Late-stage reshaping of phage-displayed libraries to macrocyclic and bicyclic landscapes using multipurpose linchpin

Abstract

Genetically-encoded libraries (GEL) are increasingly used for discovery of ligands for ‘undruggable’ targets that cannot be addressed with small molecules. However, there is currently no consensus whether ‘zero knowledge’ naïve libraries or libraries with pre-existing knowledge can offer a more effective path to discover molecular interactions. In this manuscript, we evaluated the feasibility of discovery of macrocyclic and bicyclic peptide from "non-zero knowledge" libraries. Foundational GEL platforms like phage-, yeast-, ribosome- and mRNA-display originally enabled display of libraries composed of 20 natural amino acids (20AA). Today, numerous strategies expand GEL beyond 20AA space. Upgrading libraries by incorporating unnatural amino acids (UAA) and chemical post-translational modification (cPTM) can build linear, cyclic, and bicyclic peptides that would not be achievable for 20AA-GEL. The standard operating procedure for UAA and cPTM libraries starts from a "naïve" library with 108-1012 compounds, uses target of interest and rounds of selection to narrow down to a set of receptor binding hits. However, such approach uses zero knowledge of natural peptide-receptor interactions which already exists in libraries with 20AA space. We approach this problem by late-stage chemical reshaping of phage-displayed landscape of 20AA binders to NS3aH1 protease. The re-shaping is performed under a novel multifunctional C2-symmetric linchpin, 3,5-bis(bromomethyl)benzaldehyde (termed KYL), that combines two electrophiles that react with thiols and aldehyde group that reacts with N-terminal amine. KYL diversified phage-displayed peptides into bicyclic architectures and delineates 2 distinct sequence populations which was further validated by BLI: (i) retained binding upon bicyclization (ii) lost binding once chemically modified. Our report provides a case study for discovering advanced, chemically-upgraded macrocycles and bicycles from libraries with pre-existing knowledge. The results imply that thousands of selection campaigns completed in 20AA space, in principle, can serve for late-stage reshaping and as a starting point for discovery of advanced peptide-derived ligands.