Development of computational approaches for knowledge-driven protein engineering aimed at improving thermostability
Prakash Chandra Rathi · Univ. Duesseldorf: Duesseldorfer Dokumenten- und Publikationsserver · 2015
High thermostability is a desired property for proteins, in particular for their use in industrial bio-catalysis. A bio-catalyst with elevated thermostability allows carrying out catalysis at higher temperatures and, thus, to increase the rate of a reaction. However, most proteins in nature are not optimized to withstand harsh industrial process conditions including high temperatures. Therefore, engineering existing proteins by mutagenesis is often employed in order to produce thermostable variants. However, only a minor fraction of the large number of theoretically possible mutated variants of a protein can be tested experimentally for their thermostability. Computational approaches that predict weak spots, residues that are more likely to increase a protein’s thermostability upon mutation, are hence promising for protein engineering projects aimed at improving thermostability. In this thesis, I developed such computational approaches, a task that was subdivided in three main parts: I) To identify the most significant non-covalent interactions that determine a protein’s thermostability. II) To develop and improve approaches for predicting thermostabilizing mutations on a protein based on the outcome of part I. III) To validate these approaches by their retrospective and prospective application on test systems. In this compilation thesis, I showed that hydrophobic interaction energy is the most discriminating factor between mesophilic and (hyper)thermophilic protein homologs. Using this information, I developed an approach for predicting weak spots based on residue-wise hydrophobic interaction energies (Publication I). For the first time, I showed that the size of residue clusters that are identified based on residue-wise hydrophobic interaction energies discriminates mesophilic and (hyper)thermophilic proteins much better than the existence or size of clusters of hydrophobic residues alone. Based on this finding, I improved the rigidity theory-based Constraint Network Analysis (CNA) approach for predicting protein thermostability and weak spots by modeling hydrophobic interactions in a temperature-dependent manner, in addition to performing an ensemble-based CNA (Publication II). For an easy setup and extensive analysis of CNA calculations, we developed a software package, a web service, as well as a graphical user interface that facilitate protein engineering for improving thermostability (Publications III V). Next, I used CNA to study the relation of a protein’s structural rigidity and its thermodynamic thermostability using BsLipA as a test case for which thermodynamically thermostabilized variants are reported in the literature (Publication VI). For the first time, my systematic comparative study of BsLipA variants revealed that thermodynamic thermostabilization is unequivocally accompanied by increased structural rigidity, leading to a significant and good correlation between structural rigidity and thermodynamic thermostability of these variants. Finally, in order to validate the CNA approach, I developed a computational strategy for predicting thermostabilizing mutations and applied it to lipase A from Bacillus subtilis (BsLipA) prospectively. Experimental testing confirmed the predicted thermostabilization for three out of twelve mutated BsLipA variants (Publication VII). This study demonstrated, for the first time, that CNA can be applied prospectively for enriching thermostabilizing mutations on a protein. I am confident that the speed and prediction accuracy of these new and improved computational approaches will allow to investigate the basis of protein thermostability, to predict mutations that increase thermostability, and thus to improve the efficacy and efficiency of protein engineering projects.