Distinct contribution of white matter damage to the clinical syndrome of Alzheimer's disease by Jean-Philippe Coutu
B.Eng. Engineering Physics, Ecole Polytechnique de Montreal (2010)
Submitted to the Division of Health Sciences and Technology in partial fulfillment of the requirements for the degree of
DOCTOR OF PHILOSOPHY IN MEDICAL ENGINEERING MEDICAL PHYSICS
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
February 2016
@ 2016 Massachusetts Institute of Technology. All rights reserved.
Signature redacted A uthor ...... Division of HealtN Sciences and Technology -..- Noveiber 25, 2015 Certified by...... Signature redacted David H. Salat, PhD Associate Professor in Radiology, Harvard Medical School Theds Supervisor
Accepted by Signature redacted... Emery N. Brown, MD, PhD Director, Harvard-M T Program in Health Sciences and Technology Professor of Computational Neuroscience and Health Sciences and Technology
MASSACHUSETT INSTITUTE OF TECHNOLOGY
MAR 14 2016
LIBRARIES 2 Distinct contribution of white matter damage to the clinical syndrome of Alzheimer's disease by Jean-Philippe Coutu
Submitted to the Division of Health Sciences and Technology on November 25, 2015, in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY IN MEDICAL ENGINEERING MEDICAL PHYSICS
Abstract Alzheimer's disease is a neurodegenerative disorder affecting over 5.1 million individ- uals in the United States today. The dementia exhibited with the disease is currently thought to be primarily due to amyloid plaques and neurofibrillary tangles. However, several other changes occur, including severe white matter damage that is yet to be fully understood. Such white matter damage includes white matter lesions (WML), which are more common in individuals with Alzheimer's disease than in non-demented individuals. WML are of presumed vascular origin because they show features of small-vessel disease and are more prevalent in individuals with vascular risk. It is currently unclear whether WML are linked to the neurodegenerative pathologies of Alzheimer's disease or are an independent factor that influences clinical course.
In this work, we used sensitive diffusion MRI measures to determine that the tis- sue properties of WML slightly differed microstructurally between individuals with Alzheimer's disease and non-demented controls, and were strongly related to ventric- ular enlargement. In order to further understand the role of WML, we factored the volume of WML with four other neuroimaging markers affected in Alzheimer's disease and discovered two statistically distinct factors presumed to be due to differing under- lying disease processes. One process strongly related to volume and tissue properties of WML, ventricular enlargement, age and cerebral perfusion, while the other process related to imaging markers associated with neurodegeneration. A decrease over time in the first process, interpreted as the age- and vascular-related factor, led to similar cognitive decline as the neurodegenerative factor independently, demonstrating the potential added therapeutic benefit of targeting this disease process that is distinct from the classical neurodegenerative component of the disease.
Thesis Supervisor: David H. Salat, PhD Title: Associate Professor in Radiology, Harvard Medical School
3 4 Acknowledgements
Personal acknowledgements
I want to thank first and foremost my thesis supervisor, David Salat, for his mentoring throughout the last five years. It has been a privilege, an honor and a great pleasure for me to learn from him. He has always been very supportive of my ideas and has given me plenty of room to explore new projects. Thank you for feeding my curiosity and being the greatest contributor to my scientific growth.
I also want to thank my thesis chair, Bruce Rosen, for his mentoring through his regular student meetings and my thesis committee meetings, and for making the Martinos center a very open and collaborative environment from which I have learned greatly. I also want to thank my other two committee members for their mentoring and help through my PhD, Doug Greve in particular for his precious technical insight and David Boas in particular for looking at my work from a different angle. I also want to thank several other mentors I have had, beginning with Jean Chen, who welcomed me into David's research group and helped me get started on several interesting projects. I also want to thank Diana Rosas for her clinical insight and the many opportunities she has offered me throughout and beyond my PhD. I also want to thank Bruce Fischl and members of his wonderful FreeSurfer group with whom I have interacted with throughout my PhD and who have provided both great technical insight and friendship. I also want to thank Steve Greenberg and Edip Gurol for offering their clinical perspective to my work and for teaching me about small-vessel disease.
Next, I want to thank individuals I have interacted with on a regular basis and who have contributed to my graduate work in many different ways, in particular Alison Goldblatt, Emily Lindemer, Paul Wilkens, Tyler Triggs, as well as all other members of the lab who were instrumental to the completion of my graduate work and other research projects: Emma Boyd, Jeanne Tong, Casey Callaghan, Kim Stephens, Keith Malarick, Suzanne Imbriglio, Matt Linehan, Robert McKinnis, Lisa Glukhovsky, Ali Amin-Mansour and Steven Swinford. I also want to thank my collaborators from South Korea: Seon Young Ryu, Seung Hwan Lee and Chang-Woo Ryu; as well as some of my colleagues at the Martinos center and MIT who I have enjoyed working with and from whom I have learned a lot: Juliette Selb, Phoebe Chan, Jon Polimeni, Jayashree Kalpathy-Cramer, Elfar Adalsteinsson, Maria Mody, Trey Hedden and Randy Gollub. In particular, I want to express my gratitute to members of our regular journal club: Melissa Amick, Jean Augustinack, Andre van der Kouwe, Vincent Corbo and Meghan Robinson. It was great learning with you!
I would also like to thank those who have made my journey to MIT possible, in particular Caroline Boudoux and Frederic Lesage, as well as Thomas Gervais, who have inspired me to
5 apply to MIT and have supported me and my application in various ways. I also want to thank the HST office and faculty for being extremely welcoming in the early days of my PhD and throughout, in particular Julie Greenberg, Laurie Ward, Richard Cohen, Joseph Stein, Lora Maurer, Brett Bouma and Traci Anderson. I also want to thank everyone who welcomed me at my first home in the United States, the Sidney-Pacific Graduate Residence, in particular Roger and Dottie Mark, Amy Bilton, Chelsea Curran and Anand Oza, and the many friends I have made there: Ece Gulsen, George Lan, George Tucker, Holly Johnsen, George Chen, Sumit Dutta, Ramesh Sridharan and many others. I also want to acknowledge many friends I met in the HST program and at the Martinos Center: Audrey Fan, Christin Sander, Jeffrey Stout, Pavitra Krishnaswamy, Stephanie Yaung, Kevin Chen, Dan Chonde, as well as my first roommate on this side of the border, Justin Lo, and many others. I also want to make a special mention of Louis Gagnon, a fellow French Canadian who has been very supportive of my early days in the HST program and at the Martinos center, and who has been a precious friend. I also want to acknowledge other French Canadian friends I met during my time in Boston and Cambridge: Rosalie Belanger-Rioux, Vincent Corbo, Chenjie Xia, Vincent Laverdiere, Roxane Lavoie, Vincent Beliveau, Sam Osseiran and Antoine Ramier.
I also want to thank my undergraduate friends for their support and the fun times we spent together before and during my PhD: Dany Chagnon, Alexandre Robitaille, Amelie St-Georges- Robillard, Benoit Bourrassa-Moreau, and Laurent Potvin-Trottier and Kathy Beaudette who also came to study in Boston and Cambridge and have made my journey in this foreign land feel a little bit more like home. I also want to thank my relaxation crew, who has provided me with many timely breaks to maintain my sanity throughout my journey, especially towards the end: Dany Chagnon, Patrick Lavoie, Pascale Brunet and Claude Mercure-Dansereau.
Finally, I want to acknowledge and thank members of my family: my father Yves Coutu, my mother, Liliane Samson, and my two wonderful sisters, Marie-Pier Coutu and Val6rie Coutu. You have made me who I am and I am extremely grateful that I was able to come this far thanks to you and your support. I also want to acknowledge my cat, Pepsi, for his love and purring support. Last, but not least, I want to give an enormous thank you to my fiancee, Sunny Vanderboll, who I met during my time here and who has made this journey much greater than anything I could have dreamed of. Thank you so much for your support and your love; I am looking forward to more adventures with you in the future!
Funding and data sources
My thesis research was funded by the National Institutes of Health grants R01NR010827, NS042861, NS058793 and was carried out at the Athinoula A. Martinos Center for Biomedical Imaging within the Massachusetts General Hospital (MGH), using resources provided by the Center for Functional Neuroimaging Technologies, P41RR14075, a P41 Regional Resource supported by the Biomedical Technology Program of the National Center for Research Resources (NCRR), National Institutes of Health. This work also involved the use of
6 instrumentation supported by the NCRR Shared Instrumentation Grant Program and/or High- End Instrumentation Grant Program; specifically, grant numbers S1ORRO21110, S10RR023401, S10RR019307, S10RR019254 and S10RR023043.
I am also grateful for having been funded by the Medical Engineering Medical Physics Fellowship of the MIT/Harvard Division of Health Sciences and Technology (HST), the Advanced Multimodal Neuroimaging Training Program (2T90DA022759; P.I. Bruce Rosen) of the MGH/HST Athinoula A. Martinos Center for Biomedical Imaging, the Natural Sciences and Engineering Research Council of Canada, the Fonds de Recherche du Quebec - Nature et les Technologies, the Fonds de Recherche du Quebec - Sante, and the HST IDEA2 Program supported by the Peter C. Farrell (1967) Fund.
Data used in the preparation of this thesis were obtained from the Alzheimer's disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). The ADNI was launched in 2003 as a public-private partnership, led by Principal Investigator Michael W. Weiner, MD. The primary goal of ADNI has been to test whether serial magnetic resonance imaging (MRI), positron emission tomography (PET), other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of mild cognitive impairment (MCI) and early Alzheimer's disease (AD). For up-to-date information, see www.adni-info.org.
Data collection and sharing for this project was funded by the Alzheimer's Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: AbbVie, Alzheimer's Association; Alzheimer's Drug Discovery Foundation; Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; Eurolmmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research & Development LLC.; Lumosity; Lundbeck; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical Company; and Transition Therapeutics. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (www.fnih.org). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer's Disease Cooperative Study at the University of California, San Diego. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California.
7 8 Table of Contents
A b stra ct ...... 3 A ckn o w led ge m e nts...... 5 Personal acknow ledge m e nts ...... 5 Funding and data sources ...... 6 T a b le o f C o nte nts ...... 9 Chapter 1: Motivation and background ...... 14 1.1 Alzheimer's disease: clinical profile and current state of the art ...... 14 1.1.1 Epidemiological and socio-economical perspective ...... 14 1.1.2 Cognitive decline and mild cognitive impairment ...... 15 1.1.3 Classical pathological and neuroimaging features of the disease ...... 16 1.1.4 Vascular co-morbidities and risk factors...... 18 1.1.5 Current treatment and focus of therapeutic trials ...... 19 1.2 White matter pathology in Alzheimer's disease...... 20 1.2.1 Increased prevalence and volume of white matter lesions ...... 20 1.2.2 Histological correlates of white matter lesions ...... 22 1.2.3 Epidemiological associations with risk factors for vascular disease...... 24 1.2.4 Evidence of poor perfusion in white matter lesions ...... 25 1.2.5 White matter atrophy and ventricular enlargement...... 26 1.2.6 Are the white matter findings consistent with the classical features of the disease? 27 1.3 Motivation for studying white matter in Alzheimer's disease ...... 28 1.3.1 Heterogeneity of the disease...... 28 1.3.2 Possibility of a distinct concurrent disease mechanism ...... 30 1.3.3 Potential existing avenues for delaying disease onset and for treatment...... 32 1.4 Sum m ary of thesis aim s ...... 35 Chapter 2: Investigation of diffusion tensor imaging as a sensitive tool to measure white matter m icro stru ctu re ...... 3 8 2 .1 O v e rv ie w ...... 3 8 2 .2 In tro d u ctio n ...... 3 9 2.3 M aterials and m ethods...... 45
9 2.3.1 Participants and MRI acquisition ...... 45 2 .3 .2 P re p ro cessing ...... 4 8 2.3.3 Diffusion Tensor and Kurtosis Imaging ...... 48 2.3.4 Per-voxel statistical analyses ...... 49 2.3.5 K-means clustering of region-of-interests ...... 50 2 .4 R e su lts ...... 5 2 2.4.1 Associations between age and diffusion measures of microstructure ...... 52 2.4.2 Variance and complementarity of the diffusion measures ...... 53 2.4.3 Clustering of age-effects based on multivariate diffusion patterns...... 54 2.4.4 Group differences between non-demented controls and individuals with MCI/AD .. 59 2 .5 D iscu ssio n ...... 6 0 2 .6 A p p e n d ix ...... 6 3 2.6.1 Speculation on the histological basis of the age effects observed for diffusion kurtosis imaging measures based on its biophysical model ...... 63 2.6 .2 Study lim itatio ns ...... 65 Chapter 3: Properties of white matter lesions in Alzheimer's disease do not differ from non- demented controls and poorly relate to markers of neurodegeneration...... 67 3 .1 O v e rv ie w ...... 6 7 3 .2 Intro d u ctio n ...... 6 8
3.3. M aterials and Methods...... 70 3.3.1 Participants and MRI acquisition ...... 70 3.3.2 D iffusion data processing ...... 72 3.3.3 Automated subcortical and WML segmentation...... 72 3.3.4 Registration procedures...... 75 3.3.5 Normative data calculation...... 75 3.3.6 Statistical analyses ...... 76 3 .4 R e su lts ...... 7 7 3.4.1 Group differences in prevalence and tissue properties of WIM L ...... 77 3.4.2 Associations between tissue properties of WML and neuroimaging markers of AD.. 79 3.4.3 Associations between tissue properties of parahippocampal white matter and neuroimaging markers of AD ...... 80 3.4.4 Associations between tissue properties of normal-appearing white matter and neuroim aging m arkers of A D ...... 82 3 .5 D iscu ssio n ...... 8 3
10 3.6 Appendix ...... 88 3.6.1 Effect of APOE genotype in the m odels...... 88 3.6.2 Study lim itations ...... 91 Chapter 4: Volume of white matter lesions and ventricular and hippocampal volumes cluster together as one of tw o distinct processes in Alzheim er's disease ...... 94 4.1 Overview ...... 94 4.2 Introduction ...... 95 4.3 M aterials and M ethods...... 98 4.3.1 Participants and M RI acquisition ...... 98 4.3.2 Autom ated subcortical and w hite m atter lesion segm entation ...... 100 4.3.3 Arterial spin labeling data processing...... 100 4.3.4 Factor analysis...... 102 4.3.5 Statistical analyses ...... 103 4.4 Results ...... 104 4.4.1 Classes of degenerative change...... 104 4.4.2 Associations w ith tissue properties of w hite m atter lesions...... 106 4.4.3 Associations w ith cerebral blood flow ...... 110 4.5 Discussion...... 114 4.6 Appendix ...... 119 4.6.1 Study lim itations ...... 119 Chapter 5: Distinct Alzheimer's disease processes contribute independently to cognitive decline: prelim inary assessm ent ...... 122 5.1 Overview ...... 122 5.2 Introduction ...... 123 5.3 M aterials and M ethods...... 124 5.3.1 Participants and M RI acquisition ...... 124 5.3.2 Im aging data processing ...... 126 5.3.3 Com putation of factor scores ...... 126 5.3.4 Statistical analyses ...... 126 5.4 Results ...... 127 5.4.1 Associations between MMSE decline and change in factors scores and CBF ...... 127 5.4.2 Determ inants of the change in CBF ...... 130 5.4.3 Determ inants of the change in factor scores ...... 132
11 5.5 Discussion...... 133 5.6 Appendix ...... 136 5.6.1 Study lim itations ...... 136 Chapter 6: Discussion and perspectives ...... 139 6.1 Sum m ary and future w ork ...... 139 6.2 The evidence for a distinct white matter pathway in Alzheimer's disease...... 144 6.3 Estim ated potential im pact of intervention on delaying disease onset...... 147 6.4 The future of risk prevention for Alzheim er's disease ...... 148 6.5 Conclusion ...... 149 Chapter 7: References...... 152
12 13 Chapter 1: Motivation and background
1.1 Alzheimer's disease: clinical profile and current state of the art
1.1.1 Epidemiological and socio-economical perspective
Over 5.1 million individuals over 65 years old (11%) have Alzheimer's disease in the United
States today (Alzheimer's Association, 2015a). It is estimated that this number will increase to
13.5 million individuals (16%) by 2050. While Alzheimer's disease is the sixth leading cause of
death in the United States, it is the only leading cause of death that cannot be slowed,
prevented or cured (Alzheimer's Association, 2015b). In fact, it is one of the only leading causes
of death for which there was an estimated increase in the number of deaths from 2000 to 2013,
increasing by 71% compared to a decrease of 14% for heart disease and of 23% for stroke, the
two leading causes of death in the United States. This is not taking into account the disability
that comes with Alzheimer's disease and the costly and mentally-taxing toll it takes on the large
amount of caregivers needed to cope with the disease.
The cost of Alzheimer's disease and other dementias is estimated at $226 billion for healthcare,
long-term care and hospice for the year 2015 only (Alzheimer's Association, 2015b). This cost is
nearly doubled if the unpaid care given by family and friends as well as their lost revenue is
taken into account. Furthermore, direct costs are expected to reach $1.1 trillion per year by
2050. In comparison, the estimated NIH budget to study Alzheimer's disease in 2015 is $586
million (NIH, 2015), a meager 0.3% of the current total cost to Americans.
14 It is estimated that a treatment delaying the onset of Alzheimer's disease by five years first applied in 2025 could reduce the disease toll by 5.7 million individuals or 42% in 2050 and the annual cost by 33% or $367 billion (Alzheimer's Association, 2015a). As such, studying non- conventional pathways to the disease, even those who might not necessarily be directly causative of the disease but contribute to a faster cognitive decline towards Alzheimer's disease, can be extremely beneficial socio-economically even if they do not cure the disease.
1.1.2 Cognitive decline and mild cognitive impairment
Alzheimer's disease is a neurodegenerative disorder. It affects most if not all cognitive domains, but by far its strongest and earliest effect is to diminish and remove the ability to form new memories. This is especially true of new long-term memories about important life events, defined as episodic memory, and occurs in both early-onset familial Alzheimer's disease, which is genetically determined and represents only a small percentage of cases, and late-onset
Alzheimer's disease, which constitutes the large majority of cases, and for which the cause of the disease remains unclear and under investigation (Alzheimer's Association, 2015b).
Eventually, in both cases, Alzheimer's disease robs the individual of the ability to perform normal activities of daily life, and a caregiver is needed.
In an attempt to identify and study the early stages of Alzheimer's disease, the concept of mild cognitive impairment (MCI) started to be used in the late 1990s and early 2000s as a prodromal condition or "risk state" for Alzheimer's disease (Petersen et al., 1999, 2009; Morris et al., 2001;
Gauthier et al., 2006). Many clinical tests exist to diagnose Alzheimer's disease and mild cognitive impairment, with the most practically used both in medicine and research being the
15 Mini-Mental State Examination (MMSE), Montreal Cognitive Assessment (MoCA) and the
Clinical Dementia Rating (CDR) (Folstein et al., 1975; Morris, 1993; Nasreddine and Phillips,
2005). One of the difficulties of using clinical tests is that they attempt to objectify an intrinsically subjective measure of someone's brain. While cognitive decline is the obvious consequence of the disease and, by large, attempts to diagnose it have been based on cognitive testing, ultimately, a definite diagnosis of Alzheimer's disease can only be made post-mortem with a pathological assessment of the brain (Mirra et al., 1991; Hyman and Trojanowski, 1997;
Hyman et al., 2012). However, while not used clinically to diagnose the disease at this point, several imaging measures have now been shown to be reliable biomarkers of the disease as detailed below.
1.1.3 Classical pathological and neuroimaging features of the disease
The most characteristic pathological hallmarks of AD that have been consistently found on autopsy are P-amyloid plaques and intracellular neurofibrillary tangles, commonly referred to as the 'plaques and tangles' pathology of AD (Mirra et a!., 1991; Hyman and Trojanowski, 1997;
Hyman et al., 2012). A widely-accepted model of the disease progression published in 2010 (Jr et a/., 2010) describes the appearance and accumulation of amyloid-beta (AP) into plaques as the likely earliest concrete sign of the disease, which is then sequentially followed by the appearance of intracellular neurofibrillary tangles through tau-mediated neuronal injury, changes in brain structure, decline in memory and finally decline in clinical function and the ability to perform activities of daily life.
16 The changes in brain structure are multiple. The hippocampus and medial temporal lobe have long been recognized to be the most affected structures in the brains of individuals with AD
(Seab et al., 1988), and this is also where neurofibrillary changes are observed in the earliest stages of the disease, as demonstrated by the widely-known Braak and Braak staging (Braak and Braak, 1991). Through many studies over the last several decades, including the famous case of Henry Molaison (H.M.), an epilepsy patient who underwent a bilateral removal of his hippocampi and lost all ability to form new memories, it is clear today that the degeneration of the hippocampus and medial temporal lobe is primarily responsible for the episodic memory deficits in AD (Scoville and Milner, 1957; Milner et al., 1968; Corkin, 2002). However, it is equally clear that the entire neocortex is affected throughout the disease. Indeed, while the thickness of the cortical mantle is known to generally diminish with aging (Salat et al., 2004;
Fjell et a/., 2009), its reduction is greater in individuals with Alzheimer's disease and follows a more specific spatial pattern (Lerch et al., 2005; Singh et al., 2006; Querbes et a!., 2009).
Recently, researchers at MGH have studied which cortical regions are affected in early AD and created a cortical signature that is predictive of conversion to AD in non-demented controls and individuals with MCI (Bakkour et a/., 2009; Dickerson et al., 2009, 2011).
Both hippocampal volume and regional cortical thickness have been related to the classical
'plaques and tangles' pathology and are therefore typically associated with AD neurodegeneration (Atiya et al., 2003; Kantarci and Jack, 2003; Dickerson et a!., 2009; Mormino et al., 2009; Becker et al., 2011). These features are considered classical as they are observed in the genetically-determined early-onset form of the disease, and are part of its well-known pathogenesis (Bateman et a!., 2012; Benzinger et al., 2013; Quiroz et a!., 2013). However, while
17 there has been great focus on 'plaques and tangles', there are many other changes that are observed in individuals with AD as described below, such as the increased prevalence of vascular co-morbidities and risk factors for vascular disease.
1.1.4 Vascular co-morbidities and risk factors
An increasingly recognized aspect of Alzheimer's disease is the plethora of vascular problems that seem to be more prevalent with the condition than in the general older adult population
(Kalaria, 2002; Gouw et al., 2011). Most notably, over 80% of individuals with Alzheimer's
disease have cerebral amyloid angiopathy, a condition in which there is a buildup of A140 in the
brain's blood vessels, a slightly different protein than the A342 that form the beta-amyloid
plaques (Thal et al., 2008). Atherosclerosis has been shown to be more prevalent in individuals
with AD than in non-demented controls (Hofman et al., 1997; Napoli and Palinski, 2005) and
more and worse stenoses were observed in individuals with AD, though to be the result of the
amyloid accumulation in the blood vessels (Kalback et al., 2004). In particular, the circle of
Willis, which is the interconnected vessel tree that feeds blood to our brains, suffers from
atherosclerosis in 77% of individuals with AD, compared to 47% of non-demented older adults
(Yarchoan et a/., 2012). Carotid ultrasonography also reveals lower cerebral blood flow and
longer cerebral circulation time in individuals with AD compared to controls (Puls et al., 1999;
Schreiber et a/., 2005).
In addition, several risk factors for vascular disease have been shown to increase the likelihood
of developing AD. For instance, obesity, high systolic blood pressure and high serum cholesterol
at midlife have all been associated with greater risk of developing mild cognitive impairment
18 and AD later in life (Kivipelto, E. Helkala, et aL., 2001; Kivipelto, E.-L. L. Helkala, et al., 2001;
Kivipelto et a/., 2005) or developing it earlier (Chuang et al., 2015). The greatest genetic predisposition to late-onset Alzheimer's disease is by far apolipoprotein E E4 (Saunders et al.,
1993; Bertram et al., 2007; Coon et al., 2007), which encodes a protein for cholesterol transport and may also be a genetic risk for cerebrovascular disease (McCarron et a!., 1999; Sudlow et al.,
2006; Bu, 2009). Common environmental risk factors for Alzheimer's disease include hypercholesterolemia, hypertension, hyperhomocysteinemia, diabetes, metabolic syndrome, smoking, systemic inflammation, increased fat intake and obesity (Casserly and Topol, 2004; Li et al., 2011), which are also all considered to be risk factors for vascular disease.
1.1.5 Current treatment and focus of therapeutic trials
Current treatment for Alzheimer's disease is limited to six options which temporarily increase the amount of neurotransmitters available in the brain (Alzheimer's Association, 2015b). None of the six drugs currently approved to treat Alzheimer's disease slows or stops the disease.
According to the Alzheimer's Association, the most recent drug approved in 2014 was the first to be approved in over a decade, and it was a combination of two drugs previously approved by the FDA. In comparison, 244 drugs were recently tested in clinical trials in a similar timeframe
(between 2002 and 2012).
Most of the recent pharmacological development and clinical trials have largely focused on immunotherapy or vaccination to clear amyloid from the brain of individuals with mild-to- moderate AD. Several Phase 11 clinical trials using second-generation anti-Ap antibodies have been conducted in the last few years, with pending results. In terms of passive immunotherapy,
19 bapineuzumab and solanezumab were the two leading candidates as anti-A@ monoclonal antibodies but several Phase III trials have failed to show clinical benefit (Doody et al., 2014;
Salloway et a/., 2014). Following these trials, the emphasis has shifted on using these immunotherapies earlier in the disease process to prevent it, for instance in individuals at risk for Alzheimer's disease (Panza et al., 2014).
1.2 White matter pathology in Alzheimer's disease
1.2.1 Increased prevalence and volume of white matter lesions
White matter pathology in Alzheimer's disease has been recognized as early as 1986 (Brun and
Englund, 1986). It was first described as affecting primarily the deep white matter and thought to be the result of hypoperfusion. Most importantly, it was first recognized as a feature that seems to be independent of the cortical pathology as it was observed in individuals without AD and was judged inconsistent with Wallerian degeneration.
These white matter lesions were also observed on magnetic resonance imaging (MRI) both in non-demented and demented older adults around the same time (Fazekas et al., 1987). Since then, they have been generally described as white matter hyperintensities of presumed vascular origin, due to their appearance on T 2-weighted and fluid-attenuated inversion recovery
(FLAIR) MRI and their association with vascular disease (Wardlaw et a/., 2013). They also appear hypointense on Ti-weighted MRI and have various synonymous names such as white matter signal abnormalities and white matter changes. Due to their hypodense appearance on CT, they have also been described as leukoaraiosis.
20 Many scales to characterize these lesions have been put forward, such as the Fazekas (Fazekas et al., 1987), Age-Related White Matter Changes (Wahlund et al., 2001) and Scheltens scales
(Scheltens et al., 1993), which evaluate lesions on MRI or CT with respect to their location
(periventricular versus deep or subcortical white matter) and severity, as per their extent and appearance (punctate versus confluent). However, it was found that they were all generally related to the total volume of white matter lesions (Kapeller et al., 2003; Gao et al., 2011), which is now the most studied feature.
White matter lesions are apparent in up to 95% of older adults and their prevalence and volume are known to increase exponentially with age (de Leeuw et ai., 2001; Jeerakathil et al.,
2004). Lesions have been associated with reduced cognitive abilities (Breteler et aL., 1994;
Tullberg et al., 2004; Burns et ai., 2005; Au et al., 2006; Smith et al., 2008; Lo et a/., 2012;
Marchant et al., 2012; Mortamais et ai., 2014) and have been shown to increase in volume over time faster in individuals with MCI or AD than non-demented older controls (Lo et al., 2012).
Consequently, several studies have shown that white matter lesion severity and volume is greater in Alzheimer's disease than in non-demented older controls, especially in the periventricular regions (Fazekas et al., 1987, 1996; Rezek et al., 1987; Hogervorst and Ribeiro,
2002; Capizzano et al., 2004; Burns et al., 2005; Yoshita et al., 2006; Holland et al., 2008; He et al., 2010; Polvikoski et al., 2010; Gao et a/., 2011).
The greater prevalence and proliferation of white matter lesions in AD currently lacks a solid pathophysiological explanation. Some researchers have suggested their greater prevalence might be a consequence of neurodegeneration and neuronal death in the disease. However,
21 growing evidence as detailed below suggests they might be the result of greater vascular disease in AD than is normally seen in non-demented controls. It is possible that the individuals with greater vascular disease are the ones to get a diagnosis of AD for the same pathology load of 'plaques and tangles'. Similarly, it is also possible that vascular health deteriorates after an individual becomes demented and cannot take care of themselves. These notions of circularity and causality are to be kept in mind while interpreting the results of this thesis.
1.2.2 Histological correlates of white matter lesions
White matter lesions have been associated with small-vessel changes and they have been sometimes described as incomplete infarcts (Brun and Englund, 1986). Specifically, white matter lesion severity and volume in the non-demented elderly have been correlated with increasing severity of ischemic tissue damage, which is described by a widening of the perivascular spaces (Virchow-Robin spaces), multiple small cavitations or lacunar infarctions, vascular malformations and leakage, activated macrophages, venous collagenosis, arteriolosclerosis, astrogliosis, demyelination and variable loss of fibers and oligodendrocytes
(Fazekas et al., 1993; Moody and Brown, 1997; Gouw et al., 2011; Schmidt et al., 2011). White matter lesions have also been found to have increased hypoxia-regulated proteins in a sample of donated brains from both demented and non-demented elderly individuals, supporting the idea that they are hypoxic and do not receive enough blood supply (Fernando et al., 2006).
More specifically, deep white matter lesions were associated with cerebral infarcts, capillary endothelial activation, microglial activation and arteriolar sclerosis - greater vessel wall
22 thickness and perivascular enlargement - compared to normal white matter (Fernando et al.,
2006; Polvikoski et al., 2010).
A distinction has been made for periventricular caps and lesions lining the ventricles, which
have been shown to involve demyelination related to disruption and denudation of the
ventricular lining - also called ependyma - and subependymal gliosis in both non-demented
elderly and individuals with AD (Fazekas et al., 1993; Fernando et al., 2006; Schmidt et a/.,
2011). It has been suggested that periventricular areas have greater water content, which may
come from the venous network, and may also originate following disruption of the blood-brain
barrier (Young et al., 2008; Gouw et al., 2011; Schmidt et al., 2011).
Relatively few comparative pathological studies have been conducted to differentiate white
matter lesions in individuals with AD and non-demented controls. Most of them have showed
increased demyelination and axonal loss, more severe gliosis and denudation of the ventricular
ependyma and thicker arterial adventitia - the outermost layer of arteries - in the white matter
lesions of individuals with AD compared to those of non-demented controls (Scheltens et al.,
1995; Gouw et al., 2011). Another distinction was the presence of microglial activation in the
lesions of AD (Gouw et al., 2008). Markers of Wallerian degeneration such as lipid-laden
macrophages have not been observed in white matter lesions of individuals with AD, arguing
against the idea that lesions are secondary to neurodegeneration (Pantoni and Garcia, 1997).
Overall, these studies suggest similar pathologies contribute to the histological profile observed
in lesions of individuals with AD and non-demented controls, though these pathologies may be
more severe in individuals with AD. In the current work, we look at imaging markers of tissue
23 microstructure to examine whether any qualitative or quantitative differences are observed in the white matter lesions of non-demented controls and individuals with MCI and AD, and whether they are related to imaging measures typically associated with neurodegeneration.
1.2.3 Epidemiological associations with risk factors for vascular disease
Epidemiological studies have also shown that greater volume and higher severity of white
matter lesions are associated with hypertension (Jeerakathil et al., 2004; Burns et al., 2005;
Murray et at., 2005; Stenset et al., 2006; Gottesman et at., 2010; Rostrup et a/., 2012; Brisset et
al., 2013), hyperhomocysteinemia (Hogervorst and Ribeiro, 2002; Stenset et al., 2006), history
of stroke or cardiovascular disease (Breteler eta/., 1994; Jeerakathil et al., 2004), glycated
hemoglobin level (Murray et al., 2005), body mass index and diabetes (Brisset et al., 2013),
smoking (Jeerakathil et al., 2004; Rostrup et al., 2012), dietary salt intake (Heye et al., 2015),
left ventricular hypertrophy (Jeerakathil et al., 2004), systemic markers of inflammation
(Fornage et al., 2008), arterial plaque, stiffness and endothelial dysfunction (Kearney-Schwartz
et al., 2009; Brisset et al., 2013), as well as other vascular risk factors (Breteler et al., 1994;
Murray et al., 2005; Yoshita et al., 2006), with several but not all of these associations being
independent from each other (Brisset et al., 2013). Deep or subcortical white matter lesions
have been further correlated with low-density lipoprotein (LDL), cholesterol , apolipoprotein E
E4 genotype (de Leeuw et al., 2004) and several pulmonary function measures (Murray et al.,
2005). Overall, risk factors for developing white matter lesions mirror the environmental risk
factors for developing both vascular disease and Alzheimer's disease.
