A novel recombinant adeno-associated virus vaccine reduces behavioral impairment and beta-amyloid plaques in a mouse model of Alzheimer's disease.

Abstract

Memory impairment progressing to dementia is the main clinical symptom of Alzheimer's disease (AD). Deposition of the amyloid-beta peptide (Abeta) in brain, particularly its 42-amino acid isoform (Abeta42), has been shown to play a primary and crucial role in the pathogenesis of AD. In this study we have developed a recombinant adeno-associated virus (AAV) vaccine against AD. This vaccine could express CB-Abeta42 (cholera toxin B subunit and Abeta42 fusion protein) in vivo. A single administration of the AAV-CB-Abeta42 vaccine induced a prolonged, strong production of Abeta-specific serum IgG in transgenic mice that overexpressed the London mutant of amyloid precursor protein (APP/V717I), and resulted in improved ability of memory and cognition, decreased Abeta deposition in the brain, and a resultant decrease in plaque-associated astrocytosis. Our results extended the immunological approaches for the treatment and prevention of AD to an oral, intranasal, or intramuscular route that might be better tolerated in human patients than repetitive parental immunizations in the presence of adjuvant. AAV has attracted tremendous interest as a promising vector for gene delivery. Our results raised the possibility that AAV-CB-Abeta42 vector immunization may provide the basis of a novel and promising Alzheimer's disease vaccination program.

Full text

Page 1

A novel recombinant adeno-associated virus vaccine reduces behavioral
impairment and ?-amyloid plaques in a mouse model
of Alzheimer’s disease
Jianmin Zhang,aXiaobing Wu,bChuan Qin,cJin Qi,aShibin Ma,aHuiyuan Zhang,a
Qingli Kong,aDongqing Chen,aDenian Ba,aand Wei Hea,*
aDepartment of Immunology, Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences & School of Basic Medicine,
Peking Union Medical College, Beijing 100005, People’s Republic of China
bAGTC Gene Technology Company Ltd., National High-Tech Project, R&D Base of Viral Gene Vector, No.6 Yong Chang Zhong Road,
BDA Beijing, 100176, China
cDepartment of Pathology, Institute of Experimental Animal, Chinese Academy of Medical Sciences & Peking Union Medical College,
Pan Jia Yuan Nan Li, Beijing 100005, People’s Republic of China
Received 25 April 2003; revised 8 July 2003; accepted 15 July 2003
Abstract
Memory impairment progressing to dementia is the main clinical symptom of Alzheimer’s disease (AD). Deposition of the amyloid-?
peptide (A?) in brain, particularly its 42-amino acid isoform (A?42), has been shown to play a primary and crucial role in the pathogenesis
of AD. In this study we have developed a recombinant adeno-associated virus (AAV) vaccine against AD. This vaccine could express
CB-A?42 (cholera toxin B subunit and A?42 fusion protein) in vivo. A single administration of the AAV-CB-A?42 vaccine induced a
prolonged, strong production of A?-specific serum IgG in transgenic mice that overexpressed the London mutant of amyloid precursor
protein (APP/V717I), and resulted in improved ability of memory and cognition, decreased A? deposition in the brain, and a resultant
decrease in plaque-associated astrocytosis. Our results extended the immunological approaches for the treatment and prevention of AD to
an oral, intranasal, or intramuscular route that might be better tolerated in human patients than repetitive parental immunizations in the
presence of adjuvant. AAV has attracted tremendous interest as a promising vector for gene delivery. Our results raised the possibility that
AAV-CB-A?42 vector immunization may provide the basis of a novel and promising Alzheimer’s disease vaccination program.
© 2003 Elsevier Inc. All rights reserved.
Introduction
Alzheimer’s disease (AD) is a neurodegenerative disor-
der characterized by a progressive decline of cognitive abil-
ities and by neuropathological features, including diffuse
loss of neurons in the hippocampus and neocortex, accu-
mulation of intracellular protein deposits (neurofibrillary
tangles), and accumulation of extracellular protein deposits
(amyloid or senile plaques) characteristically seen in the
associative cortices and limbic system. A main constituent
of these amyloid plaques is the amyloid-? peptide (A?), a
39–43-amino acid protein derived from the processing of a
large transmembrane protein, the ?-amyloid precursor pro-
tein (APP). It has been postulated that accumulation of A?
in the brain, resulting from abnormal processing of APP to
A? or reduced clearance of A? from brain, is the primary
cause of AD pathogenesis. The rest of the disease process,
including formation of neurofibrillary tangles containing tau
protein, is thought to result from an imbalance between A?
production and A? clearance (Hardy and Selkoe, 2002). As
a result, there is a widespread interest in developing thera-
peutic approaches to improve the balance between A? pro-
duction and A? clearance. One possible approach is to
* Corresponding author. Department of Immunology, Institute of Basic
Medical Sciences, Chinese Academy of Medical Science & School of
Basic Medicine, Peking Union Medical College, 5 Dong Dan San Tiao,
Beijing 100005, People’s Republic of China. Fax: ?86-10-65249259.
E-mail address: heweiimu@public.bta.net.cn or zjmwh@yahoo.com.cn
(W. He).
R
Available online at www.sciencedirect.com
Neurobiology of Disease 14 (2003) 365–379 www.elsevier.com/locate/ynbdi
0969-9961/$ – see front matter © 2003 Elsevier Inc. All rights reserved.
doi:10.1016/j.nbd.2003.07.005

Page 2

reduce their synthesis or deposition by targeting the en-
zymes that process APP to A? peptides. Other approaches
are aimed at increasing the clearance of A? peptides from
brain or inhibiting their aggregation to form insoluble de-
posits. Recently, immunotherapy for AD has been the sub-
ject of intense investigation. Both active and passive immu-
nization against A? could reduce the level of A?, prevent
and clear amyloid plaques, and improve cognitive behavior
in transgenic mouse models of AD (Bard et al., 2000; Chen
et al., 2000; Janus et al., 2000b; Morgan et al., 2000; Schenk
et al., 1999), in which antibodies to A? decreased the
cerebral level of the peptide by promoting microglial clear-
ance (Morgan et al., 2000; Schenk et al., 1999) and/or by
redistributing the peptide from the brain to the systemic
circulation (DeMattos et al., 2001).
These results raised the possibility of immunotherapy
against human AD. But these results also demonstrated that
high levels of anti-human A? antibody were necessary for
the effect to be detected in mice. AD is a chronic progres-
sive neurodegenerative disease that may cause suffering in
many patients for many years. It is difficult to maintain high
levels of anti-A? antibodies in the human body for such a
long time. Repeated injections of A? peptides or anti-A?
antibodies is necessary. Although active immunization with
synthetic A?42 peptide produced marked benefits in APP
transgenic mice, a phase II trial showed that roughly 5% of
the treated participants developed an inflammatory reaction
in the CNS, presumably related to the induction of TH1-type
responses (Weiner and Selkoe, 2002). On the other hand, in
passive immunization, the principal problem is the short
half-life of anti-A? antibodies in the human body. More-
over, repetitive passive immunization might give rise to
anti-antibody response. So it is necessary to develop a
method that will induce long-term, high levels of anti-A?
antibodies, but suppress cellular immunity.
In the present study, we have developed a recombinant
adeno-associated virus (rAAV) vaccine against AD. The
rAAV vector expressed a fusion protein of cholera toxin B
subunit (CB) and A?42 in vivo. CB is a nontoxic subunit of
cholera toxin and a non-TH1-inducing adjuvant (Woogen et
al., 1993). Numerous experimental animal studies have
demonstrated that oral or nasal administration of antigens
fused to CB can induce vigorous humoral immunity, but
suppress cellular immunity (McKenzie and Halsey, 1984;
Williams et al., 1999; Wu et al., 1996). Viral vector-medi-
ated gene therapies have been used in clinical trials for
human diseases. Many research groups have demonstrated
the feasibility of using rAAV vector to deliver therapeutic
genes to animal and human islets (Jindal et al., 2001; Prasad
et al., 2000; Yang and Kotin, 2000). One of the advantages
of using rAAV vector is the ability to extend the period of
transgene expression both in vitro and in vivo. Adeno-
associated virus is a small, single-stranded DNA virus lack-
ing an envelope (Koeberl et al., 1999). The virus requires a
helper virus to facilitate efficient replication. It has several
features that make it particularly useful for gene therapy.
First, the foremost feature of AAV is its safety. AAV has
been found in many animal species, including nonhuman
primates and human beings, and has never shown any
pathogenicity. This virus is the only nonpathogenic viral
vector now available and has been successfully used to
establish long-term gene expression without toxicity in vivo
in both dividing and nondividing cells. Vectors can be
generated that are completely free of helper virus (Samulski
et al., 1989). Recombinant AAV vectors, with the entire
coding sequences removed, retain only terminal repeats of
145 base pairs. These vectors, therefore, are devoid of all
viral genes, minimizing any possibility of recombination
and viral gene expression. Second, another feature of AAV
is its hardiness. AAV is resistant to temperature and pH
extremes and solvents, which makes it particularly suitable
as an orally delivered vector (During et al., 1998). Further-
more, during active infection in humans, wild-type AAV is
typically found in both respiratory and gastrointestinal tract
secretions; thus, the gut is a normal host tissue for the virus.
Finally, AAV vectors can produce long-term gene expres-
sion in a number of tissues after in vivo administration,
including muscle, gut, liver, and eye (Koeberl et al., 1999).
On the basis of these studies and the characteristics of the
AAV vector, we created a recombinant AAV vaccine to test
the hypothesis that a single administration of this AAV
vaccine to PDAPPV717Itransgenic mice, a mouse model of
AD with the human APP751 London mutation (V717I),
could produce a prolonged, efficient induction of protective
anti-A? antibodies, together with a reduction of behavioral
impairment and ?-amyloid plaques. The A?-specific hu-
moral response of the immunized animals was also evalu-
ated when using different routes of administration of the
vaccine.
Materials and methods
Construction of plasmids
As shown in Fig. 1, we constructed three AAV plasmids:
pSNAV-CB-A?42,pSNAV-A?42,
(green fluorescent protein, GFP). To construct plasmid
pSNAV-CB-A?42, which produced CB-GPGP-A?42 fu-
sion protein, the coding sequence of cholera toxin B sub-
unit, including its putative leader sequence (which codes the
signal peptide that is absent in the mature CB subunit of V.
cholerae), was prepared from the ctx AB operon in plasmid
pGEM7zf(?) (provided by Dr. Biao Kan.) by PCR ampli-
fication, using a 5? primer (5?-CCGGGGTACCCCACCAT-
GATTAAATTAAAATTTGGTG-3?) containing a KpnI re-
striction site and a 3? primer (5?-CTGCATCAGG-
ACCAGGACCATTTGCCATACTAATTGCG-3?), which
is complementary to the GPGP coding sequence and the
first 19 bp of A?42 coding sequence. A?42 coding se-
quence was PCR amplified from the cDNA library of hu-
man brain (provided by Dr. Xiaoyan Hu) with the sense
andpSNAV-GFP
366
J. Zhang et al. / Neurobiology of Disease 14 (2003) 365–379

