Strategies for the evolution of macrocyclic peptide inhibitors of protein-protein interactions

Thesis (Ph. D.)--University of Rochester. Department of Chemistry, 2018.Macrocyclic peptides are an attractive structural class for the inhibition of protein-protein interactions, and thus appealing scaffolds for design of biologically active probes and therapeutics. Macrocyclic peptides are pre-organized, structurally diverse molecules that are capable of inhibiting difficult-to-target protein-protein interactions. With macrocyclic peptides becoming increasingly important agents in the development of new bioactive molecules and therapeutics, methods to generate, diversify and screen macrocyclic peptides for the desired activity are critical. This work introduces new strategies for the efficient generation, evolution and selection of genetically encoded macrocyclic peptides as inhibitors of protein-protein interactions. This thesis explores two main approaches for the evolution and functional selection of genetically encoded macrocyclic peptides. In both systems, macrocycles are generated as stable thioether sidechain-to-sidechain linked ring structures introduced through the spontaneous reaction of cysteine with a non-canonical amino acid, O-2-bromoethyl-tyrosine, incorporated ribosomally via amber stop codon suppression. The first strategy is a medium throughput platform, in which libraries of the macrocycles are generated in E. coli with N-terminal FLAG-tags and subject to a plate based functional assay to evaluate binding strength to a target of choice. This approach was successfully used to identify a first-in-class macrocyclic peptide inhibitor of the oncogenic Sonic Hedgehog/Patched 1 protein-protein interaction. Starting from a linear peptide known to bind Sonic Hedgehog with modest affinity, macrocyclic peptide libraries were generated in E. coli and screened in a plate-based functional assay. Over a few rounds of affinity maturation, a macrocyclic peptide was identified that is able to bind Sonic Hedgehog with high (submicromolar) affinity and reduce Hedgehog signaling in mammalian cells. In a second strategy, the production of genetically encoded macrocyclic peptides was integrated with M13 phage display to obtain a high throughput functional selection platform. This system was developed by generating large macrocyclic peptide libraries as N-terminal fusions to the viral coat protein pIII of bacteriophage M13, and subjecting these libraries to affinity selection experiments against a panel of target proteins (Streptavidin, Sonic Hedgehog, and Kelch-like ECH-associated protein 1). Through this effort, unique and potent macrocyclic peptide binders for each of the targeted proteins were obtained, thus validating the value of this platform for the rapid identification of genetically encoded macrocyclic peptide inhibitors of protein-protein interactions. Because both platforms rely on the genetic incorporation of a non-canonical amino acid to mediate peptide macrocyclization, a two-tier strategy was developed to facilitate the directed evolution of orthogonal aminoacyl tRNA synthetases (AARSs) for amber stop codon suppression. Achieving efficient amber stop codon suppression is a major bottleneck in the generation of genetically encoded macrocycles due to comparatively lower efficiency of non-canonical amino acid incorporation by engineered orthogonal aminoacyl tRNA synthetases. Work described in this thesis introduces and describes an AARS evolution platform which was successfully applied to evolve a pyrrolysyl–tRNA synthetase variant with improved efficiency toward mediating the incorporation of Nϵ‐crotonyl‐lysine and other lysine analogues via amber stop codon suppression. Mutations identified in screening were shown to be transferable and improved the respective amber stop codon suppression efficiency of other evolved AARS variants