Water and proteins: a love-hate relationship.

I t is widely appreciated that water molecules play an invaluable role in governing the structure, stability, dynamics, and function of biomolecules. The hydration forces are responsible for packing and stabilization of the protein structure. Particularly, water participates in many hydrogen bond networks and screening electrostatic interactions. However, the exact range of processes mediated by water is far from being understood, and it is only in the recent years that water has been quantitatively treated as an integral component of biomolecular systems. In this issue of PNAS, Papoian et al. (1) report a significant improvement in protein structure prediction by adding a water knowledge-based potential to an established Hamiltonian for protein structure prediction. ‘‘Wetting’’ the Hamiltonian improves the predicted structures, especially for large proteins, when longrange interactions between polar or charged groups are mediated by water molecules. There are a variety of experimental and theoretical studies acknowledging the active role of solvent in protein stability and dynamics. Experimentally, xray, neutron diffraction (2), NMR (3, 4), and femtosecond fluorescence (5) measurements reveal the binding sites, structure, and dynamics of water. Theoretically, molecular dynamics simulations offer a detailed atomic description of both the biomolecule and the solvent as well as the time dependency of their dynamics (6–9). These all-atom explicit– solvent simulations of proteins, however, do not appear to provide sufficient conformational coverage to tackle many equilibrium and long-time-scale kinetic properties. Thus, a complementary approach is to adopt simplified models that trade high structural resolution of the water molecule (10, 11) (e.g., generalized born model for the water) or of both the polypeptide and the solvent (e.g., structure-based models) for enhancing conformational sampling (12, 13). Desolvation during folding processes was studied by using a structure-based (Go) model for the Src homology 3 (SH3) protein assuming that each native contact is formed only after expelling a water molecule that mediates the interaction between any two residues (13). The fully solvated unfolded chain undergoes an initial structural collapse to an overall native topological conformation that is followed by a second transition where water molecules are cooperatively squeezed out from the hydrophobic core region, resulting in a dry and packed protein. Gating protein folding by solvent, as implemented in the energetically minimally frustrated model, provides additional microscopic features of the folding events, which have been observed experimentally (13, 14). Atomistic simulation studies of SH3 and proteins A and G support the role of water as a lubricant for the packing of the hydrophobic core after the formation of the transition state (6–8). Moreover, these fully atomic simulations, which are not biased toward direct contacts between the residues, point out that the