24 1.2.4 Evidence of poor perfusion in white matter lesions
In addition to epidemiological associations with vascular risk factors, white matter lesions have also been directly shown to have lower perfusion than normal-appearing white matter using single photon emission computed tomography (Makedonov et al., 2013) and arterial spin labeling with MRI (Brickman et al., 2009). Lower perfusion and longer mean transit time were also observed in white matter lesions compared to normal-appearing white matter using dynamic contrast-enhanced MRI (Marstrand, 2002). Lower cerebrovascular reactivity was found in white matter lesions in the same study, i.e. a lower capacity to increase perfusion following a pharmacological challenge, though the increase in perfusion relative to baseline was similar to normal-appearing white matter and gray matter. White matter regions converting to lesions
have also been shown to have lower baseline perfusion than the rest of the white matter, suggesting that perfusion can predict future lesions (Promjunyakul et al., 2015).
It is unclear whether the lower perfusion observed in white matter lesions is due to the fact
lesions are present in regions that have relatively low normal cerebral perfusion compared to
the rest of the brain, as evidenced by the fact lesion frequency proportionally increases in
regions with lower normative perfusion (Holland et al., 2008). However, it remains that white
matter lesion volume is inversely proportional to its average perfusion as well as global
(Makedonov et al., 2013; Benedictus et al., 2014), regional (Crane et al., 2015) and hippocampal
perfusion (Waldemar et al., 1994), and individuals with larger confluent white matter lesions
have lower global cerebral blood flow than those with smaller lesions (Bastos-Leite et al., 2008),
suggesting they might be ischemic.
25 1.2.5 White matter atrophy and ventricular enlargement
One salient feature of the disease that oftentimes serves as an imaging hallmark but has been generally ignored from a pathophysiological perspective is the progressive enlargement of the cerebral ventricles with the disease (Carmichael et al., 2010). It is widely believed that ventricular enlargement is the consequence of brain atrophy, and more specifically of white matter atrophy, which naturally creates a space that can only be filled by cerebrospinal fluid.
Total white matter volume is known to decrease with age and is also lower in individuals with
AD compared to non-demented controls, especially in white matter regions proximal to cortical pathology, such as the parahippocampal white matter (Bigler et al., 2000; Salat et al., 2009).
However, there is no clear model of how white matter gets replaced by enlarging ventricles.
Ventricular enlargement is unlikely to be the result of increased pressure from the ventricular fluid on the parenchyma, as individuals with more advanced Alzheimer's disease might in fact have lower cerebrospinal fluid pressure than the normal population (Silverberg et Qi., 2006;
Wostyn et a/., 2009). Some early studies have also shown an association between ventricular enlargement and periventricular white matter lesion volume and/or thickness (Waldemar et al.,
1994; Fazekas et al., 1996), suggesting the possibility that white matter may progress into lesion before disintegrating and giving way to enlarged ventricles. A new body of research is suggesting that disruption of the blood-brain barrier and associated water dysregulation might play a role in ventricular enlargement, with the involvement of water channel regulators - aquaporins - in the periventricular white matter (Moftakhar et al., 2010; Anderson et al., 2011;
Taheri et al., 2011; Brinker et al., 2014). Overall, little progress has been made to understand the enlargement of the ventricles and more studies are needed to successfully integrate the
26 role of ventricular enlargement in the pathophysiology of AD. In the current work, we look at
both the tissue properties and volume of white matter lesions and whether they are related to total white matter volume and ventricular enlargement.
1.2.6 Are the white matter findings consistent with the classical features of the
disease?
Some studies have shown a correlation between greater white matter lesion volume and gray
matter atrophy and pathology in individuals with MCI and AD (Capizzano et al., 2004; Moghekar
et al., 2012; Fujishima et al., 2014), though the inverse relationship has also been found in MCI
(Jacobs et al., 2014). In non-demented individuals, the association between white matter lesion
load and brain atrophy has also sporadically been found (DeCarli et al., 1995; Wen et al., 2006),
though the correlation has been explained by cardiovascular risk factors in one study (Jouvent
et al., 2010). It is unclear whether the correlation in individuals with MCI and AD can also be
explained by cardiovascular risk factors, or whether the co-occurrence of both lower cortical
thickness and higher white matter lesion volume may lead to the correlation. White matter
lesion volume has also been shown to correlate with plasma concentrations of Ar40, the
amyloid observed in blood vessels as part of cerebral amyloid angiopathy, rather than A042,
the amyloid that makes up the plaques in AD, in MCI and AD (Gurol et al., 2006). White matter
lesions have been associated with Braak staging and neuritic plaque load (Polvikoski et al.,
2010; McAleese et al., 2015). However, the relationship between amyloid plaques and white
matter lesions was explained by age in an autopsy study of non-demented individuals (Rutten-
Jacobs et a/., 2011). Apart from these studies, several studies have shown that no significant
27 relationship between white matter lesion burden and AP is generally found (Hedden et al.,
2012; Lo et al., 2012; Marchant et al., 2012, 2013; Mortamais et al., 2014). Overall, while some studies have tried to link white matter lesions and markers of AD neurodegeneration, more
studies are needed to investigate whether they are part of the same disease process.
Furthermore, the evidence presented here as to the histological and epidemiological
underpinnings of white matter lesions highly suggests a vascular-driven pathway for these
lesions rather than one that is subsequent to neurodegeneration. These notions also suggest
that vascular disease and white matter pathology may in fact precede neurodegeneration and
potentially explain these associations in a different light than is generally assumed (Kalaria,
2002). In the current work, we investigate the relationship between tissue properties and
volume of lesions and imaging markers typically associated with neurodegeneration, such as
hippocampal volume and regional cortical thickness.
1.3 Motivation for studying white matter in Aizheimer's disease
1.3.1 Heterogeneity of the disease
The diagnosis of Alzheimer's disease is currently determined based on the presence of the
'plaques and tangles' pathology, which guides models of the pathophysiology of the disease
(Mirra et al., 1991; Hyman and Trojanowski, 1997; Jr et al., 2010; Hyman et al., 2012). However,
this first chapter has so far described a diversity of other changes that occur throughout the
disease, in particular when it comes down to the involvement of white matter pathology and its
association with vascular disease. Despite this, there remains a quest to define 'pure'
Alzheimer's disease and individuals with vascular co-morbidities or excessive burden of white
28 matter lesions are often excluded from clinical studies given the occurrence of these pathologies in more pure forms of vascular dementia (Kalaria, 2002; McKhann et al., 2011;
Hyman et al., 2012), leading to the idea that lesions are not part of the basic pathology of AD.
This quest has a long history, with many attempts to properly subdivide Alzheimer's disease into different subtypes, such as the early distinction of pre-senile AD and senile dementia, which were based on age of onset (below and above 65 years old) and have since been mostly reclassified as early-onset familial Alzheimer' disease and late-onset Alzheimer's disease
(Heston et al., 1981; Davies, 1986; Bertram et al., 2010). While the early distinction based on age remains, these two subtypes are now mostly distinguished by whether the disease is directly inherited from a parent in an autosomal dominant fashion. There have also been attempts at subdividing the prodromal phase of AD, mild cognitive impairment (MCI), in amnestic and non-amnestic forms, based on whether memory is the primary cognitive deficit
(Petersen, 2004; Gauthier et al., 2006; Petersen et al., 2009), though this distinction was not part of the most recent diagnostic guidelines for MCI (Albert et al., 2011). Despite this, large biological heterogeneity remains even in its subtypes given that it is mostly defined using cognitive testing (Nettiksimmons et al., 2014).
However, neuroimaging markers have been shown to be useful to predict who will convert to
AD despite all individuals having a similar cognitive score in MCI (Nettiksimmons et al., 2014).
Increasingly, MRI and positron emission tomography (PET) have been used to study the pathogenesis of Alzheimer's disease, but also better define its biological heterogeneity. For instance, new PET markers have confirmed and expanded the idea that the accumulation of AD occurs in non-demented older adults (Thal et a/., 2002; Wolk and Klunk, 2009). Indeed,
29 individuals considered generally cognitive intact have shown large amount of AP in their brain; however, it remains to be seen whether those individuals will develop AD or more likely may be at increased risk of developing AD (Pike et al., 2007). This notion underlies the need to reconsider the involvement of vascular and white matter pathologies, dismissed earlier due to their non-specificity to AD, at a time when amyloid pathology is the main focus of therapeutic trials.
1.3.2 Possibility of a distinct concurrent disease mechanism
Vascular co-morbidities have generally not been considered as part of the disease process but rather as an additional burden on the brain (Bowler, 2005). Furthermore, despite the fact that vascular risk factors associated with white matter lesions correspond to the same risk factors associated with the development of Alzheimer's disease, neither those vascular risk factors or white matter lesions have been integrated in the generally-acknowledged pathophysiology of
Alzheimer's disease and are still considered to be an extraneous burden to the disease.
However, an increasingly large group of individuals, mostly from the vascular disease community, have started investigating Alzheimer's disease as a primarily vascular disorder
originating in the blood vessels rather than a neurodegenerative disorder originating in the
neurons and involving the 'plaques and tangles' pathology (Torre, 2002; Mazza et al., 2011;
Kalaria et al., 2012). Indeed, while hippocampal volume is considered one of the most
predictive imaging biomarkers of the disease (Jack Jr et al., 1999; DeCarli et al., 2004; Kantarci
et a/., 2009), it is interesting to note that it is equally affected in vascular dementia (Laakso et
a/., 1996; Fein et al., 2000; Du et al., 2002), and that there is evidence that untreated
30 hypertension may lead to a greater reduction in hippocampal volume in non-demented elderly
(den Heijer et al., 2005). Furthermore, white matter lesions have been shown to be similar in individuals with AD and individuals with vascular dementia, suggesting a common pathogenesis and the existence of a pathological continuum (Gouw et al., 2011), and studies have shown that their volume and quality may also be predictive of conversion to AD (Brickman et al., 2012;
Mortamais et al., 2014; Lindemer et al., 2015). This broad spectrum of cognitive disorders
associated with vascular injury has been termed 'vascular cognitive impairment', to recognize
that cognitive impairment in later life often represents a mixture of Alzheimer's disease
pathology and microvascular damage (O'Brien et al., 2003; Bowler, 2005; Selnes and Vinters,
2006; Boss et al., 2010; Gorelick et al., 2011). Others have suggested that disruption in the
blood-brain barrier through AP accumulation in the vessels or otherwise might lead to the
reduction in cerebral blood flow and ischemia that is observed in white matter lesions (Shi et
al., 2000; Kalaria, 2002; Smith and Greenberg, 2009; Zlokovic, 2011).
Overall, the cause of Alzheimer's disease has remained unclear for decades despite the
astounding amount of research, and it has become unlikely that a single type of pathology is
responsible. An incredible diversity of models of the disease has been suggested, with evidence
to back each model. One of the models that has received recent attention is the 'two-hit'
hypothesis, suggesting that the combination and co-occurrence of two types of pathology
might lead to the clinical manifestation of the disease (Zhu et al., 2004). An obvious example in
the context of this thesis which has been suggested before (Zlokovic, 2011; Provenzano et al.,
2013) is the combination of AP accumulation with a significant vascular insult, which may or
may not be made evident through the observation of white matter lesions or other imaging
31 markers. While individuals with clinically-manifested AD might not show evidence of any vascular pathology, it may still be possible that such an insult might not be observable through current imaging means.
A few observations have been made suggesting that multiple disease mechanisms may contribute towards the clinical manifestation of AD. For instance, white matter lesions and cortical gray matter loss have been shown to have independent associations with MMSE and other cognitive measures in individuals with probable AD (Stout et al., 1996). Lower brain volume and greater white matter lesion volume have also been independently associated with reduced global and cortical cerebral blood flow in individuals with AD (Benedictus et al., 2014).
This evidence seems to make the case that white matter lesions are a separate type of damage than the classical neurodegeneration observed in AD and further study is needed in this regard.
In the current work, we investigate the possible existence of these two distinct disease
processes by using factor analysis on the volume of white matt r is d furLther
common imaging markers affected in AD, and by relating the resulting factors to global white
matter damage, cerebral blood flow and longitudinal cognitive decline.
1.3.3 Potential existing avenues for delaying disease onset and for treatment
The recognition and demonstration of a disease pathway involving vascular and white matter
pathology that is distinct from the classical 'plaques and tangles' pathology would help further
the current efforts to prevent and treat Alzheimer's disease. Indeed, many already-existing
therapies for vascular disease could be applied in individuals with Alzheimer's disease or pre-
clinical Alzheimer's disease were such a pathway clearly demonstrated, and were it shown that
32 preventing its progression also delays the clinical manifestation of the disease. Such existing drugs that have been proposed to have potential therapeutic benefit in AD with respect to reducing the vascular risk factors include angiotensin converting enzyme inhibitors, angiotensin
I blockers, peroxisomal proliferator activating receptor agonists, acyl Co-A cholesterol acyl transferase inhibitors, statins, aspirin, nonsteroidal anti-inflammatory drugs (NSAIDS), cyclo- oxygenase 2 inhibitors and thienopyridines, with the goal of acting on cholesterol homeostasis, anti-inflammatory properties, antiangiogenic properties and AD effects (Gorelick, 2004).
Individuals studies and meta-analyses have shown that statin use, antihypertensive medications, NSAIDS and estrogen actually reduce your risk of developing AD (Forette et al.,
1998; Xu et a/., 2015). Treatment of all vascular risk factors also reduced the risk of developing
AD in an MCI population, compared to only treatment of some (Li et al., 2011). The Evaluation of Vascular Care in Alzheimer's Disease (EVA) Study has also shown slower progression of white matter lesions with vascular care compared to standard care in individuals with AD in a randomized controlled clinical trial (Richard et al., 2010). However, this did not reduce global cortical or medial temporal lobe atrophy, which strengthens the notion that it might be part of a distinct disease pathway. In addition to drugs, a healthy dietary pattern, such as the
Mediterranean diet, has been shown to reduce the risk of developing AD, as well as fish consumption, folate, antioxidants (vitamin E and C) and coffee, which have also been shown to reduce the risk of vascular disease (Sofi et ai., 2010; Xu et al., 2015). The neuroprotective mechanism of a healthy diet seems to arise from improved cerebral blood flow, as demonstrated in genetically-modified mice with AD (Zerbi et al., 2014), and may also be independent from the 'plaques and tangles' pathology.
33 In order to assess and test the potential in treating and preventing AD using these avenues, there is a need to develop reliable indicators of each suggested disease component. While clinical tests have generally been used to track progress in clinical trials, and have the obvious advantage of measuring clinical function, they ultimately cannot distinguish between distinct pathologies affecting the same cognitive domains. The most useful data would be pathological in nature, but it is currently highly impractical to collect such data on a large number of subjects to determine the usefulness of a treatment or prevention strategy. Neuroimaging markers are the most practical option currently available as they can be acquired in-vivo, repeatedly over time and for almost any individual. As such, currently, hippocampal volume and cortical thickness of regions affected in early AD are typically considered good markers of classical AD
neurodegeneration. Recent advances in the field have also enabled the measurement of
amyloid-beta and tau deposition in the brain using positron emission tomography, which may
provide neuroimaging markers more representative of the 'plaques and tangles' pathology.
However, their use will be the focus of future work. On the other hand, white matter lesions, in
particular, represent one of the most useful neuroimaging markers of vascular and white
matter pathologies. White matter atrophy and ventricular enlargement are also representative,
though less studied and understood. While these features are largely recognized, their
underpinnings are often inferred to be secondary to the extensive neurodegenerative cortical
pathology but may represent a distinct disease mechanism towards the clinical manifestation of
AD.
34 1.4 Summary of thesis aims
The overarching goal of this thesis is to examine whether white matter lesions are linked to the neuroimaging markers typically associated with neurodegeneration in Alzheimer's disease (AD) and whether they independently lead to cognitive decline. In the first aim, we investigated a
Chapter 2 Chapter 3 Chapter 4 Chapter 5
Figure 1.1 Overview of thesis aims and breakdown into novel MRI method measuring the diffusion of water molecules as a sensitive tool to measure white matter microstructure in-vivo in a sample of healthy older adults. In the second aim, these sensitive measures of white matter microstructure were used to determine that the
microstructural properties of white matter lesions slightly differ between AD and non-
demented controls, and that those differences are most strongly related to ventricular volume,
not hippocampal volume and regional cortical thickness which are typically associated with AD
neurodegeneration. In the third aim, we found that the volume of white matter lesions was
strongly related to other common imaging markers of pathology in AD, such as hippocampal
volume, regional cortical thickness, ventricular volume and total white matter volume, and we
used factor analysis on these markers to determine that two underlying, distinct disease
processes seem to be responsible for this covariation. One factor related to both volume and
diffusion properties of white matter lesions, as well as both ventricular and hippocampal
volume, and was interpreted to be 'age- and vascular-associated', while the other factor was
35 related to imaging markers typically associated with neurodegeneration. Lastly, the fourth aim showed that decline in the 'age- and vascular-associated' factor longitudinally is linked to cognitive decline, independently from decline in the 'neurodegenerative' factor, demonstrating the potential added therapeutic benefit of targeting this disease pathway involving white
matter lesions that is distinct from the classical neurodegenerative component of the disease.
36 37 Chapter 2: Investigation of diffusion tensor
imaging as a sensitive tool to measure
white matter microstructure
2.1 Overview
This chapter focuses on the use of different diffusion-weighted magnetic resonance imaging
methods to explore their sensitivity to measure properties of the white matter
microenvironment. These diffusion-weighted methods in recent years have enabled the in vivo
study of tissue microstructure, in particular in the white matter, an ability that used to be only
possible through post-mortem pathology. The noninvasive nature of MRI has also enabled its
use in the study of large populations. While used here in the context of aging as a preface to
our studies in individuals with MCI and AD, it serves an informational purpose and allows us to
describe in detail the methods used later in the thesis. In addition, an advanced model of
diffusion-weighted imaging is presented, diffusion kurtosis imaging, and its advantages and
inconveniences are described in relationship with the more simple and more commonly-used
model, diffusion tensor imaging. Group differences using diffusion tensor imaging between
non-demented controls and individuals with MCI and AD are also presented. If already familiar
with these diffusion methods and their known use in generally healthy older adults and
individuals with MCI and AD, we invite the reader to skip ahead to Chapter 3. 38 2.2 Introduction
Age-associated white matter degeneration has been well-documented and is likely an
important mechanism contributing to cognitive decline in older adults and individuals with MCI
and AD as described in Chapter 1. Recent work has explored a range of non-invasive
neuroimaging procedures to differentially highlight alterations in the tissue microenvironment.
Diffusion-weighted magnetic resonance imaging has been applied across several studies as a
means to investigate the microstructural properties of white matter as well as changes due to
age and disease (Basser et al., 1994; Basser and Pierpaoli, 1996; Pfefferbaum et al., 2000; Le
Bihan et a/., 2001; Abe et al., 2002; Salat et a/., 2005). The utility of techniques such as diffusion
tensor imaging (DTI) comes from the fact that a range of microstructural properties can be
obtained from a standard acquisition and that different metrics show differential sensitivity to
effects in group comparisons. The basis of the technique relies on the sensitization of water
molecules to a specific frequency based on position, followed by a time period during which
water molecules are allowed to diffuse. An image is then acquired after this time period, and
the signal recorded diminishes in function of how much the water molecules moved since the
sensitization. The signal attenuation can then be related to the diffusivity of the water
molecules in every voxel of the image using the following formula and by making the
assumption that the water molecules diffuse according to a Gaussian distribution:
S(b) ~ S(b = 0) * exp(-bD)
where b is a variable related to acquisition parameters including the length of the time period
during which water molecules are allowed to diffuse [s/mm 2 ], S(b) is the signal recorded after
39 this period, S(b = 0) is the signal recorded without any diffusion-weighting (not influenced by the diffusion of water molecules) and D is the diffusivity we seek to estimate [mm 2 /s]. This method only allows sensitization to diffusion for one direction at a time, and therefore provides the diffusivity in only one direction for each acquisition. A common scheme is therefore to repeat the acquisition for several different orientations and diffusivity in any given direction can then be obtained following the estimation of the diffusion tensor coefficients in each voxel.
Through eigenvalue analysis, this tensor can be rotated for each voxel in a frame of reference that is more intuitive for biological interpretation. This provides us with: the axial diffusivity
(DA) which is the maximum diffusivity over all directions (first eigenvalue) and is thought to
correspond to the direction of the axons in a straight fiber bundle; the radial diffusivity (DR)
which is the average diffusivity perpendicular to the maximum diffusivity (average of the
second and third eigenvalues) and is thought to represent the diffusivity through the myelin in
a straight single fiber bundle; the overall mean diffusivity (MD); as well as the fractional
anisotropy, representing the degree of anisotropy of water diffusion and described by the
following equation:
FA =
where X1, and are the three eigenvalues of the diffusion tensor. More information can be
found in reviews of diffusion tensor imaging (Le Bihan et al., 2001; Basser and Jones, 2002;
Dong et al., 2004). Fractional anisotropy (FA) and mean diffusivity (MD) are two common
metrics calculated in DTI studies that are sensitive to changes with development and disease
(Le Bihan, 2003; Dong et al., 2004; Sundgren et aL., 2004). FA is a measure of the directional
40 dominance of water diffusion in tissue and has been loosely interpreted as an indirect quantitative metric of the density of nerve fibers and their myelin sheaths (Beaulieu, 2002;
Moseley, 2002). In contrast, MD is a measure of the overall (direction-independent) degree of water diffusion within the tissue, and has been utilized as an important marker of ischemia, edema, and cell death (Chenevert et al., 2000; Sotak, 2002). In the context of healthy aging, decreases in FA and increases in MD have been reported throughout much of the cerebral white matter (Pfefferbaum et al., 2000; Salat et al., 2005). More recently, the directional components of diffusivity, such as axial (DA) and radial (DR) diffusivity have been shown to have spatially specific and differential sensitivities to the effects of aging (Madden et al., 2009;
Bennett et al., 2010). The idea that different pathologies affect specific diffusional properties preferentially (Song et al., 2002, 2003) has potential value in the diagnosis and tracking of specific disease processes as opposed to more generic tracking of cumulative white matter damage without specific etiology. To date, however, little if any work has attempted to differentiate between various types of age-associated white matter changes based on multivariate diffusion properties.
Although DTI provides multiple indices of diffusional behavior, it is possible that composite information across a wider range of diffusional processes beyond what DTI provides would enable better classification of differing patterns of white matter damage with aging and disease. The physical model widely used to extract DTI parameters assumes that water molecules diffuse according to a Gaussian distribution, which corresponds to free, unrestricted diffusion in a homogenous environment. However, given the structural complexity of neural tissue, this assumption can only be an approximation, and appreciable differences in diffusivity
41 are expected among tissue compartments (e.g. intra- and extra-cellular) within a same volume element. In theory, these differences are better characterized using higher-order diffusions statistics. By acquiring images with multiple diffusion weightings (b-values), thus allowing for a better estimation of the water molecules' displacement distribution, the excess kurtosis of the distribution can be calculated, which is a unitless index of its non-Gaussianity (Liu et al., 2004):
1 S(b) ~ S(b = 0) * exp(-bD + b 2 D 2K) 6 where K is the excess diffusional kurtosis (shortened to simply kurtosis in the remainder of the text) and the other variables have been defined in the diffusion tensor model equation above.
The b values used in the acquisition also can be higher than what was possible with diffusion tensor imaging. Indeed, as demonstrated in Figure 2 from (Jensen and Helpern, 2010), the diffusion-weighted signal intensity can be approximated by DTI at b values around and below
1000 s/mm 2, while DKI provides a better approximation at higher b values up to about 3000 s/mm 2
Diffusional kurtosis imaging (DKI) therefore provides a novel set of in vivo microstructural
properties that describe tissue microstructure beyond the scope of traditional DTI (De Santis et
al., 2011); these properties are quantified through the mean, axial and radial diffusional .
kurtoses (MK, KA and KR, respectively). Analogous to diffusivity, diffusional kurtosis also varies
depending on the direction of diffusion weighting. KA and KR represent respectively the
diffusional kurtosis in the principal diffusion direction and the diffusional kurtosis averaged over
its perpendicular directions, based on the diffusion tensor orientation, while MK represents the
overall average diffusional kurtosis.
42 0- Figure 2 from (Jensen and Helpern, 2010) Comparison of DTI and DKI fitting models. For DTI, the logarithm of diffusion-weighted signal intensity (circles) as a function of the b-value is fit, for small b- values, to a straight line. In brain, this fit is often based on the signal for b = 0 and b = 1000s/mm 2 . For DKI, the logarithm of the signal intensity is fit, for - - DTI fit small b-values, to a parabola. In brain, this fit may be DKI fit based on the signal for b = 0, b = 1000, and b = 2000 41 1 s/mm 2 . Copyright @ 2010 John Wiley & Sons, Ltd. through license 0 5000 Permission to reproduce granted b (s/mm2) signed by Massachusetts Institute of Technology.
DKI was developed with the goal of characterizing the diffusional heterogeneity arising from multiple tissue compartments with different diffusivities (Jensen et al., 2005). Several multi- compartment models have been proposed to describe the biophysical and biological nature of diffusional kurtosis (Jensen et al., 2005; Jensen and Helpern, 2010), particularly in the white matter (Fieremans et a/., 2011; De Santis et al., 2012). For instance, assuming a multi- compartment model as described in (Jensen et al., 2005) and later refined by (Fieremans et al.,
2011) for two compartments in white matter, one intracellular and one extracellular, we obtain the following relationships for any given diffusion-weighted direction:
2 3 * var(D) 3f(1 - f)(De - Da)
D2 D2 where var(D) represents the variance of diffusivity across compartments, f represents the axonal water fraction and De and Da represent the diffusivity of the extracellular and intracellular compartments of the white matter model, respectively. While noting that they make several assumptions, these models have confirmed the idea that diffusional kurtosis represents an index of diffusional heterogeneity, defined by how variable the diffusivity index
43 varies across different cellular compartments within a voxel and by the diversity of compartments that may be present. For instance, any voxel with a predominant compartment will tend to have lower kurtosis than a voxel with multiple different compartments with various diffusivity values. Importantly, this property of diffusional heterogeneity cannot be measured with diffusion tensor imaging which only provides the average diffusivity within a given voxel.
Quantitative measures from DKI may be sensitive to developmental or disease-associated
conditions in which there is a differential alteration in diffusion and permeability properties
across cellular compartments. For instance, MK is known to vary with developmental stage in the rat (Cheung et al., 2009; Blockx et al., 2012) and human brain (Falangola et al., 2008;
Helpern et al., 2011; Latt et al., 2013), suggesting a maturational increase and subsequent
decline in white matter integrity during aging. These prior studies demonstrated coarse
changes in MK with development and aging and suggested that DKI metrics may be sensitive to
suule MiLrUsLructural Cnanges related to age.
As published in (Coutu et al., 2014), the major goals of this study were twofold: first, to examine the regional age trajectories of white-matter microstructural alterations observed through DKI
metrics in a large cross-sectional sample of generally healthy adults, and second, to determine
whether DKI provides additional, unique information compared to DTI for studying healthy
aging. In the context of this thesis, we present the relevant parts, in particular the effect of age
on all diffusion measures in white matter and the advantages and inconveniences of using
diffusion kurtosis imaging in comparison to diffusion tensor imaging in the study of white
matter properties, which will be useful for the next chapters. In addition to our publication, we
44 further present and discuss group differences in DTI measures across the white matter between individuals with MCI/AD and non-demented older controls, and compare them to the age effects.
2.3 Materials and methods
2.3.1 Participants and MRI acquisition
Two distinct datasets were used in this chapter. The first dataset included a total of 111 healthy
adults between 33 and 91 years of age (62 women, 49 men) recruited through the
Massachusetts General Hospital (MGH) and local community. This sample included healthy
individuals as well as older adults with some mild forms of vascular risk, including hypertension,
hyperlipidemia, hypercholesterolemia and diabetes. Individuals were excluded for signs of
major neurologic or psychiatric illness including dementia, high cerebrovascular disease risk or
overt disease (large vessel stroke or hemorrhage), cancer of the central nervous system, major
head trauma, and/or other neurological or psychiatric, or therapeutic conditions that may
influence cognition or imaging measures. Participants were non-demented, as assessed by a
minimum score of 24 on the Mini Mental State Exam (MMSE) (Folstein et al. 1975).
Characteristics of this first group are provided in Table 1. All participants gave informed consent
and the study protocol was approved by the Massachusetts General Hospital/Partners
Healthcare Institutional Review Board. All participants were imaged on a Siemens 3T Trio
system (Erlangen, Germany) with a 32-channel head coil. Whole-brain diffusion-weighted scans
were acquired (TR = 9250 ms, TE = 103 ins, slice thickness = 2 mm isotropic, 64 slices total,
acquisition matrix 128 x 128 (FOV = 256 mm x 256 mm), 6/8 partial Fourier, bandwidth = 1396
45 Hz/pixel, 24 non-collinear directions with b-values of 700, 1400 and 2100 s/mm 2 , single
2 average, and 10 T 2-weighted (bo) images with b-value = 0 s/mm ). The b-values were chosen to be optimal for both DTI and DKI analysis (DKI requires b-value < 2500 mm/s 2 ) and the number of directions to accommodate scan time while ensuring proper estimation of the model. The
DKI acquisition sequence used a twice-refocused balanced spin echo to reduce eddy current distortions (Reese et al. 2003). Head motion was minimized using tightly padded clamps attached to the head coil. Oblique axial slices were acquired with total scan duration of 13 min
8 s. The mean signal-to-noise ratio (SNR) at the highest b-value (2100 mm/s 2 ) across individuals was 37, obtained in the middle-brain slice by dividing the global-mean brain signal by the standard deviation of the background noise.
Table 2.1 Demographics for all participants of the first dataset. Standard error of the mean is shown in parentheses.
Participants Age Education MMSE [-]a Translation Rotation motion
(female) [years] [years] motion [mm] [millidegrees]
111 (62) 60.53 (1.09) 16.38 (0.26) 28.65 (0.14) 1.01 (0.02) 4.59 (0.18)
Standard error of the mean is shown in parentheses. a) Information not available for the youngest participant. (MMSE: Mini-Mental State Exam).
The second dataset was obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI, http://adni.loni.usc.edu) and included 74 controls, 97 participants with MCI and 48 participants with AD who underwent whole-brain MRI scanning at one or multiple visits on a 3-Tesla GE
Medical Systems scanner and had sagittal Ti-weighted 3D spoiled gradient echo images and
46 diffusion-weighted images (b = 1000 s/mm2, 41 directions) available at the time of download.
These datasets were acquired using previously described ADNI Core MRI and DTI protocols
(Jack et al., 2008). Four participants (one control, one with MCI and two with AD) were excluded because of extensive white matter damage or ventricular enlargement. Group designation of control, MCI and probable AD was determined by ADNI based on the criteria of the National Institute of Neurological and Communicative Diseases and Stroke - Alzheimer's
Disease and Related Disorders Association (McKhann et al., 1984). Participants enrolled as normal or with significant memory concern and with a Clinical Dementia Rating (Morris, 1993) of 0 were grouped together into the control group, and participants enrolled as early and late
MCI were combined into one MCI group (see ADNI 2 Procedures Manual on www.adni-info.org for more information). Clinical profiles and diagnostic information were obtained from the assessment closest in time to the MRI acquisition. Demographics for this second dataset are provided in Table 2. Written informed consent was obtained from all participants or their representatives through ADNI. The study procedures were approved by institutional review boards of all participating institutions.
Table 2.2 Demographics for all participants of the second dataset.
CN MCI AD p-value
Participants (female) 73 (46) 96 (36) 46 (17) 0.0016
Age [years] 72.98 (0.84) 73.80 (0.73) 74.58 (1.06) 0.4900
Education [years] 16.30 (0.32) 16.10 (0.28) 15.24 (0.41) 0.1086
M MSE [-]a 28.67 (0.25) 27.88 (0.22) 23.14 (0.32) <0.0001
47 Translation motion [mm] 1.32 (0.07) 1.34 (0.06) 1.35 (0.09) 0.9671
Rotation motion [millidegrees] 6.2 (0.3) 6.3 (0.3) 6.8 (0.4) 0.5621
Standard error of the mean is shown in parentheses. a) Information missing for 16 CN, 20 MCI and 11 AD. (MMSE: Mini-Mental State Exam).
2.3.2 Preprocessing
Data were processed using a combination of in-house image processing tools developed in
MATLAB (Mathworks, Natick, USA) and tools available as part of Freesurfer
(http://surfer.nmr.mgh.harvard.edu) and FSL (http://www.fmrib.ox.ac.uk/fsl). The diffusion datasets were corrected for any potential 3D head motion and eddy current distortion using FSL eddycorrect, and translation and rotation motion estimates were obtained from the registration matrices (Yendiki et a/., 2014). For individuals with multiple available datasets, we picked the one with the least average translation motion.
