UC Irvine

γ-crystallins are structural eye lens proteins responsible for focusing light into the retina. These proteins are highly stable, being capable of remaining soluble at concentratins exceeding 300 g/L for an entire lifetime. Upon a loss in solubility, either by mutation or chemical damage to the protein structure, opaque aggregates form in the eye lens, known as cataract. Understanding the conformations and interactions of the γ-crystallins in their aggregated state will provide a deeper understanding of the mechanisms behind cataract formation.The work presented in this dissertation will investigate cataract-related sequences of γS- and γD-crystallins (γS-WT and γD-WT, respectively). Using molecular dynamics and other modeling techniques, I investigate potential sites for interprotein interaction, often through the exposure of hydrophobic residues. In the congenital cataract-related G18V variant of γS-crystallin (γS-G18V), the exposure of hydrophobic patches are identified in a relatively folded protein. Simulated protein-ligand conformations of a hydrophobic probe (ANS) identify hydrophobic sites both local and allosteric to the site of mutation. In the W42R variant of γD-crystallin (γD-W42R), a 17 μs molecular dynamics simulations shows the spontaneous separation of the N- and C-terminal domains (whereas γS-WT remains stable for 50 μs). Two protein simulations show that the hydrophobic interdomain interface becomes the main site for interprotein interaction, providing strong support for a domain swapping aggregation pathway at physiological conditions. The thermal stability is analyzed for the γS-crystallins of the human and Antarctic toothfish. The less thermally stable toothfish proteins show correlated increases in backbone flexibility and decreases in the packing of the hydrophobic core, yet maintain similar structure and dynamics at their native temperature. Finally, an analysis of the domain-domain motions in γD-WT fluctuations is performed in the presence and absence of macromolecular crowding. The autocorrelation function of the interdomain distance ages over 50 μs of simulation, showing non-convergent dynamics even after significant computational sampling.