Research

JAMA Neurology | Original Investigation

Association of Perivascular Localization of Aquaporin-4

With Cognition and Alzheimer Disease in Aging Brains
Douglas M. Zeppenfeld, BS; Matthew Simon, BS; J. Douglas Haswell, BS; Daryl D’Abreo; Charles Murchison, MS; Joseph F. Quinn, MD;
Marjorie R. Grafe, MD, PhD; Randall L. Woltjer, MD, PhD; Jeffrey Kaye, MD; Jeffrey J. Iliff, PhD

Supplemental content
IMPORTANCE Cognitive impairment and dementia, including Alzheimer disease (AD), are
common within the aging population, yet the factors that render the aging brain vulnerable to
these processes are unknown. Perivascular localization of aquaporin-4 (AQP4) facilitates the
clearance of interstitial solutes, including amyloid-β, through the brainwide network of
perivascular pathways termed the glymphatic system, which may be compromised in the
aging brain.

OBJECTIVES To determine whether alterations in AQP4 expression or loss of perivascular

AQP4 localization are features of the aging human brain and to define their association with
AD pathology.

DESIGN, SETTING, AND PARTICIPANTS Expression of AQP4 was analyzed in postmortem

frontal cortex of cognitively healthy and histopathologically confirmed individuals with AD by
Western blot or immunofluorescence for AQP4, amyloid-β 1-42, and glial fibrillary acidic
protein. Postmortem tissue and clinical data were provided by the Oregon Health and Science
University Layton Aging and Alzheimer Disease Center and Oregon Brain Bank. Postmortem
tissue from 79 individuals was evaluated, including cognitively intact “young” individuals aged
younger than 60 years (range, 33-57 years), cognitively intact “aged” individuals aged older
than 60 years (range, 61-96 years) with no known neurological disease, and individuals older
than 60 years (range, 61-105 years) of age with a clinical history of AD confirmed by
histopathological evaluation. Forty-eight patient samples (10 young, 20 aged, and 18 with
AD) underwent histological analysis. Sixty patient samples underwent Western blot analysis
(15 young, 24 aged, and 21 with AD).

MAIN OUTCOMES AND MEASURES Expression of AQP4 protein, AQP4 immunoreactivity, and
perivascular AQP4 localization in the frontal cortex were evaluated.

RESULTS Expression of AQP4 was associated with advancing age among all individuals
(R2 = 0.17; P = .003). Perivascular AQP4 localization was significantly associated with AD
status independent of age (OR, 11.7 per 10% increase in localization; z = −2.89; P = .004) and
was preserved among eldest individuals older than 85 years of age who remained cognitively
intact. When controlling for age, loss of perivascular AQP4 localization was associated with Author Affiliations: Department of
increased amyloid-β burden (R2 = 0.15; P = .003) and increasing Braak stage (R2 = 0.14; Anesthesiology and Perioperative
P = .006). Medicine, Oregon Health and Science
University, Portland (Zeppenfeld,
Simon, Haswell, D’Abreo, Grafe, Iliff);
CONCLUSIONS AND RELEVANCE In this study, altered AQP4 expression was associated with Department of Neurology, Oregon
aging brains. Loss of perivascular AQP4 localization may be a factor that renders the aging Health and Science University,
brain vulnerable to the misaggregation of proteins, such as amyloid-β, in neurodegenerative Portland (Murchison, Quinn, Kaye);
Department of Pathology, Oregon
conditions such as AD.
Health and Science University,
Portland (Grafe, Woltjer); Knight
Cardiovascular Research Institute,
Oregon Health and Science
University, Portland (Iliff).
Corresponding Author: Jeffrey J.
Iliff, PhD, Department of
Anaesthesiology and Perioperative
Medicine, Oregon Health and Science
University, 3181 SW Sam Jackson Park
JAMA Neurol. 2017;74(1):91-99. doi:10.1001/jamaneurol.2016.4370 Rd, Mail Code L459, Portland, OR
Published online November 28, 2016. 97239 (iliffj@ohsu.edu).

(Reprinted) 91

Copyright 2017 American Medical Association. All rights reserved.