2.3.3 Diffusion Tensor and Kurtosis Imaging
For the first dataset, the diffusion and diffusional kurtosis tensors were fitted using the
FanDTasia toolbox implemented in MATLAB (Barmpoutis and Vemuri 2010; Barmpoutis and
Zhuo 2011). Coefficients from both tensors were robustly estimated for each voxel using a previously described homogenous polynomials approach, which enforces a positive diffusivity function, a positive-definite estimated diffusion tensor and a constrained estimated apparent
kurtosis (Barmpoutis and Vemuri 2010; Barmpoutis and Zhuo 2011). The estimated tensor coefficients were then used to interpolate the apparent diffusional kurtosis in any desired direction. Mean kurtosis (MK) maps were created by interpolating the apparent diffusional
48 kurtosis in 1000 directions based on the original 24-direction data and averaging them for every voxel. Axial kurtosis (KA) maps were obtained by estimating the apparent diffusional kurtosis in the principal diffusion direction. Radial kurtosis (KR) maps were obtained by averaging the apparent diffusional kurtosis interpolated in 1000 directions perpendicular to the principal diffusion direction, again based on the original 24-direction data. Interpolation was performed in 1000 directions for MK and KR to obtain an approximation of these metrics using the tensor coefficients while having a low directional bias (less than 0.1%, chosen directions were kept constant across individuals). In order to account for noise resulting from the lower SNR at high
b-value, each resulting DKI metric map was median-filtered with a 3x3x3 kernel to remove
potential outlying values. Maps of DTI metrics were additionally calculated based on the first b- value of the DKI dataset (i.e. b = 700 mm 2/s) since a b-value of 700 mm/s 2 has been used in ours
and other prior work showing sensitive associations with age (Damoiseaux et al. 2009;
Kochunov et al. 2007; Salat et al. 2005; Westlye et al. 2010).
For the second dataset, the DTI model was similarly fit using FSL. For individuals with multiple
available datasets, the dataset with the least average translation motion was used.
2.3.4 Per-voxel statistical analyses
Parametric maps of each diffusion metric were registered to a common FA template in MN1152
space using FSL registration tools and the Tract-Based Spatial Statistics procedure (TBSS) (Smith
et al. 2006). As a result, regression analyses were limited to voxels with mean FA values higher
than 0.2 on a standard tract skeleton, as described previously (Smith et al. 2006), minimizing
partial volume and registration confounds. General linear models included age, sex, education,
49 motion measures, as well as disease group for the analyses with individuals with MCI and AD.
DTI metrics were included as voxelwise explanatory demeaned covariates in additional age analyses to demonstrate unique age effects in DKI metrics that are unexplained by DTI metrics.
Significance testing and correction for multiple comparisons in the per-voxel analyses were achieved with 5000 permutations using FSL randomise with threshold-free cluster enhancement (TFCE), a combination of cluster-based and voxelwise thresholding proven to be
more sensitive to both spatially extended and sharply focused signals (Smith and Nichols 2009).
2.3.5 K-means clustering of region-of-interests
For the aging analyses, white matter tracts and subcortical regions on the TBSS skeleton were
segmented using a combination of the FreeSurfer Ti-based white-matter parcellation (Salat et
al. 2009) and FSL's Johns Hopkins University white matter labels (Mori and Crain 2005). These
atlases were co-registered with FSL's standard MN1152 space, and the segmented white-matter
regions were deprojected to the native space of each individual participant as in our prior work
using previously described TBSS techniques (Salat et al. 2012; Smith et al. 2007). This
deprojection allowed visual confirmation of correct translation between atlas to native space,
and the use of the native space data in the analysis minimized registration- and interpolation-
related biases. Furthermore, native voxels with FA values lower than 0.2 were excluded to
further prevent partial-volume effects, and white-matter regions with a mean native-space
skeleton volume lower than 0.95 cm 3 across both hemispheres were excluded. This resulted in
82 white matter ROls across the left and right hemispheres, and 5 commissural (intrinsically
bilateral) regions for a total of 87 regions further described in the publication associated with
50 this chapter (Coutu et al., 2014). The mean DKI/DTI measure of every ROI was extracted in the
native space. The 87 regions were combined into 46 bilateral regions (ROI values of the non-
commissural regions were averaged pairwise across hemispheres into 41 bilateral regions and
the values for the 5 remaining commissural regions were considered as is) in order to equalize
the weighting of commissural and non-commissural regions in the definition of the clusters.
This combination was justified by the general non-laterality of the age effects observed in this
study and the confirmation that there are no significant differences in the correlation
coefficients with age between hemispheres for any diffusion measure (Coutu et al., 2014).
K-means clustering was used to group regions according to their multivariate pattern of age
effect sizes across all diffusion parameters, determined by the Pearson's product-moment
correlation coefficient between age and each region's mean DTI/DKI metric. For clustering
purposes, the effect size of FA, MK, KA and KR, which are metrics that decline with age, was
inverted to be made positive and comparable to the effect size of MD, DA and DR, which
generally increase with age. The 46 regions were then clustered using MATLAB's k-means
clustering procedure based on their resulting effect size profile, or 'diffusion footprint', using
two different similarity measures, namely squared Euclidean distance and correlation. Here, we
define the concept of 'diffusion footprint' of a region by the visual pattern left by the effect size
profile, formed by the relative differences in age effect size across diffusion metrics. This notion
is important for the correlation measure, as two identical diffusion footprints will produce a
correlation coefficient of 1, even though they might not have the same average effect size
across metrics (i.e. one region may have an overall strong effect whereas the other has an
overall weak effect, but these regions still cluster together because of their effect pattern
51 across metrics as opposed to their effect size). The k-means clustering procedure was replicated with each similarity measure using an effect size profile or diffusion footprint including only the
DTI metrics in order to assess potential clustering differences with and without DKI metrics.
Repeating the procedure including only the DKI metrics was considered not practically meaningful as DTI metrics will always be available with the diffusion-weighting images required for the DKI model. This resulted in 4 different k-means procedures. All 4 procedures were used to generate 2 or 3 clusters; data are presented from the procedures generating 3 clusters and summary results are provided in (Coutu et a/., 2014) for the procedures generating 2 clusters.
2.4 Results
2.4.1 Associations between age and diffusion measures of microstructure
The results of the TBSS voxel-based age regression analyses on all diffusion metrics are shown in Figure 2.1. All results displayed are significant at p<0.01 after threshold-free cluster enhancement and correction for multiple comparisons. As observed, significant age effects were observed for all diffusion metrics across most of the white matter skeleton. There was
MD DA DR FA MK KA KR
Figure 2.1 Significant associations between all diffusion measures and age, co-varying for sex, education and motion measures. While the analyses were performed on the white matter skeleton as shown in black, the significant results have been dilated for better visualization. Correlations with age are positive for MD, DA and DR and negative for FA, MK, KA and KR. (MD: mean diffusivity; DA: axial diffusivity; DR: radial diffusivity; FA: fractional anisotropy; MK: mean kurtosis; KA: axial kurtosis; KR: radial kurtosis).
52 some variation in the spatial pattern of the effects, such as radial metrics (DR and KR) and FA generally not having significant age effects in the posterior limbs of the internal capsule, and axial metrics (mainly DA) having more spatially restricted age effects, especially centrally closer to the ventricles.
2.4.2 Variance and complementarity of the diffusion measures
Figure 2.2 shows a matrix of the squared pairwise correlation coefficients amongst individual
DTl/DKI metrics, averaged across the entire white-matter skeleton. A high degree of shared variance within DTI and within DKI metrics was observed, especially between mean and radial metrics. That is, KR was similarly related to MK as DR was to MD. A high degree of shared variance was also observed between DTI and DKI metrics (36%-78%), but was generally lower than within modality (within DTI or within DKI metrics) shared variance, and also notably always
lower than the variance shared between the commonly used MD and FA (84%).
Variance explained by individual Figure 2.2 Symmetrical matrix of the entire skeleton (%) metrics across variance explained by each individual MD DA DR FA MK KA KR metric for each other metric. The mean 100 value of each metric across the entire MD 83 97 84 71 59 70 90 white matter skeleton of each individual 80 DA 69 48 39 40 36 was correlated with the mean value of 70 each other metric and the resulting DR 92 78 62 78 60 coefficients were squared to obtain the explained variance. The correlations FA 73 51 76 50 were 40 between DKI and DTI metrics MK 75 96 emphasized by a box. (MD: mean 30 diffusivity; DA: axial diffusivity; DR: radial 61 KA 20 diffusivity; FA: fractional anisotropy; MK: 10 KR: KR mean kurtosis; KA: axial kurtosis; 0 radial kurtosis).
53 Given the shared variance across these measures, we investigated the residual age effects in
DKI metrics that are unexplained by DTI metrics. Shown in Figure 2.3a are the residual age effects in KR when accounting for DR. Although KR and DR are the DKI and DTI metrics most correlated with each other (78%), Figure 2.3a provides evidence for a difference in age effect between them. Similarly, the DTI metrics underwent the same procedure and the results for
DA, which is the DTI metric least correlated with DR (69%), are displayed in Figure 2.3b. The residual age effects in MD and FA accounting for DR were respectively spatially limited to the same areas and part of the same areas as DA (data not shown), suggesting DR accounted for the majority of age effects in all of the other DTI metrics. While complementary aging information was still obtained from multiple DTI metrics, KR showed broader differences from
DR spatially than any DTI metric. This was apparent even though the brain slices displayed in
Figure 2.3 were chosen to show as much residual DA-age effect as possible. Scatter plots of the mean KR, DA and DR values of the corresponding significant clusters from the voxelwise analyses are also displayed. In Figure 2.3a, KR shows a stronger association with age than DR, while DA (and other DTI metrics) displayed similar aging trends as DR. In Figure 2.3b, DA shows a stronger association with age than DR, while KR and DR showed similar age effects. In summary, both KR and DA showed unique variance across the white matter skeleton even when accounting for another highly correlated diffusion measure.
2.4.3 Clustering of age-effects based on multivariate diffusion patterns
We implemented a novel procedure to examine whether using diffusion kurtosis imaging metrics provides any additional information to diffusion tensor imaging in terms of the ability to
54 a) 1.50
E 1.25 E *
;1.00 tI~io* 0 pDA eC Wo o 0.75 * KR e
0.50 4 KR-age associations controlling for RD 40 50 60 io70 970 (corrected p<0.001 using TFCE) Age (years) b) 1.50 0 C4 1.25
o 1.00 C e % EDA *e 0 *KR $ 0.75 * DR
0.50
DA-age associations controlling for RD 40 50 60 70 80 90 (corrected p<0.001 using TFCE) Age (years)
Figure 2.3 Voxelwise results of general linear models for a) KR and b) DA including age and voxelwise DR as covariates. KR still had significant negative associations with age over broad areas of the white matter skeleton while controlling for DR, despite the two metrics sharing the highest variance explained between DTI and DKI metrics. DA, which is the DTI metric least correlated with DR, also had remaining age effects when controlling for DR, though they were more spatially limited. The significant results (p<0.001, corrected for multiple comparisons) have been dilated out of the skeleton for better visualization. The mean KR, DA and DR of the most significant clusters from the respective voxelwise analyses are plotted for each participant, demonstrating that the KR-age association is stronger than the DR-age association in the former, while the DA-age association is stronger than the DR-age association in the latter. (MD: mean diffusivity; DA: axial diffusivity; DR: radial diffusivity; FA: fractional anisotropy; MK: mean kurtosis; KA: axial kurtosis; KR: radial kurtosis). clustering similar effect strength or effect type. K-means clustering was used on both DTI metrics alone and on DTI+DKI metrics to group together regions with similar age effect sizes across metrics, based on two different measures of similarity: (1) squared Euclidean distance, which is the typical distance measure in metric space normally used in k-means clustering, with similar regions separated by a short distance; this measure clustered regions according to their
55 overall degree of age effect size across metrics (e.g. clustered regions showing strong effects together and regions showing weak effects together), and was unaffected by the addition of
DKI metrics (described in greater detail below); (2) correlation, which clustered together regions with a similar pattern of relative metric-age effect sizes ('diffusion footprint') as opposed to the overall effect size. This procedure was affected by the addition of DKI metrics, suggesting that DKI metrics provide unique information not present in DTI metrics.
We first clustered regions using the squared Euclidean distance measure, and each cluster center was computed as the average age effect size for each metric across all regions belonging to that cluster. The three cluster centers corresponding to DTI metrics alone and DTI+DKI
metrics are shown in Figure 2.4a and 2.4b, respectively. The regions belonging to each cluster
are colored to match their respective cluster center. As can be observed, the addition of DKI
metrics had only a negligible effect on the clustering using this similarity measure. Indeed, each
n usLI basedIU on DTI T! metrics alone had D I -age effect sizes very similar to that of a
matched center cluster based on DTI+DKI metrics. The clusters seem to represent regions with
low, medium and high age effect sizes. This suggests that the greatest distinction between
clusters seems to be their average age effect size across all diffusion metrics, and indicates that
DKI metrics do not add any significant information in that respect.
We next clustered regions based on their diffusion footprints rather than their overall age
effect size across metrics, using correlation as the similarity measure. This second clustering
approach is based on the pattern of relative age effect sizes across metrics. Shown in Figure 2.5
are the results of this clustering procedure, based on DTI metrics alone and DTI+DKI metrics
56 a) Similarity measure: squared Euclidean distance (DTI metrics only) 14 regions 15 regions 17 regions 0.5-. 0.4 I 0.2
W 0.13
Cluster I Cluster 2 0 Cluster 3 b) Similarity measure: squared Euclidean distance (DTI+DKI metrics) 9 regions 20 regions 17 regions 0.5 1
0.3
0. Cluster 1 Cluster Cluster 3 Figure 2.4 Results of the k-means clustering procedure using a squared Euclidean distance as similarity measure based on (a) DTI metrics alone and (b) with DTI+DKI metrics. The age effect size of every diffusion metric is shown for each cluster center. The sign of the age effect size, as determined by the Pearson product-moment correlation coefficient of each metric with age, was inverted for FA, MK, KA and KR for clustering purposes. A representation of each region colored according to its classification is shown on the dilated TBSS skeleton for easier visualization (analyses were solely performed on the white matter skeleton). (MD: mean diffusivity; DA: axial diffusivity; DR: radial diffusivity; FA: fractional anisotropy; MK: mean kurtosis; KA: axial kurtosis; KR: radial kurtosis; WM: white matter). respectively. Arbitrary units are used to identify this pattern as only the relative differences between metric-age effect sizes matters, not their absolute levels. The patterns of each center cluster in Figure 2.5 were therefore scaled to be similar in range to patterns shown in Figure
2.4, and any absolute value in isolation is arbitrary and therefore meaningless. The clusters now differ from those based purely on the overall strength of the metric-age associations, as shown in Figure 2.4. Indeed, in Figure 2.5a, the center clusters all shared a similar diffusion footprint, characterized with a relatively high DR-age effect size and a low DA-age effect size a's well as medium to high MD- and FA-age effect sizes. Most of these regions were therefore clustered together (cluster 1) in Figure 2.5c. Cluster 2 had few regions and was similar to cluster 1 except
57 a) Similarity measure: correlation (DTI metrics only) 33 regions 7 regions 6 regions ... 0.5-A 0.42 C S0.31 '0.2- U L~S 01 S0.0, Clusterl Cluster 2 Cluster3EN
b) Similarity measure: correlation (DTI+DKI metrics) 19 regions 19 regions 8 regions
0.4- C MO! S0.3-f- 0.2
ClusterlE Cluster 2UN Cluster3 Figure 2.5 Results of the k-means clustering procedure using correlation as similarity measure based on DT1 metrics alone (a) and with DTI+DKI metrics (b). In this case, regions in each cluster shared a diffusion footprint that was most highly correlated with the diffusion footprint of their cluster center, regardless of their overall age-association strength for all metrics. When using correlation as the similarity measure, arbitrary units are used to represent each cluster center as the diffusion footprint can be demeaned and scaled across metrics without changing its correlation coefficient with the diffusion footprint of each region. A representation of each region colored according to its classification is shown on the dilated TBSS skeleton for easier visualization (analyses were solely performed on the white matter skeleton). (MD: mean diffusivity; DA: axial diffusivity; DR: radial diffusivity; FA: fractional anisotropy; MK: mean kurtosis; KA: axial kurtosis; KR: radial kurtosis; WM: white matter). for a lower relative FA-age effect size, and cluster 3 represented regions driven mainly by a high increase in MD with age, with milder effects in FA. A key observation is that the addition of DKI metrics split cluster 1 into two distinct clusters, shown by clusters 1 and 2 in Figure 2.5b.
Indeed, the pairwise correlation coefficients between the DTI footprint of cluster 1 in Figure
2.5a and clusters 1 and 2 in Figure 2.5b were all above 0.99. The main difference therefore lied
in the DKI metrics, with both DTI and DKI metrics having similar age effects in cluster 1, and DKI
metrics having higher age effects in cluster 2. Regions from cluster 3 however were already
properly clustered using DTI metrics only, and these regions exhibited mostly isotropic changes
in diffusivity and diffusional kurtosis with a high degree of axial kurtosis changes with age.
58 2.4.4 Group differences between non-demented controls and individuals with MCI/AD
The results of the TBSS voxel-based group comparisons for all diffusion metrics are shown in
Figure 2.6. Group differences displayed are significant at p<0.05 and remaining age effects are shown significant at p<0.01 as in Figure 2.1, both after threshold-free cluster enhancement and correction for multiple comparisons. Group differences between controls and individuals with
MCI were only observed with DA broadly across the white matter skeleton. Individuals with AD on the other hand had significant differences from the control participants in all diffusivity
measures. However, MD and DR had group differences spatially distinct from those observed with DA. No significant group differences were observed for FA, except for a small part of the
Mean Diffusivity Axial Diffusivity Radial Diffusivity
MCI
Controls
AD
Controls
Remaining Age Effects a. Figure 2.6 Significant group differences between controls and individuals with MCI and AD (corrected p<0.05). The significant results have been dilated out of the TBSS skeleton where analyzed are performed for better visualization. Remaining age effects after taking into account diagnosis, sex, education and motion measures are also shown at corrected p<0.01 (MD: mean diffusivitv; DA: axial diffusivitv; DR: radial diffusivitv; FA: fractional anisotropv).
59 left internal capsule between individuals with AD and controls. The remaining associations between age and diffusion measures were observed for all diffusion metrics and with a characteristically similar spatial pattern as those observed in Figure 2.1 in our large independent sample of older non-demented participants.
2.5 Discussion
The current data demonstrate that all diffusion measures are sensitive markers of age- associated tissue alterations. With respect to the novel DKI metrics, an aging-associated decrease in the complexity of the diffusion microenvironment was found in most white matter structures, suggesting a global trend towards tissue homogeneity with age. This was primarily explained by a global decrease in KR, with prefrontal, parietal and temporal white matter showing stronger age-associations than primary motor and sensory areas, and a less pronounced decrease in KA though relatively uniformly throughout the white matter.
Importantly, DKI provided unique and complementary regional markers of microstructural changes relative to DTI. While DTI metrics were sufficient in identifying white matter regions most strongly affected by age, DKI metrics enabled a novel mapping of aging processes based on a multivariate diffusion footprint that is distinct from the spatial patterns based on the overall age effects in DTI/DKI metrics. These data additionally suggested that DKI and the multivariate combination of all diffusion metrics may reveal early microstructural changes in neurodegenerative diseases, and this concept is further used in Chapter 3 to compare the microstructure of white matter lesions of individuals with MCI and AD and non-demented controls. Group differences in diffusion measures between individuals with MCI and AD and
60 non-demented controls were observed throughout the white matter and neuroimaging markers will be used in Chapter 4 to further explain those differences and attempt to attribute them to specific underlying disease processes.
The first original goal of the study was achieved through the findings that much of the white
matter showed age-associated reduction in DKI metrics, signifying a diffusivity homogenization
across its various cellular compartments. More affected regions identified by DKI tended also to exhibit greater increases in diffusivity in aging, and coincided with regions previously
implicated in aging (Barrick et al. 2010). The spatial variation in age-associations across all
diffusion metrics within white matter seems to be in accord with previously hypothesized
models of cortical aging where prefrontal and association areas are greatly affected while
primary motor and sensory areas such as the occipital and central areas are relatively preserved
(Driscoll et al. 2009; Raz et al. 1997). Overall, diffusion measures were sensitive markers of the
white matter microstructure in vivo in a large population.
The second original goal of the study was to determine whether DKI metrics provide any
additional information on microstructural changes during aging that is unaccounted for by DTI
metrics. DKI metrics co-varied with DTI metrics to a lower extent on a global level than DTI
metrics or DKI metrics did among themselves, suggesting that the use of both DTI and DKI
metrics might enable better characterization of the white matter microstructure. Furthermore,
even the most correlated DKI and DTI metrics, KR and DR, were differentially affected by aging
over large areas of the white matter, while DR accounted for the majority of age effects in all of
the other DTI metrics. Interestingly, this was also true for DA which exhibited independent age
61 effects from DR, though to a lesser extent, confirming the complementary of DTI metrics despite their high correlation. Initial clustering experiments were driven by the overall average age effects across all diffusion metrics, suggesting that the level of microstructural damage associated with age may already be properly characterized by DTI measures, which requires shorter and simpler MRI acquisition. When trying to distinguish type of microstructural age effects using our multivariate diffusion footprint approach, DTI was mostly useful to distinguish between isotropic and anisotropic effects. It is interesting to note that the diffusion footprint approach leads to a quite different spatial pattern than that produced from the average age effect across diffusion metrics. For example, while the cerebellum and superior frontal white matter seem to undergo similar anisotropic microstructural changes, the superior frontal white matter undergoes a greater degree of these same changes than the cerebellum. Furthermore,
DKI metrics provided additional information that allowed a clustering of regions that was not possible with DTI metrics alone. Therefore, we conclude that DKI provides additional, unique complementary information about types of microstructural changes in the context of healthy aging when used in combination with DTI metrics. Additional testing showing that each diffusion measure represents unique properties can be found in the publication (Coutu et al.,
2014).
In addition to these original goals, we presented in this chapter the result of group differences between white matter DTI measures of microstructure of individuals with MCI and AD and
nondemented controls. Interestingly, the group differences were different from the age effects where most diffusion measures were affected. For instance, the only differences between
individuals with MCI and controls were observed for DA, which had age effects that were less
62 spatially broad than other metrics. Furthermore, the large majority of group differences were observed for diffusivity markers, and almost no differences were observed for FA. This is consistent with previous DTI studies in individuals with AD (Acosta-Cabronero et al., 2010; Salat et al., 2010; Douaud et al., 2011), including a study by our group that showed similar results as those presented here, though it focused on average DTI measures of whole tracts (Lee et al.,
2015). The group differences were also observed primarily in areas prone to white matter lesions. To this effect, Chapter 3 investigates whether there are any microstructural differences in the white matter lesions of individuals with AD compared to those of non-demented controls. Our available dataset did not allow the investigation of whether differences in the novel DKI markers are present in individuals with MCI and AD. However a few studies have suggested group differences in DKI markers (Falangola et al., 2013; Fieremans et a/., 2013).
2.6 Appendix
2.6.1 Speculation on the histological basis of the age effects observed for diffusion kurtosis imaging measures based on its biophysical model
The histological basis of the diffusional kurtosis cross-sectional changes reported in this chapter is currently unknown. Several biological processes may underlie the global MK and KR decline in white matter with age, such as myelin breakdown, increases in axonal membrane permeability, edema, fiber loss and shortening as well as decrease in density of myelinated axons (Bartzokis,
2004; Fazekas et al. 1993; Peters, 2002), which may all increase the tissue homogeneity and decrease the variability in diffusivity among tissue compartments. The age-effect in KA is likely to be associated with the same biological processes in crossing fibers perpendicular to the
63 principal diffusion direction. However, as cross-sectional changes are also observed in predominantly non-crossing fibers, the age-effect in KA might additionally be linked to an
increased homogeneity along the principal diffusion direction, either due to an aging-related diffusion homogenization process or to an increased extracellular space presumed to have
lower heterogeneity than axons and glia (Fieremans et al. 2010). Ischemia and chronic damage
may also result in a change in DKI metrics following tissue infarction and cell death, as recent
evidence points to an increase in KA in the acute ischemic period (Hui et al. 2012a; Hui et al.
2012b; Jensen et al. 2011). In this study, a multivariate classification of white matter regions
based on the relative differences between age effects of DTI and DKI metrics was introduced.
For instance, the first of the three clusters obtained using all diffusion metrics represents
regions with similar age-associations in diffusional kurtosis and diffusivity, and with stronger
effects in radial than axial metrics, potentially attributable to changes in the density of
myelinated axons, myelin integrity and axonal membrane permeability. The second cluster also
had a strong radial component but diffusional kurtosis changes were stronger than diffusivity
changes, indicating stronger tissue-compartment changes perpendicular to the fibers, resulting,
for instance, from an increasing similarity between the diffusivities of intra- and extra-axonal
spaces. The third cluster had a strong isotropic age effect in both DTI and DKI metrics,
potentially related to fiber loss and shortening, edema and gliosis. These interpretations are
speculative and based on limited prior work. Future studies will address the histological
validation and neuropathological assessment of DKI properties and of the potential types of
microstructural changes suggested in this study.
64 2.6.2 Study limitations
This study has limitations that will be addressed in future work. First, it is important to emphasize that the underlying tissue microstructure measured by DKI remains poorly understood and that animal studies will be required to confirm the physiological and histological underpinnings associated with the diffusional kurtosis changes observed in this study. Second, this is a cross-sectional study providing limited information about what may be expected longitudinally. Third, the low image SNR at higher b-values could have resulted in less- than-optimal fitting of the DKI parameters on which a median filter was used. To examine the influence of the median filter, MK was fitted using an in-house unconstrained nonlinear estimation algorithm (Jensen et al. 2005) and interpolation was used to replace improbable fitted kurtosis values, which resulted in the same spatial distribution of significant age effects.
Head motion is also a major potential confound in diffusion imaging studies; however, we controlled for motion parameters for each participant in the general linear models. Finally, it is important to note that the same type of change in diffusion property might be more sensitively detected by one diffusion metric than another simply due to the specific anatomy of a given region. For example, the interpretation of axial and radial metrics relies on the assumption of a single population of aligned fibers, while crossing fibers have been shown to predominate in the cerebral white matter (Jeurissen et al. 2012; Wedeen et al. 2012). Indeed, factors such as the underlying white matter structure affect diffusion measures (Wheeler-Kingshott and
Cercignani 2009) and therefore their biological interpretation is not simple. Future studies will address this issue by using the potential of DKI to resolve crossing fibers and enable characterization of different fiber populations with aging and neurodegenerative disease.
65 66 Chapter 3: Properties of white matter lesions in Alzheimer's disease do not differ from non-demented controls and poorly relate to markers of neurodegeneration
3.1 Overview
As described in the thesis introduction, white matter lesions are highly prevalent in individuals with Alzheimer's disease (AD). Although these lesions are presumed to be of vascular origin and linked to small vessel disease in older adults, little information exists about whether they differ in Alzheimer's disease and about their relationship to markers of classical AD neurodegeneration. Thus, this chapter focuses on examining differences in the diffusion properties of these white matter lesions (WML) as segmented on Tr-weighted MRI between non-demented controls and individuals with MCI and AD, as well as their link to imaging markers presumed to be altered due to primary AD neurodegenerative processes. Tissue microstructure of WML was quantified using diffusion tensor imaging which was detailed in the previous chapter. Diffusivity progressively increased non-significantly in WML of cognitively healthy older adults, individuals with mild cognitive impairment and individuals with AD,
67 respectively, after taking into account the normative anatomy. Significant associations were found between diffusivity of WML and ventricular volume, volume of WML and total WM volume. In comparison, group differences in parahippocampal white matter microstructure were found for all diffusion metrics and were largely related to hippocampal volume. These results suggest that the microstructural properties of white matter lesions are similar in aging, mild cognitive impairment and Alzheimer's disease and do not primarily relate to markers of neurodegeneration such as hippocampal volume and regional cortical thickness, suggestive of a distinct disease pathway which is further investigated in the next chapter. While we have looked at the quality of the lesions in this chapter, the relationship between the quantity or volume of white matter lesions and other neuroimaging markers affected in Alzheimer's disease is explored in the next chapter. Most of the work presented in this chapter and the next chapter has been published in the Journal of Alzheimer's Disease (Coutu et al., 2015).
SIntruction
MRI measures of hippocampal volume and cortical thickness have been shown to predict
incident Alzheimer's disease (AD) (Jack Jr et al., 1999; Whitwell et al., 2008; Bakkour et al.,
2009; Kantarci et al., 2009; Dickerson et al., 2011) and to correlate with classical
histopathological measures of AD (Jack Jr et al., 2002; Archer et al., 2006). Less widely
recognized is that total volume of white matter lesions also increases with (Scheltens et al.,
1995; Yoshita et al., 2006; Holland et al., 2008) and is predictive of (Smith et al., 2008; Brickman
et al., 2012; Provenzano et al., 2013; Mortamais et al., 2014) mild cognitive impairment (MCI)
and AD, though not in every study (DeCarli et al., 2004; Kantarci et al., 2009). These lesions, also
68 called leukoaraiosis, are typically identified in vivo as white matter signal hyperintensities of presumed vascular origin on neuroimaging (Wardlaw et al., 2013) due to their appearance on
T2-weighted and fluid-attenuated inversion recovery (FLAIR) MRI. They can also be observed as moderately hypointense regions in white matter (WM) on Trweighted MRI (however not well distinguished from infarcted tissue). Epidemiological studies demonstrate that these white matter lesions are associated with small vessel disease (Pantoni, 2010; Gouw et al., 2011),
hypertension (Jeerakathil et al., 2004; Murray et al., 2005; Stenset et al., 2006; Yoshita et al.,
2006; Gottesman et al., 2010; Rostrup et al., 2012) and other vascular risk factors (Breteler et
al., 1994; Jeerakathil et al., 2004; Murray et al., 2005; Stenset et al., 2006; Yoshita et al., 2006)
in non-demented individuals. MRI and SPECT studies confirm their lower perfusion compared to
normal-appearing WM (Markus et al., 2000; Marstrand, 2002; Bastos-Leite et a/., 2008;
Brickman et a/., 2009; Makedonov et al., 2013). While lesion volume is known to be increased in
AD (Scheltens et a/., 1995; Yoshita et al., 2006; Holland et al., 2008; Gao et a/., 2011; Brickman
et al., 2012), limited evidence exists that the white matter lesions present in AD are similar in
nature to those observed in non-demented older individuals (Leys et al., 1990; Pantoni and
Garcia, 1997; Yoshita et al., 2006). Relatively few pathological studies have been conducted,
most of them showing increased demyelination and axonal loss, and more severe gliosis and
denudation of the ventricular ependyma in the lesions of AD compared to the lesions of non-
demented controls (Scheltens et a/., 1995; Gouw et al., 2008; Barker et a/., 2014). Additionally,
little is known about how this typically vascular-associated tissue damage relates to more
classical imaging markers of AD pathology (Hirono et a!., 2000; Lo et al., 2012; Jacobs et aL.,
2014), such as cortical thickness and hippocampal volume, which could provide important
69 information about how this tissue damage fits with the classical and diagnostic pathophysiologic properties of the disease.
We segmented white matter lesions automatically with the FreeSurfer analysis stream using T- weighted images. While this segmentation was chosen for its automation and convenience, we note that this procedure does not differentiate white matter lesions typically measured as
'hyperintensities' from lacunar infarcts, though infarcts are much less prevalent and contribute a much smaller volume (DeCarli et al., 2004; Chen et a/., 2009; Duering et al., 2013). We refer to the white matter segmentation studied here as white matter lesions (WML), which may be measuring similar underlying pathology as those from standard T 2-weighted and FLAIR methods based on highly correlated volumetric results as detailed in the methods. We examined volume and tissue properties of WML in a sample of controls, MCI, and AD to better understand how these markers relate to common degenerative processes in AD. Results were compared to other types of WM damage including changes in the normal-appearing white matter (NAWM) globally and in the parahippocampal WM which exhibited microstructural changes in prior work
in AD (Salat et al., 2010) and was considered a 'pathology control' to determine whether the
effects within the lesions were truly unique and distinct from a more classical AD effect
potentially secondary to cortical degeneration.
3.3 Materials and Methods
3.3.1 Participants and MRI acquisition
The dataset used was the same as the second dataset described in the previous chapter, from
the Alzheimer's Disease Neuroimaging Initiative (ADNI, http://adni.loni.usc.edu). A subgroup of
70 individuals with a volume of WML greater than 1% of total WM volume was also examined to assure that results were not skewed by individuals with small volumes of WML. Demographics
both for this subgroup and for the entire group are provided in Table 3.1.
