Adaptive Evolution of Peptide Inhibitors for Mutating SARS-CoV-2

Abstract

The SARS-CoV-2 virus is currently causing a worldwide pandemic with dramatic societal consequences for the humankind. In the past decades, disease outbreaks due to such zoonotic pathogens have appeared with an accelerated rate, which calls for an urgent development of adaptive (smart) therapeutics. Here, a computational strategy is developed to adaptively evolve peptides that could selectively inhibit mutating S protein receptor binding domains (RBDs) of different SARS-CoV-2 viral strains from binding to their human host receptor, angiotensin-converting enzyme 2 (ACE2). Starting from suitable peptide templates, based on selected ACE2 segments (natural RBD binder), the templates are gradually modified by random mutations, while retaining those mutations that maximize their RBD-binding free energies. In this adaptive evolution, atomistic molecular dynamics simulations of the template-RBD complexes are iteratively perturbed by the peptide mutations, which are retained under favorable Monte Carlo decisions. The computational search will provide libraries of optimized therapeutics capable of reducing the SARS-CoV-2 infection on a global scale.

Keywords: adaptive evolution; molecular dynamics; peptide libraries; therapeutic peptides.

Figure 1

Mutations of SARS‐CoV‐2. a) Time‐dependent mutation tree of SARS‐CoV‐2 colored according to geographical regions of origin (through June 2020).^{[ 21 ]} b) Twenty‐five single mutations identified in RBD of the S protein. c) Five amino acid mutations on RBD in contact with the ACE2 receptor. d) Binding free energies are evaluated with the MMGB‐SA method for the ACE2‐RBD complexes, including the originally reported RBD (wild type, labeled as WT) and the five mutant RBDs listed in panel (c).

Figure 2

Modeling of peptide‐RBD complexes. a) Complexes of S protein RBD (blue) and two peptide templates (red). Locations of five S protein mutations examined in the present work are marked by blue spheres. Amino acid residues changed in singly mutated peptides are marked by yellow spheres. b) Free energies of binding, $\mathrm{\Delta }{G}_{\text{MMGB-SA}}$, between the originally reported S protein RBD and the wild type or singly mutated ACE2‐based peptides. Locations of mutated peptide amino acids are highlighted in panel a. c) Snapshots of initial and optimized template‐1 peptides binding to the original RBD. The sequence of the optimized peptide was obtained after 100 mutation attempts, with 10 ns long MD simulation after each mutation. The final peptide with the optimized sequence was further relaxed in a 175 ns MD simulation. The initial peptide is shown as a red helix, with amino acids that are subsequently mutated shown in thin faded yellow licorice. The optimized peptide is shown as an orange helix, with mutated amino acids shown in thick yellow licorice. d) Adaptive evolution of template‐1. The plot shows the binding free energies, $\mathrm{\Delta }{G}_{\text{MMGB-SA}}$, between the peptide and the original RBD, presented as a function of the performed mutation, where the peptide:RBD complexes are relaxed for 10 ns after each mutation attempt. e) The time evolution of $\mathrm{\Delta }{G}_{\text{MMGB-SA}}$ between the final optimized peptide and the original RBD. The average value, obtained from the last 75 ns of the trajectory (gray), is $\mathrm{\Delta }{G}_{\text{MMGB-SA}}=-57$ kcal mol⁻¹. The faded green line shows the data points calculated every 0.1 ns, and the dark green line shows the running average. f) Initial and optimized sequences of template‐1 peptides. The final peptides were optimized for binding to the original and mutant RBDs, with peptide‐RBD complexes relaxed in 10 ns MD simulations after each attempted mutation.

Figure 3

a–c) Adaptive evolution of template‐1 coupled with singly‐mutated RBDs. d) Adaptive evolution of template‐2 coupled with the original RBD. After attempted mutations, peptide:RBD complexes were relaxed in 10 ns simulation steps.