Project Summary/Abstract
Amyloid fibril formation is central to the disease etiology of a number of human diseases, including Alzheimer’s
disease, type 2 diabetes, and a variety of prion diseases. Although molecular structures for thousands of
amyloid fibrils have been resolved using techniques like X-ray crystallography and nuclear magnetic resonance
(NMR), the mechanism of amyloid fibril formation is largely unknown. The mechanism of primary nucleation,
whereby fibril formation begins in a solvent environment that previously did not contain any amyloid fibrils, a
crucial step in amyloid disease onset, is particularly mysterious. Dye-binding fluorescence microscopy
experiments have been used to observe the spontaneous formation fibril formation in microfluidic chambers
from individual primary nucleation sites. These experiments revealed two key mechanistic details: 1) fibril
formation propagated through solution as a traveling wave of constant velocity moving away from the primary
nucleation site, and 2) there exists a linear relationship between the lag time before fibril formation and the
inverse of volume. We hypothesize that the confinement of insulin to smaller volumes is an evolutionary
adaptation that renders amyloid fibril formation prohibitively slow, in turn, influencing the size of insulin
granules in pancreatic beta cells. We will develop novel top-down coarse-grained model that utilize a bridged
approach, whereby two representations of an ensemble of fibril-forming proteins (one purely topological
network representation and one granular representation in explicit space) exchange information as time
evolves. This approach will leverage the high computational efficiency of exponential-family random graph
models (purely topological), with improved spatial realism provided by a minimal explicit space model based on
a Lennard-Jones fluid. The models will first be fit using a threefold validation strategy whereby they will be
parameterized to simultaneously reproduce three known experimental observables: the fibril’s topological
structure (derived from structures reported in the protein data bank), fibril growth kinetics (compared to dye-
binding fluorescence experiments), and the spatial propagation patterns of fibril formation (compared to
aforementioned microfluidic experiments). Analysis of the validated models will then be used to propose
potential mechanisms for primary nucleation, the modulation of which is actively being explored for the
development of preventative treatments for amyloid diseases. The proposed work will require an innovation to
the network Hamiltonian methodology (first introduced by the PI and others), in that it will be the first to include
explicit spatial degrees of freedom. This development will facilitate the comparison of network Hamiltonian
models to experimental results and enhance the predictive power of the simulations, for both the present work
and future studies in molecular self-assembly.