Page 3

primer
CATGAC-3?) containing the coding sequence of GPGP
(glycine-proline-glycine-proline) and the antisense primer
(5?-GGAAGATCTTTACTACGCTATGACAACACCGC-
CC-3?) complementary to the last 19 bp of A?42 coding
sequence followed by a stop codon and the BglII restriction
site. The cholera toxin B subunit coding sequence and A?42
coding sequence were linked through assembly and fill-in
reactions (Sugahara et al., 1995), and then amplified by PCR
with the 5? primer (5?-CCGGGGTACCCCACCATG-3?)
containing the KpnI restriction site, and the 3? primer (5?-
GGAAGATCTTTACTACGCTATG-3?) containing the Bg-
lII restriction site. The plasmid pSNAV-CB-A?42 was gen-
erated by insertion of the KpnI-BglII-digested PCR product
in the BglII-KpnI sites of pSNAV.
For pSNAV-A?42 encoding A?42 peptide, we first syn-
thesized the leader sequence of APP (5?-CCGGGGTA-
CCCCACCATGCTGCCCGGTTTGGCACTGCTCCTG-
CTGGCCGCCTGGACGGCTCGGGCGGATGCAGAAT-
TCCGACATGAC-3?), containing the KpnI restriction site
at the 5? end and the first 21 bp of A?42 coding sequence
encoding the 1?7 aa of A?42. The A?42 coding sequence
was PCR amplified from the plasmid pSNAV-CB-A?42
with the sense primer (5?-GATGCAGAATTCCGACAT-
GAC-3?) and the antisene primer (5?-GGAAGATCTTTAC-
TACGCTATGACAACACCGCCC-3?), which is comple-
mentary to the last 19 bp of A?42 coding sequence followed
by a stop codon and the BglII restriction site. The leader
sequence of APP and A?42 coding sequence were linked
through assembly and fill-in reactions (Sugahara et al.,
1995), and amplified by PCR with the 5? primer (5?-
CCGGGGTACCCCACCATG-3?) containing the KpnI re-
striction site, and the 3? primer (5?-GGAAGATCTTTAC-
TACGCTATG-3?) containing the BglII restriction site. The
plasmid pSNAV-A?42 was generated by insertion of the
KpnI-BglII-digested PCR product in the BglII-KpnI sites of
pSNAV. For the plasmid pSNAV-GFP, the GFP cDNA
fragment was isolated from the plasmid pEGFP-N1 (Clon-
tech, Palo Alto, CA) by BglII and KpnI digestion, and then
inserted into the BglII-KpnI sites of pSNAV.
(5?-GGTCCTGGTCCTGATGCAGAATTCCGA-
Cell culture and AAV vector production
Baby hamster kidney (BHK)-21 cells were obtained
from the American Type Culture Collection and maintained
in RPMI-1640 (GibcoBRL, Grand Island, NY, USA) sup-
plemented with 10% fetal bovine serum (Hyclone, Logan,
UT, USA), 50 U/ml penicillin, and 50 ?g/ml streptomycin
at 37°C in an atmosphere of 5% CO2. The BHK-21 cells
were transfected by the calcium phosphate coprecipitation
method with the plasmid (pSNAV-CB-A?42, pSNAV-
A?42, or pSNAV-GFP) according to the protocol of Cal-
Phos ProFection Mammalian Transfection System (Pro-
mega, Madison, WI, USA). Individual BHK-21 cell clones
were isolated after selection in 0.5 mg/ml G418 (Promega).
Southern blotting was used to determine the integrity of the
plasmids incorporated into BHK-21 cell chromosomes.
The AAV containing CB-A?42, A?42, or GFP gene was
packaged by infecting G418-resistant BHK-21 cells that had
already integrated with one of the plasmids (pSNAV-CB-
A?42, pSNAV-A?42, or pSNAV-GFP) with recombinant
herpes simplex virus, which can express rep and cap genes
of wild-type AAV. After 48 h, the cells were harvested and
lysed in Tris buffer (10 mM Tris-HCl, 150 mM NaCl, pH
8.0) by three cycles of freezing and thawing. One round of
sucrose precipitation and two rounds of CsCl density-gra-
dient ultracentrifugation were sufficient to isolate the AAV
vectors from the lysates. The vector titer was determined by
quantitative DNA slot-blot hybridization of the DNase I-re-
sistant fraction.
Animals and animal immunization
PDAPPV717Itransgenic mice (their parents are C57/BL6
black) express high levels of human APP751 containing the
London (V717I) mutation, which can markedly increase the
generation of A?42 (Sturchler-Pierrat et al., 1997). These
mice have readily detectable cognitive impairment from
3?5 months onwards and their A?42 levels in brain and
those of the slightly shorter and less amyloidogenic A?40
rise dramatically during the first year of life, resulting in the
formation of multiple A? plaques by 12 months of age as
previously described (Dewachter et al., 2000). PDAPPV717I
mice were obtained from the Institute of Experimental An-
imal, Chinese Academy of Medical Sciences & Peking
Union Medical College (Beijing, China) at the age of ap-
proximately 1 month (prophylactic group) and 12 months
(therapeutic group) and housed in standardized conditions
as described elsewhere (Cavallaro et al., 1997) in an animal
room compliant with the Public Health Service Policy on
Humane Care and Use of Laboratory Animals. Mice were
divided into two groups: the prophylactic group, 1 month-
old mice without any behavioral impairment or amyloid
plaques in their brains, and the therapeutic group, 12-
month-old mice having manifested typical behavioral im-
pairment and amyloid plaques in their brains. The therapeu-
tic group included AAV-CB-A?42, AAV-A?42, AAV-
GFP treated and untreated subgroups. The prophylactic
group included three of the above subgroups with the dele-
tion of the AAV-A?42 subgroup. We used three different
routes in administering the AAV-CB-A?42 and AAV-
A?42 recombinant vectors, i.e., oral, intranasal (i.n.), and
intramuscular (i.m.). The AAV-GFP vector was adminis-
tered via the intramuscular route. For intranasal immuniza-
tion, mice were anesthetized with ethyl ether and about 20 s
later, 30 ?l of AAV vector containing 5 ? 1010particles
was dripped into a nostril. The mice were able to absorb
such preparations simply by breathing. For intramuscular
immunization, 100 ?l of AAV vector containing 5 ? 1010
particles was injected into the gastrocnemius muscle. For
oral immunization, 100 ?l of AAV vector containing 5 ?
367
J. Zhang et al. / Neurobiology of Disease 14 (2003) 365–379