Table 3.1 Demographics for all participants and for the subgroup with a volume of white matter
lesions greater than 1% of the total white matter volume.
All ADNI, n = 215
CN MCI AD p-value
Participants (female) 73 (46) 96 (36) 46 (17) 0.0016
Age [years] 72.98 (0.84) 73.80 (0.73) 74.58 (1.06) 0.4900
Education [years] 16.30 (0.32) 16.10 (0.28) 15.24 (0.41) 0.1086
MMSE [-]a 28.67 (0.25) 27.88 (0.22) 23.14 (0.32) <0.0001
APOE E4 [# alleles]b 0.32(0.10) 0.64(0.07) 0.88(0.11) 0.0006
Translation motion [mm] 1.32 (0.07) 1.34 (0.06) 1.35 (0.09) 0.9671
Rotation motion [degrees] 0.0062 (0.0003) 0.0063 (0.0003) 0.0068 (0.0004) 0.5621
ADNI subgroup with WML volume > 1% total WM volume, n = 118
CN MCI AD p-value
Participants (female) 30 (17) 54 (20) 34 (9) 0.0436
Age [years] 75.74 (1.24) 76.61 (0.93) 77.05 (1.17) 0.7356
Education [years] 16.30 (0.54) 16.22 (0.40) 15.41 (0.51) 0.3774
MMSE [-]c 28.17 (0.49) 27.64 (0.30) 22.48 (0.41) <0.0001
APOE E4 [# alleles]d 0.18 (0.16) 0.65 (0.09) 0.78 (0.12) 0.0108
71 Translation motion [mm] 1.35 (0.11) 1.46 (0.08) 1.48 (0.11) 0.6372
Rotation motion [degrees] 0.0063 (0.0005) 0.0069 (0.0004) 0.0073 (0.0005) 0.4292
All significant p-values are bolded. Standard errors are shown in parentheses. (MMSE: Mini-
Mental State Exam; CN: control; MCI: mild cognitive impairment; AD: Alzheimer's disease; WM:
white matter; WML: white matter lesions)
a. Information missing for 16 CN, 20 MCI and 11 AD.
b. Information missing for 29 CN, 5 MCI and 12 AD.
c. Information missing for 12 CN, 7 MCI and 9 AD.
d. Information missing for 13 CN, 3 MCI and 7 AD.
3.3.2 Diffusion data processing
The diffusion data was processed as described in sections 2.3.2 and 2.3.3.
3.3.3 Automated subcortical and WML segmentation
Automated subcortical and WM segmentation as well as cortical surface reconstruction were
obtained from the T-weighted images using FreeSurfer (https://surfer.nmr.mgh.harvard.edu)
(Fischl et al., 1999, 2002). The longitudinal processing stream was used to obtain more accurate
results that are not biased by the number of datasets available for each participant (Reuter et
al., 2010, 2012; Reuter and Fischl, 2011). Segmentations of the entorhinal and
parahippocampal WM were combined into a single segmentation that we referred to here as
parahippocampal WM. The automated segmentation also included a WML segmentation that is
conservative relative to T 2-weighted and FLAIR MRI and segmented only the most obvious WML
identifiable on T-weighted images. FreeSurfer mrirelabelhypointensities was used to refine
72 the WML segmentation using the surface reconstruction. We found a correlation coefficient of
0.96 (n = 112) between the volume of WML obtained with FreeSurfer and the WM hyperintensity volume obtained with FLAIR MRI and tissue priors using publicly-available values from ADNI. However, the WM hyperintensity volume obtained with FLAIR MRI was on average
1.14 times greater than the volume obtained with FreeSurfer, which was more conservative.
Examples of T-weighted hypointensities and their WML segmentation in controls and individuals with MCI and AD are provided in Figure 3.1. Total WM volume, parahippocampal
WM volume, ventricular volume (lateral ventricles) and hippocampal volume were normalized as a volume percentage of estimated total intracranial volume (eTIV) in each individual. The natural logarithm of the volume of WML divided by total WM volume was used for all statistical analyses to obtain a more normalized distribution of this typically skewed measure.
Additionally, FreeSurfer was used to extract measures of cortical thickness from cortical surface labels representing the regions that undergo thinning in early AD, previously described as the cortical signature of AD given the reliability of this effect across samples (Bakkour et a/., 2009;
Dickerson et ai., 2009, 2011). The average cortical thickness weighted by the surface area of each label has been used as a specific measure of cortical atrophy in AD and will be referred to here as the AD signature cortical thickness or simply as cortical thickness. This cortical signature did not include the hippocampus as this structure has unique anatomy compared to the regions of the cortical mantle modeled here as a two dimensional sheet for the measurement of cortical thickness. Additionally, the hippocampus is a unique structure known to be vulnerable to both AD and vascular pathology (Laakso et a/., 1996; Du et a/., 2002) and therefore may have
73 Figure 3.1 Examples of automated WML segmentation by FreeSurfer in controls and individuals with MCI and AD with a) a low volume of WML approximating 1% of total WM volume and b) a high volume of WML approximating 5% of total WM volume. The WML segmentation is represented by light purple. The rest of the segmentation corresponds to the standard color table freely available on the FreeSurfer website, though the segmentation of the inferior lateral ventricles was modified to a deep purple color like the rest of the ventricles to avoid confusion with the WML segmentation. 74 unique properties compared to cortical structures included in the AD signature calculation
(which includes neocortex as well as other types of cortex).
3.3.4 Registration procedures
The diffusion-weighted images were registered to the anatomical series using FreeSurfer boundary-based registration (Greve and Fischl, 2009). Average DTI metrics within the WML segmentations were obtained for each individual in diffusion native space. A segmentation of the normal-appearing white matter (NAWM) was created from the subtraction of the WML segmentation from the total WM segmentation in native diffusion space. Furthermore, a WM skeleton mask was created using FSL Tract-Based Spatial Statistics (Smith et al., 2006) and was used to reduce partial volume effects when calculating average DTI values coming from both
NAWM and parahippocampal WM in native diffusion space, as described in the previous chapter and the associated published work (Coutu et al., 2014).
3.3.5 Normative data calculation
Each individual's anatomical series was registered to the MN1152 common space using FSL
FLIRT and FNIRT to allow comparison of the anatomical segmentations and extract group maps of the WML prevalence in each voxel. Diffusion maps were warped to this common space using
FSL to create normative diffusion maps by averaging maps from all subjects with a volume of
WML less than 1% of total WM volume. These normative averages were then warped back to the native diffusion space of every subject. This procedure was performed to determine the difference between diffusion metrics inside WML and normative diffusion metrics in the same
75 regions for each individual, and therefore account for the varying inter-individual location of
WML in our analyses.
3.3.6 Statistical analyses
Statistical analyses were performed using JMP 10 statistical software (SAS Institute Inc., Cary,
NC, U.S.A.). Post-hoc Tukey HSD tests were performed to assess for any significant differences
between groups for the volume and DTI metrics of each structure, using age, sex, education
(Teipel et al., 2009) and motion measures (average translation and average rotation (Yendiki et
al., 2014)) as covariates. Individuals with a volume of WML less than 1% of total WM volume
were excluded from statistical analyses of DTI metrics in WML since they generally had few
WML voxels which were close to the ventricles and for which ROI averages of diffusion metrics
were more similar to the ventricles, likely due to partial-volume effects. General linear models
including group, age, sex, education, motion measures and these five measures were used to
understand which individual measures accounted for the variance in diffusion metrics
independently of all other variables. These models were applied for the average diffusion
measures in WML, parahippocampal white matter and normal-appearing white matter. Group
by marker interactions were not included as they were not significant when added to the
models. All results were corrected for multiple comparisons with Bonferroni (3 WM regions /
comparisons for group differences in volume and 3 WM regions x 4 diffusion metrics = 12
primary comparisons for all results involving DTI) and estimated parameters were provided in
the models in addition to p-values to ease interpretation. The same models including the
number of APOE E4 alleles as an additional variable are presented in the appendix.
76 3.4 Results
3.4.1 Group differences in prevalence and tissue properties of WML
Qualitative examination suggested that the spatial distribution of WML was similar across groups as previously described (Holland et al., 2008) (Figure 3.2A). The subtractions between groups (Figure 3.21B) suggested greater prevalence of WML in posterior areas for MCI compared
to controls and greater prevalence of WML in anterior areas for AD compared to MCI. However,
voxel-wise differences in prevalence were only significant between AD and controls (corrected
p < 0.05 using FSL randomise with threshold-free cluster enhancement, not shown).
Figure 3.2 A) Spatial prevalence of WML in controls (CN) and individuals with mild cognitive impairment (MCI) and Alzheimer's disease (AD). Color scale varies from red to yellow, from when at least one participant has WML to when 30% or more have WML in a given voxel. The cap of 30% allows easier comparison of the diffuse differences between groups. B) Group differences in spatial prevalence of WML. Color scale varies from light blue to blue for negative differences of -20 to -1 percentage points and from red to yellow for positive differences of 1 to 20 percentage points. All results are displayed in the common MN1152 space after using FSL FNIRT for proper nonlinear registration of the subcortical structures.
77 Group differences in volume of WML were significant between AD and controls (corrected p <
0.01, Figure 3.3A). No significant group differences were observed for diffusion metrics within
WML. Normalization of diffusion measures to account for the differential location of the lesions across individuals reduced the standard error of group averages but group differences were still
not significant after correction for multiple comparisons. This is contrasted by significant group
differences in volume and diffusion properties for the parahippocampal WM (Figure 3.3B) and a
few group differences in total WM volume and NAWM diffusion properties (Figure 3.3C).
A) 3 1.0 *MCI 0.10 E 08 E 0. 0.05 060 FA 2 0.4. 0.00 A~ II - !- 21.MD DA DR 0.2 Z 0 50 0.01 1 o -0.05 MD DA DR FA
0..4 C) 30 4 1.0 t t E 04 E.8 20 0.8 0.3 0.6 -~0.6 3.020.4 E 0.4 j~.10.2 30.2
00 0 0.0. Z 0 I 0.0 MD DA DR FA MD DA DR FA Figure 3.3 Group comparisons in volume and DTI metrics for a) WML, b) parahippocampal WM and c) NAWM. The volume of WML is normalized to the total WM volume. Parahippocampal and total WM volumes are normalized by the eTIV. Group differences were statistically assessed with the post-hoc Tukey test for each volume and DTI metric, correcting for age, sex, education and motion measures (*, ** and *** for corrected p < 0.05, 0.01 and 0.001 respectively, t for uncorrected p < 0.05). The difference between diffusion properties of WML and the corresponding diffusion properties in a non-lesioned normative brain is shown in addition to the absolute diffusion properties of WML. The analyses were limited to individuals with a volume of WML greater than 1% total WM volume. The log-transform of the normalized volume of WML was used for statistical purposes. Standard error bars are shown. (MD, DA and DR: mean, axial and radial diffusivity; FA: fractional anisotropy; WML: white matter lesions; WM: white matter; NAWM: normal-appearing white matter; eTIV: estimated total intracranial volume; CN: control; MCI: mild cognitive impairment; AD: Alzheimer's disease)
78 3.4.2 Associations between tissue properties of WML and neuroimaging markers of AD
In Table 3.2, we tested whether diffusion metrics of WML were associated with neuroimaging markers to determine if they would be uniquely related to hippocampal volume and cortical thickness. We found significant associations between increased diffusivity of WML and increased ventricular volume (MD, DA, DR: corrected p < 0.001), decreased volume of WML
(MD, DR: corrected p < 0.05; DA: corrected p < 0.01) and decreased total WM volume (DR: corrected p < 0.05), independently of each other. No associations were significant between diffusion metrics of WML and hippocampal volume or cortical thickness, with the exception of
increased FA of WML with decreased cortical thickness (corrected p < 0.05).
Table 3.2 Models of the diffusion parameters in WML with all neuroimaging markers.
WML (subgroup with volume > 1% total WM volume, n = 118)
Parameters MD (@; p-value) DA (B; p-value) DR (P; p-value) FA ($; p-value)
Group (MCI) 0.03; 0.6332 0.04; 0.5480 0.02; 0.6976 0.01; 0.8862
Group (AD) 0.11; 0.1604 0.07; 0.3995 0.14; 0.0912 -0.27; 0.0229
Age 0.07; 0.1938 0.04; 0.4654 0.08; 0.1118 -0.18; 0.0167
Sex (female) 0.04; 0.4210 0.03; 0.4923 0.04; 0.3949 -0.06; 0.3937
Education 0.10; 0.0280 0.08; 0.0631 0.11; 0.0200 -0.12; 0.0831
Cortical thickness 0.03; 0.6039 -0.01; 0.8331 0.05; 0.3521 *-0.24; 0.0029
Hippocampal vol. 0.08; 0.2628 0.07; 0.3264 0.08; 0.2439 -0.13; 0.2131
Ventricular volume 0.55; <0.0001 0.60; <0.0001 0.51; <0.0001 0.15; 0.0712
79 Volume of WIML *-0.26; 0.0015 **-0.27; 0.0007 *-0.25; 0.0026 -0.07; 0.4984
Total WIM volume -0.15; 0.0042 -0.14; 0.0085 *-0.16; 0.0036 0.12; 0.1270
Norm. properties 0.55; <0.0001 0.50; <0.0001 0.57; <0.0001 0.61; <0.0001
Translation motion 0.21; 0.0218 0.13; 0.1435 0.26; 0.0069 **-0.50; 0.0002
Rotation motion -0.19; 0.0350 -0.14; 0.1144 -0.22; 0.0187 0.29; 0.0254
All continuous variables were standardized prior to applying the model for easier comparison of parameter estimates (P). Uncorrected p-values are presented and significant associations with corrected p < 0.05 are bolded (*, ** and *** for corrected p < 0.05, 0.01 and 0.001 respectively). Associations with uncorrected p <0.05 are italicized. (WML: white matter lesions;
WM: white matter).
3.4.3 Associations between tissue properties of parahippocampal white matter and neuroimaging markers of AD
In Table 3.3, we tested in comparison whether parahippocampal WM diffusion metrics were associated with neuroimaging markers to determine if they would be primarily related to hippocampal volume and cortical thickness. Increased diffusivity of parahippocampal WM was associated with hippocampal volume (MD, DR: corrected p < 0.01; DA: corrected p < 0.05). We additionally found an association between total WM volume and FA of the parahippocampal
WM (corrected p < 0.05).
Table 3.3 Models of the diffusion parameters in parahippocampal white matter with all neuroimaging markers.
80 Parahippocampal WM (all, n = 215)
Parameters MD (1; p-value) DA (0; p-value) DR (1; p-value) FA (1; p-value)
Group (MCI) -0.08; 0.3057 -0.02; 0.8662 -0.11; 0.1425 0.13; 0.0802
Group (AD) 0.16; 0.1809 0.06; 0.6554 0.20; 0.0756 -0.23; 0.0367
Age 0.03; 0.6548 -0.00; 0.9909 0.05; 0.4887 -0.11; 0.1210
Sex (female) -0.01; 0.8554 -0.06; 0.4412 0.01; 0.8476 -0.11; 0.0633
Education 0.05; 0.4451 0.01; 0.8321 0.06; 0.2978 -0.10; 0.0900
Cortical thickness -0.11; 0.1440 -0.02; 0.7771 -0.15; 0.0391 0.19; 0.0091
Hippocampal vol. **-0.32; 0.0008 *-0.31; 0.0038 **-0.30; 0.0007 0.19; 0.0274
Ventricular volume 0.02; 0.8150 0.00; 0.9282 0.03; 0.7291 0.01; 0.9291
Volume of WML 0.13; 0.1038 0.15; 0.1004 0.11; 0.1361 -0.05; 0.5017
Total WM volume -0.02; 0.8370 0.11; 0.2064 -0.08; 0.2762 *0.23; 0.0012
Translation motion -0.01; 0.9051 0.03; 0.8128 -0.03; 0.7428 0.08; 0.4395
Rotation motion 0.11; 0.3269 0.01; 0.9165 0.16; 0.1505 -0.25; 0.0216
All continuous variables were standardized prior to applying the model for easier comparison of
parameter estimates (1). Uncorrected p-values are presented and significant associations with
corrected p < 0.05 are bolded (*, ** and *** for corrected p < 0.05, 0.01 and 0.001
respectively). Associations with uncorrected p <0.05 are italicized. (WML: white matter lesions;
WM: white matter).
81 3.4.4 Associations between tissue properties of normal-appearing white matter and neuroimaging markers of AD
In Table 3.4, we tested whether NAWM diffusion metrics were associated with neuroimaging markers to confirm if they would be most associated with the volume of WML as shown in prior work examining cognitively healthy older adults (Leritz et al., 2014), but also possibly minimally associated with hippocampal volume and cortical thickness. As expected, we found significant associations between greater volume of WML and both greater NAWM diffusivity and lower
NAWM FA (MD, DR, FA: corrected p < 0.001). However, we also found additional, independent associations between lower total WM volume and both greater DR (corrected p < 0.05) and lower FA (corrected p < 0.01) and no associations involving either hippocampal volume or cortical thickness.
Table 3.4 Models of the diffusion parameters in normal-appearing WM with all neuroimaging markers.
Normal-appearing white matter (all, n = 215)
Parameters MD (P; p-value) DA (P; p-value) DR (1; p-value) FA (P; p-value)
Group (MCI) 0.11; 0.0994 0.17; 0.0323 0.09; 0.2027 0.02; 0.8058
Group (AD) -0.09; 0.3685 -0.10; 0.4079 -0.09; 0.3804 0.07; 0.5049
Age 0.09; 0.1603 0.13; 0.0694 0.07; 0.2754 0.00; 0.9891
Sex (female) 0.05; 0.3410 0.06; 0.3216 0.05; 0.3855 -0.04; 0.5233
Education 0.10; 0.0574 0.10; 0.1040 0.10; 0.0550 -0.07; 0.2079
82 Cortical thickness -0.08; 0.2342 -0.07; 0.3574 -0.08; 0.2104 0.03; 0.6393
Hippocampal vol. -0.05; 0.5035 -0.16; 0.0741 -0.00; 0.9772 -0.09; 0.2914
Ventricular volume 0.04; 0.5349 0.11; 0.1534 0.01; 0.8824 0.11; 0.1174
Volume of WML 0.36; <0.0001 0.20; 0.0108 0.42; <0.0001 -0.51; <0.0001
Total WM volume -0.14; 0.0323 -0.02; 0.7400 *-0.19; 0.0033 **0.23; 0.0005
Translation motion -0.15; 0.1322 -0.29; 0.0100 -0.08; 0.4058 -0.16; 0.1094
Rotation motion 0.40; <0.0001 0.48; <0.0001 0.36; 0.0003 -0.16; 0.1068
**
All continuous variables were standardized prior to applying the model for easier comparison of parameter estimates (P). Uncorrected p-values are presented and significant associations with corrected p < 0.05 are bolded (*, ** and *** for corrected p < 0.05, 0.01 and 0.001 respectively). Associations with uncorrected p < 0.05 are italicized. (WML: white matter lesions;
WM: white matter).
3.5 Discussion
We observed in this study that individuals with AD had greater volumes of WML than non- demented older adults as demonstrated in prior work. Additionally, we showed that the diffusion values within WML differed between groups, though not significantly after correction for multiple comparisons. Parahippocampal WM, which may be more likely to undergo changes secondary to medial temporal cortex neurodegeneration as part of classical AD, showed consistent group differences for all diffusion metrics, which were most correlated with
83 hippocampal volume. Diffusion measures of WML correlated instead significantly with ventricular, WML and total WM volumes independently.
The current results demonstrate the need for better understanding of the increase in ventricular and WML volumes observed in MCI and AD. Indeed, ventricular enlargement better related to the diffusion measures in WML than any other variable including volume of WML and total WM volume, and was mainly related to an increase in average diffusivity. Diffusivity of
WML also seemed to be increased with decreased volume of WML; however, this effect was apparent only as an independent effect from ventricular volume and total WM volume, which were also included in the general linear model. Without those covariates, greater volume of
WML indeed related to greater diffusivity within WML (not shown). This result suggests that both ventricular volume and volume of WML might be related to the same variance in diffusivity measures within WML and Chapter 4 investigates this possibility by using factor analysis on the neuroimarging markers to identify the independent underlying processes that may explain the covariation between neuroimaging markers in this population. It is still interesting to note that some residual variance of WML diffusivity was accounted for by WML volume as an independent effect from this potential common factor, suggesting possibly two types of WML such as mild and diffuse WML and more severe, localized WML. Similarly, both total WM and ventricular volumes had independent associations with diffusivity of WML, while they are largely considered in the literature to represent the same underlying process - white matter atrophy. While there was an association between FA of WML and regional cortical thickness, it was in the opposite direction than would be expected in the case of a neurodegenerative effect, i.e. increased FA was associated with reduced cortical thickness.
84 Furthermore, no association was found between the diffusion properties of WML and hippocampal volume. Overall, these results suggest that the presumed neurodegenerative effects of AD are not directly related to the microstructural changes of the more prevalent
WML observed in AD. We further ascertain these results in Chapter 4 as described earlier by factoring the neuroimaging markers and making sure that the general lack of associations observed between diffusion measures of WML and both hippocampal volume and regional cortical thickness is not due to the fact both markers account for the same variance and therefore cannot relate to diffusivity parameters independently of each other.
Interestingly, group differences in diffusion metrics of WML could have been expected given the group differences in ventricular volume and volume of WML and their associations with diffusion metrics of WML. However, this is likely due to some degree to the residual variance not explained by the correlation between diffusion metrics of WML and both ventricular and
WML volume and to group differences in diffusion metrics of WML being simply slightly under significance threshold. Remarkably, diffusion metrics of WML, parahippocampal white matter and normal-appearing white matter tracked better with neuroimaging markers than with group differentiation, despite the fact those imaging markers were affected with disease. Indeed, group determination was not found to be significant in any model, even despite significant group differences in parahippocampal white matter diffusion measures. Furthermore, there were no significant group by imaging marker interactions when they were included in any of the models, i.e. the associations between imaging markers and diffusion metrics did not differ significantly between groups. This suggests that the diffusion properties of the white matter regions investigated relate very well to the processes affecting the neuroimaging markers
85 included in this study, and further suggests those neuroimaging markers provide a comprehensive overview of the processes affecting the white matter in AD, with the caveat that the study design is cross-sectional.
One phenomenon we speculate may account for the strong relationship found in this study between increasing ventricular volume and diffusivity of WML, independently of decreasing total WM volume, is the denudation of the ventricular ependyma, which is severe in AD
(Scheltens et al., 1995) and may permit leakage of cerebrospinal fluid into the WM tissue as suggested in prior work (Scheltens et al., 1995; Fazekas et al., 1996). Other potential causes of water dysregulation have been recently investigated in the context of AD and could account for the association between the diffusivity in WML and ventricular volume (Brinker et al., 2014). In particular, increased or dysregulated aquaporin expression of the subependymal cells and other cells lining the lateral ventricles (Moftakhar et a/., 2010) as well as blood-brain barrier disruption (Anderson et ui., 2011; Taheri et ai., 2011) may also lead concurrently to ventricular expansion and the formation of edema in WML (Zlokovic, 2011). It is possible that overall dysfunction of the ventricular lining might be a precipitating factor or provide a 'second hit' to more classical AD neurodegenerative processes in the development of clinically diagnosed AD.
However, clinical manifestation of AD in the relative absence of WML has been reportedly observed and therefore such lesions may not be a necessary component of the disorder.
To our knowledge, this is the first report that takes into account the diffusion properties of the normative anatomy. Specifically, diffusion values within WML were strongly dependent on the normative values which varied with the underlying anatomy (e.g. in regions with single straight
86 fibers vs crossing fibers), underlying the importance of considering these values when calculating the degree of tissue damage within WML. This importance is evidenced in Figure 3.4 where it can be observed that the relationship between mean diffusivity and normalized ventricular volume is much greater when taking into account the location of the lesions.
'C 2.2 u) 0.6 r E * 20 LE0.4 o.8 go 000V 1.6 0.2
M ';= 1.4 r 0.456 = 0.0 r 0.708 :E1.2 P < 0.001 M 1.0 . -0.2 $ 0 0.02 0.04 0.06 0 0 0.02 0.04 0.06 Normalized ventricular Normalized ventricular volume (% ICV) volume (% ICV) Figure 3.4 Relationship between normalized ventricular volume and mean diffusivity of the white matter lesions without and with taking into account the location of the lesions. The relationship using the difference between the mean diffusivity in the lesions and the mean diffusivity in a non-lesioned normative brain is shown in addition to the relationship with the absolute mean diffusivity of lesions. Controls, individuals with mild cognitive impairment and Alzheimer's disease are shown respectively in white, light gray and dark gray. The analyses were limited to individuals with a volume of lesions greater than 1% total WM volume. (ICV: intracranial volume) There was also a difference in the sex distribution across the three groups investigated in this
study. While sex was not a significant factor in any model, the diffusivity values in the
parahippocampal white matter differed between males and females even after correcting for
age, education and motion measures, and this will need to be investigated in future work.
There were no sex differences in either white matter lesions or normal-appearing white matter.
In the context of this thesis, the current study indicates that WML do not appear to have
different microstructural properties in individuals with MCI and AD compared to non-demented
87 controls. Furthermore, it provides no evidence that classical markers of neurodegeneration such as hippocampal volume and regional cortical thickness account for any variance in the diffusion properties of WML independently of ventricular volume, suggesting that those WML are unlikely the result of Wallerian degeneration and that they are more likely related to an independent process in the pathogenesis of AD that involves ventricular enlargement and white matter atrophy. The next chapter focus on ascertaining those results as described earlier in the discussion by determining whether there exist independent underlying processes or factors in
AD and by understanding which processes are related to the effect on white matter.
3.6 Appendix
3.6.1 Effect of APOE genotype in the models
The associations remained largely the same between the diffusion measures of WML, parahippocampal and normal-appearing WM and the neuroimaging markers when adding
APOE genotype in the models. However, it is interesting to note that the number of APOE E4 alleles was associated with reduced diffusivity in WML. The bivariate correlation (without any other covariates) is also significant with the same direction for the association (not shown). This association remains unexplained as the presence of APOE E4 alleles is considered to have a negative overall effect on the brain by increasing risk for AD and the expectation in that regard would be that more APOE E4 alleles would be linked to increased diffusivity in WML. This is not the case here and further study is needed.
Table 3.5 Models of the diffusion parameters in WML, parahippocampal and normal-appearing
WM with all neuroimaging markers and including the APOE genotype as a variable.
88 WML (subgroup with volume > 1% total WM volume, APOE genotype, n = 95)
Parameters M D (B; p-value) DA (1; p-value) DR (1; p-value) F (1; p-value)
Group (MCI) 0.01; 0.8758 0.03; 0.6721 -0.00; 0.9920 0.09; 0.3802
Group (AD) 0.18; 0.0514 0.14; 0.1317 0.20; 0.0331 -0.22; 0.1077
Age 0.04; 0.4161 0.02; 0.6485 0.05; 0.3107 -0.15; 0.0638
Sex (female) 0.02; 0.6687 0.01; 0.8442 0.03; 0.5656 -0.09; 0.2792
Education 0.06; 0.2429 0.04; 0.4547 0.07; 0.1719 -0.13; 0.0815
APOE4 (per allele) *-0.23; 0.0039 *-0.25; 0.0022 -0.22; 0.0071 -0.05; 0.6673
Cortical thickness 0.04; 0.5646 -0.01; 0.8686 0.06; 0.3212 *-0.27; 0.0034
Hippocampal vol. 0.11; 0.1389 0.10; 0.1651 0.11; 0.1377 -0.13; 0.2357
Ventricular volume 0.57; <0.0001 0.60; <0.0001 0.54; <0.0001 0.09; 0.2860
Volume of WML -0.21; 0.0162 -0.20; 0.0195 -0.22; 0.0150 -0.06; 0.6055
Total WM volume -0.12; 0.0290 -0.12; 0.0352 -0.12; 0.0304 0.09; 0.2553
Norm. properties 0.58; <0.0001 0.57; <0.0001 0.59; <0.0001 0.54; <0.0001
Translation motion *0.30; 0.0031 0.22; 0.0259 *0.34; 0.0010 *-0.46; 0.0020
Rotation motion -0.27; 0.0053 -0.21; 0.0273 *-0.30; 0.0025 0.31; 0.0273
Parahippocampal WM (subgroup with APOE genotype, n = 169)
Parameters MD (P; p-value) DA (1; p-value) DR (P; p-value) FA (0; p-value)
Group (MCI) -0.06; 0.5322 0.02; 0.8029 -0.09; 0.2743 0.13; 0.1244
89 Group (AD) 0.09; 0.5428 -0.03; 0.8377 0.14; 0.2970 -0.23; 0.0940
Age 0.03; 0.7199 -0.01; 0.9273 0.05; 0.5503 -0.10; 0.2164
Sex (female) 0.02; 0.8113 -0.04; 0.5837 0.05; 0.4932 -0.15; 0.0365
Education 0.10; 0.1641 0.05; 0.5570 0.12; 0.0784 -0.13; 0.0441
APOE4 (per allele) 0.14; 0.1976 0.06; 0.5995 0.17; 0.1010 -0.17; 0.0856
Cortical thickness -0.22; 0.0143 -0.16; 0.1022 -0.24; 0.0060 0.20; 0.0181
Hippocampal vol. -0.25; 0.0184 -0.22; 0.0555 -0.25; 0.0146 0.18; 0.0657
Ventricular volume -0.01; 0.9290 -0.06; 0.5561 0.02; 0.8308 -0.03; 0.7256
Volume of WML 0.11; 0.2120 0.11; 0.2669 0.11; 0.2163 -0.09; 0.3014
Total WM volume -0.04; 0.6445 0.08; 0.3985 -0.09; 0.2316 *0.24; 0.0025
Translation motion -0.03; 0.8415 0.06; 0.6677 -0.07; 0.5776 0.14; 0.2762
Rotation motion 0.15; 0.2346 0.05; 0.7389 0.20; 0.1091 -0.24; 0.0481
Normal-appearing white matter (subgroup with APOE genotype, n = 169)
Paramete rs MD (P; p-value) DA (1; p-value) DR (P; p-value) FA (1; p-value)
Group (M Cl) 0.06; 0.4806 0.13; 0.1583 0.02; 0.7726 0.07; 0.3592
Group (A D) -0.03; 0.8115 -0.05; 0.7202 -0.02; 0.8723 0.02; 0.8967
Age 0.08; 0.2896 0.13; 0.1263 0.05; 0.4562 0.02; 0.8026
Sex (fema le) 0.03; 0.6492 0.03; 0.6882 0.03; 0.6496 -0.02; 0.8062
Educatior 0.12; 0.0654 0.11; 0.1330 0.12; 0.0566 -0.08; 0.2076
APOE4 (p er allele) 0.01; 0.9015 0.02; 0.8180 0.01; 0.9524 0.01; 0.9120
Cortical thickness -0.13; 0.1182 -0.13; 0.1328 -0.12; 0.1347 0.05; 0.5490
90 Hippocampal vol. 0.03; 0.7405 -0.08; 0.4598 0.08; 0.3840 -0.16; 0.0976
Ventricular volume 0.05; 0.5436 0.10; 0.2563 0.02; 0.7712 0.09; 0.2716
Volume of WML 0.39; <0.0001 0.23; 0.0123 0.45; <0.0001 0.52; <0.0001
Total WM volume -0.14; 0.0545 -0.02; 0.7581 -0.19; 0.0084 *0.25; 0.0011
Translation motion -0.15; 0.2049 -0.31; 0.0209 -0.07; 0.5247 -0.21; 0.0861
Rotation motion **0.41; 0.0006 **0.48; 0.0002 *0.35; 0.0018 -0.14; 0.2327
All continuous variables were standardized prior to applying the model for easier comparison of parameter estimates (P). Uncorrected p-values are presented and significant associations with corrected p < 0.05 are bolded (*, ** and *** for corrected p < 0.05, 0.01 and 0.001 respectively). Associations with uncorrected p < 0.05 are italicized. (WML: white matter lesions;
WM: white matter).
3.6.2 Study limitations
The current work is limited in that the findings are cross-sectional and do not provide information about the mechanisms of the associations reported. Follow-up longitudinal and interventional work would be valuable to determine whether these associations continue to track with time and whether a therapeutic reduction in one type of change is followed by a reduction in one or more of the associated markers. Another limitation is the possible inclusion of lacunar infarcts in the segmentation of WML as those are also hypointense on Tr-weighted imaging. However, they have lower prevalence and much lower volume than more common
WML identified as WM hyperintensity on T 2-weighted and FLAIR imaging. While lacunar infarcts
91 may be responsible for the relationship between decreasing volume of WML and increasing diffusivity of WML, this association was independent of the key association between greater ventricular volume and greater diffusivity of WML.
