Epoxide hydrolase I from Solanum tuberosum (StEH1) and isolated variants thereof has been studied for mapping structure-function relationships with the ultimate goal of being able to in silico predict modifications needed for a certain activity or selectivity. To solve this, directed evoultion using CASTing and an ISM approach was applied to improve selectivity towards either of the enantiomeric product diols from (2,3-epoxypropyl)benzene (1). A set of variants showing a range of activites and selectivities was isolated and characterized to show that both enantio- and regioselectivity was changed thus the enrichment in product purity was not solely due to kinetic resolution but also enantioconvergence. Chosen library residues do also influence selectivity and activity for other structurally similar epoxides styrene oxide (2), trans-2-methyl styrene oxide (3) and trans-stilbene oxide (5), despite these not being selected for.    The isolated hits were used to study varying selectivity and activity with different epoxides. The complex kinetic behaviour observed was combined with X-ray crystallization and QM/MM studies, powerful tools in trying to explain structure-function relationships. Crystal structures were solved for all isolated variants adding accuracy to the EVB calculations and the theoretical models did successfully reproduce experimental data for activities and selectivities in most cases for 2 and 5.  Major findings from calculations were that regioselectivity is not always determined in the alkylation step and for smaller and more flexible epoxides additional binding modes are possible, complicating predictions and the reaction scheme further. Involved residues for the catalytic mechanism were confirmed and a highly conserved histidine was found to have major influence on activity thus suggesting an expansion of the catalytic triad to also include H104. Docking of 1 into the active site of the solved crystal structures was performed in an attempt to rationalize regioselectivity from binding. This was indeed successful and an additional binding mode was identified, involving F33 and F189, both residues targeted for engineering. For biocatalytic purpose the enzyme were was successfully immobilized on alumina oxide membranes to function in a two-step biocatalytic reaction with immobilized alcoholdehydrogenase A from Rhodococcus ruber, producing 2-hydroxyacetophenone from racemic 2.

Recent years have seen tremendous developments in methods for computational modeling of (bio-) molecular systems. Ever larger reactive systems are being studied with high accuracy approaches, and high-level QM/MM calculations are being routinely performed. However, applying high-accuracy methods to large biological systems is computationally expensive and becomes problematic when conformational sampling is needed. To address this challenge, classical force field based approaches such as free energy perturbation (FEP) and empirical valence bond calculations (EVB) have been employed in this work. Specifically: Force-field independent metal parameters have been developed for a range of alkaline earth and transition metal ions, which successfully reproduce experimental solvation free energies, metal-oxygen distances, and coordination numbers. These are valuable for the computational study of biological systems. Experimental studies have shown that the epoxide hydrolase from Solanum tuberosum (StEH1) is not only an enantioselective enzyme, but for smaller substrates, displays enantioconvergent behavior. For StEH1, two detailed studies, involving combined experimental and computational efforts have been performed: We first used trans-stilbene oxide to establish the basic reaction mechanism of this enzyme. Importantly, a highly conserved and earlier ignored histidine was identified to be important for catalysis. Following from this, EVB and experiment have been used to investigate the enantioconvergence of the StEH1-catalyzed hydrolysis of styrene oxide. This combined approach involved wildtype StEH1 and an engineered enzyme variant, and established a molecular understanding of enantioconvergent behavior of StEH1. A novel framework was developed for the Computer-Aided Directed Evolution of Enzymes (CADEE), in order to be able to quickly prepare, simulate, and analyze hundreds of enzyme variants. CADEE’s easy applicability is demonstrated in the form of an educational example. In conclusion, classical approaches are a computationally economical means to achieve extensive conformational sampling. Using the EVB approach has enabled me to obtain a molecular understanding of complex enzymatic systems. I have also increased the reach of the EVB approach, through the implementation of CADEE, which enables efficient and highly parallel in silico testing of hundreds-to-thousands of individual enzyme variants.