Conformational design of cyclic peptides

Due to their potential importance as drug molecules as well as other applications methods to design small cyclic peptides with a rigid well-defined conformation are useful. Computational methods to predict the conformation of cyclic peptides to identify those with a well-defined conformation allow for screening of potential sequences prior to expending the resources required to make the peptides. An alternative method to produce conformationally restricted cyclic peptides would be to include structural elements that prevent the peptide changing conformation. This thesis focuses on designing well-structured cyclic peptides, either through the use of computational techniques which were developed in order to help predict the structure of cyclic peptides or, through the introduction into the cyclic peptide structure of a β-turn mimic. Chapter 1 covers current methods for the synthesis of cyclic peptides as well as methods for predicting the conformations a cyclic peptide is likely to form. β-turns, including β-turn mimics are also discussed. Chapter 2 focuses on the modification of bias-exchange metadynamics (BE META) simulations, which can be used to predict the conformation of cyclic peptides, to also predict the occurrence of cis proline within proline-containing cyclic peptides. An additional replica was added into the BEMETA to allow for cis/trans isomerisation. A series of cyclic hexapeptides was synthesised and the cis to trans ratio of the proline within the peptides obtained by NMR. These results were then used to evaluate the computational predictions. It was found faster convergence in the simulations was reached using the additional replica, but the forcefield could not always accurately model the energy difference between the cis and trans proline states. Chapter 3 presents the results of the analysis of β-turns found within a database. Cyclic hexapeptides are frequently observed to form a structure composed of two overlapping β-turns. It was hypothesised information on the β-turns extracted from the database could therefore be used to help design cyclic hexapeptides. Two peptides were designed based on the database analysis. The structure of the peptides, determined by NMR, show the amino acids in the peptides occupy the predicted positions in the major conformations. Chapter 4 explores the introduction of restraints into BE-META simulations used to predict the conformations of cyclic peptides. The restrained simulations are used to infer the lowest energy structures a cyclic hexapeptide can adopt based on the backbone conformation of the peptide when a specific β-turn type is present within the peptide. The inclusion of chiral amino acids at specific positions within the cyclic peptide structure is seen to alter the most stable conformation. In Chapter 5 a Random Forest machine learning algorithm was trained to predict the β-turn type a sequence will form based on the β-turns extracted from the database in Chapter 3. The Random Forest in combination with the lowest energy conformations determined by the restrained simulations in Chapter 4 is used to predict the conformations cyclic hexapeptides will adopt. A well-structured cyclic peptide containing the biologically active RGD motif was designed using the methods developed in this chapter. The use of the Random Forest allowed for fast filtering of potential sequences to identify those predicted to form only one major conformation. Chapter 6 focuses on the incorporation of a β-turn mimic, which forms through a chemical ligation reaction, into cyclic peptides. Conditions were found to incorporate the β-turn mimic into the cyclic peptides which allow for the cyclisation of the peptide and formation of the β-turn mimic in a single step. The reaction has a broad sequence tolerance and was found to be suitable for peptide macrocycles of varying size. Chapter 7 aims to introduce additional functionality to the β-turn mimic used in Chapter 6. The structure of the β-turn mimic was modified to include a fluorescent naphthalene group. A peptide containing the modified β-turn mimic was synthesised and circular dichroism (CD) analysis shows the modified β-turn mimic retains the β-turn structure. The fluorescent properties of the modified β-turn mimic were also analysed and found to be very similar to that of tryptophan. Chapter 8 makes use of the β-turn mimic in order to design a cyclic WW Domain mimic which retains the ability to bind proline-rich ligands. Two β-strands from a WW Domain structure were cyclised using the methods developed in Chapter 6. Molecular dynamics simulations show the β-strand structure is retained in the cyclised peptide. Binding studies also demonstrate the cyclised WW domain retains the ability to bind to a ligand known to bind to the wildtype WW Domain