Pyranose Ring Puckering Thermodynamics for Glycan Monosaccharides Associated with Vertebrate Proteins

Abstract

The conformational properties of carbohydrates can contribute to protein structure directly through covalent conjugation in the cases of glycoproteins and proteoglycans and indirectly in the case of transmembrane proteins embedded in glycolipid-containing bilayers. However, there continue to be significant challenges associated with experimental structural biology of such carbohydrate-containing systems. All-atom explicit-solvent molecular dynamics simulations provide a direct atomic resolution view of biomolecular dynamics and thermodynamics, but the accuracy of the results depends on the quality of the force field parametrization used in the simulations. A key determinant of the conformational properties of carbohydrates is ring puckering. Here, we applied extended system adaptive biasing force (eABF) all-atom explicit-solvent molecular dynamics simulations to characterize the ring puckering thermodynamics of the ten common pyranose monosaccharides found in vertebrate biology (as represented by the CHARMM carbohydrate force field). The results, along with those for idose, demonstrate that the CHARMM force field reliably models ring puckering across this diverse set of molecules, including accurately capturing the subtle balance between ⁴C₁ and ¹C₄ chair conformations in the cases of iduronate and of idose. This suggests the broad applicability of the force field for accurate modeling of carbohydrate-containing vertebrate biomolecules such as glycoproteins, proteoglycans, and glycolipids.

Keywords: GalNAc; GlcNAc; Neu5Ac; fucose; galactose; glucose; glucuronate; iduronate; mannose; tetrahydropyran; xylose.

Figure A1

ΔG(α₁, α₂) from eABF simulations using incorrect versus correct dihedral force field parameters for Neu5Ac along with (α₁, α₂) data from all Neu5Ac structures in the PDB (searched 30 July 2021). ΔG(α₁, α₂) contour data are for αNeu5Ac with incorrect parameters (a), αNeu5Ac with correct parameters (b), MeαNeu5Ac with incorrect parameters (c), MeαNeu5Ac with correct parameters (d), βNeu5Ac with incorrect parameters (e), βNeu5Ac with correct parameters (f), MeβNeu5Ac with incorrect parameters (g), and MeβNeu5Ac with correct parameters (h). α₁ and α₂ are in degrees. ΔG(α₁, α₂) is from the first simulation in the triplicate and is in kcal/mol, with contours drawn every 1 kcal/mol and colored from 0–3 kcal/mol. PDB data were divided into two groups: those from α anomers and those from β anomers. Crystallographic data from the α anomers are displayed as small +’s in (a–d) and crystallographic data from the β anomers are displayed as small +’s in (e–h).

Figure 1

Compounds considered in the current study. Glc carbon atoms are numbered in blue. All other monosaccharides follow the same numbering scheme, except for Neu5Ac, which is numbered as pictured. All monosaccharides are drawn as the β anomer. The α anomer is created by inversion of the configuration at carbon 2 for Neu5Ac and at carbon 1 for all other monosaccharides. Both anomers for each monosaccharide as well as the corresponding O-methyl glycosides, formed by methylation at the anomeric carbon hydroxyl, were studied, for a total of 45 compounds (44 monosaccharides + THP).

Figure 2

Pyranose ring puckering (red) and glycosidic bond rotation (blue) are the major sources of polymer flexibility in vertebrate glycans.

Figure 3

MeαIdoA ΔG(α₁, α₂) from eABF simulation. Each panel is from a separate 200-ns simulation seeded with different initial random velocities. α₁ and α₂ are in degrees. ΔG(α₁, α₂) is in kcal/mol, with contours drawn every 1 kcal/mol, colored from 0–3 kcal/mol, and labeled every 2 kcal/mol.

Figure 4

Sampling of specific MeαIdoA ring puckering conformations during eABF simulation with the Babin-Sagui (α₁, α₂) reaction coordinate. Sampled (α₁, α₂) values are separated into those for ⁴C₁, ¹C₄, and ²S_O (blue, red, and green dots, respectively, in panel “a”) and for all other (black dots, panel “b”) puckering conformations. α₁ and α₂ are in degrees. ΔG(α₁, α₂) is in kcal/mol, with contours drawn every 1 kcal/mol from 0–5 kcal/mol and colored from 0–3 kcal/mol. Puckering data have been aggregated across the triplicate simulations, and ΔG(α₁, α₂) is from the first simulation in the triplicate.

Figure 5

MeαIdoA conformational transitions in standard (non-biased) (a), eABF (b), and CMAP-biased (c) molecular dynamics simulations. eABF and CMAP biased simulations have biasing on the Babin-Sagui (α₁, α₂) reaction coordinate. Data in each panel are from triplicate simulations (blue, red, and green) seeded with different random initial velocities.

Figure 6

MeαIdoA Cremer-Pople (θ, 𝜙) values sampled during eABF (a) and CMAP-biased (b) molecular dynamics simulations. Pyranose ring puckering regions [56] (“⁴C₁”, “northern tropical”, “²S_O”, etc.) are labeled as defined in the Materials and Methods section. Biasing was applied to the Babin-Sagui (α₁, α₂) reaction coordinate. Data in each panel are from triplicate simulations (blue, red, and green) seeded with different random initial velocities. eABF simulations were 200 ns and CMAP-biased simulations were 1000 ns.

Figure 7

Comparison of ΔG values for the ⁴C₁ to ¹C₄ equilibrium in Ido and IdoA compounds from eABF simulations, CMAP-biased simulations, and NMR experiments. Data are presented as eABF vs. NMR (a), CMAP-biased vs. NMR (b), and eABF vs. CMAP-biased (c). The specific compounds and the experimental data from NMR experiments are as detailed in Table 1. Simulation data points are averages from triplicate simulations, with error bars representing 95% confidence intervals. The solid diagonal is the line y = x, and the dotted diagonal lines are ±0.5 kcal/mol.