Page 4

1010particles was administered via an orogastric tube into
the stomach. Both in the prophylactic and in the therapeutic
groups, a group without any treatment and a group receiving
intramuscular immunization with 5 ? 1010particles of a
recombinant AAV virus expressing GFP (AAV-GFP)
served as controls. Each group consisted of 10 mice. Equal
numbers of nontransgenic C57/BL6 littermates served as
gene-background controls. One mouse in each group was
used to determine the expression of CB-A?42, A?42, or
GFP in vivo 1 month after treatment. Four PDAPPV717I
transgenic mice and two nontransgenic mice died of unre-
lated causes during the course of the study.
Identification of expression
The expression of CB-A?42 or A?42 protein in the
intestine, nostril, and muscle was detected by immunohis-
tochemistry with a commercially available A?42 primary
antibody and horseradish peroxidase (HRP)-conjugated sec-
ondary antibody 1 month, 5 months (only in the therapeutic
group), and 12 months (only in the prophylactic group) after
immunization. The expression of GFP in muscles was ex-
amined in frozen section of gastrocnemius muscle under the
fluorescence microscope.
ELISA analysis for serum anti-human A? antibodies
Samples of immune sera were taken from the tail veins of
mice 1, 2, 5, and 12 (only in the prophylactic group) months
after immunization. Anti-human A? antibody titers were
quantified by sandwich ELISA. Ninety-six-well microtiter
plates (Corning Costar Corp., Cambridge, MA, USA) were
coated overnight at 4°C with 100 ?l of 2 ?g/ml synthetic
human A?42 in 50 mM carbonate buffer, pH 9.6. The wells
were washed four times with phosphate-buffered saline
(PBS) containing 0.05% Tween 20 (PBS-T), and blocked
with 200 ?l of blocking buffer (5% goat serum and 1%
bovine serum albumin in PBS-T) for 4 h at room tempera-
ture. Mouse serum was prepared in PBS at an initial dilution
of 1:16 and subsequent two-fold dilutions were made. All
samples were run in duplicate and simultaneously incubated
at 37°C for 1 h followed by washing for six times with
PBS-T. Plates were blocked a second time with blocking
buffer for 30 min at 37°C followed by washing for five
times; 100 ?l of HRP-conjugated goat anti-mouse IgG
(Santa Cruz Biotechnologies, Santa Cruz, CA, USA) diluted
to 1:10,000 in PBS were added to the wells and incubated
for 1 h at 37°C. Plates were then washed for six times with
PBS-T and developed with 100 ?l of 3,3,5,5-tetramethyl-
benzidine solution (Roche-Boehringer-Mannheim, Mann-
heim, Germany). The coloring reaction was stopped with
100 ?l of 2 M sulphuric acid. Plates were read spectropho-
tometrically at 450 nm in a microplate reader (Labsystems
IEMS, Helsinki, Finland).
Cell culture and neutralization test by serum anti-human
A? antibody
Primary cultures of cortical neurons were prepared from
newborn mice as described previously (Dawson et al.,
1991). Cells were plated in 96-well culture dishes (4 ? 105
cells/cm2), pretreated with poly-L-lysine (10 ?g/ml in PBS),
and cultured for 6 days in vitro in NEUROBASAL medium
supplemented with 2% B-27 and 0.5 mM L-glutamine. Un-
der these conditions, neuronal cultures (up to 98% of neu-
rons) display high differentiation and survival rates (Daw-
son et al., 1991). Then, cells were cultured in a serum-free
medium supplemented with 2% B-27 prior to the neutral-
ization test.
The prevention of A?42 neurotoxicity was measured as
previously described (Frenkel et al., 2000). Briefly, 0.12
?M A?42 was incubated for a week at 37°C to produce
fibrils, and then incubated with the sera of AAV-immunized
animals at dilutions of 10:1 or 50:1 for 24 h. The reaction
mixtures were added to the wells containing primary cul-
tures of neurons and incubated at 37°C for 2 days. Cell
viability was assessed by measuring cellular redox activity
with 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium
bromide (MTT) as described previously (Frenkel et al.,
2000).
Water maze test
Water maze tests were carried out 5 months after immu-
nization in the therapeutic group and 12 months after im-
munization in the prophylactic group.
Apparatus
The water maze apparatus consisted of a circular tank
(1.2 m in diameter and 0.47 m high), manufactured accord-
ing to Janus et al. (2000a). The inner surface of the tank was
covered with white paint. During testing, the tank was filled
to a depth of 25 cm with water (24–25°C), which was made
turbid with 2 kg of powdered milk. An escape platform, 9
cm in diameter, made of white plastic was submerged 0.5
cm below the water surface in the center of one of the four
quadrants of the maze. During a probe trial the platform was
removed from the pool. In the cue learning test (a visible
platform version of the water maze), the platform position
was marked with a 10 cm high, 1 cm in diameter post
painted with black and white horizontal stripes, and fitted
with a 2.5-cm white ball at the top. The tank was placed in
a small room (3 ? 4 m) with many extramaze cues (ceiling
lights, animal cages, experimental box, and so on). The
positions of the extramaze cues were constant throughout
the study. The movements of the animal in the tank were
recorded by a video camera suspended 2.5 m above the
center of the tank and connected to a video tracking system
(VIOS-88, Biomedica, Japan), and analyzed with a com-
puter.
368
J. Zhang et al. / Neurobiology of Disease 14 (2003) 365–379

Page 5

Procedure
The procedure of the water maze test was a modified
version of the procedure previously described (Janus et al.,
2000a). Briefly, 1 day before the first training session in the
water maze, all mice underwent nonspatial pretraining
(NSP), to assess swimming abilities and familiarize mice
with the test. At first, each mouse was allowed to swim
freely for 60 s in the tank without the escape platform. Then,
they were released from one end of a 15-cm-wide alley
(made of two pieces of white plastic inserted parallel to each
other into the water) and allowed to swim to the platform at
the other end. Three alley trials, with 30-min intertrial
intervals (ITI) were carried out. Conventional place dis-
crimination training during which a hidden platform was
fixed in the center of the NW (northwestern) quadrant
(named the target quadrant, TQ) of the tank started on the
following day and continued for 6 days with four trials per
day (ITI about 30 min). Each mouse was gently lowered
into the water, facing the wall of the tank, at the SE (south-
eastern) quadrant. The release points during each session
were constant. The trial ended when a mouse climbed the
platform, or when a maximum of 60 s elapsed. If, during a
trial, a mouse failed to find a platform within the allocated
60 s, it was guided to find the platform by an experimenter.
At the end of each trial, the mouse was allowed to rest on
the platform for 10 s. After the last trial on day 6, the
platform was removed from the pool and each mouse re-
ceived one 60-s swim “probe trial.” For the probe trials, an
annulus crossing index was calculated, which represents the
number of passes over the platform site minus the mean of
passes over alternative sites in other quadrants.
One week after the end of the place discrimination train-
ing, the cue navigation tests with a visible platform marked
by a flag were carried out for 4 days with four trials per day.
Escape latency (in seconds), length of swim path (in centi-
meters), and swim speed (cm/s) were recorded using a video
tracking system.
Data analyses
The water maze data from four consecutive trials in each
daily session were pooled together into four-trial blocks for
the analysis. Preliminary analysis of data revealed no sig-
nificant difference in swim speed between the groups, and
significant correlation was found between latency and path
length to reach the escape platform. Therefore, the final
analysis of the data was performed on the basis of latency
and percentage of time spent by the mice in the target
quadrant. These measures of spatial learning were recom-
mended in cases with no difference in swim speed between
study groups (Morris, 1984).
Sample collection
After completion of the water maze tests, all mice were
anesthetized with pentobarbital, exsanguinated, and per-
fused intracardially with ice-cold 0.9% saline. The brain
was removed and divided sagittally along the interhemi-
spheric fissure. One hemisphere was immersion-fixed in
fresh, buffered 4% paraformaldehyde for 24 h at room
temperature. Frozen sections were stained for A? peptides
by immunohistochemistry. The contralateral hemispheres
were snap-frozen in a mixture of isopentane and liquid
nitrogen and stored at ?80°C for A? ELISA.
Immunohistochemical staining of mouse brains
For immunohistochemical studies, 4% paraformalde-
hyde-fixed cerebral hemispheres from transgenic and con-
trol mice were cut into 40-?m sagittal sections using a Leica
(Bannockburn, IL, USA) frozen microtome and placed on
poly-L-lysine-coated slides with two sections per slide. Ev-
ery sixth pair of sections was blocked in dilute (3%) hydro-
gen peroxide and nonimmune goat serum, and then immu-
noreacted with the monoclonal antibody 3D6 (specific for
the free amino-terminal region of A?). The specificity of
immunoreaction was confirmed by preadsorption with the
appropriate peptides as well as by checking that no signal
was detected when the primary antibody was omitted.
Quantification of A? deposits in mouse brain by image
analysis
Analysis of immunoreactive deposits for A? was done
on a computer using the public-domain program NIH Image
J (http://rsb.info.nih.gov/nih-image), by defining a specific
brain region (hippocampus or cortex) and setting a mono-
chromatic based threshold to discriminate nonspecific stain-
ing and to select pixels corresponding to immunolabelled
structure. Serial images of frozen coronal sections stained
simultaneously with A? antibody 3D6 were all captured
consecutively in one sitting. The percentage of brain region
covered by A? immunoreaction was used to measure A?
burden. For all image analyses, six sections at the level of
the dorsal hippocampus, each separated by consecutive
240-?m intervals, were evaluated for each animal. A con-
servative nonparametric statistical test, the Mann-Whitney
U test, was performed to compare the percentages of hip-
pocampal and cortical area occupied by A? in the AAV-
CB-A?42-treated mice versus the other groups (untreated,
AAV-GFP-, or AAV-A?42-treated) mice.
ELISA for brain A?
For ELISA of brain A? (Johnson-Wood et al., 1997),
frozen cerebral hemisphere was homogenized in 10 vol-
umes of ice-cold guanidine buffer (5.0 M guanidine HCl/50
mM Tris Cl, pH 8.0) and mixed for 3 to 4 h at room
temperature. The homogenates were further diluted 1:10
with ice-cold casein buffer (0.25% casein/0.05% sodium
azide/20 ?g/ml aprotinin/5 mM EDTA, pH 8.0/10 ?g/ml
leupeptin in PBS) before centrifugation (16,000 ? g for 20
369
J. Zhang et al. / Neurobiology of Disease 14 (2003) 365–379

