Lasso peptides, a unique class of natural products with interlocked conformations, offer exceptional stability and biomedical potential. However, their folding mechanisms and the role of cyclase enzymes in guiding lasso formation remain poorly understood. In this study, we employed molecular dynamics (MD) simulations integrated with deep learning (DL) approaches to elucidate the mechanisms behind lasso peptide folding and cyclase binding and engineer both peptide and cyclase.

Extensive MD simulations of 20 structurally characterized lasso peptides revealed universally low pre-folded populations and high folding free energy barriers. To overcome sampling limitations, we applied the transition-based reweighting analysis method (TRAM), combining unbiased and biased simulations to accurately capture conformational changes, thermodynamics, and kinetics. A DL-based Variational Autoencoder (VAE) clustered numerous kinetic folding pathways into distinct metastable routes, enabling identification of folding pathways in solution. Secondary structure formation emerged as a critical determinant: simulated β-sheet content in transition-state ensembles, identified by the TS-DAR neural framework, strongly correlated with experimentally validated peptide production via cell-free biosynthesis. This highlights secondary structure and loop stability are critical folding determinants.

Given the unfavorable nature of spontaneous folding in solution, we further explored cyclase-assisted lasso peptide folding. MD simulations of fusilassin bound to cyclase revealed essential contacts between loop residues and the cyclase. Engineering studies rescued previously incompatible peptide with cyclase variants, highlighting the importance of these interactions for specificity. To resolve the cyclase role, we examined two hypotheses: folding-upon-binding versus folding-before-binding. By combining structure prediction, MD simulations, and DL-based Markov State Models, our investigation supported the folding-upon-binding, showing peptide folding occurs after binding, and cyclase pocket functions as a chaperone mediating peptide formation. Finally, we leveraged DL methods such as MPNN to advance both peptide and cyclase engineering.

This work elucidates universal folding features and confirms a folding-upon-binding mechanism for lasso peptides, establishing a framework for rational peptide engineering and cyclase design, ultimately advancing development of lasso peptide therapeutics.

Diwakar Shukla

University of Illinois Urbana-Champaign College of Liberal Arts & Sciences

Appearances