92 93 Chapter 4: Volume of white matter lesions and ventricular and hippocampal volumes cluster together as one of two distinct processes in Alzheimer's disease
4.1 Overview
White matter lesions were shown to have similar diffusion properties both in non-demented aging and in Alzheimer's disease in the previous chapter. Furthermore, strong associations were observed between diffusivity of white matter lesions and ventricular volume, volume of white matter lesions and total WM volume. Given the high covariation between volume of white matter lesions, total WM volume, ventricular volume as well as hippocampal volume and regional cortical thickness, factor analysis was used in this chapter to identify any distinct disease pathways that may lead to this covariation among these neuroimaging markers. The factor analysis suggested two independent sets of covarying degenerative changes, with one factor strongly linked to aging, cerebral blood flow, ventricular expansion and both volume and tissue properties of white matter lesions, while the other factor related to classical patterns of cortical and hippocampal neurodegeneration in AD. These relationships are described in this
94 chapter using two large independent datasets of cognitively healthy older adults and individuals with mild cognitive impairment and AD, one with diffusion tensor imaging data and one with arterial spin labeling MRI data. The impact of these two potentially distinct classes of degenerative change in AD on longitudinal cognitive decline is investigated in the next chapter.
4.2 Introduction
Several neuroimaging markers are affected by Alzheimer's disease (AD). However, it is unclear to what extent those neuroimaging markers represent the same underlying process. Figure 4.1 shows the scatterplots of the interrelationships between the AD signature cortical thickness, hippocampal volume, ventricular volume, total white matter volume and volume of white matter lesions in the sample of non-demented controls and individuals with mild cognitive impairment (MCI) and AD described in the last chapter. Most of these correlations are extremely significant, suggesting that there may be only one underlying process responsible for driving the changes in those neuroimaging markers. Most of these associations have previously been observed. The volume of white matter lesions, for instance, has been correlated with gray matter atrophy and pathology in individuals with MCI and AD (Capizzano et al., 2004; Moghekar et al., 2012; Fujishima et al., 2014). Ventricular enlargement has generally been considered a consequence of brain atrophy, and more specifically of white matter atrophy. However, few studies have considered those neuroimaging markers together and there has been a focus on using hippocampal volume or some measure of gray matter atrophy alone in studies of AD, while other neuroimaging markers may have been more correlated with the object of study.
95 -2. . 2 e 0 .2 * P<0.001 * p<0.001 **. . *,. p<0.001 2 *e
Log-Normalized .3 -3 WMC Volume 4 4 . * 0 Z_J (log of %total 40 5.O. *d WM volume) e * -62. . 2 p<0.0010 , 40-. 2. 2 a 0n24 6 0.30Aea.7 20 40 401 40 0 00 1 r = -0.46 0 P< . 4 p<0.001 2S . Normalized WM Volume (% ICV)
20 . 2. 201 2.0 24 2.8 01 0.4 00 0 07 .4 5 4 .3 -2 0 2 4 6 2p<0.0 2 = -0.49 6 5 * p<0.001 0 r 0.52 e r*.. 0.2 , . . . r 30 V 4, 0410 . * e Normalized 0 Ventricular 25 - 'a . p<0.001 2 Volume (% ICV) 2 .3 0. . . 2 2* o
20 24 208 .1 .4 . 24 20 2s 30 3s 40 oAs*A .7" e 4 * 0 E r =-0.51 r 0.46 r =-0.55 0.s D,6 Normalized Hippocampal 04 0. Volume (% ICV) 03.4O 0.2
r4 0 4 -2 - 20 0 30 20 40 4 2 4 0
1 2.8 r =-0.40 2. 2.8 r -0.43 2. AD Signature 2.06. 2.4 2.- 2. Thickness (mm) 2o -20 r 0.52 20r 2. -* 0.20 2. 1 a 2 4 6 0. 0.4 0.0 s 0-7 -5 -4 3 -2 20 20 20 20 40 4 Volume AD Signature Cortical Thickness Log-Normalized WIML Volume Normalized WM Volume Normalized Ventricular Volume Normalized Hippocampal 0 Control U Mild cognitive impairment 0 Alzheimer's disease
Figure 4.1 Scatterplots of volume of white matter lesions, total WM volume, ventricular volume, hippocampal volume and AD signature cortical thickness displayed as a correlation matrix. Hippocampal, ventricular and total WM volumes were normalized by the estimated total intracranial volume. The natural logarithm of the volume of white matter lesions divided by the total WM volume was used. Pearson's correlation coefficients and their associated p-values are shown on opposite sides of the diagonal. (WMC: white matter lesions; WM: white matter; eTIV: estimated total intracranial volume; AD: Alzheimer's disease) Despite those correlations suggesting a single underlying process, there is also evidence in the
literature of two or more distinct processes that may influence those neuroimaging markers.
For instance, one process is generally attributed to classical AD neurodegenerative changes,
and co-morbid vascular disease has been suggested as a distinct, additive process.
Neuropathological staging has recently been designed to evaluate this vascular component of
96 the disease, and found that it was not correlated with classical AD neuropathological staging based on AP immunohistochemistry, Braak staging and frequency of neuritic plaques (Zea-
Sevilla et al., 2015). Cerebral blood flow as measured on MRI with arterial spin labeling has also been shown to be reduced regionally and globally in individuals with AD compared to non- demented controls (Alsop et a/., 2000; Dai et al., 2009; Schuff et al., 2009; Binnewijzend et a/.,
2013; Mattsson et al., 2014). However, it is unclear whether this may be a consequence of primary AD neurodegenerative processes leading to cell death and reduced perfusion demand, part of a distinct process that may involve vascular disease and lead to cognitive decline, or both. For example, lower brain volume and greater white matter lesion volume have been independently associated with reduced global and cortical CBF in individuals with AD
(Benedictus et al., 2014).
The models tested in the last chapter are also supportive of more than one process or mechanism underlying the changes in the neuroimaging markers; independent, significant associations of neuroimaging markers with diffusion measures of white matter integrity were found even when all other highly correlated neuroimaging markers were included in the models. These independent associations suggest the possibility of different, separate processes.
Indeed, if a single primary AD process was responsible for the variation in those five
neuroimaging markers and the imperfect correlations between markers was due to natural variation or other noise source of biological or technical origin, it would have been expected that no single neuroimaging marker would be related to the measures of white matter integrity
independently from other markers.
97 Therefore, we performed factor analysis on these five neuroimaging markers - AD signature cortical thickness, hippocampal volume, ventricular volume, total white matter volume and volume of white matter lesions - in two independent samples of non-demented controls and individuals with MCI and AD in order to determine whether there are one or more underlying processes driving the correlations between these neuroimaging markers. Two distinct factors or processes were found and replicated, and their relationships with disease status, age, cognitive ability, diffusion measures of white matter integrity and measures of cerebral blood flow were assessed. This also enabled us to confirm the analyses from the previous chapter and account for the fact that variance in diffusion metrics might be explained by a process that is not unique to any individual marker.
4.3 Materials and Methods
4.3.1 Participants and MRI acquisition
Two distinct datasets were used in this chapter. The first dataset was fully described in the previous chapter in section 3.3.1. The reader is also referred to sections 3.3.2 through 3.3.5 for complete details about the preprocessing of this dataset, including the diffusion data processing, the automated subcortical and white matter lesion segmentation, the registration procedures and the normative data calculation. The second dataset consists of another large publicly-available dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI, http://adni.loni.usc.edu) which included 113 controls, 159 participants with MCI and 92 participants with AD who underwent whole-brain MRI scanning at one or multiple visits on a 3-
Tesla Siemens scanner and had sagittal T-weighted 3D spoiled gradient echo images and
98 pulsed arterial spin labeling (ASL) images (1 reference Mo image, 52 tagged-untagged pairs) available at the time of download. One participant overlapped both datasets, though the data for that subject were obtained independently four years apart on different scanners. For individuals with multiple available ASL datasets, we picked the most recent that did not fail the
ADNI quality assurance test. These datasets were acquired using previously described ADNI
Core MRI and DTI protocols (Jack et al., 2008). Group designation of control, MCI and probable
AD was determined by ADNI based on the criteria of the National Institute of Neurological and
Communicative Diseases and Stroke - Alzheimer's Disease and Related Disorders Association
(McKhann et al., 1984). Participants enrolled as normal or with significant memory concern and with a Clinical Dementia Rating (Morris, 1993) of 0 were grouped together into the control group, and participants enrolled as early and late MCI were combined into one MCI group (see
ADNI 2 Procedures Manual on www.adni-info.org for more information). Clinical profiles and diagnostic information were obtained from the assessment closest in time to the MRI acquisition. Demographics are provided in Table 4.1. Written informed consent was obtained from all participants or their representatives through ADNI. The study procedures were approved by institutional review boards of all participating institutions.
Table 4.1 Demographics for the second dataset with arterial spin labeling MRI.
All ADNI, n = 364
CN MCI AD p-value
Participants (female) 113 (67) 159 (69) 92 (40) 0.0194
Age [years] 74.52 (0.67) 73.96 (0.60) 75.15 (0.83) 0.4790
99 Education [years] 16.62 (0.25) 16.12 (0.22) 16.14 (0.27) 0.2743
MMSE [-]a 29.08 (0.11) 27.70 (0.18) 21.56 (0.49) <0.0001
APOE E4 [# alleleS]b 0.29 (0.05) 0.52 (0.05) 0.89 (0.08) <0.0001
Translation motion [mm] 0.25 (0.01) 0.28 (0.02) 0.33 (0.02) 0.0257
Rotation motion [millidegrees] 1.7 (0.1) 1.9 (0.2) 2.5 (0.2) 0.0031
All significant p-values are bolded. Standard errors are shown in parentheses. (MMSE: Mini-
Mental State Exam; CN: control; MCI: mild cognitive impairment; AD: Alzheimer's disease) a. Information missing for 11 CN, 14 MCI and 7 AD. b. Information missing for 1 CN, 1 MCI and 1 AD.
4.3.2 Automated subcortical and white matter lesion segmentation
Automated subcortical and WM segmentation as well as cortical surface reconstruction were obtained from the T-weighted images using FreeSurfer (https://surfer.nmr.mgh.harvard.edu)
(Fischl et al., 1999, 2002), as described in section 3.3.3. Normalized WM volume, ventricular volume (lateral ventricles), hippocampal volume as well as the AD signature cortical thickness and the natural logarithm of the volume of white matter lesions divided by total WM volume were also obtained as described in section 3.3.3.
4.3.3 Arterial spin labeling data processing
The ASL dataset was corrected for 3D head motion and translation and rotation motion
estimates were obtained from the registration matrices (Yendiki et al., 2014). The tagged and
untagged images were first averaged and subtracted from each other to obtain a perfusion-
weighted image. This perfusion-weighted image (PWI) was then corrected for the longitudinal
100 relaxation of the spin tracers during data acquisition, including time lags between consecutive image slices and incomplete labeling, as described in the ADNI UCSF ASL perfusion processing methods:
PwI PWItag-corr 2aTI1 exp(Ria(T1 2 + (n - 1)r)) where a = 0.95 is the average tagging efficiency, Ria = 1/1684 kHz is longitudinal relaxation of
blood, TI, = 700 ms is the inversion time of arterial spins and TI 2 = 1900 ms is the total transit time of the spins, n is the slice number andT= 22.5 ms is the time lag between slices. In order to obtain cerebral blood flow (CBF) maps, this corrected PWI needs to be scaled by the water to blood density, as described in the ADNI UCSF ASL perfusion processing methods:
Mo Mblood =- exp(TE(R2tissue - R2blood))R
where A = 0.90 is the average blood/water ratio in tissue, R*2tissue = 1/44 kHz and R*2bIlod = 1/43 kHz are relaxation rates and TE is the blood relaxation time and may be either 12 ms or 13 ms depending on the scanner used to acquire each dataset. The CBF maps in ASL native space are finally obtained by dividing the corrected perfusion-weighted image PWilag-corr by the water to blood density MbIood and by multiplying by a factor of 6000 to convert from CBF units of milliliters per gram of tissue per second to standard CBF units of milliliters per 100 grams of tissue per minute. The model and parameters used here are also described in several publications (Buxton et al., 1998; Wong et al., 1998; Luh et al., 1999; Zhao et al., 2007; Schuff et al., 2009). Overall, while some the parameters in the model are estimates and may be inexact, they only act as a scaling factor to obtain quantitative CBF units and will not influence the
101 statistical testing of associations provided all datasets are acquired and processed in the same manner. This would not necessarily be true in models more complex than the one used in this study.
The geometric transfer matrix method was chosen for correction of partial-volume effects as it was recently found to be preferable to other methods for partial-volume correction (Greve et al., 2014). This method provides estimates of the regions-of-interest intensity averages using the linear least-squares estimate of the fit between the tissue fraction corresponding to that region and the image intensity for all voxels (Rousset et al., 1998). The anatomical segmentation was obtained from the Ti-weighted image using FreeSurfer. FreeSurfer boundary-based registration was used to register the anatomical T-weighted image to the ASL space (MO image) for each participant (Greve and Fischl, 2009). Using the resulting registration matrix, each anatomical region-of-interest was then interpolated into a regional tissue fraction
iaHip iii LI HAL sHLa aiiu the I t:Ial-vLIuIIIC II ICLIIOn was appIIeUd tUo Uobta ii anesImTdLe Of the intensity of each region-of-interest. Average CBF was finally obtained for cortical gray matter and white matter, as well as globally throughout the brain for each participant, using the average of the intensity estimates of the regions-of-interest included in those structures.
4.3.4 Factor analysis
Given the interrelationship of AD signature cortical thickness, volume of white matter lesions and hippocampal, ventricular and total WM volumes, factor analysis was performed in the first dataset, described in the previous chapter, to obtain the primary factors representing the different sources of covariation within these five neuroimaging markers of AD. The maximum
102 likelihood factoring method was used with the Varimax rotation method. Prior communality estimates were set to the squared multiple correlation coefficients of each neuroimaging marker with all other markers. Factors with an eigenvalue > 1.0 were considered significant or sufficient to represent the latent data structure. Factor scores were also obtained for each individual in the second dataset with ASL data described in this chapter using the standardization matrix and standard score coefficients extracted from the factor analysis. The same factor analysis was also replicated in the second dataset and factor scores from both analyses were compared across the individuals of the second dataset.
4.3.5 Statistical analyses
Statistical analyses were performed using JMP 10 statistical software (SAS Institute Inc., Cary,
NC, U.S.A.). As in the previous chapter, the factors were used in general linear models to understand their associations with the diffusion metrics and global and regional cerebral blood flow using age, sex, education (Teipel et al., 2009) and motion measures (average translation and average rotation (Yendiki et al., 2014)) as covariates. Group by factor interactions were not
included as they were not significant when added to the models. Results related to diffusion analysis were corrected for multiple comparisons with Bonferroni (3 WM regions / comparisons for group differences in volume and 3 WM regions x 4 diffusion metrics = 12 primary
comparisons for all results involving DTI) and estimated parameters were provided in the
models in addition to p-values to ease interpretation.
103 4.4 Results
4.4.1 Classes of degenerative change
Factor analysis in the first dataset yielded two significant factors (Table 4.2). Both factors showed a high loading from hippocampal volume. Factor 1 otherwise included high loadings (>
0.4) from volume of white matter lesions, total WM volume and ventricular volume, reflecting processes that are demonstrated in prior work to change with age and vascular disease in particular for volume of white matter lesions (Breteler et al., 1994; Jeerakathil et al., 2004;
Murray et al., 2005; Stenset et al., 2006; Yoshita et al., 2006; Gottesman et al., 2010; Pantoni,
2010; Gouw et al., 2011; Rostrup et al., 2012). For this reason and others detailed in this chapter, Factor 1 was interpreted as the 'age- and vascular-associated' factor. Factor 2 included the AD signature cortical thickness in addition to the hippocampal volume and therefore represented processes that are often used as imaging estimates of neurodegenerative changes in AD. This factor was therefore interpreted as the 'neurodegenerative' factor.
Factor analysis in the second dataset also yielded two significant factors (Factors 1' and 2') which had very similar loadings as the factors from the first factor analysis (Factors 1 and 2).
Indeed, the same rank ordering of loadings was found for each factor in this replication sample.
Factor scores from the first factor analysis (Factors 1 and 2) were obtained for each individual of the second dataset and compared with factor scores from the second factor analysis (Factors
1' and 2'). Scatterplots of the relationship between Factors 1 and 1' and between Factors 2 and
2' are shown in Figure 4.2, showing a very high correlation between factor scores of both factor analyses in the same individuals.
104 Table 4.2 Factor analysis of highly correlated neuroimaging markers in Alzheimer's disease and in the first dataset and its replication in the second dataset.
First dataset Second dataset
Parameters Factor 1 Factor 2 Factor 1' Factor 2'
Volume of WML -0.642 -0.270 -0.708 -0.321
Total WM volume 0.694 0.052 0.825 0.190
Ventricular volume -0.685 -0.288 -0.792 -0.270
Hippocampal volume 0.630 0.405 0.557 0.613
Cortical thickness 0.225 0.946 0.164 0.699
Coefficients higher than 0.40 are bolded to indicate the most important markers contributing to
each significant factor. (WML: white matter lesions; WM: white matter)
2 2 r a T x 1 1
0 0 -1 0 -1 09 U U (U U -2 LL r = 0.983 0b~j r = 0.849 -2 p < 0.001 p < 0.001 -3 -3 peooo 0
-4 -3 -2 -1 0 1 2 -4 -2 0 2 4 Factor I (-) Factor 2 (-) Figure 4.2 Scatterplots of the correspondence between the factor scores from both factor analyses. Pearson's correlation coefficients and the associated p-value are shown. Controls, individuals with mild cognitive impairment and Alzheimer's disease are shown respectively
105 Both factors were altered in individuals with AD compared to controls using factor scores from either factor analyses or samples. Factor 1 also had a stronger age effect and a weaker MMSE effect than Factor 2. While this is displayed for factor scores obtained with the first factor analysis in the first dataset in Figure 4.3, those results were also found using factor scores of any of the two factor analyses and in any of the two samples. For simplicity, and given the extreme similarity of these two factor analyses, we will use factor scores from the first factor analysis in the remaining of the thesis, unless otherwise noted.
4.4.2 Associations with tissue properties of white matter lesions
Following up on the work of Chapter 3, factor scores found in the previous section were used instead of the neuroimaging markers to confirm the previous associations but also to determine the relationship between the factors and the properties of white matter lesions, parahippocampal WM and the normal-appearing WM. All details of the statistical models are provided in Table 4.3.
First, we tested whether diffusion metrics of white matter lesions were associated with any or both factors to determine if they would be related to Factor 1 and unrelated to the Factor 2.
We found strong significant associations between diffusivity of white matter lesions and Factor
1 (MD, DA, DR: corrected p < 0.001) while DA and FA of lesions showed an association with
Factor 2 (DA: corrected p < 0.01; FA: corrected p < 0.001). Second, we tested in comparison whether parahippocampal WM diffusion metrics were associated with any or both factors to determine if they would be mainly related to Factor 2. However, significant associations were found between parahippocampal WM diffusion metrics and both Factor 1 (MD, DR, FA:
106 0 A) 0-6 0.4- 0..4 0.2 0.2 0 (~4 0 bin. 0 U -0 (U *0 U.-* ANOVA U- ANOVA -0.8 p < 0.001 *p < 0.001 CN MCI AD CN MCI AD 3 B) 2 2 1
0 0
-1 -1 U. -2 U. -2 -0A -3 -3 r -0.181
-4 p < 0.001 -4 .p < 0.01 50 60 70 80 90 50 60 70 80 90 Age (years) Age (years) 3 C) 0.295 r = 0.553 2 r = 2 P < 0.001 e p < 0.001 * C4 0 p 0 -1 .1 4.. U .2 (U -2 LL LL -3 -3 -4 -4 15 20 25 30 15 20 25 30 MMSE (-) MMSE (-) Figure 4.3 Factor scores in relation to a) group, b) age and c) mini-mental state examination (MMSE). Controls (CN), individuals with mild cognitive impairment (MCI) and Alzheimer's disease (AD) are shown respectively in white, light gray and dark gray. The presence of group differences was statistically confirmed using ANOVA and the Steiger z test confirmed that Factor 1 had a stronger age effect (p <0.001) and a weaker MMSE effect (p < 0.01) than Factor 2. Standard error bars are shown. corrected p < 0.001) and Factor 2 (DR, FA: corrected p < 0.01; MD: corrected p < 0.05), independently of each other. Finally, we tested whether NAWM diffusion metrics were associated with any or both factors to determine if they would be mainly related to Factor 1 107 and minimally associated with Factor 2. We found significant associations between all diffusion measures in NAWM and Factor 1 (MD, DR and FA: corrected p < 0.001; DA: corrected p < 0.01) and weaker associations between NAWM diffusivity and Factor 2 (MD, DR: corrected p < 0.05). Factor 2 had an effect size on diffusivity values averaging less than half the effect size of Factor 1. Of note, group determination was not found to be significant in any model. There were no significant group by imaging marker interactions when they were included in any of the models. Table 4.3 Models of the diffusion parameters in white matter lesions, parahippocampal and normal-appearing WM with factors extracted from the neuroimaging markers. White matter lesions (subgroup with volume > 1% total WM volume, n = 118) Parameters MD (P; p-value) DA (@; p-value) DR (P; p-value) FA (P; p-value) Group (MCI) 0.01; 0.8798 0.02; 0.7787 0.00; 0.9526 0.02; 0.8207 Group (AD) -0.05; 0.6085 -0.10; 0.2634 -0.01; 0.9043 -0.25; 0.0296 Age -0.00; 0.9990 -0.02; 0.7591 0.01; 0.8300 *-0.22; 0.0032 Sex (female) 0.05; 0.3867 0.05; 0.3649 0.05; 0.3834 -0.12; 0.0967 Education 0.13; 0.0202 0.11; 0.0499 0.14; 0.0129 -0.11; 0.1103 Factor 1 -0.50; <0.0001 -0.52; <0.0001 -0.47; <0.0001 -0.10; 0.2073 Factor 2 -0.17; 0.0050 -0.23; 0.0003 -0.14; 0.0262 -0.32; <0.0001 ** Norm. properties 0.88; <0.0001 0.85; <0.0001 0.88; <0.0001 0.57; <0.0001 108 Translation motion 0.06; 0.5509 -0.02; 0.8665 0.11; 0.2917 **-0.49; 0.0003 Rotation motion -0.12; 0.2770 -0.06; 0.5828 -0.15; 0.1672 0.29; 0.0360 Parahippocampal WM (all, n = 215) Parameters MD (1; p-value) DA (s; p-value) DR (0; p-value) FA ($; p-value) Group (MCI) -0.06; 0.4451 0.01; 0.9214 -0.09; 0.2168 0.12; 0.0989 Group (AD) 0.24; 0.0390 0.14; 0.2821 0.28; 0.0121 -0.28; 0.0098 Age 0.06; 0.3914 0.03; 0.7563 0.08; 0.2599 -0.14; 0.0416 Sex (female) -0.05; 0.4092 -0.11; 0.1036 -0.02; 0.7927 -0.12; 0.0294 Education 0.04; 0.5012 0.01; 0.9297 0.06; 0.3269 -0.10; 0.0909 Factor 1 -0.32; <0.0001 -0.19; 0.0317 -0.36; <0.0001 0.35; <0.0001 Factor 2 -0.22; 0.0019 -0.13; 0.0991 -0.25; 0.0002 0.24; 0.0003 * ** ** Translation motion 0.04; 0.7069 0.10; 0.4426 0.01; 0.9173 0.07; 0.5302 Rotation motion 0.07; 0.5298 -0.04; 0.7845 0.12; 0.2564 -0.24; 0.0274 Normal-appearing white matter (all, n = 215) Parameters MD (1; p-value) DA (1; p-value) DR (1; p-value) FA (1; p-value) Group (MCI) 0.12; 0.0935 0.17; 0.0243 0.09; 0.2087 0.01; 0.8535 Group (AD) -0.14; 0.1812 -0.09; 0.4349 -0.15; 0.1297 0.16; 0.1469 Age 0.13; 0.0381 0.14; 0.0409 0.12; 0.0534 -0.08; 0.2365 Sex (female) 0.06; 0.2490 0.04; 0.5131 0.07; 0.1881 -0.08; 0.1667 109 Education 0.08; 0.1431 0.09; 0.1312 0.07; 0.1834 -0.03; 0.6176 Factor 1 -0.44; <0.0001 -0.35; 0.0003 -0.46; <0.0001 0.39; <0.0001 **** ***** Factor 2 *-0.19; 0.0023 -0.18; 0.0066 *-0.18; 0.0029 0.10; 0.1356 Translation motion -0.13; 0.1977 -0.27; 0.0166 -0.06; 0.5319 -0.19; 0.0917 Rotation motion 0.40; 0.0001 0.46; <0.0001 0.35; 0.0006 -0.16; 0.1407 ** *** ** All continuous variables were standardized prior to applying the model for easier comparison of parameter estimates (P). Uncorrected p-values are presented and significant associations with corrected p < 0.05 are bolded (*, ** and *** for corrected p < 0.05, 0.01 and 0.001 respectively). Associations with uncorrected p < 0.05 are italicized. (WM: white matter). 4.4.3 Associations with cerebral blood flow While cerebral blood flow has been reported to be reduced in AD compared to non-demented controls, it is unclear whether one or both factors and their underlying biological processes may be related to this loss of perfusion. Similar models as those presented in section 4.4.2 are shown in Table 4.4 for cerebral blood flow (CBF) extracted from the entire brain, as well as from subregions (cortical gray matter and white matter). Only Factor 1 was associated with global CBF and with CBF from both subregions. However, Factor 2 showed a trend for an association with global CBF. The strongest association found was between Factor 1 and the white matter CBF. 110 Table 4.4 Models of the cerebral blood flow from the entire brain, cortical gray matter and white matter with factors extracted from the neuroimaging markers. Parameters Global CBF Cortical GM CBF WM CBF (0; p-value) (@; p-value) (0; p-value) Group (MCI) -0.01; 0.8583 0.07; 0.3056 -0.09; 0.1666 Group (AD) -0.03; 0.7150 -0.10; 0.2413 -0.01; 0.9336 Age 0.02; 0.8047 0.01; 0.9367 0.04; 0.4659 Sex (female) ***0.22; <0.0001 **0.17; 0.0012 ***0.19; 0.0003 Education -0.05; 0.3179 -0.03; 0.6094 -0.06; 0.2090 Factor 1 **0.19; 0.0040 *0.14; 0.0414 ***0.28; <0.0001 Factor 2 0.11; 0.0559 0.05; 0.4052 0.04; 0.4832 Translation motion -0.05; 0.7363 0.04; 0.7959 -0.29; 0.0541 Rotation motion 0.16; 0.2898 0.12; 0.4426 **0.44; 0.0040 All continuous variables were standardized prior to applying the model for easier comparison of parameter estimates (1). Uncorrected p-values are presented and significant associations with p < 0.05 are bolded (*, ** and *** for p < 0.05, 0.01 and 0.001 respectively). Statistical trends were italicized (p < 0.10). (CBF: cerebral blood flow; GM: gray matter; WM: white matter). Scatterplots of the relationship between each factor and global, cortical gray matter and white matter CBF are shown in Figure 4.4. As displayed, some average CBF values were found to be low (0 to 10 ml/100g/min) and some were found to be negative, showing the limitations of the technique and the need for careful interpretation of these findings. 111 50 50 40 00 Q 40 301 E 00 30 W 20 20 IL 10 10 0 0 0 0 r = 0.201 a r = 0.114 p < 0.05 -10 p < 0.001 -10 -4 -3 -2 -1 0 1 2 -4 -2 0 2 4 Factor 1 (-) Factor 2 (-) U. C 70 r =0.147 ' 70 r =0.066 0 P < 0.01 % p 0.209 60 00 60 50 E 50 0 40 40 30 U- 30 0) 20 0~ 2 20 )0 0 10 02 10 L-1 0 0 0 0a E -4 -3 -2 -1 0 1 2 o -4 -2 0 2 4 Factor 1 (-) 0 Factor 2 (-) U 40 40 30 30 20 20 00 10 C 101 0 ~0 0) 0 09 -10 -10 kl -20 0 -20 r 0.013 r = 0.250 U -30 p < 0.001 -30 p =0.810 0 0 -40 I -40 -4 -3 -2 -1 0 1 2 -4 -2 0 2 4 Factor I (-) Factor 2 (-) Figure 4.4 Factor scores in relation to global, cortical gray matter and white matter cerebral blood flow. Controls, individuals with mild cognitive impairment and Alzheimer's disease are shown respectively in white, light gray and dark gray. Pearson's correlation coefficients and associated p-values are shown. (CBF: cerebral blood flow; GM: gray matter; WM: white matter). 112 Table 4.5 shows the same general linear model for global CBF but performed in each group individually. This subgroup analysis showed that only the association between Factor 1 and global CBF is present in individuals with MCI, with an effect size three times stronger than the association between global CBF and Factor 2. In comparison, only the association between Factor 2 and global CBF was significant in individuals with AD, though the association between Factor 1 and global CBF had an equal effect size. Table 4.5 Models of the global cerebral blood flow with factors extracted from the neuroimaging markers in each subgroup. Global CBF (P; p-value) Controls (n = 113) MC I (n = 159) AD (n = 92) Age -0.05; 0.6287 0.09; 0.3595 -0.01; 0.9575 Sex (female) *0.21; 0.0138 *0.20; 0.0119 **0.31; 0.0047 Education -0.06; 0.4447 -0.11; 0.1585 -0.01; 0.9395 Factor 1 0.00; 0.9889 **0.30; 0.0020 0.22; 0.1544 Factor 2 -0.06; 0.5955 0.09; 0.3384 *0.22; 0.0341 Translation motion 0.02; 0.9478 -0.33; 0.1230 *0.73; 0.0333 Rotation motion 0.07; 0.8501 *0.51; 0.0191 *-0.69; 0.0413 All continuous variables were standardized prio r to applying the model for easier comparison of parameter estimates (0). Uncorrected p-values are presented and significant associations with p < 0.05 are bolded (*, ** and *** for p < 0.05, 0.01 and 0.001 respectively). Statistical trends 113 were italicized (p < 0.10). (CBF: cerebral blood flow; MCI: mild cognitive impairment; AD: Alzheimer's disease). 