Page 6

min at 4°C). Final dilutions were made in 0.5 M guanidine
and 0.1% bovine serum albumin and assessed for A?40 or
A?42 using commercially available ELISA kits specific for
either A?40 or A?42 and calibrated with synthetic A?
peptides (Biosource, Camarillo, CA, USA). The A?40
ELISA does not display any cross-reactivity with A?42 or
A?43, and the A?42 ELISA does not react with either
A?40 or A?43. Each brain was analyzed in duplicate, with
the average value reported for each brain.
Detecting the reactive astrocytes in mouse brain
To detect the reactive astrocytes in mouse brain, we used
immunohistochemistry to detect the expression of glial
Fig. 1. Schematic outline of the construction of the three plasmids (pSNAV-CB-A?42, pSNAV-A?42, and pSNAV-GFP).
Fig. 2. The expression of GFP, CB-A?42, and A?42 proteins in muscle 1 month after AAV-GFP, AAV-CB-A?42, and AAV-A?42 immunization,
respectively. The expression of CB-A?42 or A?42 was found in the muscle from AAV-CB-A?42 (B) or AAV-A?42-immunized mice (C), respectively,
whereas no positive signal was detected in the muscle from untreated mice (A) by immunohistochemistry with a commercially available A?42 primary
antibody and HRP-conjugated secondary antibody. The expression of GFP in the muscle from AAV-GFP-immunized mice (E) was found in frozen section
of gastrocnemius muscle under the fluorescence microscope.
370
J. Zhang et al. / Neurobiology of Disease 14 (2003) 365–379

Page 7

fibrillary acidic protein (GFAP) in astrocytes. Immunohis-
tochemical studies were performed with the biotin/avidin
system using anti-GFAP monoclonal antibody (G-A-5,
Roche Molecular Biochemicals).
Statistical analysis
Unless otherwise stated, the data in the text and figures
are expressed as mean ? SE. Statistical analysis was per-
formed using SPSS software (version 10.0, Chicago, IL,
USA). Statistical comparisons between groups were as-
sayed using ANOVA analyses for testing the significance of
values. A P value ? 0.05 was considered statistically sig-
nificant.
Results
Identification of expression
We determined the expression of GFP, CB-A?42, and
A?42 proteins in the intestine, nostril, and muscle 1 month,
5 months, and 12 months after immunization. The results
showed that A?42 proteins could be detected 1 month after
immunization, increased at 2 months, and lasted for 12
months after immunization (vide infra). Fig. 2 showed that
GFP, CB-A?42, and A?42 proteins were definitely ex-
pressed in the muscles 1 month after administration of their
respective vaccines.
A?-specific serum IgG titer
Sera samples from the vaccinated mice were analyzed
for the titers of anti-A? antibodies by ELISA using synthe-
sized A?42 peptide. Fig. 3 showed that mice vaccinated
with AAV-CB-A?42 produced a strong serum IgG response
to A?42 (reciprocal end-point titers ranging from 4,096 to
16,384 in different mice), whereas vaccination with AAV-
Fig. 3. Time course studies of humoral immunity. AAV vectors (5 ? 1010particles) were administered via different routes of vaccination. (A) Therapeutic
groups of PDAPPV717Imice. (B) Prophylactic groups of PDAPPV717Imice. (C) Therapeutic groups of nontransgenic mice. (D) Prophylactic groups of
nontransgenic mice. Titers were determined based on the highest dilution of the samples that generated an OD greater than cutoff value (mean ? 3 SD of
sera from control groups). ODs lower than cutoff value at 1:64 dilution were considered negative responses. Abbreviations: i.n., intranasal administration;
i.m., intramuscular administration: oral., oral administration.
Fig. 4. Prevention of A?42-mediated toxic effects in primary cultures of
neurons by sera from vaccinated mice with AAV-CB-A?42. Cells were
incubated with fibrillar A?42 alone, or with fibrillar A?42 that had been
incubated with serum from the third bleeding at different concentrations.
Serum of untreated mice was used as negative control. The MTT assay was
used to estimate cell survival. Results (mean ? SE) are given as the
percentage of surviving cells compared with cell alone control cultures (*P
? 0.05, compared with A?42 group;#P ? 0.05, compared with control
serum). Abbreviation: Cont., control serum.
371
J. Zhang et al. / Neurobiology of Disease 14 (2003) 365–379

Page 8

A?42 generated a lower level of A?42-specific serum IgG
(reciprocal end-point titers ranging from 128 to 512 in
different mice). In contrast, anti-A?42 antibodies were not
detectable in the untreated mice or AAV-GFP group of mice
(reciprocal end-point titer ? 64 in all mice). The anti-A?42
antibodies were detectable in the AAV-CB-A?42 vacci-
nated mice 1 month after immunization, and reached the
highest titers 2 months after immunization. High titers were
sustained for at least 12 months. Two months after immu-
nization, mice vaccinated with AAV-CB-A?42 had a 10-
fold higher antibody titer compared with mice receiving
AAV-A?42 (P ? 0.01). We also found that the intramus-
cular route of vaccination gave a significantly higher titer of
A?-specific antibody than did the other two routes (oral or
intranasal vaccination). More strikingly, the therapeutic
group of PDAPPV717Imice that were vaccinated with AAV-
CB-A?42 generated lower antibody titers of A?-specific
antibody compared with those of the prophylactic group of
PDAPPV717Imice, especially 1 or 2 months after vaccina-
tion (P ? 0.05). In AAV-CB-A?42-vaccinated mice, there
was no significant difference in the titer of anti-A? antibody
between nontransgenic mice and PDAPPV717Imice, but in
AAV-A?42-vaccinated mice, the nontransgenic mice
showed higher titers of anti-A?42 antibody compared with
PDAPPV717Imice (P ? 0.05).
Partial neutralization of serum anti-human A? antibody
Sera from AAV-CB-A?42 vaccinated mice exhibited a
partial protective effect in preventing A?42-mediated neu-
rotoxicity toward primary cultures of neuron cells. Diluted
serum (1:10) from AAV-CB-A?42 vaccinated mice pre-
vented the neurotoxicity of A?42 (70% cell viability),
whereas an unrelated serum showed no effect (Fig. 4). The
percentage of surviving cell in the diluted serum (1:10)
group was significantly increased compared with that of the
A?42 group or the other two unrelated serum groups. But
the diluted serum (1:50) group did not show any significant
effect. These results indicated that AAV-CB-A?42 vacci-
nation could elicit “therapeutic” antibody titers within 1
month.
Evaluation of the ability of cognition and memory of
vaccinated mice using the Morris water maze
Hidden water maze
To study the effect of AAV-CB-A?42 in improving the
ability of cognition and memory of PDAPPV717Imice and
nontransgenic mice, the ability of mice to acquire, process,
and recall spatial information was assessed in the modified
Morris water maze test by using escape latency as an indi-
cator of learning. The performance of mice significantly
improved over the 6 training days in all groups (Fig. 5),
although the between-group differences were clear. In the
therapeutic groups, post hoc analysis revealed that the es-
cape latency, an estimate of spatial learning and memory
capacity, was significantly shorter for transgenic AAV-CB-
A?42-vaccinated mice as opposed to the untreated group (P
? 0.01), the AAV-GFP group (P ? 0.01), or the AAV-
A?42-vaccinated group (P ? 0.01) on days 1–6, respec-
tively, whereas AAV-A?42-vaccinated mice showed no
significant difference in the escape latency compared with
the AAV-GFP group (P ? 0.05) or the untreated group (P
? 0.05) (Fig. 5A). The asymptotic level of the AAV-CB-
A?42 group on days 4–6 differed significantly from those
of the other groups (P ? 0.01, ANOVA). The respective
asymptotic levels on days 4–6 were 38.07 ? 3.39 s for the
untreated group, 38.10 ? 3.08 s for the AAV-GFP group,
35.80 ? 3.21 s for the AAV-A?42 group, and 27.86 ?
3.20 s for the AAV-CB-A?42. No significant difference
was found between the average latency of the untreated
group and that of the AAV-GFP group (Fig. 5B). There was
also no significant difference between the latencies of the
Fig. 5. Learning curves showing the average latency (mean ? SE) to find
the platform of water maze during the hidden platform water maze tests
(days 1–6). The graphs represent data from therapeutic groups (A, all of
therapeutic groups; B, untreated group and GFP treated group; C, AAV-
A?42 vector with different routes of vaccination; D, AAV-CB-A?42
vector with different routes of vaccination) and prophylactic groups (E, all
of prophylactic groups; F, AAV-CB-A?42 vector with different routes of
vaccination) of PDAPPV717Imice.
372
J. Zhang et al. / Neurobiology of Disease 14 (2003) 365–379

