly accurate prediction results when using unmodified X-ray structures. More specifically, we are able to successfully identify the proton donor in 85% of the X-ray structures when excluding structures with crystal contacts near the active site. Neither the use of the overall characteristics of the X-ray structures nor the optimization of the structure by EM, MD, or other methods improves the results of the pKa calculation algorithm. We discuss these results and their implications for the design of structure-based energy calculation algorithms in general.

Folded conformation of an immunostimulating tetrapeptide rigin: high temperature molecular dynamics simulation study

Employing high temperature quenched molecular dynamics (QMD) stimulations the conformational energy space of an immunostimulating tetrapeptide rigin: H-Gly341-Gln-Pro-Arg344-OH, is explored. Using distance dependent dielectric (epsilon =r(ij)) 31 different low energy starting structures with identical sequence were computed for their conformational preferences. According to the hypothesis of O'Connors et al. [J. Med. Chem. 35 (1992), 2870], 83 low-energy conformers resulted from unrestrained molecular dynamics (MD) simulations, could be classified into two energy minimized families: A and B, comprised of 64 (Pro C(gamma)-endo orientation) and 19 (Pro C(gamma)-exo orientation) structures, respectively. An examination of these families revealed the existence of a remarkably similar folded backbone conformation: torsion angles being phi(i+1) approximately -65 degrees, psi(i+1) approximately -65 degrees, phi(i+2) approximately -65 degrees, psi(i+2) approximately -60 degrees, characterizing a distorted type III beta-turn structure across the central Gln-Pro segment. The folded conformation of rigin is devoid of a classical 1 <-- 4 intra-molecular hydrogen bond nevertheless, the conformation is stabilized by an effective 'salt-bridge', i.e., Gly H(3)N(+)...C(alpha)OO(-) Arg interaction. Surprisingly, in both the families the unusual folded side-chain dispositions of the Gln residue favor the formation of a unique intra-residue 'main-chain to side-chain' H-bond, i.e., N(alpha)-H...N(epsilon) interaction, encompassing a seven-membered ring motif. The conformational attributes may be valuable in de novo construction of structure-based drug candidates having sufficient stimulating activity.

The Poisson-Boltzmann equation for biomolecular electrostatics: a tool for structural biology

Electrostatics plays a fundamental role in virtually all processes involving biomolecules in solution. The Poisson-Boltzmann equation constitutes one of the most fundamental approaches to treat electrostatic effects in solution. The theoretical basis of the Poisson-Boltzmann equation is reviewed and a wide range of applications is presented, including the computation of the electrostatic potential at the solvent-accessible molecular surface, the computation of encounter rates between molecules in solution, the computation of the free energy of association and its salt dependence, the study of pKa shifts and the combination with classical molecular mechanics and dynamics. Theoretical results may be used for rationalizing or predicting experimental results, or for suggesting working hypotheses. An ever-increasing body of successful applications proves that the Poisson-Boltzmann equation is a useful tool for structural biology and complementary to other established experimental and theoretical methodologies.

Distance and exposure dependent effective dielectric function

In an effort to develop a dielectric screening function for molecular dynamics simulations of biomolecules in implicit solvent, effective dielectric constants (D(eff)) for a large number of atom pairs in a typical globular protein are calculated by continuum electrostatics. Plots of D(eff) versus the intercharge distance are in general sigmoidal with the characteristics of the curve depending on the distance of the two charges from the dielectric boundary and, secondarily, on the extent to which the area surrounding each charge is occupied by solvent (the "exposure"). The D(eff) values were fitted to an empirical, analytical function of these parameters that reproduces the data reasonably well, although considerable scatter exists in the range of D(eff) from 30 to 80. In the system used for parameterization, the mean square deviation of electrostatic interaction energies with this function is 0.48 kcal/mol, compared to 1.45 for an analytical Generalized Born model and 1.52 for the linear distance-dependent dielectric model. When tested in other proteins of varying size and compactness, the present function is superior to both of the above models, except for a fully unfolded polypeptide chain, where the Generalized Born model is superior.

Focusing of the electrostatic potential at EF-hands of calbindin D(9k): titration of acidic residues

Biological functions for a large class of calmodulin-related proteins, such as target protein activation and Ca(2+) buffering, are based on fine-tuned binding and release of Ca(2+) ions by pairs of coupled EF-hand metal binding sites. These are abundantly filled with acidic residues of so far unknown ionization characteristics, but assumed to be essential for protein function in their ionized forms. Here we describe the measurement and modeling of pK(a) values for all aspartic and glutamic acid residues in apo calbindin D(9k), a representative of calmodulin-related proteins. We point out that while all the acidic residues are ionized predominantly at neutral pH, the onset of proton uptake by Ca(2+) ligands with high pK(a) under these conditions may have functional implications. We also show that the negative electrostatic potential is focused at the bidental Ca(2+) ligand of each site, and that the potential is significantly more negative at the N-terminal binding site.

Optimizing the hydrogen-bond network in Poisson-Boltzmann equation-based pK(a) calculations

pK(a) calculation methods that are based on finite difference solutions to the Poisson-Boltzmann equation (FDPB) require that energy calculations be performed for a large number of different protonation states of the protein. Normally, the differences between these protonation states are modeled by changing the charges on a few atoms, sometimes the differences are modeled by adding or removing hydrogens, and in a few cases the positions of these hydrogens are optimized locally. We present an FDPB-based pK(a) calculation method in which the hydrogen-bond network is globally optimized for every single protonation state used. This global optimization gives a significant improvement in the accuracy of calculated pK(a) values, especially for buried residues. It is also shown that large errors in calculated pK(a) values are often due to structural artifacts induced by crystal packing. Optimization of the force fields and parameters used in pK(a) calculations should therefore be performed with X-ray structures that are corrected for crystal artifacts.

Calculated pH-dependent population and protonation of carbon-monoxy-myoglobin conformers

X-ray structures of carbonmonoxymyoglobin (MbCO) are available for different pH values. We used conventional electrostatic continuum methods to calculate the titration behavior of MbCO in the pH range from 3 to 7. For our calculations, we considered five different x-ray structures determined at pH values of 4, 5, and 6. We developed a Monte Carlo method to sample protonation states and conformations at the same time so that we could calculate the population of the considered MbCO structures at different pH values and the titration behavior of MbCO for an ensemble of conformers. To increase the sampling efficiency, we introduced parallel tempering in our Monte Carlo method. The calculated population probabilities show, as expected, that the x-ray structures determined at pH 4 are most populated at low pH, whereas the x-ray structure determined at pH 6 is most populated at high pH, and the population of the x-ray structures determined at pH 5 possesses a maximum at intermediate pH. The calculated titration behavior is in better agreement with experimental results compared to calculations using only a single conformation. The most striking feature of pH-dependent conformational changes in MbCO-the rotation of His-64 out of the CO binding pocket-is reproduced by our calculations and is correlated with a protonation of His-64, as proposed earlier.