Predicting interaction sites from the energetics of isolated proteins: a new approach to epitope mapping

Abstract

An increasing number of functional studies of proteins have shown that sequence and structural similarities alone may not be sufficient for reliable prediction of their interaction properties. This is particularly true for proteins recognizing specific antibodies, where the prediction of antibody-binding sites, called epitopes, has proven challenging. The antibody-binding properties of an antigen depend on its structure and related dynamics. Aiming to predict the antibody-binding regions of a protein, we investigate a new approach based on the integrated analysis of the dynamical and energetic properties of antigens, to identify nonoptimized, low-intensity energetic interaction networks in the protein structure isolated in solution. The method is based on the idea that recognition sites may correspond to localized regions with low-intensity energetic couplings with the rest of the protein, which allows them to undergo conformational changes, to be recognized by a binding partner, and to tolerate mutations with minimal energetic expense. Upon analyzing the results on isolated proteins and benchmarking against antibody complexes, it is found that the method successfully identifies binding sites located on the protein surface that are accessible to putative binding partners. The combination of dynamics and energetics can thus discriminate between epitopes and other substructures based only on physical properties. We discuss implications for vaccine design.

Figure 1

Pictorial representation of the MLCE method. The contact map is multiplied by the simplified energy-coupling matrix. The resulting matrix reports the energetic coupling intensity of two residues in contact in space, represented as a color scale assigned to each point of the matrix. The weakest local interactions vanish in the background color: predicted epitopes are identified with circles.

Figure 2

Projection of the low-energy couplings from MLCE on their respective locations on the three-dimensional structure of all proteins analyzed. Predicted epitopes are in red, actual epitopes are in blue, and their intersection is in purple. Color code: see online version for clarity.

Figure 3

Examples of multiple antibodies binding to the same antigen (light colored), highlighting the possibility for one protein to possess multiple epitopes. The PDB code of the antigen is close to the yellow antigen, and the PDB codes of the complexes are near each respective antibody. (a) Lysozyme; (b) the von Willebrand A3 factor; and (c) human chorionic gonadotropin. Color code: see online version for clarity.