mapping neurodegenerative disease onset and...

Mapping Neurodegenerative DiseaseOnset and Progression

William W. Seeley

Memory and Aging Center, Departments of Neurology and Pathology, University of California,San Francisco, California 94143

Correspondence: [email protected]

Brain networks have been of long-standing interest to neurodegeneration researchers, in-cluding but not limited to investigators focusing on conventional prion diseases, which areknown to propagate along neural pathways. Tools for human network mapping, however,remained inadequate, limiting our understanding of human brain network architecture andpreventing clinical research applications. Until recently, neuropathological studies were theonly viable approach to mapping disease onset and progression in humans but required largeautopsy cohorts and laborious methods for whole-brain sectioning and staining. Despiteimportant advantages, postmortem studies cannot address in vivo, physiological, or longi-tudinal questions and have limited potential to explore early-stage disease except for themost common disorders. Emerging in vivo network-based neuroimaging strategies havebegun to address these issues, providing data that complement the neuropathological tradi-tion. Overall, findings to date highlight several fundamental principles of neurodegenerativedisease anatomyand pathogenesis, as well as some enduring mysteries. These principles andmysteries provide a road map for future research.

Neurodegenerative diseases are united by theinexorable and targeted spread of mis-

folded disease protein inclusions, gliosis, andsynaptic and neuronal loss. Clinical symptomsand deficits, which coalesce into recognizablesyndromes, reflect the topography of neurode-generation rather than the identity of the aggre-gating disease protein. Indeed, each protein isassociated with a handful of distinct clinicalsyndromes. Uncertainty surrounds which spe-cific aspects of each proteinopathy (i.e., “dis-ease”) drive that protein to select its uniqueanatomy in an individual patient. It has becomeclear, however, that the ultimate spatial pattern-

ing of disease is linked to the healthy brain’sconnectional architecture or “connectome.”

When discussing neurodegenerative condi-tions, it is critical to disambiguate terms thatrefer to the clinical syndrome from terms thatdescribe the underlying neuropathological en-tity giving rise to that syndrome. Throughoutthis review, I use “syndrome” when describing anamed constellation of symptoms and deficits.Examples include “behavioral variant fronto-temporal dementia” (bvFTD), “Alzheimer’sdisease (AD)-type dementia” or “corticobasalsyndrome.” In contrast, I use “disease” to referto a histopathological entity that might be

Editor: Stanley B. Prusiner

Additional Perspectives on Prion Biology available at www.cshperspectives.org

Copyright # 2017 Cold Spring Harbor Laboratory Press; all rights reserved

Advanced Online Article. Cite this article as Cold Spring Harb Perspect Biol doi: 10.1101/cshperspect.a023622

1

on April 22, 2018 - Published by Cold Spring Harbor Laboratory Press http://cshperspectives.cshlp.org/Downloaded from

mailto:[email protected]



http://www.cshperspectives.org



http://cshperspectives.cshlp.org/

found at autopsy in a patient showing a neuro-degenerative syndrome during life. Examples ofdisease terms include frontotemporal lobar de-generation (FTLD) with TAR DNA-bindingprotein 43 (TDP-43) immunoreactive inclu-sions (FTLD-TDP) Type A, Alzheimer’s disease,or corticobasal degeneration, a subtype of FTLDwith tau immunoreactive inclusions (FTLD-tau). In short, syndromes reflect where the dam-aging pathological process is, whereas diseaseterms describe what the pathological processis. Table 1 details the clinical syndromes usedto illustrate key principles throughout this re-view. Because the frontotemporal dementia(FTD), Alzheimer-type dementia, and amyotro-phic lateral sclerosis (ALS) syndromes have beenparticularly well-studied from a human networkperspective, these disorders receive more atten-tion here than several other equally importantand related disorders.

The need to separate clinical syndromicfrom neuropathologic terms stems from howfew one-to-one correspondences exist betweensyndrome and disease. These imperfect clinico-pathological correlations give rise to two relatedconcepts, “clinicoanatomical convergence” and“phenotypic diversity,” which receive extensiveconsideration in the sections that follow.

NETWORK DEGENERATION: HISTORICALMETHODS, OBSERVATIONS,AND LIMITATIONS

The stereotypical patterns of neurodegenerativedisease onset and progression have long stimu-lated ideas about a link to neuronal networks(Pearson et al. 1985; Saper et al. 1987; Braakand Braak 1991; Weintraub and Mesulam1996). The connectedness among degeneratingregions was inferred from axonal tracer studiesperformed in laboratory mammals, chiefly ro-dents and primates, and engendered diversemechanistic hypotheses ranging from spreadingprions (Prusiner 1984) to transported toxins(Saper et al. 1987), disrupted growth factors(Salehi et al. 2006), and unknown pathogens(Braak et al. 2003b). Despite these seminal per-spectives, for decades neurodegenerative diseaseanatomy was viewed through oversimplified

frameworks, divided into focal versus diffuseor subcortical versus cortical. The notion thateach disorder represents a network-based de-generation flows naturally, however, from care-ful, comprehensive postmortem neuropatho-logical localization and staging studies (Steeleet al. 1964; Brun and Gustafson 1978; Braakand Braak 1991). The great advantage of theseapproaches, which often used whole-brain orwhole-hemisphere sectioning and staining,was and remains their capacity to resolve cellulardetails in patients with defined molecular path-ological lesions. Early neuropathological hall-marks could therefore be identified in asymp-tomatic or prodromal individuals to render adetailed picture of onset and progression. Theneed to collect and process many brains, eachrequiring substantial resources, limited the useof these methods to a few laboratories. And, al-though the approach proved spectacularly suc-cessful for prevalent aging-related diseases likeAD and Lewy body disease (LBD), it can rarelycapture preclinical stages of FTLD, ALS, andother diseases too rare to be encountered bychance even in large autopsy series.

The dawn of human brain mapping in thelate 1980s, made possible by brain-wide, voxel-wise statistical methods, slowly gave rise to an erain which neurodegeneration researchers coulddetermine disease topographies in vivo withoutnarrow a priori hypotheses. This shift enabledvalidation of established patterns (such as that inAD) but was most impactful for less commondisorders, such as FTD (Rosen et al. 2002), forwhich ideas about anatomical onset and pro-gression had been difficult to derive from post-mortem data. Network-sensitive imaging ap-proaches emerged in the mid-1990s (Biswalet al. 1995; Fox and Raichle 2007) and provideda means for visualizing network organizationand degeneration in living humans (Greiciuset al. 2003, 2004; Seeley et al. 2009). Aroundthat time, complementary in vitro and animalmodel studies had begun to explore mechanismsof network-based dysfunction and disease pro-tein spread. A strong tide of empirical data nowsupports the notion that misfolded disease pro-tein conformers undergo prion-like spreadwithin and between neurons and across synapses

W.W. Seeley

2 Advanced Online Article. Cite this article as Cold Spring Harb Perspect Biol doi: 10.1101/cshperspect.a023622



Tabl

e1.

Neu

roan

atom

yofonse

tan

dpro

gres

sion

inth

em

ajor

neu

rodeg

ener

ativ

esy

ndro

mes

Syndro

me

Pro

dro

mal

sym

pto

m(s

)

Pre

senti

ng

sym

pto

m(s

)O

nse

tneu

ron(s

)

Onse

t

regi

on(s

)Ta

rget

net

work

Know

nor

candid

ate

dow

nst

ream

net

work

s

Most

like

ly

neu

ropat

holo

gica

l

dia

gnosi

s

Spre

adin

g

pro

tein

Diffe

rential

dia

gnosi

s

AD

-typ

ed

emen

tia

Dep

ress

ion

,sl

eep

dis

turb

ance

,an

xiet

y

Ep

iso

dic

mem

ory

loss

NE

5HT

Lay

erII

stel

late

pyr

amid

al

LC

DR

NE

RC

DM

NE

CN

Lan

guag

eD

AN

AD

Tau

Ab�

LB

DF

TL

D-T

DP

HS

Mix

edlv

PPA

Wo

rd-fi

nd

ing

dif

ficu

lty

Hes

itan

tsp

eech

Wo

rdfi

nd

ing

dif

ficu

ltie

s

Un

know

nD

om

inan

tA

G,

TP

JP

ost

erio

rla

ngu

age

DM

NA

DTa

uAb�

FT

LD

-TD

P

PC

AU

nkn

own

Vis

ual

chan

ges

Nav

igat

ion

Mey

ner

tce

llC

alcC

txL

at occ

ipit

al

1 ˚an

d2 ˚

visu

alD

MN

AD

Tau

Ab�

LB

D

bvF

TD

Job

loss

“Mid

life

cris

is”