Downloaded From: https://jamanetwork.com/ by b Wildberries on 02/25/2023

 Research Original Investigation
A
dvancing age is the strongest risk factor for the devel-
opment of neurodegenerative disorders such as
Alzheimer disease (AD).1 A common feature of these
diseases is the age-associated accumulation of protein aggre-
gates, including senile plaques comprised of amyloid-β (Aβ)
1-42 in AD, α-synuclein within Lewy bodies in Parkinson
disease and Lewy body dementia, and hyperphosphorylated
tau within neurofibrillary tangles in AD and chronic trau-
matic encephalopathy, 2 yet the upstream changes that
render the aging brain vulnerable to the misaggregation of
proteins remain unknown.
Since 2013, we have defined a brain wide perivascular path-
way, termed the glymphatic system, that facilitates the recir-
culation of cerebrospinal fluid (CSF) through the brain
parenchyma and supports the clearance of interstitial sol-
utes including Aβ and tau.3-7 Perivascular exchange of CSF
and interstitial fluid is dependent on the astroglial water
channel aquaporin-4 (AQP4), which is localized to peri-
vascular astrocytic endfeet that ensheathe the cerebral
vasculature.7 We demonstrated that perivascular CSF recir-
culation and Aβ clearance are impaired in the aging mouse
brain, impairment that was associated with the loss of peri-
vascular AQP4 localization. Prior studies in postmortem
human tissue show that AQP4 is up regulated8,9 and that
localization of AQP4 to the cerebral vasculature is disrupted
in the AD cortex.10 This suggests that age-related mislocal-
ization of AQP4 may slow glymphatic function and promote
protein aggregation and neurodegeneration. In this study,
we assessed AQP4 expression and perivascular localization
in brain samples from a human case series including indi-
viduals of different ages and with different cognitive and
neuropathological AD profiles.
Methods
Human Participants and Tissues
This study consisted of 79 patients from the Oregon Health and
Science University Layton Aging and Alzheimer Disease Cen-
ter and associated postmortem tissue repository, the Oregon
Brain Bank. The Oregon Health and Science University insti-
tutional review board approved the study. Volunteers signed
written informed consent. All aged participants were commu-
nity-dwelling individuals with no known neurological dis-
ease (control individuals) or with a clinical history of AD as es-
tablished by neurologic evaluation in the Layton Aging and
Alzheimer Disease Center as previously described11 and in ac-
cordance with established consensus criteria.12 Brain au-
topsy was performed on all participants after consent was ob-
tained from the next of kin and in accordance with Oregon
Health and Science University guidelines.
Neuropathological Evaluation and Evaluation
of AQP4 Expression and Localization
Brains in the Oregon Brain Bank underwent neuropathologi-
cal evaluation for Aβ plaque density, neurofibrillary pathol-
ogy, and vascular pathology. Expression of AQP4 and localiza-
tion was evaluated by Western blot and immunofluorescence
92 JAMA Neurology January 2017 Volume 74, Number 1 (Reprinted)
Copyright 2017 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ by b Wildberries on 02/25/2023
Aquaporin-4 Localization and Cognition in Aging Brains
Key Points
Question Is the expression or localization of the astroglial water
channel aquaporin-4 altered in patients with advanced age or with
Alzheimer disease?
Findings In this postmortem analysis, aquaporin-4 protein
expression and localization in the cortex of a series of aged
cognitively intact individuals and patients with Alzheimer disease
revealed statistically significant associations between aquaporin-4
expression and aging. Loss of aquaporin-4 protein localization to
perivascular astrocytic endfeet was strongly associated with
Alzheimer disease status and pathology.
Meaning Increasing aquaporin-4 expression is a feature of the
aging human brain, and mislocalization of aquaporin-4 is related to
the development of Alzheimer disease pathology.
in frozen and fixed frontal cortical tissue, respectively. De-
tailed descriptions of these methods are provided in the
eMethods in the Supplement.
Statistical Evaluation
A detailed description of statistical approaches is provided in
the eMethods in the Supplement. Multiple comparisons cor-
rected P values, means, and standard errors of the mean are
indicated in the article. All statistical analyses were carried out
using Prism 6 (GraphPad) and R 3.2.1 (R Foundation).13
Results
Demographic and Histopathological Characteristics
Expression of AQP4 and localization were evaluated from a
cohort of 79 individuals including “young” control indi-
viduals younger than 60 years of age (range, 33-57 years)
without a history of neurological disease, cognitively intact
(Clinical Dementia Rating14 = 0) “aged” individuals older
than 60 years of age (range, 61-96 years), and individuals
with histopathologically confirmed AD older than 60 years
of age (range, 61-105 years) . Among these, 48 (10 young, 20
aged, and 18 with AD) underwent histological analysis. Sixty
underwent Western blot analysis (15 young, 24 aged, and 21
with AD) of AQP4. Twenty-nine individuals (11 aged and 18
with AD) underwent both histological and Western blot
analysis. The Table shows sex balance and median and
interquartile ranges of age, years of schooling, Aβ plaque
density, and Braak stage for all participants. Most recent
Mini-Mental State Examination and Clinical Dementia Rat-
ings are provided for a subset of aged individuals and indi-
viduals with AD who underwent cognitive testing at the
Layton Aging and Alzheimer Disease Center. No significant
differences between groups were observed in sex balance
(χ2 = 0.51; P = .78), years of schooling (t = 0.39; P = .70), or
postmortem interval (F = 2.37; P = .10). Linear regression
showed no association between postmortem interval and
global AQP4-immunoreactivity (IR) or perivascular AQP4
localization assessed by immunofluorescence (eFigure 1A in
the Supplement, R 2 = 0.01; P = .44; eFigure 1B in the
jamaneurology.com
 Aquaporin-4 Localization and Cognition in Aging Brains Original Investigation Research

Table. Cohort Demographic, Cognitive, and Pathological Features

Demographic Features Cognitive Features Alzheimer Pathology

Sex
Balance PostMortem Years of Amyloid Plaque Braak Stage
Age Age, (Male/ Interval, Schooling, MMSE, CDR, Density (0-3), (0-6),
Range No. Median (IQR) Female) Median (IQR) Median (IQR) No. Median (IQR) Median (IQR) No. Median (IQR) Median (IQR)
Young 25 44.0 (14/11) 17.8 NA 0 NA NA 25 0.0 0.0
(<60 y) (37.0-51.0) (10.9-32.3) (0.0-0.0) (0.0-0.0)
Aged 33 83.0 (14/19) 13.5 14.0 15 28.0 0.0 33 0.0 1.0
(>60 y) (69.0-87.6) (5.0-21.8) (12.0-18.0) (27.0-29.0) (0.0-0.0) (0.0-1.0) (0.0-3.0)a
Alzheimer 21 81.3 (10/11) 12.0 14.0 16 18.0 1.0 21 2.0 6.0
Disease (67.0-95.4) (6.0-24.0) (12.0-16.0) (9.0-22.5) (0.5-2.0) (1.5-3.0)b,c (4.0-6.0)b,c
(>60y)
b
Abbreviations: CDR, Clinical Dementia Rating; IQR, interquartile range. Padj<.001 vs young individuals.
a c
Padj<.05. Padj<.001 vs aged individuals (Kruskal-Wallis test).

