Conformational Dynamics in Alzheimer Peptide Formation and Amyloid Aggregation

Principal Investigator:
Martin Zacharias

Lehrstuhl für Molekulardynamik, Physik-Department T38, Technische Universität München (Germany)

Local Project ID:

HPC Platform used:
SuperMUC of LRZ

Date published:

The generation and assembly of Aβ peptides into pathological aggregates is associated with neurodegenerative diseases including Alzheimer’s disease. Goal of this project was to better understand the dynamics of γ-secretase a key enzyme for the formation of Aβ peptides using large scale Molecular Dynamics simulations and how it associates with substrate molecules. Using the HPC system SuperMUC it was possible to characterize local and global motions of γ-secretase in atomic detail and how it is related to function. In addition, large scale simulations were employed to investigate the amyloid propagation mechanism at the tip of an already formed amyloid fragment. The kinetics and thermodynamics of the process were analyzed and compared to alternative amyloid secondary nucleation events.


Self-assembly of peptides into ordered amyloid fibrils is associated with several neurodegenerative diseases including Alzheimer’s disease. The key component of the pathological aggregates in case of Alzheimer’s disease is the so-called Aβ peptide resulting from the cleavage of the amyloid precursor protein by the intra-membrane γ-secretase enzyme. The primary function of the enzyme is the proteolytic degradation of membrane proteins. Recently, the structure of γ-secretase has been determined [1] which gives important insights into the complex structural arrangement of several subunits forming the active enzyme. The enzyme active site is localized in the membrane spanning presenilin subunit. The enzyme undergoes important conformational changes during the proteolytic reaction cycle. A full understanding of the enzyme function and the design of inhibitors for interfering with Aβ peptide generation and amyloid formation requires also an understanding of the enzyme conformational flexibility. To elucidate the local and global dynamics of the large γ-secretase complex we have employed extensive molecular dynamics (MD) simulations of γ-secretase using the SuperMUC parallel computer resources. In a parallel study we have also used extensive MD-simulations combined with advanced sampling techniques to study the amyloid propagation and nucleation processes at atomic detail. The results of the studies give important insights into the mechanism of amyloid peptide production and the process of peptide aggregation to form pathological amyloids.

Results and Methods

Dynamics of γ-secretase in a phospholipid membrane

The simulations on γ-secretase have been performed on the entire enzyme complex embedded in a phospholipid membrane and including explicitly the surrounding aqueous solvent (Figure 1). The use of SuperMUC resources allowed us to run simulations for several micro-seconds starting from different start structures and simulation conditions. It was possible to identify long-lived phospholipid binding sites that give hints on putative exo-binding sites for amyloid precursor proteins. The distribution of water molecules found during simulations indicates that the active site and a possible substrate binding channel are accessible for the aqueous solvent. This finding has important consequences for inhibitor design and for understanding how substrate peptides and proteins access the enzyme active site. The most dominant global motions extracted from the simulations correspond to bending and screwing motion of the extracellular nicastrin subunit relative to the membrane-spanning domains which can influence the recognition of the extracellular part of substrate proteins (Figure 1).

Amyloid propagation and secondary nucleation

In addition to the dynamics of γ-secretase repsonsible for the generation of Aβ Alzheimer amyloid peptides, we aso performed simulation of fibril formation. As a model system we used the Aβ9-40 fragment that forms amyloid fibrils in vitro. Using Umbrella Sampling simulations in combination with a replica-exchange advanced sampling method we were able to compare two important sub-processes of amyloid formation. The propagation process of an already formed fibril corresponds to the binding of monomeric amyloid peptides at the tips of an already formed fibril fragment (Figure 2). In extensive simulations we were able to obtain a free energy profile for the process along a dissociation/association reaction coordinate and obtained a binding free energy change in good agreement with experiment. Based on the simulations we derived a model for Aβ association and propagation (illustrated in Figure 2) and were able to estimate also the kinetics of the processes with a fast docking but slow locking phase [2]. Sufficient sampling of possible conformational states is a major issue for the extraction of accurate free energies of binding and for elucidating a mechanistic model of the association and propagation process which required the SuperMUC parallel computer resources.

Apart from the filament propagation at the tip of an already formed filament, a second important process to form pathogenic fibrils is the nucleation of new fibrils often promoted by already formed amyloids. Exactly this process might be responsible for the formation of toxic oligomers. Such secondary nucleation may involve the hydrophobic lateral surfaces of fibrils and we performed free energy simulations to characterize the binding of monomers and short fragments of filaments to attach to the lateral surface of a pre-formed existing filament (Figure 3). Interestingly, the calculated free energy for binding of Aβ peptides to the lateral surface of an already formed filament is similar to the calculated propagation free energy indicating that both processes may compete. However, the simulations indicated that binding of peptide monomers, dimers of trimers resulted in bound structures that deviate significantly from the conformation in a filament and hence are unlikely to form productive nuclei for initiating new filament propagation. Only filaments that included four monomers resulted in complexes with a stable filament type structure that can initiate the formation of new filaments (Figure 3).

On-going Research / Outlook

Understanding the mechanism of Aβ peptide production and fibril formation is of major biomedical importance. The large scale simulations of γ-secretase as well as umbrella sampling and replica exchange simulations on filament formation show excellent scaling on parallel super computers. In future research we plan to study the influence of mutations on domain motions of γ-secretase and how mutations affect amyloid formation and propagation both in collaboration with experimental groups at TUM. The simulations of amyloid propagation and γ-secretase dynamics were only possible by using the SuperMUC parallel computer facilities.

References and Links

[1] Bai X, Yan C, Yan G, Lu P, Ma D, Sun D, Zhou R, Scheres SHW, Shi Y. An atomic structure of human γ-secretase. Nature 525 (2015) 212-18.

[2] Schwierz N, Frost CV, Geissler PL, Zacharias M. Dynamics of Seeded Aβ40-Fibril Growth from Atomistic Molecular Dynamics Simulations: Kinetic Trapping and Reduced Water Mobility in the Locking Step. J Am Chem Soc. 138 (2016) 527-39.

[3] Schwierz N, Frost CV, Geissler PL, Zacharias M. From Aβ-Filament to Fibril: Molecular Mechanism of Surface-Activated Secondary Nucleation from All-Atom MD Simulations. J Phys Chem B 121 (2017) 671--82.

Research Team

Christina Frost, Manuel Hitzenberger, Jonathan Coles, Danial Pourjafar Dehkordi, Julian Hartmann, Florian Kandzia, Korbinian Liebl, Asman Nayis, Maria Reif, Till Siebenmorgen, Paul Westphälinger

Project Partners

Aliaksei Krukau, Gerald Mathias, Leibniz-Rechenzentrum, München

Scientific Contact:

Prof. Dr. Martin Zacharias
Lehrstuhl für Theoretische Biophysik (T38) - Molekulardynamik
Technische Universität München
James-Franck-Str. 1, D-85748 Garching (Germany)
e-mail: martin.zacharias[at]

NOTE: This report was first published in the book "High Performance Computing in Science and Engineering – Garching/Munich 2018".

LRZ project ID: pr48po

January 2019

Tags: LRZ Life Sciences