Neurobiology of Disease
Blood-Borne Amyloid-
Dimer Correlates with Clinical
Markers of Alzheimer’s Disease
Victor L. Villemagne,1,2,3 Keyla A. Perez,1,4,5 Kerryn E. Pike,1,2 W.Mei Kok,4,5,6 Christopher C. Rowe,2,3
Anthony R. White,1,5 Pierrick Bourgeat,8 Olivier Salvado,8 Justin Bedo,7 Craig A. Hutton,4,6 Noel G. Faux,1,8
Colin L. Masters,1 and Kevin J. Barnham1,4,5
1Mental Health Research Institute, The University of Melbourne, Parkville, Melbourne, Victoria 3052, Australia, 2Department of Nuclear Medicine and
Centre for PET, and 3Department of Medicine, Austin Health, Heidelberg, Victoria 3084, Australia, 4Bio21 Molecular Science and Biotechnology Institute,
Departments of 5Pathology and 6Chemistry, and 7Victorian Research Laboratory, National ICT Australia, The University of Melbourne, Parkville,
Melbourne Victoria 3010, Australia, and 8CSIRO Preventative Health National Research Flagship ICTC, The Australian e-Health Research Centre,
BioMedIA, Herston, Queensland 4029, Australia
Alzheimer’s disease (AD) is the most common age-related dementia. Unfortunately due to a lack of validated biomarkers definitive
diagnosis relieson thehistological demonstrationof amyloid-
(A
) plaquesand tauneurofibrillary tangles.A
processing is implicated
in AD progression and many therapeutic strategies target various aspects of this biology. While A
deposition is the most prominent
feature ofAD, oligomeric formsofA
havebeen implicated as the toxic species inducing theneuronal dysfunction. Currently there areno
methods allowing routine monitoring of levels of such species in living populations. We have used surface enhanced laser desorption
ionization time of flight (SELDI-TOF)mass spectrometry incorporating antibody capture to investigate whether the cellularmembrane-
containing fraction of blood provides a new source of biomarkers. There are significant differences in the mass spectra profiles of AD
compared with HC subjects, with significantly higher levels of A
monomer and dimer in the blood of AD subjects. Furthermore, levels
of these species correlatedwith clinicalmarkers ofAD includingbrainA
burden, cognitive impairment andbrain atrophy. These results
indicate that fundamental biochemical events relevant to AD can be monitored in blood, and that the species detected may be useful
clinical biomarkers for AD.
Introduction
Alzheimer’s disease (AD) is the most common age-related de-
mentia. However, due to a lack of validated biomarkers definitive
diagnosis still relies on postmortem histological demonstration
of amyloid-
(A
) plaques and tau neurofibrillary tangles (Masters
et al., 2006). Genetic studies have implicated A
processing in the
progression of AD, and while A
deposition is the most promi-
nent feature of AD, A
plaque load does not necessarily correlate
with disease severity (McLean et al., 1999). A
peptides are gen-
erated from the cleavage of the amyloid precursor protein (APP),
by the proteases
- and -secretases (Masters et al., 2006).
The clinically useful methods for monitoring A
species in
vivo are either in the CSF via ELISA methods using antibodies
raised against monomeric A
(Sjo¨gren et al., 2003) or via
positron emission tomography (PET) imaging using the radioac-
tive 11C-labeled Pittsburgh compound B (PIB) to measure the
deposited A
burden in the brain (Klunk et al., 2004; Rowe et al.,
2007). Neither the monomeric nor fibrillar forms of A
are
thought to be responsible for the toxicity associated with A

(Cappai and Barnham, 2008). Recent studies have implicated
oligomeric forms of A
as the toxic species that induce the neu-
ronal dysfunction associated with AD (Walsh et al., 2002; Walsh
and Selkoe, 2004; Shankar et al., 2008). These species include
synaptotoxic A
dimers isolated from postmortem AD brain tis-
sue (Shankar et al., 2008). To date, there is no method for mon-
itoring such species in living subjects, although an ELISA assay
for detecting oligomeric A
has recently been described (Xia et
al., 2009).
APP is a ubiquitously expressed type 1 transmembrane pro-
tein found in all tissues including blood. Traditionally, when
blood is collected and fractionated for diagnostic purposes,
plasma and/or serum is analyzed while the pellet containing cel-
lular elements (CEs); i.e., erythrocytes, leukocytes and platelets; is
discarded. Although a recent longitudinal study found that
higher plasma A
42 levels at the onset of the study were associ-
ated with a threefold increased risk of AD (Schupf et al., 2008),
attempts to use plasma A
levels as a biomarker for AD have, at
best, generated variable results (Zetterberg and Blennow, 2006).
Not only is APP a transmembrane protein but so are the pro-
teases that generate A
and a range of studies have shown that
many of the deleterious biological effects attributable to A
are
due to the interaction of oligomeric neurotoxic A
peptides with
cell membranes (Glabe and Kayed, 2006; Hung et al., 2008).
Therefore, given thatmost APP/A
pathology seems to be driven
by membrane interactions—and A
oligomers have a higher af-
Received Oct. 18, 2009; revised March 14, 2010; accepted March 19, 2010.
This work was funded by National Health and Medical Research Council of Australia (NHMRC).
Correspondence should be addressed to Kevin J. Barnham, Level 4, Bio21 Institute, 30 Flemington Road,
Parkville, VIC 3010, Australia. E-mail: kbarnham@unimelb.edu.au.
DOI:10.1523/JNEUROSCI.5180-09.2010
Copyright © 2010 the authors 0270-6474/10/306315-08$15.00/0
The Journal of Neuroscience, May 5, 2010 • 30(18):6315–6322 • 6315

