1.Introduction
Amyloid-beta (Aβ) peptides of 40 and 42 amino acids are proteolytically cleaved form amyloid precursor protein by β- and γ-secretases. Soluble Aβ dimers isolated from Alzheimer’s cortex directly induce tau hyperphosphorylation and neuritic degeneration, and small Aβ oligomers are believed to the most toxic species.1,2 The aggregation kinetics of Aβ follows a sigmoidal curve with three phases: a lag phase free of any Thioflavin T fluorescence signal, a growth phase or fibril elongation followed by a saturation phase.3,4 Aβ42, less abundant but more toxic than in its Aβ40 counterpart, is very sensitive to protein concentration upon aggregation in the bulk solution, and also pH, temperature, the presence of membrane and the addition of seeds. Characterizing the early Aβ oligomers is challenging experimentally and computationally.5-7
This is an experimental challenge due to the transient and heterogeneous ensemble of oligomer structures, the fact that most experimental observables provide time- and space-averaged properties,6 and the µs time resolution cannot be achieved yet.8 This is a challenge for computer simulations due to the accuracy of the protein and water force fields,9,10 and the time scale of primary nucleation in pure buffer which is on the order of several hours for Aβ42 at µM concentrations.4 By using atomistic molecular dynamics (MD) simulations, the current time lengths vary from 2.5 to 30 µs for 20 Aβ42 peptides with an implicit solvent model,11 and Aβ40 monomer in explicit water,12 respectively. Simulations of Aβ40/42 species with free lipids and calcium ions in aqueous solution are also currently limited to the µs time scale.13-15 Going beyond this time scale or sampling rare events has been made possible by the use of coarse-grained or mesoscopic models and enhanced sampling techniques such as path or umbrella sampling and metadynamics, among others. Our current structural view of Aβ40/42 monomers and small oligomers is random coil, with increasing β-sheet content as the oligomer size increases.16 In this study, we emphasize other metastable oligomers based on α-helix bundles that are predicted by the AlphaFold2 machine learning and are indirectly supported by low-resolution experimental data on many amyloid polypeptides.