Page 9

groups vaccinated in different routes, whether using AAV-
A?42 vaccine (Fig. 5C) or AAV-CB-A?42 vaccine (P ?
0.05, ANOVA) (Fig. 5D).
In the prophylactic groups, we found that the escape
latencies of mice over the 6 days of training were signifi-
cantly shorter than those of all the therapeutic groups, es-
pecially the untreated and AAV-GFP-treated groups. As
shown in Fig. 5E, the average escape latencies of the AAV-
CB-A?42 group over the 6 days of training were signifi-
cantly shorter than those of the control groups (AAV-GFP
or untreated group) (P ? 0.05, ANOVA). The respective
asymptotic levels on days 5–6 were 26.67 ? 2.10 s for the
untreated group, 26.61 ? 2.08 s for the AAV-GFP group,
and 21.75 ? 2.20 s for the AAV-CB-A?42 group. Fig. 5F
showed there was no significant difference in the escape
latencies of the AAV-CB-A?42 group over the 6 days of
training between the groups vaccinated in different routes
(P ? 0.05, ANOVA). All groups of nontransgenic mice
performed this task equally well, but there was no signifi-
cant difference between the groups (data not shown).
Probe test
Search of the hidden platform can be terminated by
chance contact with the target. Therefore, goal-directed
search pattern can be better distinguished in probe trials
when the animals are allowed to search the platform for 60 s
in an empty maze. Because later stages of the 1-min search
can be influenced by a failure to find the goal at the expected
location during the first half of the probe trial, the percent-
age of time spent in different quadrants of the pool during
the first 30 s was used to assess the effectiveness of the
animals’ search of the position of the goal.
In the therapeutic groups, as shown in Fig. 6A, AAV-
CB-A?42-vaccinated mice spent significantly more time in
the target quadrant (TQ) than that of the untreated group,
AAV-GFP group, or AAV-A?42 groups. Untreated mice
and AAV-GFP-treated mice showed significantly less pref-
erence for the target quadrant than did AAV-CB-A?42-
vaccinated mice, suggesting impaired memory retention. In
the prophylactic groups, AAV-CB-A?42-vaccinated mice
showed similar results as in the therapeutic groups (Fig.
6B).
A more refined parameter for the spatial bias in the place
navigation test is the number of crossings over the exact
former location of the platform (Janus et al., 2000a). We
analyzed the annulus crossing index during the probe tests.
The index expresses the spatial place preference and con-
trols for alternative search strategies without place prefer-
ences, such as circular search (Gass et al., 1998; Wehner et
al., 1990). The results showed that the AAV-CB-A?42-
vaccinated mice in the therapeutic group crossed the correct
Fig. 6. Percentage of time (mean ? SEM) spent in the target quadrant (TQ)
in probe trials on day 6 of the hidden platform water-maze training. The
graphs represent data from the therapeutic groups (A) and prophylactic
groups (B) of PDAPPV717Imice during the first 30 s of the probe tests. *P
? 0.05, **P ? 0.01, vs. the untreated group (Tukey-Kramer test).#P ?
0.05 vs. the AAV-GFP group (Student Neuman-Keuls test);@P ? 0.05 vs.
the AAV-A?42 groups.
Fig. 7. Annulus crossing index during the probe trial after the last training
trial on day 6. The graphs represent data from the therapeutic groups (A)
and prophylactic groups (B) of PDAPPV717Imice during the first 30 s of the
probe tests. A positive index indicates selective focal search of the previous
platform position; an index approaching zero reflects nonspatial or circular
search of the pool. *P ? 0.01, compared with the untreated group or
AAV-GFP group;#P ? 0.01, compared with the AAV-A?42 group.
373
J. Zhang et al. / Neurobiology of Disease 14 (2003) 365–379

Page 10

site significantly more often than did the AAV-GFP group
(P ? 0.01) or the untreated group (P ? 0.01) (Fig. 7A).
AAV-A?42-vaccinated mice showed no significant differ-
ence from the controls. In the prophylactic groups, AAV-
CB-A?42-vaccinated mice showed similar results as in
therapeutic groups (Fig. 7B).
Visible water maze
One week after the end of the hidden platform trial, a
“visible platform trial” was carried out. As shown in Fig. 8,
the results revealed significant differences in performance
between AAV-CB-A?42-vaccinated mice and AAV-GFP-
treated, untreated, or AAV-A?42-vaccinated mice in the
therapeutic group (Fig. 8A). Average latencies to reach the
cued platform on day 4 were 15.5 ? 3.0, 19.4 ? 3.5, 20.4
? 3.6, or 19.5 ? 4.1 s for AAV-CB-A?42-vaccinated,
AAV-GFP-treated, untreated, or AAV-A?42-treated mice,
respectively (Fig. 8A). But in the prophylactic groups, there
was no significant difference between the groups (Fig. 8E).
In both the therapeutic and prophylactic groups, there was
no significant difference between the latency of AAV-GFP-
treated and that of untreated mice (Fig. 8B and E), and no
significant difference was found in the average escape la-
tencies of the AAV-CB-A?42 group (Fig. 8C and F) or
AAV-A?42 group (Fig. 8D) over the 6 days of training
whether using one or the other of the three vaccination
routes (P ? 0.05, ANOVA). We also found that in the
prophylactic groups, the PDAPPV717Imice did not show
any significant difference from the nontransgenic littermates
in this test (data not shown).
In conclusion, these results from the Morris water maze
paradigm demonstrated unequivocally that the cognitive
deficits of AAV-CB-A?42-vaccinated mice were markedly
improved compared with those of the controls.
Fig. 8. Learning curves showing the average latency (mean ? SE) to find
the platform in the water maze during the hidden platform water maze tests
(days 1–6). The graphs represent data from the therapeutic groups (A, all
of therapeutic groups; B, untreated group and GFP treated group; C,
AAV-A?42 vector with different routes of vaccination; D, AAV-CB-A?42
vector with different routes of vaccination) and prophylactic groups (E, all
of prophylactic groups; F, AAV-CB-A?42 vector with different routes of
vaccination) of PDAPPV717Imice.
Table 1
Quantitative image analysis for amyloid plaques in the hippocampus and ELISAs for A?40 or A?42 in brains of all PDAPP transgenic micea
GroupsTherapeutic groupsProphylactic groups
n
Hippocampusb
A?42 (ng/g)A?40 (ng/g)
n
Hippocampusb
A?42 (ng/g)A?40 (ng/g)
AAV-CB-A?42
oral
i.n
i.m
AAV-A?42
oral
i.n
i.m
AAV-GFP
Untreated
272.17 ? 2.04c,e,g
2.41 ? 2.61
2.15 ? 1.63
1.95 ? 2.41
3.25 ? 3.23
3.29 ? 3.87
3.09 ? 3.41
3.38 ? 4.10
3.57 ? 3.69
3.63 ? 3.71
264.8 ? 26.4d,f,g
279.4 ? 31.4
261.5 ? 25.6
253.6 ? 32.7
400.9 ? 42.3
397.2 ? 45.8
391.7 ? 34.6
413.5 ? 56.1
425.9 ? 61.3
432.8 ? 51.3
573.3 ? 99.5d,f,g
603.5 ? 106.7
560.1 ? 96.4
556.4 ? 112.5
910.1 ? 139.5
916.4 ? 167.3
891.7 ? 126.1
925.3 ? 147.8
960.4 ? 154.9
954.6 ? 156.3
260.47 ? 0.34d,f
0.61 ? 0.54
0.27 ? 0.21
0.41 ? 0.30
30.98 ? 10.7c,e
33.2 ? 9.0
29.4 ? 12.4
30.2 ? 10.3
61.7 ? 12.9c,e
67.1 ? 12.6
57.8 ? 14.2
60.1 ? 13.2
9
9
9
9
9
8
26
8
9
9
8
8
9
9
2.39 ? 1.56
2.45 ? 1.02
55.3 ? 13.5
51.9 ? 10.2
127.8 ? 18.6
130.6 ? 17.9
aAbbreviations: i.n., intranasal; i.m., intramuscular.
bPercentage of A? immune reactivity (mean ? SD), representing A? plaque burden in the hippocampus.
cP ? 0.05,dP ? 0.01, versus untreated mice;eP ? 0.05,fP ? 0.01, versus AAV-GFP group;gP ? 0.05,hP ? 0.01, versus AAV-A?42-vaccinated
group.
374
J. Zhang et al. / Neurobiology of Disease 14 (2003) 365–379

Page 11

Amyloid plaque image analysis
After completion of the water maze tests, we carried out
quantitative immunohistochemical measures to determine
the extent of amyloid-? burden in the brain of mouse. As
shown in Table 1, quantitative immunohistochemical image
analysis of the A? burden in the hippocampus demonstrated
that in the therapeutic groups, although there was no sig-
nificant difference between the three routes of AAV-CB-
A?42 vaccination, the median value of the A? burden in the
hippocampus of the AAV-CB-A?42-vaccinated group were
significantly less than that of the untreated group (P ? 0.05,
Mann-Whitney U test) or the AAV-GFP-treated group (P ?
0.05, Mann-Whitney U test). In contrast, the median value
of A? deposition for the group vaccinated with AAV-A?42
was not significantly different from that of the untreated
group or AAV-GFP-treated group (P ? 0.05, Mann-Whit-
ney U test). From Fig. 9, we can see that brain tissues from
the untreated group or AAV-GFP-treated group contained
large amyloid plaques in both the retrosplenial cortex (Fig.
9a) and hippocampus (Fig. 9d) (since there was no signifi-
cant difference in the A? plaque burden between the un-
treated group and the AAV-GFP-treated group, hereafter,
we will pool them together as controls). We also observed
a similar pattern of A? deposition in mice vaccinated with
AAV-A?42 (Fig. 9b and e). But brain sections from mice
vaccinated with AAV-CB-A?42 showed very small isolated
amyloid plaques in the cortex or hippocampus (Fig. 9c and
f). In the prophylactic groups, vaccination with AAV-CB-
A?42 resulted in almost complete prevention of A? depo-
sition. Of 26 mice vaccinated with AAV-CB-A?42, 14 had
no detectable A? plaque in six brain sections of each mouse
examined. Nine mice had very small isolated A? plaques.
Only three mice had relatively large mature A? plaques. In
contrast, all of the untreated mice or AAV-GFP-treated
mice developed significant amounts of mature A? plaques
Fig. 9. Cortical and hippocampal A? deposition. Immunohistochemical
images from PDAPPV717Imice with A? burdens representative of the
median values of their respective groups. Since there was no significant
difference in the A? plaque burden between the untreated group and
AAV-GFP-treated group, here we pool them together in one panel as
controls, a, b, c, d, e, and f show A? deposition in PDAPPV717Imice in the
therapeutic groups. a, b, and c show cortical A? plaques in controls (a),
AAV-A?42-vaccinated mice (b), and AAV-CB-A?42-vaccinated mice (c).
d, e, and f show hippocampus A? plaques in controls (d), AAV-A?42-
vaccinated mice (e), and AAV-CB-A?42-vaccinated mice (f). g, h, i, and
j show A? deposition in PDAPPV717Imice of the prophylactic groups. g
and h show cortical A? plaques in controls (g) and AAV-CB-A?42-
vaccinated mice (h). i and j show hippocampus A? plaques in controls (i)
and AAV-CB-A?42-vaccinated mice (j).
Fig. 10. Cortical astrocytosis in PDAPPV717Imice. Images of plaque-
associated astrocytosis, as determined by GFP immunohistochemistry,
representing the median values of their respective groups. a, b, and c show
cortical astrocytosis from mice in the therapeutic groups. From a, b, and c,
we can see that astrocytosis in the cortex of AAV-CB-A?42-vaccinated
mice (c) were significantly less than that of controls (a) or that of AAV-
A?42-vaccinated mice (b). d and e show cortical astrocytosis in
PDAPPV717Imice in the prophylactic groups. We also found that in the
prophylactic groups, astrocytosis in the cortex of AAV-CB-A?42-vacci-
nated mice (e) was significantly less than that of controls (d).
375
J. Zhang et al. / Neurobiology of Disease 14 (2003) 365–379

