Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Vascular contributions to brain aging along the Alzheimer's disease continuum
(USC Thesis Other)
Vascular contributions to brain aging along the Alzheimer's disease continuum
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
VASCULAR CONTRIBUTIONS TO BRAIN AGING ALONG THE ALZHEIMER’S DISEASE
CONTINUUM
by
Meral A. Tubi
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulllment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(NEUROSCIENCE)
August 2022
Copyright 2022 Meral A. Tubi
Dedication
This work is dedicated to those in my life who have provided me unconditional love and support:
Kr,Regina,Yitzhak,andLaviTubi
ToKr,my husband and soul mate, your unconditional support and love over the last 11 years have
enabled my success and dreams to become reality. You never once doubted my ability, always provided
support (and pizza) whenever I needed it most, and gave me the condence I never knew I could have.
This body of work would not be possible without you. You will forever be my co-author in life.
ToReginaandYitzhak, my parents-in-law who provided me with the unconditional family support.
Although we didn’t speak the same language, you showed me what to value the most in life,
perseverance to achieve my dreams, and how to care for a family. Thank you for always cheering me on.
Regina, may your memory always be a blessing.
ToLavi, my loving and endearing son, you keep me humble and have enabled me to grow more than I
could ever imagined. Your entrance into this world has enabled me to be more productive and more
focused so I can be present with you. You have shown me that my most important contributions to this
world are those that enable future generations to live a more healthy and happy life.
ii
Acknowledgements
I want to extend my largest thank you to my incredible mentor, Meredith Braskie. Over the last 6 years
you have enabled my success and were a true exemplar of a supportive scientic mentor with immense
integrity. I am so grateful to be yourrst graduate student and couldn’t have asked for more from your
mentorship and my graduate experience. You provided me with incredible scientic writing, analytical,
and organizational skills. You gave me the freedom, yet supportive environment, to build the foundation
neededasanindependentscientist. Mostimportantly,Iamforevergratefulfortheamountofcompassion
and support you have shown me when life put me through the most dicult challenges.
Ialsowouldliketoextendahugethankyoutomycommitteemembers: MeredithBraskie,PaulThomp-
son,WendyMack,HelenaChui,andMaraMather. Youhaveprovidedincrediblyvaluablefeedback,guid-
ance, and questions that have fostered my success and scientic interests. I aspire to have a scientic
career like yours, with unprecedented impact, mentorship ability, and ingenuity. Each of you remind me
ofthereasonswhyIchoseaneurosciencecareerandthereasonswhyIwanttostayinsciencetoimprove
health and mitigate disease.
I would like to thank my parents for always pushing me to achieve my highest academic potential.
I would also like to thank my classmates and peers that helped me get through some of my toughest
academic courses andrst years of my PhD journey; Artemis, Lily, Sandhya, and Sadhna. You gave me
the support and camaraderie needed to get through the most dicult classes and exams. I would also
like to give a huge thank you to the collaborative and supportive Braskie Lab, which provided an ideal
iii
environment, fantastic team members, and comedic relief through game nights to help me get through
the pandemic. Also, thank you Nia, for being a great team member, organizational wizard, and friend in
and outside of the lab. Also, a huge thank you to Lauren Salminen for your continued support, guidance,
mentorship throughout my PhD journey. I am also incredibly thankful for the supportive environment
and personnel at the Imaging Genetics Center (IGC) and the Laboratory of Neuroimaging (LONI). The
incredible interdisciplinary nature and unprecedented computational power of the institute fostered an
ideal environment to complete this dissertation work.
I would also like to thank my funding sources. This work wasnancially supported in part by NIH
grantsT32MH111360(Levitt),F31AG059356(Tubi),R01AG041915(Thompson),andP50AG05142(Chui),
R01 AG054073 (Sid O’Bryant), R01 AG058162 (Chui, Marmarelis, Billinger, Zhang), and P01 AG055367
(Chen & Finch). Data collection and sharing for this project was funded by the Alzheimer’s Disease Neu-
roimagingInitiative(ADNI)(NationalInstitutesofHealthGrantU01AG024904)andDODADNI(Depart-
mentofDefenseawardnumberW81XWH-12-2-0012). ADNIisfundedbytheNationalInstituteofBiomed-
icalImagingandBioengineering,theNationalInstituteonAging,andthroughgeneroussupportfromthe
following: AbbVie, Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; Araclon Biotech;
BioClinica,Inc.;Biogen;Bristol-MyersSquibbCompany;CereSpir,Inc.;Cogstate;EisaiInc.;ElanPharma-
ceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Homann-La Roche Ltd and its aliated company
Genentech,Inc.;Fujirebio;GEHealthcare;IXICOLtd.;JanssenAlzheimerImmunotherapyResearch&De-
velopment,LLC.;Johnson&JohnsonPharmaceuticalResearch&DevelopmentLLC.;Lumosity;Lundbeck;
Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies; Novartis
PharmaceuticalsCorporation;PzerInc.;PiramalImaging;Servier;TakedaPharmaceuticalCompany;and
TransitionTherapeutics. TheCanadianInstitutesofHealthResearchcontributestosupportADNIclinical
sites in Canada. Private sector resources are coordinated by the Foundation for the National Institutes
of Health (www.fnih.org). The grantee organization is the Northern California Institute for Research and
iv
Education,andthestudyisorganizedbytheAlzheimer’sTherapeuticResearchInstituteattheUniversity
of Southern California. ADNI data are distributed by the Laboratory for Neuro Imaging at the University
of Southern California.
v
TableofContents
Dedication ii
Acknowledgements iii
ListofTables ix
ListofFigures xi
Abstract xiv
Chapter1: Introduction 1
1.1 Alzheimer’s disease (AD) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Clinical and Neuropathological Trajectory . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.2 Alzheimer’s Disease Neuroimaging Initiative (ADNI) . . . . . . . . . . . . . . . . . 4
1.1.3 AD Brain Biomarkers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.3.1 Cortical Thickness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.3.2 FDG-PET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2 Vascular Contributions to Dementia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.2.1 White Matter Hyperintensities (WMH) . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.2.2 Vascular Endothelial Growth Factor (VEGF) . . . . . . . . . . . . . . . . . . . . . . 14
1.3 Purpose and Aims . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Chapter2: Whitematterhyperintensitiesandtheirrelationshiptocognition:Eectsofseg-
mentationalgorithm 20
2.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3 Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3.1 Participants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3.2 Neuropsychological testing and diagnostic criteria . . . . . . . . . . . . . . . . . . 24
2.3.3 MRI scanning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3.4 CSF collection and analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.3.5 In-house WMH algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.3.5.1 Creating white matter masks . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.3.5.2 Segmenting WMHs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.3.5.3 Regional WMH segmentation . . . . . . . . . . . . . . . . . . . . . . . . 29
2.3.5.4 Intensity thresholding . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.3.6 Existing white matter hyperintensity segmentation algorithms . . . . . . . . . . . 31
vi
2.3.7 Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.3.7.1 Amyloid group dierences . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.3.7.2 In-house WMH analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.3.7.3 WMH segmentation comparison . . . . . . . . . . . . . . . . . . . . . . . 34
2.3.7.4 Executive function & memory analysis . . . . . . . . . . . . . . . . . . . 34
2.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.4.1 Between and within-amyloid group comparison . . . . . . . . . . . . . . . . . . . . 35
2.4.2 Regional relationship to diagnosis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.4.3 Intensity threshold modication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.4.4 Clinical associations detected by other WMH algorithms . . . . . . . . . . . . . . . 39
2.4.5 Executive function and memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.6 Supplementary Material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Chapter3: RegionalrelationshipsbetweenCSFVEGFlevelsandAlzheimer’sdiseasebrain
biomarkersandcognition 54
3.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.3.1 The Alzheimer’s Disease Neuroimaging Initiative (ADNI) . . . . . . . . . . . . . . 58
3.3.2 Participants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.3.3 CSF analytes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.3.4 APOE genotyping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.3.5 MRI processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.3.6 FDG-PET processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.3.7 Neuropsychological assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.3.8 Statistical analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.3.8.1 Relationship between VEGF and Amyloid levels . . . . . . . . . . . . . . 64
3.3.8.2 VEGF associations with AD brain biomarkers: FDG-PET & MRI cortical
thickness ROIs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.3.8.3 VEGFassociationswithcognitivemeasures: asecondaryreproducibility
analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.3.8.4 Mediation analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.4.1 Relationship between VEGF and Aβ levels . . . . . . . . . . . . . . . . . . . . . . . 67
3.4.2 VEGF associations with regional FDG-PET SUVR . . . . . . . . . . . . . . . . . . . 67
3.4.3 VEGF associations with regional cortical thickness . . . . . . . . . . . . . . . . . . 70
3.4.4 VEGF associations with cognitive measures: A secondary reproducibility analysis . 71
3.4.5 Mediation analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.6 Supplementary Material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Chapter4: WhitematterhyperintensityvolumemodifytheassociationbetweenCSFvascu-
larbiomarkersandregionalFDG-PETalongtheAlzheimer’sdiseasecontinuum 84
4.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.3.1 The Alzheimer’s Disease Neuroimaging Initiative (ADNI) . . . . . . . . . . . . . . 89
vii
4.3.2 Participants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.3.3 Imaging Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.3.4 White Matter Hyperintensity (WMH) Segmentation . . . . . . . . . . . . . . . . . 92
4.3.5 FDG-PET Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.3.6 CSF Analytes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.3.7 APOE Genotyping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.3.8 Statistical Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.3.8.1 Vascular CSF biomarkers relationship to FDG-PET SUVR in vascular
territories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.3.8.2 WMH interaction with CSF biomarkers on regional FDG-PET SUVR . . 96
4.3.8.3 Group dierences in the WMH interaction with CSF biomarkers on
FDG-PET SUVR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.3.8.4 Vascular CSF biomarkers relationship to FDG-PET SUVR in cortically
dened ROIs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.4.1 Vascular CSF biomarkers relationship to whole brain FDG-PET SUVR . . . . . . . 98
4.4.2 Dierences in vascular territory FDG-PET SUVR . . . . . . . . . . . . . . . . . . . 98
4.4.3 VEGF and CRP association with vascular territory FDG-PET SUVR . . . . . . . . . 98
4.4.4 WMH interaction with CSF biomarkers on FDG-PET SUVR . . . . . . . . . . . . . 100
4.4.4.1 Amyloid-positivity dierences . . . . . . . . . . . . . . . . . . . . . . . . 100
4.4.4.2 APOE4-carrier dierences . . . . . . . . . . . . . . . . . . . . . . . . . . 103
4.4.4.3 Sex Dierences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.4.5 Cortical DK Atlas ROI associations . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
4.6 Supplementary Material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
Chapter5: Conclusions 116
References 118
viii
ListofTables
2.1 Demographicfeaturesofthesampleanalyzed ...................... 24
2.2 Voxel,volume,andintensityinformationfromtheparticipantsusedtocalculate
theintensityratio ....................................... 29
2.3 Associations between in-house derived WMH volume by region and clinical
diagnosisinAβ-participants ................................. 36
2.4 Associationsbetweenin-housederivedWMHvolumebysubregionanddiagnosis
inAβ-participants ....................................... 37
2.5 Associationsbetweenin-housederivedtotalWMHvolumeandclinicaldiagnosis
byintensitythresholdinAβ-participants ......................... 38
2.6 Associations between WMH volume by segmentation method and clinical
diagnosisinAβ-participants ................................. 40
2.7 One-way ANOVA with pairwise comparisons. p-values displayed are corrected
formultiplecomparisonsusingthefalsediscoveryrate(FDR)............. 42
S2.1 Associations between in-house derived WMH volume by region and clinical
diagnosisinAβ-participantswithscannertypeaddedasacovariate......... 49
S2.2 Associations between in-house derived WMH volume by region and executive
functioninAβ-participants.................................. 50
S2.3 Associationsbetweenin-housederivedWMHvolumebyregionandmemoryin
Aβ-participants.......................................... 50
S2.4 Associations between in-house derived total WMH volume by threshold and
clinicaldiagnosisinAβ+participants ............................ 50
S2.5 Associations between in-house derived WMH volume by region and clinical
diagnosisinAβ+participants................................. 51
ix
S2.6 Associationsbetweenin-housederivedWMHvolumebysub-regionanddiagnosis
inAβ+participants ....................................... 51
S2.7 Associations between WMH volume by segmentation method and clinical
diagnosisinAβ+participants................................. 51
S2.8 AssociationsbetweenWMHvolumeandclinicaldiagnosisinAβ-participantsby
LGAlesionbeliefmapintensitythreshold......................... 52
S2.9 AssociationsbetweenWMHvolumeandclinicaldiagnosisinAβ-participantsby
BIANCAprobabilitymapthreshold ............................. 52
3.1 ParticipantCharacteristics .................................. 60
3.2 AssociationsbetweenVEGFlevelsandFDG-PETSUVRROIanalyzedineachAβ
stratum .............................................. 69
3.3 InteractionsbetweenCSFVEGFlevelsandCSFt-tauandp-tauonFDG-PETROIs
(n=158)............................................... 70
3.4 AssociationsbetweenVEGFandcorticalthicknessROIanalyzedineachAβstratum 71
3.5 RegionalmediationeectsinAβ+participants(n=116) ................. 73
S3.1 AssociationstoVEGFinallparticipants(N=310)andbyAβstratum ......... 79
S3.2 Interactionsbetween VEGFlevels andCSF t-tauand p-tauon cortical thickness
ROIs(N=310) ........................................... 80
S3.3 AssociationbetweenVEGFlevelstoADNI-EFsubtestsinAβ+participants(n=215) 81
4.1 CohortCharacteristics(n=158) ................................ 91
S4.1 VascularCSFbiomarkerassociationtoFDG-PETvascularterritoryregions(n=158) 115
x
ListofFigures
1.1 Model of AD biomarker abnormality trajectory over time. Figure published in
Jacketal. 2013........................................... 6
1.2 DKatlasdepictingthecorticalboundarydelineationsof34ROIs.Figurepublished
inDesikanetal. 2006. ..................................... 8
2.1 Flow diagram illustrating the workow of our method to segment WMH. The
intensity ratio is dened as
Minimum Intensity of WMH
Mean WM intensity without WMH
..................... 28
2.2 Image on the left depicts the coronal view of the MNI lobe map atlas from FSL
5.0.7(maxprob-thr0-1mm). Theimageontherightdepictsthelobemapafterwe
manuallyextendedtheboundariesofthelobesintothewhitematter......... 30
2.3 WMH boundary segmentation based on varying intensity thresholds of the
study-specicintensity ratio. Thefar-right imageillustrates the85%, 100%, and
105% threshold masks all overlaid on the base FLAIR image for comparison
purposes. ............................................. 38
2.4 Range of WMH severity and variation in white matter segmentation methods.
The severity was evaluated as WMH volume corrected for ICV. We dened mild
WMH volume in a participant, when the individual’s total WMH volume was less than
the mean total WMH volume across participants. Moderate WMH volume was dened as
the individual’s total WMH volume being between the mean and two standard deviations
above the mean across participants, and severe WMH volume when the individual’s
total WMH volume was greater than two standard deviations above the mean across
participants. For BIANCA, “masked” indicates that the same WM mask generated for our
in-house algorithm was used as input for the analysis . . . . . . . . . . . . . . . . . . . . . 41
S2.1 Testing of the division of periventricular and deep WMH boundaries by kernel
sizes 4, 6, and 9. Blue represents lesion classied as periventricular WMH and red
represents lesion classied as deep WMH . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
S2.2 VariationintheLGAWMHsegmentationbythresholdofthelesionbeliefmap
intensity. ............................................. 52
xi
S2.3 VariationintheBIANCAWMHsegmentationbytheprobabilitymapthreshold
andoptionalWMmaskinput. ................................ 53
3.1 AD-signatureROIsextractedfromFreeSurfermappedonaskull-strippedbrain . 62
3.2 Correlation plots between VEGF and FDG-PET SUVR in the inferior parietal
cortex,MTG,andITG,stratiedbyCSFamyloid(<192pg/mL),t-tau(>93pg/mL),
andp-taupositivity(>23pg/mL)............................... 68
S3.1 Violin plots of group dierences in CSF VEGF levels by amyloid and t-tau
positivity. Individuals who were both Aβ- and t-tau- (mean ± SD = 2.73 ± 0.12) had
signicantly higher VEGF levels than individuals who were A β+ and t-tau- (mean ± SD
= 2.63 ± 0.12; t(175.85) = 5.28,p< 0.001). Individuals who were Aβ+ and t-tau- had
signicantly lower VEGF levels than individuals who were both A β+ and tau+ (mean ±
SD = 2.73 ± 0.12;t(188.1) = 6.24,p< 0.001). No signicant dierences in VEGF levels
werefoundbetweentheAβ-t-tau-andAβ+t-tau+groups(t(179.97) = 0.44,p=0.66).
Group dierences were not evaluated in the A β- tau+ group (mean ± SD = 2.88 ± 0.09)
since the sample size was small (n=9) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
S3.2 Piecewise linear mixed eects spline regression between CSF Aβ levels (with a
knotat192pg/mL),andVEGF(logtransformed)levels,whichdemonstratesthat
evenatamyloidlevelsthatareconsidered“normal”intheAβ-group,VEGFmay
be inhibited with increasing amyloid load (i.e., lower CSF Aβ levels). While the
magnitude of the slope of continuous Aβlevels in those having Aβ levels < 192pg/mL
(i.e.,Aβ+participants)toVEGFwassmallerthantheslopeofAβinthosehavingAβlevels
> 192pg/mL (i.e., Aβ- participants) to VEGF, the slope estimates did not signicantly
dier( β(SE)=0.0004(0.0005);p=0.440). RedcolorindicatesCSFAβ+status,whilegrey
color indicates Aβ- status . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.1 Participant specic example of the FDG-PET Processing pipeline to extract
vascular territory ROIs in FDG-PET space. GM = Gray Matter, ACA = Anterior
Cerebral Artery, MCA = Middle Cerebral Artery, PCA = Posterior Cerebral Artery, SUVR
= Standard Uptake Value Ratio. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.2 FDG-PET SUVR mean dierences by GM vascular territory ROI (ACA, MCA,
PCA). ACA = Anterior Cerebral Artery, MCA = Middle Cerebral Artery, PCA = Posterior
Cerebral Artery, SUVR = Standard Uptake Value Ratio. * indicatesp< 0.05. NS. = not
signicant. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.3 Interaction plots between VEGF and CRP and WMH on FDG-PET SUVR in the
MCAterritory. InthosewithlowWMHvolumes,higherCSFVEGFwasassociatedwith
higher FDG-PET SUVR, but as WMH load increases the relationship is attenuated and
becomesnegativeinthosewiththehighestWMHload(>2SDofmean). Inthosewithlow
WMH volumes, higher CRP was associated with lower FDG-PET, with the relationship
becoming stronger with increasing WMH load. MCA = Middle Cerebral Artery, SUVR =
Standard Uptake Value Ratio, SD = Standard Deviation. . . . . . . . . . . . . . . . . . . . . 101
xii
4.4 Forest plots illustrating the beta-estimates and 95% condence intervals for the
VEGF by WMH and CRP by WMH interaction on FDG-PET SUVR in the ACA,
MCA, PCA territories in the entire cohort (Analysis 1), and by stratication of
participantsbyamyloid-betapositivity(Analysis2),APOE4carriers(Analysis3),
andsex(Analysis4). InthosewithlowWMHvolumes,higherCSFVEGFwasassociated
with higher FDG-PET SUVR, but as WMH load increases the relationship is attenuated.
InthosewithlowWMHvolumes,higherCRPVEGFwasassociatedwithlowerFDG-PET,
with the relationship becoming stronger with increasing WMH load. ACA = Anterior
Cerebral Artery, MCA = Middle Cerebral Artery, PCA = Posterior Cerebral Artery, SUVR
= Standard Uptake Value Ratio. * Indicatesp< 0.05...................... 102
4.5 ForestplotofbothmaineectsbetweenCSFBiomarkers(VEGF,CRP)andFDG-
PET SUVR and interactions between CSF Biomarkers (VEGF, CRP) and WMH
volume on FDG-PET SUVR in FreeSurfer Cortical ROIs. For ease of visualization,
ROIs are color coded by the vascular territory they are most prominent in, but ROI
vascular territory locations may vary by individual or may be apparent in more than one
region. (Red=ACA, Green=MCA, and Blue=PCA). Scales vary across each forest plot. ROI
= Region of interest, STS = Superior temporal sulcus, ACA = Anterior Cerebral Artery,
MCA = Middle Cerebral Artery, PCA = Posterior Cerebral Artery. * indicates the p-value
< 0.05 prior to FDR correction. ** indicates p-value < 0.05 after FDR correction. . . . . . . 106
xiii
Abstract
My dissertation aims to clarify how vascular dysfunction contributes to suboptimal brain aging in older
adults at risk for Alzheimer’s disease (AD). AD is a multifactorial and heterogeneous disease, with
increasing evidence demonstrating that vascular dysfunction contributes to AD and dementia risk. By
the time an individual has evidence of cognitive decline and/or an AD diagnosis, the brain has already
undergone severe structural and functional degeneration, making it essential to assess vascular factors
that may contribute to early AD brain dysfunction. While AD neuroimaging brain biomarkers (cortical
thinning,brainFDG-PET)uctuateearlyandoftenfollowaspecicspatialandtemporalsequenceinAD
disease course, heterogeneity still exists. Critically, sources of the heterogeneity in AD neuroimaging
phenotype patterns have yet to be unmasked and few studies have mapped how vascular factors
regionally relate to AD-brain biomarkers and cognition. To address this, I leveraged the Alzheimer’s
Disease Neuroimaging Initiative (ADNI) dataset (ages 55-90) to investigate how markers of vascular
dysfunction (e.g., white matter hyperintensities (WMH) and cerebral spinaluid (CSF) vascular
endothelial growth factor (VEGF)) relate to brain aging biomarkers and cognition in individuals across
the AD continuum. The work unveils sources of variability in AD biomarkers by demonstrating both
independent and interactive relationships between vascular dysfunction and AD neuropathology on
brain health. Collectively, unmasking vascular contributions to AD can help promote deeper
phenotyping of disease to enhance precision medicine and intervention eorts targeted at those who are
most at risk for cognitive decline.
xiv
Chapter1
Introduction
1.1 Alzheimer’sdisease(AD)
1.1.1 ClinicalandNeuropathologicalTrajectory
Alzheimer’s disease (AD) is the most common cause of dementia, with the burden of AD set to increase
from 58 million to 88 million people by 2050 (U.S. Census Bureau, 2021). AD is expensive, devastating,
andhasnopreventionorcure(U.S.CensusBureau,2021),creatinganurgentneedtopreventandidentify
disease-modifying solutions. A growing list of risk factors for sporadic AD and related dementias con-
found eorts to understand cause and consequences of the disease. Risk factors include genetic (APOE4
genotype), sex, environment (e.g., pollution), and vascular disease, which may all interact and complicate
the risk prole of AD further ( Silva et al., 2019; Riedel et al., 2016; Chin-Chan et al., 2015). For example,
carriers of one APOE4 allele carry 3 times the risk and a carriers of two APOE4 alleles carry up to 12
timestheriskofADthannon-carriers(innon-HispanicWhitesamples)(Loyetal.,2014;Holtzmanetal.,
2012; Michaelson, 2016). Critically, sex also interacts with APOE4 genotype to amplify risk of AD further
in females (Riedel et al., 2016). Vascular disease is often co-morbid with AD and can independently and
interactively contribute to AD pathogenesis, making it an essential area of investigation (Sweeney et al.,
1
2019). Given the interwoven nature of vascular dysfunction and AD pathogenesis, my dissertation aims
to begin to unravel how these factors relate to brain and cognition biomarkers along the AD continuum.
Historically, AD has been classied according to a clinical framework and post-mortem neuropatho-
logical reports of amyloid-beta plaques (Aβ) and neurobrillary tangles (NFTs) ( Braak and Braak, 1991).
Neurotoxic amyloid plaques accumulate outside of neurons and are derived from the accumulation of
Aβ-brils generated from mis-cut amyloid precursor protein (APP) cleavage ( O’brien and Wong, 2011).
Amyloid plaque accumulation follows a prototypical spatial sequence; amyloid rst accumulates in the
brainstem and basal frontal and temporal cortices and eventually spreads diusely throughout the cortex
(BraakandBraak,1991). Followingamyloidplaqueaccumulation,NFTsaccumulatewithintheneuron,as
a result of hyperphosphylation of tau in destabilized microtubules. The spatial progression of NFTs also
follows a prototypical spatial trajectory beginning in the transentorhinal cortex, spreading to limbic re-
gionsandeventuallydiuselytoassociationcortices(BraakandBraak,1991). WhileADneuropathological
markersarehallmarkfeaturesofAD,interventionalstudiestargetingthesemarkershaveoverwhelmingly
failedto prevent or halt disease course(Plascencia-Villaand Perry,2020). Consequently,both the ineec-
tive targeting of neuropathology to modify disease course and the dissemination of multi-modal studies
havefosteredashiftinthetheADdiagnosisframeworkfromaclinicalmodeltoaclinical-biologicalcon-
tinuum (i.e., ATN framework) (Jack Jr et al., 2018).
Recent advances in our understanding of AD have evolved from in-vivo multi-modal neuroimag-
ing and blood and cerebrospinal uid (CSF)-based studies. Amyloid and tau can also be detected in-
vivo through CSF and plasma assays, which provide a general indication of amyloid and tau burden in
the brain. However, development of positron emission tomography (PET) ligand tracers to detect amy-
loid (e.g. [11C]Pittsburgh-Compound-B, [18F]orbetaben, [18F]orbetapir) and tau (e.g., [18F]THK5317,
[18F]THK5351, [18F]AV1451, [11C]PBB3, [18F]-PI-2620) provide unprecedented ability to spatially map
2
neuropathologicalburdenandprogressionin-vivo(Krishnadasetal.,2021;Okamuraetal.,2016). Thesein-
vivomethodsofneuropathologicaldetectionhavedemonstratedthatamyloidaccumulationoccursyears,
or even decades, before symptom onset, and is followed by NFT accumulation, and eventual gray matter
(GM) neurodegeneration (Jack Jr and Holtzman, 2013). These pathophysiological changes both indepen-
dently and synergistically impact brain function that progressively manifests in the classical cognitive
prole of AD.
Clinically, cognitive diagnosis classication schemes have been used to group cognitive proles to-
gethertobettermapwhereanindividualliesontheAD-continuum. Inthesimplestschemes,participants
are classied by cognitive ability using a series of neuropsychological batteries to determine if someone
iscognitivelynormal(CN),withmildcognitiveimpairment(MCI),orprobableAD(BudsonandSolomon,
2012). Tofurthercapturethespectrumofcognitivedysfunction(e.g.,decitsinepisodicmemory,seman-
tic memory, and executive function), a broad neuropsychological test battery is used to detect cognitive
decits in each cognitive domain. Typically, the earliest signs of cognitive deterioration in AD include
episodicmemoryloss(identiedontestssuchasfreerecall,recognition,paired-associatelearning)( Arnáiz
and Almkvist, 2003). Episodic memory impairment in AD typically stems from inadequate consolidation
and storage of new information (rather than impaired retrieval of information) and functionally maps to
limbic regions (e.g., hippocampus) and prefrontal cortical regions where AD pathology is most present
(Desgrangesetal.,1998;HodgesandPatterson,1995). Theotherprimarycognitivedomainimpactedearly
inADisexecutivefunction,whichistheabilitytomentallymanipulateinformation,problemsolve,andex-
ecutegoal-directedbehavior. Executivefunctioncanbedetectedontests,suchasTrailsMakingtest-Part
B,andfunctionallymapstotheprefrontalcortex(Baudicetal.,2006). Asbothpathologyandclinicalsymp-
toms progress, cognitive domains of verbaluency, semantic categorization, and object naming typically
begin to be impaired (Hodges and Patterson, 1995). Semantic memory impairments typically map to the
temporalcortex,aswellasfrontalandparietalassociationcortices(HodgesandPatterson,1995;Chertkow
3
et al., 2008). Later in disease course AD cognitive dyfunction also impacts the ability to sustain attention,
constructional praxis, and language abilities (Baudic et al., 2006). While AD progression often follows a
prototypicalcognitiveandneuropathologicaltrajectorywithtemporalandspatialcorrelates,heterogene-
ityandindividualvariabilityinADstillexist,makingndingacure,implementingpreventioneorts,and
understanding risk even harder to map. As such, it is essential to continue to assess mechanisms linked
tovariouscognitiveandphysiologicalprocessesinAD.Toaddressthisinmydissertation,Ievaluatehow
vascular markers (e.g., white matter hyperintensities, and vascular endothelial growth factor) are related
to domain and test-specic measures of cognitive function to better understand mechanisms contribut-
ing to variability in disease. Eorts, such as these, to capture variability in AD and dissect mechanisms
contributingtoheterogeneityinADphenotypepatternsarebestsuitedforlargelongitudinalconsortium
eorts that integrate multi-modal data and incorporate diverse populations.
1.1.2 Alzheimer’sDiseaseNeuroimagingInitiative(ADNI)
TheAlzheimer’sDiseaseNeuroimagingInitiative(ADNI),pioneeredbyDr. MichaelW.Weinerin2004,is
a longitudinal multi-center (i.e., 57 site) neuroimaging study of AD risk and progression in North Amer-
ica. This large consortium integrates multi-modal neuroimaging, genetic, proteomic, biochemical, neu-
ropathological,neuropsychologicaldataofindividualsbetweentheagesof55-90todetectandtrackearly
risk factors and progression of AD. This open-access initiative provides shared access through the USC
LaboratoryofNeuroImaging’sImageandDataArchive(IDA).Theconsortiumwasoneoftherstlarge-
scale initiatives that provided open-access to multi-modal neuroimaging data to better understand and
map variability in disease risk and progression, a core need in unraveling mechanisms and consequences
of AD.
Data collection was iterated through stages (e.g., ADNI 1, ADNI GO, ADNI 2, ADNI 3) with dierent
datadomainsoffocuscollectedduringeachstage. Forexample,ADNI1wastheonlyADNIcollectionstage
4
to acquire CSF used for multi-proteomic assay analysis. Therefore, I leveraged ADNI 1 data for disserta-
tion Chapters 3 and 4 to assess the relationship between CSF biomarkers and multi-modal neuroimaging
indices. Also, since ADNI 1 did not have Fluid-Attenuated Inversion Recovery (FLAIR) imaging available,
which is sensitive to pathology (e..g, white matter hyperintensities), I used ADNI 2 data for my work in
Chapter 2.
1.1.3 ADBrainBiomarkers
By the time a clinical AD diagnosis is made, the brain has undergone massive, and likely irreversible
damage. It is therefore crucial to identify non-invasive early biomarkers in individuals most at risk for
cognitive decline. Neuroimaging biomarkers are a useful, sensitive, and often non-invasive tool to assess
physiological changes both before and during cognitive decline has begun. While amyloid, tau, and neu-
rodegeneration are the hallmarks of AD, pathophysiological changes also accompany and contribute to
disease progression (Figure 1.1). For example, early in AD progression, brain changes may include re-
gional reductions in cerebral bloodow (CBF), glucose dysmetabolism, functional connectivity changes,
gray matter atrophy, and even loss of white matter (WM) microstructural integrity (Ruan et al., 2016;
Jack Jr and Holtzman, 2013). Importantly, each of these altered physiological, functional, and structural
brain changes can be identied using multi-modal MR and PET imaging methods in-vivo and over mul-
tiple time points. In my dissertation, I will focus on 2 neuroimaging outcome markers associated with
AD progression: regional cortical thickness derived from T1-weighted structural MRI scans and regional
FDG-PET standard uptake value ratio (SUVR).
1.1.3.1 CorticalThickness
Progressive neurodegeneration is a hallmark feature of AD and occurs after neuropathology is present.
GraymatterindicesofvolumeandthicknessarecorrelatedwithgreaterADneuropathologicalburdenand
5
Figure1.1: ModelofADbiomarkerabnormalitytrajectoryovertime. FigurepublishedinJacket
al. 2013.
worse cognitive performance along the AD continuum (Jack et al., 2019; Querbes et al., 2009; Dickerson
etal.,2009;Kauretal.,2014). GraymattervolumeandthicknesscanbeeasilyidentiedusingT1-weighted
6
structural MRI, a non-invasive in-vivo imaging technique. The most common structural brain indices in-
clude volume, cortical thickness, surface area, and intracranial volume (ICV). These indices are often ex-
tracted using automated pipelines and publicly available software, such as FreeSurfer (Fischl, 2012) and
FSL (Jenkinson et al., 2012; Smith et al., 2004). FreeSurfer is an open source tool that uses a probabilistic
atlas-based segmentation method to automatically process and analyze MR images to extract template-
derived neuroimaging measures of interest. Conventionally, regions of interest (ROI) in FreeSurfer (and
other neuroimaging softwares) are automatically dened by templates based on neuroanatomical bound-
aries of cortical gyri and sulci. The most common atlas is the Desikan-Killany (DK) Atlas that extracts 34
neuroanatomicalregions(Figure1.2),whichhasbeenshowntobereliableandanatomicallyvalid(Desikan
etal.,2006). IusedthisatlastoextracttheADcorticalthinningsignatureROIsinChapter3andthecortex
wide FDG-PET regions in Chapter 4.
Cortical thickness is a more sensitive measure of early AD-related neurodegeneration than cortical
volume measurements (Gómez-Isla et al., 1996; Burggren et al., 2008a). Therefore, I selected to evaluate
regional cortical thickness instead of cortical volume measurements in Chapter 3. Also, the topographic
distribution of gray matter atrophy in AD typically occurs in medial temporal lobe structures (e.g, hip-
pocampus) and cortical regions that map to an AD cortical thinning signature (Wang et al., 2015a); the
entorhinal cortex (ERC), posterior cingulate, superior (STG), middle (MTG), and inferior temporal gyri
(ITG), fusiform gyri, superior and inferior parietal cortex, and precuneus. These regions may be more
vulnerabletoneurodegenerationbothasaresultofdirectneurotoxiceectsfromADpathology(e.g.,tau)
(Gao et al., 2018) and indirectly from other pathways, such as vascular dysfunction (Zlokovic, 2011). Ac-
cordingly, in Chapter 3 of this dissertation, I investigate how a vascular biomarker (i.e., VEGF) relates to
cortical thinning and FDG-PET signal in these AD cortical thinning signature ROIs.
7
Figure1.2: DKatlasdepictingthecorticalboundarydelineationsof34ROIs. Figurepublishedin
Desikanetal. 2006.
1.1.3.2 FDG-PET
PETisanon-invasivemolecularimagingtechniquethatisessentialforidentifyingindividualswithbiolog-
ically dened AD ( Maclin et al., 2019). The mostcommon PET tracers used in ADresearch include mark-
ers of amyloid burden (e.g., Pittsburgh Compound-B (11C-PiB), 18F-Florbetapir) and glucose metabolism
(FDG), though many other tracers are used in clinical and pre-clinical studies (Márquez and Yassa, 2019).
Alterations in FDG-PET signal are seen early in AD pathogenesis and are an indirect marker of neurode-
generation and disease progression (Jack Jr and Holtzman, 2013).
8
FDG-PET is most commonly known to be a measure of neuronal glucose metabolism. FDG-PET uses
a ligand to track endothelial-specic glucose transporter-1 (GLUT1) transport across the blood brain bar-
rier(BBB).However,FDG-PETcannotreliablytrackGLUT1throughit’sentiremetabolicfate,resultingin
variousinterpretationsofFDG-PETsignal. Forexample,FDG-PETmayreectBBBdysfunction( Sweeney
etal.,2019),microglial-inducedinammation( Choietal.,2021;Xiangetal.,2021),orastroglialglutamate
transport (Zimmer et al., 2017), reecting an age-dependent shift from neuronal to astrocytic function
and energy consumption (Jiang and Cadenas, 2014). This shift to astrocytic function may occur because
neuronsarealsosubstantiallyfueledbylactateproducedbyastrocytes,whichoccursthroughaerobicgly-
colysisthroughglutamaterecycling(JiangandCadenas,2014). FDG-PETregionalalterationsalsospatially
alignwithCBFdisturbancesderivedfromArterialSpinalLabeling(ASL)MRI,withsomestudiesevensug-
gesting that ASL can be used as a surrogate for FDG-PET (Dolui et al., 2020; Musiek et al., 2012). I make
noteofthesevariationsinsignalinterpretationthroughoutthisworkinChapters3and4. Despitesources
of variability in the interpretation of FDG-PET signal changes, AD studies demonstrate consistent reduc-
tions in FDG-PET signal early in AD progression and can reliably dierentiate individuals with AD from
cognitivelynormalcontrols andthosewithotherdementias(Salaetal.,2020;Kantarcietal.,2021;Pagani
et al., 2015).
Regional reductions in FDG-PET signal correlate with cognitive function and also exhibit a classic
signaturespanningtheparietal,temporal,andfrontalcortices(Edisonetal.,2007;Salaetal.,2020). Specif-
ically, the precuneus, posterior cingulate cortex (PCC), and lateral temporo-parietal cortex demonstrate
consistent metabolic alterations across studies (Perani, 2014; Drzezga et al., 2003; Mosconi et al., 2009).
These regions align with the brain’s default mode network (DMN), a set of functionally and anatomi-
cally connected brain regions that consistently elicit higher neuronal activity when at rest compared to
whenengagedinexternaltask(Raichleetal.,2001). Consequently,theseregionsmaybemorevulnerable
9
to hypometabolism because of higher metabolic requirements of the DMN, making them also more sus-
ceptible to neurodegeneration in AD (Vlassenko et al., 2010; Goyal et al., 2014). The regional specicity
of these alterations also illuminate the role of bioenergetic dysfunction as a possible mechanism driving
pathophysiological changes in AD.
Both in AD animal models and in humans along the the AD continuum, consistent and substantial
regionalreductionsincerebralglucosemetabolismandbioenergeticcapacityhavebeennoted(Dingetal.,
2013; Yao et al., 2009; Mosconi et al., 2008; Kato et al., 2016). Evidence from multi-modal studies also
demonstrate that the bioenergetic coupling between glucose metabolism and neural activity is impaired
in amnestic MCI and AD patients (Marchitelli et al., 2018a). The coupling of glucose metabolism and
regional cerebral bloodow (CBF) in response to increased synaptic activity is a multi-faceted, complex
processthatrequiresthecoordinationofmultiplecelltypesfromtheneurovascularunit(NVU)toworkin
congruence with various vasoactive agents and vascular protein biomarkers (Iadecola, 2004; Blonz, 2017;
Kisler et al., 2017; Lourenço et al., 2017). Dysfunction of any one of these numerous cell types, vasoactive
agents,orproteinsresponsibleforcoordinatingandmaintainingahemodynamicresponsemaycontribute
to the uncoupling of neurovascular-neuroenergetic responses seen in the pathogenesis of AD. Further,
the loss of estrogen at menopause may ignite an AD bioenergetic phenotype; peri- and post-menopausal
females(ages40-60years)havegreaterAβdepositionandmitochondrialenzymedeciency,lowerregional
brain glucose metabolism, and lower brain volume in AD-signature regions (precuneus, PCC, superior
andinferiorparietal,lateralandmedialtemporalcortices)comparedtopre-menopausalfemales(Mosconi
et al., 2017, 2018). Therefore, identifying physiological mechanisms and interactions that contribute to
alterations in neurovascular-neuroenergetic coupling, such as seen in FDG-PET, may provide valuable
insight into possible intervention strategies to combat cognitive decline.
10
While regional vulnerabilities are apparent in FDG-PET, the underlying vascular contributions and
biomarker associations linked to these changes remain unknown. Therefore, my dissertation aims to fur-
ther investigate factors linked to regional FDG-PET signal, such as CSF vascular biomarkers (Chapters
3 and 4). This body of work also aims to address whether FDG-PET signal may mediate a previously
detectedassociationbetweenaCSFvascularbiomarkers(i.e.,vascularendothelialgrowthfactor)andcog-
nition (Chapter 3). These insights will furtherilluminate the direct and indirectrole vascular dysfunction
and FDG-PET play in AD pathogenesis.
1.2 VascularContributionstoDementia
Cerebrovasculardisease(CVD)encompassesbothlargeandsmallvesselvasculardiseases(SVD)thatdis-
turb cerebral circulation and function. CVD is often co-morbid with AD pathology and may exacerbate
risk,manifestation,andprogressionofdementia(Santosetal.,2017;Tublinetal.,2019). CVDmayevenplay
acausalrole(e.g.,vascularhypothesisofAD)inAD-pathogenesisandprogression(Nelsonetal.,2016). A
primary cause of CVD is atherosclerosis, a build of cholesterol-induced plaque in vessel walls that results
in disturbed bloodow and inammation ( Soehnlein and Libby, 2021) Atherosclerosis is a vascular risk
factorassociatedwithAD.AtherosclerosisiscorrelatedwithADclinicaldiagnosisandpost-mortemneu-
ropathologicalndings, with AD patients having more severe atherosclerosis than age-matched controls
(Roher et al., 2011). Atherosclerosis modies shear stress dynamics, the frictional force against the vessel
wall,consequentlyreducingCBF,changingendothelialgeneandproteinexpression,andmodifyingdown-
stream oxidative and inammatory pathways ( Cunningham and Gotlieb, 2005). Endothelial cell induced
shear stress ne tunes local CBF ow through the release of various vasoactive metabolites. Therefore,
greaterintracranialatherosclerosismayimpairneurovascularcouplingdynamicsattheleveloflocalCBF
owandprovideaphysiologicallinktoAD.GiventheassociationbetweenatherosclerosisandAD,aswell
11
astheeectsthatatherosclerosishasonlocalCBFandneurovascular-neuroenergeticcoupling,itremains
an important point of investigation when considering mechanisms of AD-related cognitive decline.