Supplement, R 2 = 0.02; P = .29; Pearson correlation) or
postmortem interval and AQP4 expression evaluated by
Western blot (eFigure 1C in the Supplement, R 2 = 0.01;
P = .39; Pearson correlation). The mean (SE) interval
between age at most recent cognitive testing and death did
not differ between aged individuals (0.9 [0.2] years) and
individuals with AD (1.9 [0.5] years) (P = .13, independent
samples t test). Histopathological examination revealed
greater Aβ plaque density in AD compared with young or
aged individuals (AD vs young: Dunn’s z = 6.64; adjusted
P <.001; AD vs old: Dunn’s z = 5.27; adjusted P <.001;
Kruskal-Wallis test). Individuals with AD had higher Braak
scores compared with aged or young individuals (AD vs
young: Dunn’s z = 7.54; adjusted P <.001; AD vs old: Dunn’s
z = 4.58; adjusted P <.001; Kruskal-Wallis test); however,
aged individuals exhibited significantly higher Braak scores
than young individuals (Dunn’s z = 3.59; adjusted P = .005;
Kruskal-Wallis test).
General cerebrovascular pathology (VIS) was rated on the
degree of arteriolosclerosis, perivascular hemosiderin leak-
age, perivascular space dilation and myelin loss from the fron-
tal cortex, and cortical amyloid angiopathy in the occipital cor-
tex according to criteria published be Deramecourt et al.15
Median and interquartile ranges for VIS and subscores are
shown in the eTable in the Supplement. As shown in eFigure
2A in the Supplement, VIS was significantly greater in the aged
individuals and individuals with AD compared with young in-
dividuals (young vs old: Dunn’s z = −2.24; adjusted P = .04;
young vs AD: Dunn’s z = −3.99; adjusted P <.001, respec-
tively; Kruskal-Wallis test). Evaluation of the relationship be-
tween age and VIS indicated a significant linear association
(eFigure 2B in the Supplement; R2 = 0.36; P < .001, Pearson cor-
relation).
Evaluation of Perivascular AQP4 Localization
by Immunofluorescence
When AQP4 expression and localization were evaluated by im-
munofluorescence double-labeling, we observed that in the
young cortex, AQP4 expression was uniform through the cor-
tical layers spanning from the glia limitans to the subcortical
white matter (Figure 1A). Although expressed in fine pro-
cesses throughout the neuropil, AQP4-IR was greatest in peri-
vascular endfeet surrounding the cerebral vasculature
(Figure 1B, arrowhead). Expression and localization of AQP4
were dramatically altered in the aged brain. While intense AQP4
expression was maintained near the cortical surface, AQP4
expression became discontinuous below cortical layer II
(Figure 1C) where AQP4-IR remained intense within a sparse
subpopulation of astrocytes scattered throughout the deeper
cortical layers (Figure 1D). When compared between groups,
the mean (SD) AQP4-IR was significantly increased among
individuals with AD (31.57 [2.84] arbitrary units) compared
with young individuals (20.21 [0.99] arbitrary units;
P = .004; Welch-corrected 1-way ANOVA) or aged individuals
(23.78 [1.31] arbitrary units; adjusted P = .029; Welch-
corrected 1-way analysis of variance [ANOVA]) (Figure 1E).
Multiple linear regression analysis showed that when correct-
ing for the influence of age, increasing AQP4-IR was signifi-
c antly associated with increasing Aβ plaque density
(Figure 1F; unadjusted P = .006; R2 = 0.21; adjusted for age
P = .01; R 2 = 0.30; Huber-White–corrected ordinary least
squares [OLS]). In addition, increasing AQP4-IR also tended
to be associated with increasing Braak stage, a relationship
persisting even when adjusting for the effects of age
(Figure 1G; unadjusted P = .002; R2 = 0.15; adjusted for age
P = .03; R2 = 0.20; Huber-White–corrected OLS).
In parallel with increasing global AQP4-IR, mean (SE) peri-
vascular AQP4 localization was reduced in individuals with AD
(1.095 [0.007]) compared with young individuals (1.160
[0.019]; adjusted P = 0.01; Welch-corrected 1-way ANOVA) or
aged individuals (1.161 [0.016]; P = .003; Welch-corrected
1-way ANOVA; Figure 1H). Regression analysis showed that
when correcting for the influence of age, reduced perivascu-
lar AQP4 localization was associated with increasing Aβ
plaque density (Figure 1I; unadjusted P < .001; R2 = 0.15;
adjusted for age P < .001; R2 = 0.14; Huber-White–corrected
OLS) and increasing Braak stage (Figure 1J; unadjusted
P = .002; R 2 = 0.14; adjusted for age P = .004; R 2 = 0.12;
Huber-White–corrected OLS).
We next evaluated whether altered AQP4 expression or lo-
calization are general features of the aging brain. Regression
analysis showed a significant association between increasing
age and increasing AQP4-IR (Figure 2A; P = .001; ρ = 0.45,
Spearman’s correlation), while no significant association
was evident between age and alterations in perivascular AQP4
localization (Figure 2B; P = .15; ρ = −0.21, Spearman’s

jamaneurology.com (Reprinted) JAMA Neurology January 2017 Volume 74, Number 1 93

Copyright 2017 American Medical Association. All rights reserved.

Downloaded From: https://jamanetwork.com/ by b Wildberries on 02/25/2023

 Research Original Investigation Aquaporin-4 Localization and Cognition in Aging Brains

Figure 1. Loss of Perivascular Aquaporin-4 (AQP4) Localization Is Associated With Alzheimer Disease (AD) Pathology

A Young (uniform distribution) B Young (localization) C Aged AQP4 expression D Aged AQP4 localization

DAPI DAPI
DAPI AQP4 DAPI AQP4
AQP4 500 μm 50 μm 500 μm GFAP 50 μm
GFAP AQP4

E Global AQP4 expression F Global AQP4 expression G Global AQP4 expression

a
80 80 80
b

60 60 60
AQP4-IR, AU

AQP4-IR, AU
AQP4-IR, AU

P = .03
P = .01
40 40 40

20 20 20

0 0 0
Young Aged AD 0 1 2 3 0 1 2 3 4 5 6
Amyloid Plaque Density Braak Stage

H Perivascular AQP4 localization I Perivascular AQP4 localization J Perivascular AQP4 localization

d
1.4 1.4 1.4
c Cognitively intact
Alzheimer disease
1.3 1.3
AQP4 Ratio (PV:Parenchyma)