Ap

ath

yD

isin

hib

itio

nC

om

pu

lsiv

ity

VE

NFo

rkce

llA

CC

,F

ISa

lien

ceE

CN

SAN

FT

LD

-TD

PF

TL

D-t

auT

DP

-43�

Tau

FT

LD

-FU

SA

D

svP

PAW

ord

-fin

din

gd

iffi

cult

yL

oss

of

wo

rd/

ob

ject

mea

nin

gU

nkn

own

Tp

ole

SAN

Sali

ence

FT

LD

-TD

PT

DP

-43�

FT

LD

-tau

nfv

PPA

Spee

chd

iffi

cult

yN

on

flu

ent

aph

asia

Un

know

nIF

Go

An

teri

or

lan

guag

eP

ost

erio

rla

ngu

age

FT

LD

-tau

Tau

FT

LD

-TD

P

CB

SU

nkn

own

Lo

sso

fli

mb

con

tro

lU

nkn

own

Pre

mo

tor

cort

exD

ors

alp

erir

ola

nd

icE

CN

Lan

guag

eF

TL

D-t

auTa

uA

DF

TL

D-T

DP

PSP

-SA

pat

hy,

men

tal

rigi

dit

yFa

lls

Un

know

nrM

TrM

Tn

etw

ork

Do

rsal

per

iro

lan

dic

PSP

Tau

CB

D

AL

SFa

sicu

lati

on

s,at

rop

hy

Wea

knes

sU

MN

LM

N1 ˚

MC

AH

Pyr

amid

alm

oto

rE

CN

Sali

ence

AL

S-T

DP

TD

P-4

3�A

LS-

FU

SA

LS-

SOD

CJD

Var

iab

leV

aria

ble

Un

know

nV

aria

ble

Var

iab

leV

aria

ble

CJD

PrP

Sc

-

Ast

eris

kac

know

led

ges

lin

geri

ng

un

cert

ain

tyab

ou

tw

het

her

the

ind

icat

edp

rote

insp

read

sal

on

gax

on

alp

ath

way

s.

1 ˚M

C,

pri

mar

ym

oto

rco

rtex

;5-

HT,

5-h

ydro

xytr

ypta

min

e;A

CC

,an

teri

or

cin

gula

teco

rtex

;A

D,

Alz

hei

mer

’sd

isea

se;

AH

,an

teri

or

ho

rn;

AL

S,am

yotr

op

hic

late

ral

scle

rosi

s;A

G,

angu

lar

gyru

s;b

vFT

D,

beh

avio

ral

vari

ant

fro

nto

tem

po

ral

dem

enti

a;C

alcC

tx;

calc

arin

e

cort

ex;

CB

S,co

rtic

ob

asal

syn

dro

me;

CJD

,C

reu

tzfe

ldt–

Jaco

bD

isea

se;

DA

N,

do

rsal

atte

nti

on

net

wo

rk;

DM

N,

def

ault

mo

de

net

wo

rk;

DR

N,

do

rsal

rap

he

nu

cleu

s;E

CN

,ex

ecu

tive

-co

ntr

ol

net

wo

rk;

ER

C,

ento

rhin

alco

rtex

;F

I,fr

on

toin

sula

;F

TL

D,

fro

nto

tem

po

ral

lob

ar

deg

ener

atio

n;F

US,

fuse

din

sarc

om

a;IF

Go

,in

feri

or

fro

nta

lgyr

us,

op

ercu

lar

par

t;L

BD

,Lew

yb

od

yd

isea

se;L

C,l

ocu

sce

rule

us;

LM

N,l

ower

mo

tor

neu

ron

;NE

,no

rep

inep

hri

ne;

PrP

Sc ,p

rio

np

rote

insc

rap

ie;P

SP,p

rogr

essi

vesu

pra

nu

clea

rp

alsy

;rM

T,ro

stra

lmid

bra

inte

gmen

tum

;

SAN

,se

man

tic-

app

rais

aln

etw

ork

;SO

D,

sup

ero

xid

ed

ism

uta

se;

TD

P-4

3,tr

ansa

ctiv

ere

spo

nse

DN

A-b

ind

ing

pro

tein

43kD

a;T

po

le,

tem

po

ral

po

le;T

PJ,

tem

po

rop

arie

tal

jun

ctio

n;

UM

N,u

pp

erm

oto

rn

euro

n;

VE

N,

von

Eco

no

mo

neu

ron

.

Mapping Neurodegeneration

Advanced Online Article. Cite this article as Cold Spring Harb Perspect Biol doi: 10.1101/cshperspect.a023622 3



(Frost and Diamond 2010; Goedert et al. 2010;Prusiner 2012). Most recently, structural, func-tional, and molecular neuroimaging studieshave been combined to replicate the stereotyp-ical spread of AD pathological hallmarks (Chooet al. 2007; Whitwell et al. 2007; Thal et al. 2014;Johnson et al. 2016). Emerging positron emis-sion tomography (PET) ligands for pathologicalforms of the tau protein mayenable similarchar-acterizations for the non-AD tauopathies, suchas FTLD-tau, at least during their symptomaticphases. Molecular probes fora-synuclein, TDP-43, and other disease proteins remain an impor-tant target for development.

HUMAN BRAIN NETWORK MAPPING:THE METHODS

Structural and functional connectivity analysesnoninvasively map healthy large-scale networksin vivo (Greicius et al. 2003; Damoiseaux et al.2006; Fox and Raichle 2007; Biswal et al. 2010)and can detect network connectivity changes inliving patients (Greicius et al. 2004; Zhou et al.2010). The following paragraphs provide a briefoverview of the major network-sensitive struc-tural and functional magnetic resonance–basedneuroimaging methods.

Intrinsic Connectivity

With task-free functional magnetic resonanceimaging (tf-fMRI), researchers can now identi-fy functional intrinsic connectivity networks(ICNs) derived from temporally synchronous,spatially distributed, spontaneous low-frequen-cy (,0.1 Hz) blood oxygen level–dependent(BOLD) signal fluctuations (Biswal et al. 1995;Raichle et al. 2001; Fox et al. 2005; Fox andRaichle 2007). These ICNs, which may repre-sent functional connections spanning multiplesynapses, represent a conserved and robust formof organized macroscopic brain activity. Com-pared to conventional task-based fMRI studies,tf-fMRI is free of performance confounds, mak-ing it easier to apply and interpret in cognitivelyimpaired populations. To derive ICNs, seed-based analyses determine correlations amonglow-frequency BOLD fluctuations of a seed re-

gion with the rest of the brain (Biswal et al.1995). Other approaches, such as independentcomponent analysis and clustering methods,take advantage of multiple simultaneous braininteractions to identify brain networks (Beck-mann et al. 2005; Yeo et al. 2011). Ongoingefforts seek to characterize temporal dynamicsof ICNs and elucidate possible causal relation-ships (see reviews by Krajcovicova et al. 2014and Dennis and Thompson 2014). Synchroni-zation across neuronal assemblies can likewisebe computed from task-free electro- or magne-to-encephalography data.

Structural Covariance

Coordinated variations in brain structure acrosssubjects have been used as measures of the asso-ciation between regions to construct large-scale“structural covariance networks” (Mechelli et al.2005; Lerch et al. 2006; He et al. 2008; Seeleyet al.2009). This approach, which may use gray mat-ter volume or cortical thickness data, relies onthe assumption that structural covariance re-flects a shared trophic influence during devel-opment or ongoing co-trophism conferred bysynaptic coupling across regions. Mean graymatter volume or thickness of a region of inter-est is used to conduct a whole-brain voxel-wiseregression across subjects to identify those vox-els (or regions or vertices) whose magnitude iscorrelated with the region of interest. Other an-alytic approaches, such as independent compo-nent analysis and clustering, can likewise beused to derive structural covariance networks.

Structural Connectivity

The term “structural connectivity” most strictlyrefers to the axonal connections between neu-rons or brain regions. Although axonal connec-tivity remains beyond the resolution of currentneuroimaging techniques, the integrity of me-dium to large fiber tracts can be assessed in vivousing diffusion-weighted imaging methods,which map the diffusion of water moleculesand rely on the principle that diffusion is re-stricted by tissue structure (Le Bihan et al.1992), especially within highly ordered white

W.W. Seeley




matter tracts. Region of interest analysis or data-driven voxel-based analysis allows estimation ofgroup differences in fiber tract integrity or as-sociations with cognitive functioning. Fibertracking between specific region pairs can fur-ther be performed (Mori et al. 1999; Mori andZhang 2006).

