Small-Animal PET Imaging of Amyloid-Beta Plaques with
[11C]PiB and Its Multi-Modal Validation in an APP/PS1
Mouse Model of Alzheimer’s Disease
Andre ´ Manook1*, Behrooz H. Yousefi1, Antje Willuweit2¤b, Stefan Platzer3¤c, Sybille Reder1,
Andreas Voss4, Marc Huisman1¤d, Markus Settles5, Frauke Neff6¤e, Joachim Velden2¤f, Michael Schoor7,
Heinz von der Kammer2, Hans-Ju ¨rgen Wester1¤g, Markus Schwaiger1, Gjermund Henriksen1.¤a,
1Nuklearmedizinische Klinik und Poliklinik, Klinikum rechts der Isar, Technische Universita ¨t Mu ¨nchen, Munich, Germany, 2Evotec AG, Hamburg, Germany,
3Neurologische Klinik und Poliklinik, Klinikum rechts der Isar, Technische Universita ¨t Mu ¨nchen, Munich, Germany, 4Institut fu ¨r Pathologie - Biomedizinische Mikroskopie,
Helmholtz Zentrum Mu ¨nchen, Neuherberg, Germany, 5Institut fu ¨r Ro ¨ntgendiagnostik, Klinikum rechts der Isar, Technische Universita ¨t Mu ¨nchen, Munich, Germany,
6Institut fu ¨r Allgemeine Pathologie und Pathologische Anatomie - Neuropathologie, Klinkum rechts der Isar, Technische Universita ¨t Mu ¨nchen, Munich, Germany,
7TaconicArtemis GmbH, Cologne, Germany
In vivo imaging and quantification of amyloid-b plaque (Ab) burden in small-animal models of Alzheimer’s disease (AD) is a
valuable tool for translational research such as developing specific imaging markers and monitoring new therapy
approaches. Methodological constraints such as image resolution of positron emission tomography (PET) and lack of
suitable AD models have limited the feasibility of PET in mice. In this study, we evaluated a feasible protocol for PET imaging
of Ab in mouse brain with [11C]PiB and specific activities commonly used in human studies. In vivo mouse brain MRI for
anatomical reference was acquired with a clinical 1.5 T system. A recently characterized APP/PS1 mouse was employed to
measure Ab at different disease stages in homozygous and hemizygous animals. We performed multi-modal cross-
validations for the PET results with ex vivo and in vitro methodologies, including regional brain biodistribution, multi-label
digital autoradiography, protein quantification with ELISA, fluorescence microscopy, semi-automated histological
quantification and radioligand binding assays. Specific [11C]PiB uptake in individual brain regions with Ab deposition
was demonstrated and validated in all animals of the study cohort including homozygous AD animals as young as nine
months. Corresponding to the extent of Ab pathology, old homozygous AD animals (21 months) showed the highest
uptake followed by old hemizygous (23 months) and young homozygous mice (9 months). In all AD age groups the
cerebellum was shown to be suitable as an intracerebral reference region. PET results were cross-validated and consistent
with all applied ex vivo and in vitro methodologies. The results confirm that the experimental setup for non-invasive [11C]PiB
imaging of Ab in the APP/PS1 mice provides a feasible, reproducible and robust protocol for small-animal Ab imaging. It
allows longitudinal imaging studies with follow-up periods of approximately one and a half years and provides a foundation
for translational Alzheimer neuroimaging in transgenic mice.
Citation: Manook A, Yousefi BH, Willuweit A, Platzer S, Reder S, et al. (2012) Small-Animal PET Imaging of Amyloid-Beta Plaques with [11C]PiB and Its Multi-Modal
Validation in an APP/PS1 Mouse Model of Alzheimer’s Disease. PLoS ONE 7(3): e31310. doi:10.1371/journal.pone.0031310
Editor: Mark R. Cookson, National Institutes of Health, United States of America
Received November 10, 2009; Accepted January 5, 2012; Published March 9, 2012
Copyright: ? 2012 Manook et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by grants from Deutsche Forschungsgemeinschaft (DFG, German Research Foundation): HE4560/1-3 (G.H., H.J.W., A.D.),
DR445/3-1 and DR445/4-1 (A.D.), GRK333 and IRTG1373 (A.M.) and by grants from Karl-Max von Bauernfeind Association (A.M.) and Elite Graduate Network of
Bavaria (A.M.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing Interests: A.M. is a freelance trainer and application specialist for Pmod Technologies, Zurich, Switzerland. A.W. and J.V. have been employees and
H.v.d.K. is an employee at Evotec, Hamburg, Germany. M.S. is an employee at TaconicArtemis, Cologne, Germany. The transgenic mouse model ‘‘Arte10’’ was
developed at TaconicArtemis and characterized at Evotec. The animal model was used by Technische Universita ¨t Mu ¨nchen, Munich, Germany, under the premise
of an academic collaboration. This does not alter the authors9 adherence to all the PLoS ONE policies on sharing data and materials.
