Fibril Surface-Dependent Amyloid Precursors Revealed by Coarse-Grained Molecular Dynamics Simulation

Amyloid peptides are known to self-assemble into larger aggregates that are linked to the pathogenesis of many neurodegenerative disorders. In contrast to primary nucleation, recent experimental and theoretical studies have shown that many toxic oligomeric species are generated through secondary processes on a pre-existing fibrillar surface. Nucleation, for example, can also occur along the surface of a pre-existing fibril—secondary nucleation—as opposed to the primary one. However, explicit pathways are still not clear. In this study, we use molecular dynamics simulation to explore the free energy landscape of a free Abeta monomer binding to an existing fibrillar surface. We specifically look into several potential Abeta structural precursors that might precede some secondary events, including elongation and secondary nucleation. We find that the overall process of surface-dependent events can be described at least by the following three stages: 1. Free diffusion 2. Downhill guiding 3. Dock and lock. And we show that the outcome of adding a new monomer onto a pre-existing fibril is pathway-dependent, which leads to different secondary processes. To understand structural details, we have identified several monomeric amyloid precursors over the fibrillar surfaces and characterize their heterogeneity using a probability contact map analysis. Using the frustration analysis (a bioinformatics tool), we show that surface heterogeneity correlates with the energy frustration of specific local residues that form binding sites on the fibrillar structure. We further investigate the helical twisting of protofilaments of different sizes and observe a length dependence on the filament twisting. This work presents a comprehensive survey over the properties of fibril growth using a combination of several openMM-based platforms, including the GPU-enabled openAWSEM package for coarse-grained modeling, MDTraj for trajectory analysis, and pyEMMA for free energy calculation. This combined approach makes long-timescale simulation for aggregation systems as well as all-in-one analysis feasible. We show that this protocol allows us to explore fibril stability, surface binding affinity/heterogeneity, as well as fibrillar twisting. All these properties are important for understanding the molecular mechanism of surface-catalyzed secondary processes of fibril growth.

. The free energy profile for an  binding to the fibril. 3 Figure S3. Probability contact maps for the five different structural precursors on the fibrillar surface are shown to a residue-to-residue resolution. 5 Figure S4. The free energy profile for a single fibril-like ensemble Abeta11-42 monomer binding to the Abeta42 fibrillar surface (12 chains) is shown (monomer in the "back" orientation). 6 Figure S5. Distribution of sampling windows for simulation with Abeta11-42 peptide (in the fibril-like form) binding to the fibrillar surface is shown. 6 Figure S6. The free energy profile for an Abeta11-42 monomer (biased, fibril-like) binding to the fibril. 7 Figure S7. Five key structural precursors of the fibril-like monomer on the fibrillar surface and their probability contact maps are shown. 8 Figure S8. Probability contact maps for the five different structural precursors of the fibril-like monomer on the fibrillar surface are shown to a residue-to-residue resolution. 10 Figure S9. The total twist angle of protofilaments of different sizes, averaged over three independent simulation trajectories, is shown with the error bar. 11 Figure S10. The initial structure of the protofilament model of different sizes is analyzed and compared. 12 Figure S1. The distribution of sampling windows for simulation with Abeta11-42 peptide binding to the fibrillar surface is shown.
The Abeta monomer (unbiased), is put in the "front" orientation with respect to the central fibril. See the subplot in Fig.2(a) for all the six orientations. These sampling windows represent the last 900 frames of a total of 5000 frames. The data shown ensures that the free energy calculation is converged (See Fig.S2). Figure S2. The free energy profile for an  binding to the fibril.
The simulation is carried out with the monomer being put in 6 different initial orientations, with respect to the central fibril (see subplot in Fig Figure S3. Probability contact maps for the five different structural precursors on the fibrillar surface are shown to a residue-to-residue resolution. (a) C-ter surface precursor (b) N-ter surface precursor (c) Cleft-gate precursor (d) Even-end elongation precursor (e) Odd-end elongation precursor. The color bar scheme for the contact maps is the same as in Fig.3. Hydrophobic, polar, and charged residues are indicated in the sequence and are color coded. See the inset in (e). Figure S4. The free energy profile for a single fibril-like ensemble Abeta11-42 monomer binding to the Abeta42 fibrillar surface (12 chains) is shown (monomer in the "back" orientation). Figure S5. The distribution of sampling windows for simulation with Abeta11-42 peptide (in the fibril-like form) binding to the fibrillar surface is shown.
The Abeta monomer, biased in the fibril-like form, is put in the "back" orientation with respect to the central fibril. See the subplot in Fig.2(a) for all the six orientations. These sampling windows represent the last 900 frames of a total of 5000 frames. The data shown ensures that the free energy calculation is converged (See Fig.S6). Figure S6. The free energy profile for an Abeta11-42 monomer (biased, fibril-like) binding to the fibril.
The simulation is carried out with the monomer being put in 6 different initial orientations, with respect to the central fibril (see subplot in Fig.2(a)). In each panel, a bootstrapping strategy is used to show the free energy profile every 900 frames. The first 500 frames are disregarded.  (a) C-ter surface precursor (b) N-ter surface precursor (c) Even-end elongation precursor (d) Odd-end elongation precursor. The probability contact map next to each structural precursor presents the contacts formed between the monomer and the fibril. The color bar, scaled by probability, is shown on the right. The horizontal axis uses the fibril index (1 to 384; 32x12=384), which sequentially renumbers the 12 monomers in the fibril. The vertical axis describes the residue index of the free monomer by adding up the existing fibril index, 385 to 416 (384+32=416). Different structural features are labeled as "strand" or "loop". Note that in (a), a schematic monomer structure is shown in a red box, with its structure in the fibrillar form. Three strands (strands 1, 2, and 3) and two loops (loops 1 and 2) are indicated. The color scheme for the structure is the same as that in Figure 1. Figure S8. Probability contact maps for the five different structural precursors of the fibril-like monomer on the fibrillar surface are shown to a residue-to-residue resolution.
(a) C-ter surface precursor (b) N-ter surface precursor (c) Cleft-gate precursor (d) Even-end elongation precursor (e) Odd-end elongation precursor. The color scheme for the contact maps is the same as in Fig.3. Hydrophobic, polar, and charged residues are indicated in the sequence and are color-coded. See the inset in (e). Figure S9. The total twist angle of protofilaments of different sizes, averaged over three independent simulation trajectories, is shown with the error bar. Figure S10. Initial structure of the protofilament model of different sizes is analyzed and compared.
(a) The initial fibrillar structures of different sizes are displayed (from top, 12, 24, 36, 48, and 62 chains). (b) Distribution of the twist angle (θ) for the initial structure. (c) Time evolution of the averaged twist angle (θ bar) in the pre-equilibration process, with the last frame representing the structure and angle distribution shown in (a) and (b), respectively.