Connectomics

The term “connectome” refers to a comprehen-sive map of the brain’s neural connections(Sporns et al. 2005), whether the connectionsare defined on structural (MRI/diffusion) orfunctional (fMRI, electroencephalogram [EEG],magnetoencephalography [MEG]) grounds. Bymodeling networks as graphs (brain regions asnodes and node-to-node connections as edges),graph theoretical analyses offer a flexible andquantitative approach for characterizing brainnetwork topology. Several graph theoreticalmetrics quantify brain network “hubs” (i.e., re-gions with high degree centrality) (Sporns et al.2007; van den Heuvel and Sporns 2011; Zuoet al. 2012; Crossley et al. 2013), whereas othermetrics, such as clustering coefficient and pathlength, emphasize modularity or efficiency ofcommunication. “Connectomics,” then, refersto the science of brain connectivity.

NEURODEGENERATIVE DISEASE: UNIFYINGANATOMICAL PRINCIPLES

This section introduces the key concepts of neu-rodegenerative disease onset and progression.In my view, the most critical unanswered ques-tions in neurodegenerative disease research re-gard these two issues. In addition, I discuss twointerrelated neurodegenerative disease phe-nomena: clinicoanatomical convergence andphenotypic diversity. Any comprehensive mod-el of disease onset and progression must ac-count for these observations, which cut acrossthis class of human illness.

Onset

Patients with each neurodegenerative syndromeemerge from an incipient preclinical stage dur-

ing which symptoms remain absent or subtleand the lesion remains restricted to just one orfew brain regions and only to the most suscep-tible cells and microcircuits within the affectedregions. This focal onset manifests as cell-type-specific disease protein aggregation followed byquantifiable neuronal dropout (Hyman et al.1984; Graveland et al. 1985; Seeley et al. 2006;Kim et al. 2012).

Progression

What anatomical principles govern the relent-less spatiotemporal progression of each disease?Postmortem and in vivo neuroimaging studiessuggest that the pattern of regional injury re-flects a network-based landscape (Fig. 1), argu-ing against the notion that disease spreadsacross the cortical mantle via spatial contiguity(Steele et al. 1964; Brun and Gustafson 1978;Saper et al. 1987; Braak and Braak 1991; Grei-cius et al. 2004; Buckner et al. 2005; Seeley et al.2009). But what factors govern how diseasespreads from the onset node(s) to downstreamregions within and beyond the target network?At least three onset-progression scenariosshould be considered (Fig. 2).

1. Unifocal (or simultaneous oligofocal) onsetwith connectional spread. In this scenario,the later-affected regions are determined en-tirely by the axonal connections of the mostvulnerable cells within the onset region(s).

2. Staggered multifocal onset without connec-tional spread. Here, anatomical progressionreflects independent, temporally staggerederuptions of disease within multiple (notnecessarily interconnected) regions. In thisway, progression is connectivity-indepen-dent and generated by a graded hierarchy ofregional and/or cellular vulnerabilities tosome diffusely expressed pathogenic process.

3. Combined unifocal and staggered multifocalonset with connectional spread. In this mod-el, which blends aspects of the previous two,disease progression reflects not only the con-nectivity of the initial onset regions but alsothe emergence of later but independent on-





set sites and the connections of affected neu-rons within those later-affected sites.

Clinicoanatomical Convergence

Clinicoanatomical convergence describes theobservation that most clinical syndromes canbe caused by at least a few and often several un-derlying pathological entities. For example, pa-tients with bvFTD may be found to have any oneof at least 15 different underlying pathologicaldiagnoses, spanning three FTLD major molecu-lar classes (FTLD with tau, TDP-43, or FUS im-munoreactive inclusions) and AD. The key ques-tion is whether convergence occurs at thenetwork, regional, or neuronal level (Fig. 3). Inother words, distinct proteinopathies could con-verge at the network level by targeting disease-specific nodes within the same syndrome-asso-

ciated network. In this scenario, neuroimagingstudies might improve antemortem patholo-gical predictions by detecting disease-specificatrophy signatures (within the syndromic net-work). Alternatively, convergence could occurat the regional or even neuronal level, in whichcase methods capturing brain structure or func-tion would fail to discriminate between diseases,and alternative approaches, such as molecularimaging or fluid biomarkers linked to the diseaseproteins themselves, would be required.

Phenotypic Diversity

Phenotypic diversity refers to the observationthat the same histopathological entity (i.e., dis-ease) may be associated with several distinctclinical syndromes, reflecting distinct regionaldegeneration patterns (Fig. 4). For example,

AD

A

B

C

+35 +11 +14 +10 +40

L TPole

6

4.4

5

3.6

6

4.6

SDLIFG

PNFA CBS

Atrophy max= seed ROI

R PMC

R FI

R Ang

Intrinsic functional connectivity networks: healthy controls

Structural covariance networks: healthy controls

bvFTD

Syndrome-specific regional atrophy patterns: patients vs. controls

Figure 1. Neurodegenerative syndromes reflect degeneration within large-scale networks. (A) Five clinical neuro-degeneration syndromes showed distinct atrophy patterns, with atrophy maxima highlighted with white circles.Regions circled in A were used as seed regions of interest (ROIs) for task-free functional magnetic resonanceimaging (fMRI) analysis (B) and structural covariance mapping (C) in healthy controls. Both approaches showedthat the connectivity of the healthy brain mirrored the five atrophy patterns. These data showed that eachsyndrome was anatomically linked to a specific large-scale network that could be detailed in the healthy brainwith connectivity-based methods. (Reproduced, with permission, from Seeley 2016, # 2016 Oxford UniversityPress; www.oup.com.)

W.W. Seeley




Pick’s disease, a subtype of FTLD-tau, may pre-sent with bvFTD, semantic variant primary pro-gressive aphasia, nonfluent variant primaryprogressive aphasia, or corticobasal syndrome,based on the targeted regional epicenter and itsnetwork-based affiliations. This observationsuggests either that (1) each disease proteinmaintains a certain nonrandom variabilitywith regard to where it first aggregates in anindividual brain or that (2) neuropathologicaltaxonomy remains inadequately specified andthat further characterization (i.e., “splitting”)of the tau protein found in Pick’s disease will,extending the example, reveal different forms ofposttranslationally modified or misfolded tau ineach of the syndromic presentations of Pick’sdisease. Such hypothetical subtypes of a givensingle disease protein are often conceptualizedas conformer “strains” (Sanders et al. 2014). Toexplain phenotypic diversity entirely, the tauprotein strain recognized pathologically asPick’s disease would have to be further divided

into Pick-type “substrains,” one for each of thesyndromic presentations of Pick’s disease.

Based on these unifying neurodegenerationprinciples, I will discuss disease onset regionsand cell types in more detail. I will review neu-roimaging data that inform competing modelsof disease progression. I will relate competingconcepts of onset and spread to clinicoanatom-ical convergence and phenotypic diversity. Final-ly, I will consider the most important frontiersin selective vulnerability and network imaging.

MODELING ONSET: WHERE AND HOWDOES DISEASE BEGIN?

Evidence to Date

How does each neurodegenerative disease selectits initial target or targets? This question re-mains an enduring mystery for every illness,and merely identifying the early targets hasproven challenging enough. For AD and LBD,

MildOnset node

3. Combinationof 1 and 2

2. Staggered multifocal onset, no connection- based spread

1. Unifocal onset, connection- based spread

Onset

Incipient Mild

Progression

Moderate Severe

Moderate Severe

Nodal severity

Very severe

Figure 2. Neurodegenerative disease onset and progression. What is the relationship between disease onset andprogression? After a first locus of onset, progression to other regions could involve (1) connectivity-based spreadalone, (2) secondary sites of onset within or outside the target network, or (3) a combination of these models.(Reproduced, with permission, from Seeley 2016, # 2016 Oxford University Press; www.oup.com.)





early neuronal targets have been identifiedthrough cross-sectional postmortem studiesthat have included patients at all stages of thedisease process, from asymptomatic to prodro-mal to full-blown symptomatic and even end-stage (Braak et al. 1993, 2004). Regional-levelobservations made with this approach havebeen well-supported by longitudinal imagingstudies in living individuals. For example, stud-ies following older individuals from health tomild memory impairment and later AD-typedementia show early tau deposition and atrophyin the entorhinal cortex (Killianyet al. 2002; Jacket al. 2004; Varon et al. 2011; Johnson et al.

