Docking glycosaminoglycans to proteins: analysis of solvent inclusion

Abstract

Glycosaminoglycans (GAGs) are anionic polysaccharides, which participate in key processes in the extracellular matrix by interactions with protein targets. Due to their charged nature, accurate consideration of electrostatic and water-mediated interactions is indispensable for understanding GAGs binding properties. However, solvent is often overlooked in molecular recognition studies. Here we analyze the abundance of solvent in GAG-protein interfaces and investigate the challenges of adding explicit solvent in GAG-protein docking experiments. We observe PDB GAG-protein interfaces being significantly more hydrated than protein-protein interfaces. Furthermore, by applying molecular dynamics approaches we estimate that about half of GAG-protein interactions are water-mediated. With a dataset of eleven GAG-protein complexes we analyze how solvent inclusion affects Autodock 3, eHiTs, MOE and FlexX docking. We develop an approach to de novo place explicit solvent into the binding site prior to docking, which uses the GRID program to predict positions of waters and to locate possible areas of solvent displacement upon ligand binding. To investigate how solvent placement affects docking performance, we compare these results with those obtained by taking into account information about the solvent position in the crystal structure. In general, we observe that inclusion of solvent improves the results obtained with these methods. Our data show that Autodock 3 performs best, though it experiences difficulties to quantitatively reproduce experimental data on specificity of heparin/heparan sulfate disaccharides binding to IL-8. Our work highlights the current challenges of introducing solvent in protein-GAGs recognition studies, which is crucial for exploiting the full potential of these molecules for rational engineering.

Fig. 1

Dependence of water molecules abundance in GAG-protein interfaces on crystal structure resolution

Fig. 2

Time fractions of interactions per protein (dark grey) and per GAGs (light grey) residues for CD44—HA (a); Cathepsin K—CS (b) complexes

Fig. 3

Comparison of the reference docking experiments of Autodock 3, eHiTs, MOE docking and FlexX for 11 complexes: a RMSD of top scoring pose; b Lowest RMSD within 50 top poses; c Rank of the pose with the lowest RMSD in 50 top poses; d Number of correct poses in 50 top poses; e Number of correct poses in 10 top poses. ‘+Wat’ relates to the runs with explicit water molecules. ‘+ Wat3’ relates to the type 3 of FlexX water

Fig. 4

Results for the docking of Idu(2S)-GlcNAc(6S) to IL-8 with Autodock 3: 50 top docking solutions. The residues of heparin binding site are labeled and shown in licorice, the pyranose rings of disaccharides are in lines: red—Ido(2S) and green—GlcNAc(6S)