AQP4 Ratio (PV:Parenchyma)

AQP4 Ratio (PV:Parenchyma)

1.3
P = .004

1.2 P = .001 1.2

1.2

1.1 1.1 1.1

1.0 1.0 1.0

0.9 0.9 0.9

Young Aged AD 0 1 2 3 0 1 2 3 4 5 6
Amyloid Plaque Density Braak Stage

A, Perivascular AQP4 localization was evaluated by immunofluorescence. In the AQP4-IR in perivascular domains to global AQP4-IR) showing reduced
young cortex, AQP4 was distributed uniformly from the cortical surface to the perivascular localization in individuals with AD. I-J, Regression analysis shows
subcortical white matter boundary. B, High-power confocal imaging showing that when controlling for the effects of age, reduced perivascular AQP4 is
localization of AQP4 in perivascular astrocytic endfeet (arrowhead). C, In the significantly associated with both amyloid-β plaque density (age-adjusted
aged cortex (aged 60-85 years), AQP4 was intensely expressed in a sparse P < .001, R2 = 0.14) (I) and Braak stage (age-adjusted P = .004, R2 = 0.12;
population of cortical astrocytes, while wider AQP4 expression was reduced. Huber-White–corrected ordinary least squares) (J). DAPl indicates
D, High-magnification image showing altered AQP4 expression and localization L-diaminopimelate aminotransferase; GFAP indicates glial fibrillary acidic
in the aged cortex. E, Quantification of global AQP4-immunoreactivity (IR) protein; and PV indicates parenchyma volume.
showing increased AQP4 expression in the AD cortex compared with the a
P = .03.
cognitively intact young or aged cortex. F-G, Regression analysis shows that b
P = .004; Welch-corrected 1-way analysis of variance.
when controlling for the effects of age, increasing AQP4-IR is significantly
c
(P = .01, R2 = 0.30; Huber-White–corrected ordinary least squares) associated P = .01.
with amyloid-β plaque density (F), as is association with Braak stage (P = .029, d
P = .003; Welch-corrected 1-way analysis of variance.
R2 = 0.20) (G). H, Quantification of perivascular AQP4 localization (ratio of

94 JAMA Neurology January 2017 Volume 74, Number 1 (Reprinted) jamaneurology.com

Copyright 2017 American Medical Association. All rights reserved.

Downloaded From: https://jamanetwork.com/ by b Wildberries on 02/25/2023

 Aquaporin-4 Localization and Cognition in Aging Brains Original Investigation Research

Figure 2. Loss of Perivascular Aquaporin-4 (AQP4) Localization Predicts Alzheimer Disease (AD) Status

Cognitively intact (25-85 y)

A Regression analysis C Logistic regression analysis Cognitively intact − eldest (>85 y)
Alzheimer disease (>60 y)
80 80

P = .01

60 60
AQP4-IR, AU

AQP4-IR, AU
40 40

P = .001

20 20

0 0
40 60 80 100 Young Aged Eldest Aged Eldest
Age, y Cognitively Intact Alzheimer Disease

B Perivascular AQP4 localization D Perivascular AQP4 localization

1.4 1.4
P < .001

1.3 1.3
AQP4 Ratio, PV:Parenchyma
AQP4 Ratio, PV:Parenchyma

1.2 1.2

1.1 1.1

1.0 1.0
P = .02

0.9 0.9
40 60 80 100 Young Aged Eldest Aged Eldest
Age, y Cognitively Intact Alzheimer Disease

A, Regression analysis shows that among all individuals, increasing age is
significantly associated with increasing AQP4-immunoreactivty (IR) (P = .001,
ρ = 0.45, Spearman’s correlation). B, Perivascular AQP4 localization is not
significantly associated with age across all individuals; however, among
individuals with AD, increasing age was associated with decreasing perivascular
localization (P = .02, ρ = −0.57, Spearman’s correlation). Perivascular AQP4
localization among cognitively intact eldest individuals older than 85 years of
age (oval) diverged markedly from eldest individuals with AD. Results of
White-Huber–corrected logistic regression analysis indicated that AQP4-IR (C)
and perivascular AQP4 localization (D) are significant predictors of AD status
(P = .01, age-adjusted P = .29; P < .001, age-adjusted P = .02; respectively).
Significance of AQP4 as a covariate predicting cognitive status shown in the
figure. PV indicates parenchyma volume.

correlation). Strikingly, although perivascular AQP4 localiza-
tion values were similar between aged individuals and indi-
viduals with AD between 60 and 85 years of age, there was a
marked divergence between values among cognitively intact
individuals and individuals with AD older than 85 years of age
(referred to as “eldest” individuals; Figure 2B, oval). Among
individuals with AD, increasing age was significantly associ-
ated with reduced perivascular AQP4 localization (P = .02;
ρ= −0.57, Spearman’s correlation).
The results of Huber-White–corrected logistic regression
analysis modeling AD status (cognitively intact vs AD) with age
and perivascular AQP4 localization as independent variables
indicated that age was a significant predictor (adjusted P = .02;
b = 0.037; z = 2.36) and decreasing perivascular AQP4 local-
ization was significantly associated with AD status (P < .001;
b = −24.6; z = −3.58; Figure 2D). In contrast, increas-ing
AQP4-IR was significantly associated with AD status (P = .01;
b = 0.13; z = 2.45) in our cohort, while age showed no associa-
tion (P = .29; b = 0.017; z = 1.07; Figure 2C).
Although both AQP4-IR and perivascular AQP4 localiza-
tion were associated with vascular pathology (P = .05; R2 = 0.07
and P = .05; R2 = 0.07, respectively), this association disap-
peared when the effects of age were controlled for (P = .83;
R2 = 0.17 and P = .17; R2 = 0.05, respectively; Huber-White–
corrected OLS). These results suggest that increasing vascular
pathology and increasing global AQP4-IR are general features
of the aging brain with the preservation of perivascular AQP4
localization a significant predictor of preserved cognitive func-
tion, even in individuals who remain cognitively intact late
into life.

jamaneurology.com (Reprinted) JAMA Neurology January 2017 Volume 74, Number 1 95

Copyright 2017 American Medical Association. All rights reserved.