2016), consistent with classical postmortemstudies (Braak and Braak 1991; Braak et al.1993). Johnson et al. (2016) found that tau dep-osition extends beyond the medial temporallobe only in patients with cortical b-amyloiddeposition. On the other hand, in vivo brainimaging lacks the regional subnuclear and neu-ronal subtype resolution required to provide acomplete picture. This limitation is well-illus-trated by AD and Parkinson’s disease, in whichthe earliest brain neuronal protein aggregates arenow understood to emerge in brainstem nucleithat are difficult to resolve with conventionalMRI or PET: the locus coeruleus and dorsal ra-

Disease A

Clinicoanatomical convergence

Network-level convergenceNode-level convergence

Disease BDisease A Disease B

Network pattern 1Network

pattern 1ANetwork

pattern 1B

Syndrome 1

Mild Moderate Severe Very severe Epicenter

EpicenterVery severeSevereModerateMild

E E E

E

Syndrome 1

Figure 3. Clinicoanatomical convergence may occur at the neuronal, nodal, or network levels. Diseases that causeeach syndrome may converge at multiple levels to create the syndrome. Convergence at the level of specificneuronal types (not shown) or even specific network nodes (left) would be expected to create nearly identicalpatterns of network impairment. Alternatively, convergence could occur at the level of the overall network(right), with each disease targeting different nodes but, nonetheless, manifesting as the same (or nearly thesame) syndrome. Circles represent network nodes (brain regions), lines represent edges (connections betweentwo nodes), and shorter edges indicate tighter connections between node pairs. Color shading indicates theseverity of predicted regional impairment based on the onset nodes (“epicenters”) indicated by arrows. (Re-produced, with permission, from Seeley 2016, # 2016 Oxford University Press; www.oup.com.)

W.W. Seeley




phe in AD (Bondareff et al. 1981; Grinberg et al.2009; Braak and Del Tredici 2012) and the dorsalmotor nucleus of the vagus nerve in LBD (Braaket al. 2003a). In LBD, the process may begin evenmore peripherally, in the olfactory mucosa andenteric nervous system. For less common dis-eases, like FTLD, determining early neuronalsubtype selectivity has been even more difficultbecause of the diversity of FTD syndromesand the scarcity of postmortem materials frompatients with asymptomatic or prodromaldisease. The few laudable attempts to derive dis-tinct stages using cross-sectional materials havenot been able to include individuals withpresymptomatic disease (Brettschneider et al.2014; Irwin et al. 2016). Furthermore, becauseeach FTLD pathological subtype produces di-verse clinical phenotypes, it would be difficultto interpret presymptomatic FTLD materialseven if they became available. One remarkableexception comes from a patient who died of

brainstem lymphoma but was astutely noted toharbor premanifest Pick’s disease (Miki et al.2014), with Pick bodies and other Pick-typetau inclusions in an anterior cingulate-frontoin-sular pattern that almost perfectly matches theearly bvFTD regional vulnerability profile (See-ley et al. 2008).

How, then, can brain imaging studies insymptomatic patients inform our understand-ing of disease onset? Regions showing the great-est atrophy during symptomatic disease may ormay not represent the sites of initial injury, butrecent neuroimaging studies support an emerg-ing model for generating hypotheses aboutwhere each syndrome begins before it spreads.Having established that each neurodegenerativesyndrome is linked to a specific network (Fig. 3)(Seeley et al. 2009), my colleagues and I, ledby Juan (Helen) Zhou, showed that each syn-drome-associated brain network contains avulnerable “epicenter” (or epicenters), whose

Disease ADisease ADisease A

Anatomicalpattern 1

Syndrome 1

Anatomicalpattern 2

Phenotypic diversity

Syndrome 2

Anatomicalpattern 3

Syndrome 3

E2

E1

E3E1 E2

E3

Figure 4. Phenotypic diversity suggests that most diseases can produce multiple clinical syndromes, reflecting asmall portfolio of candidate onset regions (“epicenters,” E). The heterogeneity of clinical manifestations for eachdisease is illustrated here at the network level, where onset within epicenters (E1, E2, or E3) that anchor distinctnetworks gives rise to three different clinicoanatomical presentations. Network depictions follow Figure 3.





connectivity in health mirrors—and may tem-plate—the spatial patterning of each syndrome(Zhou et al. 2012). These epicenters bear closerelationships to the early clinical and anatomicaldeficits that define each syndrome. For instance,in bvFTD the identified epicenters in the rightfrontoinsula and pregenual anterior cingulatecortex are known for their co-activation aspart of a “salience network” for homeostaticbehavioral guidance (Seeley et al. 2011) and har-bor a unique class of large, bipolar projectionneurons, called von Economo neurons, that mayrepresent the initial target in bvFTD (Seeley et al.2006; Kim et al. 2012). Identifying an epicenter,as defined above, does not prove that this epi-center represents the site of initial injury; none-theless, there is a striking overlap betweenregions of peak atrophy and those that serve asepicenters (Zhou et al. 2012).

Despite the scarcity of postmortem materi-als representing presymptomatic FTLD, struc-tural and functional imaging has begun to pro-vide insights into presymptomatic inheritedFTD. In the first large study of this kind, carriersof FTLD-causing microtubule-associated pro-tein tau (MAPT) or progranulin (GRN) muta-tions showed fractional anisotropy reductionsin the right uncinate fasciculus and decreasedfunctional connectivity between key saliencenetwork hubs, the anterior mid-cingulate cor-tex and frontoinsula, compared with noncarri-ers (Dopper et al. 2013). More recently, usingregion of interest–based structural MRI, re-searchers have identified sites presumed to re-flect incipient atrophy in each of the three majorFTD-causing mutations (MAPT, GRN, andC9ORF72) (Rohrer et al. 2015). Convergingwith findings from patients with symptomaticbvFTD and with the bvFTD epicenters identi-fied by Zhou and colleagues (Zhou et al. 2012),Rohrer et al. identified the insula as a regionshowing atrophy among the youngest mutationcarriers when examining all three genetic sub-groups together. Although the insula appearedto degenerate first in a GRN mutation carriersubset analysis, other regions showed even ear-lier deficits in MAPT (hippocampus) andC9ORF72 (thalamus) carriers, as predicted bythe atrophy seen in symptomatic mutation car-

riers (Whitwell et al. 2009a,b; Mahoney et al.2012; Sha et al. 2012). These important studies,however, share several methodological limita-tions. In presymptomatic FTD gene carriers,we have no way to predict which of the severalassociated clinical syndromes will later emerge;in this way, group-level results likely represent ablend of preclinical syndromic patterns, as wellas the known anatomical heterogeneity withineach syndrome. Studies of preclinical inheritedFTD may also generalize weakly to sporadicFTD, considering the diversity of genetic mech-anisms and the known anatomical differencesseen in patients with inherited versus sporadicFTD. Finally, it remains uncertain whether theobserved gray matter volume deficits representincipient degeneration in early adulthood or anabnormal developmental trajectory that has yetto be traced back to its origins.

Relationship to ClinicoanatomicalConvergence and Phenotypic Diversity

Does clinicoanatomical convergence reflect on-set within the same vulnerable neuron popula-tion or within different neuronal constituentsof the same region or network? To addressthis question requires that we study all relevantlevels, in a single syndrome, as caused by multi-ple diseases. For example, does bvFTD begin inthe von Economo neurons whether the syn-drome is caused by FTLD-tau, TDP-43, orFUS? Some studies have provided clues towardthis cell-type-level convergence on the vonEconomo neurons (Seeley et al. 2006; Kimet al. 2012; Santillo et al. 2013; Santillo and En-glund 2014), but the studies needed to fully re-solve the issue have yet to be performed.

Principles of disease onset should also beviewed in light of phenotypic diversity. For ex-ample, although most patients with underlyingAD present with early memory loss, a signifi-cant minority presents with a nonamnestic syn-drome (Snowden et al. 2007). Patients withnonfamilial early-onset AD (EOAD, defined asonset ,65 years in most studies) show a mix ofcognitive deficits, often beginning with atten-tional or executive impairment (Frisoni et al.2007; Koedam et al. 2010). Focal syndromes

W.W. Seeley




such as posterior cortical atrophy (PCA), char-acterized by predominant visuospatial and vi-suoperceptual deficits (Crutch et al. 2012) andthe logopenic variant of primary progressiveaphasia (lvPPA), a progressive disorder of lan-guage (Gorno-Tempini et al. 2008), are alsostrongly linked to AD pathology. The factorsdriving this phenotypic diversity are not wellunderstood but could reflect an internal hierar-chy or “pecking order” of vulnerability that dif-fers between individual patients based on theirgenetic backgrounds, life experiences, region-specific stressors (trauma, seizures, vascularmalformations, etc.), or developmental anom-alies (Rogalski et al. 2013).

Lingering Questions and Uncertainties

Many key questions remain within the generalconcept of disease onset. How many cell typesand/or brain regions undergo independent(sometimes referred to as “cell autonomous”)onsets? What is the hierarchy of neuron-typevulnerabilities for each disease? Does this ordervary across individuals? Does onset occur with-in neurons, glia, or both? Can cells undergo a“reversible onset,” such as protein aggregationand dysfunction, but then revert to a healthystate? Does protein misfolding and aggregationbegin only within a select and finite group of celltypes/brain regions for each protein, or, alter-natively, does this homeostatically controlledprocess pervade the aging brain but remain incheck in all but that protein’s short list of onsetcells/regions, which are somehow ill-equippedto manage the quality control process?