* E-mail: firstname.lastname@example.org
¤a Current address: ABX advanced biochemical compounds GmbH, Radeberg, Germany
¤b Current address: Institut fu ¨r Neurowissenschaften und Medizin, Forschungszentrum Ju ¨lich GmbH, Ju ¨lich, Germany
¤c Current address: Munich, Germany
¤d Current address: Nuclear Medicine and PET research, VU University Medical Center, Amsterdam, Netherlands
¤e Current address: Institut fu ¨r Pathologie, Helmholtz Zentrum Mu ¨nchen, Neuherberg, Germany
¤f Current address: Institut fu ¨r Pathologie, Universita ¨tsklinikum Hamburg-Eppendorf, Hamburg, Germany
¤g Current address: Lehrstuhl fu ¨r Pharmazeutische Radiochemie, Technische Universita ¨t Mu ¨nchen, Munich, Germany
. These authors contributed equally to this work.
PLoS ONE | www.plosone.org1March 2012 | Volume 7 | Issue 3 | e31310
Neuritic plaques containing Ab and neurofibrillary tangles
continue to define the neuropathological entity of AD and a
definite diagnosis can still only be established post-mortem [1–4].
The increased production of certain Ab species, their aggregation
and deposition as insoluble plaques is regarded as an early and key
pathology in the development of AD, and many modern treatment
approaches are directed at the prevention or reversal of Ab plaque
deposition in the brain . Ab plaque imaging with PET has now
entered the realm of the revised criteria for diagnosis of
Alzheimer’s disease  and helps to further improve early and
specific diagnosis and treatment monitoring .
Several radiolabeled compounds with high affinity and
specificity for Ab aggregates have been developed [8–19]. Among
these compounds, [11C]6-OH-BTA-1 ([11C]PiB) is presently the
one most extensively evaluated worldwide.
Advances in PET technology have facilitated the imaging of
small animals [20–22]. Transgenic mice possessing the mutations
held responsible for familiar AD have been shown to develop Ab
deposits, tangles and synaptic dysfunction, thus, mimicking human
AD pathology [23–31]. However, the correspondence between
preclinical and clinical data on Ab imaging remains a challenge in
transgenic models of AD .
Earlier in vivo, in vitro and ex vivo analyses suggested that
[11C]PiB shows specific binding to Ab plaques in transgenic mice
. Also, high-resolution imaging studies, such as MRI [34,35]
and in vivo optical imaging [36,37] demonstrated specific binding
to Ab plaques in transgenic mouse models. Although small-animal
PET imaging could allow for the quantification of global Ab
plaque load in the brain in vivo, previous studies suggest that
detection of small differences between transgenic and healthy
control animals by PET remains a challenge [33,38,39]. This may
be due to methodological limitations like image resolution in
relation to the small size of target structures, image co-registration,
animal motion, signal-to-noise ratios and cranial tracer distribu-
tion in rodents. Furthermore, transgenic mouse models suitable for
PET imaging of Ab plaques were lacking . Only one study
showed in vivo mouse brain imaging using PET with [11C]PiB
, though very high specific activities of the tracer were
required to obtain a signal.
Here, we report the evaluation and multi-modal cross-validation
of a feasible small-animal PET imaging approach with [11C]PiB.
