Three simulations with different initial velocity distributions instead of a single trajectory were used to sample the conformational space in the protein-solvent system, because a single trajectory samples only a small fraction of the conformational space than multiple short trajectories [32–34].
The structure of CALB shows a high stability in all solvents and is therefore a useful system to examine the effect of different solvents on structure and flexibility. Multiple MD simulations of each protein-solvent system confirmed that the structures of CALB in different solvent deviated from each other by less than 0.8 Å. This structural difference is within the deviation of 0.6–0.8 Å of the three simulations of the same protein-solvent system. Our observation that the structure of CALB is independent of its environment is supported by the fact that CALB does not undergo conformational transitions  and by a comparison of different crystal structures obtained under different crystallization conditions [3, 35] which show backbone RMSD values below 0.3 Å. Also in most molecular dynamics simulations structures showed no significant changes in different solvents [20, 22, 31] with the exception of Rhizomucor miehei lipase for which a solvent-induced conformational change was observed . In addition, circular dichroism measurements of CALB in different solvents showed that its secondary structure did not change . This is also confirmed by X-ray structures of various serine proteases, crystallized in the presence of small amounts of different organic solvents, which are nearly indistinguishable from their structure in water [37–40]. Flexible and rigid regions of CALB identified in the simulations were similar to regions of high and low B-factors reported in the crystal structure . In contrast to structure, the flexibility is solvent-dependent. In organic solvents, the flexibility of proteins is decreased, which has been confirmed by different experimental techniques such as time-resolved fluorescence anisotropy , ESR , and dielectric relaxation spectroscopy . This was also observed in simulations of lipases  and subtilisin , while no significant differences between water and organic solvents have been observed in simulations of subtilisin .
It is observed that solvents with a lower dielectric constant lead to a decreased protein flexibility as shown by EPR [45, 12], which is in general consistent with the results of our simulations. However, there is one outlier. In ISO the flexibility is higher than expected from its dielectric constant. Interestingly, isopentane is the only solvent in our simulations with one freely rotatable, bond which is not considered in the rigid solvent model. We suppose that a flexible solvent model would be necessary to properly treat the effects of this solvent. It has also been suggested that the size of organic solvent molecules correlates with the protein flexibility , but no correlation was found in our simulations. A further effect of organic solvents in our simulations is a decrease in solvent-accessible surface, especially of the hydrophilic surface. In our simulations, the hydrophilic surface decreased by 600 Å2 from simulations in WAT to CHE, while the hydrophobic surface increased slightly by 50 Å2. It has been suggested that in water polar side chains orient toward the surface, thus increasing the hydrophilic surface and decreasing the polar intra-molecular interactions that mediate the rigidity of the protein , while in organic solvents the surface area is reduced which leads to improved packing and increased stability .
There are in general two types of water molecules observed in organic solvents. The 'inside class' water molecules that are bound in the interior of the protein and can play an important role for active conformation in organic solvents by a hydrogen bond network and the 'contact class' water molecules that are weakly bound to the surface of the enzyme and can be rapidly exchanged by other water molecules . The dynamical properties of surface-bound water molecules differ considerably from the bulk water as shown by X-ray crystallography  and NMR experiments . The residence times of most surface-bound water molecules are between 10 and 100 ps, bound and free water molecules at the surface of a protein are in a dynamic equilibrium . In agreement with these experiments all water molecules at the surface were rapidly exchanged during the simulation time of 2 ns. In correlation to our simulations, where an increased number of less mobile water molecules was observed at higher logP at the surface of CALB, the number of water molecules with high B-factor was decreased by an increasing hydrophobicity of alcohols at lysozyme studied by X-ray . In organic solvents the amount of water molecules in the organic solvent phase is low and therefore the probability for an exchange of bound and structured water at the surface is low. A high rate of exchange of water molecules at the surface, observed in our simulations, might be the reason for an increased flexibility, in agreement to previous observations where the protein mobility increased with an increasing amount of water .
In agreement with experimental results  in the simulations organic solvents strip just a few water molecules from the enzymes surface, while polar solvents strip most water from the surface. This stripping of water from the surface by polar solvents has been already shown in experiments  and in simulations . The solvent dielectric constant correlated with stripping of water molecules in simulations with increasing polarity of the solvent . It was shown in experiments, that desorption is independent of the kind of the protein, increasing with the dielectric constant of the solvent . Like in our simulations of CALB, a spanning water network consisting of several small water clusters was observed in previous simulations [54, 55, 22, 31] and experiments , in which non-polar solvents enhanced the formation of clusters [31, 54]. In agreement to our simulations, the water clusters in a cutinase consisted of 2 to 8 water molecules at a hydration level of 15% (w/w) , depending on the solvent, in the crystal structure of CALB the largest cluster of water molecules with B-factors lower 40 Å2 consists of 14 water molecules at most. The spanning water network resulted by a slow exchange of water molecules at the surface in organic solvents. Polar groups favor direct interactions with water molecules and form hydrogen bonds, while non-polar groups enhance interactions among water molecules and enhance the local structure of neighboring water molecules. The concept of hydrophobic hydration and the freezing of water to clusters around hydrophobic surface was previously suggested  and is supported by the solvent dependent flexibility observed in simulations.
From the results it can be concluded, that the reduced flexibility of CALB in non-polar solvents is not only a consequence of the interaction between organic solvent molecules and the protein, but also due to the interaction with the enzyme-bound water and its exchange on the surface. Despite the higher fluidity of organic solvents, the flexibility of CALB is decreased, because the water exchange at the surface is restricted.