Downloaded From: https://jamanetwork.com/ by b Wildberries on 02/25/2023

 Research Original Investigation Aquaporin-4 Localization and Cognition in Aging Brains

Figure 3. Altered Aquaporin-4 (AQP4) Expression in Deep Layers of the Figure 4. Amyloid-β (Aβ) Plaques Are Associated With Aquaporin-4
Aging Cortex (AQP4) Mislocalization

A Young (25-45 y) B Aged (60-85 y) C Eldest (>85 y) A Young (25-45 y) B Aged (60-85 y) C Alzheimer (60-85 y)

Gyrus
Gyrus Gyrus
Cortical Gyrus

500 μm 500 μm 500 μm

AQP4

D Young (25-45 y) E Aged (60-85 y) F Eldest (>85 y) Sulcus

Sulcus
Sulcus
WM
WM WM

DAPI
DAPI DAPI AQP4
1 mm 1 mm Aβ1-42 1 mm
AQP4 AQP4
Cortical Sulcus

500 μm 500 μm 500 μm

250 μm 150 μm 250 μm

Wide field confocal montages show AQP4 immunoreactivity extending from

the cortical surface (dashed line) at the crests of gyri (A-C) and the depths of
sulci (D-F). Compared with the uniform AQP4 expression throughout different D Aβ plaque–associated AQP4 E Aβ plaque–associated AQP4
cortical layers in young cortex (A and D), AQP4 labeling in cortical layers II to IV
of cognitively intact aged (60-85 years) individuals (B and E) was reduced, with
the appearance of intensely immunoreactive astrocytes throughout these
layers. These differences were not observed among cognitively intact eldest
individuals (older than 85 years) (C and F).
DAPI DAPI
AQP4 200 μm 200 μm AQP4 50 μm
50 μm
Aβ1-42 GFAP

Regional Patterns of AQP4 Localization
When evaluating the spatial pattern of AQP4 expression in the
cortex, we observed that compared with the uniform distri-
bution of AQP4 through layers II to VI of the young cortex
(Figure 3A and E), AQP4 localization was dramatically altered
in the cortex of aged individuals (Figure 3B and F). Interest-
ingly, different layer-specific patterns of AQP4 expression
were evident among aged individuals at the crests of gyri vs
the depths of sulci. At the crests of gyri of aged individuals,
AQP4 expression was lost between layers II and IV, while a
band of AQP4-IR remained in layers V to VI at the gray
matter–white matter boundary (Figure 3A-B). At the depths
of sulci of aged individuals, an intense band of AQP4-IR
remained in cortical layer II, while AQP4 expression within
deeper layers was reduced (Figure 3D-E). Figure 4A-B shows
representative multifield montages depicting the effects of
age on AQP4 expression along the length of whole gyri from
young and aged individuals. Such regional patterns of
AQP4-IR were observed in whole-gyrus images from 6 to 7
individuals per group.
Among cognitively intact eldest individuals (older than 85
years), AQP4 expression in the middle layers of the cortex was
maintained (Figure 3C). Within the depths of the sulci, the
Large wide-field confocal montages extending from the crests of sulci to the
base of gyri were generated. A, AQP4-immunoreactivity (IR) in the young brain
is uniform through the depth of the cortex to the gray matter−white matter
(WM) boundary. Insets depict a region of the montage at higher magnification.
B, In the cognitively intact aged cortex (60-85 years), AQP4-IR is reduced in
deeper cortical layers (blue arrowhead), while a band of intense IR runs along
the gray matter−WM boundary (white arrowhead) within the cortex at the crest
of the gyrus but not in the cortex in the depth of the sulcus. C, In individuals
with Alzheimer disease, AQP4-IR in deeper cortical layers is increased compared
with aged individuals, owing to the presence of AQP4-IR associated with Aβ
plaques (dashed circles in inset). D, High-power confocal microscopy shows
that Aβ plaques are surrounded by foci of AQP4-IR. E, These foci are features of
Aβ plaque-associated reactive astrocytes (glial fibrillary acidic protein).
intense band of AQP4 immunofluorescence observed super-
ficially in layer II of aged individuals was absent in the cogni-
tively intact eldest individuals (Figure 3F). Thus, the regional
patterns of altered AQP4 expression and localization that are
features of the aged brain are absent among the cognitively
intact eldest individuals.
Evaluation of AQP4 Expression by Western Blot
Expression of AQP4 in the frontal cortex was evaluated by West-
ern blot. As shown in eFigure 3A in the Supplement, no sig-
nificant differences in total AQP4 expression were observed

96 JAMA Neurology January 2017 Volume 74, Number 1 (Reprinted) jamaneurology.com

Copyright 2017 American Medical Association. All rights reserved.

