Comprehensive computational design of ordered peptide macrocycles

Abstract

Mixed-chirality peptide macrocycles such as cyclosporine are among the most potent therapeutics identified to date, but there is currently no way to systematically search the structural space spanned by such compounds. Natural proteins do not provide a useful guide: Peptide macrocycles lack regular secondary structures and hydrophobic cores, and can contain local structures not accessible with l-amino acids. Here, we enumerate the stable structures that can be adopted by macrocyclic peptides composed of l- and d-amino acids by near-exhaustive backbone sampling followed by sequence design and energy landscape calculations. We identify more than 200 designs predicted to fold into single stable structures, many times more than the number of currently available unbound peptide macrocycle structures. Nuclear magnetic resonance structures of 9 of 12 designed 7- to 10-residue macrocycles, and three 11- to 14-residue bicyclic designs, are close to the computational models. Our results provide a nearly complete coverage of the rich space of structures possible for short peptide macrocycles and vastly increase the available starting scaffolds for both rational drug design and library selection methods.

Fig. 1. Expanding the repertoire of ordered macrocycles by comprehensive sampling

(A) Torsion bin definitions (bin B = blue; A = red; Y = mirror of bin B, cyan; X = mirror of bin A, orange) overlaid on the generated nine-residue macrocycle backbone torsion angle distributions. The inversion symmetry is evident. (B) Convergence test. The number of clusters (torsion bin strings) observed for which the mirror image is not observed (yellow) or for which the mirror image is observed (cyan) was determined both for the complete set of 7,500,000 samples for nine-residue macrocycles and for randomly selected subsets of different sizes (x axis). For comparison, the less converged results obtained starting from a 20,000-sample subset and then down-sampling are shown in the inset: The fraction of single-chirality clusters is higher, and the number of both-chirality clusters does not plateau. (C) Energy landscape analysis using large-scale genKIC backbone generation followed by all-atom energy minimization. Each point is the result of an independent calculation; the energy is shown on the y axis and the root mean square deviation (RMSD) from the design model on the x axis. The extent of funneling (or convergence) of the energy landscape is quantified through the energy gap between conformations generated by stochastic sampling around the design model (red points) and high-RMSD structures, and the Boltzmann weight of the near–design model population. (D) Number of designed macrocycle clusters with distinct torsion bin strings and backbone hydrogen bond patterns with energy gaps less than −0.1 and Boltzmann weights greater than 0.85; the number of unbound peptide macrocycles in the PDB and CSD, consisting of only 20 canonical amino acids and their D-enantiomers, is indicated in pink.

Fig. 2. Recurrent local structural motifs in designed macrocycles

Top three rows: frequencies in native proteins in the PDB (red) and in the designed macrocycles (blue) indicated by vertical bars (including both enantiomers); structures on the right are colored by torsion bin as in Fig. 1A. Bottom row: example of structure propagation with residue insertion for a 7- to 10-residue macrocycle series.

Fig. 3. Seven- to eight-residue macrocycle NMR structures are very close to design models

Columns A: Design model. B: Amino acid sequence, torsion bin string, hydrogen bond pattern, and building block composition. C: Observed backbone-backbone (green), backbone–side-chain (purple), and side-chain–side-chain (orange) NOEs. D: Overlay of design model (green) on MD-refined NMR ensemble (gray; the average backbone RMSD to the NMR ensemble is indicated). E: Average decrease in the propensity to favor the designed state (P_Near, see methods) over all mutations at each position. Darker gray indicates larger decreases (P_Near values for each substitution at each position are in fig. S16); positions particularly sensitive to mutation are boxed and indicated by color in the design model in column A. F: Representative energy funnels for mutations at key positions (colored points) as compared to the design sequence (gray points). Row I, column G: Experimental SLIM data. Distribution of peak width at half height for peptide libraries with all amino substitutions at positions 4 and 5; the position 4 library has a broader distribution consistent with the computed energy landscape in column F. Rows II, IV, V, column G: Representative energy landscapes for double substitutions (red) of critical residues overlaid on the original design landscape (gray). Row III, column G: Overlay of design model on alternative structure NMR ensemble (turn flip at bottom right).

Fig. 4. Nine- to 14-residue macrocycle NMR structures are very close to design models

Rows I to III: 9- and 10-residue designs. Columns A to G are as in Fig. 3 rows II, IV, and V. Row IV: Comparison of bicyclic design models and NMR structures.