Specific binding of [11C]PiB to Ab plaques in transgenic AD
mouse brain could be demonstrated in PET using specific activities
as used in clinical routine for humans. The mouse study collective
was designed with three transgenic groups of an APP/PS1 mouse
model  which serve as examples for different AD stages. PET
distinguished animals according to their Ab plaque burden and
these in vivo findings were validated in all other experimental
Mouse brain PET with [11C]PiB
For in vivo assessment of cerebral Ab plaque deposition, 47
animals in five study groups were scanned at least once with
[11C]PiB. 35 of these animals also received an in vivo MR brain
scan on a human clinical scanner (Table 1).
Visual inspection of co-registered PET/MR images revealed
distinct activity retention in the cortex of all transgenic mice
corresponding to their study group, whereas for all control animals
the cortex appeared to be free of specific activity uptake (Figure 1).
In transgenic mouse brain the activity uptake expanded through-
out the entire cortex, with slightly stronger signal in frontal
neocortical compared to hippocampal regions and a stronger
signal in the thalamus. These findings are in good correspondence
to the earlier onset of plaque deposition in cortex and to large
plaque sizes in thalamus as observed by Willuweit et al.  ex
Time-activity curves (TACs) of target and reference tissues
showed characteristics that were common for all study groups
(Figure 1 and Figure S6). Initial cerebellar uptake was always
higher than initial cortical uptake and cortical TACs of control
animals fell below cerebellar TACs early. In contrast, for each
transgenic animal the cortical TAC showed higher values than the
cerebellar TAC from around 3 min p.i. and remained distinctly
separable. From about 10 min p.i. on, each transgenic animal
could be assigned to its study group by its neocortex-cerebellum
The individual in vivo radioligand binding was examined by
calculating the binding potential with a reference tissue approach
using the cerebellum [41–44]. Parametric images of regional Ab
plaque burden for representative animals were created with the 2-
step multilinear reference tissue model 2 (MRTM2, [43,45]) as
shown in Figure 2 and Figure S1. The binding potential values for
neocortex estimated by the same model for the whole study
collective showed highly significant separation of all transgenic
animals from controls and a clear distinction of AD animals
belonging to different study groups (Figure S10). Old homozygous
AD animals (tgtg-old) exhibited highest activity retention, followed
by old hemizygous (tg-old) and young homozygous mice (tgtg-
young). The tg-old animals were at all scans in between the
homozygous groups while their results were generally closer to
those of the tgtg-young group. In control animals we never
observed any specific tracer uptake within the entire brain.
In general, no significant differences in activity uptake were
found between old and young or female and male control animals.
Further, no significant difference was found between male and
female young homozygous animals. Also, there were no
differences between right and left hemispheric tracer uptake
observed in all animals.
The robustness and consistency of PET results was shown by
performing test-retest experiments (Figure S5), by calculating
alternative measures for radioligand binding (Table S1) and by
averaging all neocortical and all cerebellar TACs for each study
group (Figure S6).
In general, the results above show a tight correlation of visual
inspection and PET analysis with all other modalities (Figure 2 and
Autoradiography with [3H]PiB ex vivo
Extensive ex vivo autoradiography of brain slices was performed
to verify that the cortical tracer uptake values as measured by PET
represent true binding of [11C]PiB to cortical Ab plaques. All
animals in this analysis had a PET scan with [11C]PiB, before.
On visual inspection, representative slices showed a homo-
genously dotted pattern of intensive multi-focal tracer retention
throughout the cortex of transgenic mice with a fully symmetric
right-left appearance. Particularly high uptake was detected in the
neocortex, hippocampus and thalamus (Figure 2 (column 2)). The
strong uptake in the thalamus was notable due to very high tracer
retention in fewer but much larger plaques. Plaques were also
present in the olfactory bulb although smaller in size (data not
shown). Without exception, the cerebellum was free of specific
[3H]PiB uptake in autoradiography. The entire brain of control
animals did neither show focal nor diffusely increased neocortical
tracer uptake (Figure 2). A clear difference in the amount of
In Vivo PiB Imaging in AD Mice
PLoS ONE | www.plosone.org2March 2012 | Volume 7 | Issue 3 | e31310
[3H]PiB uptake, corresponding to different stages of Ab plaque
burden, was noted even visually. Hence, the representative
samples of the four major study groups showed corresponding
results of ex vivo [3H]PiB uptake in autoradiography to [11C]PiB
uptake as measured in PET. In addition, individual uptake
patterns were in full correspondence to the patterns of Thioflavin
S and anti-Ab40/42 stains done on neighboring sections (Figure 2).
