Unfolding and refolding of the native structure of bovine pancreatic trypsin inhibitor studied by computer simulations.

A new procedure for studying the folding and unfolding of proteins, with an application to bovine pancreatic trypsin inhibitor (BPTI), is reported. The unfolding and refolding of the native structure of the protein are characterized by the dimensions of the protein, expressed in terms of the three principal radii of the structure considered as an ellipsoid. A dynamic equation, describing the variations of the principal radii on the unfolding path, and a numerical procedure to solve this equation are proposed. Expanded and distorted conformations are refolded to the native structure by a dimensional-constraint energy minimization procedure. A unique and reproducible unfolding pathway for an intermediate of BPTI lacking the [30,51] disulfide bond is obtained. The resulting unfolded conformations are extended; they contain near-native local structure, but their longest principal radii are more than 2.5 times greater than that of the native structure. The most interesting finding is that the majority of expanded conformations, generated under various conditions, can be refolded closely to the native structure, as measured by the correct overall chain fold, by the rms deviations from the native structure of only 1.9-3.1 A, and by the energy differences of about 10 kcal/mol from the native structure. Introduction of the [30,51] disulfide bond at this stage, followed by minimization, improves the closeness of the refolded structures to the native structure, reducing the rms deviations to 0.9-2.0 A. The unique refolding of these expanded structures over such a large conformational space implies that the folding is strongly dictated by the interactions in the amino acid sequence of BPTI. The simulations indicate that, under conditions that favor a compact structure as mimicked by the volume constraints in our algorithm, the expanded conformations have a strong tendency to move toward the native structure; therefore, they probably would be favorable folding intermediates. The results presented here support a general model for protein folding, i.e., progressive formation of partially folded structural units, followed by collapse to the compact native structure. The general applicability of the procedure is also discussed.