Page 12

(Fig. 9g and i). Quantitative imaging of the A? burden in
the hippocampus demonstrated that the median value of the
A? burden in the hippocampus of the AAV-CB-A?42-
vaccinated group was significantly less than that of the
untreated group (P ? 0.01, Mann-Whitney U test) or the
AAV-GFP group (P ? 0.01, Mann-Whitney U test).
ELISA test for A? peptides in brains
To confirm the efficacy of AAV-CB-A?42 vaccine in
decreasing the levels of A? peptides in the brain of
PDAPPV717Imice, we performed sensitive ELISAs specific
for A?40 and A?42 on the entire contralateral (frozen)
hemisphere of brains from each group of mice. This method
measures the total A?40 or A?42 content of the brain tissue.
Both human A?40 and A?42 were detected in the brains of
PDAPPV717Imice. No signals above the background were
detected in nontransgenic animals. As shown in Table 1, in
the therapeutic groups, the mean level of A?42 or A?40 in
AAV-CB-A?42-vaccinated mice showed statistically sig-
nificant difference from that of AAV-A?42 (P ? 0.05,
ANOVA), AAV-GFP (P ? 0.01, ANOVA), or untreated
group (P ? 0.01, ANOVA). However, no significant dif-
ference was found in the mean level of A?42 or A?40 of the
AAV-A?42 group and those of the AAV-GFP (P ? 0.05,
ANOVA) and the untreated groups (P ? 0.05, ANOVA).
There was considerable spread in the A?40 and A?42 levels
in the untreated group or AAV-GFP group, with levels of
A?40 ranging from 575.1 to 1234 ng/g of wet brain tissue
and those of A?42 ranging from 215 to 728 ng/g of wet
brain tissue. When the mean A?40 and A?42 levels of
AAV-CB-A?42-vaccinated mice were compared with those
of the untreated mice, significant decreases of 39.9% A?40
(P ? 0.01, ANOVA) and 38.8% A?42 (P ? 0.01, ANOVA)
were observed. Although the mean A?40 and A?42 levels
of AAV-A?42-immunized mice were also decreased, no
significant difference was found when compared with those
of untreated mice (P ? 0.05, ANOVA) or the AAV-GFP-
treated group (P ? 0.05, ANOVA). We also found that
there was no significant difference in the mean level of
A?40 or A?42 when the AAV-CB-A?42 vaccine was given
via different routes. In the prophylactic groups, similar
results were found when compared to those of the therapeu-
tic groups. The mean A?40 and A?42 levels of AAV-CB-
A?42-vaccinated mice were significantly decreased com-
pared with those of the untreated mice (P ? 0.01, ANOVA)
or the AAV-GFP group (P ? 0.01, ANOVA). But the mean
levels of A?40 or A?42 of each prophylactic group were
remarkably lower than those of each therapeutic group.
Astrocytic activation in brains of PDAPPV717Imice and
nontransgenic mice
Astrocytosis, another hallmark of plaque-associated pa-
thology in patients with Alzheimer’s disease as well as
PDAPPV717Imice, was dramatically reduced in the brains
from AAV-CB-A?42-vaccinated mice in both the therapeu-
tic (Fig. 10c) and prophylactic groups (Fig. 10e). Brains
from mice of controls (Fig. 10a and d; since there was no
significant difference in cortical astrocytosis between the
untreated group and the AAV-GFP-treated group, here after
we will pool them together in one panel as controls) and
AAV-A?42-vaccinated mice (Fig. 10b) contained numer-
ous clusters of astrocytosis that were immunoreactive to
GFAP, a finding typical of A? plaque-associated astrocy-
tosis.
Discussion
Much evidence indicates that abnormal processing and
extracellular deposition of amyloid-? peptide is central to
the pathogenesis of Alzheimer’s disease (Steiner et al.
1999). A strong support for the unequivocal relationship
between the plaques and behavioral deficits came from the
elegant studies by Chen et al. (2000), Janus et al. (2000b),
and Morgan et al. (2000), who demonstrated that A? pep-
tide vaccination or passive administration of anti-A? anti-
body reduced behavioral impairment and amyloid burden in
the brain of animal models of AD. But recently, although a
phase I safety study in few individuals did not detect sig-
nificant side effects, a phase II safety study showed that
roughly 5% of the treated participants developed what
seemed to be an inflammatory reaction in the CNS (an
aseptic meningoencephalitis). The occurrence of the menin-
gocerebral inflammation was not correlated with either the
presence or titers of antibodies against A? among the trial
participants (Weiner and Selkoe, 2002). The mechanism of
this self-limited inflammatory reaction is unknown, but the
appearance of the inflammation before the detection of A?
antibodies in some of the recipients may suggest that a T
cell-mediated immune reaction to A?, presumably related
to the induction of TH1-type responses, was responsible
(Weiner and Selkoe, 2002).
In present study, some important strategies that induce
nonpathogenic T-cell responses had been utilized to develop
an AAV-CB-A?42 vaccine that induces long-term, high
levels of anti-A? antibodies. For example, modified autoan-
tigens (CB-A?), tolerogenic routes such as oral or nasal
administration, and a non-TH1-inducing adjuvant (CB) were
used. We assessed the efficacy of the AAV-CB-A?42 vac-
cine on PDAPPV717Itransgenic mice. A single administra-
tion of AAV-CB-A?42 vaccine induced high levels of pro-
longed A?-specific serum IgG antibody either by the i.n.,
i.m., or oral route. No detectable anti-A?42 antibody was
found in the group of AAV-GFP-treated or untreated group,
and a very much lower level of A?-specific serum IgG
antibody could be detected in mice vaccinated with AAV-
A?42 vaccine. Moreover, the behavioral abilities of AAV-
CB-A?42 vaccinated mice were significantly improved, and
the amyloid plaques and astrocytosis in the brains were also
significantly reduced compared with those of untreated or
376
J. Zhang et al. / Neurobiology of Disease 14 (2003) 365–379