Visual perception of differences in ex vivo tracer uptake
between the study groups (Figure 2) could be confirmed
quantitatively by measuring a total of 64 slides from 24 animals
of the study collective yielding 248 observations for the cortical
region (Table 1).
The neocortex-to-cerebellum ratios of [3H]PiB uptake fully
reflected the in vivo PET findings for the same target region. The
average ratios for tg-old were 1.9060.26 (range: 1.46–2.11), for
tgtg-young 1.2560.07 (range: 1.19–1.33), for tgtg-old 2.5460.27
(range: 2.13–2.71) and for ctl-old 0.9360.02 (range: 0.90–0.95).
Statistical significance of differences between groups was tested for
ctl-old against tgtg-young (p,0.001), tgtg-young against tg-old
(p=0.004) and tg-old against tgtg-old (p=0.002) corresponding to
the staging of Ab load in these groups.
No significant differences in tracer uptake were found between
right and left neocortical regions in all animals and between old
and young control animals.
The individual results and pairwise correlations (Figure S10) are
Regional brain biodistribution of [11C]PiB
Twenty animals from the homozygous transgenic study groups
and matched controls were used for ex vivo regional brain
biodistribution of [11C]PiB at 30 min p.i.. Mouse brains were
dissected into four regions: 1. telencephalon as the major target
region, 2. olfactory system for its proximity to high extracerebral
uptake regions, 3. cerebellum as the reference region and 4.
diencephalon and midbrain as the remaining brain structures. The
cerebellum was used as the reference region for ratio calculations
of individual %ID/g values.
The target-to-reference ratios for telencephalon confirmed the
in vivo PET measurements for neocortex in these study groups.
The results for the old homozygous animals showed a large
difference to the young homozygous mice (p,0.001). This
corresponded to the large differences between these two groups
as seen in all the other experimental modalities (Table 2).
The young transgenic animals could easily be separated from
the controls (p=0.012). Differences between the young and old
control groups were not significant (p=0.090). In the control
groups, it is notable that the tracer uptake for the telencephalical
region relative to cerebellum was reversed (,1).
A similar behavior of relative uptake of [11C]PiB was found in
the other two target regions. The PiB uptake in the olfactory
system in transgenic mice is specific, but considerably lower than
in cortical regions. Young transgenic animals even had no
significantly higher uptake than young controls (p=0.491)
Table 1. Mouse study collective and numbers of mice per experiment.
flavin S Ab 40/42brain cranium
21.660.0 21.960.94 32.062.7444
SUM47354 242922 22 24204
Five study groups were defined, three of them with transgenic APP/PS1 mice, the others with age- and gender-matched controls. Major subgroups were female. Old
refers to an age of about 23 (hemizygous (tg)) and 21 months (homozygous (tgtg)). Young is defined as an age of 9 months. Young homozygous study group (tgtg-
young) and both control groups (ctl) were designed to reveal possible gender effects. As an overview and orientation for this study, numbers in each cell state how
many animals per subgroup were analyzed in the corresponding experiment. Mean ages, weights and injected doses are shown for each subgroup including standard
deviation. The pairwise correlations of these modalities are shown in Figure 7 and Figure S10.
(tg: hemizygous APP/PS1, tgtg: homozygous APP/PS1, ctl: C57BL6/J control animals).
