SMP103: a de novo-engineered macrocyclic peptide binder candidate targeting the virulence-associated transpeptidase Sortase A

Abstract

The emergence of antimicrobial resistance has intensified interest in anti-virulence strategies that target bacterial pathogenicity rather than viability. Sortase A (SrtA), a transpeptidase responsible for anchoring virulence-associated surface proteins to the cell wall of Gram-positive pathogens, represents an attractive therapeutic target. In this study, we employed an integrated artificial intelligence-assisted protein engineering workflow to design de novo macrocyclic peptide binders targeting the catalytic pocket of Staphylococcus aureus SrtA. A substrate-bound SrtA model was first generated using Boltz-2 to delineate the substrate-recognition region, and the mapped residues were subsequently used to guide binding-site-conditioned peptide design with BoltzGen against the experimentally determined active SrtA structure. A library of 200 macrocyclic peptide-protein complexes was generated and prioritized through confidence-guided structural and orthogonal fold assessment using HighFold. Three prioritized candidates were further characterized using electrostatic analyses, replicated 2-μs all-atom molecular dynamics simulations, and conformational analyses, after which SMP103 was subjected to replicated steered molecular dynamics and umbrella-sampling calculations. Among the evaluated candidates, SMP103 exhibited persistent occupancy of the targeted SrtA groove, comparatively confined conformational sampling, and reproducible residue-level engagement across three independent 2-μs simulations. Replicated enhanced-sampling calculations further characterized its dissociation energetics, with the averaged umbrella-sampling PMF exhibiting a maximum free-energy rise of 15.54 kcal mol⁻¹ and a bootstrap uncertainty of ±0.37 kcal mol⁻¹. Collectively, these findings identify SMP103 as a computationally predicted de novo -engineered macrocyclic peptide binder candidate and demonstrate the utility of integrating generative artificial intelligence with multiscale molecular simulations for binder discovery targeting SrtA.