Alzheimer’s Disease (AD) is the commonest fatal neurodegenerative disease
causing dementia. It has a devastating effect on individuals and their families and
causes progressive cognitive, social and functional decline with no current cure.
AD is characterised neuropathologically by amyloid-β (Aβ) and tau protein
deposition, but this does not fully explain how the brain is affected and why
neurodegeneration occurs. Key molecular drivers of disease could include cell
stress and impaired mitochondrial oxidative phosphorylation with synaptic loss.
Elucidating key mechanisms of this disease process in vivo has the potential for
better diagnostic classification, and a better understanding of the neurobiology in
order to prioritise targets for earliest interventions to delay or reverse disease.
In this study I set out to determine whether multimodal imaging using PET
(amyloid-beta, the sigma-1 receptor, mitochondrial complex I, synaptic vesicle 2A)
complemented by MR (structural, arterial spin labelling and neurite orientation
diffusion density imaging) can determine key clinically relevant cellular
mechanisms in AD patients.
Chapter 1 provides an introduction to AD and imaging. Chapter 2 outlines the
material and methods used in this study.4
In Chapter 3 I aimed to boost the diagnostic power of one key trigger of AD -
amyloid-β, using [18F]Florbetapir PET, with machine learning applied to big data.
I created an algorithm that utilised spatial information by k-means clustering then
an optimised quadratic support vector machine based on the progression of regional
brain vulnerabilities to AD across 758 volunteers from the Alzheimer’s Disease
Neuroimaging Initiative (ADNI). I was able to develop an automated diagnostic
decision support and show that my algorithm outperforms classifiers proposed
previously for amyloid-β PET.
In Chapter 4 I aimed to understand the underlying mechanisms of AD using
multiple imaging modalities in my highly phenotyped cohort of early AD and
controls. I quantified the in vivo density of the endoplasmic reticulum cell stress
marker, the sigma 1 receptor (S1R) using [11C]SA4503 PET, as well as that of
mitochondrial complex I (MC1) with [18F]BCPP-EF and the pre-synaptic vesicular
protein SV2A with [11C]UCB-J. I also integrated these assessments with regional
brain volumes and brain perfusion (CBF) measured with MRI arterial spin labelling
and followed the AD patients longitudinally to estimate rates of change with disease
progression over 12-18 months. I generated new evidence for widespread cellular
stress and bioenergetic abnormalities in early AD and showed how they may be
clinically meaningful.
In Chapter 5 I sought to understand synaptic microstructure from my early AD
cohort in further detail. I quantified in vivo measures of microarchitecture5
evaluating extracellular free water (FISO), neurite density (NDI) and orientation
dispersion (ODI) using neurite orientation dispersion imaging (NODDI), as well as
more conventional DTI measures of fractional anisotropy (FA), mean/axial/radial
diffusivity (MD, AD, RD, respectively), and the pre-synaptic vesicular protein
SV2A with [11C]UCB-J PET in my early AD cohort compared to controls. My
results showed increased extracellular free water (FISO) using NODDI MRI could
be more sensitive to differences in pathology related to synapse loss that could be
used to support early-stage evaluations of novel therapeutics for AD
Overall, my study provides a systematic investigation of the cellular phase of AD
using PET and MR imaging. My main finding was that widespread cell stress in
AD leads to regionally vulnerable expression of pathology. If these findings were
applied to both clinical practice and drug discovery they could lead to better
diagnostic classification and prioritise targets to modulate the disease.