4.5 Discussion We demonstrated two independent classes of degenerative changes in AD through factor analysis of hippocampal volume, AD signature cortical thickness, ventricular volume, total WM volume and volume of white matter lesions. One factor was more strongly associated with age, cerebral blood flow and diffusivity and total volume of white matter lesions, which are typically presumed to be of vascular origin; yet the factor was strongly affected in individuals with AD compared to controls. This factor was interpreted as the 'age- and vascular-related' factor. The second factor was more strongly related to MMSE and imaging markers of AD neurodegeneration, such as cortical thickness, and was associated with worse parahippocampal WM microstructure. This factor was interpreted as the 'neurodegenerative' factor. Critically, a significant amount of variance in two commonly examined markers of change in AD, ventricular volume and hippocampal volume, factored with the volume of white matter lesions and therefore demonstrated a potential degenerative link between vascular conditions and changes commonly attributed to classical AD neurodegenerative processes. As hippocampal volume was retained as an important marker in both factors, we further investigated its major determinants. Greater age, reduced AD signature cortical thickness, reduced total WM volume and greater ventricular volume and volume of white matter lesions were all associated with a lower hippocampal volume and each accounted for significant, independent additional variance explained, even when taking into account group determination in the model. These 114 associations, especially with age, white matter lesions and total WM volumes, might be key to understanding the presumed vascular component of AD pathogenesis and its influence on the hippocampus (Laakso et al., 1996; Du et al., 2002; den Heijer et al., 2005) and requires further investigation. The two independent classes of degenerative changes obtained through factor analysis were replicated in a second independent dataset of non-demented controls and individuals with MCI and AD (though with one overlapping participant). In addition to a strong correlation between the factor scores of both factor analyses, cross-sectional relationships of the factors with group, age, MMSE, as well as diffusion measures of white matter integrity and cerebral blood flow were replicated using factor scores from both factor analyses. This replication of the factor analysis using the same exact method on two independent samples demonstrates the generalizability of the study findings. A few more factor analysis experiments were performed within each group and for each sex in order to understand the influence of these variables on the factor structure. The factor structure preserved the same variables for the 'age- and vascular-associated' factor regardless of group in both datasets. The 'neurodegenerative' factor also related to regional cortical thickness in each group, but did not include hippocampal volume except for individuals with MCI in the replication dataset, and included ventricular volume in individuals with AD only in the first dataset, which may be explained by the greater reduction in cortical thickness in these individuals, which leads to smaller brain volume and larger fluid spaces. Interestingly, males had white matter lesions as part of both factors in the first dataset, while no differences from the factor structure reported in this study were found in women. However, no differences in factor structure were observed in males in the replication 115 sample and therefore future work is needed. All these experiments confirmed the strong association found between white matter lesions, hippocampal volume and ventricular volume, as part of the 'age- and vascular-associated' factor. While both factors were related to age, the 'age- and vascular-associated' factor was much more strongly related to age than the 'neurodegenerative' factor. This is consistent with the notion that age is one of the most important risk factors for vascular disease (D'Agostino et al., 2008). Both factors were also strongly related to MMSE score, indicating that they may both play a role in contributing towards the degree of cognitive impairment in patients. The 'neurodegenerative' factor was more strongly related to MMSE score, and this may be explained by the fact the gray matter is more directly involved in the processing of information than the white matter. The impact of both factors on longitudinal cognitive decline is studied in the next chapter. There was at a much stronger effect size of the 'age- and vascular-associated' factor than the 'neurodegenerative' factor on the diffusivity of white matter lesions (nearly by a factor of three). Similarly, the effect size of the 'age- and vascular-associated' factor was greater than the effect size of the 'neurodegenerative' factor on diffusion measures of both parahippo.campal WM and NAWM, though to a lesser degree than on diffusion measures of white matter lesions. Despite this, the 'neurodegenerative' factor was still significantly associated with diffusion measures in all three white matter structures, suggesting a partial involvement of the classical neurodegenerative effects of AD in the degradation of the white matter, possibly through Wallerian degeneration. One significant association was counter to this interpretation, linking 116 greater neurodegenerative effects to an increase fractional anisotropy in white matter lesions, which is generally linked to greater fiber order or number, or greater myelination of fibers. This association was also observed in the previous chapter between a decreasing AD cortical thickness and an increasing fractional anisotropy of lesions, and will need further study in order to fully understand its implications. Overall, the 'age- and vascular-associated' factor was responsible for most of the effects observed in the white matter, which hints at a dissociation between the presumed neurodegenerative effects of AD and the microstructural changes of WM and especially of the more prevalent white matter lesions observed in AD, suggesting those effects are mostly independent from each other. Oddly, greater involvement of the 'neurodegenerative' factor was associated with greater fractional anisotropy in lesions, which is generally interpreted as a greater order and density of myelinated fibers. This association is the inverse of what would be expected if neurodegeneration led to poorer white matter integrity. One possibility is that some of the lesions may be in regions with crossing fibers, and selective degeneration of one tract relative to the other may lead to an increase in fractional anisotropy associated with reduced cortical thickness. This will be the focus of future work. There was also a greater effect size of the 'age- and vascular-associated' factor than the effect size of the 'neurodegenerative' factor on global cerebral blood flow, and only the effect of the 'age- and vascular-associated' factor was significant. The 'age- and vascular-associated' factor was also the only factor related to regional CBF, both cortical gray matter CBF and white matter CBF. While WM CBF is especially prone to noise and motion, it is interesting to note nonetheless that the strongest association observed was between the 'age- and vascular- associated' factor and white matter CBF. This result is also consistent with the fact this factor is 117 strongly related to greater white matter atrophy and greater volume of white matter lesions, which have poorer perfusion and have been linked to lower regional and global cerebral blood flow as described at length in the thesis introduction (Brickman et al., 2009; Makedonov et al., 2013; Benedictus et al., 2014; Crane et al., 2015). Interestingly, while both factors had a similar effect size on global CBF in individuals with AD, the effect size of the 'age- and vascular- associated' factor was three times stronger than the effect size of the 'neurodegenerative' factor on the global CBF of individuals with MCI. While this study is cross-sectional and causality cannot be determined, these results may suggest that the 'age- and vascular-associated' factor may be linked to a loss of CBF leading to cognitive decline in the prodromal phase, before dementia is diagnosed, while the association between the 'neurodegenerative' factor and CBF in individuals with AD may simply represent the CBF lost as a consequence to cell death. The next chapter uses longitudinal data and assesses whether the loss of CBF associated with the 'age- and vascular-associated' factor may be leading to neurodegeneration and cognitive decline rather than be a consequence of neurodegeneration. Finally, it is interesting to note that no group differences were observed in diffusion measures of integrity or regional or global measures of cerebral blood flow when the factors were included in the models. Both diffusion and cerebral blood flow measures tracked better with the factors than with group differentiation. In particular for cerebral blood flow measures, this suggests that the factors are representative of the processes that generally lead to the group differences observed in those measures, as previously reported in the literature (Alsop et al., 2000; Dai et al., 2009; Schuff et al., 2009; Binnewijzend et al., 2013; Mattsson et al., 2014). However, it is also possible that few or no group differences are observed because the CBF data 118 was noisy and had low and negative CBF values, showing the limitations of the technique and the need for careful interpretation of these findings. Future work will aim to reduce sources of noise in this type of data. Females also had significantly higher CBF than males, and this will need to be better studied and understood in future work. 4.6 Appendix 4.6.1 Study limitations The current work has the same limitations as reported in the last chapter. There is a possibility that the factor analysis may be driven by the differences in the technical procedures used to obtain the neuroimaging estimates. Indeed, measures of volume - volume of white matter lesions, total WM volume, ventricular volume and hippocampal volume - are obtained from the same anatomical segmentation, while regional cortical thickness is obtained through a more complicated process that may suffer from different sources of noise, such as the greater reliance on a good image contrast between the gray matter and the white matter. Such a technical difference may have resulted in the two factors, one primarily involving volumetric measures, and one primarily involving the thickness measurement. However, the replication of the factor analysis attributed a greater contribution of hippocampal volume to the second factor, showing a less important divide between the volumetric and thickness measurements. Future work will investigate this possibility and ensure that the different sources of noise are not influencing the factor analysis. A few additional limitations exist in relationship with the use of arterial spin labeling MRI. First and foremost, this technique provides noisy data and future work will be needed to reduce sources of noise in this dataset. Furthermore, the technique 119 captures an image of the cerebral blood flow observed within three seconds of the tagging of the blood, and therefore does not capture blood flow in areas where the tagged blood does not have enough time to reach. However, such a slow flow would most certainly decrease the oxygenation and perfusion available, and would probably have similar effects as an actual decrease in perfusion in the areas where the measurement of the cerebral blood flow would be reduced as a result of this technical limitation. The parameters used in the quantitative model used to infer cerebral blood flow were also not specific to gray matter or white matter, while regional values of cerebral blood flow are reported for both cortical gray matter and white matter. However, this does not have any effect on the findings reported here because those parameters are used in scaling factors and therefore would not modify any associations observed across individuals. Furthermore, there was no comparison between cortical gray matter CBF and white matter CBF in this study, where proper scaling would matter to compare quantitative CBF values. Regardless of these limitations, the current work demonstrates that white matter lesions are linked to other traditional imaging markers of AD and provide novel information about the complex inter-associational properties of several known markers of AD potentially providing information about multiple 'classes' of partially independent degenerative change to be targeted for therapeutic intervention. 120 121 Chapter 5: Distinct Alzheimer's disease processes contribute independently to cognitive decline: preliminary assessment 5.1 Overview The previous chapter described the existence of two potentially distinct classes of degenerative change in AD, with potentially age- and vascular-mediated tissue damage contributing to one factor involving white matter lesions and classical neurodegenerative changes associated with AD contributing to a second factor. While both factors were related to the Mini-Mental State Exam (MMSE) score used to assess the degree of cognitive impairment, these associations were cross-sectional and it is yet unclear whether a longitudinal change in one or both factors leads to a change in MMSE. Such a link between the change in the age- and vascular-related factor and MMSE decline, independently of the change in the neurodegenerative factor, would ultimately demonstrate the potential to prevent and reduce cognitive decline by targeting this distinct pathway involving white matter lesions. In this chapter, we describe such a link and show further evidence suggesting that preventing a change in this age- and vascular-related factor would not only prevent a decline in MMSE, but would prevent a decline in MMSE equivalent to preventing a similar change in the neurodegenerative factor. 122 5.2 Introduction A long-standing debate in the field of Alzheimer's disease (AD) is centered on the question of whether or not vascular disease is inherently part of the progression towards dementia, or simply co-morbidity. In the last chapter, we found two distinct processes that seemed to drive the co-variation in five neuroimaging markers affected by AD. One of those processes seemed to represent those vascular co-morbidities as it was more strongly related to age and cerebral blood flow, and involved white matter lesions that have strong ties to small-vessel disease and vascular disease in general. While this process may be distinct from classical AD neurodegenerative processes, it remains unclear whether the two processes have an independent effect on cognitive decline, and may both lead to clinical progression towards dementia and AD. A few previous studies have tried to distinguish their impact on MMSE. For instance, white matter lesions and cortical gray matter loss have been shown to have independent associations with MMSE, which is used to assess the degree of cognitive impairment, and other cognitive measures in individuals with probable AD (Stout et al., 1996). Previous studies have also shown a link between changes in cerebral blood flow and cognitive decline; however, it is unclear whether this link is independent from vascular-associated pathology such as white matter lesions. We have previously shown in our cross-sectional study that both processes are significantly related to Mini-Mental State Examination score despite being uncorrelated. Therefore, while the process attributed to vascular co-morbidities seems to be unrelated to classical neurodegenerative changes, it remains clinically important as it may lead to further progression 123 towards dementia as assessed with clinical outcomes. In this study, we aim to determine whether a longitudinal decrease in both factors is independently related to cognitive decline and therefore that change in both the 'age- and vascular-associated' factor and the 'neurodegenerative' factor lead to further progression towards dementia. In addition, we aim to understand whether baseline and change in cerebral blood flow are also associated with cognitive decline, as well as with changes in the 'age- and vascular-associated' factor. Predictors of change in cerebral blood flow are also determined. 5.3 Materials and Methods 5.3.1 Participants and MRI acquisition The dataset used consists of a longitudinal dataset of 122 individuals with MCI from the Alzheimer's Disease Neuroimaging Initiative (ADNI, http://adni.loni.usc.edu) who underwent whole-brain MRI scanning at least twice approximately two years apart on a 3-Tesla Siemens scanner and had sagittal Tr-weighted 3D spoiled gradient echo images and arterial spin labeling (ASL) images (1 reference Mo image, 52 tagged-untagged pairs) available at the time of download for these two visits. Some individuals in this dataset were used in the cross-sectional study detailed in the previous chapter. Five individuals were excluded from the analysis due to outlier, improbable longitudinal segmentation data (e.g. large decrease in ventricular volume). This dataset was acquired using previously described ADNI Core MRI and DTI protocols (Jack et al., 2008). Group designation of MCI was determined by ADNI based on the criteria of the National Institute of Neurological and Communicative Diseases and Stroke - Alzheimer's Disease and Related Disorders Association (McKhann et al., 1984). Participants enrolled as early 124 and late MCI were combined into one MCI group (see ADNI 2 Procedures Manual on www.adni- info.org for more information). Clinical profiles and diagnostic information were obtained from the assessment closest in time to the MRI acquisition. Seventeen individuals converted from MCI to clinical AD over the two years of longitudinal follow-up. Demographics are provided in Table 5.1, showing that individuals with MCI who converted to AD had a higher frequency of the APOE allele E4, lower MMSE at baseline and a greater decrease in MMSE than individuals with MCI who did not convert. Written informed consent was obtained from all participants or their representatives through ADNI. The study procedures were approved by institutional review boards of all participating institutions. Table 5.1 Demographics for all participants with MCI that had longitudinal data. MCI participants who Other MCI p-value converted to AD participants Participants (female) 17(9) 100 (42) 0.4023 Age at baseline (years) 72.05 (1.97) 70.74 (0.69) 0.5402 Time between scans (years) 2.04 (0.02) 2.01 (0.01) 0.2896 Education (years) 16.94 (0.57) 16.44 (0.27) 0.4313 APOE E4 (# alleles) 1.06 (0.18) 0.45 (0.06) 0.0048 MMSE at baseline (-) 27.00 (0.50) 28.33 (0.16) 0.0200 MMSE difference (-) -3.59 (0.59) -0.26 (0.20) <0.0001 Translation motion (mm) 0.29 (0.03) 0.27 (0.01) 0.5782 Rotation motion (degrees) 0.0020 (0.0002) 0.0018 (0.0001) 0.4427 125 Standard errors are shown in parentheses. (MMSE: Mini-Mental State Exam; MCI: mild cognitive impairment) 5.3.2 Imaging data processing The reader is referred to sections 4.3.2 and 4.3.3 for complete details about the preprocessing of this dataset, including the automated subcortical and white matter lesion segmentation and processing of the arterial spin labeling data to obtain volumetric and cortical thickness data as well as global values of cerebral blood flow (CBF) at each time point. 5.3.3 Computation of factor scores Factor scores were obtained for each individual at each time point using the standardization matrix and standard score coefficients extracted from both factor analyses described in the previous chapter. 5.3.4 Statistical analyses Statistical analyses were performed using JMP 10 statistical software (SAS Institute Inc., Cary, NC, U.S.A.). The factors and global CBF at baseline and their change over time were used in a general linear model of the change over time in Mini-Mental State Examination (MMSE) score using age, time between scans, sex, education (Teipel et al., 2009), number of APOE E4 alleles and average motion measures across time points (average translation and average rotation (Yendiki et al., 2014)) as covariates. The change in global and regional CBF was also modeled using the factors at baseline and their change over time, the global or regional CBF at baseline 126 as well as the same covariates. Estimated parameters were provided in the models in addition to p-values to ease interpretation. 5.4 Results 5.4.1 Associations between MMSE decline and change in factors scores and CBF Average factor score and global cerebral blood flow at baseline and longitudinal change in factor score and global cerebral blood flow are provided for individuals with MCI who converted to AD and those who did not in Table 5.2. All factor scores at baseline and all changes in factor scores were significantly different between the participants with MCI who converted and those who did not. There was evidence of a lower global CBF at baseline and a greater decrease in CBF in individuals who converted compared to those who did not, but they were not significant, probably due to the fact it is a noisy measurement. Table 5.2 Average baseline and difference in factor scores and CBF over time. MCI participants who Other MCI p-value converted to AD (n = 17) participants (n = 100) Factor 1 (baseline) -0.594 (0.230) 0.005 (0.106) 0.0270 Factor 1 (difference) -0.381 (0.053) -0.145 (0.014) 0.0004 Factor 2 (baseline) 0.286 (0.248) 0.863 (0.069) 0.0374 Factor 2 (difference) -0.434 (0.115) -0.126 (0.028) 0.0176 Global CBF (baseline) 21.65 (2.16) 24.03 (0.76) 0.3131 Global CBF (difference) -4.39 (1.84) -1.81 (0.68) 0.2043 127 Standard errors are shown in parentheses. (CBF: cerebral blood flow; MCI: mild cognitive impairment) General linear models of the difference in MMSE between time points two years apart are shown in Table 5.3. Factor scores from the primary factor analysis and its replication in an independent sample were used in two different models to confirm that the results did not differ. In all instances, both the difference in Factor 1 and the difference in Factor 2 were significantly related to the difference in MMSE independently of each other. There was a trend that the difference in global CBF may also be independently related to the difference in MMSE, (p < 0.10). Additionally to the changes in factors and CBF, both Factor 2 and MMSE at baseline were also significant independently in predicting the change in MMSE. Table 5.3 Model of the longitudinal decline in MMSE using both sets of factors in participants with MCI. Model with Factors 1 and 2 Model with Factors 1' and 2' Parameters Diff. MMSE (P; p-value) Diff. MMSE (@; p-value) Age (baseline) 0.10; 0.3428 0.09; 0.4361 Time between scans 0.01; 0.8846 0.01; 0.9387 Sex (female) -0.08; 0.3447 -0.10; 0.2527 Education 0.13; 0.1395 0.13; 0.1220 APOE E4 (# alleles) -0.13; 0.3045 -0.14; 0.2998 Factor 1/1' (baseline) 0.05; 0.7246 -0.02; 0.8807 Factor 1/1' (difference) ***0.49; 0.0001 ***0.37; 0.0005 128 Factor 2/2' (baseline) *0.21; 0.0287 *0.21; 0.0310 Factor 2/2' (difference) *0.24; 0.0106 ***0.33; 0.0005 Global CBF (baseline) -0.01; 0.9270 0.01; 0.9284 Global CBF (difference) 0.16; 0.0819 0.18; 0.0561 MMSE (baseline) **-0.27; 0.0018 **-0.29; 0.0011 Translation motion -0.26; 0.2451 -0.26; 0.2348 Rotation motion 0.19; 0.4109 0.20; 0.3755 All continuous variables were standardized prior to applying the model for easier comparison of parameter estimates (B). Uncorrected p-values are presented and significant associations with p < 0.05 are bolded (*, ** and *** for p < 0.05, 0.01 and 0.001 respectively). Associations with p < 0.10 are italicized. (MMSE: Mini-Mental State Exam; CBF: cerebral blood flow). Scatterplots of these significant relationships are shown in Figure 5.1 without any covariates. While MMSE, factor scores and global CBF generally showed a decrease as expected, a few participants showed a positive change over time in these measures. However, those were generally close to zero suggesting that they may be the result of noise in the estimation of the imaging and cognitive measures. The relationship between the difference in Factor 1 and difference in Factor 2 is also shown, indicating no correlation between them. While the absence of correlation between factors in the cross-sectional datasets is a direct result of the factor analysis, it was not necessarily expected that the change over time would be independent. 129 6 0 6 0 r< ).441 r = 0.204 4 0.001 0 4 < 0.05 0 000 p Tco 0 .. 2 0 0CDCD0 LuW 2 0 2 (0 ( 20 0 0 0 M&cZX 0 0 0 OCE -2 * 0 00 ~-o000 -4 '5 * . GCDC*c30 -6 0 * *D0 0 -6 - 0 -8 I |RI a -1 -0.8 -0.6 -0.4 -0.2 0 0.2 -1.5 -1 -0.5 0 0.5 Diff. Factor 1 (-) Diff. Factor 2 (-) 6 0 0.5 r r = 0.234 r= -0. 029 4 p < 0.05 0 p= 0. 76 00O 2 NJ- 0 LU 0 0 U0 0 0 0 Q U -0.5 '5 -2 o goo -L 4L 0 -4 U1 0 00 0 0 -6 00 *0 0 -81 ------1.5 . -30 -20 -10 0 10 20 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 Diff. Global CBF (mill OOg/min) Diff. Factor 1 (-) 0 Converted to Alzheimer's disease El Has not converted Figure 5.1 Longitudinal difference in MMSE over two years related to the difference in factor scores and global cerebral blood flow. The correlation between the difference in Factor 1 and the difference in Factor 2 is also shown. Pearson's correlation coefficients and associated p-values are shown. 5.4.2 Determinants of the change in CBF General linear models of the change in global, cortical gray matter and white matter cerebral blood flow after two years are shown in Table 5.4. Only CBF value at baseline predicted change in CBF for all structures, though both age and Factor 1 at baseline were also predictive for the changes in global and white matter CBF and as a trend for the change in cortical gray matter CBF. Younger age, lower factor score and greater CBF at baseline were related to a greater 130 decrease in CBF. However, Female sex was also associated with a lesser decrease in white matter CBF than males (trend for global CBF change). Factor 2 at baseline displayed a trend linking a lower factor score at baseline to greater decreases in global and cortical gray matter CBF. Neither the change in Factor 1 nor the change in Factor 2 was related to any CBF change. Table 5.4 Model of the longitudinal change in global and regional CBF in participants with MCI. Parameters Diff. Global CBF Diff. Cortical GM Diff. WM CBF (1; p-value) CBF 1; p-value) (P; p-value) Age (baseline) *0.27; 0.0237 0.18; 0.1356 *0.24; 0.0343 Time between scans -0.09; 0.3312 -0.03; 0.7043 -0.13; 0.1371 Sex (female) 0.18; 0.0658 0.12; 0.1940 *0.18; 0.0461 Education 0.03; 0.7770 0.00; 0.9706 0.05; 0.5602 APOE E4 (# alleles) -0.08; 0.5629 -0.08; 0.5383 -0.05; 0.6611 Factor 1 (baseline) *0.29; 0.0279 0.22; 0.0957 **0.36; 0.0060 Factor 1 (difference) 0.06; 0.5703 0.06; 0.5685 0.08; 0.4436 Factor 2 (baseline) 0.17; 0.0843 0.18; 0.0624 0.08; 0.3991 Factor 2 (difference) -0.02; 0.8280 -0.02; 0.8137 -0.08; 0.4229 CBF (baseline) ***-0.43; <0.0001 * -0.45; <0.0001 ***-0.46; <0.0001 Translation motion 0.07; 0.7792 -0.04; 0.8794 -0.11; 0.6238 Rotation motion -0.04; 0.8836 0.04; 0.8569 0.21; 0.3714 All continuous variables were standardized prior to applying the mo del for easier comparison of parameter estimates (1). Uncorrected p-values are presented and significant associations with 131 p < 0.05 are bolded (*, ** and *** for p < 0.05, 0.01 and 0.001 respectively). Associations with p < 0.10 are italicized. (MMSE: Mini-Mental State Exam; CBF: cerebral blood flow). 5.4.3 Determinants of the change in factor scores General linear models of the change in factor scores after two years are shown in Table 5.5. Only two variables predicted the change in Factor 1, and no variables used in the model predicted the change in Factor 2. Higher number of APOE E4 alleles and lower factor score at baseline were associated with a greater longitudinal reduction in Factor 1. These relationships are evidenced in Figure 5.2 without any covariates. There was also a trend for Factor 2 at baseline to predict the change in Factor 1. Table 5.5 Model of the longitudinal change in factor scores in participants with MCI. Parameters Diff. Factor 1 Diff. Factor 2 (W; p-value) (P ; -Value)1 Age (baseline) 0.09; 0.3727 0.16; 0. 1707 Time between scans -0.08; 0.3136 -0.00; 0. 9855 Sex (female) -0.10; 0.2259 -0.10; 0.3111 Education -0.11; 0.1502 -0.14; 0. 1126 APOE E4 (# alleles) **-0.38; 0.0014 -0.17; 0. 2008 Factor 1 (baseline) ***0.43; 0.0002 0.17; 0. 1623 Factor 2 (baseline) 0.16; 0.0555 0.08; 0. 3805 Global CBF (baseline) 0.00; 0.9651 -0.00; 0.9622 132 Global CBF (difference) 0.05; 0.5438 -0.03; 0.7631 Translation motion 0.05; 0.7979 -0.05; 0.8241 Rotation motion 0.04; 0.8377 -0.03; 0.9153 All continuous variables were standardized prior to applying the model for easier comparison of parameter estimates (3). Uncorrected p-values are presented and significant associations with p < 0.05 are bolded (*, ** and *** for p < 0.05, 0.01 and 0.001 respectively). Associations with p < 0.10 are italicized. (MMSE: Mini-Mental State Exam; CBF: cerebral blood flow). 0.2 0.2 .T. 0 - _1 O0 T"-0.2 o b_ -0.2 o -0.6 -0.6 0 r = 0.308 0 -0.8 P < 0.001 -0.8 -4 -2 0 2 4 No alleles One allele Two alleles Factor I at baseline (-) Converted to Alzheimer's disease Has not converted Figure 5.2 Longitudinal difference in Factor 1 over two years related to the difference in factor score at baseline and number of APOE E4 alleles. Pearson's correlation coefficient and associated p-value is shown. 5.5 Discussion Individuals with mild cognitive impairment (MCI) who converted to Alzheimer's disease (AD) during the two-year longitudinal follow-up demonstrated lower factor score at baseline and a greater decrease in factor score over time for both the 'age- and vascular-associated' and 'neurodegenerative' factors, as well as a non-significant lower global cerebral blood flow (CBF) at baseline-and greater decrease in CBF over time, when compared to individuals with MCI who 133 did not convert. Changes in the 'age- and vascular-associated' factor and in the 'neurodegenerative' factor were independently related to cognitive decline as measured with the Mini-Mental State Examination (MMSE) after the course of two years in all participants with MCI. The similar or greater effect size suggests that preventing decline in the 'age- and vascular- associated' factor may result in a slowing of the cognitive decline that is at least equivalent to treating the neurodegenerative aspect of the disease, which is the current focus of the biopharmaceutical industry. The change in global cerebral blood flow (CBF) also showed an independent trend of association with cognitive decline, indicating a potential third mechanism by which cognitive may be prevented. However, this change in global CBF was also predicted by the 'age- and vascular-associated' factor at baseline, suggesting potential mediation that will need to be further studied in future work. While changes in both factors involve structural changes that may require a longer therapeutic strategy, several treatment options as well as diet and exercise have been shown to increase cerebral blood flow which may show promise for shorter-term therapeutic intervention towards the prevention of cognitive decline. Future studies will address this possibility and investigate further the determinants of change in both factors to help devise a therapeutic strategy. While a change in the 'age- and vascular-associated' factor was related to a change in MMSE, this factor at baseline was not related to greater cognitive decline, indicating cognitive ability may have been lost before but it does not further lead to greater cognitive decline. This behavior can be likened to the effect of a stroke or of ischemic damage. While tissue death may have occurred in a brain region and may have been associated with a permanent loss of function, it may not necessarily compound further tissue death and cognitive decline. In 134 contrast, the lower the 'neurodegenerative' factor at baseline, the greater cognitive decline was observed. It may be speculated that early on, neurodegeneration may put a greater strain on surviving neural tissue directly surrounding the dying neurons to perform the functions that would otherwise be lost. As the disease progresses and as this coping mechanism struggles, it is possible that the cognitive decline becomes exponential past a certain threshold, which could explain this relationship. Previous studies have linked greater white matter lesion burden at baseline with a greater longitudinal decrease in brain perfusion in individuals with AD (Hanaoka et al., 2015) as well as in individuals with manifest arterial disease (van der Veen et al., 2015). Our results confirm this through the association between our 'age- and vascular-associated' factor at baseline, strongly linked to white matter lesion burden and their properties, and change in global and regional CBF. Interestingly, the opposite relationship was not found to be true in those studies, i.e. lower brain perfusion at baseline did not lead to a longitudinal increase in white matter lesions in these populations, though such a link has been shown before (Promjunyakul et al., 2015). The current study also failed to demonstrate an association between poorer perfusion at baseline or change in CBF and a decrease in the 'age- and vascular-associated' factor. This was unexpected as this factor was related to CBF in our previous cross-sectional study. These results suggest that the 'age- and vascular-associated' factor may be the result of an accumulation of vascular insults that concurrently lead to lower perfusion and to telltale signs such as white matter lesions. These insults seemed to be predicted by the presence of APOE E4 alleles as well as previous vascular insults, but other sources of reduction in CBF may not in themselves contribute further damage. For instance, a decrease in global and regional CBF may be the 135 result of cell death and tissue neurodegeneration and not of a reduction in perfusion of viable tissue. However, in this study, the change in the 'neurodegenerative' factor and its baseline were not significantly associated with change in CBF. No parameter explored in this study predicted the change in this 'neurodegenerative' factor, but future work will use positron emission tomography (PET) measures of amyloid and tau burden, which we hypothesize are likely to predict such a change. While it is interesting to note that the effect size of the 'age- and vascular-associated' factor at baseline on the change in white matter CBF was the greatest of all structures investigated, in line with the fact this factor is representative of white matter pathology, it is important to note that measuring CBF in the white matter is challenging. In comparison, the effect size of both factors at baseline on cortical gray matter CBF change was very similar, although only a trend and not significant. Finally, there were a few more determinants of the change in CBF. In addition to females having generally greater CBF than males as demonstrated in the previous chapter, we also found that females were more resistant to a decrease in CBF. Greater age was also associated with less of a reduction in CBF, which may result from having a lower baseline CBF. Indeed, lower CBF at baseline was also associated with less of a reduction in CBF, suggesting that there may be a limit to which CBF can be reduced. 5.6 Appendix 5.6.1 Study limitations The current work has the same limitations as reported in the previous chapter. It is possible that there was a learning effect related to the administration of the MMSE which is done once 136 every six months as part of the ADNI protocol. This may have resulted in the positive changes observed for MMSE score over time, though there is no reason to believe this learning effect differed between individuals, except possibly for the education level which was taken into account in the model. Similarly, there were positive changes recorded in some individuals both in factor scores and in CBF. While those positive changes are not theoretically impossible, a positive change in any factor score is unlikely and probably due to technical noise. On the other hand, positive changes in cerebral blood flow have been reported in prior studies, though we have not accounted for any potential determinant of such an increase in CBF in our statistical models, such as a change in medication, diet or physical exercise. Those determinants will be investigated in future studies. 