CVDisalsomarkedbyincreasedinammatoryprocesses,asatheroscleroticplaquebuild-uppromotes
inammation. Bioenergetic function may become disturbed in AD through vascular inammatory pro-
cesses(Wilkinsetal.,2014). VascularinammationhasbeenincreasinglyrecognizedasariskfactorinAD
and contributes to structural damage and functional decline in the brain by facilitating the activation of
vascularendothelialcellsofbloodvesselstointeractwithimmunecells,inammatorycytokines,andblood
products(AshbyandMack,2021;Govindpanietal.,2019). Vascularinammationinducesheightenedper-
meabilityofcerebralbloodvessels,migrationofimmunecellsintothebrainparenchymaandcontributes
tovascularremodelingthatperturbsCBFinthecapillaries(AshbyandMack,2021). Vascularinammation
can also further promote other pathological vascular processes, such as angiogenesis in which leaky and
abnormally structured blood vessels are formed and impede the cerebral circulation (Govindpani et al.,
2019). The ADNI dataset I used in my dissertation work had limited measures of CVD and no measures
of atherosclerosis. Therefore, I investigated other markers of vascular dysfunction; white matter hyper-
intensity volume, and CSF vascular inammatory markers to evaluate the intersection between vascular
dysfunction and AD neuropathology on AD brain and cognitive outcomes.
1.2.1 WhiteMatterHyperintensities(WMH)
White matter hyperintensities (WMH) are a consequence of CVD small vessel disease (SVD) and age-
relatedvasculardysfunction. WMHsincreasetheriskofstroke,cognitiveimpairment,anddeath(Debette
and Markus, 2010). Unfortunately, the prevalence of WMHs in old age is almost ubiquitous; 11-31% of
64 year olds have WMHs, and by the time an individual is 82 years old, 64-94% have WMHs present
(Ylikoski et al., 1995; Garde et al., 2000). The primary risk factors for WMHs include hypertension, with
smoking, diabetes, and cardiovascular disease also contributing to WMH accumulation (Longstreth et al.,
12
1996; Dufouil et al., 2001; Jeerakathil et al., 2004). Given that WMHs are ubiquitous in old age and WMH
riskfactorsalsooverlapwithADandrelateddementiariskfactors,itremainsdiculttountangletherole
WMHs play in AD disease progression and manifestation.
WMHs appear as a hyperintense, bright white signal on T2-weighted imaging, such as auid attenu-
atedinversionrecovery(FLAIR)sequence(Habesetal.,2016). FLAIRMRIsuppressesthesignalintensityof
CSFintheventriclestobettervisualizepathology,suchasWMHs. ThehistopathologicalfeaturesofWMHs
areheterogeneous. Whitematterlesionsaretypicallycharacterizedbydiusedemyelination,palemyelin,
axonalloss,apoptosisofoligodendrocytes,gliosis,andvacuolization(Wardlawetal.,2015). However,the
precisemechanismsthatinducethesechangeshavenotbeenfullyelucidated. Multi-modelneuroimaging
studies indicate that physiological changes precede and accompany WMH growth. For example, changes
in MRI diusion tensor imaging (DTI) derived fractional anisotropy (FA) (a measure of the degree of dif-
fusion restriction within an image voxel) is apparent in white matter regions that will eventually develop
a hyperintense signal on T2-weighted imaging scans (Jiaerken et al., 2019). Reduced FDG-PET signal and
greater amyloid deposition have also been noted in brain regions with more lesions in connecting white
matter (Glodzik et al., 2014).
WMHsexpandovertime,withsomeregionsappearingbrighterandmoreapparentthanotherregions.
As a result, various WMH assessment measures capture dierent boundaries, giving rise to further vari-
abilityinourunderstandingofhowWMHsrelatetoAD.WMHscanbequalitativelyassessedusingvisual
rating scales (e.g., Fazekas scale) and quantitatively measured using automated segmentation methods to
extractvolumemeasurements(Tubietal.,2020). WMHsarecommonlymeasuredbyvolumeandlocation
withWMHsrstappearingaroundtheventriclesandpotentiallyextendingtodeeperwhitematter. Vari-
abilityinperiventricularanddeepWMHsetiologicaloriginsandclinicalconsequencesexist. Forexample,
periventricular WMHs are more strongly linked to cognitive decits than deep WMHs ( Grianti et al.,
13
2018). Etiological dierences in periventricular and deep WMH accumulation are also apparent. For ex-
ample, periventricular WMHs may be related to arterial pressure, plasma leakage, BBB permeability, and
decline in total CBF, while deep WMHs may be associated with axonal loss, arteriolosclerosis, and body
mass index (ten Dam et al., 2007a; Haller et al., 2013; Wharton et al., 2015; Grianti et al., 2018).
DespitetheheterogeneityinWMHcauseandcomposition,measuresofWMHsareconsistentlylinked
to cognitive dysfunction. WMHs increase the risk of cognitive decline dose-dependently, with execu-
tive function and processing speed particularly vulnerable to WMH accumulation (Prins and Scheltens,
2015). AlthoughgreaterWMHaccumulationislinkedtoworsecognitivefunction,theinteractionbetween
WMHs and AD pathology on AD outcomes remains ambiguous. For example, some studies demonstrate
that WMHs may interact with AD pathology and clinical progression, potentially exacerbating cognitive
decline (Brickman et al., 2008). However, other studies demonstrate that the association between WMHs
and cognitive decline is only apparent early in AD or not apparent at all in AD (Soldan et al., 2020; Sudre
et al., 2017; Rojas et al., 2018). The conicting results in our understanding of WMHs relationship to AD
mayberelatedtothevarioussegmentationcriteriausedtodeneWMHboundaries. Therefore,Chapter2
of this body of work aims to assess the most clinically relevant WMH segmentation methods in a sample
of participants without dementia (cognitively normal, MCI). I compare ve segmentation methods that
dene WMH boundaries withslightly dierentalgorithms, capturingvariabilityin WMHsegmentations,
with AD diagnosis and composite measures of executive function and memory.
1.2.2 VascularEndothelialGrowthFactor(VEGF)
Vascularendothelialgrowthfactor-A(VEGF)isasignalingproteininvolvedinbloodvesselgrowth,oxygen
andglucosedelivery,vasodilation,andvascularpermeability. VEGFisapowerfulangiogenicregulatorthat
canbeendogenouslyupregulatedasapositiverepairresponsetocombatneurodegenerationbypromoting
glucose and oxygen delivery (Sone et al., 2000; Jais et al., 2016)(Góra-Kupilas and Jośko, 2005; Jais et al.,
14
2016; Zhao et al., 2015). For example, VEGF is upregulated as a compensatory mechanism for some of
the earliest preclinical AD biomarkers and risk factors: cerebral hypoperfusion (Benderro and LaManna,
2011) and hypometabolism (Sone et al., 2000; Jais et al., 2016). VEGF is also important in stimulating
neurogenesis(Caoetal.,2004;Fournieretal.,2012;Fabeletal.,2003),improvingsynapticplasticity(Licht
et al., 2011), maintaining cognitive function (Cao et al., 2004), enhancing glucose transport (Sone et al.,
2000; Zhao et al., 2015), and counteracting neurodegeneration (Góra-Kupilas and Jośko, 2005; Zacchigna
etal.,2008)-allessentialaspectsofADprevention. AlthoughVEGFconcentrationsandtheabilityofVEGF
tostimulateangiogenesisbothdeclinewithage(Rivardetal.,2000)andwithprogressiveAD(Tangetal.,
2013) it can be modulated endogenously and exogenously (Licht et al., 2011; Uysal et al., 2015) making
it a viable therapeutic target in older adults. For example, exercise induces VEGF-mediated hippocampal
neurogenesis and enhanced CBF (Viboolvorakul and Patumraj, 2014).
VEGFupregulationisneuroprotectiveinbothanimalin-vivo(Wangetal.,2011;Sunetal.,2003)andin-
vitrostudies(Yangetal.,2014;Bürgeretal.,2009). However,humanstudiesin-vivohavefounddiscrepant
relationshipsbetweenVEGFlevelsandADdiagnosis(Hohmanetal.,2015;Tarkowskietal.,2002)making
it an important point of investigation. This variability may be a result of intervening mechanisms from
various biological stressors that modulate VEGF activity. Specically, inammation ( Sainson et al., 2008),
atherosclerosis (Lehoux and Jones, 2016; Li et al., 2005; Rutkowski and Swartz, 2007), amyloid-beta (Aβ)
plaques(Yangetal.,2004)areAD-relevantbiological-stressorsthatareknowntomodifyVEGFsignaling.
Consequently, pre-clinical AD hypometabolism may be a result of biological-stressors interfering with
VEGFsignalingcapacity,ultimatelyimpairingglucosetransport. Thesebiological-stressorsincreasewith
age(SinghandNewman,2011;BraakandBraak,1997;Mraketal.,1997)andarepresentinvaryingdegrees
acrossindividuals. Therefore,lowerVEGFlevelsmayindicatethatVEGFisnotbeingupregulatedbecause:
1)noVEGF-inducingsignalsarepresentandnocorrespondingrepairresponseisnecessary(healthycon-
dition)or2)bio-stressorsareinterferingwithVEGFsignaling(pathogeniccondition). Thisdichotomymay
15
explain why previous studies show discrepant results regarding the directional relationship between CSF
VEGFandagingoutcomes(Hohmanetal.,2015;Tarkowskietal.,2002). Thesemodifyingfactorsmayalso
explainwhyVEGFconcentrationsandtheabilityofVEGFtostimulateangiogenesisdeclinebothwithage
(Rivardetal.,2000)andadvancedstagesofAD(Tangetal.,2013). Chapters2and3ofmydissertationaim
to address this mechanistic relationship by evaluating whether these bio-stressors (e.g., neuropathologi-
cal burden, vascular risk) modify (i.e. interacts with) CSF VEGF’s association with cognition (executive
function and memory processing) and regional AD brain biomarkers (cortical thickness, and FDG-PET).
While VEGF was the primary focus of this dissertation work, I wanted to further evaluate whether
the ndings were specic to VEGF or broadly driven by vascular inammation and seen across other
CSF vascular biomarkers. Therefore, in Chapter 4, I evaluate whether similar associations are found in
5 CSF biomarkers that are correlated with CSF amyloid levels and that have an active role in vascular
inammatory processes: brinogen, VEGF, von Willebrand Factor (vWF), C-reactive protein (CRP); and
vascular cell adhesion molecule 1 (VCAM1) (Pillai et al., 2019). As will be discussed in more detail in
Chapter 4, we found VEGF and CRP to be related to FDG-PET uptake and have interaction eects with
WMHvolume. Briey,CRPisamajorinammatorymarkerproducedbytheliverthatiswidelyrecognized
as an indicator of systemic and chronic inammation at elevated levels ( Maksimowicz-McKinnon et al.,
2004; Fernandes et al., 2020; Schmidt-Arras and Rose-John, 2016). CRP has been consistently associated
withCVD,inwhichhigherlevelsofcirculatingCRPhavebeencorrelatedwithhigherriskofischemia(van
Haelst et al., 2003; Rouhl et al., 2012) and larger volumes of WMHs (Low et al., 2019). Additionally, both
high and low levels of CRP have been associated with amyloid and tau pathology (Fernandes et al., 2020;
Brosseron et al., 2018). Together, the evaluation of CSF biomarkers, such as VEGF and CRP, in addition
to neuroimaging measures can help provide proles of the AD continuum to improve the accuracy of
diagnostic predictions, disease staging and distinguish eective therapeutic targets.
16
1.3 PurposeandAims
Bythetimeanindividualhasevidenceofcognitivedecits,thebrainhasalreadyundergoneseverestruc-
tural and functional degeneration, making it essential to assess factors (e.g., vascular dysfunction) that
may contribute to brain dysfunction early in the AD cascade. AD neuroimaging brain biomarkersuctu-
ateearlyandfollowaspecicspatialandtemporalsequenceinADdiseasecourse. Identifyingthespatial
relationship of vascular factors to early brain changes may also illuminate possible vascular mechanisms
thatcontributetoheterogeneityinADneuroimagingphenotypepatterns. Further,neuropathologicalbur-
denandCVDcanmodifytherelationshipbetweenCSFbiomarkersandADneuroimagingmarkers. Given
the complexity of the interplay between vascular contributions and AD neuropathological markers, my
dissertation aims to dissect how vascular factors (e.g., VEGF, WMH volume) contribute to variability in
AD-brain and cognition biomarkers. This dissertation work helps ll a critical gap of knowledge – by
assessingtheregionalrelationshipbetweenvascularfactors(e.g.,VEGF,WMH)andAD-brainbiomarkers
and cognition. To address this, I have three overarching aims addressed in this body of work:
Aim1(Chapter2):
• Evaluate whether the spatial specicity of WMH segmentation modies the association between
WMHvolumeandcognition(executivefunctionandepisodicmemory)toidentifythemostclinically
meaningful relationships.
• EvaluatewhethertherelationshipbetweenWMHvolumeandcognitiondiersbetweenindividuals
with and without abnormal Aβ-levels
• Hypothesis: The presence of AD-specic neuropathology and varying criteria for creating a WMH
segmentation will mask WMH eects on cognition.
17
Aim2(Chapter3):
• Evaluate whether CSF VEGF has a regional relationship to AD brain biomarkers (cortical thickness
and FDG-PET) and cognition (executive and memory function).
• Evaluate whether AD neuropathological markers (CSF Aβ, total-tau, phosphorylated-tau) modify
the relationship between VEGF and regional AD brain biomarkers.
• Evaluate whether regional FDG-PET signal mediates a relationship between VEGF and cognitive
function.
• Hypothesis: In those with higher levels of AD neuropathology, higher endogenous VEGF levels (which
may compensate for amyloid-driven VEGF inhibition) would be associated with greater FDG-PET and
cortical thickness in AD-cortical thinning signature regions. Also, FDG-PET signal will mediate the
previouslydetectedassociationbetweenVEGFandcognitioninthosealongtheADcontinuum(Aβpar-
ticipants).
Aim3(Chapter4):
• Evaluate whether CSF vascular inammatory markers (which are correlated with CSF amyloid) are
associated with whole-brain FDG-PET uptake.
• In CSF vascular inammatory biomarkers that demonstrate an association with whole brain FDG-
PET signal, evaluate whether they are associated with regional FDG-PET signal in vascular territo-
ries and cortex wide gryal boundaries.
• Evaluate whether WMH volume dampens associations between CSF vascular biomarkers and re-
gional FDG-PET uptake.
• Evaluate whether interactions between CSF vascular inammatory biomarkers and FDG-PET vary
by APOE4 genotype, amyloid load, and sex.
18
• Hypothesis: WMH volume will interact with CSF vascular inammatory biomarkers association with
regional FDG-PET, particularly in regions that are vulnerable to atherosclerosis. We hypothesize the
results will be strongest in those most at risk for AD (Aβ+, APOE4 carriers).
19
Chapter2
Whitematterhyperintensitiesandtheirrelationshiptocognition:
Eectsofsegmentationalgorithm
This section is adapted from:
Tubi, M. A., Feingold, F. W., Kothapalli, D., Hare, E. T., King, K. S., Thompson, P. M., and Braskie,
M. N. (2020). White matter hyperintensities and their relationship to cognition: Eects of segmentation
algorithm. NeuroImage, 206:116327.
2.1 Abstract
Whitematterhyperintensities(WMHs)arebrainwhitematterlesionsthatarehyperintenseonuidatten-
uated inversion recovery (FLAIR) magnetic resonance imaging (MRI) scans. Larger WMH volumes have
been associated with Alzheimer’s disease (AD) and with cognitive decline. However, the relationship be-
tweenWMHvolumesandcross-sectionalcognitivemeasureshasbeeninconsistent. Wehypothesizethat
this inconsistency may arise from 1) the presence of AD-specic neuropathology that may obscure any
WMH eects on cognition, and 2) varying criteria for creating a WMH segmentation. Manual and auto-
matedprogramsaretypicallyusedtodeterminesegmentationboundaries,butcriteriaforthoseboundaries
candier. ItremainsunclearwhetherWMHvolumesareassociatedwithcognitivedecits,andwhichseg-
mentation criteria inuence the relationships between WMH volumes and clinical outcomes.
20
Inasampleof260non-dementedparticipants(ages55–90,141males,119females)fromtheAlzheimer’s
DiseaseNeuroimagingInitiative(ADNI),wecomparedtheperformanceofveWMHsegmentationmeth-
ods, by relating the WMH volumes derived using each method to both clinical diagnosis and composite
measures of executive function and memory. To separate WMH eects on cognition from eects related
to AD-specic processes, we performed analyses separately in people with and without abnormal cere-
brospinaluid amyloid levels.
WMHvolumeestimatesthatexcludedmorediuse,lower-intensitylesionsweremorestronglycorre-
latedwithclinicaldiagnosisandcognitiveperformance,andonlyinthosewithoutabnormalamyloidlev-
els. ThesendingsmayinformbestpracticesforWMHsegmentation,andsuggestthatADneuropathology
may mask WMH eects on clinical diagnosis and cognition.
2.2 Introduction
White matter hyperintensities (WMHs) in the brain white matter are lesions having a signal intensity
brighter than the surrounding white matter on a magnetic resonance imaging (MRI) uid attenuation
inversionrecovery(FLAIR)sequence(Yoshitaetal.,2006). WMHsareassociatedwithvascularrisk(Prins
andScheltens,2015;Scottetal.,2015)andmayrepresentincreasedbloodbrainbarrierpermeability,plasma
leakage, and degeneration of axons and myelin (Haller et al., 2013). WMH volumes are associated with
older age, Alzheimer’s disease (AD), small vessel disease, and cognitive decline, making them a measure
of clinical interest (Brickman et al., 2009; Prins and Scheltens, 2015).
AD-specicprocessesmayinuencetheobservedeectofWMHsonclinicaldiagnosisandcognition.
In cross-sectional data, amyloid plaque counts do not correlate as strongly with cognition as neurobril-
lary tangle counts (Wilcock and Esiri, 1982). Still the presence of amyloid positivity in cognitively intact
olderadultsisconsideredtobeasignofpreclinicalAD(Haneetal.,2017),andisassociatedwithfasterlon-
gitudinaldeclineincognitivefunctioncomparedtothatseeninamyloid-negativeolderadults(Mortamais
21
etal.,2017). WMHsandamyloiddepositioninADmayinuenceoneanother( Grimmeretal.,2012;Scott
etal.,2015,2016),andbothmaycontributetocognitiveimpairment(Provenzanoetal.,2013;Gordonetal.,
2015). WeusedamyloidpositivityasasurrogateforAD-specicprocesses,whichmayinuencecognition
independently of, and together with, WMHs. We studied the eect of WMHs on cognition by evaluating
therelationshipseparatelyinthosewhowereamyloidpositive(Aβ+)ornegative(Aβ-)(Shawetal.,2009a).
We hypothesized that the relationship between WMH volume and cognition would be stronger in those
who were Aβ- (and thus had less cognitive variability added by AD-related processes) compared to those
who were Aβ+.
Larger WMH volumes have been associated with both decreased global cognitive function (Au et al.,
2006; Frisoni et al., 2007; Kloppenborg et al., 2014) and domain specic-cognitive impairment, including
executive function (Gunning-Dixon and Raz, 2000; Smith et al., 2011; Lampe et al., 2017; Aljondi et al.,
2020) and memory (Cees De Groot et al., 2000; Gunning-Dixon and Raz, 2000; Smith et al., 2011; Lampe
et al., 2017). However, results vary among studies that have evaluated the WMHs to cognition relation-
ship. Thisvariabilitymayarisefromdierencesintheclassicationoflesionboundariesinsegmentation
methods(Smartetal.,2011;Caligiurietal.,2015;Wangetal.,2015b;Dadaretal.,2017),whichmaycapture
physiologically dierent components (Haller et al., 2013).
Manual segmentation - the gold standard when comparing automated methods - is time consuming,
requiring multiple raters, and training to establish intra-rater and inter-rater reliability. Further, expert
reviewersindierent laboratoriesmayusedierentvisual ratingscalesormaydisagree aboutwhatcon-
stitutesaclinically-relevantWMHboundaryorlocation. Therefore,acceptableintra-studyreliabilitymay
not translate into high reliability between methods or studies (Grimaud et al., 1996; Mantyla et al., 1997;
Kapelleretal.,2003;Prinsetal.,2004;Yoshitaetal.,2005). Often,limitedinformationisprovidedinpubli-
cationstodescribethecriteriausedfordeningmanualsegmentations-suchaswhethertoincludemin-
imally hyperintense lesions, or lighter ‘halos’ around larger higher-intensity lesions. This makes ground
22
truthandreplicationacrossstudiesdicult(Firbanketal.,2004;Gibsonetal.,2010;Smartetal.,2011;Iorio
etal.,2013;Griantietal.,2018). ItisunclearwhichWMHmanualsegmentationcriteriaresultinthemost
clinically-relevant lesion assessments (Van Straaten et al., 2006).
We calculated WMH volumes using the default options forve automated WMH segmentation algo-
rithms. Ourgoalwasnottoevaluatethesoftwarepackagesthemselves,allofwhichcanbeoptimized,but
rather to create a range of typical segmentations that allowed us to identify which features strengthened
thesensitivitytodetectingarelationshipbetweenWMHvolumesandcognitivemeasuresinAβ+andAβ-
non-demented older adults.
2.3 Materialsandmethods
Data used in the preparation of this article were obtained from the Alzheimer’s Disease Neuroimaging
Initiative (ADNI) database (https://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),
otherbiologicalmarkers,andclinicalandneuropsychologicalassessmentcanbecombinedtomeasurethe
progression of mild cognitive impairment (MCI) and early Alzheimer’s disease (AD).
2.3.1 Participants
Weevaluated260non-dementedparticipants,aged55–90yearsold,fromADNI2whohadallofthefollow-
ing variables available: 1) 3T MRI T1-weighted anduid attenuated inversion recovery (FLAIR) images,
2) cerebrospinal uid (CSF) A β
42
levels (described further in the ADNI methods page http://adni.loni.
usc.edu/methods/),and3)neuropsychologicalassessment. BoththeCSFcollectionandneuropsychological
testingoccurredwithin18.5months(averageof3.2monthsand27days,respectively)oftheMRIscan. Four
supplemental participants were used for training of our in-house WMH intensity ratio method, and four
23
Table 2.1: Demographicfeaturesofthesampleanalyzed
Aβ-Aβ+
Controls MCI Total Controls MCI Total
N 54 89 143 44 73 117
Age
(years)
73.07 ± 5.53* 70.00 ± 7.29* 71.16 ±6.83** 74.68 ±7.17 73.26 ±7.64 73.79 ±7.47**
Sex (M/F) 32/22 46/43 78/65 23/21 40/33 63/54
Education
(years)
16.81 ± 2.60 16.01 ± 2.47 16.31 ± 2.54 16.70 ±2.47 16.40 ±2.61 16.51 ± 2.56
ICV
(mm
3
)
1.46 x 106 ±
1.37 x 105
1.46 x 106 ±
1.32 x 105
1.46 x 106 ±
1.33 x 105
1.47x106 ±
1.51 x 105
1.47x106 ±
1.47 x 105
1.47 x 106 ±
1.48 x 105
Shownasmean ±standarddeviation. Weevaluatedgroupleveldierences,betweenamyloidgroups(A β-
vs. Aβ+) and within amyloid groups (control vs. MCI), across age, education, and intracranial volume
(ICV),usingWelch’stwo-tailedt-tests. Weevaluatedsexusinga 2
test. *Signicantlydierentbetween
controlsandMCIwithinAβgroup,p< 0.05. **SignicantlydierentbetweenA β+andAβ-participants,
p< 0.05.
additional participants were removed after failing FreeSurfer segmentation quality control procedures.
Demographic information is tabulated in Table 2.1. Data analyzed in this study - including MRI scans,
CSF amyloid-β
1-42
(Aβ
42
) levels, and neuropsychological test scores - were downloaded from the publicly
available ADNI Image Data Archive (IDA; https://ida.loni.usc.edu). WMH volumes assessed using one
of theve algorithmsweevaluated-theintensityhistogramsalgorithm–werealsodownloadeddirectly
from the ADNI IDA.
2.3.2 Neuropsychologicaltestinganddiagnosticcriteria
ParticipantsunderwentADNIbaselineneuropsychologicaltesting-includingtestsoflong-termandwork-
ing memory, language, and executive function - within 3 months of their brain scan. Clinical diagnoses
weredeterminedbyADNIasfollows: probableADisassessedaccordingtoNINDS/ADRDAcriteria(McK-
hann et al., 1984). However, to minimize the contributions to cognition of neurodegeneration that is spe-
cictoAD,ourstudyincludedonlyparticipantswithMCI(n=162)andthosewhowerecognitivelyintact
(n=98). Participants diagnosed with MCI did not meet the diagnostic criteria for dementia, but did report
24
a memory complaint. MCI participants had objective memory loss as measured by education-adjusted
scores on the Wechsler Memory Scale-Revised - Logical Memory II (WMS-Logical Memory II; Score 8,
4, or 2 for having completed 16, 8–15, or 0–7 years of education, respectively). They also had a Clinical
Dementia Rating (CDR) scores of 0.5 (with a mandatory requirement that CDR memory box score was
0.5 or higher), an absence of signicant impairments in other cognitive domains, and preserved daily life
activities. CognitivelyintactcontrolsdidnotmeetthediagnosticcriteriaforprobableADorMCIandhad
no memory complaints. They had a Mini-Mental State Exam (MMSE) score between 24 and 30, a CDR of
0, and scored higher than the education-adjusted MCI thresholds listed above on the Wechsler Memory
Scale-Revised - Logical Memory II scores. Participants were excluded if they had a serious neurological
condition, neuropsychiatric condition (e.g., major depression, bipolar disorder, schizophrenia), or history
of brain injury. We used previously-validated ADNI composite scores for executive function (Gibbons
et al., 2012) and memory (Crane et al., 2012). The normalized composite measures of executive function
and memory were derived from an iterative process that applied item response theory and conrmatory
factoryanalysistopreviouslyacquiredADNIneuropsychologicalbattery(Craneetal.,2012;Gibbonsetal.,
2012). The executive function composite score was derived fromve clock drawing items (circle, symbol,
numbers,hands,andtime),TrailMakingTestpartsAandB,andCategoryFluency(animals). Thememory
compositescorewasderivedfromReyAuditoryVerbalLearningTest(RAVLT),ADAssessmentSchedule
- Cognition (ADAS-Cog), MMSE, and WMS-Logical Memory II.
2.3.3 MRIscanning
Participants underwent whole-brain MRI scanning on 3-Tesla scanners across 51 sites in North Amer-
ica. EachparticipantwasscannedusingananatomicalT1-weightedsequence(1.2mmthicksagittalslices;
0.9375⇥ 0.9375mm
2
in-planeresolution,256⇥ 256matrix)andaT2-weighteduidattenuatedinversion
25
recovery(FLAIR)sequence(5mmthickaxialslices;0.86⇥ 0.86mm
2
in-planeresolution). AllMRIacquisi-
tionsitespassedrigorousscannervalidationtests,andthescanprotocolswereoptimizedacrosssitesand
manufacturers(GE,Philips,Siemens). AGEscannerwasusedtoacquireMRIdataon61participantsacross
14 sites, a Philips scanner was used to acquire MRI data on 46 participants across 10 sites, and a Siemens
scanner was used to acquire MRI data on 153 participants across 27 sites. Detailed procedures on scan
acquisition and optimization are provided elsewhere (https://adni.loni.usc.edu/). All T1-weighted and
FLAIR images were visually checked for quality. We did not perform bias correction on the FLAIR scans,
because 1) our visual quality control assessment did notnd extensive FLAIReld inhomogeneities, and
2)arecentanalysis(ValdésHernándezetal.,2016)ofbiascorrectionperformanceonFLAIRwhitematter
hyper-intensity progression found that applying a biaseld correction was not recommended for FLAIR
images. Although there are advantages of correcting the magneticeld inhomogeneities seen in FLAIR,
thatstudyfoundthatbiaseldcorrectioninthismodalitymayresultindistortionofrealhyperintensities
with a specic expense of subtle intensity dierences. No T1 or FLAIR image had artifacts that were se-
vere enough to interfere with structural or WMH segmentations. A board-certied neurologist was part
of process of reviewing the FLAIR images and WMH algorithm development.
2.3.4 CSFcollectionandanalysis
Participants underwent at least one lumbar puncture to obtain CSF for assays of several biomarkers. The
samplecollectionandanalysisprocessesaredescribedinShawetal. (2009). Aβ+participantsweredened
as those who had CSF Aβ
42
levels 192pg/ml, consistent with prior guidelines (Shaw et al., 2009a).
2.3.5 In-houseWMHalgorithm
Wedevelopedasemi-automatedmethodtosegmentwhitematterhyperintensitiesusingbothT1-weighted
and FLAIR images (Figure 2.1).
26
2.3.5.1 Creatingwhitemattermasks
White matter masks were used to exclude hyperintensities other than WMHs from our segmentations.
To create these masks, we performed biaseld correction on the T1-weighted scans using the Advanced
Normalization Tools (ANTs) N4 correction (Tustison et al., 2010). We then submitted these bias-corrected
imagestoFreeSurfer(version5.3)toobtaintissue-segmentationmasksandintracranialvolumeestimation
(Fischletal.,2002). FreeSurferestimatesintracranialvolume(ICV)usingtheknownrelationshipbetween
the ICV and the linear transform of an individual brain to MNI305 template space (Buckner et al., 2004).
BecausewhitemattermasksproducedbyFreeSurfermayomitWMHs,weconstructedwhitemattermasks
by subtracting the gray matter and CSF masks from the full brain mask. Rarely, WMH were extensive
enoughthattheywerecontiguouswithgraymatterontheT1-weightedimage. Whenthathappened,the
intensityvaluesweresimilarenoughthatWMHswereincludederroneouslyinthegraymattermask. We
therefore visually inspected and manually edited all WM masks to ensure that the gray matter masks did
not include WMHs. Each participant’s resulting white matter mask was linearly transformed (6 degrees
of freedom) to the participant’s own FLAIR image using FMRIB’s Linear Image Registration Tool (FLIRT)
in FSL (Jenkinson and Smith, 2001; Jenkinson et al., 2002). This white matter mask in FLAIR space was
thennon-linearlytransformedtotheFLAIRimageusingtheANTssymmetricimagenormalization(SyN)
method (Avants et al., 2008). We examined and edited the white matter masks as needed in FLAIR space
to ensure all white matter (including WMHs) was included.
Next, for each participant, we constructed a mask of the peripheral white matter alone (which is less
likely to contain WMHs) to calculate the mean intensity for the WM that does not contain lesions. To do
this,rst,weerodedabinarywholebrainMontrealNeurologicalInstitute(MNI152)1mmtemplatebrain
mask by 63% (an arbitrary value chosen to provide a mask that excluded peripheral white matter). We
non-linearly transformed this eroded template brain into each individual’s T1-weighted image using the
ANTs SyN method. The eroded brain masks were then non-linearly transformed into each participant’s
27
Figure2.1: FlowdiagramillustratingtheworkowofourmethodtosegmentWMH.Theintensity
ratio is dened as
Minimum Intensity of WMH
Mean WM intensity without WMH
FLAIR image space using ANTs SyN and were subtracted from the participant’s complete white matter
mask in FLAIR space to create a mask that contained only the brain periphery.
2.3.5.2 SegmentingWMHs
Our in-house WMH segmentation protocol is illustrated in Figure 2.1 and detailed here. First, we created
a reference standard segmentation that was visually similar to a manual segmentation in a sub-sample
of four training participants who had minimal WMHs on the FLAIR image. We chose participants with
minimal WMHs, because for these participants, WMHs were clearly dened and unambiguous, and they
contributedminimallytotheoverallmeanwhitematterintensityforthatparticipant. Inourfourtraining
participants, we automatically identied WMHs by applying a participant-specic intensity threshold at
99thpercentileofthesignalintensityinthetotalwhitematterforeachparticipant,usingthefslmathsfunc-
tioninFSL.Wearrivedatthis99thpercentilethresholdbyvisuallyassessingwhichthresholdadequately
segmentedtheseclearlydelineatedlesionsinourtestparticipants. Ifwehadincludedparticipantshaving
extensive WMH in this training set, their mean white matter intensity would be low, because the WMHs
themselveswouldreducethemeansignalintensityintheWMmask. Therefore,inparticipantswithexten-
siveWMHs,a99thpercentileintensitythresholdwouldnotadequatelyidentifyWMHs. OncetheWMHs
were identied in these four participants, we used their data to calculate a study-specic intensity ratio
that could be used to identify high intensity WMHs, even in participants who also have more extensive
28
Table 2.2: Voxel, volume, and intensity information from the participants used to calculate the
intensityratio
Training Set
Participant
WMH Voxels WMH
Volume
WMH minimum
intensity
Mean intensity
ofWMwithout
WMH
Intensity Ratio
1 1425 5262.30 536.65 395.29 1.36
2 1855 6850.22 551.40 404.26 1.36
3 2125 7847.29 579.15 405.49 1.43
4 1966 7260.14 553.34 366.39 1.51
Average 1842.75 6804.99 555.14 392.86 1.42
Mean volume is inmm
3
.
and diuse lesions. To calculate a study-specic intensity ratio, across the four training participants, we
divided the mean minimum intensity of the WMHs by the mean intensity of the normal-appearing white
matter (excluding the WMHs). The voxel, volume, and intensity information derived from the four train-
ingparticipantsistabulatedinTable2.2. Thisresultedinastudy-specicWMHintensityratioindicating
how much greater the minimum intensity of WMHs was compared with the mean intensity of normal-
appearingWM.Wethenobtainedaparticipant-specicWMHmap,bycalculatingthemeanintensityvalue
ofeachparticipant’sFLAIRimagewithintheperipheralwhitemattermask(see2.3.5.1Creatingwhitemat-
ter masks) and multiplied it by our study-specic WMH intensity ratio to obtain a threshold, which we
then applied to that participant’s original FLAIR image.
2.3.5.3 RegionalWMHsegmentation
We investigated regional dierences in WMH accumulation across three lobes: frontal, temporal, and
parietaloutlinedbasedontheMNIlobemapatlasfromFSL5.0.7(maxprob-thr0-1mm). Totheextentthat
thestandardlobemapdidnotcovertheentirewhitematter,wemanuallyextendedthelobargraymatter
boundaries into the white matter, and visually conrmed that the segmentations were accurate. Figure
2.2depictsbeforeandafterwemanuallyextendedthelobargraymatterboundariesintothewhitematter.
29
Figure 2.2: Image on the left depicts the coronal view of the MNI lobe map atlas from FSL 5.0.7
(maxprob-thr0-1mm). Theimageontherightdepictsthelobemapafterwemanuallyextended
theboundariesofthelobesintothewhitematter.
Thelobarmaskswereregisteredtoeachparticipant’sFLAIRspace. ThisallowedustocalculatetheWMH
volume for the frontal, temporal, and parietal lobes.
We also performed an analysis to separate periventricular and deep WMHs. We did this by dilating
theventriclesegmentationmaskintheparticipant’sFLAIRspacebyaspherekernelof5.16mm(6voxels
x0.86mmresolution). Fig. S1inthesupplementalmaterialillustratestestingperformedontheseparation
of periventricular and deep WMH boundaries by kernel size. We multiplied the dilated ventricle mask by
the participant-specic WMH map to construct the periventricular WMH map. The deep WMH volume
was calculated by subtracting the participant-specic periventrucular WMH volume from that partici-
pant’sfullWMHvolume. Weperformedqualitycontroloneachofthesegmentationmaskstoensurethat
the individual mask’s boundaries were accurate and the ventricular segmentation did not have articial
enlargement.
30
2.3.5.4 Intensitythresholding
WenextinvestigatedhowvaryingtheinclusivenessoftheWMHmasks(toincludeorexcludemorediuse
signalsurroundinghyperintenselesions)aectedtherelationshipofWMHvolumetocognition. Todothis
we created masks based on dierent percentages of the constructed intensity ratio. We calculated WMH
volumesderivedfromthresholdingat85%,90%,95%,and105%oftheintensityratio. Todeterminethenew
thresholdforeachpercentagewemultipliedthepercentbytheunadjustedintensityratioandappliedthe
adjustedvaluetothemeansignalintensityintheperipheralmask. Lowerthresholdpercentagesprovided
a more ‘lenient’ WMH map - that included more diuse lesions - while higher values included only the
highest intensity voxels in the white matter, often associated with more discrete lesions.
2.3.6 Existingwhitematterhyperintensitysegmentationalgorithms
We also evaluated how WMH volumes related to cognition using four WMH algorithms other than our
own: 1)anintensityhistogram-basedalgorithm(DeCarlietal.,1995)twoalgorithmsthatarepartofSPM’s
lesion segmentation tool (LST): 2) the lesion growth algorithm (LGA) (Schmidt et al., 2012) and 3) the le-
sionpredictionalgorithm(LPA)(Schmidt,2017)(http://www.applied-statistics.de/lst.html);and4)FSL’s
brainintensityabnormalityclassicationalgorithm(BIANCA)( Griantietal.,2016). Weusedthedefault
settings of each algorithm.
1) The intensity histograms algorithm is the standard method used in ADNI to calculate the WMH
volume. This algorithm uses a Bayesian probabilistic method to generate likelihood estimate values for
WMH at each voxel in the white matter. These likelihoods are thresholded at three standard deviations
above the mean to construct the binary WMH mask. 2) LGA was implemented in the LST toolbox, ver-
sion2.0.15(http://www.statistical-modelling.de/lst.html)forSPM.T1-weightedandFLAIRimageswere
used as inputs. The algorithm selects an initial lesion map and subsequently grows along voxels that are
hyperintense relative to surrounding tissue. 3) LPA was implemented in the LST toolbox, version 2.0.15,
31
forSPM.WeusedonlyFLAIRimagesasinput. Thealgorithmisabinaryclassierusingalogisticregres-
sionmodeltrainedondatafrom53participantswithsevereMultipleSclerosis(MS).Themodelcovariates
include a similar belief map used in the LGA algorithm above and a spatial covariate that accounts for
voxel-specic changes in lesion probability. Thetted model parameters are implemented to segment le-
sionsofnovelimagesbyestimatingthelesionprobabilityacrosseachvoxel,outputtingalesionprobability
map. 4)BIANCAwasimplementedusingFSL.WeusedaT1-weightedimage,FLAIRimage,andthesame
training set as we used for our in-house algorithm. BIANCA classies each voxel based on intensity and
spatialfeaturestooutputtheprobabilityofthatvoxelbeinginaWMH.WeusedBIANCA’sdefaultsettings
andimplementedthedefaultprobabilitymapthresholdof0.9(probabilityofavoxelbeingaWMH),which
historically has optimized the voxel WMH classication false positives and false negative detection rate
(Grianti et al., 2016). We ran BIANCA both with and without including as inputs the same individual
WM masks created using our in-house algorithm. Using a WM mask to exclude non-white matter has
been shown to reduce false positives (Grianti et al., 2016).
2.3.7 Statistics
2.3.7.1 Amyloidgroupdierences
WestratiedthecohortbyCSFamyloidlevel(A β-,Aβ+)andevaluateddemographicmeasuresbothwithin-
and between-amyloid group. In our within-amyloid group analyses, we assessed dierences between di-
agnostic groups (cognitively intact controls or MCI). We used Welch’s t and a 2
test to evaluate group
dierences in sex.
We covaried for age, sex, years of education, and ICV in all subsequent analyses. Adding scanner
manufacturerasacovariatedidnotmodifytherelationshipbetweenWMHvolumeandclinicaldiagnosis
and did not signicantly contribute to our analyses, so we did not include manufacturer in the statistical
32
models reported throughout the paper (Supplementary Table S2.1). For all statistical models with a bi-
nary dependent variable (such as diagnosis), we performed logistic regression. For all statistical models
with a continuous dependentvariable (such as cognitive composite scores), weperformed multiple linear
regression. We used the composite cognitive measures available on the ADNI website. However, when
we further evaluated the contribution to our eects of individual neuropsychological subtests within the
composite measures, we used Z-score transformed values in our analyses.
All statistical analyses were performed in R version 3.5.1 (University of Auckland, Auckland, New
Zealand) (R Core Team, 2013).
2.3.7.2 In-houseWMHanalysis
WeusedlogisticregressiontotestourhypothesisthattotalWMHvolumewouldbemoreassociatedwith
clinicaldiagnosisinAβ-participants. ToevaluatewhetherourresultswerespecictoA β-participants,we
also used logistic regression to relate WMH volume to diagnosis in Aβ+ participants. WMH volume was
signicantly related to diagnosis in A β- participants only. Therefore, all subsequent analyses presented
here were performed only in Aβ- participants. Additional analyses performed in Aβ+ participants can be
found in the supplementary material (Table S2.4, Table S2.5, Table S2.6, Table S2.7).
To further investigate our signicant results in A β- participants we examined whether regional dif-
ferences in WMH accumulation (in the frontal, temporal, and parietal lobes as well as periventricular
(PVWMH)anddeepWMH(DWMH;regionsacrosslobes)wereassociatedwithclinicaldiagnosisandex-
ecutivefunctionandmemory. Wecorrectedformultiplecomparisonsusingthefalsediscoveryrate(FDR)
approach, and report FDR-adjusted p-values (Yekutieli and Benjamini, 1999).
We assessed whether changing the threshold of our in-house WMH algorithm (i.e., including or ex-
cluding more diuse, lower-intensity voxels to WMHs) modied the relationship between WMH volume
and clinical diagnosis. To do this, we used logistic regression to test the association in Aβ- participants
33
between clinical diagnosisand WMHvolume calculatedusingdierent intensityratios (at85%, 90%,95%,
and105%oftheoriginalintensityratio). Weperformedadditionalanalysesinsegmentationmethodsthat
had an available threshold option (LST, LGA, and BIANCA). Using both the LST LGA method and the
BIANCAmethod,weevaluatedtherelationshipbetweenWMHvolume,derivedfromvaryingthresholds,
and clinical diagnosis in Aβ- participants (Supplementary Tables S2.8 and Table S2.9; Supplementary Fig-
ureS2.2andFigureS2.3). Becauseperipheral(deep)WMHmaybelessintensethanperiventricularWMH,
wefurtherinvestigatedtherelationshipbetweendeepWMHanddiagnosisinAβ-participantswithamore
lenient threshold (85% of the intensity ratio) to deep WMH, using our in-house method. We then related
deep WMH volume using the 85% intensity threshold to diagnosis in Aβ- participants. This analysis was
meant to evaluate possible separate eects of location and intensity of WMHs.
