Stapled by design, new peptide-based therapeutic leads targeting protein-protein interactions

Abstract

Interest in the development of peptide-based therapeutics has increased in recent years, mainly due to the relatively low toxicity of both the peptides and their metabolites, and the ready availability of building blocks and ease of synthesis, which allows for facile structural variation. Peptide binding sites generally provide larger contact surfaces than small molecules, and sometimes more exposed interaction sites, making the regulation of large protein-protein interfaces (PPIs) possible. However, in contrast to antibodies and other proteins, in most cases natural short peptides (10 to 20-mers) do not retain a stable secondary conformation, which can lead to a decreased affinity. Indeed, most of the peptide drugs that have reached the market are exceptions, with the ability to retain a compact, stable conformation (for example macrocyclic peptides or peptides stabilised with disulphides bridges). Thus, constrained peptides, which are forced to retain a rigid conformation, might be better candidates for protein binding. ‘Stapling’ peptides can induce an α-helical structure by introducing ’stapled’ residues acting as a synthetic brace, holding two residues of adjacent helix-turns close together. Interest in the development of stapled peptides has markedly increased over the last decade, due to early successes and the increasing availability of staple precursors. The handful of successful strategies to design these novel peptides that have already emerged are discussed in Chapter 1, together with significant examples that have influenced research in the field.

Among computer-aided drug design techniques, Molecular Dynamics (MD) simulations and associated methods have earned increased attention within the pharmaceutical sector due to improvements made in the last decades. However, there are still limitations to the use of these methods for peptide development and design, due to the lack of protocols for using unnatural amino acids in MD workflows. In this thesis, MD simulations are used to design peptides containing non-proteinogenic amino acids and assess their properties. In Chapter 2, the fundamental principles of computational chemistry and MD used in this project are introduced.

Chapter 3 describes the development of a pipeline to generate and validate parameter sets automatically, which can be applied in MD simulations for modelling unnatural amino acids commonly used in staple chemistry. The parameters were found to be in good agreement with quantum mechanics data and fully compatible with the AMBER14SB forcefield.

The unnatural amino acids also reproduced the behaviour of natural amino acids in terms of psi/phi distribution. In Chapter 4, these parameters were integrated into a custom solute tempering protocol to sample the conformational space of stapled peptides. This method was benchmarked against the more ‘traditional’ T-REMD and was found to be computationally less intensive while producing similar sampling. A set of well-documented stapled peptides, based on a helical peptide from the tumour suppressor protein p53, which binds the negative regulator MDM2, was used to probe the accuracy of the MD simulations. It was found that the experimentally determined helical content of these stapled peptides, reported in the literature, could be successfully reproduced.

This first validation of the new forcefields for stapled residues, allowed the development of a computational platform to design helical peptides from known PPIs called ‘Stapline’. This pipeline extracts key binding interactions of known protein binders and builds models that replicate these interactions using the side chains of stapled helical peptides. The application of the ‘Stapline’ platform to the in-silico design of two classes of stapled peptides that selectively recognise the receptor binding domain of the “Spike” protein on the SARS-CoV-2 virus is described in Chapter 5. The work was extended to a third class of stapled peptides, using studies of coil systems to design peptides to the tail of the Spike protein. The binding affinity of these fluorescent, stapled peptides to the SARS-CoV-2 virus were assessed against known Spike-protein binding peptides in biophysical assays.

This ‘Stapline’ procedure was then integrated into a second peptide design workflow to optimise the binding selectivity of IL1-R inhibitors as described in Chapter 6. IL1-R is a protein involved in many diseases including rheumatoid arthritis, as it is one of the key proteins in the inflammatory pathway. Peptide inhibitors of the IL-R1/IL-1 interaction developed to date have nanomolar potency but have short half-life. Thus, there is a need for new peptides with improved drug-like properties. Following MD simulations, a MMPBSA (Molecular Mechanics Poisson-Boltzmann Surface Area) and MMGBSA (Molecular Mechanics Generalised Born Surface Area) protocol was employed. The efficiency of this protocol in assessing the binding of the peptides to their target, as well as predicting their interactions was evaluated. By combining MD descriptors to quantify peptide stability and relative free energy of binding obtained from MMPBSA and MMGBSA, we suggested favourable stapled peptide designs. A selection of highest-ranked stapled peptide inhibitors were characterised using biophysical techniques, including FRET, Biacore and a cell-based assay, leading to the successful identification of novel potent binders.