Page 13

AAV-GFP-treated mice. In contrast, the administration of
AAV-A?42 vaccine did not show these improvements al-
though it could induce lower levels of A?-specific serum
IgG. We also found that although the titer of A?-specific
serum IgG antibody of mice vaccinated with AAV-CB-
A?42 by injection is higher than that by oral or intranasal
administration, there was no significant difference in A?
deposition or plaques between different routes of adminis-
tration. The behavioral ability between the three routes of
administration was also not significantly different. These
data indicate that high titer of anti-A? antibody was neces-
sary to improve the behavioral abilities and to reduce the
amyloid plaques or astrocytosis in the brain of mice, al-
though this effect plateaued off after the titers reached a
certain level.
It is important to note that the humoral immunological
responses were observed specifically in the AAV-CB-
A?42-vaccinated mice that showed significant reductions in
cerebral A? as measured by quantitative imaging and A?
ELISA. In other words, we found that a significant decrease
in the mean A? burden only in the AAV-CB-A?42-vacci-
nated mice that developed high titers of anti-human A?
antibody. No difference was found in the mean A? burden
between the AAV-A?42-vaccinated mice, which developed
only low titers of anti-human A? antibody, and the un-
treated mice or AAV-GFP-treated mice, which showed no
anti-human A? antibody responses. Importantly, the lack of
immune response to AAV-GFP was associated with no
decrease in the mean cerebral A? burden as determined by
image analyses and ELISA. Therefore, several notes should
be discussed, as below.
First, the higher titers of A?-specific serum IgG antibody
of mice vaccinated with AAV-CB-A?42 compared with
that with AAV-A?42 showed that CB played an important
role in stimulating the immune response to A?42. This
proved that the immunogenicity of A?42 without any ad-
juvant is very low. This may be especially important when
one hopes to use in human beings, because A?42 is a
protein endogenous to human beings. So, the first feature of
AAV-CB-A?42 vaccine is that the A?42 DNA fragment
was linked to the CB gene whose expression product (CB)
greatly increased the immunogenicity of A?42 in vivo.
Previous studies reported that CB possessed strong adjuvant
effects when coadministered with antigens. It is usually
fused to antigens in developing vaccines for inducing hu-
moral immunity (Ploix et al., 1999). This effect may be due
to its activating effect on B cells, including the upregulation
of MHC class II molecules (Nashar et al., 1997), or to the
induction of regulatory Th2-type CD4?T cells (Ploix et al.,
1999). Additionally, conjugation with CB may greatly fa-
cilitate antigen delivery and presentation to the gut-associ-
ated lymphoid tissues (GALT) due to its affinity for the cell
surface GM-ganglioside receptors, including the membra-
nous cells as well as enterocytes (McGhee et al., 1992). Our
findings raised the possibility that bacterial toxin-based ge-
netic adjuvant may be a safe and effective means to enhance
the potency of DNA vaccines for both prophylactic and
therapeutic purposes.
The second feature is that we used the AAV-mediated
delivery system. This makes it possible that the CB-A?42
fusion protein could be continuously expressed and the titer
of anti-A?42 antibody could be maintained for a long pe-
riod of time. AAV seems to be a promising vector in clinical
gene therapy. Xin et al. (2001, 2002) reported that oral
administration of a single dose of an AAV vector expressing
human immunodeficiency virus type 1 env gene (AAV-
HIV) induced both systemic and mucosal immunity against
HIV. During et al. (2000) reported that an AAV vaccine
generated autoantibodies that targeted a specific brain pro-
tein, the NR1 subunit of the N-methyl-D-aspartate (NMDA)
receptor. A single oral dose of the AAV vaccine produced
strong antiepileptic and neuroprotective activity 1 to 5
months following vaccination in rat models for both kain-
ate-induced seizure and middle cerebral artery occlusion
stroke. Thus, a vaccination strategy targeting brain proteins
is feasible and may have therapeutic potential for neurolog-
ical disorders. Furthermore, the induction of immune re-
sponses by mucosal administration of an AAV vaccine is
relatively safe and can be administered for long periods of
time, making it most desirable for chronic diseases such as
AD.
In this study, we took advantage of the merits of AAV as
described above and the adjuvant activity of CB. Our results
showed that although oral or intranasal administration of
AAV-CB-A?42 vaccine could also induce strong humoral
immune response, the titers of anti-A? antibody induced
were lower than that of intramuscular administration. This
may be due to the fact that the AAV-CB-A?42 vector, when
given intramuscularly, can gain ready access to the blood
circulation, thus facilitating the products of expression (CB-
A?42) contact with antigen-presenting cells and lympho-
cytes. Another reason may be in that the expression of
CB-A?42 varies in difference tissues.
In the present study, we also found that the anti-A?
antibody response of the therapeutic groups of PDAPPV717I
mice induced by AAV-CB-A?42 vaccine was weaker than
that of the prophylactic groups, but no difference was found
in the nontransgenic mice of both groups. This suggests that
the older PDAPPV717Itransgenic mice with higher level of
amyloid deposition in the brains had partial impairment or
disturbance of the immune system.
Although the mechanism of A? immunotherapy remains
unclear, it has opened up a new area of research to gain
insight into why such an approach can lead to the elimina-
tion of amyloid deposits in the brains of transgenic mice that
developed AD amyloidosis. Bard reported that the antibod-
ies could cross the blood-brain barrier (BBB), enter the
central nervous system (CNS), bind to amyloid plaques,
activate microglial cells mediated by Fc receptor, and in-
duce the clearance of preexisting amyloid (Bard et al.,
2000). However, DeMattos et al. (2002) demonstrated that
the presence of antibody (m266) in the peripheral circula-
377
J. Zhang et al. / Neurobiology of Disease 14 (2003) 365–379

Page 14

tion directly facilitated net A? efflux from the brain, acting
as a “peripheral sink.” They also showed that long-term
peripheral administration of antibody (m266) to PDAPP
mice markedly reduced the A? burden without the antibody
actually crossing the blood-brain barrier and binding to A?
deposits in the brain. They suggested that the likely mech-
anism is that by altering the dynamic equilibrium of A?
between brain, CSF, and plasma, a reduction of plasma A?
can lead to an efflux of brain A? to the cerebral spinal fluid
and into the circulation. Their hypothesis was supported by
some recent studies showing that exogenous A?40 can be
transported rapidly from CSF to plasma (Ghersi-Egea et al.,
1996; Shibata et al., 2000; Zlokovic et al., 1996).
Recently, Bacskai et al. (2002) demonstrated that F(ab’)2
fragments of 3d6 (which lack the Fc region of the antibody)
also led to clearance of the deposits, similar to the results
obtained with full-length 3d6 antibody. This suggests that
the clearance of amyloid deposits in vivo may involve, in
addition to Fc-dependent clearance, a non-Fc-mediated dis-
ruption of plaque structure (Bacskai et al., 2002). Additional
work is required to resolve how A? immunotherapy works.
In summary, we demonstrated that a single administra-
tion with AAV-CB-A?42 to PDAPPV717Itransgenic mice
could induce high levels of anti-A? antibody, and result
subsequently in improved ability of memory and cognition,
decreased A? accumulation and deposition in the brain, and
a resultant decrease in plaque-associated astrocytosis. These
effects were specifically associated with an anti-A? anti-
body response. Our results extended the immunological
approach for the treatment and prevention of AD to a single
oral, intranasal, or intramuscular administration of our new
vaccine that might be better tolerated by human patients
than routine repetitive immunizations in the presence of
adjuvant. Our results also raised the possibility that AAV-
CB-A?42 vector immunization may provide the basis for a
novel and promising Alzheimer’s disease vaccination strat-
egy.
Acknowledgments
This work was supported by the grants (CMB 99-699)
from the China Medical Board of New York, USA and
(2001CB510009) from the National Program for Key Basic
Research Projects, Ministry of Science and Technology,
China, P.R. We thank Dr. Biao Kan for plasmid
pGEM7zf(?) and Dr. Xiaoyan Hu for cDNA library of
human brain. We also acknowledge Dr. Huiyuan Luo for
verbal correction.
References
Bacskai, B.J., Kajdasz, S.T., McLellan, M.E., Games, D., Seubert, P.,
Schenk, D., Hyman, B.T., 2002. Non-Fc-mediated mechanisms are
involved in clearance of amyloid-beta in vivo by immunotherapy.
J. Neurosci. 22, 7873–7878.
Bard, F., Cannon, C., Barbour, R., Burke, R.L., Games, D., Grajeda, H.,
Guido, T., Hu, K., Huang, J., Johnson-Wood, K., Khan, K., Kholo-
denko, D., Lee, M., Lieberburg, I., Motter, R., Nguyen, M., Soriano, F.,
Vasque ´z, N., Weiss, K., Welch, B., Seubert, P., Schenk, D., Yednock,
T., 2000. Peripherally administered antibodies against amyloid beta-
peptide enter the central nervous system and reduce pathology in a
mouse model of Alzheimer disease. Nat. Med. 6, 916–919.
Cavallaro, S., Meiri, N., Yi, C.L., Musco, S., Ma, W., Goldberg, J., Alkon,
D.L., 1997. Late memory-related genes in the hippocampus revealed by
RNA fingerprinting. Proc. Natl. Acad. Sci. USA 94, 9669–9673.
Chen, G., Chen, K.S., Knox, J., Inglis, J., Bernard, A., Martin, S.J., Justice,
A., McConlogue, L., Games, D., Freedman, S.B., Morris, R.G., 2000.
A learning deficit related to age and beta-amyloid plaques in a mouse
model of Alzheimer’s disease. Nature 408, 975–979.
Dawson, V.L., Dawson, T.M., London, E.D., Bredt, D.S., Snyder, S.H.,
1991. Nitric oxide mediates glutamate neurotoxicity in primary cortical
cultures. Proc. Natl. Acad. Sci. USA 88, 6368–6371.
DeMattos, R.B., Bales, K.R., Cummins, D.J., Dodart, J.C., Paul, S.M.,
Holtzman, D.M., 2001. Peripheral anti-A beta antibody alters CNS and
plasma A beta clearance and decreases brain A beta burden in a mouse
model of Alzheimer’s disease. Proc. Natl. Acad. Sci. USA 98, 8850–
8855.
DeMattos, R.B., Bales, K.R., Cummins, D.J., Paul, S.M., Holtzman, D.M.,
2002. Brain to plasma amyloid-betefflux: a measure of brain amyloid
burden in a mouse model of Alzheimer’s disease. Science 295, 2264–
2267.
Dewachter, I., van Dorpe, J., Spittaels, K., Tesseur, I., Van Den, H.C.,
Moechars, D., Van Leuven, F., 2000. Modeling Alzheimer’s disease in
transgenic mice: effect of age and of presenilin 1 on amyloid biochem-
istry and pathology in APP/London mice. Exp. Gerontol. 35, 831–841.
During, M.J., Symes, C.W., Lawlor, P.A., Lin, J., Dunning, J., Fitzsimons,
H.L., Poulsen, D., Leone, P., Xu, R., Dicker, B.L., Lipski, J., Young,
D., 2000. An oral vaccine against NMDAR with efficacy in experi-
mental stroke and epilepsy. Science 287, 1453–1460.
During, M.J., Xu, R., Young, D., Kaplitt, M.G., Sherwin, R.S., Leone, P.,
1998. Peroral gene therapy of lactose intolerance using an adeno-
associated virus vector. Nat. Med. 4, 1131–1135.
Frenkel, D., Katz, O., Solomon, B., 2000. Immunization against Alzhei-
mer’s beta-amyloid plaques via EFRH phage administration. Proc.
Natl. Acad. Sci. USA 97, 11455–11459.
Gass, P., Wolfer, D.P., Balschun, D., Rudolph, D., Frey, U., Lipp, H.P.,
Schutz, G., 1998. Deficits in memory tasks of mice with CREB muta-
tions depend on gene dosage. Learn. Mem. 5, 274–288.
Ghersi-Egea, J.F., Gorevic, P.D., Ghiso, J., Frangione, B., Patlak, C.S.,
Fenstermacher, J.D., 1996. Fate of cerebrospinal fluid-borne amyloid
beta-peptide: rapid clearance into blood and appreciable accumulation
by cerebral arteries. J. Neurochem. 67, 880–883.
Hardy, J., Selkoe, D.J., 2002. The amyloid hypothesis of Alzheimer’s
disease: progress and problems on the road to therapeutics. Science
297, 353–356.
Janus, C., D’Amelio, S., Amitay, O., Chishti, M.A., Strome, R., Fraser, P.,
Carlson, G.A., Roder, J.C., George-Hyslop, P., Westaway, D., 2000a.
Spatial learning in transgenic mice expressing human presenilin (PS1)
transgenes. Neurobiol. Aging 21, 541–549.
Janus, C., Pearson, J., McLaurin, J., Mathews, P.M., Jiang, Y., Schmidt,
S.D., Chishti, M.A., Horne, P., Heslin, D., French, J., Mount, H.T.,
Nixon, R.A., Mercken, M., Bergeron, C., Fraser, P.E., George-Hyslop,
P., Westaway, D., 2000b. A beta peptide immunization reduces behav-
ioural impairment and plaques in a model of Alzheimer’s disease.
Nature 408, 979–982.
Jindal, R.M., Karanam, M., Shah, R., 2001. Prevention of diabetes in the
NOD mouse by intra-muscular injection of recombinant adeno-associ-
ated virus containing the preproinsulin II gene. Int. J. Exp. Diabetes
Res. 2, 129–138.
Johnson-Wood, K., Lee, M., Motter, R., Hu, K., Gordon, G., Barbour, R.,
Khan, K., Gordon, M., Tan, H., Games, D., Lieberburg, I., Schenk, D.,
378
J. Zhang et al. / Neurobiology of Disease 14 (2003) 365–379