2.3.7.3 WMHsegmentationcomparison
To evaluate dierences across WMH volumes derived from various algorithms, we related WMH volume
(predictor variable) calculated using each of theve segmentation algorithms, to clinical diagnosis (out-
come variable), adjusting for age, sex, years of education, and ICV. To test model dierences between the
WMH segmentation algorithms, we performed a one-way ANOVA with pairwise comparisons, applying
FDR to correct for multiple comparisons.
2.3.7.4 Executivefunction&memoryanalysis
To assess whether WMH volume had any cognitive domain-specic eects, we performed multiple lin-
ear regression to relate WMH volume, derived from the segmentation algorithm that detected strongest
associations, to composite scores of executive function and memory (Crane et al., 2012; Gibbons et al.,
2012). To further investigate any signicantndings between WMH volume and executive function and
memory, we evaluated the relationship between WMH volume and the composite score subtests. In this
34
neuropsychological subtest analysis, we corrected for multiple comparisons by applying false discovery
rate (FDR) and reported FDR adjusted p-values (Yekutieli and Benjamini, 1999).
2.4 Results
2.4.1 Betweenandwithin-amyloidgroupcomparison
In a within-amyloid group analysis, we found that in Aβ- participants, cognitively intact controls were
signicantly older than those with MCI (t=0.850;p=0.005). In Aβ+ participants, no statistically sig-
nicant dierences were found. When diagnosis was not considered, A β+ participants were signicantly
older than the Aβ- participants (t=2.941;p=0.004; Table 2.1). We controlled for age in all further
analyses along with sex, years of education, and estimated ICV.
2.4.2 Regionalrelationshiptodiagnosis
InAβ-participantsonly,highertotalWMHvolumederivedfromourin-housealgorithmwassignicantly
associated with worse clinical diagnosis (z =2.373, WMH volume partial p=0.018). In this model,
higher age (z =3.417, partialp< 0.001) and lower educational level (z = 11.842, partial p=0.065)
werealsosignicantlyassociatedwithpoorerclinicaldiagnosis. Additionally,largerfrontal,parietal,and
periventricularWMHvolumewereassociatedwithworseclinicaldiagnosisinAβ-participants(Table2.3).
ResultsfromthetotalWMHandregionalanalysiscanbefoundinTable2.3. Inafollow-upanalysisinAβ-
participants,werelatedregionalWMHvolumetoexecutivefunctionandmemoryscores. Noneoftheindi-
vidualregionswererelatedtoexecutivefunction(SupplementaryTableS2.2)ormemory(Supplementary
Table S2.3).
We performed a follow-up analysis to further identify regional specicity of periventricular and deep
WMeects,andfoundthatlargerWMHvolumesinthefrontalperiventricularandparietalperiventricular
regions were signicantly associated with worse clinical diagnosis in A β- participants (Table 2.4).
35
Table 2.3: Associationsbetweenin-housederivedWMHvolumebyregionandclinicaldiagnosis
inAβ-participants
Region
Controls MCI
Mean
volume ±
SD
Median (IQR) Mean
volume ±
SD
Median (IQR) z-
score
Partial
p-
value
FDR
ad-
justed
p-
value
Total 2839 ±2684 2084(943-3763) 4379 ±6609 2145(1118-4118) 2.373 0.018* –
Frontal 584 ± 893 262 (119-622) 1389 ±2639 418 (139-1512) 3.057 0.002* 0.010*
Parietal 739 ± 866 456 (152-944) 1489 ±3032 356 (149-1146) 2.303 0.021* 0.035*
Temporal 117 ± 135 61 (18-188) 144 ± 200 65 (26-183) 1.448 0.148 0.148
PV 1940 ±1706 137 (807-2720) 2823 ±3353 1667 (797-3285) 2.727 0.006* 0.015*
Deep 899 ± 1445 352 (92-972) 1555 ±3570 383 (124-945) 1.757 0.079
•
0.099
•
Meanvolumeisinmm
3
. Eachrelationshipwasevaluatedusingalogisticregression,adjustedforage,sex,
years of education, and ICV. Multiple comparison correction was applied to frontal, parietal, temporal,
periventricular, and deep WMH volume analyses, using FDR adjusted values. Clinical diagnoses: MCI =
1; control = 0. PV=Periventricular; SD = Standard Deviation; IQR = Interquartile Range. *p< 0.05.
•
p< 0.10, indicating a trend level association.
No signicant associations were found between total or regional WMH and clinical diagnosis in A β+
participants. AllanalysesinAβ+participantscanbefoundinthesupplementalmaterial(TablesS2.4,S2.5,
S2.6, S2.7).
2.4.3 Intensitythresholdmodication
We next evaluated whether including less intense/more diuse WMH voxels in the WMH volume mea-
sure aected the relationship we saw between WMH volume and clinical diagnosis in A β- participants.
We did this by adjusting the intensity ratio thresholds used to dene WMHs. In A β- participants, higher
total WMH volume derived from both the unadjusted WMH threshold (i.e., 100%) and the 105% inten-
sity threshold (which further excluded lower-intensity voxels) were signicantly associated with poorer
clinical diagnosis (Table 2.5, Figure 2.3). Thresholds of 85%, 90%, and 95% of the original WMH thresh-
oldincludedlowerintensityvoxelscharacteristicofdiuselesions;WMHvolumescalculatedusingthese
more inclusive thresholds were not signicantly associated with clinical diagnosis (Table2.5). For each
36
Table 2.4: Associations between in-house derived WMH volume by subregion and diagnosis in
Aβ-participants
Region
Controls MCI
Mean vol-
ume ± SD
Median(IQR) Mean vol-
ume ± SD
Median (IQR) z-
score
Partial
p-
value
FDR
ad-
justed
p-
value
Frontal PV 470 ± 725 193 (78-538) 954 ± 1488 351 (69-128) 3.252 0.001* 0.007*
Parietal PV 453 ± 423 373 (144-618) 774 ± 1217 298 (129-787) 2.440 0.015* 0.044*
Temporal PV 93 ± 112 48 (10-141) 105 ± 152 44 (15-138) 1.177 0.239 0.239
Frontal Deep 114 ± 252 34 (8-109) 435 ± 1297 54 (17-211) 1.907 0.057
•
0.085
•
Parietal Deep 286 ± 593 40 (2-250) 715 ± 1989 39 (2-242) 1.931 0.053
•
0.085
•
Temporal Deep 25 ± 51 6 (0-27) 39 ± 77 8 (0-44) 1.327 0.184 0.221
Meanvolumeisinmm
3
. Eachrelationshipwasevaluatedusingalogisticregression,adjustedforage,sex,
years of education, and ICV. Multiple comparison correction was applied to frontal, parietal, temporal,
periventricular, and deep WMH volume analyses, using FDR adjusted values. Clinical diagnoses: MCI =
1; control = 0. PV=Periventricular; SD = Standard Deviation; IQR = Interquartile Range. *p< 0.05.
•
p< 0.10, indicating a trend level association.
intensity threshold we tested, the covariate of older age was signicantly associated with poorer clinical
diagnosis and lower educational level attained had a trend level association with poorer clinical diagno-
sis. Modication of thresholds using the LGA and BIANCA methods can be found in the supplementary
material (Table S2.8, Figure S2.2, Table S2.9, Figure S2.3). In a follow-upanalysis we investigated whether
applying a more lenient threshold to deep WMH resulted in a larger relationship between deep WMH
volume and clinical diagnosis. We found that, -when thresholded at 85% of the intensity ratio to allow
theinclusionoflower-intensitylesions,deepWMHvolumestillwasnotsignicantlyrelatedtodiagnosis
(z = 0.307,p=0.759).
37
Figure 2.3: WMH boundary segmentation based on varying intensity thresholds of the study-
specicintensityratio. Thefar-rightimageillustratesthe85%,100%,and105%thresholdmasks
alloverlaidonthebaseFLAIRimageforcomparisonpurposes.
Table 2.5: Associations between in-house derived total WMH volume and clinical diagnosis by
intensitythresholdinAβ-participants
Controls MCI
Intensity Thresh-
old Percentage
(Ratio Number)
Mean volume
± SD
Median
(IQR)
Mean volume
± SD
Median
(IQR)
z-
score
Partial
p-value
85% (1.207) 18206±13717 14906 (7098-
25585)
18524±14935 14959 (8663-
21507)
0.904 0.366
90% (1.278) 8830 ± 7714 6990 (3015-
11760)
10145±10898 6655 (3931-
10841)
1.431 0.152
95% (1.349) 4750 ± 4324 3576 (1598-
6191)
6426 ± 8417 3651 (1961-
6329)
2.109 0.035
100% (1.42) 2839 ± 2684 2084 (943-
3763)
4379 ± 6609 2145 (1118-
4118)
2.373 0.018*
105% (1.491) 1799 ± 1831 1245 (563-
2395)
3066 ± 5208 1338 (645-
2781)
2.388 0.017*
Mean volume is in mm
3
. Each relationship was evaluated using a logistic regression, adjusted for age,
sex, years of education, and ICV. Clinical diagnoses were coded as MCI = 1; control = 0. SD = Standard
Deviation; IQR = Interquartile Range. *p< 0.05
38
2.4.4 ClinicalassociationsdetectedbyotherWMHalgorithms
WithinourAβ-group,weassessedtheassociationbetweendiagnosisandtotalWMHvolume. Wedidthis
usingsixlogisticregressionanalyses,oneforeachWMHsegmentationmethod: 1)ourin-houseintensity-
based algorithm; 2) a previously-published WMH segmentation based on mathematical modeling of MR
pixel intensity histograms (DeCarli et al., 1995): and three methods freely available online - 3) LST - LGA
(Schmidtetal.,2012),4)LST-LPA,(Schmidt,2017),5)FSL-BIANCAusinganoptionalWMmaskasinput,
and 6) FSL – BIANCA without using an optional WM mask as input (Grianti et al., 2016)Figure2.4.
In non-demented Aβ- participants, greater total WMH volume was signicantly associated with MCI
diagnosis, calculated using four algorithms: 1) our in-house algorithm (p=0.018), 2) an algorithm based
on MR pixel intensity histograms (p=0.001)(DeCarli et al., 1995), 3) LGA (p< 0.001)(Schmidt et al.,
2012), and 4) BIANCA using the optional WM mask as input (p=0.032)(Grianti et al., 2016). Total
WMHvolumewasnotsignicantlyassociatedwithdiagnosisusingLPA(p=0.086)orBIANCAwithout
using the optional WM mask as input (p=0.088), using the default options (Table 2.6).
39
Table2.6: AssociationsbetweenWMHvolumebysegmentationmethodandclinicaldiagnosisin
Aβ-participants
Controls MCI
WMH Seg-
mentation
Method
Mean volume
± SD
Median (IQR) Mean volume
± SD
Median (IQR) z-score Partial
p-value
In-house 2839 ± 2684 2084 (943-
3763)
4379 ± 6609 2145 (1118-
4118)
2.373 0.018*
Intensity
histograms
3133 ± 2788 2340 (1359-
4076)
5955 ± 8686 2713 (1243-
6664)
3.231 0.001*
LGA 2658 ± 3478 1525 (575-
3076)
5457 ± 8473 1864 (315-
6903)
3.533 <0.001*
LPA 20163±19443 12338 (6713-
28781)
20617±21128 11772 (5057-
30651)
1.715 0.086•
BIANCA
(masked)
4272 ± 4063 2404 (1494-
6386)
5442 ± 7020 2705 (1257-
6564)
2.145 0.032*
BIANCA
(unmasked)
14165 ± 5568 13610 (10209-
17196)
16147 ± 8050 15285 (9845-
20916)
1.705 0.088
•
Meanvolumeisinmm
3
. Eachrelationshipwasevaluatedusingalogisticregression,adjustedforage,sex,
years of education, and ICV. For BIANCA, “masked” indicates that a WM mask was used as input for the
analysis. Clinical diagnoses: MCI = 1; control = 0. SD = Standard Deviation; IQR = Interquartile Range.
*p< 0.05,
•
p< 0.10, indicating a trend level association.
Weperformedaone-wayANOVAwithpairwisecomparisons(Table2.7)todeterminewhetherWMH
volumewassignicantlydierentacrossalgorithms. WefoundthatWMHvolumescalculatedusingLPA
were signicantly dierent from WMH volumes using all other methods. Our in-house method, LGA,
40
BIANCA (masked), and the intensity histogram method did not provide WMH volumes that were signi-
cantly dierent from one another.
Figure 2.4: Range of WMH severity and variation in white matter segmentation methods. The
severitywasevaluatedasWMHvolumecorrectedforICV.WedenedmildWMHvolumeinapar-
ticipant, when the individual’s total WMH volume was less than the mean total WMH volume across
participants. Moderate WMH volume was dened as the individual’s total WMH volume being between
themeanandtwostandarddeviationsabovethemeanacrossparticipants,andsevereWMHvolumewhen
theindividual’stotalWMHvolumewasgreaterthantwostandarddeviationsabovethemeanacrosspar-
ticipants. ForBIANCA,“masked”indicatesthatthesameWMmaskgeneratedforourin-housealgorithm
was used as input for the analysis
41
Table2.7:One-wayANOVAwithpairwisecomparisons. p-valuesdisplayedarecorrectedformul-
tiplecomparisonsusingthefalsediscoveryrate(FDR)
In-House Intensity Histogram LGA LPA
Intensity Histogram p=0.66–– –
LGA p=0.78 p=0.78 ––
LPA p< 0.001* p< 0.001* p< 0.001*–
BIANCA (masked) p=0.66 p=0.93 p=0.78 p< 0.001*
For BIANCA, “masked” indicates that a WM mask was used as input for the analysis. *p <0.05
2.4.5 Executivefunctionandmemory
Forsimplicityofpresentation,insubsequentanalyses,weusedtheWMHsegmentationmethodthatpro-
duced the strongest association to clinical diagnosis (LGA) to further evaluate the relationship between
WMH volume and cognition – specically, neuropsychological composite measures of memory or exec-
utive function, although it is important to note that LGA WMH volumes were not signicantly dierent
from those calculated using the in-house, BIANCA (masked), or intensity histogram algorithms, all of
whichweresignicantlyassociatedwithdiagnosis(Table2.7). Usingmultiplelinearregression,wefound
thatgreaterLGA-derivedWMHvolumesweresignicantlyassociatedwithlowerexecutivefunctioncom-
posite scores (omnibusp< 0.001; WMH volume partial t=2.33;p=0.021). LGA-derived WMH vol-
umeswerenotsignicantlycorrelatedwithcompositememoryscores(omnibusp< 0.001;WMHvolume
partialt=1.629;p=0.106).
We further investigated our signicant result to determine whether certain neuropsychological sub-
tests may be driving the association between LGA-derived WMH volumes and executive function. We
42
found that greater WMH volume was signicantly associated with lower Category Fluency score (om-
nibusp< 0.001; WMH volume partialt=3.12; FDR adjustedp=0.013). LGA-derived WMH volumes
were not associated with Trail Making Test Part A (omnibusp< 0.001; WMH volume partial t=1.96;
FDR corrected p=0.157) or Part B (omnibusp< 0.001; WMH volume partial t=0.79; FDR adjusted
p=0.516), or any of the three clock drawing subscores: symbol (omnibusp< 0.001; WMH volume
partial t=1.389; FDR adjusted p=0.516); numbers (WMH volume partial t=1.389; FDR adjusted
p=0.330); or time (omnibusp< 0.001; WMH volume partialt=0.151; FDR adjustedp=0.880). We
didnotexaminetherelationshipbetweenWMHvolumeandtheclockdrawingcircleorhandscorestest,
astherewasceilingeectonthesetests–ontheclockdrawingcirclesubtest,all143oftheA β-participants
received a perfect score and on the hand subtest, 142 of the 143 participants received a perfect score.
2.5 Discussion
We investigated, in a sample of non-demented Aβ- older adults, the most clinically relevant features of
WMH boundary selection, by relating WMH volume (using 5 dierent algorithms) to clinical diagnosis
andcognitivefunction. Wefoundthat1)largertotal,frontal,parietal,andperiventricularWMHvolumes,
derived from our in-house algorithm, were signicantly associated with a worse clinical diagnosis in A β-
older adults, 2) limiting WMH boundaries to voxels having the highest- intensity thresholds strength-
ened the relationship between WMH volume and clinical diagnosis, and 3) the most clinically relevant
WMH segmentation algorithms (LGA, intensity histogram, our in-house method, and BIANCA with the
WM mask option) were methods that limited boundary selection to the most high-intensity areas of the
WMHs. Our study is therst to compare multiple WMH segmentation methods using clinical diagnosis
and cognitive composite measures to assess clinical relevance.
WefoundasignicanteectbetweenWMHvolumeanddiagnosisonlyinA β-participants. Here,Aβ
positivitymayreectAD-relatedneuropathologicalchangesmorebroadly,whichmaybeassociatedwith
43
cognitiveeects,evenincognitivelyintactolderadults(Braskieetal.,2010;Amariglioetal.,2012;Hoetal.,
2018). Variability from such AD-related eects on cognition could add statistical noise to WMH-related
cognitive eects in A β+ participants, making those eects harder to detect. The relationship between
cerebral amyloidosis and WMHs is currently debated (Roseborough et al., 2017), although emerging evi-
dence indicates that Aβ and WMHs may have both independent and interactive eects (Scott et al., 2016;
Schreiner et al., 2018). WMH accumulation in Aβ- participants may represent an increased vulnerability
to developing abnormal levels of Aβ later, although future longitudinal studies are needed to clarify this
possibility. Additionally, the interaction between amyloid and WMHs may also make it more dicult to
detect an eect on cognition that is specically attributable to WMHs in A β+ participants. It is possible
that Aβ may interact with the eect of WMH accumulation on cognition dierently when evaluating co-
horts with a broader range of diagnoses, such as those with symptomatic AD (Provenzano et al., 2013)
However, within our cohort of non-demented older adults, the eect of subtle increases in WMH volume
on cognition was only detectable in individuals without abnormal amyloid levels, suggesting that AD-
specic processes may mask the eect of WMH accumulation before clinical onset of AD. Thesendings
are consistent with our hypothesis that the relationship between WMH volume and clinical diagnosis is
dependent upon both WMH boundary selection and amyloid-positivity status.
Wefoundthatlargertotal,frontal,andparietalWMHvolumesweresignicantlyassociatedwithworse
clinical diagnosis, while WMH in the temporal lobe were not. We also found that periventricular, but not
deepWMHweresignicantlyassociatedwithclinicaldiagnosis,whichisconsistentwithpastndingsre-
lating periventricular WMH to global cognition (Kim et al., 2008; Bolandzadeh et al., 2012; Grianti et al.,
2018). However, our study found trend level signicance when relating deep WMH to clinical measures.
The ability to detect a robust eect could be a result of small discrete lesions having a dierent intensity
distribution from larger lesions, with deep WMHs appearing lighter than periventricular WMHs. There-
fore,deepWMHscouldbeunder-segmentedbyvarioussegmentationmethods,resultinginperiventricular
44
WMHs appearing to have a stronger relationship to clinical variables than deep WMHs. To test this, we
performedanadditionaltestinwhichweusedamorelenientthresholdtosegmentlesshyperintensedeep
WMHs. WefoundthatwhenwesegmentedthelighterregionsofdeepWMHs,therelationshiptoclinical
diagnosisbecameweaker,suggestingthatboththelocationandintensityoflesionsareimportanttoclin-
icalrelevance. PeriventricularanddeepWMHsarebothassociatedwithseveremyelinlossandincreased
microglia activity (Simpson et al., 2007), but may also have dierent etiological origins. Periventricular
WMHs may be related to arterial pressure, plasma leakage, blood brain barrier permeability, and decline
intotalcerebralbloodow,whiledeepWMHsmaybeassociatedwithaxonalloss,arteriolosclerosis,and
body mass index (ten Dam et al., 2007b; Haller et al., 2013; Wharton et al., 2015; Grianti et al., 2018).
Although future work is needed to illuminate the mechanisms of thesendings, our work suggests that
disruption of global cognitive processes may be related to region-specic changes.
WefoundthatWMHboundaryselectionwasanimportantalgorithmfeaturethatmodiedthedegree
to which logistic regression could capture the relationship between WMH volume and clinical diagno-
sis. The WMH volumes calculated such that only the most hyperintense voxels were included, were best
associated with clinical diagnosis. We further validated our intensity thresholdndings by determining
that the WMH segmentation methods most associated with diagnosis (using default settings) were the
four algorithms that most limited WMHs to highest intensity voxels (LGA, in-house method, the inten-
sity histogram method, and BIANCA with the WM mask included as an input). The segmentations that
weresignicantlyassociatedwithdiagnosisprovidedsmallerWMHvolumesforthesamescans. Avisual
review suggests that these segmentations captured the most discrete and highly intense regions Figure
2.4). Additionally,whenwetesteddierentthresholdsforselectingvoxelsusingthethreealgorithmsthat
allowed such adjustments, the thresholds that resulted in smaller WMH volumes composed of the most
45
intense voxels were most closely related to clinical diagnosis. Our results suggest that optimization of al-
gorithm parameters to capture the most intense WMH voxels will yield more robust classication results
relevant to clinical diagnosis.
Usingonlydefaultoptions,theWMHvolumesthatresultedinthelargestvolumeswerederivedfrom
LPA and BIANCA without the WM mask and were not signicantly associated with clinical diagnosis.
LPA included lighter and more diuse hyperintense regionsin addition to brighter, more discrete lesions.
UsingadditionaloptionalparametersforLPAmayhaveproducedsignicantassociationswithdiagnosis.
BIANCAwithouttheoptionalWMmaskprovidedWMHestimatesthatappearedvisuallysimilartotheless
inclusive in-house, histogram, and LGA methods, but also included some non-white matter hyperintense
regions, such as in the cortex, cerebellum, and brainstem regions. Applying the BIANCA option to input
a white matter mask prevented the inclusion of erroneous non-WM voxels in the WMH map, and WMH
volume estimated using BIANCA with a WM mask as input was signicantly associated with clinical
diagnosis. Weimplementedeachalgorithmusingthedefaultparameterstoprovidevaryingsegmentation
results among segmentation methods allowing us to better investigate which WMH characteristics were
most clinically relevant. Our purpose was not to recommend any one software package over another.
Rather, our ndings highlight the importance of limiting the WMH search to white matter regions and
segmenting only the most hyperintense voxels, regardless of the algorithm used.
In follow-up analyses of Aβ- older adults, greater WMH volumes were associated with lower execu-
tive function composite scores. Although previous literature relating WMH volume to cognitive function
isvariable(PrinsandScheltens,2015),mountingevidencedemonstratesthatWMHvolumehasbothbroad
and specic eects on cognitive function ( Hedden et al., 2012; Kloppenborg et al., 2014). Globally, WMHs
have been associated with future cognitive decline (Boyle et al., 2016), impacting multiple neuropsycho-
logical domains (Au et al., 2006; Gunning-Dixon and Raz, 2000). However, WMHs most consistently have
46
been associated decits in processing speed and executive function ( Debette et al., 2010; Murray et al.,
2010; Kloppenborg et al., 2014; Lampe et al., 2017), consistent with our currentndings.
The association we found between WMH volume and executive function was driven primarily by
decits in category uency – a type of verbal uency test that here involves freely generating as many
animal names as possible within a set time period. Category uency, a sensitive marker for cognitive
impairment,isimpactedbyfrontallobeWMHaccumulation(Gootjesetal.,2004),asitrecruitsbothfrontal
andtemporallobebrainregions(Mummeryetal.,1996;Gootjesetal.,2004;Baldoetal.,2006;Peteretal.,
2016). Although temporal lobe WMH volume was not associated with category uency measures, the
eect of WMH volume on category uency may be attributed to disruption of more global structural
cortical connections, such as between the frontal and temporal lobes (Wiseman et al., 2018).
Our study included data only from non-demented older participants, and therefore may not be gen-
eralizable to participants with Alzheimer’s disease and other diseases aecting the white matter, such as
multiplesclerosis. WerelatedtotalandregionalWMHvolumetocognitioninourstudy,butdidnotevalu-
atehowthenumberoftotalorregionalWMHsineachbrainrelatedtocognition. Suchananalysiswould
be interesting and may yield dierent results. Use of intensity threshold to calculate WMHs, as in our
method, maydierently capture smalland largelesions, assmall discretelesionsmay havea dierentin-
tensity distribution from larger more diuse lesions. We used a visual quality control assessment of each
WMmask,andmanualeditingasneededforaccuracy. BecauseourWMmaskwascreatedbysubtracting
thegraymatterandCSFmasksfromthewholebrainontheT1-weightedimages,editingwasonlyrequired
whentheWMHandgraymatter,whichhavesimilarintensitiesontheT1-weightedimages,werecontigu-
ous, in which case, the automatic segmentation may include WMH erroneously in the gray matter mask.
ThiswasnotacommonoccurrenceintheADNIcohort,whoseparticipantsdonottendtohaveextensive
WMpathologyintheperiphery. However,inacohortthatincludesmanyparticipantswithveryextensive
WMH pathology, this manual editing step may be more time consuming. Additionally, we used a limited
47
setofopen-source WMHtoolboxeswhichmaynotcaptureallthepossiblevariabilityofWMHboundary
segmentations. These automated WMH segmentation methods have optional parameters that use dier-
ent variations of location and intensity as inputs into either a linear or nonlinear classier. We used the
default settings on the various packages in order to arrive at variable segmentations, but optimization of
these parameters may have resulted in signicant associations between the WMH segmentation volumes
andclinicaldiagnosis. Ourintentherewasnottoevaluatethesoftwarepackagesperse,buttodetermine
whattypeofsegmentationwouldbemostclinicallyrelevant. Ourconvergingresultssuggestthatmultiple
algorithms may generate useful segmentations.
Overall,ourstudysoughttosystematicallyassessautomaticWMHsegmentationstoidentifythemost
clinically meaningful results. Our ndings suggest that WMH segmentations that exclude the lightest
and most diuse hyperintensities have the strongest clinical relevance and that this relationship is most
evidentonlyinAβ-olderadults. ThissuggeststhatAD-specicprocesses,suchasamyloidaccumulation,
maymaskthecognitiveconsequencesofWMHs. However,evaluationofhigherintensityWMHvolumes
is a useful metric to classify global cognitive function and assess domain-specic changes in executive
function in older adults. Our work is an initial step toward harmonizing WMH segmentation protocols,
allowing for more robust and reliable investigations on how WMHs mechanistically relate to cognition
and sub-optimal brain aging.
2.6 SupplementaryMaterial
48
TableS2.1:Associationsbetweenin-housederivedWMHvolumebyregionandclinicaldiagnosis
inAβ-participantswithscannertypeaddedasacovariate
Region WMH Volume β WMH Volume z-
value
WMH Volume p-
value
WMH Volume
FDR-Corrected
p-value
Total 1.7 x 10
-4
2.614 0.009* –
Frontal 9.3 x 10
-4
3.378 < 0.001* 0.004*
Parietal 4.3 x 10
-4
2.451 0.014* 0.024*
Temporal 1.7 x 10
-3
1.481 0.139 0.139
Periventricular 3.0 x 10
-4
2.937 0.003* 0.008*
Deep 2.3 x 10
-4
1.929 0.054
•
0.067
•
Meanvolumeisinmm
3
. Eachrelationshipwasevaluatedusingalogisticregression,adjustedforage,sex,
yearsofeducation,ICV,andscannermanufacturer. Multiplecomparisoncorrectionwasappliedtofrontal,
parietal, temporal, periventricular, and deep WMH volume analyses, using FDR adjusted values. Scanner
type was not signicant in any of the models. Clinical diagnoses: MCI = 1; control = 0. SD = Standard
Deviation; IQR = Interquartile Range. *p< 0.05.
•
p< 0.10, indicating a trend level association.
FigureS2.1: TestingofthedivisionofperiventricularanddeepWMHboundariesbykernelsizes
4,6,and9. Blue represents lesion classied as periventricular WMH and red represents lesion classied
as deep WMH
49
Table S2.2: Associations between in-house derived WMH volume by region and executive func-
tioninAβ-participants
Region β t-value p-value FDR p-value
Frontal -5.2 x 10
-5
1.917 0.057 0.197
Parietal -2.7 x 10
-5
1.152 0.251 0.419
Temporal 8.7 x 10
-5
0.276 0.783 0.783
Periventricular -3.7 x 10
-5
1.770 0.078
•
0.197
Deep -1.5 x 10
-5
0.801 0.424 0.530
Mean volume is inmm
3
. Each relationship was evaluated using a linear regression, adjusted for age, sex,
years of education, and ICV. Multiple comparison correction was applied to frontal, parietal, temporal,
periventricular, and deep WMH volume analyses, using FDR adjusted values. Clinical diagnoses: MCI =
1; control = 0. SD = Standard Deviation; IQR = Interquartile Range.
•
p < 0.10, indicating a trend level
association.
Table S2.3: Associations between in-house derived WMH volume by region and memory in Aβ-
participants.
Region β t-value p-value FDR p-value
Frontal -2.5 x 10
-5
0.964 0.337 0.836
Parietal -1.2 x 10
-5
0.525 0.600 0.836
Temporal 5.7 x 10
-5
0.188 0.852 0.852
Periventricular -1.7 x 10
-5
0.825 0.411 0.836
Deep -7.9 x 10
-6
0.429 0.669 0.836
Mean volume is inmm
3
. Each relationship was evaluated using a linear regression, adjusted for age, sex,
years of education, and ICV. Multiple comparison correction was applied to frontal, parietal, temporal,
periventricular,anddeepWMHvolumeanalyses,usingFDRadjustedvalues. Clinicaldiagnoses: MCI=1;
control = 0. SD = Standard Deviation.
Table S2.4: Associationsbetweenin-housederivedtotalWMHvolumebythresholdandclinical
diagnosisinAβ+participants
Threshold Mean ± SD Median (IQR) β z-value p-value
85% 23031 ± 16705 19338 (12049-29575) -5.7 x 10
-6
-0.488 0.626
90% 13204 ± 12852 9351 (5368-16346) -6.8 x 10
-6
-0.448 0.654
95% 8611 ± 10786 5524 (2646-10142) -9.1 x 10
-6
-0.499 0.618
100% 6104 ± 9378 3520 (1500-6845) -1.3 x 10
-5
-0.611 0.541
105% 4493 ± 8241 2265 (886-4782) -1.8 x 10
-5
-0.712 0.476
Mean volume is in mm
3
. Each relationship was evaluated using a logistic regression, adjusted for age,
sex, years of education, and ICV. Clinical diagnoses were coded as MCI = 1; control = 0. SD = Standard
Deviation; IQR = Interquartile Range
50
TableS2.5:Associationsbetweenin-housederivedWMHvolumebyregionandclinicaldiagnosis
inAβ+participants
Region Mean ± SD Mean (IQR) β z-value p-value FDR p-value
Total 6104 ± 9378 3520 (1500-6845) -1.3 x 10
-5
-0.611 0.541 –
Frontal 1795 ± 3051 621 (209-2241) -2.7 x 10
-5
-0.411 0.681 0.681
Parietal 2202 ± 4770 794 (280-2005) -2.6 x 10
-5
-0.630 0.528 0.681
Temporal 229 ± 549 91 (32-222) -2.4 x 10
-4
-0.661 0.508 0.681
Periventricular 3815 ± 4118 2570 (1193-5103) -2.5 x 10
-5
-0.502 0.616 0.681
Deep 2289 ± 5715 643 (201-2043) -2.2 x 10
-5
-0.640 0.522 0.681
Meanvolumeisinmm
3
. Eachrelationshipwasevaluatedusingalogisticregression,adjustedforage,sex,
years of education, and ICV. Multiple comparison correction was applied to frontal, parietal, temporal,
periventricular,anddeepWMHvolumeanalyses,usingFDRadjustedvalues. Clinicaldiagnoses: MCI=1;
control = 0. SD = Standard Deviation; IQR = Interquartile Range
Table S2.6: Associationsbetweenin-housederivedWMHvolumebysub-regionanddiagnosisin
Aβ+participants
Region Mean ± SD Median (IQR) β z-
value
p-
value
FDR
p-value
Frontal Periventricular 1253 ± 1728 529 (147-1785) -4.9 x 10
-5
-0.413 0.679 0.882
Parietal Periventricular 1039 ± 1331 586 (219-1277) -1.1 x 10
-5
-0.072 0.943 0.943
Temporal Periventricular 172 ± 356 70 (20-169) -2.0 x 10
-4
-0.370 0.711 0.882
Frontal Deep 542 ± 1617 84 (22-269) -4.1 x 10
-5
-0.338 0.735 0.882
Parietal Deep 1163 ± 3640 153 (17-911) -4.4 x 10
-5
-0.767 0.443 0.882
Meanvolumeisinmm
3
. Eachrelationshipwasevaluatedusingalogisticregression,adjustedforage,sex,
years of education, and ICV. Multiple comparison correction was applied to frontal, parietal, temporal,
periventricular,anddeepWMHvolumeanalyses,usingFDRadjustedvalues. Clinicaldiagnoses: MCI=1;
control = 0. SD = Standard Deviation; IQR = Interquartile Range
Table S2.7: Associations between WMH volume by segmentation method and clinical diagnosis
inAβ+participants
Method Mean ± SD Median (IQR) β z-value p-value
In-house 6104 ± 9378 3520 (1500-6845) -1.3 x 10
-5
-0.611 0.541
Intensity Histogram 8064 ± 12885 4404 (1795-9685) -7.8 x 10
-3
-0.505 0.613
LGA 8554 ± 11812 4658 (1157-12638) -5.3 x 10
-6
-0.310 0.757
LPA 33542 ± 31110 25078 (10611-45116) 2.0 x 10
-6
0.258 0.797
BIANCA (masked) 2412 ± 5889 754 (153-2368) -2.4 x 10
-5
-0.690 0.490
BIANCA (unmasked) 60372 ± 30869 53941 (34987-82547) 1.7 x 10
-6
0.266 0.790
Meanvolumeisinmm
3
. Eachrelationshipwasevaluatedusingalogisticregression,adjustedforage,sex,
years of education, and ICV. For BIANCA, “masked” indicates that a WM mask was used as input for the
analysis. Clinical diagnoses: MCI = 1; control = 0. SD = Standard Deviation; IQR = Interquartile Range
51
TableS2.8:AssociationsbetweenWMHvolumeandclinicaldiagnosisinAβ-participantsbyLGA
lesionbeliefmapintensitythreshold
Threshold Mean Volume ± SD Median (IQR) β z-value p-value
0.1 79689 ± 8882 4860 (2435-9853) 1.52 x 10
-4
3.538 < 0.001*
0.3 (Default) 4400 ± 7131 1656 (5121-398) 2.14 x 10
-4
3.533 < 0.001*
0.5 3341 ± 6199 906 (130-3567) 2.52 x 10
-4
3.478 < 0.001*
Mean volume is in mm
3
. Each relationship was evaluated using a logistic regression, adjusted for age,
sex,yearsofeducation,andICV.Clinicaldiagnoses: MCI=1;control=0. SD=StandardDeviation;IQR=
Interquartile Range. *p< 0.05
Figure S2.2: Variation in the LGA WMH segmentation by threshold of the lesion belief map in-
tensity.
Table S2.9: Associations between WMH volume and clinical diagnosis in Aβ- participants by
BIANCAprobabilitymapthreshold
Threshold Mean Volume ± SD Median (IQR) β z-value p-value
0.85 (mask) 5999 ± 6908 3124 (1674-8020) 6.1 x 10
-5
1.866 0.062
•
0.9 (masked, default) 5000 ± 6085 2547 (1365-6379) 8.4 x 10
-5
2.145 0.032*
0.95 (masked) 4268 ± 5333 2135 (1126-5362) 1.0 x 10
-4
2.232 0.026*
0.9 (unmasked, default) 15399 ± 7256 14388 (10023-19267) 4.7 x 10
-5
1.705 0.088
•
Meanvolumeisinmm
3
. Eachrelationshipwasevaluatedusingalogisticregression,adjustedforage,sex,
years of education, and ICV. “Masked” indicates that a white matter mask was used as an input. Clinical
diagnoses: MCI = 1; control = 0. SD = Standard Deviation; IQR = Interquartile Range. Default indicates
default threshold. *p< 0.05.
•
p< 0.10, indicating a trend level association.
52
Figure S2.3: VariationintheBIANCAWMHsegmentationbytheprobabilitymapthresholdand
optionalWMmaskinput.
53
Chapter3
RegionalrelationshipsbetweenCSFVEGFlevelsandAlzheimer’s
diseasebrainbiomarkersandcognition
This section is adapted from:
Tubi, M. A., Kothapalli, D., Hapenney, M., Feingold, F. W., Mack, W. J., King, K. S., Thompson, P.
M., Braskie, M. N., for Alzheimer’s Disease Neuroimaging Initiative (2021). Regional relationships be-
tween CSF VEGF levels and Alzheimer’s disease brain biomarkers and cognition. Neurobiology of Aging,
105:241–251.
3.1 Abstract
Vascularendothelialgrowthfactor(VEGF)isacomplexsignalingproteinthatsupportsvascularandneu-
ronalfunction. Alzheimer’sdisease(AD)-neuropathologicalhallmarksinterferewithVEGFsignalingand
modify previously detected positive associations between cerebral spinaluid (CSF) VEGF and cognition
andhippocampalvolume. However,itremainsunknown1)whetherregionalrelationshipsbetweenVEGF
andglucosemetabolismandcorticalthinningexist,and2)whetherAD-neuropathologicalhallmarks(CSF
Aβ,t-tau,p-tau)alsomodifytheserelationships. Weaddressedthisin310Alzheimer’sDiseaseNeuroimag-
ing Initiative (ADNI) participants (92 cognitively normal, 149 mild cognitive impairment, 69 AD; 215 CSF
Aβ+, 95 CSF Aβ-) with regional cortical thickness and cognition measurements and 158 participants with
54
FDG-PET. In Aβ+ participants (CSF A 42
192 pg/mL), higher CSF VEGF levels were associated with
greaterFDG-PETsignalintheinferiorparietal,andmiddleandinferiortemporalcortices. AbnormalCSF
amyloidandtaulevelsstrengthenedthepositiveassociationbetweenVEGFandregionalFDG-PETindices.
VEGF also had both direct associations with semantic memory, as well as indirect associations mediated
by regional FDG-PET signal to cognition.
3.2 Introduction
Vascularendothelialgrowthfactor-A(VEGF),asignalingproteinencodedbytheVEGF-Agene,isinvolved
in vascular and metabolic physiological processes including blood vessel growth,oxygen and glucose de-
livery, vasodilation, and vascular permeability (Lange et al., 2016). Homeostatic VEGF signaling is also
integral to the maintenance of cognitive function (Cao et al., 2004), as VEGF promotes neurogenesis (Cao
et al., 2004; Fournier et al., 2012), improves synaptic plasticity (Licht et al., 2011), and counteracts neu-
rodegeneration(Góra-KupilasandJośko,2005;Zacchignaetal.,2008)-allessentialtothepreservationof
cognitive ability in older adults.
Greater VEGF availability is neuroprotective in pre-clinical Alzheimer’s disease (AD) models (Religa
et al., 2013; Spuch et al., 2010; Wang et al., 2011), but in-vivo human studies have found varying relation-
ships between VEGF levels and AD diagnosis and cognition. Specically, in small independent studies,
cerebrospinaluid(CSF)VEGFlevelsdidnotdierbydiagnosis( Chakrabortyetal.,2018)orwerehigher
inADpatientscomparedtocontrols(Tarkowskietal.,2002)However,inlargerAlzheimer’sDiseaseNeu-
roimaging Iniative (ADNI) studies, higher CSF VEGF levels were associated with less cognitive decline
(Hohman et al., 2015; Paterson et al., 2014) and this relationship was stronger in those with abnormal
levels of AD-neuropathological hallmarks - CSF Aβ
42
and CSF t-tau (Hohman et al., 2015).
55
TheassociationbetweenVEGFandcognitionmaybestrongerinthepresenceofADpathology,possi-
bly due to bidirectional eects between VEGF and amyloid and tau on brain function. Amyloid can mod-
ify homeostatic VEGF signaling; for instance, in AD human post-mortem temporal cortex tissue samples,
VEGFco-accumulatedwithpre-aggregatedAβplaqueswithahighanityandspecicitytoA βpeptides,
ultimatelydepletingVEGFbioavailability(Yangetal.,2004). AβcanalsointerferewithVEGFsignalingby
modifyingVEGFreceptorexpression(Angometal.,2019;Choetal.,2017). Conversely,higherVEGFlevels,
generated through exogenous application of VEGF, can also rescue vascular damage, attenuate memory
impairment when it exists, and reduce amyloid and hyperphosphorylated tau load (Religa et al., 2013).
Thus, presence of higher VEGF levels may compensate for amyloid-driven VEGF inhibition, resulting in
better brain function and cognitive ability. While the relationship between VEGF and tau has not been
fullyelucidated,arecentanimaltauopathymodeldemonstratedthattauinducedbloodvesselabnormali-
tieswithanassociatedup-regulationofVEGF-Ageneexpression(Bennettetal.,2018). Collectively,these
studies highlight both independent and synergistic relationships between VEGF and AD neuropathologi-
cal hallmarks, but it remains unknown whether these interactionsexist for early brain biomarkers on the
AD cascade, such as cerebral glucose metabolism and cortical thickness.
By the time an individual has evidence of cognitive decline and/or an AD diagnosis, the brain has
alreadyundergoneseverestructuralandfunctionaldegeneration,makingitessentialtoassessfactors,such
as VEGF, that may contribute to brain biomarker changes early in the AD cascade. Neuroimaging brain
biomarkersuctuate early and follow a specic spatial and temporal sequence in disease course ( Jack Jr
and Holtzman, 2013). Identifying the spatial relationship of vascular factors to brain measures may also
illuminatepossiblevascularmechanismsthatcontributetoheterogeneityinADneuroimagingphenotype
patterns (Risacher et al., 2017). Critically, few studies have mapped how human VEGF levels are related
to early AD-brain biomarkers. Data derived from the ADNI indicate that higher CSF VEGF levels are
associated with less hippocampal atrophy and greater total cerebral glucose metabolism (Hohman et al.,
56
2015; Wang et al., 2018), but regional cortical indices of glucose metabolism and gray matter thickness
have yet to be identied. Mapping these regional relationships will help clarify how VEGF relates to the
topographical trajectory of AD-related pathology and disease progression, informing future mechanistic
and prevention studies.