finity for lipid membranes (Hung et al., 2008)—we investigated
whether the usually discarded membrane-rich CE fraction may
contain oligomeric A
species that correlate with markers of AD
severity.
Materials andMethods
Participants. The blood of 118 participants was analyzed for the study.
Fifty-two elderly individuals with well documented normal cognitive
function, 43 patients with mild to moderate AD, and 23 subjects with
MCI were recruited for the study (Table 1). Clinical classification was
based on the clinical interview,Mini-Mental State Examination (MMSE)
and clinical dementia rating (CDR). All AD patients met National Insti-
tute of Neurological and Communicative Dis-
orders and Stroke-Alzheimer’s Disease and
Related Disorders Association criteria for
probableAD (McKhann et al., 1984).MCI sub-
jects met the Petersen criteria of subjective and
objective cognitive difficulties, predominantly
affecting memory, in the absence of dementia
or significant functional loss (Petersen et al.,
1999). All patients were recruited from the
Austin Health Memory Disorders and Neu-
robehavioural Clinics or the Healthy Aging
Study (Mental Health Research Institute).
None of the participants were receiving, nor
ever received, anti-A
medication. Written in-
formed consent was obtained before participa-
tion. Approval for the study was obtained from
the Austin Health Human Research Ethics
Committee.
Blood sample preparation. Whole blood was
collected by phlebotomy in EDTA vacutainers (6ml of K3EDTA, Greiner
Bio-One) and processed within 20 min of procurement. Vacutainers
were spun at 3500 rpm (1.9 g) at 4°C for 30 min. The upper layer of
plasma was then removed and small aliquots of 100 l were made and
store at
80°C for possible future use. Thematerial remaining in the tube
was then homogenized using a vortex and after this 100 l aliquots were
made and stored at
80°C for future use.
Each aliquot was used just once by combining 10 l of this material
(CEs)with 10l of urea 8M, and 120l of 0.5%TritonX-100 in PBS, and
placed for 15 min in an ultra-sound bath with ice (each sample was done
by duplicate) 130 l of this mixture were added per spot to be analyzed.
Surface enhanced laser desorption ionization-time of flight mass spec-
trometry. PS10 ProteinChip arrays (Bio-Rad) were used for the SELDI-
TOF (surface enhanced laser desorption ionization time of flight) mass
spectrometry (MS). Specific antibodies (2l of either 4G8 orWO2)were
added to the arrays in PBS (0.25 mg/ml). To confirm that the binding
observed was not due to nonspecific binding control spectra using a
nonspecific IgG antibody were also obtained. Chips were then incubated
overnight at 4°C in a humidity chamber.
Excess antibodies were then removed and blocking buffer (0.5 M eth-
anolamine in PBS) was added (5 l) and arrays were incubated for 30
min. After the removal of the blocking buffer, each array was washed for
5 min with 50 l of 0.5% Triton X-100/PBS (wash-buffer). The solvent
was then removed, and the arrays were washed for 5 min with 50 l of
PBS. All biological samples were analyzed in triplicate. Individual sam-
ples (130 l) were added to each spot and the arrays were incubated at
room temperature for 3 h. The excess samplewas then removed, and each
spot was washed twice with 100 l of wash-buffer for 10 s, followed by a
wash with 100 l of PBS twice for 10 s as well. Finally, the arrays were
washed twice with 100 l of HEPES 1 mM for 10 s. The array was then
air-dried. One microliter of sinapinic acid (SPA, 50% saturated in 50%
(v/v) acetonitrile and 0.5% in TFA) was applied to each spot twice. The
array was air-dried between each application. All incubations andwashes
were performed on a shaking table.
Chips were analyzed in a PBSIIC, SELDI-TOFmass spectrometer, and
spectra analyzed using Ciphergen ProteinChip software 3.1. The distri-
butions of the mass to charge ratio (m/z) peak intensities in the spectra
showed skewness to either left or right. By taking the logarithm of the
peak intensities, skewness was substantially reduced, and the distribu-
tions met criteria for normality. To demonstrate that the detected peak
intensities are dependent on sample concentration, linear standard
curves of concentration versus peak intensity were constructed for syn-
thetic A
and the A
1-42Met35(O) dimer (see supplemental material,
available at www.jneurosci.org).
Preparation of A
1-42Met35(O) dimer. Resin-bound A
11-42Met35(O)
was prepared according to standardmethods (Tickler et al., 2001; Barnham
et al., 2003). Dityrosine was prepared according to the previously re-
portedmethod (Barnhamet al., 2004; Skaff et al., 2005). Fmoc protection
of dityrosine and incorporation into SPPS of the A
1-42Met35(O) dimer
was performed according to the previously reported method (Kok et al.,
2009).
Genotyping. ApoE genotype was determined by PCR amplification of
genomic DNA.
Neuropsychological assessments. All subjects undertook a variety of
neuropsychological tasks, designed to assess a broad range of cognitive
Figure 1. Representative SELDI-TOF/MS spectra. Samples extracted from the CE of blood fromanAD subject (top) and a normal
age-matched control (bottom). In these examples the antibody WO2 was used. Peaks marked withF are A
42 and the corre-
sponding dimer (*), respectively, are elevated in AD. In contrast, the 9962 Da peak (†) is elevated in HC.
Table 1. Demographics: participants means (SD) for groups by classification
HC (n 52) MCI (n 23) AD (n 43)
Age (years) 73.2 (7.7) 70.9 (9.6) 72.7 (10.7)
MMSE 29.0 (1.1) 26.3 (2.4)* 21.8 (4.1)*
Male/female 25/27 9/14 20/23
% ApoE 4 29% 48%† 73%†
Composite memory
0.1 (0.9)
2.4 (1.0)*
3.4 (0.6)*
Composite frontal 0.1 (0.7)
0.9 (0.6)*
1.8 (0.9)*
Gray matter volume 0.31 (0.01) 0.29 (0.01)* 0.29 (0.01)*
Neocortical PiB SUVR 1.4 (0.4) 1.9 (0.6)* 2.3 (0.4)*
*Significantly different from controls ( p 0.05).
†Significantly different from controls (Fisher’s exact test p 0.05).
Table 2. Mass-to-charge ratios of peaks identified using the antibodies WO2 and
4G8 in the SELDI-TOF MS that were different between AD and HC subjects
WO2 4G8
Peak m/z p Peak m/z p
4529 0.07
4625 0.003
5289 0.028
9058 0.001 9058 0.001
9962 0.002
10293 0.004 10292 0.034
11310 0.001 11310 0.044
11346 0.001
11432 0.014
12310 0.002 12310 0.024
12768 0.003
12834 0.008
15315 0.07
15524 0.06
6316 • J. Neurosci., May 5, 2010 • 30(18):6315–6322 Villemagne et al. • Blood-Borne A
Dimer in Alzheimer’s Disease