Downloaded From: https://jamanetwork.com/ by b Wildberries on 02/25/2023

 Aquaporin-4 Localization and Cognition in Aging Brains
between young individuals, aged individuals, or individuals
with AD (P = .23, 1-way ANOVA). The discrepancy between
these findings and the age-related differences in AQP4-IR
(Figure 2A-B) may stem from the fact that our immunofluo-
rescence analysis specifically omitted analysis of cortical layer
I and the glial limitans, which exhibit pronounced AQP4-IR in
the aged and AD cortex (Figure 3 and Figure 4). Aquaporin-4
is expressed in the human brain as 2 alternative isoforms,
AQP4-M1 and AQP4-M23, the relative expression of which has
been proposed to drive perivascular localization.16 When we
evaluated expression of AQP4-M1 and AQP4-M23 separately,
we did not observe any specific differences in expression of
either isoform between groups. However, when the ratio of
AQP4-M1 to AQP4-M23 expression was evaluated, we ob-
served that the relative expression of AQP4-M1 (the AQP4-M1
to AQP4-M23 ratio) significantly declined in aged cortex (mean
[SD], 0.10 [0.09]; adjusted P = .05, 1-way ANOVA) and AD
cortex (mean [SD], 0.13 [0.02]; adjusted P = .01, 1-way ANOVA)
(eFigure 3B in the Supplement).
Differences in AQP4 Localization
Associated With Senile Plaques
Participants with AD exhibited increased AQP4-IR in cortex
(Figure 1E) in addition to reduced perivascular AQP4 local-
ization (Figure 1H). Wide field immunofluorescence imaging
showed that in contrast to the pattern of AQP4-IR in the
aged cortex (Figure 4B), AQP4-IR in the AD cortex remained
high through all cortical depths (Figure 4C). Confocal
microscopy showed that this was attributable to intense
AQP4-IR associated within glial fibrillary acidic protein–
positive reactive astrocytes surrounding cortical senile
plaques (Figure 4C-E).
Discussion
We have evaluated the association between differences in AQP4
expression and localization and Alzheimer pathology across
a wide range of aged human samples, including those from cog-
nitively intact individuals and individuals with AD. We dis-
covered that AQP4 expression increases in the aging brain and
that loss of perivascular AQP4 localization is associated with
worsening AD pathology. We observed that AQP4 expression
increased with age across all individuals. Regional patterns of
AQP4 expression and localization were dramatically altered in
the aged cortex (60-85 years of age) yet were preserved among
cognitively intact individuals older than 85 years. Perivascu-
lar AQP4 localization was reduced among individuals with AD
and reduced localization was strongly associated with in-
creased neurofibrillary and Aβ pathology, even when control-
ling for the effect of age on AD pathology. In line with these
findings, logistic regression analysis revealed that perivascu-
lar AQP4 localization was strongly associated with AD status
independent of age.
Analysis of Aβ kinetics within the CSF of clinical partici-
pants showed that Aβ clearance is impaired among par-
ticipants with late-onset AD, suggesting that slowing Aβ
clearance rather than increasing Aβ production underpins
jamaneurology.com
Copyright 2017 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ by b Wildberries on 02/25/2023
Original Investigation Research
Aβ plaque deposition.17 A similar analysis in 2015 demon-
strated that Aβ clearance slows with increasing age, includ-
ing in individuals prior to the deposition of Aβ plaques as
detected by Pittsburgh compound B positron emission
tomographic scans.18 The changes in the aging human brain
that underlie the slowing of Aβ clearance have not yet been
defined.
Amyloid β clearance from the brain appears to involve sev-
eral different processes, including uptake and degradation by
brain cells such as microglia; efflux across the blood-brain bar-
rier by transporters, such as low-density lipoprotein receptor-
related protein-1; and clearance by bulk flow along perivascu-
lar pathways. 19-21 Because each takes place within the
perivascular compartment, it appears likely that processes such
as perivascular bulk clearance and blood-brain barrier efflux
interact to support Aβ clearance from the brain.4,22 The pre-
cise nature and extent of this interaction, and how these re-
lationships may change as the brain ages, remain unknown and
are an important subjects for future research.
In the rodent brain, perivascular CSF–interstitial fluid ex-
change and interstitial Aβ clearance are markedly reduced in
the absence of the astroglial water channel AQP4,7 while in a
transgenic mouse model of AD, deletion of the Aqp4 gene in-
creased Aβ plaque formation and worsened cognitive
impairment.23 In the aging mouse cortex, impairment of peri-
vascular CSF–interstitial fluid exchange and interstitial Aβ clear-
ance was associated with reduced perivascular AQP4
localization.4 These experimental findings suggest that loss of
perivascular localization in the aging rodent brain slows Aβ
clearance and promotes Aβ deposition. Our findings here dem-
onstrate that similar age-related changes to astroglial AQP4 lo-
calization are occurring in the human as in the rodent brain,
and suggest that AQP4 mislocalization may be 1 of the factors
underlying the slowing of Aβ clearance and promoting Aβ
plaque formation in the aging human brain.
Our findings are consistent with prior studies carried out
in human postmortem cortical tissue showing that while AQP4
expression evaluated by Western blot does not significantly
change in the AD cortex,24 AQP4 immunoreactivity is substan-
tially increased in the AD cortex.8,9 Additionally, Wilcock et al10
reported that AQP4 localization to the microvasculature is re-
duced among individuals with AD. In prior studies, disrup-
tion of AQP4 expression and localization have been reported
to be associated with both Aβ plaques and vascular Aβ
deposits,9,24,25 suggesting that changes in AQP4 expression and
localization are features of the aging brain. This study ex-
tends these previous findings to include what is, to our knowl-
edge, the largest clinical case series to date to assess the rela-
tionship between changes in AQP4 expression and localization,
aging, and AD pathology. Importantly, to our knowledge, this
is the first study to explicitly quantify perivascular AQP4 lo-
calization and to evaluate whether changes in this feature,
which is believed to underlie perivascular Aβ clearance along
the glymphatic pathway,4,5,7,26 are significantly associated with
AD status or AD pathology as measured by Braak stage or Aβ
plaque density.
We did not observe evidence of Aβ plaque–associated dif-
ferences in AQP4 expression or localization in cognitively
(Reprinted) JAMA Neurology January 2017 Volume 74, Number 1 97
 Research Original Investigation Aquaporin-4 Localization and Cognition in Aging Brains

