Computational Design of a Model for the Platelet Integrin αIIbβ3.

Integrins, the principal receptors linking the cytoskeleton to extracellular ligands, are responsible, in large part, for platelet adhesion and aggregation. Like many membrane spanning proteins, the molecular structure of integrins has been difficult to determine, although crystal and NMR structures are available for water soluble portions of αVβ3 and αIIbβ3. In addition, models for the αIIb and β3 transmembrane (TM) heterodimer have been obtained using computational methods, but with conflicting results. Accordingly, we employed two different modeling strategies to distinguish between existing models. First, we used a Monte Carlo-simulated annealing (MCSA) protocol that features a selective advantage for structures that are consistent with previously characterized αIIb and β3 TM domain point mutations. During each step of the protocol, we calculated the effects of selected mutations and assigned a penalty to the step if it produced a structure that was inconsistent with experimental results. Second, we threaded the αIIb and β3 TM sequences onto a database of membrane helix dimers and rank-ordered the structures based on their calculated energies and their agreement with the cysteine crosslinking experiments reported by Luo et al (PLoS Biol 2:776–86, 2004). The MCSA and threaded TM models were similar with a root mean squared deviation (RMSD) of 1.5 Å. Moreover, they differed from a model reported by Kay Gottschalk (Structure 13:703–12, 2005) most notably by an ~50° rotation of the αIIb TM helix about its helical axis. This positions αIIb residue Leu980 in the TM interface and in van der Waals contact with β3 residue Gly708, a result consistent with mutagenesis experiments. Further, the αIIb and β3 TM helices pack along conserved glycine residues (Gly972 and Gly976) in the GXXXG motif of αIIb with a right-handed crossing angle of −18°. To create a model for the full-length integrin, the N-termini of the TM model were connected to the crystal structure of the αVβ3 extracellular domain and its C-termini were connected to the corresponding NMR structures of the αIIb and β3 cytosolic domains. The model predicts an αIIb Arg995-β3 Asp723 salt bridge in the membrane proximal cytosolic domains and a change in the secondary structure of the extracellular domains from an extended to a helical conformation that occurs at Pro965 in αIIb and Pro693 in β3, just proximal to the TM domains. Finally, the full length model provides insight into distal interactions since its extracellular portion is held close to the membrane and is therefore sterically inaccessible to large ligands such as fibrinogen, whereas cytosolic regulatory sites such as the talin binding site in β3 are accessible and not sequestered by α/β interactions. In conclusion, we have modeled the αIIbβ3 TM heterodimer using two fundamentally different computational strategies, both of which converge upon similar structures that were used to construct a model for the full-length integrin.