MODELING PROGRESSION: HOWDOES DISEASE MOVE BEYOND THE CELLSAND REGIONS WHERE IT BEGINS?

Evidence to Date

That each neurodegenerative syndrome reflectsa large-scale network breakdown has now beenestablished through data that converge acrossdiseases, methods, and research groups. Earlynetwork-based imaging support for this princi-ple came from studies of AD-type dementia,

which features an anatomical profile stronglylinked to the default mode network (Greiciuset al. 2003, 2004; Buckner et al. 2008). Next, itwas shown that AD and four distinct FTD syn-dromes are each associated with atrophy reflect-ing a healthy human intrinsic connectivity andstructural covariance network (Fig. 1) (Seeleyet al. 2009). But how does disease progressfrom the onset stage to render a network-basedspatial pattern? At least four disease-general hy-potheses have been put forth and can be sum-marized as (1) “nodal stress,” in which regionssubject to heavy network traffic (i.e., “hubs”)undergo activity-related “wear and tear” thatgives rise to or worsens disease (Buckner et al.2009; Saxena and Caroni 2011); (2) “trans-neuronal spread,” in which some toxic agentpropagates along network connections, perhapsthrough “prion-like” templated conformationalchange (Prusiner 1984; Baker et al. 1994; Ridleyet al. 2006; Walker et al. 2006; Frost et al. 2009;Frost and Diamond 2010; Lee et al. 2010; Juckerand Walker 2011); (3) “trophic failure,” in whichdisruption of network connectivity underminesinternodal trophic factor support, acceleratingdisease within nodes lacking collateral trophicsources (Appel 1981; Salehi et al. 2006); and (4)“shared vulnerability,” in which networkedregions feature a common gene or protein ex-pression signature (Richiardi et al. 2015) thatconfers relatively disease-specific susceptibility,evenly distributed throughout the network.These nonmutually exclusive candidate networkdegeneration mechanisms make competingpredictions about how healthy network archi-tecture should influence disease-associatedregional vulnerability. Although “network de-generation” is often understood to mean “net-work-based spread,” only the “transneuronalspread” model proposes that progression repre-sents physical spreading of a pathological pro-cess along axons connecting individual neurons.

To date, most efforts to investigate mecha-nisms of disease progression have relied oncross-sectional data. In the study by Zhou andcoworkers (2012), we identified epicenterswhose normal connectivity profiles most re-sembled the syndrome-associated atrophy pat-terns, as described above. We then used graph





theoretical analyses in healthy subjects to showthat regions with higher total connectional flowand, more consistently, shorter functional pathsto the epicenters showed greater syndrome-as-sociated vulnerability. The relationship betweenregional network “traffic” and vulnerabilitysuggests that activity-dependent mechanisms,such as oxidative stress, local extracellular mi-lieu fluctuations, or glia-dependent phenome-na might influence regional vulnerability; thisinfluence might be a key factor in determiningsites of initial or secondary onset. Becausenodes with shorter connectional paths to anepicenter also showed greater vulnerability, itappears that “connectional closeness” is anotherkey vulnerability factor, an observation mostparsimoniously explained by physical, trans-synaptic spreading of a toxic agent. Epicenter in-filtration by disease may provide privileged butgraded and connectivity-driven access across thenetwork that determines where the disease willarrive next. Predictions made by the trophic fac-tor insufficiency hypothesis were not consistentwith our data. Although a shared gene or proteinexpression profile across networked regions mayinfluence sites of onset, our findings were diffi-cult to reconcile with predictions made by the“shared expression” model. We further exam-ined connectivity-vulnerability relationshipswithin the “off-target” networks to determinehow nodal characteristics influence downstreamvulnerability. Here, overwhelmingly, the evi-dence supported the transneuronal spreadmodel. In summary, the findings best fit a modelin which initial vulnerability may in part reflecta node’s centrality (i.e., “hubness”) within thetarget network, whereas downstream vulnera-bility within and beyond the target networkmore closely relates to a node’s connectionalproximity to the most vulnerable epicenters.

In AD, innovative studies have begun to linkregional connectivity profiles to hallmark ADmolecular lesions, which can now be localizedin vivo with molecular PET imaging, and diseaseprogression. In a study describing an “epidemicspreading model,” the investigators consideredaxonal propagation of amyloid protein along thehealthy structural connectome and regionalclearance mechanisms. The model was able to

explain roughly 50% of the variance in measuredamyloid deposition on amyloid PET (Iturria-Medina et al. 2014) based on the connectionalmodel, supporting the general hypothesis thatregional amyloid deposition in part reflects theconnectional distance from specific outbreak re-gions, which may lie in the anterior paramedianand posterior cingulate cortices. In AD, clearly,progression models need to account for twostages of the illness, one in which amyloid-bdeposition is a key factor and another in whichintraneuronal tau spreading takes over anddrives the clinical and anatomical deficit pat-tern. In a recent longitudinal study of prodromalAD and AD-type dementia, the healthy brain’sstructural connectome was used to predict theprogression of regional atrophy by modelingprogression as simple diffusion along fiber tracts(Raj et al. 2015). This model makes no assump-tions about where the diffusive process begins, afeature that may allow the model to accommo-date the known heterogeneity in onset sitesacross patients.

Relationship to ClinicoanatomicalConvergence and Phenotypic Diversity

How do emerging principles of disease progres-sion relate to clinicoanatomical convergence? Ifprogression is driven by connectional spread,then brain-wide anatomical convergence couldmerely reflect a shared population of onsetneurons. Alternatively, distinct onset sites with-in the same network could, via connectionalspread, produce convergent involvement ofthe overall network. In other words, there maybe alternative anatomical pathways to the samesyndrome. A particularly clear example of thisnotion comes from bvFTD. In the subset of pa-tients who carry the C9ORF72 hexanucleotiderepeat expansion, salience network dysfunctionresembles that seen in sporadic bvFTD, but theloss of network integrity is linked to a strategiclesion of the medial pulvinar thalamus (Leeet al. 2014). This mechanism of network break-down differs from that seen in sporadic bvFTD,where the salience network is disrupted by earlyinvolvement of anterior cingulate and frontoin-sular cortices. Thus, in bvFTD, the clinical def-

W.W. Seeley




icits may reflect disruption of the same networkby damage to distinct onset nodes.

The phenotypic diversity produced by ADraises the question of whether each clinical ADvariant can be linked to a distinct large-scalenetwork or onset site. A recent study testedthis hypothesis by assessing intrinsic functionalconnectivity in healthy subjects, seeding regionscommonly or specifically atrophied in early-on-set AD, lvPPA, or posterior cortical atrophy(Lehmann et al. 2013). The investigators foundthat the connectivity maps derived from com-monly atrophied regions of interest resembledthe default mode network, which was affected inall AD variants, whereas seeding regions specif-ically atrophied in each AD variant revealed dis-tinct, syndrome-specific connectivity patternsin the healthy brain. These findings indicatethat the syndrome-specific neurodegenerativepatterns in AD variants are driven by the in-volvement of specific networks outside the de-fault mode network. One might predict thatspread into these distinct networks reflects dif-ferences in the precise localization of onset inthe three variants; where exactly (in which re-gions and neuronal subtypes) these syndromesbegin remains uncertain, but meticulous neu-roanatomical studies suggest that PCA may be-gin with neurofibrillary tangle formation andneuronal loss within large, long-range projec-tion neurons in the primary visual cortex, suchas the layer 5 Meynert cells (Hof et al. 1997).

Lingering Questions and Uncertainties

Many questions about the mechanisms of dis-ease progression remain unanswered, and manyof those questions are daunting. Consideringthe three hypothetical progression scenarios(Fig. 2), what is the balance between connec-tion-based spread versus secondary sites of on-set? Does spread within the local microcircuitryoccur via contiguity (such as release of diseaseprotein by dying cells and uptake by others), oris it governed by axo-dendritic (or dendro-den-dritic) synapses? What better predicts diseaseprogression: a patient’s current, “personalized”(i.e., diseased) connectome, that patient’s pre-morbid connectome, or a normative connec-

tome from young or older subjects? How dogenetic risk factors interact with the connec-tome to influence disease progression? Resolv-ing these questions may help to facilitate de-velopment of individualized treatment andprevention trials.

