Molecular origin of the differential stabilities of the protofilaments in different polymorphs: molecular dynamics simulation and deep learning

J Biomol Struct Dyn. 2024 Nov 17:1-17. doi: 10.1080/07391102.2024.2427364. Online ahead of print.

Abstract

Fragments of α-synuclein, an intrinsically disordered protein, whose misfolding and aggregation are responsible for diseases like Parkinson's disease and others, can co-exist in different polymorphs like 'rod' and 'twister'. Their apparently stable structures have different degrees of tolerance to perturbations like point mutations. The molecular basis of this is investigated using molecular dynamics-based conformational sampling. A charge-swapping mutation, E46K, known to be a reason for the early onset of Parkinson's disease, has differential impact on two polymorphs, and its molecular reason has been probed by investigating the intra-fibril interaction network that is responsible for stabilizing the aggregates. Two different quaternary level arrangement of the peptides in two polymorphs, establishing two different types of interrelations between residues of the peptide monomers, form the basis of their differential stabilities; a Deep Neural Network (DNN)-based analysis has extracted different pairs of residues and their spatial proximities as features to distinguish the states of two polymorphs. It has revealed that difference in these molecular arrangements intrinsically assigns key roles to different sets of residues in two different forms, like a feedback loop from quaternary structure to sequence level; an important insight into the sequence-structure relationship in general. Such atomic level insights were substantiated with the proof of differences in the dynamic correlation between residue pairs, altered mobilities of the sidechains that affects packing and redistribution of the weightage of different principal modes of internal motions in different systems. The identification of key residues with altered significance in different polymorphs is likely to benefit the planned design of fibril breaking molecules.

Keywords: deep learning; fibril; molecular dynamics; principal component analysis; α-synuclein.