Devof a elopment new simulation-guided approach to design antimicrobial peptides with high selectivity towards individual bacterial species

The goal of this project is to develop new molecular dynamics (MD) simulation based peptide-design methods and apply them to design and optimize novel antimicrobial peptides (AMPs) that target individual bacterial species with high selectivity. Developing new antibacterials to tackle rapidly rising antibiotic resistance is one of the most pressing and critical unmet health care needs. This is reflected by the UK's Antimicrobial Resistance Strategy, which is part of a BBSRC focus area. AMPs are extremely promising pharmacophores, but their chemical diversity, flexible nature, and prohibitive number of possible configurations, combined with the lack of suitable design and optimization tools has hindered their translation into the clinic. My group and I have been working on developing quantitatively accurate experimentally validated molecular simulation methods to design membrane-active peptide for over 15 years. This proposal builds on this expertise and will deliver new experimentally validated physical simulation methods for rational design and optimization of new AMPs.

Current antibiotics are increasingly failing due to the emergence of drug-resistant bacterial strains. At the same time the pharmaceutical industry is retreating from the development of new antibiotics. This is because the extremely high cost of developing new antibiotics cannot be recuperated, as any new drugs will need to be held in reserve to treat drug-resistant infections. Hence there is not only a critical unmet need to develop new antibiotic therapies, but also a need for methods that bring down the cost of antibacterial pharmacophore design and validation. AMPs present one of the most promising classes of antimicrobial pharmacophores. They are more resilient than antibiotics against antimicrobial resistance formation, easy to produce and modify, and offer a near-infinite chemical and structural reservoir, which remains largely untapped. In our preliminary work (patent pending), we show that AMPs can be designed to provide a key feature desirable for next-generation antimicrobials: precise and exclusive targeting of specific pathogenic microbes without harming symbiotic bacteria or human tissues.

Here we will build on this work to develop new simulation-guided methods to design AMPs that target specific bacterial species with high selectivity. First we apply unbiased atomic detail peptide partitioning MD simulations to design sequences that bind efficiently to specific membrane models of bacteria and human cells. Next we will combine genetic algorithms and molecular assembly simulations to introduce pore-forming mutations into these membrane-targeting sequences. These simulation approaches, developed by us, utilize sophisticated algorithms and methodologies that will allow us to iteratively simulate tens of thousands of sequence mutations to actively select for peptides that partition and form pores in specific membranes with high selectivity.

These designed functional sequences will then serve as templates for experimental optimization using a combinatorial peptide library approach. This combined simulation-guided design and experimental optimization approach is equivalent to screening millions of peptide sequences, in a fraction of the cost and time. We will validate the functional characteristics and targeted antibacterial activity of the designed AMPs by in vitro screening against live bacteria and human cell lines.

The methods developed here will not only provide urgently needed design tools to realize the potential of AMPs as pharmacophores, but also fundamentally advance our understanding of how specific peptide sequences target bilayers with particular lipid compositions, and shed new light on the molecular mechanisms driving the formation of peptide pores in cellular membranes.

The four main objectives and methods of the proposed project are:

Objective 1 involves development of algorithms to automatically analyse simulations and propose and rank mutations. Algorithms to automatically detect peptide folding and partitioning into bilayers will be based on atomic coordinates and equilibrium population based free energy calculations. Protein assemblies will be detected by modifying cluster algorithms to include topological and geometric information. Genetic and artificial intelligence algorithms will be used to propose and rank membrane binding-inducing and membrane pore-stabilising mutations.

Objective 2 involves the simulation guided-design of functional template sequences for experimental optimisation. We will first construct atomic detail lipid bilayer models and then study the partitioning of peptides into these bilayers using unbiased equilibrium folding-partitioning simulations. Atomic detail pore formation will be simulated using a generalized Born implicit membrane model that treats the peptide environment as a continuum.

Objective 3 involves experimental synthesis and optimisation of the simulation-designed template sequences. Peptide libraries based on the template designs will be solid-phase synthesized using a resin split-and-pool approach that results in a unique peptide sequence per resin bead. Peptides will be screened for binding and pore formation using tryptophan and dye-leakage fluorescence assays.

Objective 4 involves the assessment of peptide activity in vitro and simulation based mechanistic characterisation. Peptides will be titrated onto live cells to assess their antibacterial activity and toxicity towards human cells using standard in vitro assays. The mechanisms of pore formation will be determined using long-timescale unbiased molecular dynamics simulation using a method recently developed by us.

Academia and not-for-profit research
In addition to the work of our collaborators, our method development will directly impact research in the field of membrane-active peptides and proteins. The focus on automation and modelling of fundamental interactions driving pore assembly will solidify the foundation of our research, as well as that of many others. A large number of research groups all over the world work on antimicrobial peptides (AMPs) and other membrane-active peptides. Our results and software will be presented at international meetings, seminars, workshops, and in peer-reviewed publications. Our algorithms and methods will be most relevant to researchers investigating antimicrobial, membrane-active, and pore-forming peptides, as well as membrane protein folding, via either computational or experimental approaches. All method and algorithm developments, as well as experimental calibration datasets will be made freely available, including annotated source code, via our website at KCL. We will also provide detailed instructions of how to apply these methods in our publications and talks, as well as provide sample data and analyses.

Industry and Business
This project will provide an important set of design methods that will facilitate and guide the development of next-generation precision antibacterials. In addition, the methods developed here will allow detailed in silico characterisation of putative peptide sequences with their native lipid environment. Despite the enormous potential of biologicals such as peptides, proteins, and antibodies as drugs, delivery vehicles, and biomarkers, their translation into the clinic is impeded by a poor understanding of their molecular mechanisms, and a lack of experimentally validated predictive modelling tools and characterisation technology. Our project proposes to address all of these needs, focusing on peptides for the moment, but with a vision to extend these methods to proteins, antibodies, nanoparticles, and other biologicals in the future. Our tools and methods will enable studying the activity of membrane-active peptides with diverse biomedical and industrial applications, such as AMP design (as proposed here), studying the mechanisms of bacterial toxins, studying viral fusion peptides, etc..

General public
We will give talks and classes at schools (such as the ones I have previously given at the American Community School in Egham), and we will use the King's College open days to showcase our research in a generally understandably manner. Being able to better visualise and clearly describe nanoscale biological processes will provide a much clearer mental image of these processes, which we hope will aid educational goals by leading to improved interest and understanding of the machinery of life. In addition, we will provide training and practical research experiences at all levels. My lab has previously hosted two high-school students for 6 weeks each. The pupils were trained in designing peptides via molecular modelling and simulation, and then proceeded to synthesize and characterise these designed peptides experimentally. One of these students has now moved on to study at Princeton University with the long-term goal to becoming a researcher. In addition, my lab has hosted 4 undergraduates doing research projects, 3 of them are now pursing graduate studies.