FUTURE DIRECTIONS

To aid in the search for treatments, connectivi-ty-based neuroimaging methods will need todetect early disease in individuals or track pro-gression over time. For most sporadic diseases,presymptomatic detection remains a distant re-ality because either the right tools are lacking orthe disease is too infrequent to facilitate large-scale population screening without a sensitiveand affordable test. Efforts to monitor diseasewith connectivity metrics have, so far, been lim-ited, with most evidence coming from cross-sectional correlations with disease severity(Zhang et al. 2010; Zhou et al. 2010). One lon-gitudinal study showed reduced intrinsic con-nectivity in the posterior default mode networkand increased connectivity in anterior and ven-tral default mode subnetworks in AD comparedto healthy controls at baseline (Damoiseaux etal. 2012). At follow-up, patients showed wors-ening connectivity across all default modesubsystems, consistent with a network-baseddegeneration model in which disease firstspreads from its “epicenters” to interconnectednodes within the target network (Zhou et al.2012). An alternative model based on diffusionwithin the white matter architecture (i.e., struc-tural connectome) showed that the model couldpredict progression in subjects with mild cog-nitive impairment and AD-type dementia (Rajet al. 2015). Systematic collection and analysisof multicenter multimodal imaging and bio-marker data, including functional and struc-tural connectivity metrics, will be required toassess the value of imaging biomarkers for diag-nosis, prognosis, and disease monitoring.

ACKNOWLEDGMENTS

I thank Juan (Helen) Zhou for helpful dis-cussions and for contributing to the illustra-





tions. I also thank the study participants at theUniversity of California, San Francisco, for theirinvaluable contributions to neurodegenerationresearch.

REFERENCES

Appel SH. 1981. A unifying hypothesis for the cause ofamyotrophic lateral sclerosis, parkinsonism, and Alz-heimer disease. Ann Neurol 10: 499–505.

Baker HF, Ridley RM, Duchen LW, Crow TJ, Bruton CJ.1994. Induction of b (A4)-amyloid in primates by injec-tion of Alzheimer’s disease brain homogenate. Compar-ison with transmission of spongiform encephalopathy.Mol Neurobiol 8: 25–39.

Beckmann CF, DeLuca M, Devlin JT, Smith SM. 2005. In-vestigations into resting-state connectivity using inde-pendent component analysis. Philos Trans R Soc Lond BBiol Sci 360: 1001–1013.

Biswal B, Yetkin FZ, Haughton VM, Hyde JS. 1995. Func-tional connectivity in the motor cortex of resting humanbrain using echo-planar MRI. Magn Reson Med 34: 537–541.

Biswal BB, Mennes M, Zuo XN, Gohel S, Kelly C, Smith SM,Beckmann CF, Adelstein JS, Buckner RL, Colcombe S, etal. 2010. Toward discovery science of human brain func-tion. Proc Natl Acad Sci 107: 4734–4739.

Bondareff W, Mountjoy CQ, Roth M. 1981. Selective loss ofneurones of origin of adrenergic projection to cerebralcortex (nucleus locus coeruleus) in senile dementia. Lan-cet 1: 783–784.

Braak H, Braak E. 1991. Neuropathological staging ofAlzheimer-related changes. Acta Neuropathol 82: 239–259.

Braak H, Del Tredici K. 2012. Where, when, and in whatform does sporadic Alzheimer’s disease begin? Curr OpinNeurol 25: 708–714.

Braak H, Braak E, Bohl J. 1993. Staging of Alzheimer-relatedcortical destruction. Eur Neurol 33: 403–408.

Braak H, Del Tredici K, Rub U, de Vos RA, Jansen Steur EN,Braak E. 2003a. Staging of brain pathology related tosporadic Parkinson’s disease. Neurobiol Aging 24: 197–211.

Braak H, Rub U, Gai WP, Del Tredici K. 2003b. Idiopath-ic Parkinson’s disease: Possible routes by which vulner-able neuronal types may be subject to neuroinvasionby an unknown pathogen. J Neural Transm 110: 517–536.

Braak H, Ghebremedhin E, Rub U, Bratzke H, Del Tredici K.2004. Stages in the development of Parkinson’s disease-related pathology. Cell Tissue Res 318: 121–134.

Brettschneider J, Del Tredici K, Irwin DJ, Grossman M,Robinson JL, Toledo JB, Lee EB, Fang L, Van DeerlinVM, Ludolph AC, et al. 2014. Sequential distributionof pTDP-43 pathology in behavioral variant fronto-temporal dementia (bvFTD). Acta Neuropathol 127:423–439.

Brun A, Gustafson L. 1978. Limbic lobe involvement inpresenile dementia. Arch Psychiatr Nervenkr 226: 79–93.

Buckner RL, Andrews-Hanna JR, Schacter DL. 2008. Thebrain’s default network: Anatomy, function, and rele-vance to disease. Ann N Y Acad Sci 1124: 1–38.

Buckner RL, Snyder AZ, Shannon BJ, LaRossa G, Sachs R,Fotenos AF, Sheline YI, Klunk WE, Mathis CA, Morris JC,et al. 2005. Molecular, structural, and functional charac-terization of Alzheimer’s disease: Evidence for a relation-ship between default activity, amyloid, and memory. JNeurosci 25: 7709–7717.

Buckner RL, Sepulcre J, Talukdar T, Krienen FM, Liu H,Hedden T, Andrews-Hanna JR, Sperling RA, JohnsonKA. 2009. Cortical Hubs revealed by intrinsic functionalconnectivity: Mapping, assessment of stability, and rela-tion to Alzheimer’s disease. J Neurosci 29: 1860–1873.

Choo IH, Lee DY, Youn JC, Jhoo JH, Kim KW, Lee DS, Lee JS,Woo JI. 2007. Topographic patterns of brain functionalimpairment progression according to clinical severitystaging in 116 Alzheimer disease patients: FDG-PETstudy. Alzheimer Dis Assoc Disord 21: 77–84.

Crossley NA, Mechelli A, Vertes PE, Winton-Brown TT, Pa-tel AX, Ginestet CE, McGuire P, Bullmore ET. 2013. Cog-nitive relevance of the community structure of the hu-man brain functional coactivation network. Proc NatlAcad Sci 110: 11583–11588.

Crutch SJ, Lehmann M, Schott JM, Rabinovici GD, RossorMN, Fox NC. 2012. Posterior cortical atrophy. LancetNeurol 11: 170–178.

Damoiseaux JS, Prater KE, Miller BL, Greicius MD. 2012.Functional connectivity tracks clinical deteriorationin Alzheimer’s disease. Neurobiol Aging 33: 828.e19–828.e30.

Damoiseaux JS, Rombouts SARB, Barkhof F, Scheltens P,Stam CJ, Smith SM, Beckmann CF. 2006. Consistent rest-ing-state networks across healthy subjects. Proc Natl AcadSci 103: 13848–13853.

Dennis EL, Thompson PM. 2014. Functional brain connec-tivity using fMRI in aging and Alzheimer’s disease. Neu-ropsychol Rev 24: 49–62.

Dopper EG, Rombouts SA, Jiskoot LC, Heijer T, de Graaf JR,Koning I, Hammerschlag AR, Seelaar H, Seeley WW, VeerIM, et al. 2013. Structural and functional brain connec-tivity in presymptomatic familial frontotemporal de-mentia. Neurology 80: 814–823.

Fox MD, Raichle ME. 2007. Spontaneous fluctuations inbrain activity observed with functional magnetic reso-nance imaging. Nat Rev Neurosci 8: 700–711.

Fox MD, Snyder AZ, Vincent JL, Corbetta M, Van Essen DC,Raichle ME. 2005. The human brain is intrinsically orga-nized into dynamic, anticorrelated functional networks.Proc Natl Acad Sci 102: 9673–9678.

Frisoni GB, Pievani M, Testa C, Sabattoli F, Bresciani L,Bonetti M, Beltramello A, Hayashi KM, Toga AW,Thompson PM. 2007. The topography of grey matterinvolvement in early and late onset Alzheimer’s disease.Brain 130: 720–730.

Frost B, Diamond MI. 2010. Prion-like mechanisms in neu-rodegenerative diseases. Nat Rev Neurosci 11: 155–159.

Frost B, Ollesch J, Wille H, Diamond MI. 2009. Conforma-tional diversity of wild-type tau fibrils specified by tem-plated conformation change. J Biol Chem 284: 3546–3551.

W.W. Seeley




Goedert M, Clavaguera F, Tolnay M. 2010. The propagationof prion-like protein inclusions in neurodegenerative dis-eases. Trends Neurosci 33: 317–325.

Gorno-Tempini ML, Brambati SM, Ginex V, Ogar J,Dronkers NF, Marcone A, Perani D, Garibotto V, CappaSF, Miller BL. 2008. The logopenic/phonological variantof primary progressive aphasia. Neurology 71: 1227–1234.

Graveland GA, Williams RS, DiFiglia M. 1985. Evidence fordegenerative and regenerative changes in neostriatalspiny neurons in Huntington’s disease. Science 227:770–773.

Greicius MD, Krasnow B, Reiss AL, Menon V. 2003. Func-tional connectivity in the resting brain: a network anal-ysis of the default mode hypothesis. Proc Natl Acad Sci100: 253–258.