In Vivo PiB Imaging in AD Mice
PLoS ONE | www.plosone.org3March 2012 | Volume 7 | Issue 3 | e31310
Figure 1. Small-animal [11C]PiB PET/MRI overview. [11C]PiB PET co-registered to in vivo 1.5T cranial MRI of the same mouse. Overview of cranial
tracer uptake shows images of four representative animals from the major study groups in radiological orthogonal perspective (20–30 min frame). (A)
23 month old female hemizygous APP/PS1 mouse (weight: 20.8 g, injected dose: 14.7 MBq, color scale 37–144 kBq/cc), (B) 9 month old female
homozygous APP/PS1 mouse (weight: 22.2 g, injected dose: 15.2 MBq, color scale 60–350 kBq/cc) (C) 21 month old female homozygous APP/PS1
mouse (weight: 24.5 g, injected dose: 24.2 MBq, color scale 73–280 kBq/cc), (D) 23 month old female C57BL/6J control mouse (weight: 29.9 g,
injected dose: 15.1 MBq, color scale: 66–300 kBq/cc). Columns from left to right show horizontal (1), coronal (2) and sagittal (3) views. The right
column (4) shows corresponding neocortical (yellow) and cerebellar (magenta) time-activity curves (TACs). Inset (5) shows initial tracer dynamics on a
smaller time scale (1 to 3 min) to delineate the peak of uptake required for quantification of PET data. Difference between transgenic and control
animals is significant for each study group visibly and analytically. For the young homozygous animal, it is seen in the lower color scale range. Cortex
in B2shows uptake towards blue and cyan. Same structures show lowest uptake in D2(magenta, corresponding to cerebellum). TACs confirm visual
perception: neocortex TAC in B4intersects cerebellum TAC and stays above it (neocortex-to-cerebellum ratio .1) while neocortex TAC in D4remains
below the cerebellum TAC (ratio ,1). PET color look-up-table is UCLA (Pmod) with lower thresholds set to still visualize the cerebellum. Arrowheads
indicate slice positions. Slice coordinates (corresponding to Paxinos atlas) are: horizontal Bregma 21.90 mm, coronal Bregma 20.10 mm and sagittal
0.65 mm lateral. Image scale is double size of reality. Further results for these animals are shown in Figure 7.
In Vivo PiB Imaging in AD Mice
PLoS ONE | www.plosone.org4 March 2012 | Volume 7 | Issue 3 | e31310
In Vivo PiB Imaging in AD Mice
PLoS ONE | www.plosone.org5March 2012 | Volume 7 | Issue 3 | e31310
corresponding to a low Ab plaque load in the olfactory bulb at
younger ages. The relative tracer uptake in old homozygous
animals was slightly higher than in young ones (p=0.027) but by
far not as distinct as for the telencephalon. This confirmed
previous reports  and is consistent with our observations of
smaller and fewer plaques in the olfactory bulb. This result was
also notable in the context of unspecific tracer binding. The
olfactory bulb reaches in between extracerebral structures with
high unspecific tracer uptake but was not a contributing region for
these high uptakes.
Relative [11C]PiB uptake of the remaining basal brain structures
(diencephalon and midbrain) was already significantly higher in
young transgenic animals than in controls (p=0.031)) (Table 2).
However, in old homozygous animals it was not significantly
higher to the young ones (p=0.318). In our experience and in
consistency with previous reports in the same or similar models
[31,46,47], the neuroanatomical structure with the major
contribution to uptake in this region was the thalamus. The
difference between young and old control groups was not
significant (p=0.114). For this region, it is notable that the
relative tracer uptake was not reversed in the control groups (.1).
The relative [11C]PiB uptake behavior is also shown graphically in
Extracerebral tracer retention in proximity to brain
We observed considerable [11C]PiB retention in regions of the
mouse head that appeared to be extracerebral, possibly around
nasal and eye cavities, but very close to the brain (Figure 1). To
distinguish specific PiB uptake in brain from probably unspecific
extracerebral uptake and for further validation of our PET
imaging and co-registration protocol, we performed variants of the
general in vivo and ex vivo experiments.
Sequential [11C]PiB/[18F]FDG PET for five old homozygous
and five control animals while keeping the animal in place
provided automatic overlay of both PET images and clear spatial
localization of the brain. Thus, it was possible to confirm that the
high frontal [11C]PiB retention was indeed located outside the
brain (Figure S2).
Additionally, the in vivo PET protocol was modified to a two-
step in vivo/ex vivo PET protocol, in which complete heads
without brains of four male tgtg-young and four ctl-young animals
were scanned from about 35 min to 65 min p.i. (Figure 3B). The
remaining [11C]PiB retention in exclusively extracerebral ana-
tomical structures clearly shows the same uptake pattern as seen in
PET in vivo.