intact young or aged individuals. The relationship between in- Limitations

creasing AQP4-IR and advancing age and between differ-
ences in AQP4 localization and increasing Aβ plaque deposi-
tion and neurofibrillary pathology observed even among
cognitively intact individuals argues that the loss of perivas-
cular AQP4 localization is not the result of advancing AD pa-
thology. However, we cannot rule out the possibility that sub-
clinical Aβ plaque burden or neurofibrillary pathology, which
is present both among cognitively intact individuals and those
with mild cognitive impairment,27 contributed to differences
in AQP4 expression or localization in this study. However, the
absence of a significant association between cerebrovascular
pathology and differences in AQP4 expression or localization
when controlling for effects of age suggest that these effects
are likely not driven by vascular injury.
Age-related cognitive decline in the absence of frank de-
mentia is commonly viewed as an inherent concomitant of
normal aging.28 Yet among the nondemented aging popula-
tion, an important subset harbor a high prevalence of the neu-
ropathologies that are commonly associated with frank
dementia.29 The substrate of this resistance to age-related pa-
thologies and cognitive decline is not well understood but
has been attributed to aspects of greater brain reserve or
resilience.30 In this study, we observed loss of perivascular
AQP4 localization with increasing age among individuals with
AD, while perivascular AQP4 localization was preserved in in-
dividuals older than 85 years of age. Similarly, regional differ-
ences in AQP4 expression and localization observed in aging
individuals 60 to 85 years of age were generally absent among
these cognitively intact individuals older than 85 years of age.
Whether preservation of perivascular AQP4 localization
throughout the aging process provides protection or a re-
serve against the development of AD pathology or whether
other factors contribute independently to both reduced AD pa-
thology and increased perivascular AQP4 localization cannot
be ascertained from this data.
In experimental studies in rodents, the role of AQP4 in glym-
phatic system function has been best characterized in the cor-
tex, while the spread of amyloid and neurofibrillary pathol-
ogy throughout the frontal cortex is believed to play a key role
in the development of cognitive decline in the progression of
AD. Therefore, this initial study was limited to analysis of the
expression and localization of AQP4 in the frontal cortex.
Whether these age-associated effects also occur in other brain
areas relevant to cognitive decline, such as the hippocampus
and entorhinal cortex, is an important consideration. Be-
cause this study is reliant on postmortem tissue, it is impos-
sible from these data to verify whether AQP4 mislocalization
is itself the causative feature rendering the aging brain sus-
ceptible to neurodegenerative processes or merely a parallel
consequence of an independent upstream pathological pro-
cess. Additionally, whether preservation of perivascular AQP4
localization throughout the aging process provides protec-
tion or a reserve against development of AD pathology or
whether other factors contribute independently to both re-
duced AD pathology and increased perivascular AQP4 local-
ization cannot be ascertained from these data. Further in-
quiry into these questions is warranted.
Conclusions
These histological findings are consistent with the notion that
loss of perivascular AQP4 localization is one of the factors that
renders the aging brain vulnerable to Aβ aggregation and neu-
rodegeneration. If substantiated, this would potentially place
age-related changes in perivascular AQP4 localization mecha-
nistically upstream of processes of protein misaggregation and
would suggest that targeting AQP4 localization may provide
a therapeutic approach to intervene in the AD pathogenesis up-
stream of Aβ plaque or neurofibrillary tangle formation.

ARTICLE INFORMATION
Accepted for Publication: September 9, 2016.
Published Online: November 28, 2016.
doi:10.1001/jamaneurol.2016.4370
Author Contributions: Dr Iliff had full access to all
the data in the study and takes responsibility for the
integrity of the data and the accuracy of the data
analysis.
Concept and design: Zeppenfeld, Haswell, Iliff.
Acquisition, analysis, or interpretation of data: All
Authors.
Drafting of the manuscript: Zeppenfeld, Haswell,
D'Abreo, Murchison, Grafe, Iliff.
Critical revision of the manuscript for important
intellectual content: Zeppenfeld, Simon, Murchison,
Quinn, Grafe, Woltjer, Kaye, Iliff.
Statistical analysis: D'Abreo, Murchison, Iliff.
Administrative, technical, or material support:
Zeppenfeld, Simon, Grafe, Woltjer, Kaye, Iliff.
Study supervision: Zeppenfeld, Grafe, Iliff.
Conflict of Interest Disclosures: Dr Iliff reports
serving as a consultant for Shire and receiving
honoraria from Shire Pharmaceuticals and
Genentech. No other disclosures are reported.
Funding/Support: This work was supported by
funding from the American Heart Association grant
12SDG11820014 (Dr Iliff), the Oregon Partnership
for Alzheimer’s Research (Dr Iliff), grants from the
Research and Development Office of the
Department of Veterans Affairs (Merit Review
Grant, Dr Kaye), and grant NS089709 from the
National Institutes of Health (Dr Iliff), including
Alzheimer’s Disease Center grant AG08017 from
the National Institute on Aging that supported the
longitudinal follow-up and subsequent brain
autopsies providing the human brain samples used
in this study.
Role of the Funder/Sponsor: The funding sources
had no role in the design and conduct of the study;
collection, management, analysis, and
interpretation of the data; and preparation, review,
or approval of the manuscript; and decision to
submit the manuscript for publication.
REFERENCES
1. Alzheimer’s Association. 2015 Alzheimer’s
disease facts and figures. Alzheimers Dement. 2015;
11(3):332-384.
2. Ross CA, Poirier MA. Protein aggregation and
neurodegenerative disease. Nat Med. 2004;10
(suppl):S10-S17.
3. Plog BA, Moll KM, Kang H, et al. A novel
technique for morphometric quantification of
subarachnoid hemorrhage-induced microglia
activation. J Neurosci Methods. 2014;229:44-52.
4. Kress BT, Iliff JJ, Xia M, et al. Impairment of
paravascular clearance pathways in the aging brain.
Ann Neurol. 2014;76(6):845-861.
5. Xie L, Kang H, Xu Q, et al. Sleep drives
metabolite clearance from the adult brain. Science.
2013;342(6156):373-377.
6. Iliff JJ, Lee H, Yu M, et al. Brain-wide pathway for
waste clearance captured by contrast-enhanced
MRI. J Clin Invest. 2013;123(3):1299-1309.
7. Iliff JJ, Wang M, Liao Y, et al. A paravascular
pathway facilitates CSF flow through the brain
parenchyma and the clearance of interstitial
solutes, including amyloid β. Sci Transl Med. 2012;4
(147):147ra111.
8. Moftakhar P, Lynch MD, Pomakian JL, Vinters
HV. Aquaporin expression in the brains of patients

98 JAMA Neurology January 2017 Volume 74, Number 1 (Reprinted) jamaneurology.com

Copyright 2017 American Medical Association. All rights reserved.