Greicius MD, Srivastava G, Reiss AL, Menon V. 2004. De-fault-mode network activity distinguishes Alzheimer’sdisease from healthy aging: Evidence from functionalMRI. Proc Natl Acad Sci 101: 4637–4642.

Grinberg LT, Rub U, Ferretti RE, Nitrini R, Farfel JM, Poli-chiso L, Gierga K, Jacob-Filho W, Heinsen H, BrazilianBrain Bank Study G. 2009. The dorsal raphe nucleusshows phospho-tau neurofibrillary changes before thetransentorhinal region in Alzheimer’s disease. A preco-cious onset? Neuropathol Appl Neurobiol 35: 406–416.

He Y, Chen Z, Evans A. 2008. Structural insights into aber-rant topological patterns of large-scale cortical networksin Alzheimer’s disease. J Neurosci 28: 4756–4766.

Hof PR, Vogt BA, Bouras C, Morrison JH. 1997. Atypicalform of Alzheimer’s disease with prominent posteriorcortical atrophy: A review of lesion distribution and cir-cuit disconnection in cortical visual pathways. Vision Res37: 3609–3625.

Hyman BT, Damasio AR, Van Hoesen GW, Barnes CL. 1984.Alzheimer’s disease: Cell-specific pathology isolates thehippocampal formation. Science 298: 83–95.

Irwin DJ, Brettschneider J, McMillan CT, Cooper F, Olm C,Arnold SE, Van Deerlin VM, Seeley WW, Miller BL, LeeEB, et al. 2016. Deep clinical and neuropathological phe-notyping of Pick disease. Ann Neurol 79: 272–287.

Iturria-Medina Y, Sotero RC, Toussaint PJ, Evans AC. 2014.Epidemic spreading model to characterize misfoldedproteins propagation in aging and associated neurode-generative disorders. PLoS Comput Biol 10: e1003956.

Jack CR Jr, Shiung MM, Gunter JL, O’Brien PC, WeigandSD, Knopman DS, Boeve BF, Ivnik RJ, Smith GE, ChaRH, et al. 2004. Comparison of different MRI brain at-rophy rate measures with clinical disease progression inAD. Neurology 62: 591–600.

Johnson KA, Schultz A, Betensky RA, Becker JA, Sepulcre J,Rentz D, Mormino E, Chhatwal J, Amariglio R, Papp K, etal. 2016. Tau positron emission tomographic imagingin aging and early Alzheimer disease. Ann Neurol 79:110–119.

Jucker M, Walker LC. 2011. Pathogenic protein seeding inAlzheimer disease and other neurodegenerative disor-ders. Ann Neurol 70: 532–540.

Killiany RJ, Hyman BT, Gomez-Isla T, Moss MB, Kikinis R,Jolesz F, Tanzi R, Jones K, Albert MS. 2002. MRI measuresof entorhinal cortex vs hippocampus in preclinical AD.Neurology 58: 1188–1196.

Kim EJ, Sidhu M, Gaus SE, Huang EJ, Hof PR, Miller BL,DeArmond SJ, Seeley WW. 2012. Selective frontoinsularvon Economo neuron and fork cell loss in early behav-ioral variant frontotemporal dementia. Cereb Cortex 22:251–259.

Koedam EL, Lauffer V, van der Vlies AE, van der Flier WM,Scheltens P, Pijnenburg YA. 2010. Early-versus late-onsetAlzheimer’s disease: More than age alone. J AlzheimersDis 19: 1401–1408.

Krajcovicova L, Marecek R, Mikl M, Rektorova I. 2014. Dis-ruption of resting functional connectivity in Alzheimer’spatients and at-risk subjects. Curr Neurol Neurosci Rep 14:491.

Le Bihan D, Turner R, Douek P, Patronas N. 1992. DiffusionMR imaging: Clinical applications. AJR Am J Roentgenol159: 591–599.

Lee JK, Jin HK, Endo S, Schuchman EH, Carter JE, Bae JS.2010. Intracerebral transplantation of bone marrow-de-rived mesenchymal stem cells reduces amyloid-beta dep-osition and rescues memory deficits in Alzheimer’s dis-ease mice by modulation of immune responses. StemCells 28: 329–343.

Lee SE, Khazenzon AM, Trujillo AJ, Guo CC, Yokoyama JS,Sha SJ, Takada LT, Karydas AM, Block NR, Coppola G, etal. 2014. Altered network connectivity in frontotemporaldementia with C9orf72 hexanucleotide repeat expansion.Brain 137: 3047–3060.

Lehmann M, Madison CM, Ghosh PM, Seeley WW, Mor-mino E, Greicius MD, Gorno-Tempini ML, Kramer JH,Miller BL, Jagust WJ, et al. 2013. Intrinsic connectivitynetworks in healthy subjects explain clinical variabilityin Alzheimer’s disease. Proc Natl Acad Sci 110: 11606–11611.

Lerch JP, Worsley K, Shaw WP, Greenstein DK, Lenroot RK,Giedd J, Evans AC. 2006. Mapping anatomical correla-tions across cerebral cortex (MACACC) using corticalthickness from MRI. Neuroimage 31: 993–1003.

Mahoney CJ, Beck J, Rohrer JD, Lashley T, Mok K, Shake-speare T, Yeatman T, Warrington EK, Schott JM, Fox NC,et al. 2012. Frontotemporal dementia with the C9ORF72hexanucleotide repeat expansion: Clinical, neuroana-tomical and neuropathological features. Brain 135:736–750.

Mechelli A, Friston KJ, Frackowiak RS, Price CJ. 2005. Struc-tural covariance in the human cortex. J Neurosci 25:8303–8310.

Miki Y, Mori F, Tanji K, Kurotaki H, Kakita A, Takahashi H,Wakabayashi K. 2014. An autopsy case of incipient Pick’sdisease: Immunohistochemical profile of early-stage Pickbody formation. Neuropathology 34: 386–391.

Mori S, Zhang J. 2006. Principles of diffusion tensor imag-ing and its applications to basic neuroscience research.Neuron 51: 527–539.

Mori S, Crain BJ, Chacko VP, Van Zijl PCM. 1999. Three-dimensional tracking of axonal projections in the brainby magnetic resonance imaging. Ann Neurol 45: 265–269.

Pearson RC, Esiri MM, Hiorns RW, Wilcock GK, Powell TP.1985. Anatomical correlates of the distribution of thepathological changes in the neocortex in Alzheimer dis-ease. Proc Natl Acad Sci 82: 4531–4534.





Prusiner SB. 1984. Some speculations about prions, amy-loid, and Alzheimer’s disease. N Engl J Med 310: 661–663.

Prusiner SB. 2012. Cell biology. A unifying role for prions inneurodegenerative diseases. Science 336: 1511–1513.

Raichle ME, MacLeod AM, Snyder AZ, Powers WJ, GusnardDA, Shulman GL. 2001. A default mode of brain func-tion. Proc Natl Acad Sci 98: 676–682.

Raj A, LoCastro E, Kuceyeski A, Tosun D, Relkin N, WeinerM. 2015. Network diffusion model of progression pre-dicts longitudinal patterns of atrophy and metabolism inAlzheimer’s disease. Cell Rep 10: 359–369.

Richiardi J, Altmann A, Milazzo AC, Chang C, ChakravartyMM, Banaschewski T, Barker GJ, Bokde AL, Bromberg U,Buchel C, et al. 2015. BRAIN NETWORKS. Correlatedgene expression supports synchronous activity in brainnetworks. Science 348: 1241–1244.

Ridley RM, Baker HF, Windle CP, Cummings RM. 2006.Very long term studies of the seeding of beta-amyloidosisin primates. J Neural Transm (Vienna) 113: 1243–1251.

Rogalski E, Weintraub S, Mesulam MM. 2013. Are theresusceptibility factors for primary progressive aphasia?Brain Lang 127: 135–138.

Rohrer JD, Nicholas JM, Cash DM, van Swieten J, Dopper E,Jiskoot L, van Minkelen R, Rombouts SA, Cardoso MJ,Clegg S, et al. 2015. Presymptomatic cognitive and neu-roanatomical changes in genetic frontotemporal demen-tia in the Genetic Frontotemporal Dementia Initiative(GENFI) study: A cross-sectional analysis. Lancet Neurol14: 253–262.

Rosen HJ, Gorno-Tempini ML, Goldman WP, Perry RJ,Schuff N, Weiner M, Feiwell R, Kramer JH, Miller BL.2002. Patterns of brain atrophy in frontotemporal de-mentia and semantic dementia. Neurology 58: 198–208.

Salehi A, Delcroix JD, Belichenko PV, Zhan K, Wu C, VallettaJS, Takimoto-Kimura R, Kleschevnikov AM, Samba-murti K, Chung PP, et al. 2006. Increased App expressionin a mouse model of Down’s syndrome disrupts NGFtransport and causes cholinergic neuron degeneration.Neuron 51: 29–42.

