Design of protein-protein interactions via beta-strand pairing

The design of new protein-protein interfaces is a test of our understanding of protein interaction biophysics and can provide new tools to understand cell biology. Methods to accurately design high-affinity interactions have not been established, making it necessary to devise new strategies to facilitate the design process. Solvent exposed main chain atoms on beta-strands are prone to interact with other exposed strands and could serve as the basis for the design of a novel interaction. This dissertation describes the application of beta-strand pairing to design homodimeric and heterodimeric complexes. It also addresses the successes and failures in computational interface design to determine how design methods need to be improved.;One of the most common ways that proteins interact is the formation of symmetric ho- modimer. A way to test the hypothesis that beta-strand mediated interactions can be accurately designed is to redesign a monomeric protein to form a symmetric homodimer via beta-strand pairing. A computational method in Rosetta was used to find monomeric proteins with exposed beta-strands then redesign them to form a symmetric homodimer by pairing exposed beta-strands to form an intermolecular beta-sheet. A crystal structure of one designed complex closely matches the computational model (RMSD = 1.0 A). This work demonstrates that beta-strand pairing can be used to computationally design new interactions with high accuracy;After successful design of a homodimer, beta-strand pairing can be extended design to heterodimers. A computational protocol is described that identifies proteins with exposed strand capable of pairing with an exposed strand on a target protein. The interface of the identified protein is then redesigned to allow it to bind to the target protein. Experimental testing of proteins designed to bind RalA and PCSK9 show that no interaction is made. Directed evolution of the scaffold proteins could allow them to bind to their target.;Most computational protein interface designs from our laboratory and others fail to form when tested experimentally. Successful and failed protein interface designs were examined to see if they provided answers about what works in interface design. Successful designs were, in general, more hydrophobic than failed designs and had few designed hydrogen bonds buried at the interface. Rosetta designed hydrogen bonds were found not to match hydrogen bond distributions observed in high resolution crystal structures. The hydrogen bonding portion of the energy function needs to be improved to allow for design of polar interfaces similar to those found in native proteins.