New atomic resolution insights into dynamic protein-carbohydrate interactions enabled by high-performance computing

Abstract

Protein-carbohydrate interactions are a critical component of cellular structure and function. However, the inherent flexibility of biological carbohydrate polymers and the microheterogeneity resulting from their non-template-based biosynthesis complicate their study using experimental methods. Therefore, simulation approaches, and all-atom explicit-solvent molecular dynamics (MD) simulations in particular, provide an enabling technology for advancing the understanding of protein-carbohydrate interactions at the atomic level of resolution. Here, we detail our recent MD studies on the CD44 receptor performed using the highly-parallel NAMD MD engine with the CHARMM all-atom force field on the Kraken supercomputer. With these technologies, simulation time lengths of hundreds of nanoseconds are routinely reached. Combining both regular unbiased MD and advanced MD methods that bias sampling to important degrees of freedom, new insights are obtained into the function of CD44, which is both a receptor for large carbohydrate molecules in the extracellular matrix and whose own function is modulated by covalent attachment of branched carbohydrates to make CD44 a glycoprotein. In particular, the simulations explain, for the first time, the molecular mechanism of the experimentally-observed order-to-disorder transition in CD44 that is known to enhance its carbohydrate binding affinity. Additionally, the simulations explain, again for the first time, the molecular mechanism through which particular monosaccharides in branched carbohydrates covalently attached to CD44 have been experimentally observed to block CD44 binding to extracellular matrix carbohydrates. These insights expand the understanding of how CD44 performs its biological functions, which include cell adhesion, migration, and vascular trafficking. Importantly, these new insights enabled by leading-edge simulation and computing technologies have not been accessible by existing experimental methods.