Computational Studies of Glycan Conformations in Glycoproteins

N-glycans refer to oligosaccharide chains covalently attached to the side chain of asparagine (Asn) residues, and the majority of proteins synthesized in the endoplasmic reticulum (ER) are N-glycosylated. N-glycans can modulate the structural properties of proteins due to their close proximity to their parent proteins and their interactions between the glycan and the protein surface residues. In addition, N-glycans provide specific regions of recognition for cellular and molecular recognition. Despite their biological importance, the structural understanding of glycans and the impact of glycosylation to glycan or protein structure are lacking. I have explored the conformational freedom of glycans and their conformational preferences in different environments using structural databases and computer simulations. First, I have developed an algorithm to reliably annotate a given atomic structure of glycans. This algorithm is important because many glycan molecules in the crystal structure database are misannotated or contain errors. Using the algorithm, a database of glycans found in the PDB is constructed and available to the public. Second, the impact of glycosylation on the glycan conformation has been examined. Contrary to the common belief that the glycan conformations are independent to the protein structure, it appears that the protein structure can significantly affect the glycan structure upon glycosylation. This observation is significant because it may provide insight into protein-glycan interaction and opens up the possibility of a template-based glycan modeling approach. Third, the differences in conformational preference between glycans in solution and in glycoproteins has been examined. Using molecular dynamics (MD) simulations, the conformational preference of N-glycan pentassacharide in solution is exhaustively studied. Surprisingly, the conformational distribution is dominated by a single major conformational state and several minor conformational states. The dominant conformational state adopts a more extended conformation, thus it appears that entropy plays an important role in determining the conformational state. On the other hand, in glycoproteins, glycans can interact with surrounding protein side chains and, as a result, several conformational states are more equally populated. Based on these observations, a protocol is proposed for modeling the glycan portion of a known protein structure. It is typically more managable to acquire an atomic resolution structure or aglycoprotein (glycoprotein without glycan). In addition, the glycoform and the glycosylation site can be identified independently by mass spectrometry or NMR. The proposed modeling protocol assumes the glycosylation site, glycoform, and aglycoprotein structure are already known, and builds glycan structure models on top of the known aglycoprotein structure. The performance of the modeling protocol is greatly improved by using appropriate template structures. This protocol can be used to generate the initial model for MD simulations or refinement of low resolution models from experiments (small angle X-ray scattering and electron microscopy)