To confirm the findings from in vivo and ex vivo PET, the 20
animals from the regional brain biodistribution study (Table 2)
were also used to measure [11C]PiB uptake in various cranial
organs (Figure 3C). The organs with the most prominent [11C]PiB
uptake were the harderian and parotid gland and the eyebulbs. In
general, tracer retention varied unsystematically between animals
of the same groups and no differences could be detected between
transgenic and control animals.
To further validate unspecific extracerebral tracer retention, ex
vivo [3H]PiB autoradiographs of a complete transgenic homozy-
gous mouse head showed the exact locations of unspecific tracer
retention in various anatomical structures very accurately
(Figure 3A). Exposition time needed to be shortened to achieve
good resolution of extracerebral tissues. For this reason, only few
plaques can be seen. The analogy of the unspecific extracerebral
uptake pattern in ex vivo [3H]PiB autoradiography and ex vivo
[11C]PiB PET of the head can be seen well.
As the olfactory bulb reaches in between the anatomical
structures that have been characterized with high unspecific tracer
uptake, it was included in [11C]PiB regional brain biodistribution,
[3H]PiB autoradiography, Thioflavin S and Ab40/42 histological
analyses whenever possible. In general, it showed smaller and
fewer plaques and lower uptake values, confirming that it was not
involved in the higher tracer retention regions around it.
The spectrum of results, above, validated the high unspecific
tracer uptake to be extracerebral. The proximity of frontal brain
parts to extracerebral anatomical structures, in particular present
within the eye cavities, confirmed the importance of very accurate
image co-registration for reliable PET analyses described above.
The principle and high quality of our image co-registration
method is presented in Figure S3.
Thioflavin S and Ab40/42 antibodies for plaque
Brain sections were stained with Thioflavin S (29 animals) and
with double immunofluorescence against Abx–40(anti-Ab40) and
Abx–42(anti-Ab42) (22 animals) for the histological quantification
Figure 2. Mouse Ab plaque pathology in vivo and ex vivo. PET binding potential maps for [11C]PiB and corresponding autoradiography and
fluorescence microscopy images of neighboring horizontal brain sections showing data from the same animals presented in Figure 1. Left brain
halves are shown. Frontal cortex is at top and cerebellum at bottom of each panel. (A) 23 month old female hemizygous APP/PS1 mouse, (B) 9 month
old female homozygous APP/PS1 mouse, (C) 21 month old female homozygous APP/PS1 mouse, (D) 23 month old female C57BL/6J control mouse.
Column (1): Binding potential maps for [11C]PiB (BPND, MRTM2) matched to MRI. Shown is the same horizontal level as in Figure 1 and S1. Color table
is UCLA (Pmod). Width of color scale represents 3 mm in reality. Column (2): Digital [3H]PiB ex vivo autoradiograph with optical image (gray) of a brain
section of the same animal, killed 1 hour p.i.. Color table is Red Hot (ImageJ). Column (3): Double immunofluorescence microscopy for Ab40 (green)
and Ab42 (red). Anatomical reference (gray) is provided by control channel (Cy3). Column (4): Thioflavin S fluorescence (FITC excitation, cyan).
Anatomical reference (gray) is provided by DAPI fluorescence. Right column: Identical Ab plaque constellations of adjacent sections (as marked by white
rectangle in columns (1) to (3). Top panel (5): magnification of digital autoradiograph as seen in column (2). Middle panel (6): corresponding magnified
viewofAb40/Ab42stainasseen incolumn(3).Bottompanel(7):correspondingmagnifiedviewofThioflavin Sstainasseenincolumn(4).Columns(2)to
(4) show directly neighboring 10 mm thick sections of the left brain half from bottom to top of skull at about 1.9 mm below Bregma. Width of zoom
panels in rightmost column represents 350 mm in reality. Complete orthogonal views for binding potential maps are shown in Figure S1.
Table 2. Regional brain biodistribution of [11C]PiB.
tgtg-young 1.2760.24 1.8760.581.7960.57
Mouse brain was dissected into four regions (olfactory system, telencephalon,
cerebellum and remaining brain structures) 30 min p.i.. Results show mean
[11C]PiB uptake ratios (6 SD) of the three target regions relative to cerebellum
(initially measured as %ID/g) for the homozygous study groups and both
control groups. Data are reported graphically in Figure 3 as reference to
extracerebral [11C]PiB distribution.
