This thesis describes the development of a molecular simulation methodology to study properties of enzymes in non-aqueous media at fixed thermodynamic water activities. The methodology is applied in a molecular dynamics study of the industrially important enzyme Candida antarctica lipase B (CALB) in water and organic solvents. The effects of solvent on structural and dynamical enzyme properties are studied, and special attention is given to how enzyme properties in organic solvents are affected by the hydration level, which is shown to be related to the water activity. In experimental studies of enzyme kinetics in non-aqueous media, it has been a fruitful approach to fix the enzyme hydration level by controlling the water activity of the medium. In this work, a protocol is therefore developed for determining the water activity in non-aqueous protein simulations. The method relies on determining the concentration of water in a region of the simulation box far from the protein surface. In order to evaluate the corresponding activity, a previously developed methodology based on fluctuation solution theory is employed to compute the excess Gibbs energy of the water/organic solvent mixture. This requires that separate simulations of this mixture are carried out at different compositions, and that the total correlation function integrals, i.e. spatial integrals of the pair radial distribution functions (RDFs), are evaluated. A main challenge is that the total correlation function integrals do not converge within the system size of the simulation box generally used in simulation. Therefore, a method is developed for extending the RDFs to arbitrary distances so that the integrals can be evaluated. The method, which was first used in the classical study of the Lennard-Jones fluid by Verlet (Verlet (1968), Phys. Rev., 165, 201–214), is here extended for application to simulations of molecular fluid mixtures. It extends the RDFs by enforcing that the corresponding direct correlation functions follow a certain approximation at long distances. This approximation is here derived in terms of statistical mechanical fluid theory. An extensive set of numerical tests are carried out for validating the method, and it is found that thermodynamic properties of good accuracy are obtained from the integrals of the extended RDFs. The method is also shown to be at least as good as existing methods for correlation function integration, while for small systems, it seems to be even better. The method is applied to compute the excess Gibbs energy of the mixtures of water and organic solvents used in the simulations of CALB. This allows to determine the water activity of the simulated systems and thus to compare protein properties in different organic solvents at fixed water activities. The study bridges therefore the previously used simulation approach where properties were compared at similar hydration levels (Yang et al (2004), Biophys. J., 87, 812–821); Micaêlo and Soares (2007), FEBS J., 274, 2424–2436; Trodler and Pleiss (2008), BMC Struct. Biol., 8) and the approach to fix the water activity which often is used in experimental studies. The water activity is shown to have a profound effect on the structure and dynamics of CALB. Conformational flexibility, for instance, increases with increasing hydration in acetone, t-butanol, methyl t-butyl ether and hexane, but not in methanol. A consequence of this is that hydration needs to be carefully considered in simulation studies of proteins in organic media. The organic solvent is also shown to affect structure and dynamics of CALB. The effects on flexibility can partially be attributed to the mobility of the hydration water, as proposed in a previous study (Trodler and Pleiss (2008), BMC Struct. Biol., 8). The present results indicate that flexibility may also be affected by adsorption of organic solvent molecules to the enzyme surface. This seems in particular to be the case in t-butanol in which the lowest flexibility of CALB is observed. Future applications of the methodology may lead to an improved understanding of enzyme properties in non-aqueous media, which may have significant impact on the development of rational strategies for solvent selection in biocatalysis.
Main Research Area:
Peters, Günther H.J., Abildskov, Jens
Technical University of Denmark, Department of Chemical Engineering, 2011