domains commonly affected by AD and aging. In addition to the MMSE
and CDR, episodic memory was assessed using delayed recall of the Cal-
ifornia Verbal Learning Test–Second edition (CVLT-II), and the Rey
Complex Figure Test (RCFT), while executive function was measured
using letter fluency, category fluency, verbal fluency switching task, and
the incongruent condition of the Stroop task. Using the neuropsycholog-
ical test results of 65 healthy older people with negative PiB and normal
MRI scans as the reference, a composite episodic memory score was
generated by taking the average of the Z scores
for the memory tasks (Pike et al., 2007), and a
composite executive function score was gener-
ated by taking the average of theZ scores for the
executive function tasks.
Neuroimaging. All subjects underwent a 3D
spoiled gradient echo (SPGR) T1-weighted
MRI acquisition for screening, quantification
of gray matter atrophy, and subsequent coreg-
istration with the PET images. As described
previously (Ourselin et al., 2001), T1-weighted
MR images for each subject were classified into
gray matter (GM), white matter (WM), and
CSF using an implementation of the expecta-
tion maximization segmentation algorithm.
The Montreal Neurological Institute (MNI)
single-subject MRI brain template (Collins et
al., 1998) and corresponding Automated Ana-
tomical Labeling (AAL) template (Tzourio-
Mazoyer et al., 2002) and tissues priors were
spatially normalized to each participant to au-
tomatically obtain a parcellation for each se-
lected atlas into GM, WM, and CSF. The
measured gray matter volumes were normal-
ized for head size using the total intracranial
volume, defined as the sum of GM, WM, and
CSF volumes. The volume results are presented
as the proportion of total intracranial volume.
Production of 11C-PiB and PiB-PET scans
were performed at the Centre for PET, Austin
Hospital, as previously described (Rowe et al.,
2007). Briefly, a 30 min acquisition emission
PET scan was acquired starting at 40 min after
the administration of 370MBq of 11C-PiB. Re-
gional Standardized Uptake Value (SUV), de-
fined as the decay-corrected brain radioactivity
concentration, corrected for injected dose and
body weight, was normalized to the cerebellar cortex to obtain SUV
Ratios (SUVR) (Rowe et al., 2007). Neocortical A
burdenwas expressed
as the average SUVR of the area-weighted mean for the following
cortical regions of interest: frontal (consisting of dorsolateral pre-
frontal, ventrolateral prefrontal, and orbitofrontal regions), superior
parietal, lateral temporal, lateral occipital, and anterior cingulate and
posterior cingulate.
Statistical analysis.Continuous variables for the groups were tested for
normality of distribution using the Shapiro–Wilk test and visual inspec-
tion of variable histograms. Statistical evaluations were performed
using ANOVA for all spectrum peaks, using false discovery rate (FDR)
(Benjamini and Hochberg, 1995) to select the peaks that were different
between groups. The selected peaks for each group were then compared
using ANOVA, followed by a Dunnet’s test to compare each group with
controls, and a Tukey-Kramer HSD test to establish differences between
groupmeans. Categorical differences were evaluated using Fisher’s exact
test. Pearson product-moment correlation analyses were conducted be-
tween the different variables. Assessment of the robustness of the corre-
lations was performed via tenfold leave-10%-out cross-validation. To
investigate the patterns and interrelationships between the series of spec-
trum peaks, a hierarchical cluster analysis using an average linkage
method was performed (Sokal and Michener, 1958). In all instances
statistical significance was defined as p  0.05. Multiple comparisons
were controlled for with a FDR. Data are presented as mean SD unless
otherwise stated.
Results
Demographic, clinical, neuropsychological, and neuroimaging
characteristics of the 118 participants are reported in Table 1.
There were significant differences between the healthy older con-
trol (HC), AD, andMild Cognitive Impairment (MCI) groups in
Figure 2. Distribution of blood SELDI-TOF MS values in regards to mass to charge ratio and clinical classification in 118 partic-
ipants. Hierarchical cluster analysis showing two distinct classes of peaks associated with the normal or abnormal processing of
APP.A, At the left end, peak 9962Da, a species thatwas higher in HC than in AD. On the right end of the cluster are peaks due to the
A
monomer and dimer that were higher in AD. The pattern of the cluster analysis is consistent with two different processing
pathways for APP: an amyloidogenic pathway that is elevated in AD and a non-amyloidogenic pathway that is elevated in the
control subjects. Box-and-whiskersplots comparing the intensities ofA
monomer, dimer, and thepeakat9962Da inHC,MCI, and
AD subjects. B, A good separation is observed between the AD and HC groups (Cohen’s d: 0.41, 0.76 and 0.73 for monomer, dimer
and 9962 Da peak, respectively).
Figure 3. Correlation between A
42 monomer and dimer. The intensities of the peaks
assigned to themonomericA
42and the correspondingdimer from118participants arehighly
correlated (r 0.79, p 0.0001). HC, Blue circle; MCI, green
; AD, red circle.
Villemagne et al. • Blood-Borne A
Dimer in Alzheimer’s Disease J. Neurosci., May 5, 2010 • 30(18):6315–6322 • 6317

