| Proteins interact in solution to perform their biological function, thus studies of solvent impact on protein-protein interactions are crucial for understanding many processes in the cell. However, though solvent is known to be a very important player in protein structure, dynamics and energetics, it is often disregarded due to the complexity inherent to detailed solvent description, which requires computationally demanding approaches. As a consequence, protein interfaces may not be properly described due to the exclusion of interfacial water-mediated interactions. We have developed a database (SCOWLP), which classifies all interfacial protein residues of the PDB into three classes based on physico-chemical properties of the interacting atoms including solvent: dry (direct interaction), dual (direct and water-mediated interactions), and wet spots (residues interacting only through one water molecule). We have observed that wet spots display similar characteristics to residues contacting water molecules in cores or cavities of proteins, and therefore represent an important proportion of protein interfaces, therefore solvent should not be excluded in protein-protein interactions studies. By using a MD approach we have characterized properties of water-mediated protein interactions at residue and solvent level for a representative set of protein complexes. We have shown that wet spots are energetically and dynamically very similar to other interfacial residues, and that water molecules mediating these protein interactions have essentially longer residence time than surface solvent. Calculated free energies indicate that these water molecules significantly affect protein-protein complex formation and stability. An interesting observation is that water molecules in protein interfaces may contribute to the conservation of interactions, therefore, allowing certain sequence variability in the interacting partners. We have aimed to find out if an explicit introduction of a solvent term into a sequence-based protein contacts prediction method could improve the results obtained by this approach. For this, we have used the data on interfacial solvent from the SCOWLP database and implemented it into the concept of correlated mutations, which assumes that interacting protein residues co-evolve so that a mutation in one of the interacting counterparts is compensated by a mutation in the other. We have compared predictions of both intra- and interdomain protein contacts carried out with and without solvent consideration. We have found that the introduction of solvent improves predictions for both datasets significantly. Our results confirm an indispensable role of water in protein interactions, suggesting that taking into account solvent is essential for obtaining accurate description of protein interfaces in important implementations for energetics, folding, docking and rational design. |
|