To address these gaps, we evaluated whether CSF VEGF levels were related to regional measures of
1)urodeoxyglucose(FDG)positronemissiontomography(PET)and2)andmagneticresonanceimaging
(MRI) derived cortical thickness in AD-signature regions (Wang et al., 2015a), by assessing these associ-
ations in both Aβ+ and Aβ- participant subgroups (Shaw et al., 2009b) in an ADNI cohort (N = 310). We
alsoevaluatedwhethertheseregionalassociationsweremodiedbyCSFphosphorylated
181
tau(p-tau),an
AD-specic marker of tau phosphorylation, and total tau (t-tau), a non-specic marker of neuronal dam-
agethatiscloselylinkedtocognitivedysfunctioninearlystagesofAD(Blennowetal.,2010;Heddenetal.,
2013; Holtzman, 2011). We hypothesized that in those along the AD-continuum (i.e., Aβ+ participants),
higher endogenous VEGF levels (which may compensate for amyloid-driven VEGF inhibition) would be
associated with greater glucose metabolism and cortical thickness in AD-signature regions. Whereas in
Aβ-participants,wehypothesizedthathigherendogenousVEGFlevelswouldnotbepositivelyassociated
with AD-brain biomarkers, since a lower amyloid load would likely not interfere with VEGF availability.
Sinceabnormaltauaccumulationismorecloselylinkedtochangesinbrainfunctionandcognitivedecline
inADthanamyloid(Heddenetal.,2013;Ossenkoppeleetal.,2019),wehypothesizedthattheassociation
betweenVEGFlevelsandAD-relevantbrainneuroimagingbiomarkerswouldbestrongestinthepresence
of higher CSF p-tau and t-tau levels.
In a secondary analysis, we reproduced a previous analysis by Hohman et al., (Hohman et al., 2015)
that evaluated the relationships between VEGF levels and cognitive domain measures of executive func-
tion (ADNI-EF) and episodic memory (ADNI-MEM). We further expanded on the analysis by assessing
test-specic eects in cognitive domain measures. We also evaluated whether the previously-detected
57
association between VEGF and cognition was mediated by regional FDG-PET associations. We hypothe-
sized that regional FDG-PET signal would mediate the detected association between VEGF and cognition
in those along the AD continuum (Aβ+ participants).
3.3 Methods
3.3.1 TheAlzheimer’sDiseaseNeuroimagingInitiative(ADNI)
Data used in the preparation of this article were obtained from the Alzheimer’s Disease Neuroimaging
Initiative(ADNI)database(https://ida.loni.usc.edu). TheADNIwaslaunchedin2004asapublic-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), PET, other biological markers, and clinical and
neuropsychologicalassessmentcanbecombinedtomeasuretheprogressionofmildcognitiveimpairment
(MCI)andearlyAD.Forup-to-dateinformation,seewww.adni-info.org. Alldatautilizedinthisstudywere
acquired from the publicly available ADNI Image Data Archive (https://ida.loni.usc.edu).
3.3.2 Participants
WeselectedparticipantsfromtheADNIcohortwhohadavailableCSFVEGF,1.5TstructuralMRIscan,and
neuropsychological assessment within 1 year of each other, resulting in 310 participants (92 cognitively
intact(CI),149mildcognitiveimpairment(MCI),69AD),aged56to89yearsold. CSFwasacquiredwithina
mean±standarddeviation(SD)of28.6±28.5days(max=237days)fromtheMRIscan. Neuropsychological
assessments were acquired within a mean ± SD of 18.6 ± 24.5 days (max = 237 days) from the MRI scan.
Inasubsetofthecohort,we analyzed158participantswhoalsounderwentFDG-PETscanning(39CI,80
MCI, 39 AD) acquired within a mean ± SD of 30.0 ± 26.7 (max = 203 days) days from the MRI scan.
DetailedinformationonADNI’sparticipantselectionanddiagnosticcriteriaareoutlinedintheADNI
protocol (http://www.adni-info.org/Scientists/AboutADNI.aspx). Briey, participants were excluded from
58
the ADNI cohort if they had a serious neurological or neuropsychiatric condition, or history of brain
injury. Participants were classied into diagnostic categories of probable AD, MCI, and CI. Criteria for
an AD diagnosis included: 1) a memory complaint, 2) objective memory dysfunction identied by an
education-adjustedWechslerMemoryScale-Revised(WMS-R)-LogicalMemoryII(LM-II)R(WMS-RLM-
II),3)Mini-MentalStateExam(MMSE)scorebetween20-26(inclusive),4)ClinicalDementiaRating(CDR)
0.5), and 5) NINDS/ADRDA criteria for probable AD (McKhann et al., 1984). MCI diagnosis was made
when 1) a reported memory complaint was present, but the diagnostic criteria for dementia was not met,
or 2) when they scored between 24-30 (inclusive) on the MMSE and 0.5 on the CDR (with CDR memory
box score 0.5 or higher). CI participants did not meet the criteria for probable AD or MCI and had no
memory complaints. Participant characteristics are detailed in Table 3.1.
3.3.3 CSFanalytes
AllCSFanalytemeasurementswerepreviouslyanalyzedbyothersandaccessedthroughADNI’spublicly
available data repository (http://adni.loni.usc.edu). In this analysis, CSF analytes used were VEGF-A,
amyloid-beta
42
(Aβ), total-tau (t-tau), and phosphorylated
181
-tau (p-tau). CSF samples were acquired at
thebaselinevisitandwerebatchprocessedusingLuminexMulti-AnalyteProling(xMAP)immunoassay
technology (Myriad Rules Based Medicine; RBM, Austin, TX), with standard ADNI protocols (Spellman
et al., 2015). We used the ADNI CSF Aβand total t-tau quality-checked median scaled analyte measure-
ments. Clinical performance of these analytes was previously assessed in a precision analysis conducted
in an inter-laboratory standardization study across seven centers usingve CSF pools from 3 cognitively
normalolderadultsand2ADpatients(Shawetal.,2011). Allsampleswereruninduplicate. Amyloidpos-
itivity was dened by cut-os outlined in prior work; A β+ individuals had CSF A 42
< 192pg/mL and
Aβ- participants had CSFA 42
levels 192pg/mL (Shaw et al., 2009b). Quality control (QC) procedures
for CSFA 42
, CSF p-tau, and t-tau are described elsewhere (Shaw et al., 2009b, 2011).
59
Table 3.1: ParticipantCharacteristics
Total Cohort Aβ+Aβ- p-value
N 310 215 95
Age (years) 74.8 ± 6.8 74.6 ± 6.9 75.5 ± 6.6 0.60
Sex 0.65
Males 187 (60.3%) 132 (61.4%) 55 (57.9%)
Females 123 (39.7%) 83 (38.6%) 40 (42.1%)
Diagnosis <0.001*
A
Cognitively intact (CI) 92 (29.7%) 33 (15.3%) 59 (62.1%)
Mild cognitive impairment (MCI) 149 (48.1%) 117 (54.4%) 32 (33.7%)
Alzheimer’s disease (AD) 69 (22.2%) 65 (30.2%) 4 (4.2%)
Apolipoprotein e4 (APOE4) allele count <0.001*
A
0 APOE4 alleles 159 (51.3%) 74 (34.4%) 85 (89.5%)
1 APOE4 alleles 115 (37.1%) 105 (48.8%) 10 (10.5%)
2 APOE4 alleles 36 (11.6%) 36 (16.7%) 0 (0.0%)
Education (years completed) 15.7 ± 3.0 15.6 ± 3.0 15.8 ± 2.8 0.88
MMSE score 26.8 ± 2.6 26.1 ± 2.6 28.4 ± 1.7 <0.001*
A
CSF VEGF (natural log of pg/mL) 2.70 ± 0.13 2.69 ± 0.13 2.74 ± 0.13 <0.001*
A
CSF t-tau (pg/mL) 98.4 ± 52.1 113.8 ± 54.4 63.4 ± 20.5 <0.001*
A
CSF p-tau (pg/mL) 33.9 ± 17.4 39.7 ± 17.3 20.9 ± 7.6 <0.001*
A
Continuousvariablesshownasmean ±standarddeviation. Categoricalvariablesshownasfrequency(%).
Amyloid group level dierences (A β- vs. Aβ+) were evaluated on continuous variables using a Welch’s
two-samplet-testandcategoricalvariablesusinga 2
test. *indicatesp< 0.05groupdierencebetween
Aβ- (n=95) vs. Aβ+ (n=215) participants.
A
indicatesp< 0.05 between Aβ- (n=42) vs. Aβ+ (n=116)
participants in subset with available FDG-PET data.
We downloaded CSF VEGF values from ADNI Biomarkers Consortium CSF QC Multiplex data sheet.
The CSF samples were acquired from participants after an overnight fast at the ADNI 1 baseline visit.
The samples were processed, aliquoted, and stored at -80°C in accordance with ADNI Biomarker Core
Laboratory Standard Operating Procedures (http://adni.loni.usc.edu/wp-content/uploads/2010/09/CSF_
Biomarker_Test_Instr.pdf). TheQCproceduresperformedbyADNIontheCSFVEGFassay(andallADNI
CSFproteomicassays)includedtest/retestonasubsampleof16CSFsamplestotestfortheprecision,cross-
reactivity,spike-recovery(accuracy),andreliability. NomissingdataandnooutlierswerereportedforCSF
VEGF values, according to ADNI Biomarkers Consortium CSF QC multiplex data. VEGF measurements
were natural log-transformed by ADNI (Box and Cox, 1964).
60
3.3.4 APOE genotyping
We adjusted for Apolipoprotein "4(APOE4) genotype in all analyses. APOE4 genotype count was coded
as0forAPOE4non-carriers,1forcarriersofoneAPOE4allele,and2forcarriersoftwoAPOE4alleles. We
codedbydosebecauseofpossibledose-dependenteectsofAPOE4oncognitiveability,vascularfunction,
andMRIbiomarkers(Makkaretal.,2020;Hobeletal.,2019). APOEgenotypingwasperformedbyCogenics
from a 3mL aliquot of blood. PCR amplication and Hhal restriction enzyme digestion was performed,
which was then resolved on 4% Metaphor Gel and viewed by ethidium bromide staining (Saykin et al.,
2010). More information on the ADNI APOE genotyping protocol can be found here: http://adni.loni.
usc.edu/methods.
3.3.5 MRIprocessing
AllparticipantsunderwentMRIscanningofasagittalT1-weighted3DMagnetizationPreparedRapidAc-
quisitionGradientEcho(MPRAGE)at1.5T,acquiredacross55sites. Scanswereconductedandcoordinated
across sites for optimal comparability according to ADNI protocols (http://adni.loni.usc.edu/methods/
documents/mri-protocols/)andpassedqualitycontrol(QC)(Wymanetal.,2013). WeprocessedMRIswith
FreeSurfer 5.3 (http://surfer.nmr.mgh.harvard.edu/)(Dale et al., 1999; Fischl and Dale, 2000) to calculate
regional cortical thickness in a-priori selected regions-of-interest (ROI) in an AD-cortical thinning signa-
ture: theentorhinalcortex(ERC),posteriorcingulate,superior(STG),middle(MTG),andinferiortemporal
gyri(ITG),fusiformgyri,superiorandinferiorparietalcortex,andprecuneus(Wangetal.,2015a). ADsig-
natureROIsarevisualizedinFigure3.1. TheseROIswereselectedastheydetectsubtleandearlyneuronal
loss, correlate spatially with tau progression and cognitive decline (Burggren et al., 2008b; Querbes et al.,
2009;Wangetal.,2015a). Allindividualregionswerevisuallyqualitycheckedandregionalcorticalthick-
ness measures that failed our QC protocol were excluded from the study. Raters (n = 2) performed a test
61
Figure 3.1: AD-signatureROIsextractedfromFreeSurfermappedonaskull-strippedbrain
for QC on a known training set and were required to obtain reliability indicating substantial agreement
(Cohen’s> 0.60) to the answer key before performing quality checks in the current study.
3.3.6 FDG-PETprocessing
158 participants (39 CI, 80 MCI, 39 AD) underwent a fasting FDG-PET scan at their baseline visit, using
a standard protocol (six 5-minute frames collected 30 minutes after a 5 mCi [18F]-FDG injection) (Jagust
etal.,2010). Wedownloadedpre-processedFDG-PETscans(Co-registered,Averaged,StandardizedImage
and Voxel Size, Uniform Resolution), from the LONI IDA website (https://ida.loni.usc.edu). FDG-PET
pre-processing was performed according to standardized protocols to reduce inter-scanner eects (Joshi
62
et al., 2009) and enable data pooling. Pre-processing steps included averaging the frames and spatially
aligning (via interpolation) to a standard voxel size and resolution.
We constructed a study-specic minimum deformation template (MDT) to reduce the total deforma-
tioneortneededtoco-registerimages. TheMDTwasconstructedfromT1-weighted3DMPRAGEscans
derivedfrom26non-demented(11CI,15MCI)participantsfromoursample. MDTconstructionhasbeen
previouslydescribed(Braskieetal.,2014). Briey,welinearlytransformedtheskull-strippedT1-weighted
MPRAGE scans into a standard template space. Next, an ane template was created by taking the voxel-
wise average of the aligned scans. The scans were then iteratively, nonlinearly transformed to the ane
template (Yanovsky et al., 2009, 2008).
FDG-PET voxelwise intensity values were divided by a pons-vermis mean reference region value -
dened as the top 50% of the pons and cerebellar vermis on corresponding slices, as recommended for
cross-sectional FDG-PET analyses (Landau et al., 2012). We manually segmented the reference region in
MNI standard template space, guided by an atlas (Cerebellum MNIfnirt-maxprob-thr50-1mm) and subse-
quently performed linear and non-linear transformations using FMRIB’s Linear Image Registration Tool
(FLIRT) in FSL (Jenkinson and Smith, 2001; Jenkinson et al., 2002) and ANTs symmetric image normal-
ization (SyN) method (Avants et al., 2008, 2011). ROIs were dened by the same FreeSurfer AD-signature
ROIs listed in Figure 3.1 and co-registered to the participant’s FDG-PET scan through linear and nonlin-
ear transformations. After all co-registrations passed visual QC inspection, the mean FDG-PET standard
uptake value ratio (SUVR) of each AD-signature ROI was extracted.
3.3.7 Neuropsychologicalassessment
Validated neuropsychological composite measures of executive function (ADNI-EF) and memory (ADNI-
MEM)wereusedtoassessdomain-speciccognitivefunction( Craneetal.,2012;Gibbonsetal.,2012). The
composite measures were developed using item-response theory (IRT) and were normalized to a mean of
63
0 and variance of 1. ADNI-EF was derived from Trail Making Test parts A and B, Digit Span Backwards
tests from the Wechsler Adult Intelligence Scale-Revised (WAIS-R), Digit Symbol Substitution, Category
Fluency - Animals, Category Fluency – Vegetables, and the Clock Drawing test. ADNI-MEM was derived
from the Rey Auditory Verbal Learning Test (RAVLT), the AD Assessment Schedule- Cognition (ADAS-
COG), the MMSE, and Wechsler Logical Memory I and II. While composite measures have the benet of
limiting multiple comparisons, thus boosting the power to detect eects, composite measures also have
limitations. Forinstance,ADNI-EFmaymaskeectsofinterest,suchasprocessingspeed,thatoccurwhen
a cognitive domain includes heterogeneous components (Jurado and Rosselli, 2007). Therefore, our study
soughttoaddressthelimitationsofdomain-specicassessmentandexpandontheanalysisbetweenADNI
VEGF levels and neuropsychological composite measures (Hohman et al., 2015) to identify test-specic
associations in each subtest included in the composite measure.
3.3.8 Statisticalanalyses
3.3.8.1 RelationshipbetweenVEGFandAmyloidlevels
WeconductedallstatisticalanalysesinRStudioversion1.0.136(http://www.R-project.org/)(Universityof
Auckland, Auckland, New Zealand) (R Core Development Team, 2010). We stratied the cohort by CSF
amyloid level (Aβ-, Aβ+) and evaluated amyloid group dierences in participant characteristics using a
2
test for categorical variables and a Welch’s two-sample t-test for continuous variables (Table 3.1). We
also compared VEGF levels by amyloid and t-tau positivity groups (Supplementary Figure S3.1). We also
evaluated whether CSF Aβlevels were signicantly associated with VEGF (log-transformed) values using
linearregressionwithsplinemodelingwithaknotat192pg/ml (i.e.,thethresholdforamyloidpositivity)
(Supplementary S3.2). We included age, sex, diagnosis, years of education, CSF t-tau, and APOE4 allele
countinthemodeltoidentifyassociationsbetweenthesevariablesandVEGF(logtransformed)valuesin
64
theentirecohort(N=310),aswellasinAβ+(n=215),andAβ-groups(n=95)separately(Supplementary
Table S3.1).
3.3.8.2 VEGFassociationswithADbrainbiomarkers: FDG-PET&MRIcorticalthicknessROIs
We performed all primary analyses separately in Aβ- and Aβ+ participants to separate individuals who
are on the AD continuum and to identify associations relevant to AD pathogenesis. Statistical rationale
also supported analyzing associations separately by Aβ- stratum because 1) CSF Aβ levels in this study
have a bi-model distribution, and 2) there are far fewer Aβ- participants than there are Aβ+ participants.
Moreover,sinceinteractiontestsarenotoriouslyoflowstatisticalpower,wesoughttodetectmoresubtle
dierencesinassociationsbetweengroupsbyanalyzingA βgroupsseparately. However,wealsoformally
tested for statistical interactions using continuous Aβ values.
In each Aβstratum analysis, we used a linear mixed eects model to evaluate regional associations
betweenVEGF(logtransformed)valuesandeachFDG-PETSUVRROI(Table3.2)andeachcorticalthick-
nessROI(Table3.4). Wespeciedarandomeectforacquisitionsite. Wecovariedforage,sex,diagnosis,
years of education, CSF t-tau, and APOE4 allele count. We also tested whether adding the sampling date
dierence between the date of CSF collection and the date of the MRI scan as a covariate in the statistical
model modied the results. After adding sampling date dierence as a covariate in the interaction anal-
yses between CSF VEGF and CSF Aβ and t-tau on FDG-PET and cortical thickness, we did notnd any
dierenceinsignicanteectsorinterpretation. Therefore,wedidnotaddsamplingdatedierenceinthe
statistical models for the reportedndings. We used the false discovery rate (FDR) to correct for multiple
comparisons (multiple ROIs).
Usingthesamemixedeectsmodelingstrategy,wealsoevaluatedwhethercontinuousCSFt-tauand
p-tau modied the association between VEGF and regional FDG-PET (Table3.3) and cortical thickness
(Supplementary Table S3.2). We added a product interaction term between VEGF and continuous CSF
65
t-tau and p-tau (separately) on each FDG-PET and cortical thickness ROI and additionally covaried for
continuous Aβ.
3.3.8.3 VEGFassociationswithcognitivemeasures: asecondaryreproducibilityanalysis
We reproduced a cross-sectional analysis Hohman et al. (2015) in a similar ADNI dataset and expanded
on the analysis. We reproduced the interaction analysis between VEGF and continuous Aβ on ADNI-EF
and ADNI-MEM, but also evaluated whether an association would be detected when stratied by A β-
positivity. Weusedalinearmixedeectsmodel(covaryingforage,sex,diagnosis,yearsofeducation,CSF
t-tau, and APOE4 allele count) and included random eects for acquisition site. This statistical approach
variedslightlyfromthatusedinHohmanetal.(2015);Hohmanetal.(2015)usedagenerallinearmodelto
test associations, whereas we additionally covaried for site, APOE4 allele count, as well as CSF t-tau and
continuous Aβ in the interaction analyses.
Weexpandedupontheanalysisandaddressedlimitationsofthecognitivecompositemeasurestested
byevaluatingassociationsbetweenVEGFandeachsubtestincorporatedinthecompositemeasure. Since
associations were only signicant at the composite level (ADNI-EF) in the A β+, but not the Aβ- stratum,
weusedalinearmixedeectsmodeltocorrelateVEGFwitheachADNI-EFcompositemeasuresubtestin
the Aβ+ stratum. Each neuropsychological subtest that comprised the cognitive composite measure was
Z-scoretransformedusingtheentirecohort(N=310)toobtainthemeanandSDforthetransformation. In
neuropsychological subtests that are binary (e.g., clock drawing items), we performed logistic regression
using a generalized linear mixed eects model with a logit link function.
3.3.8.4 Mediationanalysis
WeperformedmediationanalysestoevaluatethedirectandindirectassociationsbetweenVEGFandcog-
nitioninAβ+participants. Weevaluatedtheextenttowhichthepreviouslyreportedassociationbetween
66
VEGF and ADNI-EF was direct or mediated by regional FDG-PET SUVR. We performed three causal me-
diation analyses (Table 3.5) across the three regions (inferior parietal cortex, MTG and ITG) in which we
found a signicant relationship between VEGF levels and FDG-PET signal in A β+ participants. The R
package mediation 4.4.6 (simulations = 1000) was used to perform all analyses.
3.4 Results
3.4.1 RelationshipbetweenVEGFandAβlevels
A summary of participant characteristics in the entire cohort (N = 310) and separately in Aβ+ (n = 215)
andAβ-(n=95)stratumcanbefoundinTable3.1,wherewealsocomparedamyloidgroup(Aβ+vs. Aβ-)
dierencesinthesevariables. Dierencesbetweengroupswerefoundindiagnosisand APOE4allelecount
and Aβ+ individuals had signicantly lower VEGF levels, lower MMSE scores, and higher t-tau and p-tau
levels than Aβ- individuals (Table 3.1). In a separate analysis of CSF VEGF levels by AD neuropathology
load,individualswhowereAβ+andt-tau-(n=91)hadsignicantlylowerVEGFlevelsthanthosewhowere
both Aβ- and t-tau- (n = 86;p< 0.001) and those who were both Aβ+ and t-tau+ (n = 124,p< 0.001)
(Supplementary S3.1). VEGF levels did not dier between the A β- t-tau- and Aβ+ t-tau+ groups. We also
evaluatedtheassociationbetweencontinuousCSFAβlevels(usingapiecewiselinearmixedeectsspline
modelwithaknotat192pg/mL;cutoforA β-positivity)andVEGF.Thefulltableofstatisticalassociations
can be found in the Supplementary Table S3.1. Briey, older age, MCI diagnosis, male sex, higher CSF t-
tau, and greater CSF Aβ levels in both individuals with Aβ 192pg/mL (Aβ+) and in those with Aβ
> 192pg/mL (Aβ-) were associated with higher VEGF levels.
3.4.2 VEGFassociationswithregionalFDG-PETSUVR
All associations between VEGF levels and FDG-PET ROIs by Aβ stratum are in Table 3.2. In Aβ+ partici-
pants (n = 116), higher VEGF levels were associated with a higher mean bilateral FDG-PET signal in the
67
Figure 3.2: Correlation plots between VEGF and FDG-PET SUVR in the inferior parietal cortex,
MTG,andITG,stratiedbyCSFamyloid(<192pg/mL),t-tau(>93pg/mL),andp-taupositivity
(>23pg/mL)
68
Table 3.2: AssociationsbetweenVEGFlevelsandFDG-PETSUVRROIanalyzedineachAβstra-
tum
Aβ+ (n=116) Aβ- (n=42)
ROI β SE p-value FDR-p-value β SE p-value FDR-p-value
Superior Temporal Gyrus 0.15 0.08 0.065 0.15 -0.07 0.16 0.65 0.73
Middle Temporal Gyrus 0.25 0.09 0.006 0.027* -0.15 0.19 0.44 0.73
Inferior Temporal Gyrus 0.26 0.09 0.003 0.027* 0.04 0.17 0.81 0.81
Fusiform Gyrus 0.11 0.09 0.19 0.22 0.09 0.16 0.57 0.73
Entorhinal Cortex -0.05 0.08 0.48 0.48 0.10 0.16 0.52 0.73
Superior Parietal Cortex 0.17 0.12 0.15 0.22 -0.17 0.20 0.41 0.73
Inferior Parietal Cortex 0.32 0.12 0.009 0.027* -0.18 0.22 0.43 0.73
Posterior Cingulate 0.16 0.11 0.17 0.22 -0.30 0.28 0.30 0.73
Precuneus 0.19 0.14 0.20 0.22 -0.17 0.26 0.52 0.73
Each association was evaluated using a linear mixed eects model. Fixed eects included CSF VEGF, age,
sex, years of education, and CSF t-tau, APOE4, and diagnosis. Site was modeled using random eects. SE
= Standard Error; FDR = False Discovery Rate. * indicatesp< 0.05 signicance
inferior parietal cortex (VEGF partial β (SE) = 0.32 (0.12); FDR-adjusted p=0.027), MTG (VEGF partial
β (SE) = 0.25 (0.09); FDR-adjusted p=0.027), and ITG (VEGF partial β (SE) = 0.26 (0.09); FDR-adjusted
p=0.027). No signicant association between VEGF levels and FDG-PET SUVR ROIs were found in
Aβ-participants(n=42). Thebeta-estimatesbetweenAβ+andAβ-participantsweresignicantlydier-
ent in the posterior cingulate cortex (FDR interaction p=0.0496), but no other region had a signicant
interactive eect that passed FDR correction (FDR interactionp-values > 0.21).
We evaluated interaction eects between VEGF and CSF t-tau and p-tau on regional FDG-PET SUVR
(Table 3.3). After FDR correction, interactions between VEGF and t-tau and VEGF and p-tau were both
signicant in the MTG and inferior parietal cortex (FDR-adjustedp-values < 0.037). The ITG, superior
parietalcortex,andprecuneusalsowereregionstohavesignicantinteractionsbetweenVEGFandp-tau
(allFDR-adjustedp-values=0.029)andtrendlevelinteractionsbetweenVEGFandt-tau(allFDR-adjusted
p-values=0.051),wherethosewithhigherlevelsoftauhadastrongerpositiveassociationbetweenVEGF
andregionalFDG-PETSUVRthanthosewithlowertau. Figure3.2illustratesplotsofinteractionsbetween
VEGF levels and CSF Aβ, t-tau, and p-tau on FDG-PET SUVR in the inferior parietal, MTG, and ITG.
69
Table 3.3: Interactions between CSF VEGF levels and CSF t-tau and p-tau on FDG-PET ROIs
(n=158)
VEGF x t-tau VEGF x p-tau
ROI β SE p-value FDR p-value β SE p-value FDR p-value
Superior Temporal Gyrus 0.001 0.001 0.225 0.29 0.005 0.004 0.202 0.227
Middle Temporal Gyrus 0.003 0.001 0.008 0.037* 0.010 0.002 0.015 0.029*
Inferior Temporal Gyrus 0.003 0.001 0.028 0.051 0.011 0.004 0.010 0.029*
Fusiform Gyrus 0.0009 0.001 0.44 0.495 0.007 0.004 0.085 0.127
Entorhinal Cortex -0.00003 0.001 0.97 0.97 0.004 0.004 0.270 0.270
Superior Parietal Cortex 0.003 0.001 0.026 0.051 0.01 0.005 0.009 0.029*
Inferior Parietal Cortex 0.004 0.002 0.008 0.037* 0.013 0.005 0.013 0.029*
Posterior Cingulate 0.003 0.002 0.067 0.10 0.008 0.006 0.153 0.197
Precuneus 0.004 0.002 0.028 0.051 0.016 0.006 0.016 0.029*
Each association was evaluated using a linear mixed eects model. Fixed eects included CSF VEGF, age,
sex, years of education, and CSF Aβ, APOE4, and diagnosis. Site was modeled using random eects. SE =
Standard Error; FDR = False Discovery Rate. * indicatesp< 0.05 signicance
3.4.3 VEGFassociationswithregionalcorticalthickness
In Aβ+ participants, higher VEGF levels were nominally associated with greater MTG cortical thickness
(VEGF partial β (SE) = 0.24 (0.12); p=0.049), but after FDR correction VEGF was not associated with
any cortical thickness ROI in Aβ+ participants. In Aβ- participants, higher VEGF was associated with a
thinner superior parietal cortex (VEGF partial β (SE) = -0.41 (0.14); FDR-adjusted p=0.032) and trend
level associated with a thinner cortex in the STG, fusiform, and inferior parietal cortex (all regions VEGF
FDR-adjustedpartialp=0.062). Thebeta-estimatesnominallydieredbetweenA β+andAβ-participants
in the MTG (interactionp=0.031) and superior parietal cortex (interactionp=0.015), but did not pass
FDRcorrection(FDR-adjustedinteractionp-values=0.14). NosignicantinteractionsbetweenVEGFand
t-tau or p-tau were found on cortical thickness ROIs (Supplementary Table S3.2).
70
Table 3.4: AssociationsbetweenVEGFandcorticalthicknessROIanalyzedineachAβstratum
VEGF x t-tau VEGF x p-tau
ROI β SE p-value
FDR p-value
β SE p-value
FDR p-value
Superior Temporal Gyrus 0.23 0.12 0.051 0.15 -0.35 0.15 0.028 0.062
Middle Temporal Gyrus 0.24 0.12 0.049 0.15 -0.23 0.15 0.13 0.17
Inferior Temporal Gyrus 0.20 0.12 0.098 0.22 -0.22 0.14 0.12 0.17
Fusiform Gyrus 0.23 0.12 0.051 0.15 -0.35 0.15 0.028 0.062
Entorhinal Cortex 0.46 0.34 0.18 0.29 -0.42 0.49 0.39 0.43
Superior Parietal Cortex -0.09 0.10 0.39 0.44 -0.41 0.14 0.004 0.032*
Inferior Parietal Cortex 0.12 0.11 0.28 0.35 -0.34 0.14 0.021 0.062
Posterior Cingulate 0.13 0.10 0.19 0.29 -0.11 0.14 0.43 0.43
Precuneus 0.13 0.10 0.99 0.99 -0.20 0.12 0.10 0.17
Each relationship was evaluated using a linear mixed eects model. Fixed eects included VEGF, age,
sex, years of education, and CSF t-tau, APOE4, and diagnosis. Site was modeled using random eects.
FDR=False Discovery Rate. SE = Standard Error. * indicatesp< 0.05 signicance.
3.4.4 VEGFassociationswithcognitivemeasures:Asecondaryreproducibilityanalysis
Hohman et al. (2015) found that higher VEGF levels (in an interaction analysis with continuous CSF Aβ
andt-tau)werenotassociatedwithbaselineADNI-EForADNI-MEM,butwereassociatedwithlessADNI-
EF and ADNI-MEM longitudinal decline. In our reproduction of the Hohman et al. 2015 baseline cross-
sectional analysis between VEGF and ADNI-EF and ADNI-MEM, we found similar negative results that
neuropathological hallmarks do not interact with VEGF on ADNI-EF (Aβ interaction p =0.65; t-tau
interactionp=0.43)andADNI-MEM(Aβinteractionp=0.45;t-tauinteractionp=0.09). Whenweran
the analysis without an interaction term, we found that higher VEGF levels were associated with greater
71
ADNI-EF(VEGFpartialβ(SE)=1.14(0.38);FDR-adjustedp=0.006)andADNI-MEMscores(VEGFpartial
(SE) = 0.67 (0.30); FDR-adjustedp=0.026), whereas Hohman et al. did notnd a main eect of VEGF
on composite measures cross-sectionally. In our stratum-specic analysis, we found in A β+ participants
only (n = 215), higher VEGF levels were signicantly associated with better ADNI-EF scores (CSF VEGF
partial β (SE) = 1.19 (0.49); FDR-adjusted p=0.031). The association between VEGF levels and ADNI-
MEM in Aβ+ participants had trend-level signicance (VEGF partialβ (SE) = 0.67 (0.37); FDR-adjusted
p=0.070). In Aβ- participants, VEGF was not related to ADNI-EF (VEGF partial β (SE) = 0.29 (0.61);
FDR-adjusted p=0.635) or ADNI-MEM (VEGF partial β (SE) = 0.59 (0.56); FDR-adjusted p=0.593).
Although the eect size for the association of VEGF to ADNI-EF was larger in the A β+ stratum than in
theAβ-stratum,thisdierencewasnotsignicant(interaction p-value=0.22). ThesamplesizeoftheAβ-
stratumwassignicantlysmallerthantheA β+stratumandtherefore,wemaybeunderpoweredtodetect
a true interaction if one exists.
To further address limitations of domain specic cognitive measures and evaluate cognitive compo-
nentsdrivingtheassociationbetweenVEGFandexecutivefunctioninAβ+participants,weevaluatedthe
relationshipbetweenVEGFandeachoftheneuropsychologicalteststhatcomprisetheADNI-EFcompos-
itescoreinAβparticipants. HigherVEGFlevelswereassociatedwithbetterCategoryFluency–vegetables
(VEGFpartial β(SE)=1.90(0.52);p< 0.001)andCategoryFluency–animalsscores(VEGFpartial β(SE)
= 1.28 (0.53); p=0.018). No other ADNI-EF subtests were associated with VEGF levels (Supplementary
Table S3.3).
3.4.5 Mediationanalysis
WenextevaluatedwhethertheassociationbetweenVEGFandADNI-EFwasdirectormediatedbyregional
FDG-PET signal. We focused only on those regions in which we found a signicant relationship between
72
VEGF and FDG-PET signal: mean bilateral ITG, MTG, and inferior parietal cortex (Table 3.5). We found
that FDG-PET signal in all three regions mediated the relationship between VEGF and ADNI-EF.
Table 3.5: RegionalmediationeectsinAβ+participants(n=116)
Inferior Temporal Gyrus Middle Temporal Gyrus Inferior Parietal Cortex
Mediational Process β 95% CI p-value β 95% CI p-value β 95% CI p-value
CSF VEGF! FDG-PET SUVR ROI! ADNI-EF
Indirect eect 0.54 (0.14, 1.08) 0.006* 0.35 (0.05, 0.78) 0.016* 0.59 (0.14, 1.22) 0.010*
Direct eect 1.11 (-0.03, 2.16) 0.060 1.25 (0.16, 2.41) 0.026* 1.07 (0.07, 2.12) 0.036*
Total eect 1.65 (0.51, 2.73) 0.002* 1.60 (0.48, 2.80) 0.006* 1.66 (0.59, 2.84) 0.002*
Proportion Mediated 0.33 (0.09, 1.05) 0.008* 0.21 (0.03, 0.70) 0.022* 0.36 (0.10, 0.90) 0.012*
ROI = Region of Interest. *p< 0.05 signicance
3.5 Discussion
OurstudyevaluatedtherelationshipbetweenCSFVEGFlevelsandregionalglucosemetabolismandcorti-
calthicknessbyevaluatingwhetherrelationshipsweremodiedbyCSFA β,t-tau,andp-taulevels. Wealso
evaluated whether regional FDG-PET signal mediates an association between VEGF and cognition. Our
resultsrevealedthat1)VEGFhasregionally-specicstructuralandmetabolicassociationsinAD-signature
regions, particularly in lateral temporal-parietal cortices, 2) the association between VEGF and FDG-PET
ROIs was stronger in those with abnormal levels of AD neuropathology (CSF Aβ , t-tau and p-tau), 3) in
Aβ+ individuals, higher VEGF levels were associated with better scores on temporal-mediated language
measures (i.e., category uency) and 4) VEGF had both a direct relationship with cognitive function in
individuals along the AD continuum and an indirect relationship, mediated by regional FDG-PET signal.
In Aβ+ participants, higher VEGF levels were associated with greater mean bilateral glucose meta-
bolism in the inferior parietal cortex, and in the middle and inferior temporal gyri, which were also re-
gions that exhibited an interactive eect between VEGF and t-tau and p-tau. Previous work in the same
73
cohortthatonlyevaluatedtheassociationbetweenVEGFandglobalFDG-PETSUVR(butnotregionalin-
dices)foundasimilardirectionofassociation(Wangetal.,2018). Thisworkalsoalignswithotherstudies
demonstratingthatAD-relatedpathologyispreferentiallyassociatedwithregionaldecreasesinFDG-PET
SUVRandcorticalthicknessinthelateraltemporalregions(Dowlingetal.,2015;Malpasetal.,2018). The
temporal lobe may be more vulnerable to hypometabolism because it is preferentially impacted by early
Aβaccumulation(BraakandBraak,1991)andreducedVEGFexpression,asseeninpost-mortemADbrains
with high amyloid plaque load (Provias and Jeynes, 2014). However, the regions in which we found sig-
nicantassociationsdonotalignexclusivelywiththespatialtrajectoryofAD-associatedneuropathology.
Therefore,othervascularfactorsmaycontributetothespatialspecicityofVEGF,warrantingfurtherin-
vestigation. Forexample,lowercerebralbloodow(CBF)andhigherCSFsolubleplatelet-derivedgrowth
factor (i.e., measures of vascular function) have been recently associated with greater tau PET burden
in similar temporo-parietal regions to those found in our study, suggesting that these regions are par-
ticularly vulnerable to vascular interactions with AD neuropathology (Albrecht et al., 2020). Also, the
inferior parietal, middle and inferior temporal gyri comprise vascular territory watershed regions, which
are susceptible to hypoperfusion in AD (Huang et al., 2018). Future work should evaluate these potential
mechanisms driving the topographical correspondence of VEGF levels.
Our identied positive associations between VEGF levels and FDG-PET signal could also represent a
response to early blood brain barrier (BBB) damage, where VEGF levels increase to help restore BBB’s
glucosetransporter-1(GLUT1)levels(whichareresponsibleforcarryingFDGacrosstheBBBinFDG-PET
scanningofthebrain). Evidenceofthiscompensatoryrelationshiphasbeenseeninanimalmodels,where
GLUT1 levels decrease in response to a high fat diet, which results in a compensatory increase in VEGF
expressiontohelprestoreBBBvascularendothelialcellGLUT1levels,glucoseuptakeandcognitivefunc-
tion (Jais et al., 2016). Further, FDG-PET signal reductions could also be more representative of impaired
GLUT1-faciliated BBB transport (i.e., vascular dysfunction) than glucose metabolism, since FDG does not
74
getmetabolizedbydownstreammetabolicpathwaysandonlytracksglucosetransportshortlyafterbrain
uptake (Sweeney et al., 2019).
The direction and strength of the relationship between VEGF and AD brain biomarkers varied by
AD-neuropathological marker abnormality. Higher VEGF levels were associated with greater regional
FDG-PET SUVR and nominally related to a thicker cortex in Aβ+ participants, possibly representing a
neuroprotective eect of VEGF. Although the relationship between VEGF and cortical thickness has not
been evaluated previously, other measures of vascular dysfunction (e.g., capillary mean transit time, CSF
vascularadhesionmolecule-1,CSFintercellularadhesionmolecule-1)havebeenassociatedwithtemporo-
parietal cortical thinning and worse cognition (Janelidze et al., 2018; Nielsen et al., 2017), consistent with
Aβ+ groupndings. We also found a signicant interaction between VEGF and CSF t-tau and p-tau on
regional FDG-PET, where higher levels of tau strengthened the positive association between VEGF and
FDG-PET,evenafteradjustingforamyloid. Whilebotht-tauandp-tauhadinteractiveeectswithVEGFon
FDG-PETSUVRintheMTGandinferiorparietalcortex,p-taualsomodiedtheassociationbetweenVEGF
and FDG-PET in the ITG, superior parietal cortex and precuneus. This may reect more of an interaction
between vascular dysfunction and tau phosphorylation (i.e., p-tau) early in the AD cascade (particularly
in watershed regions that are vulnerable to vascular decits), rather than later after neurodegeneration
(reectedbyt-taulevels)hasbegun( Blennowetal.,2010). TheseresultsalsosuggestthatAβandtauhave
independenteectsontherelationshipbetweenVEGFandFDG-PET.However,asacross-sectionalstudy,
we cannot determine whether VEGF, amyloid, or tau is modifying the relationship with FDG-PET.
In Aβ- participants, VEGF levels were not related to regional FDG-PET SUVR and higher VEGF was
associatedwithathinnersuperiorparietalcortex. ThisunexpectedndingmaysuggestthathigherVEGF
levels (when amyloid levels are minimal) signal an injury response driven by something other than AD-
neuropathology. For instance, injury could result from dysfunctional vascular and metabolic processes
thatpromoteinammation,vesselpermeability,leakageandneuronalloss( VanDykenandLacoste,2018).
75
Variability in the directionality of thendings may also be related to possible interactive eects between
Aβand tau on VEGF levels. In our group analysis of VEGF levels, individuals with the lowest (Aβ- and
t-tau-) and highest (Aβ+ and t-tau+) neuropathological burden had signicantly higher VEGF levels than
those who were only Aβ+ (and t-tau-), which may reect a U-shaped trajectory of VEGF levels in AD
pathogenesis. However, as a cross-sectional study we cannot evaluate these mechanisms formally.
In our reproducibility analysis of the interaction between VEGF and CSF Aβ on cognitive domain
measures, we did notnd a signicant interaction between VEGF and A β on ADNI-EF and ADNI-MEM
compositemeasures-similartoHohmanetal. 2015. However,inourAβstratiedanalysis,wefoundthat
higher VEGF levels were associated with better ADNI-EF scores. We may have found this cross-sectional
association between VEGF and ADNI-EF because 1) dierences in the sample size of the A β stratums
may mask a formal interaction if a true one exists, and 2) we also adjusted for CSF t-tau levels, APOE4
genotype, and acquisition site in our statistical model. The executive function measure we used (ADNI-
EF)isaheterogeneousconstructthatactuallyincludesbothtestsofexecutivefunctionandlanguage. Our
subtestanalysisrevealedthattheassociationbetweenhigherVEGFlevelsandhigherADNI-EFwasdriven
by categoryuency (vegetables and animals) scores. Although categoryuency tasks have a generation
component, they are typically considered measures of language and semantic knowledge, not executive
function measures per se. Thisnding may reect changes in temporal lobe-mediated language function
(rather than frontal-mediated EF), and suggests that VEGF may have an eect on temporal lobe function
independent of tau (which we controlled for in our analysis). While few studies have correlated VEGF
with dierent cognitive measures, a multi-cohort study in AD and cognitively intact participants found
thatindividualswithbettercognitive(MMSE)scoreshadhigherVEGFlevels(Leungetal.,2015),asimilar
direction of eect we found in our study.