MMSE and cognitive composite scores.
The AD group had a higher prevalence
(70%) of ApoE 4 allele carriers. There
was no significant difference in age or in
gender distribution between groups.
Mass spectra profiles for cellular
elements are different between AD
and control
Blood was collected from all participants
and fractionated into plasma and CEwhich
were then extracted with an aqueous solu-
tion of urea and Triton X-100. The purpose
of adding a denaturant anddetergentwas to
break up any potential protein/protein or
protein/membrane interactions involving
any APP/A
fragments. A variety of dena-
turants and detergents were assessed over a
large concentration range to identify the
conditions that reproducibly gave the best
signal-to-noise in the mass spectra. The ex-
tractedmaterial was then analyzed in tripli-
cate by SELDI-TOF MS, with the operator
blinded to the disease status of the subjects,
using antibody capture (WO2 epitope A

residues 4–8 and 4G8 epitope A
residues
17–21). These are generic anti-A
antibod-
ies able to capture most A
species.
The spectra of the CEmaterial from all
subjects contained a large number of
peaks; with m/z ranging from 3.5 to 16.0
kDa; consistent with a variety of APP/A

fragments present in the CE. Moreover,
the spectra obtained from AD subjects
were substantially different to that of the
HC, and 10 peaks showed significant
differences between the two groups (Fig. 1
shows representative examples). The m/z
values of the peaks with intensities that were found to be signifi-
cantly different between theHCandADgroups are listed inTable
2. SELDI-TOF MS of the plasma fraction did not resolve any
peaks that were significantly different between the control and
disease groups; nor were any peaks due to species normally asso-
ciated with AD (i.e., A
) observed (data not shown).
A dimers are elevated in AD blood
Hierarchical clustering of the CEWO2 dataset is shown in Figure
2A. At one end of the cluster analysis are a number of peaks that
were elevated in AD with m/z ratios that are consistent with spe-
cies prominent in the amyloidogenic pathway commonly associ-
ated with AD, including A
42 (them/z of 4529 is consistent with
an oxidized form of A
; i.e., the calculated molecular weight of
A
42 is 4513Da plus 16 from an oxygen atom). The intensities of
the peaks due to A
42 are 16% higher in the AD subjects com-
paredwith the controls (Fig. 2B); while there is a strong trend this
increase in monomeric A
levels did not reach statistical signifi-
cance. However, the peakwith am/z of 9058Da corresponding to
themass of anA
42 dimerwas found to be significantly increased
in AD compared with control subjects (35% higher p  0.001)
(Fig. 2B). There was a strong correlation between the amount of
monomeric A
42 detected and the corresponding dimer (r 
0.79, p  0.0001) (Fig. 3). The dimer peak was also significantly
higher in the CE of ADpatients when 4G8was used as the capture
antibody (Table 2). While definitive identification of the species
responsible for the various peaks will require high resolution
MS/MS data, to more confidently characterize that the peaks at
m/z 4529 and 9058 corresponded to A
42 species, SELDI-TOF
spectra were collected using the antibodies G210 (A
40 specific)
and G211 (A
42 specific) (Ida et al., 1996). As shown in Figure
4A, the peaks that were assigned to A
42 and the correspond-
ing dimer were detected by three different A
antibodies (4G8,
WO2, G211) but were not detected by the A
40-specific anti-
body G210. To further characterize the A
dimer we synthe-
sized an oxidized A
dimer (the sulfur atom of Met35 is
oxidized to a sulfoxide) where the two A
peptide chains are
covalently cross-linked with a dityrosine moiety at residue
number 10. As can be seen from Figure 4B the SELDI MS of
the Mwt of this synthetic dimer is the same as the m/z of the
dimer elevated in AD blood.
At the opposite end of the cluster analysis (Fig. 2A) are a group
of peaks that were decreased in AD compared with control; the
largest difference being for a peak with a m/z of 9962 Da which
was significantly decreased in AD compared with HC by 56%
( p 0.002) (Fig. 2B). Thismolecularweight is too small to be the
result of
-secretase activity, and too big to be the result of
-secretase activity, suggesting an alternative, non-amyloido-
genic processing pathway. The observed distribution in the clus-
ter analysis (Fig. 2A) is consistent with the notion that there are
Figure 4. SELDI-TOFMS of CEs from an AD subject.A, Peaks due tomonomeric (*) and dimeric (**) A
42 are detected by three
different antibodies: WO2 (epitope 4–8), 4G8 (epitope 17–21), and G211 (C terminus of A
42), but not by G210 (C terminus of
A
40).B, Comparison of SELDI-TOFMS profile of the A
dimer detected in the CEs with that of a synthetic A
dimer in which the
A
peptide chains contain a sulfoxide at residue M35 and are covalently crosslinked by a dityrosine moiety.
6318 • J. Neurosci., May 5, 2010 • 30(18):6315–6322 Villemagne et al. • Blood-Borne A
Dimer in Alzheimer’s Disease

differential processing pathways for APP between diseased and
healthy subjects.
APP/A biomarkers correlate with clinical measures of AD
Not only are the A
monomer and dimer elevated in AD blood
compared with controls, but there are correlations between the
levels of these species in the blood and other clinical, neuropsy-
chometric, and biological markers of AD. These include MMSE
(r  0.35 and 0.36, p  0.0002 and 0.0001, for monomeric and
dimeric A
, respectively), memory impairment (r  0.27 and
0.37, p  0.004 and 0.0001, for monomer and dimer respec-
tively), executive function (r  0.30 and 0.35, p  0.0012 and
0.0002, for monomer and dimer respectively), gray matter vol-
ume (r
0.33, p 0.011, for the dimer), and brain A
burden
as measured by PET imaging using 11C-PiB (r  0.19 and 0.22,
p  0.04 and 0.02, for monomer and dimer respectively) (Fig.
5A). Cross-validation of these results showed the robustness of
the findings, with cross-validated correlation coefficients (q2) be-
ing on average 7% and 15% worse than significant non-cross-
validated correlation coefficients (r2) values for monomer and
dimer ratios, respectively, being better for neuropsychometric
measures than for neuroimaging parameters (6% vs 23%, respec-
tively). No correlations were foundwhen the clinical groups were
examined separately.
Furthermore, all the subjects including the MCI group were
separated into PiB-positive andPiB-negative groups using aNeo-
cortical SUVR threshold of 1.45, obtained by unbiased statistical
approaches such as hierarchical cluster analysis or partitionmod-
els for the determination of a Neocortical PiB ‘cutoff’ level ap-
plied to the HC group. Using this threshold, 98% of AD, 57% of
MCI and 35% of HC were classified as PiB-positive. Both the
monomer ( p 0.013) and dimer ( p 0.0002) are significantly
elevated within the PiB-positive group when compared with the
negative group (Fig. 5B).
There are also significant correlations between m/z 9962 Da
and the clinical neuropsychometric (respectively r
0.26 ( p
0.006); 0.25 ( p 0.008);
0.21 ( p 0.028) forMMSE,memory
impairment, and executive function) and brain A
burden r 

