Computationally Guided Design of Helical Peptide Interaction Reagents

Protein-protein interactions regulate all cellular processes and are attractive targets for therapeutic inhibition. The long-term goal of the proposed work is to accelerate the discovery of modified helical peptides that can be used as protein-protein interaction inhibitors in research, diagnosis and therapy. The short-term goals are to develop new, integrated computational and experimental methods that will deliver potent and selective inhibitors of Bcl-2 proteins. Anti-apoptotic Bcl-2 proteins are important in many cancers, where their over- expression counteracts cell-death signaling. Bcl-2 proteins provide resistance to chemotherapy, making them high-priority oncology targets. Many Bcl-2 protein interactions involve a well-conserved binding groove that engages short alpha helices of ~20 residues, called BH3 helices, in partner proteins. Synthetic peptides that mimic BH3 helices can inhibit anti-apoptotic function and lead to cell death. However, there are multiple members of the Bcl-2 family, and not all BH3 peptides are equally effective inhibitors of all Bcl-2 proteins. An important goal is to discover high-affinity and selective inhibitors for each family member. Another challenge is that engineered peptides are highly susceptible to proteases and have trouble crossing cell membranes, limiting their utility as reagents. Recent work has shown that chemical modifications that stabilize helices can improve their properties. The specific aims of this proposal are organized around tightly coupled computational and experimental techniques that will deepen our understanding of what makes a good helical-peptide inhibitor and help us discover useful molecules more efficiently.

The first step will be to use computational structure- based methods to design peptides predicted to bind tightly and selectively to Bcl-2 family members Bfl-1 and BHRF1. This information will be used to design combinatorial libraries of ~107 peptides focused on high-priority candidates. Libraries will be screened for molecules with desired properties in a yeast-surface display procedure that will provide feedback about the quality of the computational library design methods. The best peptides from yeast display will be further characterized using biophysical measurements in solution and x-ray crystallography. Computational model building and analysis will help establish determinants of binding affinity and specificity. Finally, the best peptides resulting from these procedures will be further optimized using chemical techniques that introduce stabilizing crosslinks into helices. Current insights into what makes good vs. poor crosslinking modifications are limited. In this work, detailed molecular dynamics simulations of modified and unmodified peptides will be carried out to build our understanding of how altered peptide structure affects binding. Overall, this work wil deliver new molecules that target important cancer-regulating proteins, new computational methods that will speed the discovery of selective peptide binders, and a better understanding of the biophysical determinants of helical-peptide interactions.