76
VEGFhadbothdirecteectsonADNI-EFscores,andalsoindirecteects,mediatedbyregionalFDG-
PET signal. Past studies have reported comparable ndings; regional temporal lobe FDG-PET hypo-
metabolismwasassociatedwithEFinAD(Habecketal.,2012)andalsomediatedtheassociationbetween
baseline AD CSF biomarkers and subsequent cognitive decline (Dowling et al., 2015). The regional speci-
city of FDG-PET on ADNI-EF may be driven by verbaluency, as the measure recruits both frontal and
temporal lobe brain regions (Melrose et al., 2009). Particularly, worse performance on category naming
tests and retrieval of semantic knowledge is associated with region-specic metabolic decreases in the
inferior temporal lobes (Melrose et al., 2009), a region we also identied to be associated with VEGF. Our
ndings could also be attributed to disruption of more global structural cortical connections, such as be-
tween the frontal and temporal lobes (Wiseman et al., 2018). The presence of both direct and indirect
associations suggests that multiple AD brain biomarkers and pathways contribute to cognitive impair-
ment.
Future work should evaluate the temporal and spatial correspondence between VEGF levels and AD
brainbiomarkers,andfurtherintegratecomplexmodelsandinteractionsinlargersamplesthatcanaddress
the multitude of factors (e.g., inammation, atherosclerosis) that modify VEGF signaling. Our study has
somelimitations. First,weincludedVEGFlevelsacquiredatonlyonetimepoint. VEGFlevelsmayuctuate
from the time of CSF collection to the time of the brain scans and neuropsychological testing. We also
cannotdirectlyevaluatewhetherVEGFisupregulatedandthereforecannotinferdirectionalityorcausality
of the relationships. Although we had a priori hypotheses regarding the temporal ordering of variables
used in the mediation analysis, future work would benet from longitudinal analysis to assess mediation
eects. Additionally, the spatial specicity of amyloid and VEGF in the brain is also unknown, since both
variableswereassessedbyaglobalCSFmeasure. SamplesizesafterstratifyingFDG-PETparticipantswere
also small, particularly in the Aβ- stratum, limiting our statistical power to formally detect interaction
eects. Since samples were also uneven in sex (females < males) and diagnostic groups (AD < CI < MCI),
77
future large-scale initiatives should aim to balance sex and diagnostic group sample sizes. Despite these
limitations,thisstudyisthelargeststudy-to-dateevaluatingregionalrelationshipsbetweenCSFVEGFand
AD brain biomarkers. Further, this study is an important step in evaluating mechanisms and populations
most at-risk for suboptimal brain aging.
Our study helps clarify the relationship between endogenous CSF VEGF and Aβin humans and ex-
pands on previously detected associations between VEGF and cognition (Hohman et al., 2015) and global
FDG-PET (Wang et al., 2018) by 1) identifying regional relationships between VEGF and FDG-PET and
cortical thickness – indices that are altered early in AD pathogenesis and provide spatial specicity to
heterogeneityinAD,and2)assessingwhethertheassociationbetweenVEGFandcognitionwasmediated
by FDG-PET signal. Our results suggest that at dierent amyloid and tau loads, there is variability in the
relationship between VEGF and regionally specic brain measures. Higher endogenous VEGF levels may
signalavascularinjuryresponsewhenAD-neuropathologicalloadislow,butserveacompensatoryneuro-
protectiverolewhenAD-neuropathologicalburdenishigh. VEGFmayalsoserveasavaluablebiomarker
in understanding vascular and metabolic contributions to variability in AD neuroimaging phenotypes.
Lastly, VEGF is modiable, such as through exercise ( Viboolvorakul and Patumraj, 2014), making it a po-
tentially therapeutic target in AD and warranting future investigation on its contribution to modulating
AD risk.
78
3.6 SupplementaryMaterial
Table S3.1: AssociationstoVEGFinallparticipants(N=310)andbyAβstratum
All Participants (N=310) Aβ+ (n=215) Aβ- (n=95)
β SE
Partial p-value
β SE
Partial p-value
β SE
Partial p-value
CSF Aβ 0.001 0.0001 <0.001* 0.0006 0.0003 0.048* 0.0009 0.0004 0.035*
Age 0.004 0.001 <0.001* 0.006 0.001 <0.001* 0.0002 0.002 0.92
Sex -0.06 0.01 <0.001* -0.06 0.02 <0.001* -0.06 0.02 0.004*
Years of Education -0.002 0.002 0.33 -0.002 0.002 0.44 -0.001 0.004 0.75
CSF t-tau 0.001 0.0001 <0.001* 0.001 0.0001 <0.001* 0.004 0.0005 <0.001*
APOE4 0.01 0.01 0.21 0.01 0.01 0.28 -0.021 0.03 0.55
MCI Diagnosis -0.06 0.01 <0.001* -0.06 0.02 0.001* -0.007 0.04 0.85
AD Diagnosis -0.01 0.01 0.28 -0.005 0.01 0.67 0.01 0.03 0.76
Fixed eects included VEGF, age, sex, years of education, and CSF t-tau,APOE4, and diagnosis. Random
eects included site. CI diagnosis was used as reference for diagnosis (as an ordered factor) in the linear
mixed eects model. SE = Standard Error. * indicatesp< 0.05 signicance.
79
Table S3.2: InteractionsbetweenVEGFlevelsandCSFt-tauandp-tauoncorticalthicknessROIs
(N=310)
VEGF x t-tau VEGF x p-tau
ROI β SE p-value FDR p-value β SE p-value FDR p-value
Superior Temporal Gyrus -0.0004 0.001 0.80 0.84 0.001 0.005 0.76 0.80
Middle Temporal Gyrus 0.002 0.001 0.15 0.48 0.007 0.030 0.15 0.44
Inferior Temporal Gyrus 0.001 0.001 0.47 0.70 0.002 0.005 0.70 0.80
Fusiform Gyrus -0.0004 0.001 0.80 0.84 0.001 0.005 0.76 0.80
Entorhinal Cortex 0.006 0.004 0.17 0.48 0.020 0.013 0.20 0.44
Superior Parietal Cortex 0.0 0.001 0.21 0.48 0.007 0.004 0.10 0.43
Inferior Parietal Cortex 0.003 0.001 0.045 0.41 0.007 0.004 0.09 0.43
Posterior Cingulate -0.0002 0.001 0.84 0.84 0.001 0.004 0.80 0.80
Precuneus 0.001 0.001 0.37 0.67 0.003 0.004 0.41 0.73
Each association was evaluated using a linear mixed eects model. Fixed eects included VEGF, age, sex,
yearsofeducation,andCSFamyloid, APOE4,anddiagnosis. Sitewasmodeledusingrandomeects. SE=
Standard Error; FDR = False Discovery Rate
80
Table S3.3: AssociationbetweenVEGFlevelstoADNI-EFsubtestsinAβ+participants(n=215)
Neuropsychological Test β SE VEGF partial p-value
Category Fluency – Animals 1.28 0.53 0.018
Category Fluency – Vegetables 1.90 0.52 <0.001
WAIS-R Digit Symbol 0.62 0.65 0.34
Digit Span Backwards 0.36 0.60 0.55
Trail Making Test Part A -1.03 0.75 0.17
Trail Making Test Part B -0.17 0.72 0.81
Clock Drawing - Symbol 0.09 1.55 0.95
Clock Drawing - Numbers 0.73 1.81 0.69
Clock Drawing - Time 0.26 1.59 0.87
Clock Drawing - Hands – – –
Clock Drawing - Circle – – –
Each model with a continuous dependent variable (neuropsychological test) was evaluated using a linear
mixed eects model. Each model with a binary dependent variable (neuropsychological test) was evalu-
ated using a logistic regression using a generalized linear mixed eects model with a logit link function.
Fixed eects included VEGF, age, sex, years of education, and CSF t-tau,APOE4, and diagnosis. Site was
modeled using random eects. CD Hands and CD Circle do not have reported eects, as ceiling eects
were observed. Clock drawing WAIS-R = Weschler Adult Intelligence Scale-revised. SE = Standard Error.
81
***
NS.
***
2.50
2.75
3.00
Amyloid−
Tau−
n=86
Amyloid+
Tau−
n=91
Amyloid+
Tau+
n=124
Amyloid−
Tau+
n=9
CSF VEGF
(LOG TRANSFORMED)
FigureS3.1: ViolinplotsofgroupdierencesinCSFVEGFlevelsbyamyloidandt-taupositivity.
Individuals who were both Aβ- and t-tau- (mean ± SD = 2.73 ± 0.12) had signicantly higher VEGF levels
than individuals who were Aβ+ and t-tau- (mean ± SD = 2.63 ± 0.12; t(175.85) = 5.28,p < 0.001).
IndividualswhowereAβ+andt-tau-hadsignicantlylowerVEGFlevelsthanindividualswhowereboth
Aβ+andtau+(mean ±SD=2.73 ±0.12;t(188.1) = 6.24,p< 0.001). NosignicantdierencesinVEGF
levels were found between the Aβ- t-tau- and Aβ+ t-tau+ groups (t(179.97) = 0.44,p=0.66). Group
dierenceswere notevaluated inthe A β-tau+ group(mean ± SD= 2.88 ± 0.09)since thesample sizewas
small (n=9)
82
FigureS3.2:PiecewiselinearmixedeectssplineregressionbetweenCSFAβlevels(withaknotat
192pg/mL),andVEGF(logtransformed)levels,whichdemonstratesthatevenatamyloidlevels
thatareconsidered“normal”intheAβ-group,VEGFmaybeinhibitedwithincreasingamyloid
load(i.e.,lowerCSFAβlevels). WhilethemagnitudeoftheslopeofcontinuousAβlevelsinthosehaving
Aβ levels< 192pg/mL (i.e., Aβ+ participants) to VEGF was smaller than the slope of Aβ in those having
Aβ levels> 192pg/mL (i.e., Aβ- participants) to VEGF, the slope estimates did not signicantly dier ( β
(SE) = 0.0004 (0.0005); p=0.440). Red color indicates CSF Aβ+ status, while grey color indicates Aβ-
status
83
Chapter4
Whitematterhyperintensityvolumemodifytheassociationbetween
CSFvascularbiomarkersandregionalFDG-PETalongtheAlzheimer’s
diseasecontinuum
4.1 Abstract
InthepresenceofAlzheimer’sdisease(AD)neuropathology,highercerebralspinaluid(CSF)vascularen-
dothelialgrowthfactor(VEGF;amulti-functionvascularpermeabilityproteinoftenupregulatedbyinam-
mation)isassociatedwithhigherFDG-PETsignal. However,itisnotyetknownwhetherthisassociation
is specic to VEGF or broadly driven by vascular inammation and whether white matter hyperintensity
(WMH)volume,amarkerofvasculardysfunction,modiesthisrelationship. Further,itremainsunknown
whethertheseinteractionsvarybyamyloidbetalevel,APOE4genotype,andsex. Toaddressthisquestion,
we evaluated whether ve vascular inammatory (VI) CSF biomarkers were related to FDG-PET signal
and whether white matter hyperintensity (WMH) volume, a marker of cerebrovascular disease, modied
theseassociationin158Alzheimer’sDiseaseNeuroimagingInitiative(ADNI)participants(55-90yearsold,
27 cognitively normal, 76 with mild cognitive impairment, 37 with AD). Since the link between vascular
inammationandFDG-PETmaybedrivenbydysfunctionwithinthecerebralcirculation,ratherthancor-
tical gyral and sulcal boundaries, we dened regions by vascular territories: the anterior cerebral artery
84
(ACA), middle cerebral artery (MCA), posterior cerebral artery (PCA). We tested whether these associa-
tionsbetweenCSFVIbiomarkersandregionalFDG-PETsignaldieredbyA β-positivity,APOE4genotype,
and sex. We hypothesized that CSF VI biomarkers would be associated with regional FDG-PET and this
relationship would be modied by WMH volume. We found that CSF VEGF and CRP were associated
withFDG-PETsignalintheMCAandPCAregions. GreaterWMHburdenmaskedrelationshipsbetween
CSF biomarkers and FDG-PET, particularly in the ACA territory. Associations were driven by Aβ+ par-
ticipants and APOE4 carriers suggesting that vascular pathology (e.g., cerebral amyloid angiopathy) may
alter immunometabolism relationships in those most at risk for AD. Our ndings were also specic to
males, suggesting that sex is an essential biological variable to consider in unveiling mechanisms and
heterogeneity in AD brain phenotypes.
4.2 Introduction
Alzheimer’s disease (AD) is a multiplex condition characterized by neurotoxic amyloid and tau accumu-
lation, neurodegeneration, and gradual cognitive decline. Individuals with AD and in those at-risk for
cognitive decline, demonstrate early and substantial regional reductions in cerebral uorodeoxyglucose
(FDG)-PET and bioenergetic capacity, which can vary by sex (Mosconi, 2005; Hu et al., 2013; Mosconi,
2013; Mosconi et al., 2017; Marchitelli et al., 2018b). While early FDG-PET signal changes are a hallmark
pathophysiologicalfeatureofAD,heterogeneityinFDG-PETsignalandclinicalmanifestationcontinueto
hindertreatmenteorts(Anchisietal.,2005;Levinetal.,2021). SourcesofvariabilitymayincludetheFDG-
PET signal itself, as it is an indirect measure of cerebral metabolic rate of glucose that can be inuenced
by synaptic function, energy metabolism, blood brain barrier permeability, and vascular and microglial
related inammation ( Jiang and Cadenas, 2014; Chiaravalloti et al., 2016; Yin et al., 2016; Sweeney et al.,
2019; Choi et al., 2021; Xiang et al., 2021). Vascular inammation may disturb neurovascular support for
85
synaptic energy demands in AD by promoting the pathological formation of leaky and abnormally struc-
turedbloodvessels(Govindpanietal.,2019),heightenedpermeabilityofcerebralbloodvessels,migration
of immune cells into the brain parenchyma, and vascular remodeling that perturbs CBF in the capillaries
(Ashby and Mack, 2021). Vascular inammation has been increasingly recognized as a risk factor for AD
and may also contribute to FDG-PET signal changes (Wilkins et al., 2014), but evidence remains scarce
on how circulating biomarkers relate to FDG-PET signal. Thus, investigating how vascular inammatory
markersrelatetoFDG-PETmayhelpunravelvascularsourcesofheterogeneityinFDG-PETsignalinAD
topromotedeeperphenotypingofdiseasetobettertargetinterventionstrategiesforthosemostatriskfor
cognitive decline.
Cerebrospinaluid (CSF) soluble biomarkers molecules with an active role in vascular inammatory
processesandinthepathogenesisofADmayprovidebiomolecularinsightintomechanismsdrivinglower
FDG-PET signal early in AD progression. We recently demonstrated that in individuals with abnormal
levelsofADneuropathology,higherCSFvascularendothelialgrowthfactor(VEGF)levelswereassociated
with greater FDG-PET signal in regions susceptible to AD-cortical thinning (Tubi et al., 2021). However,
fewotherstudieshaveexaminedtheassociationbetweenCSFmarkersofvascularinammationandFDG-
PET in AD samples, despite mounting evidence that 1) vascular risk factors (e.g., hypertension, white
matter hyperintensities) are independently associated with FDG-PET(Lo and Jagust, 2012; Jiaerkenet al.,
2019; Zhou et al., 2020), 2) vascular dysfunction and inammation are reected components of FDG-PET
signal interpretation (Sweeney et al., 2019; Xiang et al., 2021), and 3) vascular inammatory markers are
associatedwithADneuropathology(Janelidzeetal.,2018;Pillaietal.,2019). Forexample,inanAlzheimer’s
DiseaseNeuroimagingInitiative(ADNI)datasetofparticipantswithmildcognitiveimpairment(MCI),the
CSFbiomarkersthatcorrelatedwithCSFamyloidlevelsandhadthelargesteectsizewerethoserelating
to vascular inammation: brinogen, VEGF, von Willebrand Factor (vWF), C-reactive protein (CRP); and
vascular cell adhesion molecule 1 (VCAM1) (Pillai et al., 2019). To address this, we investigated whether
86
the CSF vascular inammatory markers that were linked to early amyloid accumulation were associated
with regional FDG-PET signal.
TheassociationofCSFvascularinammatorymarkerswithFDG-PETmaybemodiedbyage-related
cerebrovasculardisease,whichcanbeindirectlyindexedbywhitematterhyperintensity(WMH)volume.
WMHsappearashyperintensesignalonT2-weightedMRIandareaconsequenceofcerebralsmallvessel
disease(SVD).WhilepreciseetiologicaloriginsofWMHsareunclear,theymostcommonlyreectpathol-
ogyresultingfromchroniccerebralischemiathatpromotesdemyelination,axonalloss,bloodbrainbarrier
(BBB) breakdown, and microglial and endothelial activation (Fernando et al., 2004; Wardlaw et al., 2015).
Higher WMH volumes are associated with lower FDG-PET signal (No et al., 2019; Gaubert et al., 2021)
and greater carotid (De Leeuw et al., 2000; Pico et al., 2002; Ammirati et al., 2017; Zhang et al., 2021) and
intracranialatherosclerosis(Leeetal.,2011;Parketal.,2015;Nametal.,2017). Importantly,atherosclerosis
can induce changes in vascular protein expression (e.g., VEGF) by modifying shear stress dynamics (i.e.,
the frictional force against the vessel wall) resulting in alterations in CBF, endothelial gene expression,
and activation of downstream oxidative and inammatory pathways ( Cunningham and Gotlieb, 2005).
Moreover, vascular disease may induce variability in vascular protein levels and, consequently, modify
theircorrelativeassociationwithFDG-PET.NostudieshaveexaminedtheinteractionbetweenCSFmark-
ersandWMHvolumeonFDG-PET,butWMHinteractswithADbrainandneuropathologicalmarkerson
cognitivefunction,althoughthedirectionofeectsremainunclear. WhilesomestudiessuggestthatWMH
volume synergistically interacts with AD-related brain atrophy to accelerate cognitive decline (Brickman
etal.,2008),othersdemonstratethattheassociationbetweenWMHsandcognitionandsymptomonsetis
mostapparentinthosewithminimalornoADpathology(Soldanetal.,2020;Tubietal.,2020). Variability
in WMH interaction eects may also bedrivenbyAPOE4genotype, as APOE4carriers havelarger WMH
volumes and faster accumulation than non-carriers (Brickman et al., 2014; Sudre et al., 2017; Rojas et al.,
87
2018). APOE4 also modulates various cerebrovascular cell types (e.g., astrocytes, pericytes brain endothe-
lial cells) and predisposes the vascular system to damage, resulting in dysfunction of CBF, neurovascular
uncoupling,BBBpermeability,andcerebralamyloidangiopathy(CAA)(Taietal.,2016). Sexmayalsocon-
tribute dierently to these relationships, as WMH volumes increases the risk for AD disease progression
in males, but not females (Kim et al., 2015; Burke et al., 2019; Salminen et al., 2022). Given that vascular
dysfunction may interact with or even be exacerbated by amyloid-beta (Aβ) levels, APOE4 genotype, and
sex in AD pathogenesis, we performed subsequent analyses in our study stratied by these risk factors.
The association between CSF biomarkers and regional FDG-PET signal may vary by cerebral circu-
lation boundaries, as the distribution of vascular disease varies across vascular territories. For example,
atherosclerosisismostcommonlyidentiedintheanteriorcirculation,whichincludestheanterior(ACA)
andmiddlecerebralarteries(MCA)fedprimarilybytheinternalcarotidarteries,whiletheposteriorcere-
bralarteries(PCA)arepartoftheposteriorcirculationfedprimarilybythevertebralarteries(Roheretal.,
2011; Denswil et al., 2016). Higher levels of subclinical atherosclerosis have been associated with lower
whole brain FDG-PET signal (Cortes-Canteli et al., 2021), but regional associations between markers of
vascularfunctionandFDG-PEThaveyettobemapped. Giventheregionalvulnerabilityofatherosclerosis
deposition, brain regions supplied by the anterior circulation may be more vulnerable to disturbances of
vascular function and result in downstream disruptions in FDG-PET signal. As a result, it remains essen-
tialtoassesswhethertheinteractiverelationshipsbetweenvascularinammatorybiomarkersandWMH
volume on FDG-PET signal are strongest in regions dened by cerebral circulation, which have dierent
vulnerabilities to vascular disease.
To address these gaps, we investigated whether CSF levels of 5 vascular inammatory biomarkers
molecules most associated with amyloid levels in an ADNI cohort (VCAM, VEGF, CRP,brinogen, vWF)
(Pillaietal.,2019)relatetoFDG-PETsignal. WealsoassessedwhetherhigherWMHvolume,ameasureof
88
cerebrovascularrisk,modulatestherelationshipsbetweenthesevascularinammatoryproteinsandFDG-
PETsignalandwhethertheserelationshipswerebestdemonstratedinconventionallydenedanatomical
regions of interestor regionsdenedbyvascularterritory. Wealsostudiedwhetherthestrengthofthese
eectsvariedbyCSFamyloid-beta(A β)levels,APOE4genotype,andsex. Wehypothesizedthattheinter-
actionbetweenCSFvascularinammatorybiomarkersandWMHvolumeonregionalFDG-PETwouldbe
most apparent in regions most vulnerable to vascular disease (i.e., anterior circulation) and in those most
at risk for AD (Aβ+, APOE4+). Collectively, our study aims to provide more regional specicity to how
vascular dysfunction contributes to variability in FDG-PET signal to better understand heterogeneity in
risk and progression of brain aging along the AD continuum.
4.3 Methods
4.3.1 TheAlzheimer’sDiseaseNeuroimagingInitiative(ADNI)
Data used in the preparation of this article 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 part-
nership,ledbyPrincipalInvestigatorMichaelW.Weiner,MD.TheprimarygoalofADNIhasbeentotest
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. All data utilized in this study were acquired from the publicly available ADNI Image
Data Archive (https://ida.loni.usc.edu).
4.3.2 Participants
Ourstudyselectedthe158ADNIparticipants(75.1 ±6.8yearsold,agerange: 57to88yearsold,53females,
105males)whohadacerebrospinaluid(CSF)assay,1.5TMRIacquiredwithin1yearofabaseline,anda
89
FDG-PETscan. ParticipantshadalumbarpuncturetoacquireaCSFassaywithamean±standarddeviation
(SD) of -0.49 ± 24.48 days from the FDG-PET scan. Participants had a 1.5T magnetization prepared rapid
gradient echo (MP-RAGE) structural MRI scan acquired with a mean ± SD -29.99 ± 26.76 days from the
FDG-PET scan. Participants also underwent neuropsychological testing, medical history evaluation, and
blood draws at their baseline visit.
In our selected cohort of 158 participants, 39 were cognitively normal (CN), 80 had evidence of mild
cognitiveimpairment(MCI),and39werediagnosedwithprobableAD.AprobableADdiagnosisoccurred
whentheparticipanthadeitheramemorycomplaintorobjectivememorydysfunction. Objectivememory
dysfunctionwasassessedinaseriesofneuropsychologicalbatteries,whichincludedtheWechslerMemory
Scale-Revised (WMS-R) - Logical Memory II (LM-II) R (WMS-R LM-II) using an education-adjustment,
Mini-MentalStateExam(MMSE;scorebetween20-26,inclusive),ClinicalDementiaRatingscale(CDR; 0.5), or the NINDS/ADRDA criteria for probable AD (McKhann et al., 1984) MCI diagnosis was classied
whenamemorycomplaintwaspresent,butthecriteriafordementiawasnotmetorwhenparticipantshad
MMSEscore between24-30 (inclusive)oraCDRscoreof0.5(withCDRmemoryboxscore0.5orhigher).
ParticipantsweredeemedCNwhentheydidnotmeettheabovecriteriaforprobableADorMCIandhad
no memory complaints. Further information on ADNI’s participant selection and diagnostic criteria are
outlinedintheADNIprotocol(http://www.adni-info.org/Scientists/AboutADNI.aspx). ADNIalsoexcluded
individuals with any serious neurological or neuropsychiatric condition, or history of brain injury. The
demographic information of the cohort is outlined in Table 4.1.
4.3.3 ImagingAcquisition
ThefastingFDG-PETscanwasperformedusingastandardprotocol,whichincludedsix5-minuteframes
collected30minutesaftera5mCi[18F]-FDGinjection(Jagustetal.,2010). Allparticipantsalsounderwent
a1.5TMRIscantoacquireasagittalT1-weighted3DMagnetizationPreparedRapidAcquisitionGradient
90
Table 4.1: CohortCharacteristics(n=158)
Age 75.1 ± 6.8
Sex
Male 105 (66.5%)
Female 53 (33.5%)
Education (years completed) 15.8 ± 3.1
Diagnosis
Cognitively Normal (CN) 39 (24.7%)
Mild Cognitive Impairment (MCI) 80 (50.6%)
Alzheimer’s disease (AD) 39 (24.7%)
Ethnicity
Non-Hispanic or Latino 155 (98.1%)
Hispanic or Latino 2 (0.013%)
Unknown 1 (0.006%)
APOE4 carrier status
0 Alleles 78 (49.4%)
1 Allele 61 (38.6%)
2 Alleles 19 (12.0%)
CSF Biomarkers (natural log of pg/mL)
VEGF 2.7 ± 0.1
CRP -2.9 ± 0.6
Fibrinogen -3.4 ± 0.4
VCAM1 1.2 ± 0.1
von Willebrand Factor (vWF) -1.4 ± 0.2
AD Neuropathology Biomarkers
Aβ(pg/mL) 166.4 ± 53.9
Aβ- 42 (26.6%)
Aβ+ 116 (73.4%)
Total tau (pg/mL) 99.8 ± 54.0
Tau- 90 (57.0%)
Tau+ 68 (43.0%)
Continuousvariablesshownasmean ±standarddeviation. Categoricalvariablesshownasfrequency(%).
Echo (MPRAGE) structural MRI scan. ADNI MRI scan protocols were coordinated across sites for opti-
mal comparability according to ADNI published protocols (http://adni.loni.usc.edu/methods/documents/
mri-protocols/) and have passed quality control (QC) assessments (Wyman et al., 2013).
91
4.3.4 WhiteMatterHyperintensity(WMH)Segmentation
Participant WMH volumes (n=156) were downloaded from ADNI (https://ida.loni.usc.edu). Briey, the
methodderivesWMHsegmentationsusingaprobabilisticmodelusingspatialandcontextualpriorsbased
onprotondensity(PD),T1,andT2intensities(DeCarlietal.,1995). ThealgorithmusesaBayesianproba-
bilistic method to generate likelihood estimate values for WMH at each voxel in the white matter. These
likelihoods are thresholded at three standard deviations above the mean to construct the binary WMH
mask.
4.3.5 FDG-PETProcessing
Pre-processed FDG-PET scans were downloaded from the LONI IDA website (https://ida.loni.usc.edu).
Todiminishinter-scannereects(Joshietal.,2009)andenabledatapoolingoftheFDG-PETscans,ADNI
co-registered and averaged the frames, and spatially aligned (via interpolation) scans to a standard voxel
size with a uniform resolution, according to standardized protocols.
Conventionally,regionsofinterest(ROI)inneuroimagingstudiesareautomaticallydenedbyatlases
that use templates derived from cortical gyral and sulcal boundaries, such as the Desikan-Killany (DK)
Atlas (Desikan et al., 2006). While this method is valuable, as it provides 34 delineated regions that map
onto previously detected anatomical-functional mapping studies, a vascular territory template may more
closely align with vascular eects of interest. Specically, variability in atherosclerosis distribution may
contribute to the regional specicity of eects. For example, atherosclerosis is most commonly identied
intheanteriorcirculation,whichincludestheACAandMCAfedprimarilybytheinternalcarotidarteries,
while the PCA is part of the posterior circulation fed primarily by the vertebral arteries. Therefore, we
dened regions by both DK-atlas ROIs and vascular territories.
Wecalculatedthestandarduptakevalueratio(SUVR)intheapproximatevascularterritoryROIs(ACA,
MCA, and PCA). Vascular territory ROIs were derived from an MNI template developed by (Schirmer
92
et al., 2019) (more information detailed here: https://zenodo.org/record/3379848#.YXVT6dlBx58). Briey,
theycreatedthetemplatefrom16(meanage ±SD=69.6 ±8.2years,37.5%female)non-demented,stroke-
free participants, 12 of whom had sporadic cerebral amyloid angiopathy (CAA). An expert neurologist
outlined each of the vascular territories using a T1-weighted atlas image. We binarized each vascular
territoryROIandmultipliedeachROIbyaFreeSurfer-derivedcorticalgraymatter(GM)ribbon,asshown
in Figure 4.1.
The vascular territory MNI template was co-registered to each participant’s FDG-PET scan. To opti-
mize registration, through a series of linear and non-linear transformations, the MNI template wasrst
registered to a study-specic mean deformation template (MDT), then registered to participant-specic
T1-space,andnallytotheparticipants’FDG-PETnativespace,whereanalyseswereperformed. Wealso
extracted all 34 FreeSurfer 5.3 DK atlas regions and performed the same series of linear and non-linear
registrations to extract FDG-PET ROIs in gray matter cortical regions of interest. All registrations passed
visual QC inspection to ensure accurate alignment for co-registered images.
The FDG-PET reference region was dened as the upper half of the pons and cerebellar vermis on
correspondingslices,asrecommendedforcross-sectionalFDG-PETanalyses(Landauetal.,2012). Detailed
informationontheFDG-PETreferenceregionandco-registrationstepscanbefoundin(Tubietal.,2021).
Thestandardizeduptakevalueratio(SUVR)foreachFDG-PETROIwascalculatedbyobtainingthemean
value of that region and dividing by the mean signal in the pons-cerebellar vermis reference region.
4.3.6 CSFAnalytes
All CSF analyte measurements were acquired, processed, and quality assessed by ADNI. CSF QC values
were accessed through ADNI’s publicly available data repository (http://adni.loni.usc.edu). CSF sam-
ples wereacquiredatthebaselinevisitandwerebatchprocessedusingLuminexMulti-AnalyteProling
93
Figure 4.1: Participant specic example of the FDG-PET Processing pipeline to extract vascular
territoryROIsinFDG-PETspace. GM=GrayMatter,ACA=AnteriorCerebralArtery,MCA=Middle
Cerebral Artery, PCA = Posterior Cerebral Artery, SUVR = Standard Uptake Value Ratio.
(xMAP)immunoassaytechnology(MyriadRulesBasedMedicine;RBM,Austin,TX),usingstandardADNI
protocols (Spellman et al., 2015).
We a-priori selected for CSF vascular inammatory biomarkers that were most related to amyloid
levels in the ADNI cohort to better detect associations relevant to AD-pathogenesis; VCAM, VEGF, CRP,
brinogen, vWF) ( Pillai et al., 2019). CSF protein biomarker values were derived from ADNI Biomarkers
ConsortiumCSFQCMultiplexdatasheet,whichwerelogtransformedbyADNI(BoxandCox,1964)and
available online through ADNI’s publicly available data repository (htttp://adni.loni.usc.edu). At the
participant’sbaselinevisitafteranovernightfast,CSFsampleswereacquiredtheprocessed,aliquoted,and
storedat-80°CinalignmentwithADNIBiomarkerCoreLaboratoryStandardOperatingProcedures(http:
//adni.loni.usc.edu/wp-content/uploads/2010/09/CSF_Biomarker_Test_Instr.pdf). Further QC procedures
performed by ADNI on the CSF proteomic assays included test/retest on a subsample of 16 CSF samples
94
to test for the precision, cross-reactivity, spike-recovery (accuracy), and reliability. ADNI CSF Aβ and
total-tau quality-checked median scaled analyte measurements were selected for this study.
Amyloid positivity was dened by cut-os outlined in prior work; A β+ individuals had CSF Aβ42
< 192pg/mL and Aβ- participants had CSF Aβ42 levels 192pg/mL , t-tau+ individuals had CSF t-tau
93pg/mL, and t-tau- individuals had CSF t-tau< 93pg/mL (Shaw et al., 2009b). Detailed descriptions
of QC procedures for CSF Aβ42, CSF p-tau, and t-tau can be found elsewhere (Shaw et al., 2009b, 2011)
4.3.7 APOE Genotyping
APOE genotyping was performed by Cogenics from a 3mL aliquot of blood. PCR amplication and Hhal
restrictionenzymedigestionwasperformed,whichwasthenresolvedon4%MetaphorGelandviewedby
ethidiumbromidestaining(Saykinetal.,2010). MoreinformationontheADNIAPOEgenotypingprotocol
can be found here: http://adni.loni.usc.edu/methods.
Becauseofpossibledose-dependenteectsofAPOE4onvascularfunctionandMRIbiomarkers(Hobel
et al., 2019; Makkar et al., 2020), we covaried for APOE4 genotype count, which was coded in a dose-
dependent manner (0 for APOE4 non-carriers, 1 for carriers of one APOE4 allele, and 2 for carriers of two
APOE4 alleles) in our regression models. In our APOE stratied analysis (in section4.4.4.2 APOE4-carrier
dierences),westratiedthecohortbycarriers(either1or2alleles)andnon-carriers(0alleles)tooptimize
group sample size distribution and to retain power.
4.3.8 StatisticalAnalyses
WeconductedallstatisticalanalysesinRStudioversion1.0.136(http://www.R-project.org/)(Universityof
Auckland, Auckland, New Zealand) (R Core Development Team, 2010).
95
4.3.8.1 VascularCSFbiomarkersrelationshiptoFDG-PETSUVRinvascularterritories
First, we evaluated which vascular-inammatory CSF biomarkers were related to whole brain FDG-PET
SUVR. We evaluated CSF VEGF,brinogen, CRP, VCAM1, and vWF. For all statistical analyses we used a
linear mixed eects model to include the random eect of neuroimaging acquisition site. In all models in
this analysis, we also covaried for age, sex, cognitive diagnosis, APOE4 genotype, and CSF t-tau.
For any CSF biomarkers that demonstrated a signicant or trend level (p< 0.10) association with
whole-brainFDG-PETSUVRpriortoFDRcorrection,weperformedfurtheranalysestoinvestigateifthese
eectswereregion-specic(ACA,MCA,PCAterritories). Weusedthefalsediscoveryrate(FDR)tocorrect
for multiple comparisons (ROIs).
4.3.8.2 WMHinteractionwithCSFbiomarkersonregionalFDG-PETSUVR
Forallinteractionanalyses,wemean-centeredthecontinuousmeasuresofCSFCRP,CSFVEGF,andWMH
volumes for easier interpretation of interaction eects. In separate linear mixed eects models, we eval-
uated the interaction between VEGF and CRP (natural log transformed) and WMH volume on FDG-PET
SUVR in gray matter (GM) vascular territory ROIs. In all models for this analysis, we covaried for age,
sex,cognitivediagnosis,APOE4genotype,CSFt-tau,andintracranialvolume(ICV)andtherandomeect
of acquisition site. In a sensitivity analysis to evaluate whether the results were driven by WMH volume
outliers, we also used a robust linear mixed model regressed (rlmer in R) to evaluate the associations de-
scribed above. Since similar interaction beta estimates were found between VEGF and WMH volume on
FDG-PETROIs,wereportallresultsinthestudyusingastandardlinearmixedmodelregression(lmer in
R).
96
4.3.8.3 GroupdierencesintheWMHinteractionwithCSFbiomarkersonFDG-PETSUVR
TofurtherinvestigatewhetherthendingswerespecictothoseontheAD-continuum,westratiedthe
cohort by Aβ positivity (Aβ+ < 192 mg/dL) and performed the analyses as described above. The Aβ posi-
tivitycohortstraticationwasunequalinsamplesize(resultinginlimitedinterpretabilityofresultsdueto
powerissues). However,approximatelyequalsamplessizeswereavailableforAPOE4carrierstatus(APOE4
carrier n = 80, APOE4 non-carriers = 76) to address whether the interaction between CSF biomarkers and
FDG-PET diered by AD genetic risk. In separate analyses, we stratied the group byAPOE genotype
(carriers vs. non-carriers), and by sex (males vs. females) and performed the same analyses to identify
whether thesendings were population specic. For the amyloid stratied analyses, we covaried for age,
sex,cognitivediagnosis,APOE4genotype,CSFt-tau,andintracranialvolume(ICV)andtherandomeect
of acquisition site. For the sex stratied analyses, we covaried for age, cognitive diagnosis,APOE4 geno-
type, CSF t-tau, and intracranial volume (ICV) and the random eect of acquisition site. For theAPOE4
genotypeanalyses,wecovariedforage,sex,cognitivediagnosis,CSFt-tau,andintracranialvolume(ICV)
and the random eect of acquisition site. We also evaluated whether amyloid level contributed toAPOE4
resultsandrantheanalysiswithandwithoutamyloidlevels;betaestimatesweresimilarbetweenanalyses,
so we only reported the analysis without amyloid as a covariate.
4.3.8.4 VascularCSFbiomarkersrelationshiptoFDG-PETSUVRincorticallydenedROIs
Inseparatelinearmixedeectsmodels,weevaluatedthelinearassociationbetweenVEGFandCRP(nat-
ural log transformed) on FDG-PET SUVR in all 34 cortically dened ROIs. We also separately ran linear
mixed models to assess interaction eects between CSF biomarkers (VEGF, CRP) and WMH volume on
FDG-PETSUVRallcorticalROIs. WeusedtheFDRtocorrectformultiplecomparisons(ROIs)ineachCSF
biomarker analysis.
97
4.4 Results
4.4.1 VascularCSFbiomarkersrelationshiptowholebrainFDG-PETSUVR
WeevaluatedwhichvascularCSFbiomarkers(VEGF,CRP,brinogen,VCAM,vWF)wereassociatedwith
meangraymatterFDG-PETSUVR,andcovariedforage,sex,diagnosis,CSFt-tau,APOE4,andsite(random
eect). Prior to FDR correction, higher CSF VEGF levels (partialβ (SE)= 0.13 (0.07),p=0.056) and lower
CRP levels (partial β (SE) = -0.03 (0.01), p=0.054) were nominally associated with higher whole brain
meanFDG-PETSUVR.However,neitherCSFanalyte(VEGF,CRP)passedFDRcorrectionwhenaccounting
forallbiomarkers(FDRp-values=0.14). CSFbrinogen,VCAM,andvWFwerenotassociatedwithFDG-
PET SUVR (p-values > 0.62). To evaluate whether VEGF and CRP were correlated, we ran a correlation
analysisandfoundthatthesetwobiomarkersshowedalowlevelofcorrelationbetweeneachotherusing
the Pearson’s correlation coecient (r=0.067).
4.4.2 DierencesinvascularterritoryFDG-PETSUVR
To evaluate whether there were any baseline dierences in FDG-PET SUVR between vascular territory
regions, we compared mean values in each region using Welch’s Two Sample t-test. The FDG-PET SUVR
in the ACA (mean ± SD: 1.32 ± 0.13;p< 0.001) was signicantly higher than the MCA (mean± SD: 1.27
± 0.11) and the PCA (mean ± SD: 1.29 ± 0.11; p=0.019). No signicant dierence between MCA and
PCA FDG-PET SUVR was found (p=0.12). FDG-PET SUVR vascular territory dierences are illustrated
in Figure 4.2.
4.4.3 VEGFandCRPassociationwithvascularterritoryFDG-PETSUVR
Next,wefurtherinvestigatedwhetherthe2biomarkers(VEGF,CRP)thathadanominalassociationwith
whole brain FDG-PET SUVR had region-specic FDG-PET SUVR associations in any vascular territory
region. We adjusted for age, sex, diagnosis, t-tau, APOE4 genotype, and site (random eect). After FDR
98
Figure 4.2: FDG-PET SUVR mean dierences by GM vascular territory ROI (ACA, MCA, PCA).
ACA=AnteriorCerebralArtery,MCA=MiddleCerebralArtery,PCA=PosteriorCerebralArtery,SUVR
= Standard Uptake Value Ratio. * indicatesp< 0.05. NS. = not signicant.
correction, higher VEGF levels were signicantly associated with greater FDG-PET SUVR in the MCA
(VEGF β (SE) = 0.17 (0.07), p = 0.017) and a trend level positive association in the PCA territory (VEGF β
(SE) = 0.12 (0.07), p=0.062). After FDR correction, higher CRP levels were associated with lower FDG-
PET SUVR in the PCA (CRP β (SE) = -0.03 (0.01), p=0.016), with a trend level negative association in
the MCA territory (CRP β (SE) = -0.03 (0.01), p=0.056). Neither VEGF nor CRP were associated with
FDG-PET SUVR in the ACA territory (p-values > 0.20). The summary statistics between CSF biomarkers
(VEGF, CRP) on FDG-PET in each vascular territory can be found in Supplementary Table S4.1.
99
4.4.4 WMHinteractionwithCSFbiomarkersonFDG-PETSUVR
BecausevasculardiseasemayinducevariabilityinvascularproteinlevelsaswellasFDG-PETsignal,wein-
vestigatedwhetherWMHvolumemodulatestherelationshipbetweenCSFbiomarkersandregionalFDG-
PET SUVR for the two CSF analytes that were associated with FDG-PET SUVR. We performed separate
interaction analyses between the two CSF biomarkers (VEGF, CRP) and WMH volume on each FDG-PET
ROI.HigherWMHvolumeloadinteractedwithVEGF(allROIp-values<0.009)andCRP(allROIp-values
< 0.044) on FDG-PET SUVR in all vascular territory ROIs. In those with low WMH volumes, higher CSF
VEGFwasassociatedwithhigherFDG-PETSUVR,butasWMHloadincreasedtherelationshipwasatten-
uatedandbecamenegativeinthosewiththehighestWMHload(>2SDofmean). InthosewithlowWMH
volumes, higher CRP was associated with lower FDG-PET, with the relationship becoming stronger with
increasing WMH load. Figure 4.3 illustrates the interaction association plots between VEGF and CRP and
WMH on FDG-PET SUVR in the MCA territory. Forest plots illustrating the beta-estimates and 95% con-
dence interval (CI) of the interaction terms can be found in Figure4.4. We also assessed whether WMH
volume was directly associated with CSF VEGF and CRP covarying for age, sex, APOE carrier status, di-
agnosis, t- tau, ICV, and site (random eect). Higher WMH volume was trend level associated with lower
CSF VEGF(WMH β(SE)=-0.01 (0.003),p=0.063)andlowerCRP(WMH β(SE)=-0.03(0.02),p=0.076).