0.28 ( p  0.003) (Fig. 5A). As with the monomer and dimer
peaks, no correlation was found when the clinical groups were
examined separately. The 9962 m/z peak was significantly ( p 
0.005) elevated in the PiB-negative group (Fig. 5B). As this peak is
Figure 5. Relationship between blood SELDI-TOF MS mass to charge ratios and clinical and neuroimaging parameters. A
monomer, dimer and 9962 Da are highly correlated with clinical,
neuropsychometric and biological markers, such as MMSE, memory performance, executive function, gray matter volume, and brain A
burden as measured by PiB-PET, underlying their
interrelationship. A, These graphs reflect the balance in APP processing between the amyloidogenic and non-amyloidogenic pathway that defines AD. HC, Blue circle; MCI, green
; AD, red circle.
Box-and-whiskers plots of the intensities of A
monomer, dimer, and 9962 with brain A
burden as measured by PiB PET. B, A PiB SUVR threshold of 1.45 was used to separate the groups in
PiB-positive (PiB-pos) and PiB-negative (PiB-neg).
Villemagne et al. • Blood-Borne A
Dimer in Alzheimer’s Disease J. Neurosci., May 5, 2010 • 30(18):6315–6322 • 6319

higher in controls relative to diseased subjects, the correlations
are of opposite sign to those of monomeric and dimeric A
.
As the 9962 m/z peak and the A
dimer peak reflect a balance
between two different APP processing pathways, i.e., an amyloi-
dogenic pathway and a non-amyloidogenic pathway, we further
examined the ability of the ratio of themonomer anddimer to the
9962 Da peak to discriminate between AD and controls. The
distinction betweenADand controls for themonomer and dimer
ratios was better than when using the peaks intensities indepen-
dently ( p 0.0007 and0.0001, for monomer and dimer ratios
respectively) (Fig. 6A), and the correlations with clinical markers
of ADwere also improved:MMSE (r 0.39 and 0.41, p 0.0001,
for monomer and dimer ratios, respectively), memory impair-
ment (r  0.36 and 0.42, p  0.0001, for monomer and dimer
ratios, respectively), executive function (r  0.34 and 0.38, p 
0.0003 and 0.0001, for monomer and dimer ratios, respec-
tively), gray matter volume (r
0.18 and
0.31, p 0.16 and
0.016, formonomer and dimer ratios, respectively), and brainA

burden as measured by 11C-PiB PET (r  0.33 and 0.35, p 
0.0003 and 0.0002, for monomer and dimer ratios, respectively)
(Fig. 6B). This improvement was also reflected in a larger effect
size for the ratio compared with the one obtained with the A

dimer alone (1.03 and 0.76, respectively).
Discussion
The genetics of familial forms of AD implicates the A
peptide as
playing a pivotal role in the development of AD and A
deposi-
tion in the form of amyloid plaque is one of the defining patho-
logical hallmarks of AD. Yet one of the conundrums of AD
research is that plaque burden does not necessarily correlate with
disease progression, a finding that has been supported by data
using 11C-PiB PET imaging to show that A
burden does not
correlate with cognitive impairment in AD (Rowe et al., 2007),
and that 20–30% of seemingly healthy older controls are defined
as PiB positive indicating significant amyloid pathology is present
(Mintun et al., 2006; Rowe et al., 2007; Aizenstein et al., 2008).
However, it has been shown that relatively low abundance soluble
oligomers better correlate with disease progression (Lue et al.,
1999; McLean et al., 1999) and that synaptotoxic A
dimers are
elevated in AD brains (Shankar et al., 2008). As a result, there is
much interest in developing therapeutic and diagnostic strategies
targeting soluble oligomers of A
.
Before this study there has been no direct detection of A

oligomers in blood which probably reflects limitations in the
technology used and in the tissue being examined. None of the
A
biomarker protocols currently used clinically, e.g., PiB PET
imaging or A
ELISAs were designed to specifically detect such
species, although there has been a recent report of an ELISA
method against A
oligomers (Xia et al., 2009).
We have previously shown that SELDI-MS technology is able
to detect an array of A
oligomers and that these oligomers have
a high affinity for lipid membranes (Hung et al., 2008). There-
fore, while most attempts at identifying blood-based biomarkers
Figure 6. Characteristics and relationship of the ratio of the blood SELDI-TOF MS values with clinical and neuroimaging parameters in 118 participants. Box-and-whiskers plots comparing the
respective A
monomer and dimer to 9962 Da ratios in HC,MCI, and AD subjects.A, Amuch better separation between the AD andHC groups is obtained (Cohen’s d: 0.76 and 1.03 formonomer and
dimer ratios, respectively) thanwhen the peak intensities are examined separately. A
monomer and dimer to 9962 Da ratios are highly correlatedwith clinical, neuropsychometric, and biological
markers. B, The correlations are better than when either the monomer, the dimer, or the 9962 Da peaks are examined separately. HC, Blue circle; MCI, green
; AD, red circle.
6320 • J. Neurosci., May 5, 2010 • 30(18):6315–6322 Villemagne et al. • Blood-Borne A
Dimer in Alzheimer’s Disease

for AD have concentrated on plasma with disappointing results
(Zetterberg andBlennow, 2006), it is likely that theA
oligomers,
which have a high affinity for lipid binding, would be found
associatedwith the lipidmembranes of the usually discardedCEs.
Using SELDI-TOF MS technology it is possible to detect a
dimeric form of A
in human blood and show that the levels of
the dimer are significantly elevated in AD (Fig. 2B) and correlate
with clinical markers of the disease (Fig. 5).
In the search for potential biomarkers for AD it has been
found that autoantibodies against A
oligomers are decreased in
the blood of AD subjects (Moir et al., 2005; Britschgi et al., 2009).
In the study by Moir et al. (2005), the decreased autoantibodies
were reported to target a specific formof oxidativelymodified A