Page 15

Seubert, P., McConlogue, L., 1997. Amyloid precursor protein pro-
cessing and A beta42 deposition in a transgenic mouse model of
Alzheimer disease. Proc. Natl. Acad. Sci. USA 94, 1550–1555.
Koeberl, D.D., Bonham, L., Halbert, C.L., Allen, J.M., Birkebak, T.,
Miller, A.D., 1999. Persistent, therapeutically relevant levels of human
granulocyte colony-stimulating factor in mice after systemic delivery
of adeno-associated virus vectors. Hum. Gene Ther. 10, 2133–2140.
McGhee, J.R., Mestecky, J., Dertzbaugh, M.T., Eldridge, J.H., Hirasawa,
M., Kiyono, H., 1992. The mucosal immune system: from fundamental
concepts to vaccine development. Vaccine 10, 75–88.
McKenzie, S.J., Halsey, J.F., 1984. Cholera toxin B subunit as a carrier
protein to stimulate a mucosal immune response. J. Immunol. 133,
1818–1824.
Morgan, D., Diamond, D.M., Gottschall, P.E., Ugen, K.E., Dickey, C.,
Hardy, J., Duff, K., Jantzen, P., DiCarlo, G., Wilcock, D., Connor, K.,
Hatcher, J., Hope, C., Gordon, M., Arendash, G.W., 2000. A beta
peptide vaccination prevents memory loss in an animal model of
Alzheimer’s disease. Nature 408, 982–985.
Morris, R., 1984. Developments of a water-maze procedure for studying
spatial learning in the rat. J. Neurosci. Methods 11, 47–60.
Nashar, T.O., Hirst, T.R., Williams, N.A., 1997. Modulation of B-cell
activation by the B subunit of Escherichia coli enterotoxin: receptor
interaction up-regulates MHC class II. B7, CD40, CD25 and ICAM-1.
Immunology 91, 572–578.
Ploix, C., Bergerot, I., Durand, A., Czerkinsky, C., Holmgren, J., Thivolet,
C., 1999. Oral administration of cholera toxin B-insulin conjugates
protects NOD mice from autoimmune diabetes by inducing CD4?
regulatory T-cells. Diabetes 48, 2150–2156.
Prasad, K.M., Yang, Z., Bleich, D., Nadler, J.L., 2000. Adeno-associated
virus vector mediated gene transfer to pancreatic beta cells. Gene Ther.
7, 1553–1561.
Samulski, R.J., Chang, L.S., Shenk, T., 1989. Helper-free stocks of recom-
binant adeno-associated viruses: normal integration does not require
viral gene expression. J. Virol. 63, 3822–3828.
Schenk, D., Barbour, R., Dunn, W., Gordon, G., Grajeda, H., Guido, T.,
Hu, K., Huang, J., Johnson-Wood, K., Khan, K., Kholodenko, D., Lee,
M., Liao, Z., Lieberburg, I., Motter, R., Mutter, L., Soriano, F., Shopp,
G., Vasquez, N., Vandevert, C., Walker, S., Wogulis, M., Yednock, T.,
Games, D., Seubert, P., 1999. Immunization with amyloid-beta atten-
uates Alzheimer-disease-like pathology in the PDAPP mouse. Nature
400, 173–177.
Shibata, M., Yamada, S., Kumar, S.R., Calero, M., Bading, J., Frangione,
B., Holtzman, D.M., Miller, C.A., Strickland, D.K., Ghiso, J., Zlok-
ovic, B.V., 2000. 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. 106, 1489–1499.
Steiner, H., Capell, A., Leimer, U., Haass, C., 1999. Genes and mecha-
nisms involved in beta-amyloid generation and Alzheimer’s disease.
Eur. Arch. Psychiatry Clin. Neurosci. 249, 266–270.
Sturchler-Pierrat, C., Abramowski, D., Duke, M., Wiederhold, K.H., Mistl,
C., Rothacher, S., Ledermann, B., Burki, K., Frey, P., Paganetti, P.A.,
Waridel, C., Calhoun, M.E., Jucker, M., Probst, A., Staufenbiel, M.,
Sommer, B., 1997. Two amyloid precursor protein transgenic mouse
models with Alzheimer disease-like pathology. Proc. Natl. Acad. Sci.
USA 94, 13287–13292.
Sugahara, T., Pixley, M.R., Minami, S., Perlas, E., Ben Menahem, D.,
Hsueh, A.J., Boime, I., 1995. Biosynthesis of a biologically active
single peptide chain containing the human common alpha and chori-
onic gonadotropin beta subunits in tandem. Proc. Natl. Acad. Sci. USA
92, 2041–2045.
Wehner, J.M., Sleight, S., Upchurch, M., 1990. Hippocampal protein
kinase C activity is reduced in poor spatial learners. Brain Res. 523,
181–187.
Weiner, H.L., Selkoe, D.J., 2002. Inflammation and therapeutic vaccination
in CNS diseases. Nature 420, 879–884.
Williams, N.A., Hirst, T.R., Nashar, T.O., 1999. Immune modulation by
the cholera-like enterotoxins: from adjuvant to therapeutic. Immunol.
Today 20, 95–101.
Woogen, S.D., Turo, K., Dieleman, L.A., Beagley, K.W., Elson, C.O.,
1993. Inhibition of murine T cell activation by cholera toxin B subunit
is not mediated through the phosphatidylinositol second messenger
system. J. Immunol. 150, 3274–3283.
Wu, H.Y., Nikolova, E.B., Beagley, K.W., Russell, M.W., 1996. Induction
of antibody-secreting cells and T-helper and memory cells in murine
nasal lymphoid tissue. Immunology 88, 493–500.
Xin, K.Q., Ooki, T., Mizukami, H., Hamajima, K., Okudela, K., Hashi-
moto, K., Kojima, Y., Jounai, N., Kumamoto, Y., Sasaki, S., Klinman,
D., Ozawa, K., Okuda, K., 2002. Oral administration of recombinant
adeno-associated virus elicits human immunodeficiency virus-specific
immune responses. Hum. Gene Ther. 13, 1571–1581.
Xin, K.Q., Urabe, M., Yang, J., Nomiyama, K., Mizukami, H., Hamajima,
K., Nomiyama, H., Saito, T., Imai, M., Monahan, J., Okuda, K.,
Ozawa, K., Okuda, K., 2001. A novel recombinant adeno-associated
virus vaccine induces a long-term humoral immune response to human
immunodeficiency virus. Hum. Gene Ther. 12, 1047–1061.
Yang, Y.W., Kotin, R.M., 2000. Glucose-responsive gene delivery in
pancreatic Islet cells via recombinant adeno-associated viral vectors.
Pharmacol. Res. 17, 1056–1061.
Zlokovic, B.V., Martel, C.L., Matsubara, E., McComb, J.G., Zheng, G.,
McCluskey, R.T., Frangione, B., Ghiso, J., 1996. Glycoprotein 330/
megalin: probable role in receptor-mediated transport of apolipoprotein
J alone and in a complex with Alzheimer disease amyloid beta at the
blood-brain and blood-cerebrospinal fluid barriers. Proc. Natl. Acad.
Sci. USA 93, 4229–4234.
379
J. Zhang et al. / Neurobiology of Disease 14 (2003) 365–379