4.4.4.1 Amyloid-positivitydierences
Westratiedthecohortbyamyloidbetapositivitystatustoevaluatewhetherthepreviouslydetectedasso-
ciations and interactions between CSF biomarkers (VEGF, CRP) and WMH volume on regional FDG-PET
SUVRwasspecictothoseontheADcontinuum. InA β+participantsonly(n=115)aninteractionbetween
VEGFandWMHvolumewasidentiedforFDG-PETSUVRinallROIs: ACA(interactionβ(95%CI)=-0.08
(-0.15, -0.01), p=0.026), MCA (interaction β 95% CI) = -0.08 (-0.14, -0.01), p=0.022) and PCA (interac-
tion β (95% CI) = -0.08 (-0.14, -0.02),p=0.016); where higher WMH volume load dampened the positive
100
Figure 4.3: Interaction plots between VEGF and CRP and WMH on FDG-PET SUVR in the MCA
territory. InthosewithlowWMHvolumes,higherCSFVEGFwasassociatedwithhigherFDG-PETSUVR,
butasWMHloadincreasestherelationshipisattenuatedandbecomesnegativeinthosewiththehighest
WMH load (> 2 SD of mean). In those with low WMH volumes, higher CRP was associated with lower
FDG-PET, with the relationship becoming stronger with increasing WMH load. MCA = Middle Cerebral
Artery, SUVR = Standard Uptake Value Ratio, SD = Standard Deviation.
association between VEGF and FDG-PET. Also, in Aβ+ participants only, CRP and WMH volume had a
signicant interaction on FDG-PET SUVR in ACA (interactionβ (95% CI) = -0.04 (-0.07, -0.01),p=0.018)
and MCA (interaction β (95% CI) = -0.04 (-0.06, -0.01),p=0.022) territories and trend level interaction in
the PCA territory (interaction β (95% CI) = -0.03 (-0.05, 0.002),p=0.078) FDG-PET SUVR. No signicant
interaction was found between VEGF (all p-values > 0.51) or CRP (all p-values > 0.58) on regional FDG-
PET SUVR in Aβ- participants (n=41), although the sample size was small. Moreover, in amyloid positive
participants with low WMH volumes, higher CSF VEGF was associated with higher FDG-PET SUVR, but
as WMH load increases the relationship was dampened. In amyloid positive participants with low WMH
volumes,higherCRPVEGFwasassociatedwithlowerFDG-PET,withtherelationshipbecomingstronger
101
Figure4.4: Forestplotsillustratingthebeta-estimatesand95%condenceintervalsfortheVEGF
by WMH and CRP by WMH interaction on FDG-PET SUVR in the ACA, MCA, PCA territories
intheentirecohort(Analysis1),andbystraticationofparticipantsbyamyloid-betapositivity
(Analysis 2), APOE4 carriers (Analysis 3), and sex (Analysis 4). In those with low WMH volumes,
higherCSFVEGFwasassociatedwithhigherFDG-PETSUVR,butasWMHloadincreasestherelationship
is attenuated. In those with low WMH volumes, higher CRP VEGF was associated with lower FDG-PET,
with the relationship becoming stronger with increasing WMH load. ACA = Anterior Cerebral Artery,
MCA=MiddleCerebralArtery,PCA=PosteriorCerebralArtery,SUVR=StandardUptakeValueRatio. *
Indicatesp< 0.05
withincreasingWMHload. Forestplotsillustratingthebeta-estimatesand95%condenceinterval(CI)of
the interaction terms can be found in Figure 4.4.
102
To evaluate whether these ndings were driven by baseline dierences in WMH volume and CSF
values, we performed Welch’s two sample t-tests. Aβ+ participants had signicantly higher VEGF values
(p=0.001) and trend level higher CRP values (p=0.051) and WMH volumes (p=0.072) compared
to Aβ- participants. Moreover, dierences found between A β+ and Aβ- participants could be a result of
dierences in sample sizes between groups and/or baseline dierences in CSF values and WMH volumes.
4.4.4.2 APOE4-carrierdierences
WhiletheAβpositivitycohortstraticationwasunequalinsamplesize(resultinginlimitedinterpretabil-
ityofresultsduetopowerissues),approximatelyequalsamplessizeswereavailableforAPOE4carriersta-
tus(APOE4carriern=80, APOE4non-carriers=76). 75(93.8%)ofthe80 APOE4carrierparticipantswere
amyloid positive, and 41 (52.6%) of the 78 APOE4 non-carriers were amyloid positive. We also performed
Welch’s two sample t-tests and found APOE4-carriers had trend level higher WMH volumes (p=0.051)
and no dierence between VEGF (p=0.54) and CRP values (p=0.16) compared to APOE4 non-carrier
participants.
Inourinteractionanalyses,wefoundinAPOE4carriersonly,aninteractionbetweenVEGFandWMH
volume on FDG-PET ROIs in the ACA territory (interaction β (95% CI) = -0.09 (-0.16, -0.01), p=0.030)
and trend level interaction in the MCA (interaction β (95% CI) = -0.07 (-0.14, -0.004),p=0.050) and PCA
territories (interaction β (95% CI) = -0.07 (-0.15, -0.003), p=0.058). In APOE4 carriers only, the positive
associationofVEGFwithFDG-PETwasattenuatedinparticipantswithhigherWMHvolumes. InAPOE4
non-carriers, no signicant interactions were found between VEGF and WMH volume on FDG-PET ROIs
(p-values > 0.59).
In the interaction analysis of CRP and WMH volume on FDG-PET SUVR, we found that in APOE4
carriersonly,higherWMHvolumeattenuatedthenegativeassociationbetweenCRPandFDG-PETSUVR
intheACA(interactionβ(95%CI)=-0.04(-0.08,-0.003),p=0.045)andMCAterritory(interactionβ(95%
103
CI) = -0.04 (-0.07, -0.0004), p=0.038), but not in the PCA territory (interaction β (95% CI) = -0.03 (-0.06,
0.01), p=0.18). In APOE4 non-carriers, no signicant interactions were found between CRP and WMH
volumeonFDG-PETROIs(p-values>0.17). NeitherVEGFandCRPhadaninteractionwithWMHvolume
on regional FDG-PET SUVRS in APOE4 non-carriers, although condence intervals overlapped between
APOE4 carriers and APOE4 non-carriers. A summary of beta-estimates and 95% CI are visually depicted
in Figure 4.4.
4.4.4.3 SexDierences
Westratiedthecohortbysextoevaluatewhethersexdierenceswerepresentintheinteractionbetween
CSFbiomarkers(VEGF,CRP)andWMHvolumeonregionalFDG-PETSUVR.First,weperformedWelch’s
twosamplet-testsandfoundthatmaleshadtrendlevelhigherVEGFvalues(p=0.058)andnodierence
in CRP values (p=0.57) and WMH volumes (p=0.25) compared to females. In males only (n=103),
WMH volume interacted with VEGF on FDG-PET SUVR in all vascular territory ROIs: ACA (interaction
β (95% CI) = -0.08 (-0.16,-0.002),p=0.055), MCA (interaction β (95% CI) = -0.08 (-0.14, -0.01),p=0.040),
andPCA(interaction β(95%CI)=-0.09(-0.15,-0.02),p=0.015). Anominaleectinmaleswasalsofound
in the interaction between CRP and WMH volume on FDG-PET SUVR in all vascular territory ROIS (p-
values < 0.10). In females (n=53), no signicant interactive eects of CSF biomarkers (VEGF, CRP) and
WMHonFDG-PETSUVRwasfound(p-values>0.55),althoughthesamplesizewassmall. Allinteraction
beta-estimates and 95% CIs are found in Figure 4.4.
4.4.5 CorticalDKAtlasROIassociations
Inthemaineectanalysis,higherlevelsofVEGFwereassociatedwithhigherFDG-PETsignalinthemean
bilateralfusiformgyrus(β(95%CI)=0.17(0.05,0.3),p=0.009),isthmuscingulate(β(95%CI)=0.22(0.0,
0.43), p=0.042), inferior temporal (β (95% CI) = 0.28 (0.14, 0.41),p< 0.001), opercularis (β (95% CI) =
104
0.20 (0.02,0.38), p=0.031), the banks of the superior temporal sulcus (β (95% CI) = 0.28 (0.11, 0.45), p=
0.001), the superior temporal (β (95% CI) = 0.14 (0.02, 0.027), p=0.026), middle temporal (β (95% CI) =
0.25 (0.11, 0.39),p=0.001), supramarginal (β (95% CI) = 0.23 (0.08, 0.38),p=0.003), and inferior parietal
gyrus(β(95%CI)=0.27(0.10,0.45),p=0.003)priortoFDRcorrection. AfterFDRcorrection,higherlevels
of VEGF remained associated with higher FDG-PET signal in the banks of the superior temporal sulcus,
inferior temporal, middle temporal, supramarginal, and inferior parietal gyrus.
In the interaction analysis between VEGF and WMH volume on regional FDG-PET SUVR, most ROIs
exhibited a signicant interaction (FDR p-values < 0.049), or a trend level eect (banks of the superior
temporal sulcus, frontal pole, medial orbitalfrontal gyrus, caudal anterior cingulate, posterior cingulate,
entorhinal cortex, superior parietal, superior temporal FDR p-values < 0.076), except the temporal pole
and paracentral gyrus (FDR p-values > 0.21). Specically, we found that in those with low WMH volume,
higherVEGFlevelswereassociatedwithhigherregionalFDG-PETSUVRthroughoutthecortex,andthis
relationship was weakened in those with larger WMH volumes.
In the main eect analysis, lower levels of CRP were associated with higher FDG-PET signal in the
temporal pole (β (95% CI) = -0.03 (-0.05, -0.003), p=0.029), insula (β (95% CI) = -0.04 (-0.07, -0.01), p =
0.014), superior temporal (β (95% CI) = -0.03 (-0.06, 0.00),p=0.0498), transverse temporal gyrus (β (95%
CI)=-0.05(-0.09,-0.002),p=0.043),parahippocampal(β(95%CI)=-0.04(-0.06,-0.02),p=0.002),lingual
(β(95%CI)=-0.04(-0.07,-0.003),p=0.035),andlateralorbitofrontalgyrus(β(95%CI)=-0.04(-0.07,0.00),
p=0.0496) prior to FDR correction. No regions passed FDR.
IntheinteractionanalysisbetweenCRPandWMHonregionalFDG-PETSUVR,wefoundsignicant
interactions throughout the cortex after FDR correction 4.5; orbitofrontal (medial and lateral), inferior
frontal (orbitalis, triangularis, opercularis), middle frontal (rostral and caudal), cingulate (rostral, caudal,
posterior, isthmus), precuneus, pre and post-central, superior parietal, supramarginal, insula, transverse
105
temporal, entorhinal, and parahippocampal gyrus (FDR p-values < 0.049). For detailed plot of beta esti-
mates and 95% condence intervals, see Figure4.5.
Figure 4.5: ForestplotofbothmaineectsbetweenCSFBiomarkers(VEGF,CRP)andFDG-PET
SUVR and interactions between CSF Biomarkers (VEGF, CRP) and WMH volume on FDG-PET
SUVR in FreeSurfer Cortical ROIs. For ease of visualization, ROIs are color coded by the vascular
territory they are most prominent in, but ROI vascular territory locations may vary by individual or may
be apparent in more than one region. (Red=ACA, Green=MCA, and Blue=PCA). Scales vary across each
forest plot. ROI = Region of interest, STS = Superior temporal sulcus, ACA = Anterior Cerebral Artery,
MCA = Middle Cerebral Artery, PCA = Posterior Cerebral Artery. * indicates the p-value < 0.05 prior to
FDR correction. ** indicates p-value < 0.05 after FDR correction.
106
4.5 Discussion
Ourndings demonstrate that 1) CSF markers of vascular inammation (VEGF and CRP) have indepen-
dent associations withregional FDG-PET,2)delineatingboundariesbyvascularterritorymaybeauseful
approachinunderstandingvariabilityinvascularphenotypepatternsinthosealongtheADcontinuum,3)
WMHvolumehasdiuseinteractiveeectsonCSFbiomarkerassociationswithFDG-PETthroughoutthe
cortex,and4)WMHvolume,APOE4genotype,amyloidstatus,andsexareimportantvariablestoconsider
in uncovering mechanisms driving heterogeneity in early FDG-PET signaling and AD brain phenotype
patterns. Collectively, our results suggest that vascular dysfunction contributes to immunometabolism
relationships in those most at risk for AD.
In the whole sample, higher CSF VEGF levels and lower CSF CRP levels were associated with higher
whole-brain FDG-PET signal. While similar positive associations between VEGF and FDG-PET uptake
have previously been reported in the same ADNI cohort (Wang et al., 2018; Tubi et al., 2021), no studies
have reported an association between CSF CRP levels and regional FDG-PET signal in an older adult or
ADcohort. ThesebiomarkersalsohadinteractiveeectswithWMHvolumeonFDG-PETuptake. Higher
WMH load dampened the positive relationship between VEGF and FDG-PET and strengthened the nega-
tiverelationshipbetweenCRPandFDG-PET.Theseresultsindicatethatinbrainswiththemostapparent
vascular pathology, CSF vascular inammatory levels shift toward a negative association with FDG-PET
levels, possibly reecting a shift in FDG-PET signal composition to be more representative of vascular
dysfunction rather than glucose metabolism. Changes in FDG-PET signal indicate various physiological
processesbeyondglucosemetabolism,asFDG-PETcannotcapturetheentiremetabolismofGLUT1across
the endothelium to the uptake into neurons. As a result, alterations in signal may reect BBB breakdown
(Chiaravallotietal.,2016;Yinetal.,2016;Sweeneyetal.,2019),anage-dependentshifttofromneuronalto
astrocytic function (Jiang and Cadenas, 2014), and microglial-induced inammatory activity ( Choi et al.,
2021;Xiangetal.,2021). Moreover,whenvasculardiseaseisabsent,therelationshipbetweenCSFvascular
107
biomarkers with FDG-PET signal may reect a closer coupling of vascular inammatory protein markers
to synaptic and neuronal energy metabolism. However, when vascular disease is severe, the negative
association of CSF vascular biomarkers with FDG-PET signal, may reect a loss of coupling to energy
metabolismandashifttoacouplingofvasculardysfunctionandBBBbreakdown. Ourcurrentstudycan-
not assess this, and future work should use multi-modal techniques to dissect what is driving FDG-PET
signal changes in low and high conditions of cerebrovascular disease.
RegionalassociationsbetweenVEGFandCRPandFDG-PETwereidentiedinbothvascularterritories
and in cortical gyral boundaries. Before accounting for WMH volume interaction eects, higher VEGF
levels were associated with greater FDG-PET SUVR in the MCA and PCA territories, while lower CRP
levels were associated with greater FDG-PET SUVR in the PCA territory, after adjusting for age, sex,
CSF total-tau level, diagnosis, APOE4 genotype and site. CSF biomarkers may be related to FDG-PET
in the PCA, because previous work has shown that cerebral small vessel disease (microbleeds, amyloid
angiopathy;CAA)ismoreprominentinthePCAterritory(Jiaetal. 2014). WefoundthatCSFbiomarkers
were not directly associated with FDG-PET SUVR in the ACA territory, which was the vascular territory
to exhibit signicantly higher mean SUVR compared to both MCA and PCA FDG-PET SUVR, suggesting
that the ACA may be more resistant to perturbations in metabolism. However, when we accounted for
theinteractionbetweenWMHvolumeandCSFbiomarkersonFDG-PETsignalintheACAterritory,there
wasasignicantinteractiveeect;greaterWMHvolumedampenedthepositiveassociationbetweenVEGF
andFDG-PETsignal. ThissuggeststhatvariabilityinWMHvolumemasksassociationsbetweenvascular
inammatory biomarkers and FDG-PET signal and demonstrates that the cortex is diusely sensitive to
WMH accumulation. Further, greater WMH volume, a surrogate marker of vascular disease, may more
severely uncouple neurovascular biomolecular signaling from neuroenergetic synaptic needs in AD. Of
note, participants with lower WMH volume had signicantly higher FDG-PET uptake than those with
lowerWMHvolumes,implyingthatthelinkbetweenCSFmeasures(e.g.,VEGF)andFDG-PETsignalmay
108
just be most apparent when FDG-PET levels are higher and before brain decline has begun. Together,
thesendings illuminate the need to account for vascular disease in modeling associations between CSF
biomarkers and AD-neuroimaging measures.
Inourcortex-wideassessmentofFDG-PETsignalusingcorticallydenedROIs,wefoundregionalas-
sociationsbetweenVEGFonFDG-PETSUVR.HigherVEGFlevelswerealsoassociatedwithgreaterFDG-
PET SUVR in the superior parietal, supramarginal, middle temporal, and inferior temporal lobe, aligning
with work from a recent study that found subclinical carotid plaque burden was associated with voxel-
wiseFDG-PETuptakeintheparietotemporalregionsandcingulategyrus(Cortes-Cantelietal.,2021). The
detected regions are also primarily in the MCA territory and within the previously detected AD cortical
thinning signature (Wang et al., 2015a; Tubi et al., 2021). The regions also partially align with regions
activated in the default mode network (DMN), a set of cortical anatomical and functional hubs with high
metabolicdemands(Vlassenkoetal.,2010;Passowetal.,2015). Potentially,theconvergenceoftheregions
withspatiallydistinctvulnerabilitytovasculardisease(e.g.,atherosclerosis),highenergeticdemands(i.e.,
DMN), and susceptibility to hypoperfusion (i.e., vascular territory borderzones) make them more vulner-
able to age- and AD-related changes to inadequate hemodynamic and synaptic-energy demands in the
presence of vascular dysfunction (Huang et al., 2018).
WMH volume interacted with VEGF levels on FDG-PET signal diusely across the cortex, regardless
of vascular territory or cortical gyri boundary delineation. Specically, the positive association between
VEGFlevelsandFDG-PETsignalwereattenuatedinparticipantswithgreaterWMHburden. Wealsofound
thatgreaterWMHvolumewastrendlevelassociatedwithlowerVEGFlevels,afteradjustingforage,sex,
diagnosis, total-tau, APOE4 carrier status, and ICV. While no studies have investigated the interaction
between vascular disease and VEGF before, animal models have demonstrated that VEGF expression is
modied by vascular burden and vascular forces. VEGF expression is modied by altered shear stress
dynamics induced by atherosclerosis (Ghaari et al., 2017; Lehoux and Jones, 2016). VEGF levels may
109
also indirectly be modied by inammation. For example, VEGF signaling may be up- or down-regulated
depending on the presence and duration of inammatory signals, such as tumor-necrosis-factor (TNF)-
alpha (Sainson et al., 2008). However, our study was cross-sectional, and we cannot infer causation or
directionality of the interactions.
AsignicantinteractionbetweenCRPandWMHvolumeonregionalFDG-PETsignalwasalsofound
throughout the cortex. In participants with larger WMH volumes, higher CRP was associated with lower
FDG-PETglucosemetabolism. WhilenostudieshaveinvestigatedtheinteractionbetweenCRPandWMH
on AD brain biomarkers, CRP and WMH volume may have bi-directional eects driven by genetics and
circulating protein levels. For example, carriers of the CRP-286T allele have an increased inammatory
response and larger WMH volumes compared to other allelic combinations (Raz et al., 2012). Also, in
middle aged to older adults, higher plasma CRP levels were associated with larger WMH volumes and
worse white matter microstructural integrity (Wersching et al., 2010; Raz et al., 2012; Walker et al., 2018).
Higher serum CRP levels have also been associated with other regional AD brain biomarkers, such as
cortical thinning, in a study evaluating how metabolic risk, physical activity, and inammation relate to
cognitive aging (Corlier et al., 2018).
The interactions between CSF biomarkers and WMH volume on FDG-PET uptake were strongest in
APOE4 carriers compared to non-carriers. This aligns with work demonstrating that CSF biomarker dif-
ferencesaremostapparentinthoseatriskforAD,suchasinAPOE4carriers. Forexample,CRPlevelsare
lowerinAPOE4carrierscomparedtonon-carriers(Duarte-Gutermanetal.,2020;Konijnenbergetal.,2020;
Wangetal.,2022),andalsohaveinteractiveeectsonADriskandAD-relatedtemporallobebrainatrophy
and hippocampal volume (Tao et al., 2018). Interestingly, Tao et al., found that the interactive eects be-
tween CRP levels and AD-related brain changes were strongest in those without cerebrovascular disease,
whereas we found that the negative association between CSF CRP and FDG-PET was strongest in those
withhighWMHvolume(i.e.,greatervasculardisease). Whilenoin-vivohumanstudieshaveinvestigated
110
whether VEGF levels interact with APOE4 genotype and vascular disease on AD-brain outcomes, animal
model and genetic studies indicate that APOE4 genotype modulates VEGF genetic pathways. For exam-
ple, in APOE4 carriers, higher VEGFA expression is associated with worse cognitive performance, while
no association between VEGFA expression and cognitive performance was found in APOE4 non-carriers
(Mooreetal.,2020). APOE4carriers(vs. non-carriers)andamyloidpositive(vs. negative)participantshad
higherWMHvolumescomparedtotheirrespectivecounterparts,suggestingthatgroupdierencescould
also be just most apparent in those with high WMH volumes.
The interaction between CSF biomarkers and WMH volume on FDG-PET ROIs was also stronger in
Aβ+ participants compared to Aβ- participants. This nding aligns with previously published reports
fromsimilarADNIcohorts,whereassociationsbetweenVEGFandAD-neuroimagingmarkers(e.g.,FDG-
PETSUVR,hippocampalvolume)werestrongestinthosewithabnormallevelsofADpathology(i.e.,CSF
Aβ, total-tau, and phosphorylated-tau) (Hohman et al., 2015; Tubi et al., 2021). VEGF associations may
also be more apparent in those with high amyloid levels, because Aβ and VEGF have a bi-directional
relationship,withtheabilityofamyloidplaquestoco-accumulateandsequesterVEGFwithahighanity
(Yang et al., 2004; Religa et al., 2013; Angom et al., 2019; Martin et al., 2021). Amyloid accumulation can
also contribute to alterations in FDG-PET uptake in white matter pathology; FDG-PET SUVR in WMH
regions is lower in Aβ+ participants compared to Aβ- participants (Kalheim et al., 2016). Higher plasma
CRPlevelshavealsobeenassociatedwithlargerWMHvolumes,withCRPlevelsinteractingwithAβlevels
on markers of cerebral small vessel disease (e.g., microbleeds, enlarged perivascular spaces) (Hilal et al.,
2018). AssociationsmayalsobemoreapparentinAβ+thanAβ-participantsbecausewea-prioriselected
forCSFvascularbiomarkersthatweremostrelatedtoamyloidlevelstobetterdetectassociationsrelevant
toAD-pathogenesis. However,ourstudycannotdistinguishiftheassociationsaredrivenbyparenchymal
versusvasculardepositionofamyloid(i.e.,CAA),sinceabnormalCSFAβlevelscanreectbothCAAand
parenchymal amyloid (Banerjee et al., 2020). Given that APOE4 genotype is a major risk factor for CAA
111
(Shinohara et al., 2016), and CAA is a risk factor for WMH accumulation (Gurol et al., 2013) and is linked
with neurovascular decoupling (Peca et al., 2013), it is possible that these interactions may be driven by
CAA (rather than parenchymal amyloid), although our study could not directly assess this.
ThesignicantinteractionbetweenCSFVEGFandWMHvolumeonFDG-PETsignalintheMCAand
PCAterritorywasstrongerinmalescomparedtofemaleparticipantsafterstratication. Thisisconsistent
withpastndingsdemonstratingthatinmales,butnotfemales,greaterWMHvolumesincreasestherisk
for AD disease progression (Kim et al., 2015; Burke et al., 2019; Salminen et al., 2022). The sex-dierences
may also be driven by higher levels of VEGF found in males compared to females in our study. Alterna-
tively, the results may be a result of inadequate power, since there was an unequal distribution between
groups (male n = 103, female n = 53). Outside of our study, few studies have examined sex dierences in
CSF biomarkers as they relate to AD-neuroimaging markers (Salminen et al., 2022) and the studies that
have examined sex dierences in VEGF levels have collectively been inconclusive. For example, VEGF
expression in female endothelial progenitor cell dierentiation is signicantly higher compared to male
cells (Randolph et al., 2019). However, in an in-vivo sample of 443 participants (31-75 years old), VEGF
serum levels were signicantly higher in males than females, with male smoker’s serum levels signi-
cantly higher than non-smokers serum levels, suggesting that risk factors for vascular disease augment
VEGFlevels(Kimuraetal.,2007). Altogether,thesendingssuggestthatsex,hormonevariation,andvas-
culardiseaseareimportantbiologicalvariablestoconsiderinunveilingheterogeneityinADphenotypes,
but future work needs to include expansive datasets that can evaluate how circulating CSF biomarkers
relate to AD neuroimaging biomarkers by sex.
WefoundnoassociationbetweenCSFbrinogen,vWF,andVCAMandwholebrainFDG-PETsignal.
WhileitispossiblethatVCAM,vWF,andbrinogendonotrelatetoFDG-PETsignalingeneral,itisalso
possible that they are not associated with FDG-PET in our study because the sensitivity of their eects
112
maybemoreapparentatadierenttemporalpointinADpathogenesis. Forexample,alterationsinFDG-
PETsignaloccurearlyinAD,butchangesintheexpressionandsensitivityoftheseCSFbiomarkerscould
occur earlier or later than AD-related changes in FDG-PET signal (Wolters et al., 2018). Medication use
may also modify CSF protein levels, such as VCAM, potentially masking eects of interest (Rezaie-Majd
et al., 2003; Hattori et al., 2006; Hocaoglu-Emre et al., 2017).
While our study was novel in many domains, several limitations in our study are present. Our study
wasthersttouseapreviouslyestablishedMNI-templateofvascularterritoriestodelineateboundariesof
interestinFDG-PETsignal. However,wecouldnotassesstheactualvascularterritoriesoftheparticipants.
While our study was also therst to evaluate how CSF biomarkers implicated in vascular inammation
relatetoregionalFDG-PETSUVR(denedbybothvascularterritoriesandcortexwidegyralboundaries),
the study was cross-sectional. The cross-sectional nature of the study limits the interpretability of the
causal and temporal scale of relationships between CSF biomarkers and FDG-PET. In our study, sample
sizesofgroupswereunequal betweenamyloidandsexgroups,limitingtheinterpretabilityoftheresults.
However, similar results were seen between APOE4 carriers vs non-carriers, which had similar sample
sizesbetweengroups. WeusedCSFAβlevelstoindicatewhetheraparticipantwasontheAD-continuum,
but abnormal CSF Aβ levels may also reect vascular, rather parenchymal amyloid accumulation. We
also used WMH volume as a general index of vascular disease severity and did not have spatial maps of
WMH segmentations to evaluate whether the eects varied by brain location (e.g., periventricular, deep).
Thus, larger WMHs may be more likely to include deep WMHs and may drive detected interactions, as
periventricular and deep WMHs have dierent etiological origins and clinical consequences. Also, the
ADNI dataset is not a population-based sample, and the distribution and extent of WMH volumes and
CVD may be larger and more variable in the general population. We also did not account for medication
useinourstudydesign,whichmaymodifyproteinlevels. Toaddresstheselimitations,futureworkshould
expandcohortsizestoaugmentpowertodetectvariabilityinpopulationsatriskforADbyincorporating
113
equal group distributions that are also ethnically and racially diverse and account for medication use.
Additionally, future work in this domain would benet from evaluating vascular territory-derived CBF
and FDG-PET in the same cohort, include a broader range of vascular risk variables (e.g., hypertension,
atherosclerosis, etc.), and evaluate whether results are similar for other bio-energetic markers potentially
driving neurovascular-neuroenergetic uncoupling in AD.
Our results demonstrate that the vascular inammatory markers VEGF and CRP are both indepen-
dently associated with regional FDG-PET signal in regions dened by vascular territories and cortical
boundaries. Greater WMH volume, a marker of vascular disease, interacted with CSF biomarkers and
FDG-PET signal diusely throughout the cortex, unmasking previously undetected associations between
CSFbiomarkersandregionalFDG-PETsignal. OurresultshadthestrongesteectsinA β+participants(vs.
Aβ- participants) and APOE4 carriers (vs. APOE4 non-carriers), suggesting that vascular pathology may
alter immunometabolism relationships in those most at risk for AD. Ourndings were also strongest in
males (vs. females), suggesting that sex is an essential biological variable to consider in unveiling mecha-
nismsandheterogeneityinvascularADbrainphenotypes. Furtherevaluationofthecomplexrelationships
between vascular dysfunction, CSF biomarkers and neuroimaging measures can help provide proles of
the AD continuum to improve the accuracy of diagnostic predictions, disease staging and distinguish ef-
fective therapeutic targets.
4.6 SupplementaryMaterial
114
Table S4.1: VascularCSFbiomarkerassociationtoFDG-PETvascularterritoryregions(n=158)
VEGF CRP
ROI β (SE) p-value β (SE) p-value
ACA 0.08 (0.08) 0.29 -0.02 (0.02) 0.20
MCA 0.17 (0.07) 0.017** -0.03 (0.01) 0.056
PCA 0.12 (0.07) 0.062 -0.03 (0.01) 0.016**
Modelled variables included VEGF, age, sex, diagnosis, APOE4, CSF t-tau, ICV, and site (random eect) in
a linear mixed eects model. SE = Standard Error. *p < 0.05 before FDR correction. ** p < 0.05 after FDR
correction
115
Chapter5
Conclusions
Collectively, my research aims toll an essential gap in our understanding of how vascular dysfunction
relatestoregionalADbrainbiomarkers,cognition,andphenotypepatterns. Thisbodyofworkattemptsto
unravelcomplexinteractionsbetweencirculatingphysiologicalbiomarkersandbrainindicesofstructure
andfunction. Myresearchthusfardemonstratesbothindependentandinteractiverelationshipsbetween
vascular dysfunction and AD neuropathology on brain health, clarifying mechanisms and populations
most at-risk for suboptimal brain aging. Specically, my dissertation focused on understanding vascu-
lar contributions to suboptimal brain aging in older adults along the AD continuum. I focused on two
markers of vascular function: 1) White matter hyperintensities (WMH), which is a neuroimaging marker
that generally reects vascular burden and is linked to abnormal arterial pressure, plasma leakage, blood
brain barrier (BBB) permeability, axonal loss, arteriolosclerosis, and greater body mass index (BMI). We
found that the clinically relevant WMH segmentation methods were those that excluded the lightest and
most diuse hyperintensities. Further, this relationship was only present in participants with low levels
of amyloid in the brain, suggesting that AD-pathological burden may mask cognitive consequences of
WMHs. Our work was also an initial step toward harmonizing WMH segmentation protocols, helping
promotemorerobustandreliableinvestigationsonWMHs. 2)Vascularendothelialgrowthfactor(VEGF)
116
is a multi-purpose signaling protein involved in blood vessel growth, oxygen and glucose delivery, va-
sodilation,andvascularpermeability,andmodiablethroughexercise,makingitapotentiallytherapeutic
target in AD. We identied regional relationships between VEGF and FDG-PET and cortical thickness –
indices that are altered early in AD pathogenesis and provide spatial specicity to heterogeneity in AD.
Further, we found that FDG-PET signal mediates the relationship between endogenous VEGF levels and
cognition. We also found that CSF VEGF and CRP are related to FDG-PET signal in both regions dened
by cerebral circulation (ACA, MCA, and PCA territories) and cortically dened regions primarily located
in the AD-cortical thinning signature. Our results suggest that higher VEGF levels may signal a vascular
injury response when AD-neuropathological load is low but serve a compensatory neuroprotective role
when AD-neuropathological burden is high. We also found that the positive relationship between VEGF
andFDG-PETuptakeisattenuatedinthosewithhigherlevelsofWMHaccumulation,andthenegativere-
lationshipbetweenCRPandFDG-PETisstrengthenedinthosewithhigherlevelsofWMHaccumulation.
Further, these interaction eects were most apparent in those with AD pathology,APOE4 carriers, and
males. Thiscollectionofworkrevealsthatthatatdierentamyloid,tau,andvasculardiseaseloads,there
is variability in the relationship between measures of vascular function and AD-measures of brain and
cognitive aging. This remains an essential area of focus, as targeting vascular dysfunction is a modiable
risk factor that can reduce disease risk and slow disease progression.
Inthebroaderscopeofmyfutureresearchaims,Iplantoclarifyothermediatingandmoderatingpath-
waysinADphenotypepatternstofurtherrevealmechanismsdrivingdierencesinADriskandprogres-
sion. For example, this work demonstrates that sex dierences, neuropathology levels,APOE4 genotype,
and vascular burden all contribute to variability in relationships to AD brain and cognition biomarkers.
Without proper interaction tests, larger sample sizes, and independent cohort analyses, many complex
eectsmaybemissedbythesephysiologicalprocesses. Anothercriticalmeasurethatwasnotincludedin
this body of work is how race and ethnicity interact with these relationships - a critical area of focus on
117
my future research endeavors. To address this, rst steps after this dissertation work include analyzing
howVEGFrelatestovoxelwisewhitemattermicrostructuralchangesinanindependentcohort. Evidence
ofbothwhitematter(WM)micro-andmacro-structuralabnormalitiesoccurinAD.Fractionalanisotropy
(FA), an index of microstructural integrity and diusion restriction within an MRI voxel, augments the
power to detect AD-specic decits. However, mechanisms driving loss of WM microstructural integrity
in AD are poorly understood. Importantly, VEGF provides neurotrophic support via glia and is also se-
creted by glial cells essential for maintaining myelin, and may be an important factor in understanding
mechanisms driving WM integrity loss in AD. However, in humans, the relationship between VEGF and
WMmicrostructuralintegrityhasyettobemapped. IalsoplantoincorporatehowADpathology,APOE4
genotype, sex, vascular disease, and ethnicity/race interact with the relationship between CSF VEGF on
white matter microstructural integrity to better understand factors that may additively or synergistically
contribute to WM integrity loss in AD.
BymappingcomplexpathophysiologicalrelationshipsinAD,wecanmovetowardsprecisionmedicine
tobetterprevent,halt,andslowdiseaseprogression. Myfutureresearchaimstocreateamulti-dimensional
map of how vascular risk proles relate to AD pathogenesis to better target populations most at risk for
cognitivedecline. Afurtheraspirationalgoalistodesignaprogrammaticlineofresearchthatgoesbeyond
associations between risk/protective factors and biomarkers of AD to elucidating mechanisms by which
vascular factors impact neuronal metabolism and functional activity (e.g., neurovascular-neuroenergetic
uncoupling)inthesettingofAD.Ultimately,Iaspireformypastandcurrentcollectionsofworktopromote
theprogressofscienceandimprovenationalhealthbybothresearchingfactorsthatcontributetodisease
vulnerability and by promoting resiliency in and outside of the laboratory.
118
References
Albrecht, D., Isenberg, A. L., Stradford, J., Monreal, T., Sagare, A., Pachicano, M., Sweeney, M., Toga, A.,
Zlokovic, B., Chui, H., et al. (2020). Associations between vascular function and tau pet are associated
with global cognition and amyloid. Journal of Neuroscience, 40(44):8573–8586.
Aljondi,R.,Szoeke,C.,Steward,C.,Gorelik,A.,andDesmond,P.(2020).Theeectofmidlifecardiovascular
risk factors on white matter hyperintensity volume and cognition two decades later in normal ageing
women. Brain imaging and behavior, 14(1):51–61.
Amariglio,R.E.,Becker,J.A.,Carmasin,J.,Wadsworth,L.P.,Lorius,N.,Sullivan,C.,Maye,J.E.,Gidicsin,
C., Pepin, L. C., Sperling, R. A., et al. (2012). Subjective cognitive complaints and amyloid burden in
cognitively normal older individuals. Neuropsychologia, 50(12):2880–2886.
Ammirati, E., Moroni, F., Magnoni, M., Rocca, M. A., Messina, R., Anzalone, N., De Filippis, C., Scotti, I.,
Besana,F.,Spagnolo,P.,etal.(2017). Relationbetweencharacteristicsofcarotidatheroscleroticplaques
and brain white matter hyperintensities in asymptomatic patients. Scientic reports , 7(1):1–11.
Anchisi, D., Borroni, B., Franceschi, M., Kerrouche, N., Kalbe, E., Beuthien-Beumann, B., Cappa, S., Lenz,
O., Ludecke, S., Marcone, A., et al. (2005). Heterogeneity of brain glucose metabolism in mild cognitive
impairment and clinical progression to alzheimer disease. Archives of neurology, 62(11):1728–1733.
Angom,R.S.,Wang,Y.,Wang,E.,Pal,K.,Bhattacharya,S.,Watzlawik,J.O.,Rosenberry,T.L.,Das,P.,and
Mukhopadhyay, D. (2019). Vegf receptor-1 modulates amyloid 1–42 oligomer-induced senescence in
brain endothelial cells. The FASEB Journal, 33(3):4626–4637.
Arnáiz,E.andAlmkvist,O.(2003). Neuropsychologicalfeaturesofmildcognitiveimpairmentandpreclin-
ical alzheimer’s disease. Acta Neurologica Scandinavica, 107:34–41.
Ashby, J. W. and Mack, J. J. (2021). Endothelial control of cerebral bloodow. The American Journal of
Pathology, 191(11):1906–1916.
Au, R., Massaro, J. M., Wolf, P. A., Young, M. E., Beiser, A., Seshadri, S., D’Agostino, R. B., and DeCarli, C.
(2006). Association of white matter hyperintensity volume with decreased cognitive functioning: the
framingham heart study. Archives of neurology, 63(2):246–250.
Avants, B. B., Epstein, C. L., Grossman, M., and Gee, J. C. (2008). Symmetric dieomorphic image regis-
tration with cross-correlation: evaluating automated labeling of elderly and neurodegenerative brain.
Medical image analysis, 12(1):26–41.
Avants,B.B.,Tustison,N.J.,Song,G.,Cook,P.A.,Klein,A.,andGee,J.C.(2011).Areproducibleevaluation
of ants similarity metric performance in brain image registration. Neuroimage, 54(3):2033–2044.
Baldo, J. V., Schwartz, S., Wilkins, D., and Dronkers, N. F. (2006). Role of frontal versus temporal cortex
in verbal uency as revealed by voxel-based lesion symptom mapping. Journal of the International
Neuropsychological Society, 12(6):896–900.
119
Banerjee, G., Ambler, G., Keshavan, A., Paterson, R. W., Foiani, M. S., Toombs, J., Heslegrave, A., Dick-
son, J. C., Fraioli, F., Groves, A. M., et al. (2020). Cerebrospinal uid biomarkers in cerebral amyloid
angiopathy. Journal of Alzheimer’s disease, 74(4):1189–1201.
Baudic, S., Dalla Barba, G., Thibaudet, M. C., Smagghe, A., Remy, P., and Traykov, L. (2006). Executive
function decits in early alzheimer’s disease and their relations with episodic memory. Archives of
clinical neuropsychology, 21(1):15–21.
Benderro,G.F.andLaManna,J.C.(2011). Hypoxia-inducedangiogenesisisdelayedinagingmousebrain.
Brain research, 1389:50–60.
Bennett, R. E., Robbins, A. B., Hu, M., Cao, X., Betensky, R. A., Clark, T., Das, S., and Hyman, B. T. (2018).
Tau induces blood vessel abnormalities and angiogenesis-related gene expression in p301l transgenic
mice and human alzheimer’s disease. Proceedings of the National Academy of Sciences, 115(6):E1289–
E1298.
Blennow,K.,Hampel,H.,Weiner,M.,andZetterberg,H.(2010). Cerebrospinaluidandplasmabiomarkers
in alzheimer disease. Nature Reviews Neurology, 6(3):131–144.
Blonz,E.R.(2017). Alzheimer’sdiseaseastheproductofaprogressiveenergydeciencysyndromeinthe
centralnervoussystem: theneuroenergetichypothesis. JournalofAlzheimer’sDisease,60(4):1223–1229.
Bolandzadeh, N., Davis, J. C., Tam, R., Handy, T. C., and Liu-Ambrose, T. (2012). The association between
cognitivefunctionandwhitematterlesionlocationinolderadults: asystematicreview. BMCneurology,
12(1):1–10.
Box, G. E. P. and Cox, D. R. (1964). An analysis of transformations. Journal of the Royal Statistical Society.
Series B (Methodological), 26(2):211–252.
Boyle, P. A., Yu, L., Fleischman, D. A., Leurgans, S., Yang, J., Wilson, R. S., Schneider, J. A., Arvanitakis,
Z., Arfanakis, K., and Bennett, D. A. (2016). White matter hyperintensities, incident mild cognitive
impairment, and cognitive decline in old age. Annals of clinical and translational neurology, 3(10):791–
800.
Braak,H.andBraak,E.(1991). Neuropathologicalstageingofalzheimer-relatedchanges. Actaneuropatho-
logica, 82(4):239–259.
Braak,H.andBraak,E.(1997). Frequencyofstagesofalzheimer-relatedlesionsindierentagecategories.
Neurobiology of aging, 18(4):351–357.
Braskie,M.,Boyle,C.,Rajagopalan,P.,Gutman,B.,Toga,A.,Raji,C.,Tracy,R.,Kuller,L.,Becker,J.,Lopez,
O., et al. (2014). Physical activity, inammation, and volume of the aging brain.Neuroscience, 273:199–
209.
