How Structural Details Influence the Result of pKa Calculations in Proteins
Tim Meyer · Universitätsbibliothek der FU Berlin Hochschulschriftenstelle u. Dokumentenserver · 2015
The pH dependent protonation of amino acids is important for many functions and properties of proteins. Titratable residues determine the pH stability of a protein, the efficiency of enzymatic reactions and they play an important role in proton and ion transport. To understand the function of many proteins on a molecular level in detail it is therefore crucial to have computational tools available that allow a reliable computation of the protonation behavior of these residues; a property usually described in terms of pKa values. The pKa of a residue is mainly influenced by electrostatic interactions, which in turn are sensitive to structural details of the protein in the environment of the residue. Within this thesis two new procedures have been developed that address the challenge of accurate pKa computation in proteins. Usually the interior of a protein is hydrophobic. Nevertheless in some proteins there are deep pockets or cavities that are filled with water molecules. These water- filled cavities offer two difficulties for pKa computation. First, crystal structures of proteins are an unreliable source of information about these buried waters, since waters are often not resolved in crystal structures since they are disordered. Second, waters that were found in the protein structures have not been considered correctly in electrostatic energy calculations. The first new procedure Karlsberg+(cav) solves these problems by introducing an algorithm that is capable of reliably locating protein cavities and correcting the electrostatic energy calculations performed by the software Karlsberg+ based on the detected cavities. This procedure significantly improved the accuracy of the pKa computations for a set of SNase variants, of which many do contain buried water molecules in the corresponding crystal structures. While the first procedure was a correction for Karlsberg+, the second new procedure KB2+MD is a complete rework of its underlying concept. To account for the protonation dependent conformational variability of a protein, Karlsberg+ generates a set of structures that are subtle modeled crystal structures, each representing the protein at a different pH interval. The new KB2+MD procedure generalizes this idea by performing short molecular dynamic simulations for a set of different protonation patterns. Structures taken from these simulations are then analyzed with electrostatic energy calculations to obtain pKa values. Extensive benchmark calculations on 194 residues in 13 standard proteins have been performed to optimize the procedure. It was found that the exclusion of intramolecular 1-2, 1-3 and 1-4 interactions, the so called “non-bonded exclusion”, is essential to obtain good agreement between calculated and measured pKa values. Furthermore the use of an alternative set of vdW radii to define the molecular surface of a protein, as well as energy minimization of the whole water box with a dielectric constant of ε = 4 prior to electrostatic energy calculations increased the accuracy of the pKa computations. With the optimized KB2+MD procedure the accuracy could be significantly improved compared to Karlsberg+ and the result is now on a pair with the latest version of the widely used empirical prediction scheme PropKa, yielding a pKa RMSD of 0.79. Since both new procedures address different aspects of the pKa prediction protocol, they could be combined directly. A reduced set of SNase variants has been used to compare the new procedures individually to the combination of both. While each of the new procedures vastly improved the accuracy of pKa calculations, the combination of them yielded an even better agreement with the experimental values. It turned out that for the combined procedure the details of the cavity determination are less important than they are for Karlsberg+(cav). While the resulting pKa RMSD of 2.3 is still far from the value obtained for the standard benchmark set, the results mark an important step forward in the effort to compute precise pKa values for the challenging SNase variants. In a collaboration project the influenza virus protein hemagglutinin was studied. Hemagglutinin triggers the fusion of virus and host cell membranes in response to acidification of the late endosomes. Also being thought to play an important role in the strategy of the virus to adapt to different hosts and transmission modes, the pH-sensing mechanism of the protein is still not understood. Based on experimental studies and computational modeling a specific histidine was identified to be one of the key elements of this mechanism. Additionally, a detailed model on how this histidine regulates the pH dependent fusion was developed.