caused by a reaction with copper and resulting in covalently
cross-linked amyloid protein species, the so-called CAPS. The
reaction of A
and copper has been shown to lead to the forma-
tion of dityrosine cross-linked oligomers of A
(Atwood et al.,
2004; Barnham et al., 2004). One consequence of the reduction in
autoantibodies is greater accumulation of these A
oligomers in
the blood. Interestingly, we were able to show that the A
dimer
detected in blood had the sameMS profile (Fig. 3B) as a synthetic
A
dimer were the A
peptide chains that contain a sulfoxide at
residue M35 are covalently crosslinked by a dityrosine moiety.
A peak detected at a m/z of 9962 Da is lower in diseased sub-
jects (Fig. 2B) and is inversely correlated with clinical markers of
AD (Fig. 5). This molecular weight does not correspond to any
obvious APP fragment—it is too small to be a fragment resulting
from
-secretase activity, and too big to be the result of
-secretase activity—nor does the mass correspond to any A
-
like aggregate. This suggests that the fragment is generated via an
alternative, non-amyloidogenic processing pathway. While de-
finitive identification of this fragment will require isolation and
amino acid sequencing, it has previously been reported that the
activity of cathepsin D is decreased in the blood of AD subjects
(Straface et al., 2005). Cathepsin D is an aspartyl protease that
cleaves APP at a number of different sites (Higaki et al., 1996),
including at Ser627, Phe765, Glu766, and Met768. Cleavage at
these sites would give rise to a number of 15 kDa fragments;
subsequent -cleavage of these fragments by -secretase at
Met722 would give rise to a 10 kDa fragment. There is also a
cathepsin D cleavage site at Val669, subsequent cleavage by
-secretase at the  site would give rise to a 5 kDa fragment. As
can be seen from Table 2 and the cluster analysis (Fig. 2A), a
number of related fragments with similar masses are detected as
being elevated in the blood of the control subjects including the
peak at 9962 Da.
The spectrum of APP fragments observed by SELDI-TOF MS
in the cluster analysis is consistent with there being two distinct
processing pathways for APP, an amyloidogenic pathway which
predominates in AD and a non-amyloidogenic pathway which
predominates in healthy subjects (Fig. 2A). Both these pathways
occur physiologically and it could be argued that the progression
to AD is the result of a shift in the processing of APP from the
non-amyloidogenic to the amyloidogenic pathway. The genetics
of early onset AD support this model (Fassbender et al., 2001).
The recent publication that
-secretase activity in platelets is in-
creased in AD compared with controls (Zainaghi et al., 2007;
Johnston et al., 2008) is consistent with the amyloidogenic path-
way being “favored” in AD.
Given that the onset of the disease may predate clinical symp-
toms of AD by many years (Price and Morris, 1999), the lack of
valid biomarkers has hampered the development and evaluation
of effective therapies for AD (Clark et al., 2008). A number of
potential AD therapeutic strategies targeting A
and its oli-
gomers (so called disease-modifying drugs) are currently being
investigated, including immunotherapy designed to promote A

clearance (Nicoll et al., 2006), secretase inhibitors, which prevent
A
generation (Olson and Albright, 2008), scyllo-inositol, which
is reported to inhibit toxic A
oligomers binding to membranes
(Nitz et al., 2008), and PBT2—a second generation MPAC that
inhibits the formation of toxic A
oligomers (Adlard et al., 2008).
The assessment of outcomes of the clinical trials is often difficult
to define as they rely on highly variable neuropsychometric tests.
To overcome the variability that is inherent in these tests, large
sample sizes and long timeframes are required to observe subtle
changes in subjects’ performance, dramatically increasing the
cost of these trials. The ability to detect preclinical or early stage
disease through reliable laboratory and neuroimaging biomark-
ers for AD would enable more efficient clinical trials to be de-
signed and monitored. Ideally a biomarker should reflect a
disease-specific process and be detected in an easily collected
tissue.
The most easily accessed tissue is blood and the fractionation
procedures we used were deliberately kept simple to reflect the
standard protocol used in clinical laboratories worldwide to per-
form plasma-based assays. In this instance, however, we analyzed
the usually discarded membrane-rich CE fraction. The data pre-
sented here establishes that disease relevant APP/A
-based bi-
omarkers are likely to be found in the membrane fraction of
blood. Given that the blood borne biomarkers correlate with dis-
ease progression they hold the promise of providing a simple yet
effective way of monitoring the success or otherwise of the vari-
ous disease modifying therapies targeting A
/APP processing.
Because the molecular changes occur well before the pheno-
typicalmanifestation of disease, identification of specific biomar-
kers for particular traits of the pathological process will permit
early intervention with disease-modifying medications. Further
characterization of the different species in AD and controls is
warranted, while ongoing longitudinal studies will help elucidate
how these markers change over time and how they relate to cog-
nitive decline.
References
Adlard PA,ChernyRA, FinkelsteinDI, Gautier E, RobbE, CortesM,Volitakis
I, Liu X, Smith JP, Perez K, Laughton K, Li QX, Charman SA, Nicolazzo
JA, Wilkins S, Deleva K, Lynch T, Kok G, Ritchie CW, Tanzi RE, et al.
(2008) Rapid restoration of cognition in Alzheimer’s transgenic mice
with 8-hydroxy quinoline analogs is associated with decreased interstitial
Abeta. Neuron 59:43–55.
Aizenstein HJ, Nebes RD, Saxton JA, Price JC, Mathis CA, Tsopelas ND,
Ziolko SK, James JA, Snitz BE, Houck PR, Bi W, Cohen AD, Lopresti BJ,
DeKosky ST, Halligan EM, KlunkWE (2008) Frequent amyloid deposi-
tion without significant cognitive impairment among the elderly. Arch
Neurol 65:1509–1517.
Atwood CS, Perry G, Zeng H, Kato Y, Jones WD, Ling KQ, Huang X, Moir
RD, Wang D, Sayre LM, Smith MA, Chen SG, Bush AI (2004) Copper
mediates dityrosine cross-linking of Alzheimer’s amyloid-beta. Biochem-
istry 43:560–568.
Barnham KJ, Ciccotosto GD, Tickler AK, Ali FE, Smith DG, Williamson NA,
LamYH, CarringtonD, TewD, Kocak G, Volitakis I, Separovic F, Barrow
CJ, Wade JD, Masters CL, Cherny RA, Curtain CC, Bush AI, Cappai R
(2003) Neurotoxic, redox-competent Alzheimer’s beta-amyloid is re-
leased from lipid membrane by methionine oxidation. J Biol Chem
278:42959–42965.
Barnham KJ, Haeffner F, Ciccotosto GD, Curtain CC, Tew D, Mavros C,
Beyreuther K, Carrington D, Masters CL, Cherny RA, Cappai R, Bush AI
(2004) Tyrosine gated electron transfer is key to the toxic mechanism of
Alzheimer’s disease beta-amyloid. FASEB J 18:1427–1429.
Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: a
Villemagne et al. • Blood-Borne A
Dimer in Alzheimer’s Disease J. Neurosci., May 5, 2010 • 30(18):6315–6322 • 6321

