Simulation-guided design of membrane active peptide-based therapeutics for biomedical applications

Student thesis: Doctoral ThesisDoctor of Philosophy


Molecular dynamics (MD) simulations is a powerful tool for protein engineering, and it provides atomic details of the protein structure and dynamic events. However, MD simulations usually require an initial frame of the protein structure. It was difficult to predict the secondary structure and function from an unstructured peptide sequence due to the limited simulation timescale by the computing power. Recent developments have significantly improved the computer performance and the accuracy of the force field to make it possible for de novo protein design. 
Here, I explore how polypeptides fold and assemble in aqueous buffer and how membrane-active peptides (MAPs) interact and assemble in membrane of different compositions, using a combination of MD simulations and experimental techniques. First, the MD simulations were challenged with several known protein fragments and peptides to confirm its accuracy. The experimentally validated MD simulations was utilized to develop a proline-rich attachment domain (PRAD) to facilitate the tetramer formation of the butyrylcholinesterase (BChE) C-terminal fragment in the aqueous condition by using hydrophobic effect. However, protein folding in a two-phase lipid-water system is more challenging because the peptide can fold differently in water and lipid and between the interface. Several MAPs were studied using experimentally validated MD simulations to understand how they interact with the cell membranes: melittin, hylaseptin P1 (HSP1), maculatin, and transactive response DNA-binding protein 43 (TDP-43) C-terminal fragment. The results showed that the MD simulations can be compared with the experiments and are reliable. 
Second, the MD simulations were applied to study the mechanisms of MAP folding and pore assembly in a model bacterial membrane for de novo antimicrobial peptide (AMP) design. The atomic detail information of how pores form was used to develop a novel AMP: LDKA. This simple peptide is composed of a small number of amino acids and shows powerful pore-forming properties. Third, statistical analyses of several thousand AMPs were performed to optimize the LDKA template sequence, synthesized ~3,000 peptides using solidphase peptide synthesis, and applied liposome dye leakage assays in each model bacterial and mammalian membrane vesicles to screen for potent peptides that have significant selectivity to target the cells. The size of the fluorescent dyes and the lipid compositions can be controlled, which allowed us to fine-tune the pore size and the binding selectivity for different membrane types. From this screen, nine LDKA analogues with different membrane selectivity and pore sizes were identified, and the selected peptides were sequenced and characterized by in-vitro experimental techniques and MD simulations. 
Finally, I synthesized a small combinatorial peptide library (36 peptides) based on the LDKA template, fine-tuning the charge distribution on the polar face, while maintaining the positions of the previously-optimized hydrophobic residues. These new peptides were evaluated against several cell lines to assess their efficacy as potential cancer therapeutics: MCF-10A (human breast epithelial cell); HMLER (human mammary cancer epithelial cell); HMLER-shEcad (human mammary cancer epithelial stem cell); U2OS (human bone osteosarcoma epithelial cell); and HEK293T (human embryonic kidney cell). Measurements of the half maximal inhibitory concentrations (IC50) shows that several peptides are effective against cancer cells and mammospheres at low micro-molar concentrations and have selectivity toward cancer cells. Remarkably, the peptides of which were composed of D-form amino acids can further improve peptide potency against cancer cells at nanomolar concentrations. Furthermore, these peptides can improve anticancer activity of conventional anticancer agents and reduce their cytotoxicity to normal cell lines. 
Overall, this study provides a proof-of-concept for de novo peptide design of combining molecular dynamics and combinatorial peptide library screening. I present two different peptide libraries against each type of drug-resistant bacteria and cancer cells. This shows that these peptides are powerful, with selectivity toward specific target cells, and their activity is comparable or even superior to commercial drugs. This demonstrates that simulation-guided design is a powerful strategy for the de novo development of membraneactive peptides and proteins, and this method has a broad spectrum of applications to address different biomedical needs.
Date of Award1 Jul 2020
Original languageEnglish
Awarding Institution
  • King's College London
SupervisorMartin Ulmschneider (Supervisor)

Cite this