137 138 Chapter 6: Discussion and perspectives 6.1 Summary and future work The first major contribution of this thesis is the finding that white matter lesions of individuals with Alzheimer's disease (AD) had slightly greater diffusivity than lesions of non-demented controls when controlling for the various locations of lesions across individuals, and that the diffusivity of lesions was strongly related to ventricular enlargement across all individuals. While diffusivity seemed to increase progressively with mild cognitive impairment and AD, it no longer differed significantly after correction for multiple comparisons. This relates to the histopathological findings showing that these lesions display similar types of pathology in individuals with AD and non-demented controls, though may be more severe in individuals with AD, as described in the introduction (Scheltens et al., 1995; Gouw et al., 2011). Furthermore, the spatial distribution of lesions was on average the same across groups, except for an overall increase in prevalence and volume in individuals with AD compared to non-demented controls. More importantly, the strong link discovered between diffusion properties of lesions and ventricular volume may speak to this non-significant increased diffusivity of lesions in individuals with AD. Indeed, greater diffusivity within the lesions, and therefore poorer microstructural integrity, was related to increased ventricular volume, which is known to be larger in individuals with AD. A potential biophysical explanation that has been offered in this thesis is that the denudation of the ventricular ependyma may allow leakage of cerebrospinal fluid into the white matter, leading concurrently to greater diffusivity within the lesions and 139 further expansion of the ventricles. Indeed, such denudation is known to be more severe in periventricular white matter lesions of individuals with AD (Scheltens et al., 1995; Gouw et al., 2011), and may be responsible for the water dysregulation and astrogliosis observed in the lesions (Brinker et al., 2014). A similar hypothesis can be made for non-periventricular lesions, which have been shown to have widened perivascular spaces, a part of a newly discovered water regulation 'glympathic' system in the brain (Xie et al., 2013; Brinker et al., 2014). More research needs to be done in order to understand the link between the histopathological features of ischemic damage and the features hypothesized to be the consequences of this damage, such as the denudation of the ventricular ependyma, widened perivascular spaces and gliosis. Disruption of the blood-brain barrier and other vascular insults have been proposed to be responsible for this link, but further research on the topic is required (Wardlaw et a/., 2003; Young et al., 2008; Gouw et al., 2011; Schmidt et al., 2011). An important contribution of this thesis is therefore an in vivo observation of the relationship between the diffusivity of white matter lesions and ventricular volume, an observation that can only be made in vivo using neuroimaging, is supported by histopathology, furthers our understanding of the pathogenesis of white matter lesions and provides potential avenues of investigation in animal models. The absence of a link between diffusion properties of lesions and either hippocampal volume or regional cortical thickness independent of the association with ventricular volume also furthers the case previously made using histopathology that lesions are not secondary to neurodegeneration (Pantoni and Garcia, 1997). A novel technical contribution originating from this work is the consideration of normative anatomy when evaluating the diffusion properties of white matter lesions. Indeed, the location of white matter lesions is varied across 140 participants, and may encompass different white matter tracts. While it is known that diffusion properties greatly vary across the brain, and in particular between single straight white matter fibers and crossing fibers, few to no studies have controlled for this. In this thesis, we presented a technique involving the nonlinear registration and averaging of individuals without lesions to create a normative atlas of each diffusion measure throughout the brain. This atlas was then registered back to each individual with lesions, and average normative diffusion properties were calculated within their segmented lesions and subtracted from the diffusion properties of the lesions. This allowed us to obtain diffusion measures approximating the change the tissue has undergone from a non-lesioned state to a lesioned state, and therefore control for the normative anatomy underlying each lesion. The second major contribution of this thesis is the underlining of the great covariation between neuroimaging markers of AD and their separation into two statistically distinct processes using factor analysis. While we previously explored the relationship between the diffusion properties of lesions and neuroimaging markers, we did not consider that those neuroimaging markers were highly correlated, and therefore could not truly ascertain whether the diffusion properties were simply more closely associated with ventricular volume than hippocampal volume or regional cortical thickness. Indeed, we could only confirm whether each individual neuroimaging marker explained unique variance in the diffusion properties of lesions when controlling for the other neuroimaging markers. In order to understand whether a single process or multiple processes were at the origin of the covariation between the neuroimaging markers, factor analysis was performed on the volume of white matter lesions, total white matter volume, ventricular volume, hippocampal volume and regional cortical thickness. This 141 resulted in two statistically distinct factors presumed to be due to differing pathological processes: the first process involved primarily the volume of white matter lesions, total white matter volume as well as the ventricular and hippocampal volumes, while the second process involved both hippocampal volume and regional cortical thickness, imaging markers typically associated with neurodegeneration. Such a clustering of the volume of white matter lesions and total white matter volume with ventricular volume was not surprising given the results previously found, tying the diffusion properties of lesions to ventricular volume, and given the well-understood correlation between white matter atrophy and ventricular enlargement. The clustering of hippocampal volume with this factor may underline the fact that it is reduced in other primarily vascular disorders, such as vascular dementia (Laakso et ai., 1996; Du et ai., 2002), and also the fact that white matter lesion severity has previously been correlated with hippocampal volume (den Heijer et al., 2005). It is interesting to observe that the factor analysis may have partitioned the contribution of two distinct pathologies - presumed vascular and neurodegenerative - on the degeneration of the hippocampus. This factor analysis was replicated with great precision in a near-independent sample (one subject overlap), confirming the generalizability of the analysis. The first factor was interpreted as the 'age- and vascular- associated' factor, due to its stronger relationship with age, an important risk factor for vascular disease (D'Agostino et al., 2008), as well as with cerebral perfusion and both diffusion properties and volume of white matter lesions which are of presumed vascular origin. The second factor was interpreted as the 'neurodegenerative' factor due to its inclusion of both hippocampal volume and regional cortical thickness and due to its stronger cross-sectional relationship with the degree of cognitive impairment. Diffusion properties of lesions and white 142 matter in general were much more strongly related to the 'age- and vascular-associated' factor, though the influence of the 'neurodegenerative' factor was still significant, especially in the parahippocampal white matter, demonstrating that the prior analysis with simply neuroimaging markers was not sufficient. However, an important contribution of this thesis is the demonstration that the great majority of the white matter damage observed in AD is linked to a distinct, independent process from the process represented by imaging markers typically associated with neurodegeneration, and is most likely related to vascular dysfunction. The final major contribution of this thesis is the demonstration that a decrease over time in the 'age- and vascular-associated' factor results in cognitive decline that is at least equally important and independent to the cognitive decline associated with a decrease in the 'neurodegenerative' factor, suggesting an additional potential avenue of preventative treatment for individuals in the pre-clinical stages of AD. While the previous findings allowed us to identify two statistically distinct processes in AD, the clinical significance of each factor remained unclear. This final set of preliminary analyses, in a longitudinal sample of individuals with MCI, allowed us to determine that both factors were independently related to cognitive decline. Furthermore, a decrease in cerebral blood flow was also associated with cognitive decline, though it was only a trend when the change in both factors was included in the model. The 'age- and vascular-associated' factor score at baseline predicted a decrease in cerebral blood flow, further linking this factor to vascular disease, and future work will assess whether the effect of the 'age- and vascular-associated' factor on cognition is mediated by a change in cerebral blood flow. Future work will also investigate determinants of both factors. For instance, it is expected that a significant vascular insult such as a stroke would result in a 143 significant decrease of the 'age- and vascular-associated' factor score both in individuals with AD and in non-demented controls. As for the 'neurodegenerative' factor, no obvious determinant was found which is consistent with the fact the origin of these changes may not have been captured by the neuroimaging modalities used in this thesis. Indeed, future work involving positron emission tomography (PET) measures of amyloid and tau deposition will help confirm our hypothesis that the 'neurodegenerative' factor is driven by those pathologies, as observed by its effect on the cerebral cortex. However, it is also possible that both factors may be influenced by those pathologies. One of the strengths of the factor analysis method was the disentanglement of the presumed-vascular and neurodegenerative contributions to the reduction in hippocampal volume. The effect of vascular pathology on the hippocampus has been hinted at previously in the literature, but the degeneration of the hippocampus in AD has generally been considered to be primarily a result of amyloid and tau pathology. Similarly, amyloid deposition in the blood vessels, called cerebral amyloid angiopathy, is one of the major vascular pathologies observed in AD and therefore amyloid pathology may affect both factors to varying extents. Further research is also needed to more closely examine the histopathological correlates of both factors. 6.2 The evidence for a distinct white matter pathway in Alzheimer's disease Overall, the current results suggest new opportunities for prevention of cognitive decline in individuals with MCI, by further understanding and treating the underlying process identified by the 'age- and vascular-associated' factor. In our preliminary analyses, such prevention suggests a reduction in cognitive decline on a scale similar or greater to treating the 'neurodegenerative' 144 component of the disease, which is the current focus of therapeutic trials. Throughout the thesis, we have demonstrated that this factor is greatly responsible for the white matter damage observed in AD and pre-clinical AD, and that this damage may be vascular in nature. Previous studies have made the case of such a distinct pathway, with some suggesting that it may even be required for the full development of AD (Zlokovic, 2011). However, such a 'two- hit' hypothesis is unlikely, given the existence of AD cases where no appreciable white matter or vascular damage was noticeable. While it is possible that the presence of such damage may not be measurable, it is much more likely that the presence of vascular disease may not be necessary but highly increases the chances of developing full-blown AD dementia. In this regard, a large recent neuropathological study looking at the presence or absence of several different types of pathologies in the post-mortem brain of individuals of 90 years of age or older found that only half of individuals with moderate to high AD pathology alone actually develop dementia (Kawas et al., 2015). However, nearly all individuals (91%) who had both moderate to high AD pathology and the presence of a vascular pathology were demented prior to death. Such a vascular pathology included the presence of macroinfarcts, microinfarcts, or subcortical arteriolosclerotic leukoencephalopathy in the periventricular area, which shares similar pathological features as those observed in white matter lesions observed on neuroimaging. Indeed, it can be inferred from previous literature that these pathologies are closely related to white matter lesions observed on neuroimaging (Chen et al., 2009; Jellinger, 2013). While a large fraction of the demented cases still only had moderate to high AD pathology, indicating that a second 'hit' or pathology may not be necessary for dementia onset, it is important to mention that the scope of evaluation of those vascular pathologies was 145 limited to specific areas of the brain, given the labor intensive work of assessing the entirety of the brain. Regardless, the fact that both pathologies observed together seemed to nearly guarantee the clinical pre-mortem diagnosis of dementia indicates a significant role of vascular pathology towards the clinical expression of AD, even if not present in all demented cases. The work presented in this thesis, in particular related to the 'age- and vascular-associated' factor, may help track this vascular component in vivo, though studies investigating the histopathological correlates of this factor are needed. While the evidence is great to support the existence of an additional pathway involving white matter and vascular damage, it remains possible that the two factors discovered in this thesis may not necessarily represent two simultaneously-occurring processes, but the occurrence of one single process at two different points in time. For instance, a possible scenario would be that cortical degeneration and reduction in cortical thickness may happen early; this may then I%_ LU e Ld L LI LL 1aterLUE5stg %ILUe dIsease LO ani ass1IC a dLU inIr1=Se In w 11 Lt IIId LatI InUIIs, WhIie matter atrophy and ventricular enlargement which may happen closer to clinical conversion. This could generate the imperfect correlations observed between the neuroimaging markers, and enable the finding of two different factors. However, the current longitudinal data does not provide any evidence of this phenomenon, as the change in both factors were uncorrelated. If the effect of one factor preceded the effect of the other, an inverse correlation may be expected, when the change in one factor does not happen at the same time as the change in the other factor. Furthermore, and perhaps more convincingly, there were plenty of individuals with MCI that had a low 'age- and vascular-associated' factor score and a high 'neurodegenerative' factor score at baseline, and vice-versa, suggesting that no specific 146 neuroimaging marker or factor is necessarily affected first. Analysis of longitudinal data over a longer period of time will be necessary to fully rule out this possibility. 6.3 Estimated potential impact of intervention on delaying disease onset One of the final results of the thesis demonstrate the equal if not greater impact of the 'age- and vascular-associated' factor on cognitive decline, when compared to the effect of the 'neurodegenerative' factor. However, it is not clear from the information provided thus far how long the disease onset may be delayed if the decrease in the 'age- and vascular-associated' factor would be prevented. While this will require further study, we can make some approximation using the current data and a few assumptions. One such assumption is that decline in this factor may be fully prevented. It is indeed possible that there is little room for prevention and that this factor represents inevitable 'wear and tear' damage given the significant relationship between this factor and age. Another assumption is that the general linear model estimates may be used individually for each parameter of interest, and that they are accurate. It also assumes that all other parameters in the model are unchanged, which is unlikely. Using non-standardized variables in the general linear model, we obtain the following estimates: 5.75 MMSE unit change per unit of factor score change for the 'age- and vascular- associated' factor, and 1.49 MMSE unit change per unit of factor score change for the 'neurodegenerative' factor. Using the average change in both factor scores for converters, available in Figure 5.2, this translates to an average decrease of 2.19 MMSE units attributable to the change in the 'age- and vascular-associated' factor and to an average decrease of 0.65 MMSE units attributable to the change in the 'neurodegenerative' factor, in both cases over 147 two years. If the contribution of the 'age- and vascular-associated' factor was entirely prevented, and therefore that the change in factor score was reduced to zero, it would prevent an average decrease of 2.19 MMSE units over two years. Assuming conversion happens after a loss of 3.59 MMSE units, which is the average decrease in MMSE score over two years observed in converters (see Figure 5.1), it would take approximately a little over 3 additional years for conversion to happen, more than doubling the remaining time to conversion from baseline. Alternatively, if the contribution of the 'neurodegenerative' factor was prevented, it would delay clinical onset by only a little less than an additional half-year. Obviously, those estimates are extremely simplistic and likely inaccurate, but demonstrate the potential benefit on delaying disease onset by seeking therapeutic options for this 'age- and vascular-associated' component of the disease, rather than only focusing on the 'neurodegenerative' aspect of the disease. Such a delay in disease onset has been shown to be extremely important socio- economically, as described in the introduction, even if it does not ultimately prevent or cure the disease. 6.4 The future of risk prevention for Alzheimer's disease Regardless of which pathology is targeted, it has now become clear that the future of treatment for AD lies in preventing onset or progression of the disease in the pre-clinical stages. Indeed, past therapeutic trials in individuals with AD have failed, as described in the introduction, and current and future trials increasingly target populations that are at risk for AD rather than individuals with a clinical diagnosis of AD. One such example is the A4 study, which investigates whether anti-amyloid treatment can slow memory loss in individuals that are at risk based on 148 their amyloid load as assessed with PET, but display no memory impairment when recruited (http://a4study.org). While clinical outcomes remain the standard of current clinical trials, trials in individuals at risk of progression towards AD may not experience measurable cognitive decline, and will necessitate greater use of neuroimaging biomarkers such as those described in this thesis. The factors uncovered in this thesis may represent effective trackers of the longitudinal progression of the disease in individuals in the pre-clinical stages and may help reduce the heterogeneity of changes that occur in those stages by ascribing them to two distinct processes. Depending on the therapeutic strategy, it is likely that only one of these two processes will be targeted and of interest, and therefore the other process can be controlled accordingly. An advantage of using the factors is that they can be obtained from a simple T1- weighted image. However, they have been obtained from research-quality scans as part of the work of this thesis and the segmentation and cortical reconstruction algorithms required to obtain the five neuroimaging markers underlying the factors may not be readily available from clinical-quality scans. Future work will be needed to investigate the possibility of using them as biomarkers in clinical trials. 6.5 Conclusion In summary, we have found evidence that Alzheimer's disease and its pre-clinical stages may harbor two distinct disease processes measurable with neuroimaging markers which are both independently responsible for cognitive decline towards the clinical onset of the disease. As part of this work, we discovered that one process was related the volume and diffusion properties of white matter lesions, a frequent type of white matter damage in individuals with 149 AD, as well as to ventricular enlargement and white matter atrophy. This process was linked to increasing age, declining cerebral perfusion and was mainly responsible for the microstructural deterioration of the white matter, which was much less related to imaging markers typically associated with neurodegeneration, forming the basis of the second process representing classical AD pathology. Our results provided hints of the involvement of periventricular damage and water dysregulation in this 'age- and vascular-associated' process, potentially secondary to disruption in blood-brain permeability and vascular insults which will need to be further studied in animal studies. Overall, this 'age- and vascular-associated' process resulted in similar cognitive decline as the 'neurodegenerative' factor after a longitudinal follow-up of two years, and therefore represents an alternative therapeutic opportunity to slow the cognitive decline associated with AD, and potentially delay the clinical onset of dementia. 150 151 Chapter 7: References Abe 0, Aoki S, Hayashi N, Yamada H, Kunimatsu A, Mori H, et al. Normal aging in the central nervous system: quantitative MR diffusion-tensor analysis. Neurobiol. Aging 2002; 23: 433-41. Acosta-Cabronero J, Williams GB, Pengas G, Nestor PJ. Absolute diffusivities define the landscape of white matter degeneration in Alzheimer's disease. Brain 2010; 133: 529-39. Albert MS, DeKosky ST, Dickson D, Dubois B, Feldman HH, Fox NC, et al. The diagnosis of mild cognitive impairment due to Alzheimer's disease: Recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimer's Dement. 2011; 7: 270-279. Alsop DC, Detre JA, Grossman M. Assessment of cerebral blood flow in Alzheimer's disease by spin-labeled magnetic resonance imaging. Ann. Neurol. 2000; 47: 93-100. Alzheimer's Association. Changing the trajectory of Alzheimer's disease: how a treatment by 2025 saves lives and dollars. 2015. a. Alzheimer's Association. Alzheimer's disease facts and figures. 2015. b. Anderson VC, Lenar DP, Quinn JF, Rooney WD. The Blood-Brain Barrier and Microvascular Water Exchange in Alzheimer's Disease. Cardiovasc. Psychiatry Neurol. 2011 Archer H a., Edison P, Brooks DJ, Barnes J, Frost C, Yeatman T, et al. Amyloid load and cerebral atrophy in Alzheimer's disease: An 11C-PIB positron emission tomography study. Ann. Neurol. 2006; 60: 145-147. Atiya M, Hyman BT, Albert MS, Killiany R. Structural magnetic resonance imaging in established and prodromal Alzheimer disease: a review. Alzheimer Dis. Assoc. Disord. 2003; 17: 177-95. Au R, Massaro JM, Wolf PA. Association of white matter hyperintensity volume with decreased cognitive functioning: the Framingham Heart Study. Arch. Neurol. 2006; 63: 246-250. Bakkour A, Morris JC, Dickerson BC. The cortical signature of prodromal AD: regional thinning predicts mild AD dementia. Neurology 2009; 72: 1048-55. Barker R, Ashby EL, Wellington D, Barrow VM, Palmer JC, Kehoe PG, et al. Pathophysiology of white matter perfusion in Alzheimer's disease and vascular dementia. Brain 2014; 137: 1524- 32. Basser PJ, Jones DK. Diffusion-tensor MRI: theory, experimental design and data analysis - a technical review. NMR Biomed. 2002; 15: 456-467. Basser PJ, Mattiello J, LeBihan D. MR diffusion tensor spectroscopy and imaging. Biophys. J. 1994; 66: 259-67. 152 Basser PJ, Pierpaoli C. Microstructural and physiological features of tissues elucidated by quantitative-diffusion-tensor MRI. J. Magn. Reson. 1996; 111: 209-219. Bastos-Leite AJJ, Kuijer JPAPA, Rombouts SARBARB, Sanz-Arigita E, van Straaten ECC, Gouw AAA, et al. Cerebral Blood Flow by Using Pulsed Arterial Spin-Labeling in Elderly Subjects with White Matter Hyperintensities. Am. J. Neuroradiol. 2008; 29: 1296-1301. Bateman RJ, Xiong C, Benzinger TLS, Fagan AM, Goate A, Fox NC, et al. Clinical and biomarker changes in dominantly inherited Alzheimer's disease. N. Engl. J. Med. 2012; 367: 795-804. Beaulieu C. The basis of anisotropic water diffusion in the nervous system - a technical review. NMR Biomed. 2002; 15: 435-55. Becker JA, Hedden T, Carmasin J, Maye J, Rentz DM, Putcha D, et al. Amyloid-@ associated cortical thinning in clinically normal elderly. Ann. Neurol. 2011; 69: 1032-42. Benedictus MR, Binnewijzend M a a, Kuijer JP a, Steenwijk MD, Versteeg A, Vrenken H, et al. Brain volume and white matter hyperintensities as determinants of cerebral blood flow in Alzheimer's disease. Neurobiol. Aging 2014: 1-6. Bennett IJ, Madden DJ, Vaidya CJ, Howard D V, Howard JH. Age-related differences in multiple measures of white matter integrity: A diffusion tensor imaging study of healthy aging. Hum. Brain Mapp. 2010; 31: 378-90. Benzinger TLS, Blazey T, Jack CR, Koeppe R a, Su Y, Xiong C, et al. Regional variability of imaging biomarkers in autosomal dominant Alzheimer's disease. Proc. Natl. Acad. Sci. U. S. A. 2013; 110: E4502-9. Bertram L, Lill CM, Tanzi RE. The Genetics of Alzheimer Disease: Back to the Future. Neuron 2010; 68: 270-281. Bertram L, McQueen MB, Mullin K, Blacker D, Tanzi RE. Systematic meta-analyses of Alzheimer disease genetic association studies: the AlzGene database. Nat. Genet. 2007; 39: 17-23. Bigler ED, Lowry CM, Anderson C V, Johnson SC, Terry J, Steed M. Dementia, quantitative neuroimaging, and apolipoprotein E genotype. AJNR. Am. J. Neuroradiol. 2000; 21: 1857-68. Le Bihan D, Mangin J-FF, Poupon C, Clark CA, Pappata S, Molko N, et al. Diffusion tensor imaging: concepts and applications. J. Magn. Reson. imaging 2001; 13: 534-46. Le Bihan D. Looking into the functional architecture of the brain with diffusion MRI. Nat. Rev. Neurosci. 2003; 4: 469-80. Binnewijzend M a a, Kuijer JP a, Benedictus MR, van der Flier WM, Wink AM, Wattjes MP, et al. Cerebral blood flow measured with 3D pseudocontinuous arterial spin-labeling MR imaging in Alzheimer disease and mild cognitive impairment: a marker for disease severity. Radiology 2013;267:221-30. Blockx 1, De Groof G, Verhoye M, Van Audekerke J, Raber K, Poot D, et al. Microstructural changes observed with DKI in a transgenic Huntington rat model: evidence for abnormal neurodevelopment. Neuroimage 2012; 59: 957-67. 153 Boss HM, van den Berg-Vos RM, Scheltens P, Weinstein HC. Vascular cognitive impairment. Ned. Tijdschr. Geneeskd. 2010; 154: A1907. Bowler J V. Vascular cognitive impairment. J. Neurol. Neurosurg. Psychiatry 2005; 76 Suppl 5: v35-v44. Braak H, Braak E. Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol. 1991; 82: 239-259. Breteler MM, van Swieten JC, Bots ML, Grobbee DE, Claus JJ, van den Hout JH, et al. Cerebral white matter lesions, vascular risk factors, and cognitive function in a population-based study: the Rotterdam Study. Neurology 1994; 44: 1246-52. Brickman AM, Provenzano FA, Muraskin J, Manly JJ, Blum S, Apa Z, et al. Regional white matter hyperintensity volume, not hippocampal atrophy, predicts incident Alzheimer disease in the community. Arch. Neurol. 2012; 69: 1621-1627. Brickman AM, Zahra A, Muraskin J, Steffener J, Holland CM, Habeck C, et al. Reduction in cerebral blood flow in areas appearing as white matter hyperintensities on magnetic resonance imaging. Psychiatry Res. Neuroimaging 2009; 172: 117-20. Brinker T, Stopa E, Morrison J, Klinge P. A new look at cerebrospinal fluid circulation. Fluids Barriers CNS 2014; 11: 10. Brisset M, Boutouyrie P, Pico F, Zhu Y, Zureik M, Schilling S, et al. Large-vessel correlates of cerebral small-vessel disease. Neurology 2013; 80: 662-669. Brun A, Englund E. A white matter disorder in dementia of the Alzheimer type: a pathoanatomical study. Ann. Neurol. 1986; 19: 253-62. Bu G. Apolipoprotein E and its receptors in Alzheimer's disease: pathways, pathogenesis and therapy. Nat. Rev. Neurosci. 2009; 10: 333-44. Burns JM, Church JA, Johnson DK, Xiong C, Marcus D, Fotenos AF, et al. White matter lesions are prevalent but differentially related with cognition in aging and early Alzheimer disease. Arch. Neurol. 2005; 62: 1870-1876. Buxton RB, Frank LR, Wong EC, Siewert B, Warach S, Edelman RR. A general kinetic model for quantitative perfusion imaging with arterial spin labeling. Magn. Reson. Med. 1998; 40: 383-96. Capizzano AA, Acion L, Bekinschtein T, Furman M, Gomila H, Martinez A, et al. White matter hype rintensities are significantly associated with cortical atrophy in Alzheimer's disease. J. Neurol. Neurosurg. Psychiatry 2004; 75: 822-827. Carmichael OT, Kuller LH, Lopez OL, Thompson PM, Dutton RA, Lu A, et al. Ventricular volume and dementia progression in the Cardiovascular Health Study. Neurobiol. Aging 2010; 28: 389- 397. Casserly 1, Topol E. Convergence of atherosclerosis and Alzheimer's disease: inflammation, cholesterol, and misfolded proteins. Lancet 2004; 363: 1139-1146. Chen X, Wen W, Anstey KJ, Sachdev PS. Prevalence, incidence, and risk factors of lacunar 154 infarcts in a community sample. Neurology 2009; 73: 266-72. Chenevert TL, Stegman LD, Taylor JMG, Robertson PL, Greenberg HS, Rehemtulla A, et al. Diffusion magnetic resonance imaging: an early surrogate marker of therapeutic efficacy in brain tumors. J. Nati. Cancer Inst. 2000; 92: 2029-2036. Cheung MM, Hui ES, Chan KC, Helpern J a, Qi L, Wu EX. Does diffusion kurtosis imaging lead to better neural tissue characterization? A rodent brain maturation study. Neuroimage 2009; 45: 386-92. Chuang Y-F, An Y, Bilgel M, Wong DF, Troncoso JC, O'Brien RJ, et al. Midlife adiposity predicts earlier onset of Alzheimer's dementia, neuropathology and presymptomatic cerebral amyloid accumulation. Mol. Psychiatry 2015: 1-6. Coon KD, Ph D, Myers AJ, Craig DW, Webster JA, Pearson J V, et al. A high-density whole- genome association study reveals that APOE is the major susceptibility gene for sporadic late- onset Alzheimer's disease. J. Clin. Psychiatry 2007: 614-619. Corkin S. What's new with the amnesic patient H.M.? Nat. Rev. Neurosci. 2002; 3: 153-160. Coutu J-P, Chen JJ, Rosas HD, Salat DH. Non-Gaussian water diffusion in aging white matter. Neurobiol. Aging 2014; 35: 1412-21. Coutu J-P, Goldblatt A, Rosas HD, Salat DH. White Matter Changes are Associated with Ventricular Expansion in Aging, Mild Cognitive Impairment, and Alzheimer's Disease. J. Alzheimers. Dis. 2015; in press Crane DE, Black SE, Ganda A, Mikulis DJ, Nestor SM, Donahue MJ, et al. Gray matter blood flow and volume are reduced in association with white matter hyperintensity lesion burden: a cross- sectional MRI study. Front. Aging Neurosci. 2015; 7: 1-8. D'Agostino RB, Vasan RS, Pencina MJ, Wolf PA, Cobain M, Massaro JM, et al. General Cardiovascular Risk Profile for Use in Primary Care: The Framingham Heart Study. Circulation 2008; 117: 743-753. Dai W, Lopez OL, Carmichael OT, Becker JT, Kuller LH, Carmichael OT, et al. Mild cognitive impairment and alzheimer disease: patterns of altered cerebral blood flow at MR imaging. Radiology 2009; 250: 856-866. Davies P. The genetics of Alzheimer's disease: a review and a discussion of the implications. Neurobiol. Aging 1986; 7: 459-466. DeCarli C, Mungas D, Harvey D, Reed B, Weiner MW, Chui H, et al. Memory impairment, but not cerebrovascular disease, predicts progression of MCI to dementia. Neurology 2004; 63: 220- 227. DeCarli C, Murphy DGM, Tranh M, Grady CL, Haxby J V., Gillette J a., et al. The effect of white matter hyperintensity volume on brain structure, cognitive performance, and cerebral metabolism of glucose in 51 healthy adults. Neurology 1995; 45: 2077-2084. Dickerson BC, Bakkour A, Salat DH, Feczko E, Pacheco J, Greve DN, et al. The cortical signature 155 of Alzheimer's disease: regionally specific cortical thinning relates to symptom severity in very mild to mild AD dementia and is detectable in asymptomatic amyloid-positive individuals. Cereb. Cortex 2009; 19: 497-510. Dickerson BC, Stoub TR, Shah RC, Sperling R a, Killiany RJ, Albert MS, et al. Alzheimer-signature MRI biomarker predicts AD dementia in cognitively normal adults. Neurology 2011; 76: 1395- 402. Dong Q, Welsh RC, Chenevert TL, Carlos RC, Maly-Sundgren P, Gomez-Hassan DM, et al. Clinical applications of diffusion tensor imaging. J. Magn. Reson. Imaging 2004; 19: 6-18. Doody RS, Thomas RG, Farlow M, lwatsubo T, Vellas B, Joffe S, et al. Phase 3 trials of solanezumab for mild-to-moderate Alzheimer's disease. N. Engl. J. Med. 2014; 370: 311-21. Douaud G, Jbabdi S, Behrens TEJ, Menke R a, Gass A, Monsch AU, et al. DTI measures in crossing-fibre areas: increased diffusion anisotropy reveals early white matter alteration in MCI and mild Alzheimer's disease. Neuroimage 2011; 55: 880-90. Du AT, Schuff N, Laakso MP, Zhu XP, Jagust WJ, Yaffe K, et al. Effects of subcortical ischemic vascular dementia and AD on entorhinal cortex and hippocampus. Neurology 2002; 58: 1635- 1641. Duering M, Csanadi E, Gesierich B, Jouvent E, Herve D, Seiler S, et al. Incident lacunes preferentially localize to the edge of white matter hyperintensities: insights into the pathophysiology of cerebral small vessel disease. Brain 2013; 136: 2717-26. Falangola MF, Jensen JH, Babb JS, Hu C, Castellanos FX, Di Martino A, et al. Age-related non- Gaussian diffusion patterns in the prefrontal brain. J. Magn. Reson. Imaging JMRI 2008; 28: 1345-1350. Falangola MF, Jensen JH, Tabesh A, Hu C, Deardorff RL, Babb JS, et al. Non-Gaussian diffusion MRI assessment of brain microstructure in mild cognitive impairment and Alzheimer's disease. Magn. Reson. Imaging 2013; 31: 840-6. Fazekas F, Chawluk JB, Alavi A. MR signal abnormalities at 1.5 T in Alzheimer's dementia and normal aging. Am. J. Neuroradiol. 