practical and powerful approach to multiple testing. J R Stat Soc B
57:289–300.
Britschgi M, Olin CE, Johns HT, Takeda-Uchimura Y, LeMieux MC,
Rufibach K, Rajadas J, Zhang H, Tomooka B, Robinson WH, Clark CM,
Fagan AM, Galasko DR, Holtzman DM, Jutel M, Kaye JA, Lemere CA,
Leszek J, Li G, Peskind ER, et al. (2009) Neuroprotective natural anti-
bodies to assemblies of amyloidogenic peptides decrease with normal
aging and advancing Alzheimer’s disease. Proc Natl Acad Sci U S A
106:12145–12150.
Cappai R, Barnham KJ (2008) Delineating the mechanism of Alzheimer’s
disease A beta peptide neurotoxicity. Neurochem Res 33:526–532.
Clark CM, Davatzikos C, Borthakur A, Newberg A, Leight S, Lee VM, Tro-
janowski JQ (2008) Biomarkers for early detection of Alzheimer pathol-
ogy. Neurosignals 16:11–18.
CollinsDL, ZijdenbosAP, KollokianV, Sled JG, KabaniNJ,HolmesCJ, Evans
AC (1998) Design and construction of a realistic digital brain phantom.
IEEE Trans Med Imaging 17:463–468.
Fassbender K, Masters C, Beyreuther K (2001) Alzheimer’s disease: molec-
ular concepts and therapeutic targets. Naturwissenschaften 88:261–267.
Glabe CG, Kayed R (2006) Common structure and toxic function of amy-
loid oligomers implies a commonmechanismof pathogenesis. Neurology
66:S74–78.
Higaki J, Catalano R, Guzzetta AW, Quon D, Nave´ JF, Tarnus C,
D’Orchymont H, Cordell B (1996) Processing of beta-amyloid precur-
sor protein by cathepsin D. J Biol Chem 271:31885–31893.
Hung LW, Ciccotosto GD, Giannakis E, TewDJ, Perez K,Masters CL, Cappai
R, Wade JD, Barnham KJ (2008) Amyloid-beta peptide (Abeta) neuro-
toxicity ismodulated by the rate of peptide aggregation: Abeta dimers and
trimers correlate with neurotoxicity. J Neurosci 28:11950–11958.
Ida N, Hartmann T, Pantel J, Schro¨der J, Zerfass R, Fo¨rstl H, Sandbrink R,
Masters CL, Beyreuther K (1996) Analysis of heterogeneous A4 peptides
in human cerebrospinal fluid and blood by a newly developed sensitive
Western blot assay. J Biol Chem 271:22908–22914.
Johnston JA, Liu WW, Coulson DT, Todd S, Murphy S, Brennan S, Foy CJ,
Craig D, Irvine GB, Passmore AP (2008) Platelet beta-secretase activity
is increased in Alzheimer’s disease. Neurobiol Aging 29:661–668.
KlunkWE, EnglerH,Nordberg A,WangY, Blomqvist G,Holt DP, Bergstro¨m
M, Savitcheva I, Huang GF, Estrada S, Ause´n B, Debnath ML, Barletta J,
Price JC, Sandell J, Lopresti BJ, Wall A, Koivisto P, Antoni G, Mathis CA,
Långström B (2004) Imaging brain amyloid in Alzheimer’s disease with
Pittsburgh Compound-B. Ann Neurol 55:306–319.
KokWM, Scanlon DB, Karas JA, TewD,Miles LA, ParkerMW, BarnhamKJ,
Hutton CA (2009) Solid-phase synthesis of homodimeric peptides:
preparation of covalently-linked dimers of amyloid-beta peptide. Chem
Commun (Camb) 41:6228–6230.
Lue LF, Kuo YM, Roher AE, Brachova L, Shen Y, Sue L, Beach T, Kurth JH,
Rydel RE, Rogers J (1999) Soluble amyloid beta peptide concentration
as a predictor of synaptic change in Alzheimer’s disease. Am J Pathol
155:853–862.
Masters CL, Cappai R, Barnham KJ, Villemagne VL (2006) Molecular
mechanisms for Alzheimer’s disease: implications for neuroimaging and
therapeutics. J Neurochem 97:1700–1725.
McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan EM
(1984) Clinical Diagnosis of Alzheimer’s Disease: report of theNINCDS-
ADRDA Work Group under the auspices of Department of Health and
Human Services Task Force on Alzheimer’s Disease. Neurology
34:939–944.
McLean CA, Cherny RA, Fraser FW, Fuller SJ, SmithMJ, Beyreuther K, Bush
AI, Masters CL (1999) Soluble pool of A
amyloid as a determinant of
severity of neurodegeneration in Alzheimer’s disease. Ann Neurol
46:860–866.
MintunMA, LarossaGN, Sheline YI,DenceCS, Lee SY,MachRH,KlunkWE,
Mathis CA, DeKosky ST,Morris JC (2006) [11C]PIB in a nondemented
population: potential antecedentmarker of Alzheimer disease. Neurology
67:446–452.
Moir RD, Tseitlin KA, Soscia S, Hyman BT, Irizarry MC, Tanzi RE (2005)
Autoantibodies to redox-modified oligomeric Abeta are attenuated in the
plasma of Alzheimer’s disease patients. J Biol Chem 280:17458–17463.