Braskie, M. N., Klunder, A. D., Hayashi, K. M., Protas, H., Kepe, V., Miller, K. J., Huang, S.-C., Barrio, J. R.,
Ercoli, L. M., Siddarth, P., et al. (2010). Plaque and tangle imaging and cognition in normal aging and
alzheimer’s disease. Neurobiology of aging, 31(10):1669–1678.
Brickman, A. M., Honig, L. S., Scarmeas, N., Tatarina, O., Sanders, L., Albert, M. S., Brandt, J., Blacker, D.,
andStern,Y.(2008). Measuringcerebralatrophyandwhitematterhyperintensityburdentopredictthe
rate of cognitive decline in alzheimer disease. Archives of neurology, 65(9):1202–1208.
120
Brickman, A. M., Muraskin, J., and Zimmerman, M. E. (2009). Structural neuroimaging in alzheimer’s
disease: do white matter hyperintensities matter? Dialogues in clinical neuroscience, 11(2):181.
Brickman, A. M., Schupf, N., Manly, J. J., Stern, Y., Luchsinger, J. A., Provenzano, F. A., Narkhede, A.,
Razlighi, Q., Collins-Praino, L., Artero, S., et al. (2014). Apoe ε4 and risk for alzheimer’s disease: do
regionally distributed white matter hyperintensities play a role? Alzheimer’s & Dementia, 10(6):619–
629.
Brosseron,F.,Traschütz,A.,Widmann,C.N.,Kummer,M.P.,Tacik,P.,Santarelli,F.,Jessen,F.,andHeneka,
M. T. (2018). Characterization and clinical use of inammatory cerebrospinaluid protein markers in
alzheimer’s disease. Alzheimer’s research & therapy, 10(1):1–14.
Buckner,R.L.,Head,D.,Parker,J.,Fotenos,A.F.,Marcus,D.,Morris,J.C.,andSnyder,A.Z.(2004). Auni-
ed approach for morphometric and functional data analysis in young, old, and demented adults using
automated atlas-based head size normalization: reliability and validation against manual measurement
of total intracranial volume. Neuroimage, 23(2):724–738.
Budson, A. E. and Solomon, P. R. (2012). New criteria for alzheimer’s disease and mild cognitive impair-
ment: Implications for the practicing clinician. The neurologist, 18(6):356.
Bürger,S.,Noack,M.,Kirazov,L.P.,Kirazov,E.P.,Naydenov,C.L.,Kouznetsova,E.,Yafai,Y.,andSchliebs,
R.(2009). Vascularendothelialgrowthfactor(vegf)aectsprocessingofamyloidprecursorproteinand
-amyloidogenesis in brain slice cultures derived from transgenic tg2576 mouse brain. International
Journal of Developmental Neuroscience, 27(6):517–523.
Burggren, A. C., Zeineh, M., Ekstrom, A. D., Braskie, M. N., Thompson, P. M., Small, G. W., and
Bookheimer, S. Y. (2008a). Reduced cortical thickness in hippocampal subregions among cognitively
normal apolipoprotein e e4 carriers. Neuroimage, 41(4):1177–1183.
Burggren, A. C., Zeineh, M., Ekstrom, A. D., Braskie, M. N., Thompson, P. M., Small, G. W., and
Bookheimer, S. Y. (2008b). Reduced cortical thickness in hippocampal subregions among cognitively
normal apolipoprotein e e4 carriers. Neuroimage, 41(4):1177–1183.
Burke, S. L., Hu, T., Fava, N. M., Li, T., Rodriguez, M. J., Schuldiner, K. L., Burgess, A., and Laird, A.
(2019). Sex dierences in the development of mild cognitive impairment and probable alzheimer’s dis-
ease as predicted by hippocampal volume or white matter hyperintensities. Journal of women & aging,
31(2):140–164.
Caligiuri, M. E., Perrotta, P., Augimeri, A., Rocca, F., Quattrone, A., and Cherubini, A. (2015). Automatic
detection of white matter hyperintensities in healthy aging and pathology using magnetic resonance
imaging: a review. Neuroinformatics, 13(3):261–276.
Cao,L.,Jiao,X.,Zuzga,D.S.,Liu,Y.,Fong,D.M.,Young,D.,andDuring,M.J.(2004).Vegflinkshippocampal
activity with neurogenesis, learning and memory. Nature genetics, 36(8):827–835.
Cees De Groot, J., De Leeuw, F.-E., Oudkerk, M., Van Gijn, J., Hofman, A., Jolles, J., and Breteler, M. M.
(2000). Cerebral white matter lesions and cognitive function: the rotterdam scan study. Annals of
Neurology: Ocial Journal of the American Neurological Association and the Child Neurology Society,
47(2):145–151.
Chakraborty, A., Chatterjee, M., Twaalfhoven, H., Milan, M. D. C., Teunissen, C. E., Scheltens, P., Fontijn,
R. D., van Der Flier, W. M., and De Vries, H. E. (2018). Vascular endothelial growth factor remains
121
unchangedincerebrospinaluidofpatientswithalzheimer’sdiseaseandvasculardementia.Alzheimer’s
research & therapy, 10(1):1–7.
Chertkow, H., Whatmough, C., Saumier, D., and Duong, A. (2008). Cognitive neuroscience studies of
semantic memory in alzheimer’s disease. Progress in brain research, 169:393–407.
Chiaravalloti, A., Fiorentini, A., Francesco, U., Martorana, A., Koch, G., Belli, L., Torniolo, S., Di Pietro,
B., Motta, C., and Schillaci, O. (2016). Is cerebral glucose metabolism related to blood–brain barrier
dysfunctionandintrathecaliggsynthesisinalzheimerdisease?: A18f-fdgpet/ctstudy. Medicine,95(37).
Chin-Chan,M.,Navarro-Yepes,J.,andQuintanilla-Vega,B.(2015). Environmentalpollutantsasriskfactors
for neurodegenerative disorders: Alzheimer and parkinson diseases. Frontiers in cellular neuroscience,
9:124.
Cho, S.-J., Park, M. H., Han, C., Yoon, K., and Koh, Y. H. (2017). Vegfr2 alteration in alzheimer’s disease.
Scientic reports , 7(1):1–11.
Choi, H., Choi, Y., Lee, E. J., Kim, H., Lee, Y., Kwon, S., Hwang, D. W., and Lee, D. S. (2021). Hippocam-
pal glucose uptake as a surrogate of metabolic change of microglia in alzheimer’s disease. Journal of
neuroinammation , 18(1):1–15.
Corlier,F.,Hafzalla,G.,Faskowitz,J.,Kuller,L.H.,Becker,J.T.,Lopez,O.L.,Thompson,P.M.,andBraskie,
M. N. (2018). Systemic inammation as a predictor of brain aging: Contributions of physical activity,
metabolic risk, and genetic risk. Neuroimage, 172:118–129.
Cortes-Canteli, M., Gispert, J. D., Salvadó, G., Toribio-Fernandez, R., Tristão-Pereira, C., Falcon, C., Oliva,
B., Mendiguren, J., Fernandez-Friera, L., Sanz, J., et al. (2021). Subclinical atherosclerosis and brain
metabolism in middle-aged individuals: the pesa study. Journal of the American College of Cardiology,
77(7):888–898.
Crane, P. K., Carle, A., Gibbons, L. E., Insel, P., Mackin, R. S., Gross, A., Jones, R. N., Mukherjee, S., Curtis,
S. M., Harvey, D., et al. (2012). Development and assessment of a composite score for memory in the
alzheimer’s disease neuroimaging initiative (adni). Brain imaging and behavior, 6(4):502–516.
Cunningham,K.S.andGotlieb,A.I.(2005). Theroleofshearstressinthepathogenesisofatherosclerosis.
Laboratory investigation, 85(1):9–23.
Dadar,M.,Maranzano,J.,Misquitta,K.,Anor,C.J.,Fonov,V.S.,Tartaglia,M.C.,Carmichael,O.T.,Decarli,
C., Collins, D. L., Initiative, A. D. N., et al. (2017). Performance comparison of 10 dierent classication
techniques in segmenting white matter hyperintensities in aging. NeuroImage, 157:233–249.
Dale,A.M.,Fischl,B.,andSereno,M.I.(1999). Corticalsurface-basedanalysis: I.segmentationandsurface
reconstruction. Neuroimage, 9(2):179–194.
De Leeuw, F.-E., De Groot, J., Bots, M., Witteman, J., Oudkerk, M., Hofman, A., Van Gijn, J., and Breteler,
M. (2000). Carotid atherosclerosis and cerebral white matter lesions in a population based magnetic
resonance imaging study. Journal of neurology, 247(4):291–296.
Debette,S.,Beiser,A.,Homann,U.,DeCarli,C.,O’Donnell,C.J.,Massaro,J.M.,Au,R.,Himali,J.J.,Wolf,
P. A., Fox, C. S., et al. (2010). Visceral fat is associated with lower brain volume in healthy middle-aged
adults. Annals of neurology, 68(2):136–144.
122
Debette, S. and Markus, H. (2010). The clinical importance of white matter hyperintensities on brain
magnetic resonance imaging: systematic review and meta-analysis. Bmj, 341.
DeCarli, C., Murphy, D., Tranh, M. a., Grady, C., Haxby, J., Gillette, J., Salerno, J., Gonzales-Aviles, A.,
Honvitz, B., Rapoport, S., et al. (1995). The eect of white matter hyperintensity volume on brain
structure, cognitive performance, and cerebral metabolism of glucose in 51 healthy adults. Neurology,
45(11):2077–2084.
Denswil,N.P.,vanderWal,A.C.,Ritz,K.,deBoer,O.J.,Aronica,E.,Troost,D.,andDaemen,M.J.(2016).
Atherosclerosis in the circle of willis: spatial dierences in composition and in distribution of plaques.
Atherosclerosis, 251:78–84.
Desgranges, B., Baron, J.-C., and Eustache, F. (1998). The functional neuroanatomy of episodic memory:
the role of the frontal lobes, the hippocampal formation, and other areas. Neuroimage, 8(2):198–213.
Desikan,R.S.,Ségonne,F.,Fischl,B.,Quinn,B.T.,Dickerson,B.C.,Blacker,D.,Buckner,R.L.,Dale,A.M.,
Maguire, R. P., Hyman, B. T., et al. (2006). An automated labeling system for subdividing the human
cerebral cortex on mri scans into gyral based regions of interest. Neuroimage, 31(3):968–980.
Dickerson, B. C., Bakkour, A., Salat, D. H., Feczko, E., Pacheco, J., Greve, D. N., Grodstein, F., Wright,
C. I., Blacker, D., Rosas, H. D., et al. (2009). The cortical signature of alzheimer’s disease: regionally
speciccorticalthinningrelatestosymptomseverityinverymildtomildaddementiaandisdetectable
in asymptomatic amyloid-positive individuals. Cerebral cortex, 19(3):497–510.
Ding, F., Yao, J., Rettberg, J. R., Chen, S., and Brinton, R. D. (2013). Early decline in glucose transport and
metabolismprecedesshifttoketogenicsysteminfemaleagingandalzheimer’smousebrain: implication
for bioenergetic intervention. PloS one, 8(11):e79977.
Dolui,S.,Li,Z.,Nasrallah,I.M.,Detre,J.A.,andWolk,D.A.(2020).Arterialspinlabelingversus18f-fdg-pet
to identify mild cognitive impairment. NeuroImage: Clinical, 25:102146.
Dowling, N. M., Johnson, S. C., Gleason, C. E., Jagust, W. J., Initiative, A. D. N., et al. (2015). The medi-
ational eects of fdg hypometabolism on the association between cerebrospinal uid biomarkers and
neurocognitive function. Neuroimage, 105:357–368.
Drzezga, A., Lautenschlager, N., Siebner, H., Riemenschneider, M., Willoch, F., Minoshima, S., Schwaiger,
M.,andKurz,A.(2003). Cerebralmetabolicchangesaccompanyingconversionofmildcognitiveimpair-
mentintoalzheimer’sdisease: apetfollow-upstudy. Europeanjournalofnuclearmedicineandmolecular
imaging, 30(8):1104–1113.
Duarte-Guterman, P., Albert, A. Y., Inkster, A. M., Barha, C. K., Galea, L. A., Initiative, A. D. N., et al.
(2020). Inammation in alzheimer’s disease: do sex and apoe matter? Journal of Alzheimer’s Disease,
78(2):627–641.
Dufouil, C., de Kersaint-Gilly, A., Besancon, V., Levy, C., Auray, E., Brunnereau, L., Alperovitch, A., and
Tzourio,C.(2001). Longitudinalstudyofbloodpressureandwhitematterhyperintensities: theevamri
cohort. Neurology, 56(7):921–926.
Edison, P., Archer, H., Hinz, R., Hammers, A., Pavese, N., Tai, Y., Hotton, G., Cutler, D., Fox, N., Kennedy,
A., et al. (2007). Amyloid, hypometabolism, and cognition in alzheimer disease: an [11c] pib and [18f]
fdg pet study. Neurology, 68(7):501–508.
123
Fabel, K., Fabel, K., Tam, B., Kaufer, D., Baiker, A., Simmons, N., Kuo, C. J., and Palmer, T. D. (2003). Vegf
is necessary for exercise-induced adult hippocampal neurogenesis. European Journal of Neuroscience,
18(10):2803–2812.
Fernandes, A., Tábuas-Pereira, M., Duro, D., Lima, M., Gens, H., Santiago, B., Durães, J., Almeida, M. R.,
Leitão, M. J., Baldeiras, I., et al. (2020). C-reactive protein as a predictor of mild cognitive impairment
conversion into alzheimer’s disease dementia. Experimental gerontology, 138:111004.
Fernando,M.,O’brien,J.,Perry,R.,English,P.,Forster,G.,McMeekin,W.,Slade,J.,Golkhar,A.,Matthews,
F.,Barber,R.,etal.(2004).Comparisonofthepathologyofcerebralwhitematterwithpost-mortemmag-
neticresonanceimaging(mri)intheelderlybrain. Neuropathologyandappliedneurobiology,30(4):385–
395.
Firbank,M.J.,Lloyd,A.J.,Ferrier,N.,andO’Brien,J.T.(2004).Avolumetricstudyofmrisignalhyperinten-
sitiesinlate-lifedepression. TheAmericanjournalofgeriatricpsychiatry: ocialjournaloftheAmerican
Association for Geriatric Psychiatry, 12(6):606–612.
Fischl, B. (2012). Freesurfer. Neuroimage, 62(2):774–781.
Fischl, B. and Dale, A. M. (2000). Measuring the thickness of the human cerebral cortex from magnetic
resonance images. Proceedings of the National Academy of Sciences, 97(20):11050–11055.
Fischl, B., Salat, D. H., Busa, E., Albert, M., Dieterich, M., Haselgrove, C., Van Der Kouwe, A., Killiany, R.,
Kennedy,D.,Klaveness,S.,etal.(2002).Wholebrainsegmentation: automatedlabelingofneuroanatom-
ical structures in the human brain. Neuron, 33(3):341–355.
Fournier,N.M.,Lee,B.,Banasr,M.,Elsayed,M.,andDuman,R.S.(2012). Vascularendothelialgrowthfac-
tor regulates adult hippocampal cell proliferation through mek/erk-and pi3k/akt-dependent signaling.
Neuropharmacology, 63(4):642–652.
Frisoni,G.B.,Galluzzi,S.,Pantoni,L.,andFilippi,M.(2007). Theeectofwhitematterlesionsoncognition
in the elderly—small but detectable. Nature Clinical Practice Neurology, 3(11):620–627.
Gao, Y.-L., Wang, N., Sun, F.-R., Cao, X.-P., Zhang, W., and Yu, J.-T. (2018). Tau in neurodegenerative
disease. Annals of translational medicine, 6(10).
Garde, E., Mortensen, E. L., Krabbe, K., Rostrup, E., and Larsson, H. B. (2000). Relation between age-
related decline in intelligence and cerebral white-matter hyperintensities in healthy octogenarians: a
longitudinal study. The Lancet, 356(9230):628–634.
Gaubert,M.,Lange,C.,Garnier-Crussard,A.,Köbe,T.,Bougacha,S.,Gonneaud,J.,deFlores,R.,Tomadesso,
C., Mézenge, F., Landeau, B., et al. (2021). Topographic patterns of white matter hyperintensities are
associated with multimodal neuroimaging biomarkers of alzheimer’s disease. Alzheimer’s Research &
Therapy, 13(1):1–11.
Ghaari,S.,Leask,R.L.,andJones,E.A.(2017).Bloodowcansignalduringangiogenesisnotonlythrough
mechanotransduction, but also by aecting growth factor distribution.Angiogenesis, 20(3):373–384.
Gibbons, L. E., Carle, A. C., Mackin, R. S., Harvey, D., Mukherjee, S., Insel, P., Curtis, S. M., Mungas, D.,
and Crane, P. K. (2012). A composite score for executive functioning, validated in alzheimer’s disease
neuroimaginginitiative(adni)participantswithbaselinemildcognitiveimpairment. Brainimagingand
behavior, 6(4):517–527.
124
Gibson,E.,Gao,F.,Black,S.E.,andLobaugh,N.J.(2010). Automaticsegmentationofwhitematterhyper-
intensitiesintheelderlyusingairimagesat3t.Journalofmagneticresonanceimaging,31(6):1311–1322.
Glodzik, L., Kuceyeski, A., Rusinek, H., Tsui, W., Mosconi, L., Li, Y., Osorio, R., Williams, S., Randall,
C., Spector, N., et al. (2014). Reduced glucose uptake and a in brain regions with hyperintensities in
connected white matter. Neuroimage, 100:684–691.
Gómez-Isla,T.,Price,J.L.,McKeelJr,D.W.,Morris,J.C.,Growdon,J.H.,andHyman,B.T.(1996). Profound
lossoflayeriientorhinalcortexneuronsoccursinverymildalzheimer’sdisease.JournalofNeuroscience,
16(14):4491–4500.
Gootjes, L., Teipel, S., Zebuhr, Y., Schwarz, R., Leinsinger, G., Scheltens, P., Möller, H.-J., and Hampel, H.
(2004). Regionaldistributionofwhitematterhyperintensitiesinvasculardementia,alzheimer’sdisease
and healthy aging. Dementia and geriatric cognitive disorders, 18(2):180–188.
Góra-Kupilas, K. and Jośko, J. (2005). The neuroprotective function of vascular endothelial growth factor
(vegf). Folia Neuropathologica, 43(1):31–39.
Gordon, B. A., Najmi, S., Hsu, P., Roe, C. M., Morris, J. C., and Benzinger, T. L. (2015). The eects of
white matter hyperintensities and amyloid deposition on alzheimer dementia. NeuroImage: Clinical,
8:246–252.
Govindpani,K.,McNamara,L.G.,Smith,N.R.,Vinnakota,C.,Waldvogel,H.J.,Faull,R.L.,andKwakowsky,
A.(2019). Vasculardysfunctioninalzheimer’sdisease: apreludetothepathologicalprocessoraconse-
quence of it? Journal of clinical medicine, 8(5):651.
Goyal, M. S., Hawrylycz, M., Miller, J. A., Snyder, A. Z., and Raichle, M. E. (2014). Aerobic glycolysis
in the human brain is associated with development and neotenous gene expression. Cell metabolism,
19(1):49–57.
Grianti, L., Jenkinson, M., Suri, S., Zsoldos, E., Mahmood, A., Filippini, N., Sexton, C. E., Topiwala, A.,
Allan, C., Kivimäki, M., et al. (2018). Classication and characterization of periventricular and deep
white matter hyperintensities on mri: a study in older adults. Neuroimage, 170:174–181.
Grianti,L.,Zamboni,G.,Khan,A.,Li,L.,Bonifacio,G.,Sundaresan,V.,Schulz,U.G.,Kuker,W.,Battaglini,
M., Rothwell, P. M., et al. (2016). Bianca (brain intensity abnormality classication algorithm): A new
tool for automated segmentation of white matter hyperintensities. Neuroimage, 141:191–205.
Grimaud, J., Lai, M., Thorpe, J., Adeleine, P., Wang, L., Barker, G., Plummer, D., Tofts, P., McDonald, W.,
and Miller, D. (1996). Quantication of mri lesion load in multiple sclerosis: a comparison of three
computer-assisted techniques. Magnetic resonance imaging, 14(5):495–505.
Grimmer,T.,Faust,M.,Auer,F.,Alexopoulos,P.,Förstl,H.,Henriksen,G.,Perneczky,R.,Sorg,C.,Youse,
B. H., Drzezga, A., et al. (2012). White matter hyperintensities predict amyloid increase in alzheimer’s
disease. Neurobiology of aging, 33(12):2766–2773.
Gunning-Dixon,F.M.andRaz,N.(2000). Thecognitivecorrelatesofwhitematterabnormalitiesinnormal
aging: a quantitative review. Neuropsychology, 14(2):224.
Gurol, M. E., Viswanathan, A., Gidicsin, C., Hedden, T., Martinez-Ramirez, S., Dumas, A., Vashkevich, A.,
Ayres,A.M.,Auriel,E.,vanEtten,E.,etal.(2013). Cerebralamyloidangiopathyburdenassociatedwith
leukoaraiosis: apositronemissiontomography/magneticresonanceimagingstudy. Annalsofneurology,
73(4):529–536.
125
Habeck, C., Risacher, S., Lee, G. J., Glymour, M. M., Mormino, E., Mukherjee, S., Kim, S., Nho, K., DeCarli,
C.,Saykin,A.J.,etal.(2012). Relationshipbetweenbaselinebrainmetabolismmeasuredusing[18f]fdg
petandmemoryandexecutivefunctioninprodromalandearlyalzheimer’sdisease. Brainimagingand
behavior, 6(4):568–583.
Habes, M., Erus, G., Toledo, J. B., Zhang, T., Bryan, N., Launer, L. J., Rosseel, Y., Janowitz, D., Doshi, J.,
VanderAuwera,S.,etal.(2016). Whitematterhyperintensitiesandimagingpatternsofbrainageingin
the general population. Brain, 139(4):1164–1179.
Haller, S., Kövari, E., Herrmann, F. R., Cuvinciuc, V., Tomm, A.-M., Zulian, G. B., Lovblad, K.-O., Gian-
nakopoulos, P., and Bouras, C. (2013). Do brain t2/air white matter hyperintensities correspond to
myelin loss in normal aging? a radiologic-neuropathologic correlation study. Acta neuropathologica
communications, 1(1):1–7.
Hane, F. T., Robinson, M., Lee, B. Y., Bai, O., Leonenko, Z., and Albert, M. S. (2017). Recent progress in
alzheimer’sdiseaseresearch,part3: diagnosisandtreatment. JournalofAlzheimer’sDisease,57(3):645–
665.
Hattori,Y.,Suzuki,K.,Hattori,S.,andKasai,K.(2006). Metformininhibitscytokine-inducednuclearfactor
b activation via amp-activated protein kinase activation in vascular endothelial cells. Hypertension,
47(6):1183–1188.
Hedden, T., Mormino, E. C., Amariglio, R. E., Younger, A. P., Schultz, A. P., Becker, J. A., Buckner, R. L.,
Johnson, K. A., Sperling, R. A., and Rentz, D. M. (2012). Cognitive prole of amyloid burden and white
matterhyperintensitiesincognitivelynormalolderadults. JournalofNeuroscience,32(46):16233–16242.
Hedden,T.,Oh,H.,Younger,A.P.,andPatel,T.A.(2013). Meta-analysisofamyloid-cognitionrelationsin
cognitively normal older adults. Neurology, 80(14):1341–1348.
Hilal, S., Ikram, M. A., Verbeek, M. M., Franco, O. H., Stoops, E., Vanderstichele, H., Niessen, W. J., and
Vernooij,M.W.(2018). C-reactiveprotein,plasmaamyloid- levels,andtheirinteractionwithmagnetic
resonance imaging markers. Stroke, 49(11):2692–2698.
Ho, J. K., Nation, D. A., Initiative, A. D. N., et al. (2018). Neuropsychological proles and trajectories in
preclinical alzheimer’s disease. Journal of the International Neuropsychological Society, 24(7):693–702.
Hobel, Z., Isenberg, A. L., Raghupathy, D., Mack, W., Pa, J., Initiative, A. D. N., et al. (2019). Apoe ε4
gene dose and sex eects on alzheimer’s disease mri biomarkers in older adult s with mild cognitive
impairment. Journal of Alzheimer’s Disease, 71(2):647–658.
Hocaoglu-Emre, F. S., Saribal, D., Yenmis, G., and Guvenen, G. (2017). Vascular cell adhesion molecule 1,
intercellularadhesionmolecule1,andclusterofdierentiation146levelsinpatientswithtype2diabetes
with complications. Endocrinology and Metabolism, 32(1):99–105.
Hodges, J. R. and Patterson, K. (1995). Is semantic memory consistently impaired early in the course of
alzheimer’s disease? neuroanatomical and diagnostic implications. Neuropsychologia, 33(4):441–459.
Hohman, T. J., Bell, S. P., Jeerson, A. L., Initiative, A. D. N., et al. (2015). The role of vascular endothelial
growth factor in neurodegeneration and cognitive decline: exploring interactions with biomarkers of
alzheimer disease. JAMA neurology, 72(5):520–529.
Holtzman, D. M. (2011). Csf biomarkers for alzheimer’s disease: current utility and potential future use.
Neurobiology of aging, 32:S4–S9.
126
Holtzman, D. M., Herz, J., and Bu, G. (2012). Apolipoprotein e and apolipoprotein e receptors: normal
biology and roles in alzheimer disease. Cold Spring Harbor perspectives in medicine, 2(3):a006312.
Hu, Y., Xu, Q., Li, K., Zhu, H., Qi, R., Zhang, Z., and Lu, G. (2013). Gender dierences of brain glucose
metabolic networks revealed by fdg-pet: evidence from a large cohort of 400 young adults. PLoS One,
8(12):e83821.
Huang, C.-W., Hsu, S.-W., Chang, Y.-T., Huang, S.-H., Huang, Y.-C., Lee, C.-C., Chang, W.-N., Lui, C.-C.,
Chen,N.-C.,andChang,C.-C.(2018). Cerebralperfusioninsuciencyandrelationshipswithcognitive
decits in alzheimer’s disease: a multiparametric neuroimaging study.Scientic reports , 8(1):1–14.
Iadecola, C. (2004). Neurovascular regulation in the normal brain and in alzheimer’s disease. Nature
Reviews Neuroscience, 5(5):347–360.
Iorio,M., Spalletta,G., Chiapponi,C., Luccichenti,G., Cacciari,C., Orfei,M.D., Caltagirone,C., andPiras,
F. (2013). White matter hyperintensities segmentation: a new semi-automated method. Frontiers in
aging neuroscience, 5:76.
Jack, C. R., Wiste, H. J., Therneau, T. M., Weigand, S. D., Knopman, D. S., Mielke, M. M., Lowe, V. J.,
Vemuri, P., Machulda, M. M., Schwarz, C. G., et al. (2019). Associations of amyloid, tau, and neurode-
generationbiomarkerproleswithratesofmemorydeclineamongindividualswithoutdementia.Jama,
321(23):2316–2325.
Jack Jr, C. R., Bennett, D. A., Blennow, K., Carrillo, M. C., Dunn, B., Haeberlein, S. B., Holtzman, D. M., Ja-
gust,W.,Jessen,F.,Karlawish,J.,etal.(2018). Nia-aaresearchframework: towardabiologicaldenition
of alzheimer’s disease. Alzheimer’s & Dementia, 14(4):535–562.
JackJr,C.R.andHoltzman,D.M.(2013). Biomarkermodelingofalzheimer’sdisease. Neuron,80(6):1347–
1358.
Jagust, W. J., Bandy, D., Chen, K., Foster, N. L., Landau, S. M., Mathis, C. A., Price, J. C., Reiman, E. M.,
Skovronsky, D., Koeppe, R. A., et al. (2010). The alzheimer’s disease neuroimaging initiative positron
emission tomography core. Alzheimer’s & Dementia, 6(3):221–229.
Jais, A., Solas, M., Backes, H., Chaurasia, B., Kleinridders, A., Theurich, S., Mauer, J., Steculorum, S. M.,
Hampel,B.,Goldau,J.,etal.(2016). Myeloid-cell-derivedvegfmaintainsbrainglucoseuptakeandlimits
cognitive impairment in obesity. Cell, 165(4):882–895.
Janelidze,S.,Mattsson,N.,Stomrud,E.,Lindberg,O.,Palmqvist,S.,Zetterberg,H.,Blennow,K.,andHans-
son,O.(2018). Csfbiomarkersofneuroinammationandcerebrovasculardysfunctioninearlyalzheimer
disease. Neurology, 91(9):e867–e877.
Jeerakathil, T., Wolf, P. A., Beiser, A., Massaro, J., Seshadri, S., D’Agostino, R. B., and DeCarli, C.
(2004). Stroke risk prole predicts white matter hyperintensity volume: the framingham study.Stroke,
35(8):1857–1861.
Jenkinson, M., Bannister, P., Brady, M., and Smith, S. (2002). Improved optimization for the robust and
accurate linear registration and motion correction of brain images. Neuroimage, 17(2):825–841.
Jenkinson, M.,Beckmann,C. F.,Behrens,T.E.,Woolrich,M.W.,andSmith,S.M.(2012). Fsl. Neuroimage,
62(2):782–790.
127
Jenkinson, M. and Smith, S. (2001). A global optimisation method for robust ane registration of brain
images. Medical image analysis, 5(2):143–156.
Jiaerken,Y.,Luo,X.,Yu,X.,Huang,P.,Xu,X.,Zhang,M.,and(ADNI),A.D.N.I.(2019).Microstructuraland
metabolicchangesinthelongitudinalprogressionofwhitematterhyperintensities. JournalofCerebral
Blood Flow & Metabolism, 39(8):1613–1622.
Jiang, T. and Cadenas, E. (2014). Astrocytic metabolic and inammatory changes as a function of age.
Aging cell, 13(6):1059–1067.
Joshi,A.,Koeppe,R.A.,andFessler,J.A.(2009). Reducingbetweenscannerdierencesinmulti-centerpet
studies. Neuroimage, 46(1):154–159.
Jurado, M. B. and Rosselli, M. (2007). The elusive nature of executive functions: A review of our current
understanding. Neuropsychology review, 17(3):213–233.
Kalheim,L.F.,Selnes,P.,Bjørnerud,A.,Coello,C.,Vegge,K.,andFladby,T.(2016).Amyloiddysmetabolism
relates to reduced glucose uptake in white matter hyperintensities. Frontiers in Neurology, 7:209.
Kantarci,K.,Boeve,B.F.,Przybelski,S.A.,Lesnick,T.G.,Chen,Q.,Fields,J.,Schwarz,C.G.,Senjem,M.L.,
Gunte, J. L., Jack, C. R., et al. (2021). Fdg pet metabolic signatures distinguishing prodromal dlb and
prodromal ad. NeuroImage: Clinical, 31:102754.
Kapeller,P.,Barber,R.,Vermeulen,R.,Ader,H.,Scheltens,P.,Freidl,W.,Almkvist,O.,Moretti,M.,DelSer,
T., Vaghfeldt, P., et al. (2003). Visual rating of age-related white matter changes on magnetic reso-
nance imaging: scale comparison, interrater agreement, and correlations with quantitative measure-
ments. Stroke, 34(2):441–445.
Kato,T.,Inui,Y.,Nakamura,A.,andIto,K.(2016). Brainuorodeoxyglucose(fdg)petindementia.Ageing
research reviews, 30:73–84.
Kaur,B.,Himali,J.J.,Seshadri,S.,Beiser,A.S.,Au,R.,McKee,A.C.,Auerbach,S.,Wolf,P.A.,andDeCarli,
C. S. (2014). Association between neuropathology and brain volume in the framingham heart study.
Alzheimer disease and associated disorders, 28(3):219.
Kim, K. W., MacFall, J. R., and Payne, M. E. (2008). Classication of white matter lesions on magnetic
resonance imaging in elderly persons. Biological psychiatry, 64(4):273–280.
Kim, S., Kim, M. J., Kim, S., Kang, H. S., Lim, S. W., Myung, W., Lee, Y., Hong, C. H., Choi, S. H., Na,
D. L., et al. (2015). Gender dierences in risk factors for transition from mild cognitive impairment to
alzheimer’s disease: A credos study. Comprehensive Psychiatry, 62:114–122.
Kimura, K., Hashiguchi, T., Deguchi, T., Horinouchi, S., Uto, T., Oku, H., Setoyama, S., Maruyama, I.,
Osame,M.,andArimura,K.(2007).Serumvegf—asaprognosticfactorofatherosclerosis.Atherosclerosis,
194(1):182–188.
Kisler, K., Nelson, A. R., Montagne, A., and Zlokovic, B. V. (2017). Cerebral blood ow regulation and
neurovascular dysfunction in alzheimer disease. Nature Reviews Neuroscience, 18(7):419–434.
Kloppenborg,R.P.,Nederkoorn,P.J.,Geerlings,M.I.,andvandenBerg,E.(2014).Presenceandprogression
of white matter hyperintensities and cognition: a meta-analysis. Neurology, 82(23):2127–2138.
128
Konijnenberg, E., Tijms, B. M., Gobom, J., Dobricic, V., Bos, I., Vos, S., Tsolaki, M., Verhey, F., Popp, J.,
Martinez-Lage, P., et al. (2020). Apoe"4 genotype-dependent cerebrospinaluid proteomic signatures
in alzheimer’s disease. Alzheimer’s Research & Therapy, 12(1):1–11.
Krishnadas, N., Villemagne, V. L., Doré, V., and Rowe, C. C. (2021). Advances in brain amyloid imaging.
Seminars in Nuclear Medicine, 51(3):241–252.
Lampe, L., Kharabian-Masouleh, S., Kynast, J., Arelin, K., Steele, C. J., Löer, M., Witte, A. V., Schroeter,
M. L., Villringer, A., and Bazin, P.-L. (2017). Lesion location matters: the relationships between white
matterhyperintensitiesoncognitioninthehealthyelderly. JournalofCerebralBloodFlow&Metabolism,
39(1):36–43.
Landau,S.M.,Mintun,M.A.,Joshi,A.D.,Koeppe,R.A.,Petersen,R.C.,Aisen,P.S.,Weiner,M.W.,Jagust,
W. J., and Initiative, A. D. N. (2012). Amyloid deposition, hypometabolism, and longitudinal cognitive
decline. Annals of neurology, 72(4):578–586.
Lange, C., Storkebaum, E., De Almodóvar, C. R., Dewerchin, M., and Carmeliet, P. (2016). Vascular en-
dothelial growth factor: a neurovascular target in neurological diseases. Nature Reviews Neurology,
12(8):439–454.
Lee,S.-J.,Kim,J.-S.,Chung,S.-W.,Kim,B.-S.,Ahn,K.-J.,andLee,K.-S.(2011).Whitematterhyperintensities
(wmh)areassociatedwithintracranialatherosclerosisratherthanextracranialatherosclerosis. Archives
of gerontology and geriatrics, 53(2):e129–e132.
Lehoux, S. and Jones, E. A. (2016). Shear stress, arterial identity and atherosclerosis. Thrombosis and
haemostasis, 115(03):467–473.
Leung,Y.Y.,Toledo,J.B.,Nefedov,A.,Polikar,R.,Raghavan,N.,Xie,S.X.,Farnum,M.,Schultz,T.,Baek,Y.,
Van Deerlin, V. M., et al. (2015). Identifying amyloid pathology–related cerebrospinaluid biomarkers
foralzheimer’sdiseaseinamulticohortstudy. Alzheimer’s&Dementia: Diagnosis,Assessment&Disease
Monitoring, 1(3):339–348.
Levin,F.,Ferreira,D.,Lange,C.,Dyrba,M.,Westman,E.,Buchert,R.,Teipel,S.J.,andGrothe,M.J.(2021).
Data-drivenfdg-petsubtypesofalzheimer’sdisease-relatedneurodegeneration. Alzheimer’sresearch&
therapy, 13(1):1–14.
Li, Y.-S. J., Haga, J. H., and Chien, S. (2005). Molecular basis of the eects of shear stress on vascular
endothelial cells. Journal of biomechanics, 38(10):1949–1971.
Licht, T., Goshen, I., Avital, A., Kreisel, T., Zubedat, S., Eavri, R., Segal, M., Yirmiya, R., and Keshet, E.
(2011). Reversible modulations of neuronal plasticity by vegf. Proceedings of the National Academy of
Sciences, 108(12):5081–5086.
Lo,R.Y.andJagust,W.J.(2012). Vascularburdenandalzheimerdiseasepathologicprogression. Neurology,
79(13):1349–1355.
Longstreth,W.,Manolio,T.A.,Arnold,A.,Burke,G.L.,Bryan,N.,Jungreis,C.A.,Enright,P.L.,O’Leary,D.,
andFried,L.(1996). Clinicalcorrelatesofwhitematterndingsoncranialmagneticresonanceimaging
of 3301 elderly people: the cardiovascular health study. Stroke, 27(8):1274–1282.
Lourenço,C.F.,Ledo,A.,Barbosa,R.M.,andLaranjinha,J.(2017). Neurovascular-neuroenergeticcoupling
axis in the brain: master regulation by nitric oxide and consequences in aging and neurodegeneration.
Free Radical Biology and Medicine, 108:668–682.
129
Low, A., Mak, E., Rowe, J. B., Markus, H. S., and O’Brien, J. T. (2019). Inammation and cerebral small
vessel disease: a systematic review. Ageing research reviews, 53:100916.
Loy, C. T., Schoeld, P. R., Turner, A. M., and Kwok, J. B. (2014). Genetics of dementia. The Lancet,
383(9919):828–840.
Maclin,J.M.A.,Wang,T.,andXiao,S.(2019).Biomarkersforthediagnosisofalzheimer’sdisease,dementia
lewy body, frontotemporal dementia and vascular dementia. General psychiatry, 32(1).
Makkar, S. R., Lipnicki, D. M., Crawford, J. D., Kochan, N. A., Castro-Costa, E., Lima-Costa, M. F., Diniz,
B.S., Brayne,C., Stephan,B., Matthews,F.,etal.(2020). Apoe"4andtheinuenceofsex,age,vascular
riskfactors,andethnicityoncognitivedecline. The Journals of Gerontology: Series A,75(10):1863–1873.
Maksimowicz-McKinnon, K., Bhatt, D. L., and Calabrese, L. H. (2004). Recent advances in vascular in-
ammation: C-reactive protein and other inammatory biomarkers.Current opinion in rheumatology,
16(1):18–24.
Malpas, C. B., Saling, M. M., Velakoulis, D., Desmond, P., Hicks, R. J., Zetterberg, H., Blennow, K., and
O’Brien,T.J.(2018). cerebrospinaluidbiomarkersaredierentiallyrelatedtostructuralandfunctional
changes in dementia of the alzheimer’s type. Journal of Alzheimer’s Disease, 62(1):417–427.
Mantyla, R., Erkinjuntti, T., Salonen, O., Aronen, H. J., Peltonen, T., Pohjasvaara, T., and Standertskjold-
Nordenstam, C.-G. (1997). Variable agreement between visual rating scales for white matter hyperin-
tensities on mri: comparison of 13 rating scales in a poststroke cohort. Stroke, 28(8):1614–1623.
Marchitelli,R.,Aiello,M.,Cachia,A.,Quarantelli,M.,Cavaliere,C.,Postiglione,A.,Tedeschi,G.,Montella,
P., Milan, G., Salvatore, M., et al. (2018a). Simultaneous resting-state fdg-pet/fmri in alzheimer disease:
relationship between glucose metabolism and intrinsic activity. Neuroimage, 176:246–258.
Marchitelli,R.,Aiello,M.,Cachia,A.,Quarantelli,M.,Cavaliere,C.,Postiglione,A.,Tedeschi,G.,Montella,
P., Milan, G., Salvatore, M., et al. (2018b). Simultaneous resting-state fdg-pet/fmri in alzheimer disease:
relationship between glucose metabolism and intrinsic activity. Neuroimage, 176:246–258.
Márquez,F.andYassa,M.A.(2019). Neuroimagingbiomarkersforalzheimer’sdisease. Molecularneurode-
generation, 14(1):1–14.
Martin, L., Bouvet, P., Chounlamountri, N., Watrin, C., Besançon, R., Pinatel, D., Meyronet, D., Honnorat,
J.,Buisson,A.,Salin,P.-A.,etal.(2021). Vegfcounteractsamyloid- -inducedsynapticdysfunction. Cell
Reports, 35(6):109121.
McKhann, G., Drachman, D., Folstein, M., Katzman, R., Price, D., and Stadlan, E. M. (1984). Clinical diag-
nosisofalzheimer’sdisease: Reportofthenincds-adrdaworkgroup*undertheauspicesofdepartment
of health and human services task force on alzheimer’s disease. Neurology, 34(7):939–939.
Melrose, R. J., Campa, O. M., Harwood, D. G., Osato, S., Mandelkern, M. A., and Sultzer, D. L. (2009). The
neural correlates of naming anduency decits in alzheimer’s disease: An fdg-pet study.International
Journal of Geriatric Psychiatry: A journal of the psychiatry of late life and allied sciences, 24(8):885–893.
Michaelson, D. (2016). Apoe4: The most prevalent yet understudied risk factor for alzheimer’s disease.
Neurobiology of Aging, 1(39):S24.
Moore, A. M., Mahoney, E., Dumitrescu, L., De Jager, P. L., Koran, M. E. I., Petyuk, V. A., Robinson, R. A.,
Ruderfer,D.M.,Cox,N.J.,Schneider,J.A.,etal.(2020). Apoe"4-specicassociationsofvegfgenefamily
expression with cognitive aging and alzheimer’s disease. Neurobiology of aging, 87:18–25.
130
Mortamais, M., Ash, J. A., Harrison, J., Kaye, J., Kramer, J., Randolph, C., Pose, C., Albala, B., Ropacki, M.,
Ritchie, C. W., et al. (2017). Detecting cognitive changes in preclinical alzheimer’s disease: A review of
its feasibility. Alzheimer’s & dementia, 13(4):468–492.
Mosconi, L. (2005). Brain glucose metabolism in the early and specic diagnosis of alzheimer’s disease.
European journal of nuclear medicine and molecular imaging, 32(4):486–510.