1987; 8: 421-426. Fazekas F, Kapeller P, Schmidt R, Offenbacher H, Payer F, Fazekas G. The relation of cerebral magnetic resonance signal hyperintensities to Alzheimer's disease. J. Neurol. Sci. 1996; 142: 121-125. Fazekas F, Kleinert R, Offenbacher H, Schmidt R, Kleinert G, Payer F, et al. Pathologic correlates of incidental MRI white matter signal hyperintensities. Neurology 1993; 43: 1683-9. Fein G, Di Sclafani V, Tanabe J, Cardenas V, Weiner MW, Jagust WJ, et al. Hippocampal and cortical atrophy predict dementia in subcortical ischemic vascular disease. Neurology 2000; 55: 1626-1635. Fernando MS, Simpson JE, Matthews F, Brayne C, Lewis CE, Barber R, et al. White matter lesions in an unselected cohort of the elderly: Molecular pathology suggests origin from chronic hypoperfusion injury. Stroke 2006; 37: 1391-1398. 156 Fieremans E, Benitez a, Jensen JH, Falangola MF, Tabesh a, Deardorff RL, et al. Novel White Matter Tract Integrity Metrics Sensitive to Alzheimer Disease Progression. AJNR. Am. J. Neuroradiol. 2013 Fieremans E, Jensen JH, Helpern J a. White matter characterization with diffusional kurtosis imaging. Neuroimage 2011; 58: 177-88. Fischl B, Salat DH, Busa E, Albert M, Dieterich M, Haselgrove C, et al. Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain. Neuron 2002; 33: 341- 55. Fischl B, Sereno M, Dale A. Cortical surface-based analysis. I. Segmentation and surface reconstruction. Neuroimage 1999; 9: 179-194. Fjell AM, Westlye LT, Amlien I, Espeseth T, Reinvang I, Raz N, et al. High consistency of regional cortical thinning in aging across multiple samples. Cereb. cortex 2009; 1.9: 2001-12. Folstein M, Folstein S, McHugh P. 'Mini-mental state': a practical method for grading the cognitive state of patients for the clinician. J. Psychiatr. Res. 1975; 12: 189-198. Forette F, Seux ML, Staessen J a., Thijs L, Birkenhager WH, Babarskiene MR, et al. Prevention of dementia in randomised double-blind placebo-controlled Systolic Hypertension in Europe (Syst- Eur) trial. Lancet 1998; 352: 1347-1351. Fornage M, Chiang YA, O'Meara ES, Psaty BM, Reiner AP, Siscovick DS, et al. Biomarkers of Inflammation and MRI-Defined Small Vessel Disease of the Brain: The Cardiovascular Health Study. Stroke. 2008; 39: 1952-9. Fujishima M, Maikusa N, Nakamura K, Nakatsuka M, Matsuda H, Meguro K. Mild cognitive impairment, poor episodic memory, and late-life depression are associated with cerebral cortical thinning and increased white matter hyperintensities. Front. Aging Neurosci. 2014; 6: 1-12. Gao FQ, Swartz RH, Scheltens P, Leibovitch FS, Kiss A, Honjo K, et al. Complexity of MRI white matter hyperintensity assessments in relation to cognition in aging and dementia from the Sunnybrook Dementia Study. J. Alzheimers. Dis. 2011; 2: 679-688. Gauthier S, Reisberg B, Zaudig M, Petersen RC, Ritchie K, Broich K, et al. Mild cognitive impairment. Lancet 2006; 367: 1262-1270. Gorelick PB, Scuteri A, Black SE, Decarli C, Greenberg SM, ladecola C, et al. Vascular contributions to cognitive impairment and dementia: a statement for healthcare professionals from the american heart association/american stroke association. Stroke. 2011; 42: 2672-713. Gorelick PB. Risk factors for vascular dementia and Alzheimer disease. Stroke. 2004; 35: 2620- 2. Gottesman RF, Coresh J, Catellier DJ, Sharrett a. R, Rose KM, Coker LH, et al. Blood pressure and white-matter disease progression in a biethnic cohort: Atherosclerosis risk in communities (ARIC) study. Stroke 2010; 41: 3-8. 157 Gouw AA, Seewann A, Flier WM Van Der, Barkhof F, Rozemuller AM, Scheltens P, et al. Heterogeneity of small vessel disease : a systematic review of MRI and histopathology correlations. J Neurol Neurosurg Psychiatry 2011; 82: 126-135. Gouw AA, Seewann A, Vrenken H, van der Flier WM, Rozemuller JM, Barkhof F, et al. Heterogeneity of white matter hyperintensities in Alzheimer's disease: post-mortem quantitative MRI and neuropathology. Brain 2008; 131: 3286-98. Greve DN, Fischl B. Accurate and robust brain image alignment using boundary-based registration. Neuroimage 2009; 48: 63-72. Greve DN, Svarer C, Fisher PM, Feng L, Hansen AE, Baare W, et al. Cortical surface-based analysis reduces bias and variance in kinetic modeling of brain PET data. Neuroimage 2014; 92: 225-236. Gurol ME, Irizarry MC, Smith EE, Raju S, Diaz-Arrastia R, Bottiglieri T, et al. Plasma beta-amyloid and white matter lesions in AD, MCI, and cerebral amyloid angiopathy. Neurology 2006; 66: 23- 9. Hanaoka T, Kimura N, Aso Y, Takemaru M, Kimura Y, Ishibashi M, et al. Relationship between white matter lesions and regional cerebral blood flow changes during longitudinal follow up in Alzheimer's disease. Geriatr. Gerontol. Int. 2015: n/a-n/a. He J, losif AM, Lee DY, Martinez 0, Chu S, Carmichael 0, et al. Brain Structure and Cerebrovascular Risk in Cognitively Impaired Patients. Arch. Neurol. 2010; 67: 1231-1237. Hedden T, Mormino EC, Amariglio RE, Younger AP, Schultz AP, Becker JA, et al. Cognitive profile of amyloid burden and white matter hyperintensities in cognitively normal older adults. J. Neurosci. 2012; 32: 16233-42. den Heijer T, Launer U, Prins ND, van Dijk EJ, Vermeer SE, Hofman A, et al. Association between blood pressure, white matter lesions, and atrophy of the medial temporal lobe. Neurology 2005; 64: 263-267. Helpern J a, Adisetiyo V, Falangola MF, Hu C, Martino A Di, Williams K, et al. Preliminary Evidence of Altered Gray and White Matter Microstructural Development in the Frontal Lobe of Adolescents With Attention-Deficit Hyperactivity Disorder : A Diffusional Kurtosis Imaging Study. J. Magn. Reson. Imaging 2011; 23: 17-23. Heston LL, Mastri AR, Anderson VE, White J. Dementia of the Alzheimer type. Clinical genetics, natural history, and associated conditions. Arch. Gen. Psychiatry 1981; 38: 1085-90. Heye AK, Thrippleton MJ, Chappell FM, Valdes Hernandez MDC, Armitage P a, Makin SD, et al. Blood pressure and sodium: association with MRI markers in cerebral small vessel disease. J. Cereb. Blood Flow Metab. 2015: 1-7. Hirono N, Kitagaki H, Kazui H, Hashimoto M, Mori E. Impact of White Matter Changes on Clinical Manifestation of Alzheimer's Disease: A Quantitative Study. Stroke 2000; 31: 2182-2188. Hofman a, Ott a, Breteler MM, Bots ML, Slooter a J, van Harskamp F, et al. Atherosclerosis, apolipoprotein E, and prevalence of dementia and Alzheimer's disease in the Rotterdam Study. 158 Lancet 1997; 349: 151-4. Hogervorst E, Ribeiro HM. Plasma homocysteine levels, cerebrovascular risk factors, and cerebral white matter changes (leukoaraiosis) in patients with Alzheimer disease. Arch. Neurol. 2002; 59: 787-793. Holland CM, Smith EE, Csapo I, Gurol ME, Brylka DA, Killiany RJ, et al. Spatial distribution of white-matter hyperintensities in Alzheimer disease, cerebral amyloid angiopathy, and healthy aging. Stroke. 2008; 39: 1127-33. Hyman BT, Phelps CH, Beach TG, Bigio EH, Cairns NJ, Carrillo MC, et al. National Institute on Aging-Alzheimer's Association guidelines for the neuropathologic assessment of Alzheimer's disease. Alzheimer's Dement. 2012; 8: 1-13. Hyman BT, Trojanowski JQ. Consensus recommendations for the postmortem diagnosis of Alzheimer disease from the National Institute on Aging and the Reagan Institute Working Group on diagnostic criteria for the neuropathological assessment of Alzheimer disease. J. Neuropathol. Exp. Neurol. 1997; 56: 1095-1097. Jack CR, Bernstein MA, Fox NC, Thompson P, Alexander G, Harvey D, et al. The Alzheimer's Disease Neuroimaging Initiative (ADNI): MRI methods. J. Magn. Reson. Imaging 2008; 27: 685- 91. Jack Jr CR, Dickson DW, Parisi JE, Xu YC, Cha RH, O'Brien PC, et al. Antemortem MRI Findings Correlate with Hippocampal Neuropathology in Typical Aging and Dementia. Neurology 2002; 58: 750-757. Jack Jr CR, Petersen RC, Xu YC, O'Brien PC, Smith GE, Ivnik RJ, et al. Prediction of AD with MRI- based hippocampal volume in mild cognitive impairment. Neurology 1999; 52: 1397-1397. Jacobs HIL, Clerx L, Gronenschild EHBM, Aalten P, Verhey FRJ. White matter hyperintensities are positively associated with cortical thickness in Alzheimer's disease. J. Alzheimers. Dis. 2014; 39: 409-22. Jeerakathil T, Wolf PA, Beiser A, Massaro J, Seshadri S, D'Agostino RB, et al. Stroke risk profile predicts white matter hyperintensity volume: the Framingham Study. Stroke. 2004; 35: 1857- 61. Jellinger K a. Pathology and pathogenesis of vascular cognitive impairment-a critical update. Front. Aging Neurosci. 2013; 5: 1-19. Jensen JH, Helpern JA, Ramani A, Lu H, Kaczynski K. Diffusional kurtosis imaging: the quantification of non-gaussian water diffusion by means of magnetic resonance imaging. Magn. Reson. Med. 2005; 53: 1432-1440. Jensen JH, Helpern JA. MRI quantification of non-Gaussian water diffusion by kurtosis analysis. NMR Biomed. 2010; 23: 698-710. Jouvent E, Viswanathan A, Chabriat H. Cerebral atrophy in cerebrovascular disorders. J. Neuroimaging 2010; 20: 213-8. 159 Jr CRJ, Knopman DS, Jagust WJ, Shaw LM, Aisen PS, Weiner MW, et al. Hypothetical model of dynamic biomarkers of the Alzheimer's pathological cascade. Lancet Neurol. 2010; 9: 119-128. Kalaria R, Akinyemi R, Ihara M. Does vascular pathology contribute to Alzheimer changes? J. Neurol. Sci. 2012; 322: 141-147. Kalaria R. Small vessel disease and Alzheimer's dementia: pathological considerations. Cerebrovasc. Dis. 2002; 13: 48-52. Kalback W, Esh C, Castaho EM, Rahman A, Kokjohn T, Luehrs DC, et al. Atherosclerosis, vascular amyloidosis and brain hypoperfusion in the pathogenesis of sporadic Alzheimer's disease. Neurol. Res. 2004; 26: 525-39. Kantarci K, Jack C. Neuroimaging in Alzheimer disease: an evidence-based review. Neuroimaging Clin. N. Am. 2003; 13: 197-209. Kantarci K, Weigand SD, Przybelski S a., Shiung MM, Whitwell JL, Negash S, et al. Risk of dementia in MCI: Combined effect of cerebrovascular disease, volumetric MRI and 1H MRS. Neurology 2009; 72: 1519-1525. Kapeller P, Barber R, Vermeulen RJ, Ader H, Scheltens P, Freidl W, et al. Visual rating of age- related white matter changes on magnetic resonance imaging: Scale comparison, interrater agreement, and correlations with quantitative measurements. Stroke 2003; 34: 441-445. Kawas CH, Kim RC, Sonnen JA, Bullain SS, Trieu T, Corrada MM. Multiple pathologies are common and related to dementia in the oldest-old. Neurology 2015; 85: 535-542. Kearney-Schwartz A, Rossignol P, Bracard S, Felblinger J, Fay R, Boivin JM, et al. Vascular Structure and Function Is Correlated to Cognitive Performance and White Matter Hypi tILeIsILies in Oider Hypertensive Patients With Subjective Memory Complaints. Stroke 2009 Kivipelto M, Helkala E, Laakso MP, Hanninen T, Hallikainen M, Alhainen K, et al. Midlife vascular risk factors and Alzheimer's disease in later life : longitudinal, population based study. BMJ 2001; 322: 1447-1451. Kivipelto M, Helkala E-LL, Hanninen T, Laakso MP, Hallikainen M, Alhainen K, et al. Midlife vascular risk factors and late-life mild cognitive impairment: A population-based study. Neurology 2001; 56: 1683-1689. Kivipelto M, Ngandu T, Fratiglioni L, Viitanen M, Kareholt I, Winblad B, et al. Obesity and vascular risk factors at midlife and the risk of dementia and Alzheimer disease. Arch. Neurol. 2005; 62: 1556-1560. Laakso M, Partanen K, Riekkinen P, Lehtovirta M, Helkala E-L, Hallikainen M, et al. Hippocampal volumes in Alzheimer's disease, Parkinson's disease with and without dementia, and in vascular dementia. Neurology 1996; 46: 678-81. Latt J, Nilsson M, Wirestam R, Stahlberg F, Karlsson N, Johansson M, et al. Regional values of diffusional kurtosis estimates in the healthy brain. J. Magn. Reson. imaging 2013; 37: 610-8. 160 Lee SH, Coutu JP, Wilkens P, Yendiki A, Rosas HD, Salat DH. Tract-based analysis of white matter degeneration in Alzheimer's disease. Neuroscience 2015; 301: 79-89. de Leeuw F-EE, de Groot JC, Achten E, Oudkerk M, Ramos LM, Heijboer R, et al. Prevalence of cerebral white matter lesions in elderly people: a population based magnetic resonance imaging study. The Rotterdam Scan Study. J. Neurol. Neurosurg. Psychiatry 2001; 70: 9-14. de Leeuw F-EE, Richard F, de Groot JC, van Duijn CM, Hofman A, Van Gijn J, et al. Interaction between hypertension, apoE, and cerebral white matter lesions. Stroke. 2004; 35: 1057-60. Lerch JP, Pruessner JC, Zijdenbos A, Hampel H, Teipel SJ, Evans AC. Focal decline of cortical thickness in Alzheimer's disease identified by computational neuroanatomy. Cereb. Cortex 2005; 15: 995-1001. Leritz EC, Shepel J, Williams VJ, Lipsitz L a, McGlinchey RE, Milberg WP, et al. Associations between T(1) white matter lesion volume and regional white matter microstructure in aging. Hum. Brain Mapp. 2014; 35: 1085-1100. Leys D, Soetaert G, Petit H, Fauquette A, Pruvo J-P, Steinling M. Periventricular and white matter magnetic resonance imaging hyperintensities do not differ between Alzheimer's disease and normal aging. Arch. Neurol. 1990; 47: 524-7. Li J, Wang Y, Zhang M, Xu Z, Gao C, Fang C, et al. Vascular risk factors promote conversion from mild cognitive impairment to Alzheimer disease. Neurology 2011; 76: 1485-91. Lindemer ER, Salat DH, Smith EE, Nguyen K, Fischl B, Greve DN. White matter signal abnormality quality differentiates mild cognitive impairment that converts to Alzheimer's disease from nonconverters. Neurobiol. Aging 2015; 36: 2447-2457. Liu C, Bammer R, Acar B, Moseley ME. Characterizing non-Gaussian diffusion by using generalized diffusion tensors. Magn. Reson. Med. 2004; 51: 924-37. Lo RY, Jagust WJ, Weiner MW, Aisen P. Vascular burden and Alzheimer disease pathologic progression. Neurology 2012; 79: 1349-55. Luh WM, Wong EC, Bandettini P a., Hyde JS. QUIPSS II with thin-slice T11 periodic saturation: A method for improving accuracy of quantitative perfusion imaging using pulsed arterial spin labeling. Magn. Reson. Med. 1999; 41: 1246-1254. Madden DJ, Bennett IJ, Song AW. Cerebral white matter integrity and cognitive aging: contributions from diffusion tensor imaging. Neuropsychol. Rev. 2009; 19: 415-35. Makedonov I, Black SE, Macintosh BJ. Cerebral small vessel disease in aging and Alzheimer's disease: a comparative study using MRI and SPECT. Eur. J. Neurol. 2013; 20: 243-50. Marchant NL, Reed BR, DeCarli CS, Madison CM, Weiner MW, Chui HC, et al. Cerebrovascular disease, 0-amyloid, and cognition in aging. Neurobiol. Aging 2012; 33: 1006.e25-36. Marchant NL, Reed BR, Sanossian N, Madison CM, Kriger S, Dhada R, et al. The Aging Brain and Cognition: Contribution of Vascular Injury and AP to Mild Cognitive Dysfunction. JAMA Neurol. 2013; 70: 488-495. 161 Markus HS, Lythgoe DJ, Ostegaard L, Sullivan MO, Williams SCR. Reduced cerebral blood flow in white matter in ischaemic leukoaraiosis demonstrated using quantitative exogenous contrast based perfusion MRI. J. Neurol. Neurosurg. Psychiatry 2000; 69: 48-53. Marstrand JR. Cerebral Perfusion and Cerebrovascular Reactivity Are Reduced in White Matter Hyperintensities. Stroke 2002; 33: 972-976. Mattsson N, Tosun D, Insel PS, Simonson A, Jack CR, Beckett L a., et al. Association of brain amyloid-P with cerebral perfusion and structure in Alzheimer's disease and mild cognitive impairment. Brain 2014; 137: 1550-1561. Mazza M, Marano G, Traversi G, Bria P, Mazza S. Primary cerebral blood flow deficiency and Alzheimer's disease: shadows and lights. J. Alzheimers. Dis. 2011; 23: 375-89. McAleese KE, Firbank M, Dey M, Colloby SJ, Walker L, Johnson M, et al. Cortical tau load is associated with white matter hyperintensities. Acta Neuropathol. Commun. 2015; 3: 60. McCarron MO, Delong D, Alberts MJ. APOE genotype as a risk factor for ischemic cerebrovascular disease: A meta-analysis. Neurology 1999; 53: 1308-1308. McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan EM. Clinical diagnosis of Alzheimer's disease: Report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer's Disease. Neurology 1984; 34: 939-944. McKhann GM, Knopman DS, Chertkow H, Hyman BT, Jack CR, Kawas CH, et al. The diagnosis of dementia due to Alzheimer's disease: Recommendations from the National Institute on Aging- Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimer's Dement. 2011; 7: 263-269. Milner B, Corkin S, Teuber H-L. Further analysis of the hippocampal amnesic syndrome: 14-year follow-up study of H.M. Neuropsychologia 1968; 6: 215-234. Mirra SS, Heyman A, McKeel D, Sumi SM, Crain BJ, Brownlee LM, et al. The Consortium to Establish a Registry for Alzheimer ' s Disease (CE RAD). Neurology 1991; 41: 479-486. Moftakhar P, Lynch MD, Pomakian JL, Vinters H V. Aquaporin Expression in the Brains of Patients With or Without Cerebral Amyloid Angiopathy. J. Neuropathol. Exp. Neurol. 2010; 69: 1201-1209. Moghekar A, KrautM, Elkins W, Troncoso J, Zonderman AB, Resnick SM, et al. Cerebral white matter disease is associated with Alzheimer pathology in a prospective cohort. Alzheimer's Dement. 2012; 8: S71-S77. Moody D, Brown W. Cerebral Microvascular Alterations in Aging, Leukoaraiosis, and Alzheimer's Disease. Ann. N. Y. Acad. Sci. 1997: 103-116. Mormino EC, Kluth JT, Madison CM, Rabinovici GD, Baker SL, Miller BL, et al. Episodic memory loss is related to hippocampal-mediated beta-amyloid deposition in elderly subjects. Brain 2009; 132: 1310-23. 162 Morris JC, Storandt M, Miller JP, McKeel DW, Price JL, Rubin EH, et al. Mild cognitive impairment represents early-stage Alzheimer disease. Arch. Neurol. 2001; 58: 397-405. Morris. The clinical dementia rating (CDR): current version and scoring rules. Neurology 1993; 43: 2412. Mortamais M, Artero S, Ritchie K. White Matter Hyperintensities as Early and Independent Predictors of Alzheimer's Disease Risk. J. Alzheimers. Dis. 2014; 42: S393-S400. Moseley ME. Diffusion tensor imaging and aging - a review. NMR Biomed. 2002; 15: 553-60. Murray AD, Staff RT, Shenkin SD, Deary IJ, Starr JM, Whalley U. Brain White Matter Hyperintensities: Relative Importance of Vascular Risk Factors in Nondemented Elderly People. Radiology 2005; 237: 251-257. Napoli C, Palinski W. Neurodegenerative diseases: insights into pathogenic mechanisms from atherosclerosis. Neurobiol. Aging 2005; 26: 293-302. Nasreddine Z, Phillips N. The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild cognitive impairment. J. ... 2005; 53: 695-699. Nettiksimmons J, DeCarli C, Landau S, Beckett L. Biological heterogeneity in ADNI amnestic mild cognitive impairment. Alzheimers. Dement. 2014; 10: 511-521.el. NIH. Estimates of Funding for Various Research, Condition, and Disease Categories. 2015 O'Brien JTJT, Erkinjuntti T, Reisberg B, Roman G, Sawada T, Pantoni L, et al. Vascular cognitive impairment. Lancet Neurol. 2003; 2: 89-98. Pantoni L, Garcia JH. Pathogenesis of Leukoaraiosis: A Review. Stroke 1997; 28: 652-9. Pantoni L. Cerebral small vessel disease : from pathogenesis and clinical characteristics to therapeutic challenges. Lancet Neurol. 2010; 9: 689-701. Panza F, Logroscino G, Imbimbo BP, Solfrizzi V. Is there still any hope for amyloid-based immunotherapy for Alzheimer's disease? Curr. Opin. Psychiatry 2014; 27: 128-37. Petersen RC, Roberts RO, Knopman DS, Boeve BF, Geda YE, Ivnik RJ, et al. Mild cognitive impairment: ten years later. Arch. Neurol. 2009; 66: 1447-1455. Petersen RC, Smith GE, Waring S, Ivnik RJ, Tangalos EG, Kokmen E. Mild cognitive impairment: clinical characterization and outcome. Arch. Neurol. 1999; 56: 303-309. Petersen RC. Mild cognitive impairment as a diagnostic entity. J. Intern. Med. 2004; 256: 183- 194. Pfefferbaum A, Sullivan E V, Hedehus M, Lim KO, Adalsteinsson E, Moseley ME. Age-related decline in brain white matter anisotropy measured with spatially corrected echo-planar diffusion tensor imaging. Magn. Reson. Med. 2000; 44: 259-268. Pike KE, Savage G, Villemagne VL, Ng S, Moss S a., Maruff P, et al. @-amyloid imaging and memory in non-demented individuals: Evidence for preclinical Alzheimer's disease. Brain 2007; 130: 2837-2844. 163 Polvikoski TM, van Straaten ECW, Barkhof F, Sulkava R, Aronen HJ, Niinist6 L, et al. Frontal lobe white matter hyperintensities and neurofibrillary pathology in the oldest old. Neurology 2010; 75: 2071-8. Promjunyakul N, Lahna D, Kaye J a., Dodge HH, Erten-Lyons D, Rooney WD, et al. Characterizing the white matter hyperintensity penumbra with cerebral blood flow measures. Neurolmage Clin. 2015; 8: 224-229. Provenzano FA, Muraskin J, Giuseppe T, Narkhede A, Wasserman BT, Griffith EY, et al. White matter hyperintensities and cerebral amyloidosis: necessary and sufficient for clinical expression of Alzheimer disease? JAMA Neurol. 2013; 70: 455-61. Puls I, Hauck K, Demuth K, Horowski a., Schlie er M, Dorfler P, et al. Diagnostic Impact of Cerebral Transit Time in the Identification of Microangiopathy in Dementia : A Transcranial Ultrasound Study. Stroke 1999; 30: 2291-2295. Querbes 0, Aubry F, Pariente J, Lotterie J-A, Demonet J-F, Duret V, et al. Early diagnosis of Alzheimer's disease using cortical thickness: impact of cognitive reserve. Brain 2009; 132: 2036- 2047. Quiroz YT, Stern CE, Reiman EM, Brickhouse M, Ruiz A, Sperling R a, et al. Cortical atrophy in presymptomatic Alzheimer's disease presenilin 1 mutation carriers. J. Neurol. Neurosurg. Psychiatry 2013; 84: 556-61. Reuter M, Fischl B. Avoiding asymmetry-induced bias in longitudinal image processing. Neuroimage 2011; 57: 19-21. Reuter M, Rosas HD, Fischl B. Highly accurate inverse consistent registration: a robust approach. Neuroimage 2010; 53: 1181-96. Reuter M, Schmansky NJ, Rosas HD, Fischl B. Within-subject template estimation for unbiased longitudinal image analysis. Neuroimage 2012; 61: 1402-18. Rezek DL, Morris JC, Fulling KH, Gado MH. Periventricular white matter lucencies in senile dementia of the Alzheimer type and in normal aging. Neurology 1987; 37: 1365. Richard E, Gouw A a, Scheltens P, Gool WA Van, van Gool W a. Vascular care in patients with Alzheimer disease with cerebrovascular lesions slows progression of white matter lesions on MRI: the evaluation of vascular care in Alzheimer's disease (EVA) study. Stroke 2010; 41: 554-6. Rostrup E, Gouw AA, Vrenken H, van Straaten ECW, Ropele S, Pantoni L, et al. The spatial distribution of age-related white matter changes as a function of vascular risk factors--results from the LADIS study. Neuroimage 2012; 60: 1597-607. Rousset OG, Ma Y, Evans AC. Correction for partial volume effects in PET: principle and validation. J. Nucl. Med. 1998; 39: 904-911. Rutten-Jacobs LCA, de Leeuw F-E, Geurts-van Bon L, Gordinou de Gouberville MC, Schepens- Franke AN, Dederen PJ, et al. White Matter Lesions Are Not Related to -Amyloid Deposition in an Autopsy-Based Study. Curr. Gerontol. Geriatr. Res. 2011; 2011: 826862. 164 Salat DH, Buckner RL, Snyder AZ, Greve DN, Desikan RSR, Busa E, et al. Thinning of the cerebral cortex in aging. Cereb. Cortex 2004; 14: 721-730. Salat DH, Greve DN, Pacheco JL, Quinn BT, Helmer KG, Buckner RL, et al. Regional white matter volume differences in nondemented aging and Alzheimer's disease. Neuroimage 2009; 44: 1247-58. Salat DH, Tuch DS, Greve DN, Van Der Kouwe AJW, Hevelone ND, Zaleta AK, et al. Age-related alterations in white matter microstructure measured by diffusion tensor imaging. Neurobiol. Aging 2005; 26: 1215-1227. Salat DH, Tuch DS, van der Kouwe AJW, Greve DN, Pappu V, Lee SY, et al. White matter pathology isolates the hippocampal formation in Alzheimer's disease. Neurobiol. Aging 2010; 31: 244-56. Salloway S, Sperling R, Fox NC, Blennow K, Klunk W, Raskind M, et al. Two phase 3 trials of bapineuzumab in mild-to-moderate Alzheimer's disease. N-. Engl. J. Med. 2014; 370: 322-33. De Santis S, Assaf Y, Jones DK. Using the biophysical CHARMED model to elucidate the underpinnings of contrast in diffusional kurtosis analysis of diffusion-weighted MRI. MAGMA 2012; 25: 267-76. De Santis S, Gabrielli A, Palombo M, Maraviglia B, Capuani S. Non-Gaussian diffusion imaging: a brief practical review. Magn. Reson. Imaging 2011; 29: 1410-6. Saunders AM, Strittmatter WJ, Schmechel D, George-Hyslop PH, Pericak-Vance MA, Joo SH, et al. Association of apolipoprotein E allele epsilon 4 with late-onset familial and sporadic Alzheimer's disease. Neurology 1993; 43: 1467-1472. Scheltens P, Barkhof F, Leys D, Pruvo JP, Nauta JJ, Vermersch P, et al. A semiquantitative rating scale for the assessment of signal hyperintensities on magnetic resonance imaging. J. Neurol. Sci. 1993; 114: 7-12. Scheltens PH, Barkhof F, Leys D, Wolters ECH, Ravid R, Kamphorst W. Histopathologic correlates of white matter changes on MRI in Alzheimer's disease and normal aging. Neurology 1995; 45: 883-888. Schmidt R, Schmidt H, Haybaeck J, Loitfelder M, Weis S, Cavalieri M, et al. Heterogeneity in age- related white matter changes. Acta Neuropathol. 2011; 122: 171-85. Schreiber SJ, Doepp F, Spruth E, Kopp U a, Valdueza JM. Ultrasonographic measurement of cerebral blood flow, cerebral circulation time and cerebral blood volume in vascular and Alzheimer's dementia. J. Neurol. 2005; 252: 1171-7. Schuff N, Matsumoto S, Kmiecik J, Studholme C, Du AT, Ezekiel F, et al. Cerebral Blood Flow in Ischemic Vascular Dementia and Alzheimer's Disease By Arterial Spin Labeling MRI. Alzheimer's Dement. 2009; 5: 454-462. Scoville WB, Milner B. Loss of recent memory after bilateral hippocampal lesions. J. Neurol. Neurosurg. Psychiatry 1957; 20: 11-21. 165 Seab JP, Jagust WJ, Wong ST, Roos MS, Reed BR, Budinger TF. Quantitative NMR measurements of hippocampal atrophy in Alzheimer's disease. Magn. Reson. Med. 1988; 8: 200-208. Selnes 0 a, Vinters H V. Vascular cognitive impairment. Nat. Clin. Pract. Neurol. 2006; 2: 538- 547. Shi J, Perry G, Smith MA, Friedland RP. Vascular abnormalities: the insidious pathogenesis of Alzheimer's disease. Neurobiol. Aging 2000; 21: 357-361. Silverberg G, Mayo M, Saul T, Fellmann J, McGuire D. Elevated cerebrospinal fluid pressure in patients with Alzheimer's disease. Cerebrospinal Fluid Res. 2006; 3: 7. Singh V, Chertkow H, Lerch JP, Evans AC, Dorr AE, Kabani NJ. Spatial patterns of cortical thinning in mild cognitive impairment and Alzheimer's disease. Brain 2006; 129: 2885-2893. Smith EE, Egorova S, Blacker D, Killiany RJ, Muzikansky A, Dickerson BC, et al. Magnetic resonance imaging white matter hyperintensities and brain volume in the prediction of mild cognitive impairment and dementia. Arch. Neurol. 2008; 65: 94-100. Smith EE, Greenberg SM. Beta-amyloid, blood vessels, and brain function. Stroke. 2009; 40: 2601-6. Smith SM, Jenkinson M, Johansen-Berg H, Rueckert D, Nichols TE, Mackay CE, et al. Tract-based spatial statistics: voxelwise analysis of multi-subject diffusion data. Neuroimage 2006; 31: 1487-505. Sofi F, Macchi C, Abbate R, Gensini GF, Casini A. Effectiveness of the mediterranean diet: Can it help delay or prevent Alzheimer's disease? J. Alzheimer's Dis. 2010; 20: 795-801. Song S-K, Sun S-W, Ju W-K, Lin S-J, Cross AH, Neufeld AH. Diffusion tensor imaging detects and differentiates axon and myelin degeneration in mouse optic nerve after retinal ischemia. Neuroimage 2003; 20: 1714-1722. Song S-K, Sun S-W, Ramsbottom MJ, Chang C, Russell J, Cross AH. Dysmyelination Revealed through MRI as Increased Radial (but Unchanged Axial) Diffusion of Water. Neuroimage 2002; 17: 1429-1436. Sotak CH. The role of diffusion tensor imaging in the evaluation of ischemic brain injury - a review. NMR Biomed. 2002; 15: 561-9. Stenset V, Johnsen L, Kocot D, Negaard A, Skinningsrud A, Gulbrandsen P, et al. Associations between white matter lesions, cerebrovascular risk factors, and low CSF Abeta42. Neurology 2006; 67: 830-3. Stout JC, Jernigan TL, Archibald SL, Salmon DP. Association of dementia severity with cortical gray matter and abnormal white matter volumes in dementia of the Alzheimer type. Arch. Neurol. 1996; 53: 742-9. Sudlow C, Martinez Gonzalez NA, Kim J, Clark C. Does apolipoprotein E genotype influence the risk of ischemic stroke, intracerebral hemorrhage, or subarachnoid hemorrhage? Systematic review and meta-analyses of 31 studies among 5961 cases and 17,965 controls. Stroke. 2006; 166 37: 364-70. Sundgren PC, Dong Q, G6mez-Hassan D, Mukherji SK, Maly P, Welsh R. Diffusion tensor imaging of the brain: review of clinical applications. Neuroradiology 2004; 46: 339-50. Taheri S, Gasparovic C, Huisa BN, Adair JC, Edmonds E, Prestopnik J, et al. Blood-brain barrier permeability abnormalities in vascular cognitive impairment. Stroke. 2011; 42: 2158-63. Teipel SJ, Meindl T, Wagner M, Kohl T, Burger K, Reiser MF, et al. White matter microstructure in relation to education in aging and alzheimer's disease. J. Alzheimer's Dis. 2009; 17: 571-583. Thal DR, Griffin WST, de Vos R a 1, Ghebremedhin E. Cerebral amyloid angiopathy and its relationship to Alzheimer's disease. Acta Neuropathol. 2008; 115: 599-609. Thal DR, Rub U, Orantes M, Braak H. Phases of AD-deposition in the human brain and its relevance for the development of AD. Neurology 2002; 58: 1791-1800. Torre J de Ia. Alzheimer disease as a vascular disorder nosological evidence. Stroke 2002; 33: 1152-1162. Tullberg M, Fletcher E, DeCarli C, Mungas D, Reed BR, Harvey DJ, et al. White matter lesions impair frontal lobe function regardless of their location. Neurology 2004; 63: 246-53. van der Veen PH, Muller M, Vincken KL, Hendrikse J, Mali WPTM, van der Graaf Y, et al. Longitudinal Relationship Between Cerebral Small-Vessel Disease and Cerebral Blood Flow: The Second Manifestations of Arterial Disease-Magnetic Resonance Study. Stroke 2015; 46: 1233- 1238. Wahlund LO, Barkhof F, Fazekas F, Bronge L, Augustin M, Sjogren M, et al. A new rating scale for age-related white matter changes applicable to MRI and CT. Stroke 2001; 32: 1318-22. Waldemar G, Christiansen P, Larsson HBW, Hogh P, Laursen H, Lassen NA, et al. White matter magnetic resonance hyperintensities in dementia of the Alzheimer type: morphological and regional cerebral blood flow correlates. J. Neurol. Neurosurg. Psychiatry 1994; 57: 1458-1465. Wardlaw JiM, Sandercock PAG, Dennis MS, Starr J. Is breakdown of the blood-brain barrier responsible for lacunar stroke, leukoaraiosis, and dementia? Stroke 2003; 34: 806-811. Wardlaw JiM, Smith EE, Biessels GJ, Cordonnier C, Fazekas F, Frayne R, et al. Neuroimaging standards for research into small vessel disease and its contribution to ageing and neurodegeneration. Lancet Neurol. 2013; 12: 822-38. Wen W, Sachdev PS, Chen X, Anstey K. Gray matter reduction is correlated with white matter hyperintensity volume: a voxel-based morphometric study in a large epidemiological sample. Neuroimage 2006; 29: 1031-9. Whitwell JL, Shiung MM, Przybelski SA, Weigand SD, Knopman DS, Boeve BF, et al. MRI patterns of atrophy associated with progression to AD in amnestic mild cognitive impairment. Neurology 2008; 70: 512-520. Wolk D a, Klunk W. Update on amyloid imaging: from healthy aging to Alzheimer's disease. Curr. Neurol. Neurosci. Rep. 2009; 9: 345-352. 167 Wong EC, Buxton RB, Frank LR. Quantitative imaging of perfusion using a single subtraction (QUIPSS and QUIPSS II). Magn. Reson. Med. 1998; 39: 702-708. Wostyn P, Audenaert K, De Deyn PP. More advanced Alzheimer's disease may be associated with a decrease in cerebrospinal fluid pressure. Cerebrospinal Fluid Res. 2009; 6: 14. Xie L, Kang H, Xu Q, Chen MJ, Liao Y, Thiyagarajan M, et al. Sleep drives metabolite clearance from the adult brain. Science 2013; 342: 373-7. Xu W, Tan L, Wang H-F, Jiang T, Tan M-S, Tan L, et al. Meta-analysis of modifiable risk factors for Alzheimer's disease. J. Neurol. Neurosurg. Psychiatry 2015: jnnp-2015-310548. Yarchoan M, Xie SX, Kling M a, Toledo JB, Wolk D a, Lee EB, et al. Cerebrovascular atherosclerosis correlates with Alzheimer pathology in neurodegenerative dementias. Brain 2012; 135: 3749-56. Yendiki A, Koldewyn K, Kakunoori S, Kanwisher N, Fischl B. Spurious group differences due to head motion in a diffusion MRI study. Neuroimage 2014; 88: 79-90. Yoshita M, Fletcher E, Harvey D, Ortega M, Martinez 0, Mungas DM, et al. Extent and distribution of white matter hyperintensities in normal aging, MCI, and AD. Neurology 2006; 67: 2192-8. Young VG, Halliday GM, Kril JJ. Neuropathologic correlates of white matter hyperintensities. Neurology 2008; 71: 804-11. Zea-Sevilla MA, Fern ndez-Blizquez MA, Calero M, Bermejo-Velasco P, R bano A. Combined Alzheimer's disease and cerebrovascular staging explains advanced dementia cognition. Alzheimer's Dement. 2015: 1-9. Zerbi V, Jansen D, Wiesmann M, Fang X, Broersen LM, Veltien A, et al. Multinutrient diets improve cerebral perfusion and neuroprotection in a murine model of Alzheimer's disease. Neurobiol. Aging 2014; 35: 600-613. Zhao JM, Clingman CS, Njrvsinen MJ, Kauppinen R a., van ZijI PCM. Oxygenation and hematocrit dependence of transverse relaxation rates of blood at 3T. Magn. Reson. Med. 2007; 58: 592-597. Zhu X, Raina AK, Perry G, Smith M a. Alzheimer's disease: The two-hit hypothesis. Lancet Neurol. 2004; 3: 219-226. Zlokovic B V. Neurovascular pathways to neurodegeneration in Alzheimer's disease and other disorders. Nat. Rev. Neurosci. 2011; 12: 723-738. 168