Downloaded From: https://jamanetwork.com/ by b Wildberries on 02/25/2023

 Aquaporin-4 Localization and Cognition in Aging Brains
with or without cerebral amyloid angiopathy.
J Neuropathol Exp Neurol. 2010;69(12):1201-1209.
9. Hoshi A, Yamamoto T, Shimizu K, et al.
Characteristics of aquaporin expression
surrounding senile plaques and cerebral amyloid
angiopathy in Alzheimer disease. J Neuropathol Exp
Neurol. 2012;71(8):750-759.
10. Wilcock DM, Vitek MP, Colton CA. Vascular
amyloid alters astrocytic water and potassium
channels in mouse models and humans with
Alzheimer’s disease. Neuroscience. 2009;159(3):
1055-1069.
11. Nelson JW, Young JM, Borkar RN, et al. Role of
soluble epoxide hydrolase in age-related vascular
cognitive decline. Prostaglandins Other Lipid Mediat.
2014;113-115:30-37.
12. Tierney MC, Fisher RH, Lewis AJ, et al. The
NINCDS-ADRDA Work Group criteria for the clinical
diagnosis of probable Alzheimer’s disease:
a clinicopathologic study of 57 cases. Neurology.
1988;38(3):359-364.
13. R: A Language and Environment for Statistical
Computing [computer program]. Vienna, Austria: R
Foundation for Statistical Computing; 2013.
14. Morris JC. The Clinical Dementia Rating (CDR):
current version and scoring rules. Neurology. 1993;
43(11):2412-2414.
15. Deramecourt V, Slade JY, Oakley AE, et al.
Staging and natural history of cerebrovascular
pathology in dementia. Neurology. 2012;78(14):
1043-1050.
jamaneurology.com
Copyright 2017 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ by b Wildberries on 02/25/2023
16. Crane JM, Tajima M, Verkman AS. Live-cell
imaging of aquaporin-4 diffusion and interactions in
orthogonal arrays of particles. Neuroscience. 2010;
168(4):892-902.
17. Mawuenyega KG, Sigurdson W, Ovod V, et al.
Decreased clearance of CNS beta-amyloid in
Alzheimer’s disease. Science. 2010;330(6012):1774.
18. Patterson BW, Elbert DL, Mawuenyega KG,
et al. Age and amyloid effects on human central
nervous system amyloid-beta kinetics. Ann Neurol.
2015;78(3):439-453.
19. Tarasoff-Conway JM, Carare RO, Osorio RS,
et al. Clearance systems in the brain-implications
for Alzheimer disease. Nat Rev Neurol. 2015;11(8):
457-470.
20. Zlokovic BV, Deane R, Sagare AP, Bell RD,
Winkler EA. Low-density lipoprotein
receptor-related protein-1: a serial clearance
homeostatic mechanism controlling Alzheimer’s
amyloid β-peptide elimination from the brain.
J Neurochem. 2010;115(5):1077-1089.
21. Simon MJ, Iliff JJ. Regulation of cerebrospinal
fluid (CSF) flow in neurodegenerative,
neurovascular and neuroinflammatory disease.
Biochim Biophys Acta. 2016;1862(3):442-451.
22. Shibata M, Yamada S, Kumar SR, et al.
Clearance of Alzheimer’s amyloid-ss (1-40) peptide
from brain by LDL receptor-related protein-1 at the
blood-brain barrier. J Clin Invest. 2000;106(12):
1489-1499.
23. Xu Z, Xiao N, Chen Y, et al. Deletion of
aquaporin-4 in APP/PS1 mice exacerbates brain Aβ
(Reprinted) JAMA Neurology January 2017 Volume 74, Number 1
Original Investigation Research
accumulation and memory deficits. Mol
Neurodegener. 2015;10(1):58.
24. Pérez E, Barrachina M, Rodríguez A, et al.
Aquaporin expression in the cerebral cortex is
increased at early stages of Alzheimer disease.
Brain Res. 2007;1128(1):164-174.
25. Yang J, Lunde LK, Nuntagij P, et al. Loss of
astrocyte polarization in the tg-ArcSwe mouse
model of Alzheimer’s disease. J Alzheimers Dis.
2011;27(4):711-722.
26. Iliff JJ, Goldman SA, Nedergaard M.
Implications of the discovery of brain lymphatic
pathways. Lancet Neurol. 2015;14(10):977-979.
27. Chételat G, La Joie R, Villain N, et al. Amyloid
imaging in cognitively normal individuals, at-risk
populations and preclinical Alzheimer’s disease.
Neuroimage Clin. 2013;2:356-365.
28. Institute of Medicine (US), Committee on the
Public Health Dimensions of Cognitive Aging, Blazer
DG, Yaffe K, Liverman CT, eds. Cognitive Aging:
Progress in Understanding and Opportunities For
Action. Washington, DC: The National Academies
Press; 2015.
29. Sonnen JA, Santa Cruz K, Hemmy LS, et al.
Ecology of the aging human brain. Arch Neurol.
2011;68(8):1049-1056.
30. Erten-Lyons D, Woltjer RL, Dodge H, et al.
Factors associated with resistance to dementia
despite high Alzheimer disease pathology. Neurology.
2009;72(4):354-360.
99