Enumerative Discovery of Noncanonical Polypeptide Secondary Structures

Abstract

Energetically favorable local interactions can overcome the entropic cost of chain ordering and cause otherwise flexible polymers to adopt regularly repeating backbone conformations. A prominent example is the α helix present in many protein structures, which is stabilized by i, i + 4 hydrogen bonds between backbone peptide units. With the increased chemical diversity offered by unnatural amino acids and backbones, it has been possible to identify regularly repeating structures not present in proteins, but to date, there has been no systematic approach for identifying new polymers likely to have such structures despite their considerable potential for molecular engineering. Here we describe a systematic approach to search through dipeptide combinations of 130 chemically diverse amino acids to identify those predicted to populate unique low-energy states. We characterize ten newly identified dipeptide repeating structures using circular dichroism spectroscopy and comparison with calculated spectra. NMR and X-ray crystallographic structures of two of these dipeptide-repeat polymers are similar to the computational models. Our approach is readily generalizable to identify low-energy repeating structures for a wide variety of polymers, and our ordered dipeptide repeats provide new building blocks for molecular engineering.

Figure 1

Computational pipeline for novel secondary structure discovery. (A) Process flow diagram for computational discovery of (AB)_n secondary structures from noncanonical amino acids. (B) Representative examples of noncanonical amino acids considered in this study. The full list of residues that were considered in this study is in Table S1. (C) Example potential energy surface of the monomeric l-Alanine residue sampled with the VABLAS protocol. (D) Schematic representation of (AB)_n repeating peptide and the chemical structure of a representative example. (E) Example of a computational conformational ensemble shown as the delta energy to the lowest energy state versus the RMSD with respect to the backbone atoms of the lowest energy state. The free energy gap between the lowest energy state and the next lowest energy state is highlighted with a black bar, and the 100 lowest energy conformers from the ensemble are enclosed in a green box and shown in (F) aligned by their backbone atoms. The repeat unit is highlighted in color.

Figure 2

Designed (AB)₄ secondary structures. (A–F) Chemical structure of the DPR unit, conformational ensemble, and predicted low energy structure for six computationally predicted examples of (AB)₄ secondary structures, DPR1–DPR6. Nonpolar hydrogens were omitted for clarity, and hydrogen bonds are depicted as dashed lines. Some side chains were excluded for clarity from linear schematic. Backbone dihedral angles are rounded to the nearest 5°.

Figure 3

Experimental and calculated CD spectra. Low energy predicted ensemble (with nonpolar hydrogens excluded for clarity), experimental CD spectra (solid lines) at 5 °C in 2:1 ACN/H₂O (A–C) and 100% ACN (D), and average theoretical spectra (dotted lines), with low energy spectra in blue and high energy spectra in red for (A) DPR1, (B) DPR2, (C) DPR3, and (D) DPR4. (E) Near UV CD spectrum of DPR1 1× (red), 3× (green), 5× (blue) at 5 °C (solid lines) and 75 °C (dotted lines). (F) Far UV spectra of 10× DPR2 at temperatures from 5 to 75 °C at 10 °C intervals (colored blue to red), showing increased unfolding at high temperatures, together with spectra of largely disordered 1× (gray dotted lines) and 3× (black dotted line) at 5 °C.

Figure 4

3D structure determination. (A) DPR1 dipeptide representation with l-Tyr and d-Pip. (B) The 0.9 Å crystal structure of 3× DPR1 (purple) overlaid with the design model (green, Cα RMSD of 0.46 Å). (C) Two symmetry-related helices from the crystal lattice in an extended head-to-tail arrangement resulting from the screw axis of symmetry (top) and rotated 90° to show the flat geometry of this secondary structure. (D) Schematic of DPR1 with repeating β-turns. (E) Design model of 4× DPR1 in stick representation with selected hydrogens shown. Key NOEs (i, i + 1 and i, i + 3) observed in the DPR1 5× 2D ¹H–¹H NOESY data at 10 °C shown as dotted yellow lines. Tyr OHs are excluded for clarity in (B,E). (F) Single dipeptide segment of the model structure shown as sticks (top) and as space filling models (bottom). The polar CH−aromatic interaction identified by the ring current shifts in NMR is shown as a solid black line. (G) Temperature and length-dependence of Pip Hδ2 chemical shifts in DPR1, color-coded by polymer length as in Figure 3E (i.e., 1× – red, 3× – green, 5× – blue). In the 5× ¹H–¹³C HSQC NMR spectrum (Figure S13), some Pip Cδ−Hδ2 cross peaks overlap and therefore the proton counts for each peak are given in parentheses (Figure S15B). The percent shifted (y-axis) was calculated for each resolved proton by subtracting the 1× shift and dividing by the maximum expected ring current shift of 2.0 ppm. (H–K) are the corresponding data for DPR2. (I) NOEs of 10× DPR2 at 10 °C. (K) Temperature and length-dependence of AMPA Hα2 and Hα3 shifts in DPR2.