On the satisfaction of backbone-carbonyl lone pairs of electrons in protein structures

Abstract

Protein structures are stabilized by a variety of noncovalent interactions (NCIs), including the hydrophobic effect, hydrogen bonds, electrostatic forces and van der Waals' interactions. Our knowledge of the contributions of NCIs, and the interplay between them remains incomplete. This has implications for computational modeling of NCIs, and our ability to understand and predict protein structure, stability, and function. One consideration is the satisfaction of the full potential for NCIs made by backbone atoms. Most commonly, backbone-carbonyl oxygen atoms located within α-helices and β-sheets are depicted as making a single hydrogen bond. However, there are two lone pairs of electrons to be satisfied for each of these atoms. To explore this, we used operational geometric definitions to generate an inventory of NCIs for backbone-carbonyl oxygen atoms from a set of high-resolution protein structures and associated molecular-dynamics simulations in water. We included more-recently appreciated, but weaker NCIs in our analysis, such as n→π* interactions, Cα-H bonds and methyl-H bonds. The data demonstrate balanced, dynamic systems for all proteins, with most backbone-carbonyl oxygen atoms being satisfied by two NCIs most of the time. Combinations of NCIs made may correlate with secondary structure type, though in subtly different ways from traditional models of α- and β-structure. In addition, we find examples of under- and over-satisfied carbonyl-oxygen atoms, and we identify both sequence-dependent and sequence-independent secondary-structural motifs in which these reside. Our analysis provides a more-detailed understanding of these contributors to protein structure and stability, which will be of use in protein modeling, engineering and design.

Keywords: bioinformatics; hydrogen bonding; noncovalent interactions; n→π* interactions; protein folding; protein stability; protein structure.

Figure 1

Backbone‐carbonyl‐oxygen non‐covalent interaction (NCI_C=O) considered in this analysis. (A) “Standard” hydrogen bonds, as exemplified by NH_i→C=O_i−4 hydrogen bonds found in an α‐helix (NH_bb, dNH⋯O ≤ 2.44 Å25; ω ≥ 90°; ρ ≥ 90°). Other donor groups include (i) side‐chain NH, e.g., from lysine or arginine, (NH_sc, parameters as for NH_bb); and (ii), side‐chain hydroxyl groups (OH_sc, d(O)H⋯O ≤ 2.31 Å25; ω ≥ 90°; ρ ≥ 90°). (B) Hydrogen bonds with a Cα—H group donor (CαH, d(Cα)H⋯O ≤ 2.68 Å25; ω ≥ 90°; ρ ≥ 90°, elevation angle < 50°), or alternatively donated by other methyl or ethyl groups from protein sidechains9 (CH_X, parameters as for CαH). (C) n→π* interactions, shown with a main‐chain carbonyl group acceptor (dC⋯O ≤ 3.22 Å; 95° ≥ θ ≥ 125°; Cα⋯C⋯O⋯H dihedral χ ≥ 120°57); but these can also have a side‐chain acceptor, e.g., asparagine or glutamine, (n→π* _sc, parameters as for n→π*). (D) Hydrogen bonds made with water (HOH, d(O)H⋯O ≤ 2.31 Å25; ω ≥ 90°; ρ ≥ 90°).

Figure 2

The percentages of NCI_C=Os per residue made across all residues in proteins. These were measured in three ways: across all residues in the initial, unsolvated high‐resolution crystal structures (black bars); across all residues and snapshots from the last 81ns of a molecular‐dynamics simulation (gray bars); from the distribution of modal averages of all residues across the same set of molecular‐dynamics simulation snapshots (diagonal bars); across all residues and snapshots for those residues that spend at least half of their molecular‐dynamics simulation at their modal average number of NCI_C=O (white bars).

Figure 3

Distributions of types of NCI_C=O made in different secondary structure. (A) Where 2 × NCI_C=O are made per residue; (B) 1 × NCI_C=O; and (C) 3 × NCI_C=O. For clarity, only those combinations of NCI representing at least 2% of all residues are shown in (A), which accounts for 89% of residues overall. Key for panel (A): black bars, 2 × HOH; red, 1 × n→π* plus 1 × NH_bb; orange, 1 × CαH plus 1 × NH_bb; yellow, 1 × n→π* plus 1 × HOH; green, 1 × NH_bb plus 1 × CH_X; turquoise, 1 × HOH plus 1 × CH_X; dark blue, 1 × NH_bb plus 1 × HOH; purple, 1 × NH_sc plus 1 × HOH. (B and C) Residues were included in the plots for panels (B) and (C) if their modal average number of NCI_C=O was 1 or 3, and spent at least 50% of the duration of MD‐simulation in these categories. Key for panel (B): red bars, 1 × NH_bb; orange, 1 × NH_sc; yellow, 1 × n→π*; green, 1 × CαH; turquoise, 1 × OH_sc, blue, 1 × CH_X. Key for panel (C): red bars, 1 v NH_bb, 1 × n→π*, 1 × CH_X; orange, 2 × NH_bb plus 1 × n→π*; yellow, 1 × NH_bb, 1 × n→π*, 1 × HOH; green, 1 × n→π*, 1 × OH_sc, 1 × NH_bb; turquoise, 1 × CαH, 1 × HOH, 1 × NH_bb; blue, 1 × n→π*, 1 × CαH, 1 × NH_bb; purple, 1 × NH_bb, 1 × NH_sc, 1 × n→π*; gray, 1 × NH_bb, 1 × CαH, 1 × CH_X; white, 2 × NH_bb, 1 × CH_X; mint green, 1 × NH_bb, 1 × NH_sc, 1 × CαH. (D, E) The most‐common NCI_C=O combinations identified in the two most‐prevalent secondary structure types. (D) β‐Strand residues with a backbone NH hydrogen bond (NH_bb, •) plus a Cα–H hydrogen bond (Cα–H, ★), (PDB 1G66, residues A6, A84‐A85). (E) α‐Helical residues residues with a NH_bb (•) plus an n→π* interaction (), (PDB 1G66, residues A26‐A30). Secondary structures were assigned by Promotif,42 which uses a modified version of the Kabsch and Sander DSSP algorithm.58 Categories “E” and “B” were combined into a single β‐structure category.

Figure 4

Local structures with over‐satisfied backbone‐carbonyl‐oxygen atoms, i.e., with 3 × NCI_C=O. (A) α‐helical motifs with 1 × NH_bb (•), 1 × n→π* interaction () and one 1 × CH_X (▪). The residue providing the CHx has been truncated for clarity. (B) Motifs at the α‐helical Ntermini with 2 × NHbb (•) plus 1 × n→π* interaction (). (C) α‐helical C‐termini with 1 × OH_sc, (▲), 1 × NH_bb (•) and 1 × n→π* interaction (), and associated WebLogos⁵⁹ indicating the amino‐acid frequencies from sequences in our dataset that display this motif. Structural images prepared with PyMOL (http://www.pymol.org). PDB codes and residue identifiers for each example can be found in the Supporting Information.