Nicoll JA, Barton E, Boche D, Neal JW, Ferrer I, Thompson P, Vlachouli C,
Wilkinson D, Bayer A, Games D, Seubert P, Schenk D, Holmes C (2006)
Abeta species removal after abeta42 immunization. J Neuropathol Exp
Neurol 65:1040–1048.
Nitz M, Fenili D, Darabie AA, Wu L, Cousins JE, McLaurin J (2008) Mod-
ulation of amyloid-beta aggregation and toxicity by inosose stereoiso-
mers. FEBS J 275:1663–1674.
OlsonRE, Albright CF (2008) Recent progress in themedicinal chemistry of
gamma-secretase inhibitors. Curr Top Med Chem 8:17–33.
Ourselin S, Roche A, Subsol G, Pennec X, Ayache N (2001) Reconstructing
a 3D structure from serial histological sections. Image Vis Comput
19:25–31.
Petersen RC, Smith GE, Waring SC, Ivnik RJ, Tangalos EG, Kokmen E
(1999) Mild cognitive impairment: clinical characterization and out-
come. Arch Neurol 56:303–308.
Pike KE, Savage G, Villemagne VL, Ng S, Moss SA, Maruff P, Mathis CA,
Klunk WE, Masters CL, Rowe CC (2007) Beta-amyloid imaging and
memory in non-demented individuals: evidence for preclinical Alzhei-
mer’s disease. Brain 130:2837–2844.
Price JL, Morris JC (1999) Tangles and plaques in nondemented aging and
“preclinical” Alzheimer’s disease. Ann Neurol 45:358–368.
Rowe CC, Ng S, Ackermann U, Gong SJ, Pike K, Savage G, Cowie TF,
Dickinson KL, Maruff P, Darby D, Smith C, Woodward M, Merory J,
Tochon-Danguy H, O’Keefe G, Klunk WE, Mathis CA, Price JC, Masters
CL, Villemagne VL (2007) Imaging beta-amyloid burden in aging and
dementia. Neurology 68:1718–1725.
Schupf N, Tang MX, Fukuyama H, Manly J, Andrews H, Mehta P, Ravetch J,
Mayeux R (2008) Peripheral Abeta subspecies as risk biomarkers of Alz-
heimer’s disease. Proc Natl Acad Sci U S A 105:14052–14057.
Shankar GM, Li S, Mehta TH, Garcia-Munoz A, Shepardson NE, Smith I,
Brett FM, Farrell MA, Rowan MJ, Lemere CA, Regan CM, Walsh DM,
Sabatini BL, Selkoe DJ (2008) Amyloid-beta protein dimers isolated di-
rectly from Alzheimer’s brains impair synaptic plasticity and memory.
Nat Med 14:837–842.
Sjo¨gren M, Andreasen N, Blennow K (2003) Advances in the detection of
Alzheimer’s disease-use of cerebrospinal fluid biomarkers. Clin Chim
Acta 332:1–10.
Skaff O, Jolliffe KA, Hutton CA (2005) Synthesis of the side chain cross-
linked tyrosine oligomers dityrosine, trityrosine, and pulcherosine. J Org
Chem 70:7353–7363.
Sokal RR, Michener CD (1958) A statistical method for evaluating system-
atic relationships. Univ Kans Sci Bull 38:1409–1438.
Straface E, Matarrese P, Gambardella L, Vona R, Sgadari A, Silveri MC,
Malorni W (2005) Oxidative imbalance and cathepsin D changes as pe-
ripheral blood biomarkers of Alzheimer disease: a pilot study. FEBS Lett
579:2759–2766.
Tickler AK, Barrow CJ, Wade JD (2001) Improved preparation of amyloid-
beta peptides usingDBU asNalpha-Fmoc deprotection reagent. J Pept Sci
7:488–494.
Tzourio-Mazoyer N, Landeau B, Papathanassiou D, Crivello F, Etard O,
Delcroix N,Mazoyer B, JoliotM (2002) Automated anatomical labeling
of activations in SPM using a macroscopic anatomical parcellation of the
MNI MRI single-subject brain. Neuroimage 15:273–289.
Walsh DM, Selkoe DJ (2004) Deciphering the molecular basis of memory
failure in Alzheimer’s disease. Neuron 44:181–193.
Walsh DM, Klyubin I, Fadeeva JV, Cullen WK, Anwyl R, Wolfe MS, Rowan
MJ, Selkoe DJ (2002) Naturally secreted oligomers of amyloid beta pro-
tein potently inhibit hippocampal long-term potentiation in vivo. Nature
416:535–539.
Xia W, Yang T, Shankar G, Smith IM, Shen Y, Walsh DM, Selkoe DJ (2009)
A specific enzyme-linked immunosorbent assay for measuring beta-
amyloid protein oligomers in human plasma and brain tissue of patients
with Alzheimer disease. Arch Neurol 66:190–199.
Zainaghi IA, Forlenza OV, Gattaz WF (2007) Abnormal APP processing in
platelets of patients with Alzheimer’s disease: correlations with mem-
brane fluidity and cognitive decline. Psychopharmacology (Berl) 192:
547–553.
Zetterberg H, Blennow K (2006) Plasma Abeta in Alzheimer’s disease—up
or down? Lancet Neurol 5:638–639.
6322 • J. Neurosci., May 5, 2010 • 30(18):6315–6322 Villemagne et al. • Blood-Borne A
Dimer in Alzheimer’s Disease