UCSF

This work describes the development and application of computational models for the investigation of disease and non-disease protein aggregation. We demonstrate how the aggregate equilibrium, formation kinetics, and structural ensembles are influenced by the structural and folding properties of the monomer units. Using a coarse-grained model, we examine the influence of folding rates and mechanisms of non-disease proteins L and G on their aggregation structure and kinetics. We demonstrate that the number and spatial distribution of contacts in the denatured state correlate with aggregation rates and the identity of inter-protein contacts, and that fast forming intermediates may inhibit aggregation through the burial of "sticky" regions. To examine the driving forces underlying the transition from amorphous aggregates to cross-beta ordered amyloid fibrils, we extend this physical model and investigate the aggregation of Alzheimer's Abeta1-40 peptide. We find that the critical nucleus, the highest free-energy species on the aggregation pathway, is composed of ten ordered peptides, the minimum number necessary to stabilize the interfilament contacts defining a fibril axis that enable fast growth of a fibril. Once past the critical nucleus, the model fibril elongates by efficiently incorporating monomers at only one of two asymmetric ends, connecting local structure differences to biased elongation. Familial Alzheimer's Disease (FAD) mutations represented in the model alter the number of peptides necessary to form the critical nucleus as well as the fibril stability, suggesting a molecular mechanism for the spectrum of in vitro aggregation kinetics and morphologies associated with dramatically different FAD clinical outcomes. The region of these mutations is examined in detail through atomistic simulations and NMR analysis of the monomeric Abeta21-30 peptide. Our simulations reproduce relaxation times and ROESY cross-peaks to interpret NMR experimental population averages, but we find no evidence of majority folded structures of this fragment as previously reported. By combining coarse-grained and atomistic simulations with experimental observables, we describe the fast-forming, poorly-populated and disordered states that drive aggregation at a level of detail unattainable by experimental techniques alone, elucidating the link between atomic level properties and aggregation outcomes for the protein L and G and the Alzheimer's Abeta systems.