Methods for the generation of large combinatorial macrocycle libraries

Macrocycles are an attractive class of molecules due to their good binding properties and yet rather small size that allows, in many cases, crossing membranes to reach intracellular targets. In comparison to classical small molecule drugs, macrocycles are generally better at binding to challenging targets such as proteins having flat, featureless surfaces or protein-protein interactions. However, the development of macrocycle-based ligands has been challenging due to the lack of sufficiently large libraries of macrocyclic compounds for screening. The preparation of large numbers of macrocycles is limited due to the chromatographic purification that is typically required following low-yielding cyclization reactions. As a result, the full potential of macrocycles cannot be exploited in drug discovery efforts.

The goal of my thesis was to develop new methods for the efficient generation of large macrocycle libraries. Towards this end, I envisioned to establish a new macrocycle library synthesis principle in which a large panel of m small cyclic peptides are combinatorially ligated in microwell plates with a large number of n chemical fragments, and the resulting m x n products screened without purification. For accessing large libraries, I planned to perform the reactions and screens at a nanomole scale and using contact-less liquid transfer.

In a first project, I developed a method for obtaining readily pure macrocyclic peptides directly from solid phase. Peptide synthesis took place on a disulfide linker to solid phase, which allowed for side chain deprotection with acid without simultaneous peptide cleavage. When a thiol group was included at the N-terminus, treatment with base in DMSO resulted in an intramolecular disulfide exchange to release peptides cyclized by a disulfide bond. The method was tolerant of diverse peptide sequences, and released all tested compounds in high purity. With this method, I was able to synthesize cyclic peptides in 96-well plates, and thus hundreds of peptides in parallel with a moderate effort. I screened a test library and identified a weak inhibitor of thrombin (Ki = 13 ± 1 µ­M), which validated the method.

In the second project, I intended to combinatorially diversify the above macrocyclic peptides with chemical fragments in nanoliter volumes, as described above. For chemical ligation, I chose to acylate amines built into the macrocyclic peptides with carboxylic acid building blocks. Following proof-of-concept experiments, I diversified a 384-member library with 10 carboxylic acids in nanoliter volumes to give 3,840 macrocyclic compounds, which I screened against thrombin. I identified a potent thrombin inhibitor (Ki = 44 ± 1 nM), and a co-crystal structure was obtained which confirmed binding to the active site of the protease. In a final application, I generated and screened a library of 19,968 macrocycles against MDM2, an important oncology target forming a protein-protein interaction with p53. The screen