Sanders DW, Kaufman SK, DeVos SL, Sharma AM, MirbahaH, Li A, Barker SJ, Foley AC, Thorpe JR, Serpell LC, et al.2014. Distinct tau prion strains propagate in cells andmice and define different tauopathies. Neuron 82:1271–1288.

Santillo AF, Englund E. 2014. Greater loss of von Economoneurons than loss of layer II and III neurons in behavioralvariant frontotemporal dementia. Am J Neurodegener Dis3: 64–71.

Santillo AF, Nilsson C, Englund E. 2013. von Economo neu-rones are selectively targeted in frontotemporal demen-tia. Neuropathol Appl Neurobiol 39: 572–579.

Saper CB, Wainer BH, German DC. 1987. Axonal and trans-neuronal transport in the transmission of neurologicaldisease: Potential role in system degenerations, includingAlzheimer’s disease. Neuroscience 23: 389–398.

Saxena S, Caroni P. 2011. Selective neuronal vulnerability inneurodegenerative diseases: From stressor thresholds todegeneration. Neuron 71: 35–48.

Seeley WS. 2016. Brain circuits: Pathway for NDD. In Neu-rodegenerative diseases: Unifying principles, Chap. 7, pp.98-122. Oxford University Press, New York.

Seeley WW, Carlin DA, Allman JM, Macedo MN, Bush C,Miller BL, Dearmond SJ. 2006. Early frontotemporal de-mentia targets neurons unique to apes and humans. AnnNeurol 60: 660–667.

Seeley WW, Crawford R, Rascovsky K, Kramer JH, WeinerM, Miller BL, Gorno-Tempini ML. 2008. Frontal para-limbic network atrophy in very mild behavioral variantfrontotemporal dementia. Arch Neurol 65: 249–255.

Seeley WW, Crawford RK, Zhou J, Miller BL, Greicius MD.2009. Neurodegenerative diseases target large-scale hu-man brain networks. Neuron 62: 42–52.

Seeley WW, Zhou J, Kim EJ. 2011. Frontotemporal demen-tia: What can the behavioral variant teach us about hu-man brain organization? Neuroscientist 18: 373–385.

Sha SJ, Takada LT, Rankin KP, Yokoyama JS, Rutherford NJ,Fong JC, Khan B, Karydas A, Baker MC, DeJesus-Her-nandez M, et al. 2012. Frontotemporal dementia due toC9ORF72 mutations: Clinical and imaging features. Neu-rology 79: 1002–1011.

Snowden JS, Stopford CL, Julien CL, Thompson JC, David-son Y, Gibbons L, Pritchard A, Lendon CL, RichardsonAM, Varma A, et al. 2007. Cognitive phenotypes in Alz-heimer’s disease and genetic risk. Cortex 43: 835–845.

Sporns O, Tononi G, Kotter R. 2005. The human connec-tome: A structural description of the human brain. PLoSComput Biol 1: e42.

Sporns O, Honey CJ, Kotter R. 2007. Identification andclassification of hubs in brain networks. PLoS One 2:e1049.

Steele JC, Richardson JC, Olszewski J. 1964. Progressivesupranuclear palsy. Arch Neurol 10: 333–360.

Thal DR, Attems J, Ewers M. 2014. Spreading of amyloid,tau, and microvascular pathology in Alzheimer’s disease:Findings from neuropathological and neuroimagingstudies. J Alzheimers Dis 42: S421–S429.

van den Heuvel MP, Sporns O. 2011. Rich-club organizationof the human connectome. J Neurosci 31: 15775–15786.

Varon D, Loewenstein DA, Potter E, Greig MT, Agron J, ShenQ, Zhao W, Celeste Ramirez M, Santos I, Barker W, et al.2011. Minimal atrophy of the entorhinal cortex and hip-pocampus: Progression of cognitive impairment. DementGeriatr Cogn Disord 31: 276–283.

Walker LC, Levine H 3rd, Mattson MP, Jucker M. 2006.Inducible proteopathies. Trends Neurosci 29: 438–443.

Weintraub S, Mesulam MM. 1996. From neuronal networksto dementia: Four clinical profiles. La demence: pourquoi,pp. 75–97.

Whitwell JL, Przybelski SA, Weigand SD, Knopman DS,Boeve BF, Petersen RC, Jack CR Jr. 2007. 3D maps frommultiple MRI illustrate changing atrophy patterns as sub-jects progress from mild cognitive impairment to Alz-heimer’s disease. Brain 130: 1777–1786.

Whitwell JL, Jack CR Jr, Boeve BF, Senjem ML, Baker M,Ivnik RJ, Knopman DS, Wszolek ZK, Petersen RC, Rade-makers R, et al. 2009a. Atrophy patterns in IVS10 þ 16,IVS10 þ 3, N279K, S305N, P301L, and V337M MAPTmutations. Neurology 73: 1058–1065.

W.W. Seeley




Whitwell JL, Jack CR Jr, Boeve BF, Senjem ML, Baker M,Rademakers R, Ivnik RJ, Knopman DS, Wszolek ZK,Petersen RC, et al. 2009b. Voxel-based morphometry pat-terns of atrophy in FTLD with mutations in MAPT orPGRN. Neurology 72: 813–820.

Yeo BT, Krienen FM, Sepulcre J, Sabuncu MR, Lashkari D,Hollinshead M, Roffman JL, Smoller JW, Zollei L, Poli-meni JR, et al. 2011. The organization of the humancerebral cortex estimated by intrinsic functional connec-tivity. J Neurophysiol 106: 1125–1165.

Zhang HY, Wang SJ, Liu B, Ma ZL, Yang M, Zhang ZJ, TengGJ. 2010. Resting brain connectivity: Changes during theprogress of Alzheimer disease. Radiology 256: 598–606.

Zhou J, Greicius MD, Gennatas ED, Growdon ME, Jang JY,Rabinovici GD, Kramer JH, Weiner M, Miller BL, SeeleyWW. 2010. Divergent network connectivity changes inbehavioural variant frontotemporal dementia and Alz-heimer’s disease. Brain 133: 1352–1367.

Zhou J, Gennatas ED, Kramer JH, Miller BL, Seeley WW.2012. Predicting regional neurodegeneration from thehealthy brain functional connectome. Neuron 73:1216–1227.

Zuo XN, Ehmke R, Mennes M, Imperati D, Castellanos FX,Sporns O, Milham MP. 2012. Network centrality in thehuman functional connectome. Cereb Cortex 22: 1862–1875.





published online March 13, 2017Cold Spring Harb Perspect Biol William W. Seeley Mapping Neurodegenerative Disease Onset and Progression

Subject Collection Prion Biology

Neurodegenerative DiseasesClinical Neurology and Epidemiology of the Major

GeschwindMichael G. Erkkinen, Mee-Ohk Kim and Michael D.

Transgenesis in MiceNeurodegenerative Disease Transmission and

CarlsonBrittany N. Dugger, Daniel P. Perl and George A.

Sclerosis and Potential TherapyPrion Properties of SOD1 in Amyotrophic Lateral

Caroline Sibilla and Anne Bertolotti

ScToward the Atomic Structure of PrP

EisenbergJose A. Rodriguez, Lin Jiang and David S.

Genetic PrP Prion Diseases

et al.Mee-Ohk Kim, Leonel T. Takada, Katherine Wong, Progression

Mapping Neurodegenerative Disease Onset and

William W. SeeleyBioassays and Inactivation of Prions

al.Kurt Giles, Amanda L. Woerman, David B. Berry, et

Erratum: Functional Prions in the BrainJoseph B. Rayman and Eric R. Kandel

Functional Prions in the BrainJoseph B. Rayman and Eric R. Kandel

Pathology of Neurodegenerative DiseasesBrittany N. Dugger and Dennis W. Dickson

Human DiseaseThe Amyloid Phenomenon and Its Links with

Christopher M. Dobson

TIA-1 Is a Functional Prion-Like ProteinJoseph B. Rayman and Eric R. Kandel

Tau Positron Emission Tomography ImagingHartmuth C. Kolb and José Ignacio Andrés Dementias

Molecular Genetics of Neurodegenerative

Flora I. Hinz and Daniel H. Geschwind

InflammationPrion-Like Polymerization in Immunity and

Xin Cai, Hui Xu and Zhijian J. ChenSequence Domains

Polymerization of Low ComplexityβCross-

Masato Kato and Steven L. McKnight

http://cshperspectives.cshlp.org/cgi/collection/ For additional articles in this collection, see

Copyright © 2017 Cold Spring Harbor Laboratory Press; all rights reserved


http://cshperspectives.cshlp.org/cgi/collection/

http://cshperspectives.cshlp.org/cgi/collection/


mapping neurodegenerative disease onset and...

Documents