Mosconi, L. (2013). Glucose metabolism in normal aging and alzheimer’s disease: methodological and
physiological considerations for pet studies. Clinical and translational imaging, 1(4):217–233.
Mosconi, L., Berti, V., Guyara-Quinn, C., McHugh, P., Petrongolo, G., Osorio, R. S., Connaughty, C., Pupi,
A., Vallabhajosula, S., Isaacson, R. S., et al. (2017). Perimenopause and emergence of an alzheimer’s
bioenergetic phenotype in brain and periphery. PloS one, 12(10):e0185926.
Mosconi, L., Mistur, R., Switalski, R., Tsui, W. H., Glodzik, L., Li, Y., Pirraglia, E., De Santi, S., Reisberg,
B.,Wisniewski,T.,etal.(2009). Fdg-petchangesinbrainglucosemetabolismfromnormalcognitionto
pathologicallyveriedalzheimer’sdisease.Europeanjournalofnuclearmedicineandmolecularimaging,
36(5):811–822.
Mosconi, L., Pupi, A., and De Leon, M. J. (2008). Brain glucose hypometabolism and oxidative stress in
preclinical alzheimer’s disease. Annals of the New York Academy of Sciences, 1147(1):180–195.
Mosconi, L., Rahman, A., Diaz, I., Wu, X., Scheyer, O., Hristov, H. W., Vallabhajosula, S., Isaacson, R. S.,
de Leon, M. J., and Brinton, R. D. (2018). Increased alzheimer’s risk during the menopause transition: a
3-year longitudinal brain imaging study. PloS one, 13(12):e0207885.
Mrak, R. E., Grin, W. S. T., and Graham, D. I. (1997). Aging-associated changes in human brain. Journal
of Neuropathology & Experimental Neurology, 56(12):1269–1275.
Mummery, C. J., Patterson, K., Hodges, J. R., and Wise, R. J. (1996). Generating ‘tiger’as an animal name
or a word beginning with t: dierences in brain activation.Proceedings of the Royal Society of London.
Series B: Biological Sciences, 263(1373):989–995.
Murray, M. E., Senjem, M. L., Petersen, R. C., Hollman, J. H., Preboske, G. M., Weigand, S. D., Knopman,
D. S., Ferman, T. J., Dickson, D. W., and Jack, C. R. (2010). Functional impact of white matter hyperin-
tensities in cognitively normal elderly subjects. Archives of neurology, 67(11):1379–1385.
Musiek, E. S., Chen, Y., Korczykowski, M., Saboury, B., Martinez, P. M., Reddin, J. S., Alavi, A., Kimberg,
D. Y., Wolk, D. A., Julin, P., et al. (2012). Direct comparison of uorodeoxyglucose positron emission
tomography and arterial spin labeling magnetic resonance imaging in alzheimer’s disease. Alzheimer’s
& Dementia, 8(1):51–59.
Nam, K.-W., Kwon, H.-M., Jeong, H.-Y., Park, J.-H., Kim, S. H., Jeong, S.-M., Yoo, T. G., and Kim, S. (2017).
Cerebral white matter hyperintensity is associated with intracranial atherosclerosis in a healthy popu-
lation. Atherosclerosis, 265:179–183.
Nelson, A. R., Sweeney, M. D., Sagare, A. P., and Zlokovic, B. V. (2016). Neurovascular dysfunction and
neurodegenerationindementiaandalzheimer’sdisease. BiochimicaetBiophysicaActa(BBA)-Molecular
Basis of Disease, 1862(5):887–900.
Nielsen, R. B., Egefjord, L., Angleys, H., Mouridsen, K., Gejl, M., Møller, A., Brock, B., Brændgaard, H.,
Gottrup, H., Rungby, J., et al. (2017). Capillary dysfunction is associated with symptom severity and
neurodegeneration in alzheimer’s disease. Alzheimer’s & Dementia, 13(10):1143–1153.
131
No, H.-J., Yi, H.-A., Won, K. S., Chang, H. W., and Kim, H. W. (2019). Association between white matter
lesions and cerebral glucose metabolism in patients with cognitive impairment. Revista Española de
Medicina Nuclear e Imagen Molecular (English Edition), 38(3):160–166.
O’brien,R.J.andWong,P.C.(2011).Amyloidprecursorproteinprocessingandalzheimer’sdisease.Annual
review of neuroscience, 34:185–204.
Okamura, N., Harada, R., Furukawa, K., Furumoto, S., Tago, T., Yanai, K., Arai, H., and Kudo, Y. (2016).
Advances in the development of tau pet radiotracers and their clinical applications. Ageing research
reviews, 30:107–113.
Ossenkoppele,R.,Smith,R.,Ohlsson,T.,Strandberg,O.,Mattsson,N.,Insel,P.S.,Palmqvist,S.,andHans-
son,O.(2019). Associationsbetweentau,a ,andcorticalthicknesswithcognitioninalzheimerdisease.
Neurology, 92(6):e601–e612.
Pagani, M., De Carli, F., Morbelli, S., Öberg, J., Chincarini, A., Frisoni, G., Galluzzi, S., Perneczky, R.,
Drzezga, A., Van Berckel, B., et al. (2015). Volume of interest-based [18f]uorodeoxyglucose pet dis-
criminatesmciconvertingtoalzheimer’sdiseasefromhealthycontrols.aeuropeanalzheimer’sdisease
consortium (eadc) study. NeuroImage: Clinical, 7:34–42.
Park,J.-H.,Kwon,H.-M.,Lee,J.,Kim,D.-S.,andOvbiagele,B.(2015).Associationofintracranialatheroscle-
roticstenosiswithseverityofwhitematterhyperintensities. EuropeanJournalofNeurology,22(1):44–e3.
Passow, S., Specht, K., Adamsen, T. C., Biermann, M., Brekke, N., Craven, A. R., Ersland, L., Grüner, R.,
Kleven-Madsen,N.,Kvernenes,O.-H.,etal.(2015).Acloselinkbetweenmetabolicactivityandfunctional
connectivity in the resting human brain. EJNMMI physics, 2(1):1–2.
Paterson, R., Bartlett, J., Blennow, K., Fox, N., Shaw, L., Trojanowski, J., Zetterberg, H., and Schott, J.
(2014). Cerebrospinaluid markers including trefoil factor 3 are associated with neurodegeneration in
amyloid-positive individuals. Translational psychiatry, 4(7):e419–e419.
Peca, S., McCreary, C. R., Donaldson, E., Kumarpillai, G., Shobha, N., Sanchez, K., Charlton, A., Steinback,
C. D., Beaudin, A. E., Flück, D., et al. (2013). Neurovascular decoupling is associated with severity of
cerebral amyloid angiopathy. Neurology, 81(19):1659–1665.
Perani, D. (2014). Fdg-pet and amyloid-pet imaging: the diverging paths. Current Opinion in Neurology,
27(4):405–413.
Peter,J.,Kaiser,J.,Landerer,V.,Köstering,L.,Kaller,C.P.,Heimbach,B.,Hüll,M.,Bormann,T.,andKlöppel,
S. (2016). Category and designuency in mild cognitive impairment: performance, strategy use, and
neural correlates. Neuropsychologia, 93:21–29.
Pico, F., Dufouil, C., Lévy, C., Besançon, V., de Kersaint-Gilly, A., Bonithon-Kopp, C., Ducimetière, P.,
Tzourio, C., and Alpérovitch, A. (2002). Longitudinal study of carotid atherosclerosis and white matter
hyperintensities: the eva-mri cohort. Cerebrovascular diseases, 14(2):109–115.
Pillai, J. A., Maxwell, S., Bena, J., Bekris, L. M., Rao, S. M., Chance, M., Lamb, B. T., Leverenz, J. B., and
Initiative,A.D.N.(2019). Keyinammatorypathwayactivationsinthemcistageofalzheimer’sdisease.
Annals of clinical and translational neurology, 6(7):1248–1262.
Plascencia-Villa,G.andPerry,G.(2020).Statusandfuturedirectionsofclinicaltrialsinalzheimer’sdisease.
International review of neurobiology, 154:3–50.
132
Prins,N.,VanStraaten,E.,VanDijk,E.,Simoni,M.,VanSchijndel,R.,Vrooman,H.,Koudstaal,P.,Scheltens,
P., Breteler, M., and Barkhof, F. (2004). Measuring progression of cerebral white matter lesions on mri:
visual rating and volumetrics. Neurology, 62(9):1533–1539.
Prins, N. D. and Scheltens, P. (2015). White matter hyperintensities, cognitive impairment and dementia:
an update. Nature Reviews Neurology, 11(3):157–165.
Provenzano, F. A., Muraskin, J., Tosto, G., Narkhede, A., Wasserman, B. T., Grith, E. Y., Guzman, V. A.,
Meier,I.B.,Zimmerman,M.E.,Brickman,A.M.,etal.(2013).Whitematterhyperintensitiesandcerebral
amyloidosis: necessary and sucient for clinical expression of alzheimer disease? JAMA neurology,
70(4):455–461.
Provias,J.andJeynes,B.(2014). Reductioninvascularendothelialgrowthfactorexpressioninthesuperior
temporal, hippocampal, and brainstem regions in alzheimer’s disease. Current neurovascular research,
11(3):202–209.
Querbes, O., Aubry, F., Pariente, J., Lotterie, J.-A., Démonet, J.-F., Duret, V., Puel, M., Berry, I., Fort, J.-C.,
Celsis,P.,etal.(2009).Earlydiagnosisofalzheimer’sdiseaseusingcorticalthickness: impactofcognitive
reserve. Brain, 132(8):2036–2047.
Raichle, M. E., MacLeod, A. M., Snyder, A. Z., Powers, W. J., Gusnard, D. A., and Shulman, G. L. (2001). A
default mode of brain function. Proceedings of the National Academy of Sciences, 98(2):676–682.
Randolph, L. N., Bao, X., Oddo, M., and Lian, X. L. (2019). Sex-dependent vegf expression underlies varia-
tionsinhumanpluripotentstemcelltoendothelialprogenitordierentiation.Scienticreports ,9(1):1–8.
Raz, N., Yang, Y., Dahle, C. L., and Land, S. (2012). Volume of white matter hyperintensities in healthy
adults: contributionofage,vascularriskfactors,andinammation-relatedgeneticvariants.Biochimica
et Biophysica Acta (BBA)-Molecular Basis of Disease, 1822(3):361–369.
Religa,P.,Cao,R.,Religa,D.,Xue,Y.,Bogdanovic,N.,Westaway,D.,Marti,H.H.,Winblad,B.,andCao,Y.
(2013). Vegf signicantly restores impaired memory behavior in alzheimer’s mice by improvement of
vascular survival. Scientic reports , 3(1):1–9.
Rezaie-Majd,A.,Prager,G.W.,Bucek,R.A.,Schernthaner,G.H.,Maca,T.,Kress,H.-G.,Valent,P.,Binder,
B.R.,Minar,E.,andBaghestanian,M.(2003). Simvastatinreducestheexpressionofadhesionmolecules
in circulating monocytes from hypercholesterolemic patients. Arteriosclerosis, thrombosis, and vascular
biology, 23(3):397–403.
Riedel, B. C., Thompson, P. M., and Brinton, R. D. (2016). Age, apoe and sex: triad of risk of alzheimer’s
disease. The Journal of steroid biochemistry and molecular biology, 160:134–147.
Risacher,S.L.,Anderson,W.H.,Charil,A.,Castelluccio,P.F.,Shcherbinin,S.,Saykin,A.J.,Schwarz,A.J.,
Initiative,A.D.N.,etal.(2017). Alzheimerdiseasebrainatrophysubtypesareassociatedwithcognition
and rate of decline. Neurology, 89(21):2176–2186.
Rivard,A.,Berthou-Soulie,L.,Principe,N.,Kearney,M.,Curry,C.,Branellec,D.,Semenza,G.L.,andIsner,
J. M. (2000). Age-dependent defect in vascular endothelial growth factor expression is associated with
reduced hypoxia-inducible factor 1 activity. Journal of Biological Chemistry, 275(38):29643–29647.
Roher, A. E., Tyas, S. L., Maarouf, C. L., Daugs, I. D., Kokjohn, T. A., Emmerling, M. R., Garami, Z., Be-
lohlavek,M.,Sabbagh,M.N.,Sue,L.I.,etal.(2011). Intracranialatherosclerosisasacontributingfactor
to alzheimer’s disease dementia. Alzheimer’s & Dementia, 7(4):436–444.
133
Rojas, S., Brugulat-Serrat, A., Bargallo, N., Minguillon, C., Tucholka, A., Falcon, C., Car-
valho, A., Morán, S., Esteller, M., Gramunt, N., et al. (2018). Higher prevalence of
cerebral white matter hyperintensities in homozygous apoe-"4allelecarriersaged45 75 :
Resultsfromthealfastudy.JournalofCerebralBloodFlow&Metabolism,38(2) : 250 261.
Roseborough, A., Ramirez, J., Black, S. E., and Edwards, J. D. (2017). Associations between amyloid and
white matter hyperintensities: a systematic review. Alzheimer’s & Dementia, 13(10):1154–1167.
Rouhl, R. P., Damoiseaux, J. G., Lodder, J., Theunissen, R. O., Knottnerus, I. L., Staals, J., Henskens, L. H.,
Kroon, A. A., de Leeuw, P. W., Tervaert, J. W. C., et al. (2012). Vascular inammation in cerebral small
vessel disease. Neurobiology of aging, 33(8):1800–1806.
Ruan,Q.,D’Onofrio,G.,Sancarlo,D.,Bao,Z.,Greco,A.,andYu,Z.(2016). Potentialneuroimagingbiomarkers
of pathologic brain changes in mild cognitive impairment and alzheimer’s disease: a systematic review.
BMC geriatrics, 16(1):1–9.
Rutkowski,J.M.andSwartz,M.A.(2007). Adrivingforceforchange: interstitialowasamorphoregulator.
Trends in cell biology, 17(1):44–50.
Sainson,R.C.,Johnston,D.A.,Chu,H.C.,Holdereld,M.T.,Nakatsu,M.N.,Crampton,S.P.,Davis,J.,Conn,
E., and Hughes, C. C. (2008). Tnf primes endothelial cells for angiogenic sprouting by inducing a tip cell
phenotype. Blood, The Journal of the American Society of Hematology, 111(10):4997–5007.
Sala,A.,Caprioglio,C.,Santangelo,R.,Vanoli,E.G.,Iannaccone,S.,Magnani,G.,andPerani,D.(2020). Brain
metabolic signatures across the alzheimer’s disease spectrum. European journal of nuclear medicine and
molecular imaging, 47(2):256–269.
Salminen, L. E., Tubi, M. A., Bright, J., Thomopoulos, S. I., Wieand, A., and Thompson, P. M. (2022). Sex is a
deningfeatureofneuroimagingphenotypesinmajorbraindisorders.HumanBrainMapping,43(1):500–
542.
Santos, C. Y., Snyder, P. J., Wu, W.-C., Zhang, M., Echeverria, A., and Alber, J. (2017). Pathophysiologic
relationship between alzheimer’s disease, cerebrovascular disease, and cardiovascular risk: a review and
synthesis. Alzheimer’s & Dementia: Diagnosis, Assessment & Disease Monitoring, 7:69–87.
Saykin,A.J.,Shen,L.,Foroud,T.M.,Potkin,S.G.,Swaminathan,S.,Kim,S.,Risacher,S.L.,Nho,K.,Huentel-
man,M.J.,Craig,D.W.,etal.(2010). Alzheimer’sdiseaseneuroimaginginitiativebiomarkersasquantita-
tive phenotypes: Genetics core aims, progress, and plans. Alzheimer’s & dementia, 6(3):265–273.
Schirmer,M.D.,Giese,A.-K.,Fotiadis,P.,Etherton,M.R.,Cloonan,L.,Viswanathan,A.,Greenberg,S.M.,Wu,
O.,andRost,N.S.(2019). Spatialsignatureofwhitematterhyperintensitiesinstrokepatients. Frontiersin
neurology, 10:208.
Schmidt, P. (2017). Bayesian inference for structured additive regression models for large-scale problems with
applications to medical imaging. PhD thesis, lmu.
Schmidt,P.,Gaser,C.,Arsic,M.,Buck,D.,Förschler,A.,Berthele,A.,Hoshi,M.,Ilg,R.,Schmid,V.J.,Zimmer,
C., et al. (2012). An automated tool for detection of air-hyperintense white-matter lesions in multiple
sclerosis. Neuroimage, 59(4):3774–3783.
Schmidt-Arras, D. and Rose-John, S. (2016). Il-6 pathway in the liver: from physiopathology to therapy.
Journal of hepatology, 64(6):1403–1415.
134
Schreiner, S. J., Kirchner, T., Narkhede, A., Wyss, M., Van Bergen, J. M., Steininger, S. C., Gietl, A., Leh, S. E.,
Treyer, V., Buck, A., et al. (2018). Brain amyloid burden and cerebrovascular disease are synergistically
associated with neurometabolism in cognitively unimpaired older adults. Neurobiology of Aging, 63:152–
161.
Scott, J. A., Braskie, M. N., Tosun, D., Maillard, P., Thompson, P. M., Weiner, M., DeCarli, C., Carmichael,
O. T., et al. (2016). Cerebral amyloid is associated with greater white-matter hyperintensity accrual in
cognitively normal older adults. Neurobiology of aging, 48:48–52.
Scott, J. A., Braskie, M. N., Tosun, D., Thompson, P. M., Weiner, M., DeCarli, C., Carmichael, O. T., Initiative,
A.D.N.,etal.(2015). Cerebralamyloidandhypertensionareindependentlyassociatedwithwhitematter
lesions in elderly. Frontiers in aging neuroscience, 7:221.
Shaw, L., Vanderstichele, H., Knapik-Czajka, M., Clark, C., Aisen, P., Petersen, R., Blennow, K., Soares, H.,
Simon, A., Lewczuk, P., et al. (2009a). Cerebrospinal uid biomarker signature in alzheimer’s disease
neuroimaging initiative subjects. Ann Neurol, 65(4):403–13.
Shaw, L. M., Vanderstichele, H., Knapik-Czajka, M., Clark, C. M., Aisen, P. S., Petersen, R. C., Blennow, K.,
Soares, H., Simon, A., Lewczuk, P., et al. (2009b). Cerebrospinaluid biomarker signature in alzheimer’s
disease neuroimaging initiative subjects. Annals of neurology, 65(4):403–413.
Shaw, L. M., Vanderstichele, H., Knapik-Czajka, M., Figurski, M., Coart, E., Blennow, K., Soares, H., Simon,
A. J., Lewczuk, P., Dean, R. A., et al. (2011). Qualication of the analytical and clinical performance of csf
biomarker analyses in adni. Acta neuropathologica, 121(5):597–609.
Shinohara,M.,Murray,M.E.,Frank,R.D.,Shinohara,M.,DeTure,M.,Yamazaki,Y.,Tachibana,M.,Atagi,Y.,
Davis,M.D.,Liu,C.-C.,etal.(2016).Impactofsexandapoe4oncerebralamyloidangiopathyinalzheimer’s
disease. Acta neuropathologica, 132(2):225–234.
Silva, M. V. F., Loures, C. d. M. G., Alves, L. C. V., de Souza, L. C., Borges, K. B. G., and Carvalho, M. d. G.
(2019). Alzheimer’sdisease: riskfactorsandpotentiallyprotectivemeasures. Journalofbiomedicalscience,
26(1):1–11.
Simpson, J., Fernando, M., Clark, L., Ince, P., Matthews, F., Forster, G., O’brien, J., Barber, R., Kalaria, R.,
Brayne,C.,etal.(2007). Whitematterlesionsinanunselectedcohortoftheelderly: astrocytic,microglial
and oligodendrocyte precursor cell responses. Neuropathology and applied neurobiology, 33(4):410–419.
Singh, T. and Newman, A. B. (2011). Inammatory markers in population studies of aging.Ageing research
reviews, 10(3):319–329.
Smart, S. D., Firbank, M. J., and O’Brien, J. T. (2011). Validation of automated white matter hyperintensity
segmentation. Journal of aging research, 2011.
Smith,E.,Salat,D.,Jeng,J.,McCreary,C.,Fischl,B.,Schmahmann,J.,Dickerson,B.,Viswanathan,A.,Albert,
M.,Blacker,D.,etal.(2011). Correlationsbetweenmriwhitematterlesionlocationandexecutivefunction
and episodic memory. Neurology, 76(17):1492–1499.
Smith, S. M., Jenkinson, M., Woolrich, M. W., Beckmann, C. F., Behrens, T. E., Johansen-Berg, H., Bannister,
P.R.,DeLuca,M.,Drobnjak,I.,Flitney,D.E.,etal.(2004). Advancesinfunctionalandstructuralmrimage
analysis and implementation as fsl. Neuroimage, 23:S208–S219.
Soehnlein,O.andLibby,P.(2021). Targetinginammationinatherosclerosis—fromexperimentalinsightsto
the clinic. Nature Reviews Drug Discovery, 20(8):589–610.
135
Soldan, A., Pettigrew, C., Zhu, Y., Wang, M.-C., Moghekar, A., Gottesman, R. F., Singh, B., Martinez, O.,
Fletcher, E., DeCarli, C., etal. (2020). White matter hyperintensities andcsf alzheimer disease biomarkers
in preclinical alzheimer disease. Neurology, 94(9):e950–e960.
Sone, H., Deo, B. K., and Kumagai, A. K. (2000). Enhancement of glucose transport by vascular endothelial
growth factor in retinal endothelial cells. Investigative ophthalmology & visual science, 41(7):1876–1884.
Spellman, D. S., Wildsmith, K. R., Honigberg, L. A., Tueerd, M., Baker, D., Raghavan, N., Nairn, A. C.,
Croteau, P., Schirm, M., Allard, R., et al. (2015). Development and evaluation of a multiplexed mass spec-
trometry based assay for measuring candidate peptide biomarkers in alzheimer’s disease neuroimaging
initiative (adni) csf. PROTEOMICS–Clinical Applications, 9(7-8):715–731.
Spuch, C., Antequera, D., Portero, A., Orive, G., Hernández, R. M., Molina, J. A., Bermejo-Pareja, F., Pedraz,
J.L.,andCarro,E.(2010). Theeectofencapsulatedvegf-secretingcellsonbrainamyloidloadandbehav-
ioral impairment in a mouse model of alzheimer’s disease. Biomaterials, 31(21):5608–5618.
Sudre, C. H., Cardoso, M. J., Frost, C., Barnes, J., Barkhof, F., Fox, N., Ourselin, S., Initiative, A. D. N., et al.
(2017). Apoe "4 status is associated with white matter hyperintensities volume accumulation rate inde-
pendent of ad diagnosis. Neurobiology of aging, 53:67–75.
Sun, Y., Jin, K., Xie, L., Childs, J., Mao, X. O., Logvinova, A., Greenberg, D. A., et al. (2003). Vegf-induced
neuroprotection, neurogenesis, and angiogenesis after focal cerebral ischemia. The Journal of clinical in-
vestigation, 111(12):1843–1851.
Sweeney,M.D.,Montagne,A.,Sagare,A.P.,Nation,D.A.,Schneider,L.S.,Chui,H.C.,Harrington,M.G.,Pa,
J.,Law,M.,Wang,D.J.,etal.(2019). Vasculardysfunction—thedisregardedpartnerofalzheimer’sdisease.
Alzheimer’s & Dementia, 15(1):158–167.
Tai, L. M., Thomas, R., Marottoli, F. M., Koster, K. P., Kanekiyo, T., Morris, A. W., and Bu, G. (2016). The role
of apoe in cerebrovascular dysfunction. Acta neuropathologica, 131(5):709–723.
Tang, H., Mao, X., Xie, L., Greenberg, D. A., and Jin, K. (2013). Expression level of vascular endothelial
growthfactorinhippocampusisassociatedwithcognitiveimpairmentinpatientswithalzheimer’sdisease.
Neurobiology of aging, 34(5):1412–1415.
Tao, Q., Ang, T. F. A., DeCarli, C., Auerbach, S. H., Devine, S., Stein, T. D., Zhang, X., Massaro, J., Au, R., and
Qiu, W. Q. (2018). Association of chronic low-grade inammation with risk of alzheimer disease in apoe4
carriers. JAMA network open, 1(6):e183597–e183597.
Tarkowski, E., Issa, R., Sjögren, M., Wallin, A., Blennow, K., Tarkowski, A., and Kumar, P. (2002). Increased
intrathecal levels of the angiogenic factors vegf and tgf- in alzheimer’s disease and vascular dementia.
Neurobiology of aging, 23(2):237–243.
ten Dam, V. H., van den Heuvel, D. M., de Craen, A. J., Bollen, E. L., Murray, H. M., Westendorp, R. G.,
Blauw,G.J.,andvanBuchem,M.A.(2007a). Declineintotalcerebralbloodowislinkedwithincreasein
periventricular but not deep white matter hyperintensities. Radiology, 243(1):198–203.
ten Dam, V. H., van den Heuvel, D. M., de Craen, A. J., Bollen, E. L., Murray, H. M., Westendorp, R. G.,
Blauw, G. J., and van Buchem, M. A. (2007b). Decline in total cerebral bloodow is linked with increase
in periventricular but not deep white matter hyperintensities. Radiology, 243(1):198–203.
136
Tubi,M.A.,Feingold,F.W.,Kothapalli,D.,Hare,E.T.,King,K.S.,Thompson,P.M.,andBraskie,M.N.(2020).
Whitematterhyperintensitiesandtheirrelationshiptocognition: Eectsofsegmentationalgorithm.Neu-
roImage, 206:116327.
Tubi, M. A., Kothapalli, D., Hapenney, M., Feingold, F. W., Mack, W. J., King, K. S., Thompson, P. M., and
Braskie, M. N. (2021). Regional relationships between CSF VEGF levels and Alzheimer’s disease brain
biomarkers and cognition. Neurobiology of Aging, 105:241–251.
Tublin, J. M., Adelstein, J. M., Del Monte, F., Combs, C. K., and Wold, L. E. (2019). Getting to the heart of
alzheimer disease. Circulation research, 124(1):142–149.
Tustison, N. J., Avants, B. B., Cook, P. A., Zheng, Y., Egan, A., Yushkevich, P. A., and Gee, J. C. (2010). N4itk:
improved n3 bias correction. IEEE transactions on medical imaging, 29(6):1310–1320.
U.S. Census Bureau (2021). 2014 national population projections datasets.
Uysal, N., Kiray, M., Sisman, A., Camsari, U., Gencoglu, C., Baykara, B., Cetinkaya, C., and Aksu, I. (2015).
Eectsofvoluntaryandinvoluntaryexerciseoncognitivefunctions,andvegfandbdnflevelsinadolescent
rats. Biotechnic & Histochemistry, 90(1):55–68.
Valdés Hernández, M. d. C., González-Castro, V., Ghandour, D. T., Wang, X., Doubal, F., Muñoz Maniega, S.,
Armitage, P. A., and Wardlaw, J. M. (2016). On the computational assessment of white matter hyperin-
tensityprogression: dicultiesinmethodselectionandbiaseldcorrectionperformanceonimageswith
signicant white matter pathology.Neuroradiology, 58(5):475–485.
Van Dyken, P. and Lacoste, B. (2018). Impact of metabolic syndrome on neuroinammation and the blood–
brain barrier. Frontiers in neuroscience, page 930.
vanHaelst,P.L.,Liem,A.,vanBoven,A.J.,Veeger,N.J.,vanVeldhuisen,D.J.,Tervaert,J.W.C.,Gans,R.O.,
and Zijlstra, F. (2003). Usefulness of elevated neopterin and c-reactive protein levels in predicting cardio-
vascular events in patients with non–q-wave myocardial infarction. The American journal of cardiology,
92(10):1201–1203.
Van Straaten, E. C., Fazekas, F., Rostrup, E., Scheltens, P., Schmidt, R., Pantoni, L., Inzitari, D., Waldemar,
G., Erkinjuntti, T., Mantyla, R., et al. (2006). Impact of white matter hyperintensities scoring method on
correlations with clinical data: the ladis study. Stroke, 37(3):836–840.
Viboolvorakul, S. and Patumraj, S. (2014). Exercise training could improve age-related changes in cerebral
bloodow and capillary vascularity through the upregulation of vegf and enos.BioMed research interna-
tional, 2014.
Vlassenko, A. G., Vaishnavi, S. N., Couture, L., Sacco, D., Shannon, B. J., Mach, R. H., Morris, J. C., Raichle,
M. E., and Mintun, M. A. (2010). Spatial correlation between brain aerobic glycolysis and amyloid- (a )
deposition. Proceedings of the National Academy of Sciences, 107(41):17763–17767.
Walker, K. A., Windham, B. G., Power, M. C., Hoogeveen, R. C., Folsom, A. R., Ballantyne, C. M., Knopman,
D. S., Selvin, E., Jack Jr, C. R., and Gottesman, R. F. (2018). The association of mid-to late-life systemic
inammation with white matter structure in older adults: The atherosclerosis risk in communities study.
Neurobiology of aging, 68:26–33.
Wang, L., Benzinger, T. L., Hassenstab, J., Blazey, T., Owen, C., Liu, J., Fagan, A. M., Morris, J. C., and Ances,
B. M. (2015a). Spatially distinct atrophy is linked to -amyloid and tau in preclinical alzheimer disease.
Neurology, 84(12):1254–1260.
137
Wang, P., Xie, Z.-H., Guo, Y.-J., Zhao, C.-P., Jiang, H., Song, Y., Zhu, Z.-Y., Lai, C., Xu, S.-L., and Bi, J.-Z.
(2011). Vegf-inducedangiogenesisamelioratesthememoryimpairmentinapptransgenicmousemodelof
alzheimer’s disease. Biochemical and biophysical research communications, 411(3):620–626.
Wang, R., Li, C., Wang, J., Wei, X., Li, Y., Zhu, Y., and Zhang, S. (2015b). Automatic segmentation and
volumetric quantication of white matter hyperintensities onuid-attenuated inversion recovery images
using the extreme value distribution. Neuroradiology, 57(3):307–320.
Wang,Y.,Grydeland,H.,Roe,J.M.,Pan,M.,Magnussen,F.,Amlien,I.K.,Watne,L.O.,Idland,A.-V.,Bertram,
L.,Gundersen,T.E.,etal.(2022).Associationsofcirculatingc-reactiveproteins,apoe"4,andbrainmarkers
foralzheimer’sdiseaseinhealthysamplesacrossthelifespan. Brain,behavior,andimmunity,100:243–253.
Wang, Z., Zhou, W., Zhou, B., Zhang, J., Initiative, A. D. N., et al. (2018). Association of vascular endothe-
lial growth factor levels in csf and cerebral glucose metabolism across the alzheimer’s disease spectrum.
Neuroscience Letters, 687:276–279.
Wardlaw, J. M., Valdés Hernández, M. C., and Muñoz-Maniega, S. (2015). What are white matter hyperin-
tensitiesmadeof? relevancetovascularcognitiveimpairment. Journal of the American Heart Association,
4(6):e001140.
Wersching, H., Duning, T., Lohmann, H., Mohammadi, S., Stehling, C., Fobker, M., Conty, M., Minnerup,
J., Ringelstein, E., Berger, K., et al. (2010). Serum c-reactive protein is linked to cerebral microstructural
integrity and cognitive function. Neurology, 74(13):1022–1029.
Wharton, S. B., Simpson, J. E., Brayne, C., and Ince, P. G. (2015). Age-associated white matter lesions: The
mrc c ognitive f unction and a geing s tudy. Brain pathology, 25(1):35–43.
Wilcock, G. and Esiri, M. (1982). Plaques, tangles and dementia: a quantitative study. Journal of the neuro-
logical sciences, 56(2-3):343–356.
Wilkins,H.M.,Carl,S.M.,Greenlief,A.,Festo,B.W.,andSwerdlow,R.H.(2014). Bioenergeticdysfunction
and inammation in alzheimer’s disease: a possible connection.Frontiers in aging neuroscience, 6:311.
Wiseman,S.J.,Booth,T.,Ritchie,S.J.,Cox,S.R.,MuñozManiega,S.,ValdésHernández,M.d.C.,Dickie,D.A.,
Royle, N. A., Starr, J. M., Deary, I. J., et al. (2018). Cognitive abilities, brain white matter hyperintensity
volume, and structural network connectivity in older age. Human Brain Mapping, 39(2):622–632.
Wolters,F.J.,Boender,J.,deVries,P.S.,Sonneveld,M.A.,Koudstaal,P.J.,deMaat,M.P.,Franco,O.H.,Ikram,
M. K., Leebeek, F. W., and Ikram, M. A. (2018). Von willebrand factor and adamts13 activity in relation to
risk of dementia: a population-based study. Scientic reports , 8(1):1–10.
Wyman,B.T.,Harvey,D.J.,Crawford,K.,Bernstein,M.A.,Carmichael,O.,Cole,P.E.,Crane,P.K.,DeCarli,
C., Fox, N. C., Gunter, J. L., et al. (2013). Standardization of analysis sets for reporting results from adni
mri data. Alzheimer’s & Dementia, 9(3):332–337.
Xiang, X., Wind, K., Wiedemann, T., Blume, T., Shi, Y., Briel, N., Beyer, L., Biechele, G., Eckenweber, F.,
Zatcepin, A., et al. (2021). Microglial activation states drive glucose uptake and fdg-pet alterations in
neurodegenerative diseases. Science translational medicine, 13(615):eabe5640.
Yang,J.,Yao,Y.,Chen,T.,andZhang,T.(2014). Vegfamelioratescognitiveimpairmentininvivoandinvitro
ischemia via improving neuronal viability and function. Neuromolecular medicine, 16(2):376–388.
138
Yang,S.-P.,Bae,D.-G.,Kang,H.J.,Gwag,B.J.,Gho,Y.S.,andChae,C.-B.(2004). Co-accumulationofvascular
endothelial growth factor with -amyloid in the brain of patients with alzheimer’s disease. Neurobiology
of aging, 25(3):283–290.
Yanovsky, I., Leow, A. D., Lee, S., Osher, S. J., and Thompson, P. M. (2009). Comparing registration methods
for mapping brain change using tensor-based morphometry. Medical image analysis, 13(5):679–700.
Yanovsky,I.,Thompson,P.M.,Osher,S.,andLeow,A.D.(2008). Asymmetricandsymmetricunbiasedimage
registration: statisticalassessmentofperformance. In2008IEEEComputerSocietyConferenceonComputer
Vision and Pattern Recognition Workshops, pages 1–8. IEEE.
Yao,J.,Irwin,R.W.,Zhao,L.,Nilsen,J.,Hamilton,R.T.,andBrinton,R.D.(2009). Mitochondrialbioenergetic
decit precedes alzheimer’s pathology in female mouse model of alzheimer’s disease. Proceedings of the
National Academy of Sciences, 106(34):14670–14675.
Yekutieli,D.andBenjamini,Y.(1999). Resampling-basedfalsediscoveryratecontrollingmultipletestproce-
dures for correlated test statistics. Journal of Statistical Planning and Inference, 82(1-2):171–196.
Yin, F., Sancheti, H., Patil, I., and Cadenas, E. (2016). Energy metabolism and inammation in brain aging
and alzheimer’s disease. Free Radical Biology and Medicine, 100:108–122.
Ylikoski, A., Erkinjuntti, T., Raininko, R., Sarna, S., Sulkava, R., and Tilvis, R. (1995). White matter hyperin-
tensitiesonmriintheneurologicallynondiseasedelderly: analysisofcohortsofconsecutivesubjectsaged
55 to 85 years living at home. Stroke, 26(7):1171–1177.
Yoshita,M.,Fletcher,E.,andDeCarli,C.(2005). Currentconceptsofanalysisofcerebralwhitematterhyper-
intensities on magnetic resonance imaging. Topics in magnetic resonance imaging: TMRI, 16(6):399.
Yoshita, M., Fletcher, E., Harvey, D., Ortega, M., Martinez, O., Mungas, D. M., Reed, B. R., and DeCarli
(2006). Extent and distribution of white matter hyperintensities in normal aging, mci, and ad. Neurol-
ogy, 67(12):2192–2198.
Zacchigna, S., Lambrechts, D., and Carmeliet, P. (2008). Neurovascular signalling defects in neurodegenera-
tion. Nature Reviews Neuroscience, 9(3):169–181.
Zhang, L., Zhou, Q., Shao, L. H., Wen, J., and Xia, J. (2021). Association of carotid atherosclerosis with
white matter hyperintensity in an asymptomatic japanese population: A cross-sectional study. Frontiers
in cardiovascular medicine, 8:366.
Zhao,F.,Deng,J.,Yu,X.,Li,D.,Shi,H.,andZhao,Y.(2015). Protectiveeectsofvascularendothelialgrowth
factor in cultured brain endothelial cells against hypoglycemia. Metabolic Brain Disease, 30(4):999–1007.
Zhou,R.,Chen,H.,Ye,F.,Huang,S.,andZhang,J.(2020). Inuenceofhypertensiononlongitudinalchanges
inbrainglucosemetabolismwasmodiedbytheapoe4alleleamongcognitivelynormalolderindividuals.
Frontiers in Aging Neuroscience, 12:85.
Zimmer, E. R., Parent, M. J., Souza, D. G., Leuzy, A., Lecrux, C., Kim, H.-I., Gauthier, S., Pellerin, L., Hamel,
E., and Rosa-Neto, P. (2017). [18 f] fdg pet signal is driven by astroglial glutamate transport. Nature
neuroscience, 20(3):393–395.
Zlokovic, B. V. (2011). Neurovascular pathways to neurodegeneration in alzheimer’s disease and other dis-
orders. Nature Reviews Neuroscience, 12(12):723–738.
139
Abstract (if available)
Abstract
My dissertation aims to clarify how vascular dysfunction contributes to suboptimal brain aging in older adults at risk for Alzheimer’s disease (AD). AD is a multifactorial and heterogeneous disease, with increasing evidence demonstrating that vascular dysfunction contributes to AD and dementia risk. By the time an individual has evidence of cognitive decline and/or an AD diagnosis, the brain has already undergone severe structural and functional degeneration, making it essential to assess vascular factors that may contribute to early AD brain dysfunction. While AD neuroimaging brain biomarkers (cortical thinning, brain FDG-PET) fluctuate early and often follow a specific spatial and temporal sequence in AD disease course, heterogeneity still exists. Critically, sources of the heterogeneity in AD neuroimaging phenotype patterns have yet to be unmasked and few studies have mapped how vascular factors regionally relate to AD-brain biomarkers and cognition. To address this, I leveraged the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset (ages 55-90) to investigate how markers of vascular dysfunction (e.g., white matter hyperintensities (WMH) and cerebral spinal fluid (CSF) vascular endothelial growth factor (VEGF)) relate to brain aging biomarkers and cognition in individuals across the AD continuum. The work unveils sources of variability in AD biomarkers by demonstrating both independent and interactive relationships between vascular dysfunction and AD neuropathology on brain health. Collectively, unmasking vascular contributions to AD can help promote deeper phenotyping of disease to enhance precision medicine and intervention efforts targeted at those who are most at risk for cognitive decline.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Using neuroinformatics to identify genomic and proteomic markers of suboptimal aging and Alzheimer's disease
PDF
Neuroimaging markers of risk & resilience to brain aging and dementia
PDF
Insulin sensitivity in cognition, Alzheimer's disease and brain aging
PDF
Alzheimer’s disease: dysregulated genes, ethno-racial disparities, and environmental pollution
PDF
Age-related androgen depletion and the development of Alzheimer's disease
PDF
Bioenergetic fuel shift of reproductive aging: implication for late-onset Alzheimer's disease
PDF
Blood-brain barrier pathophysiology in cognitive impairment and injury
PDF
Relationships between lifetime chronic stress exposure, vascular risk, cognition, and brain structure in HIV
PDF
Estrogen and progesterone-based hormone therapy and the development of Alzheimer's disease
PDF
Blood-brain barrier breakdown and vessel integrity in Alzheimer’s disease
PDF
Affective neuropsychiatric symptoms and neural connectivity in the early stages of Alzheimer’s disease
PDF
Longitudinal neurocognitive profiles of empirically-derived Alzheimer’s disease variants
PDF
Brainstem structural integrity in the progression of Alzheimer's disease
PDF
Characterizing brain aging with neuroimaging, health, and genetic data
PDF
Neuroinflammation and ApoE4 genotype in at-risk female aging: implications for Alzheimer's disease
PDF
Long-term blood pressure variability across the clinical and biomarker spectrum of Alzheimer’s disease
PDF
The role of the locus coeruleus in Alzheimer’s disease and cerebrovascular function: insights from neuroimaging, neuropsychology, and biofluid markers
PDF
The association of cerebrovascular disease risk factors with brain structure and its modification by genetic variation
PDF
Development of biomarker profiles for early detection of women with an at-risk for Alzheimer's disease phenotype
PDF
Discovery of mitochondria-DNA-encoded microproteins in neurodegeneration
Asset Metadata
Creator
Tubi, Meral Ayten
(author)
Core Title
Vascular contributions to brain aging along the Alzheimer's disease continuum
School
Keck School of Medicine
Degree
Doctor of Philosophy
Degree Program
Neuroscience
Degree Conferral Date
2022-08
Publication Date
07/19/2022
Defense Date
05/31/2022
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
aging,Alzheimer's disease,brain,cortical thickness,metabolic,MRI,neuroimaging,OAI-PMH Harvest,PET,vascular,vascular endothelial growth factor,white matter hyperintensities
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Thompson, Paul (
committee chair
), Braskie, Meredith (
committee member
), Chui, Helena (
committee member
), Mack, Wendy (
committee member
)
Creator Email
meral.tubi@gmail.com,tubi@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-oUC111373229
Unique identifier
UC111373229
Legacy Identifier
etd-TubiMeralA-10864
Document Type
Dissertation
Format
application/pdf (imt)
Rights
Tubi, Meral Ayten
Type
texts
Source
20220719-usctheses-batch-956
(batch),
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Repository Email
cisadmin@lib.usc.edu
Tags
Alzheimer's disease
brain
cortical thickness
metabolic
MRI
neuroimaging
PET
vascular
vascular endothelial growth factor
white matter hyperintensities