(515i) Quantifying Peptide Configurational Landscapes and Aggregation through Sequence-Transferable Bottom-up Coarse-Grained Models

Molecular insight into protein aggregation is important for understanding aggregation-related diseases such as Alzheimer’s, Pick’s, chronic traumatic encephalopathy, and many others. The propensity for a particular intrinsically disordered protein to form aggregates, and its resulting aggregate morphologies, depends significantly on the nature of its sequence. While simulations can provide detailed pictures into such processes, correctly reproducing such sequence-dependent cooperative behavior leading to formation of ordered amyloid fibrils necessitates an advanced multiscale modeling approach with sequence-accurate rather than toy simulation models. Bottom-up coarse-graining methods, such as relative entropy minimization, provide encouraging avenues towards sequence-specific coarse-grained (CG) models capable of such investigations, through optimization of CG models from reference atomistic simulations, without experimental structure information that may be unavailable for any particular intrinsically disordered protein.

Using relative entropy coarse-graining, we have shown that this approach produces an accurate model for the PHF6 (VQIVYK) peptide. This sequence motif appears in the fibrillar cores of all known pathological structures of the tau protein, implicated in Alzheimer’s disease and other tauopathies. The CG model, based on reference atomistic ensembles of only 3 interacting PHF6 chains, remarkably reveals fibril structure and phase behavior in large-scale simulations, and provides quantitative insights into the conformational landscape of oligomers. Moving forward, we extend this case study to wider alphabets of amino acids for more general use, using an expanded ensemble approach to incorporate information from reference systems of multiple sets of varied sequences of interacting peptides. Importantly, we demonstrate a method to optimize the reference simulation conditions for optimal CG model fidelity and transferability of the models.

We show that the resulting CG models are remarkably sequence-transferable given their coarseness, able to accurately predict configurations and secondary structure propensities. We then screen sequences using a combination of direct molecular dynamics simulations of assembly along with replica exchange to probe oligomer and aggregate structures, demonstrating sequence effects on assembly into compact β-sheet-rich aggregates vs. ordered, extended, amyloid fibrils. We also use this approach to expand the study of tau fragments to a larger 19-residue subsequence of tau found within the cores of multiple known disease-associated tau folds. We investigate the ability of this fragment to template growth of particular fibril structures, giving insight into shape propagation mechanisms governing tau aggregation.