Computational Design of Peptides and Proteins for Cyclic Peptide Applications

Abstract

Cyclic peptides are a fascinating class of compounds to use in multiple design applications due to their small size and ability to be chemically synthesized. Non-canonical amino acids can be easily incorporated into cyclic peptides to expand their chemistries and make them resistant to proteolysis. Additionally, cyclic peptides are modular and can be computationally designed to adopt specific membrane permeable conformations. During my PhD, I used computationally designed cyclic peptides for two main applications, 1) designing peptide metal-organic frameworks, and 2) designing peptide inducible dimeric proteins. First, I explored the use of designed symmetric cyclic peptides to generate new metal-organic frameworks (MOFs). MOFs have shown great potential in a wide variety of applications such as small molecule separation, drug delivery, and catalysis. However, these materials have generally been limited to using small molecule linkers and short two to four residue peptides. I aimed to computationally design peptide MOFs using symmetric cyclic peptides. Peptide based MOFs provide advantages since they can bind specifically and selectively to different substrates and can be used for catalysis. We used the Rosetta Software Suit to dock and design peptides into metal mediated 3D lattices, then experimentally screened them for crystal formation. We solved the structures of six peptide materials using single crystal X-ray diffraction. Although these structures do not match the design models, they demonstrate fundamental thermodynamic and kinetic rules that govern the formation of such materials. Our computational pipeline is the first step to computationally design peptide based metal-organic frameworks and our experimental data provides information to further refine the computational protocol for more accurate modeling. Second, to gain insight into the effects of protein oligomerization on cellular functions, I employed cyclic peptides to induce the formation of protein oligomers. Most chemically inducible dimeric systems in the literature use clinically approved drugs to induce protein oligomerization. However, these drugs have off-target effects which makes the results ambiguous. In order to make CID systems that are orthogonal to cellular machinery, I used computationally designed de novo proteins and de novo cyclic peptides to make CIDs. I used Rosetta, Alphafold2, and ProteinMPNN to design and filter peptide binding proteins. The binding was validated using isothermal calorimetry, and equilibrium dialysis leading to the identification of multiple nanomolar and micromolar peptide binders. Finally, we can use the NanoBiT assay to optimize peptide-dependent protein oligomerization.