Molecular modeling of protein-glycosaminoglycan interactions

Forty-nine regions in 21 proteins were identified as potential heparin-binding sites based on the sequence organizations of their basic and nonbasic residues. Twelve known heparin-binding sequences in vitronectin, apolipoproteins E and B-100, and platelet factor 4 were used to formulate two search strings for identifying potential heparin-binding regions in other proteins. Consensus sequences for glycosaminoglycan recognition were determined as [-X-B-B-X-B-X-] and [-X-B-B-B-X-X-B-X-] where B is the probability of a basic residue and X is a hydropathic residue. Predictions were then made as to the heparin-binding domains in endothelial cell growth factor, purpurin, and antithrombin-III. Many of the natural sequences conforming to these consensus motifs show prominent amphipathic periodicities having both alpha-helical and beta-strand conformations as determined by predictive algorithms and circular dichroism studies. The heparin-binding domain of vitronectin was modeled and formed a hydrophilic pocket that wrapped around and folded over a heparin octasaccharide, yielding a complementary structure. We suggest that these consensus sequence elements form potential nucleation sites for the recognition of polyanions in proteins and may provide a useful guide in identifying heparin-binding regions in other proteins. The possible relevance of protein-glycosaminoglycans interactions in atherosclerosis is discussed.