In Vivo PiB Imaging in AD Mice
PLoS ONE | www.plosone.org6March 2012 | Volume 7 | Issue 3 | e31310
In Vivo PiB Imaging in AD Mice
PLoS ONE | www.plosone.org7March 2012 | Volume 7 | Issue 3 | e31310
of Ab plaque load and plaque size distribution (Table 1) by
applying a computerized image analysis and object recognition
algorithm similar as reported, previously .
Thioflavin S has not been used with the applied semi-automated
method of Ab plaque quantification in this mouse model, before
. To validate Thioflavin S staining as a robust and comparably
easy method for Ab plaque quantification we first analyzed the
relation of Thioflavin S sensitivity to anti-Ab40/42 sensitivities for
Ab plaque detection in a pairwise manner. The correlations
between Thioflavin S quantification with each of the antibodies
and with their compound signal are presented in Figure S7. In
addition, Thioflavin S recognized a similar amount of relative Ab
plaque load as the anti-Ab40/42 compound signal (Figure 4A)
which seems to parallel the comparable affinities of PiB for both
Ab species . Furthermore, the anti-Ab40/42 compound result
(Figure 4A) reflects that the anti-Ab40 and anti-Ab42 measure-
ments are mostly co-localized.
Thioflavin S staining was excellent for analyzing tissues with Ab
deposits. However, it showed some unspecific binding in Ab-free
regions depending on the neuroanatomical location, in contrast to
the specific Ab antibodies. Among the regions we have measured,
the highest unspecific values were found in thalamus of control
animals. We expect this to be mostly related to the texture of the
regional brain tissue. It may also be method-related as exposition
and measurement parameters were kept identical for each stain
and were adjusted to measurements in the neocortex.
Ab plaque area and plaque size distribution
Binding of [11C]PiB to Ab loaded brain regions is probably
influenced both by total plaque volume as well as by individual
plaque size and type [35,48]. Hence, we measured relative plaque
areas and plaque size distributions using a large dataset for a
robust analysis of neocortex with Thioflavin S (643 observations)
and Ab40/42 antibodies (158 observations) (Figure 4 and Figure
The results from histological Ab plaque quantification in
neocortex also were consistent with in vivo PET results for the
same region. Highest values were observed for the tgtg-old group,
followed by tg-old and tgtg-young (see Figure S10 for pairwise
The results of plaque load as measured by Thioflavin S are
reported, here. Their values in target regions with Ab deposits
were representative for the analysis with Ab40/42 antibodies as
shown above (Figure S7). Plaque load based on Thioflavin S
binding in neocortex was measured for tg-old as 7.7261.03%
(range: 6.27–9.07%), for tgtg-young as 4.6860.70% (range: 3.53–
5.37%), for tgtg-old as 11.7861.63% (range: 9.88–13.31%) and
for ctl-old as 0.6160.17% (range: 0.34–0.79%). Differences
between groups were tested corresponding to the staging of Ab
load in these groups (ctl-old against tgtg-young, tgtg-young against
tg-old and tg-old against tgtg-old). The difference between all
groups was highly significant (p,0.001). For all relative plaque
load observations, no differences were found between right and left
brain sides in all animals and between old and young control
Plaque sizes considerably increased with age in hemizygous and
homozygous animals (Figure 4B). After histogramming the
individual plaque sizes for each transgenic group, and estimating
kernel density functions, we could calculate that the differences
between plaque size distributions of all transgenic study groups
were highly significant (p,0.001). While relative plaque areas and
size compositions were significantly different between the old study
groups, the size composition of plaques in neocortex at old ages
appeared to have a similarity independent of genotype.
Cerebellum as reference region
Willuweit et al. reported that the cerebellum seems to be free of
plaques in this animal model . For our studies, we used the
cerebellum as a reference region in PET, biodistribution and
autoradiography. Therefore, we applied the histological quantifi-
cation of Ab plaques to the cerebellum, also, to analyze whether it
qualifies as a reference region for imaging purposes in this animal
model. For this, the same large dataset as above was used for the
Thioflavin S (599 observations) and Ab40/42 antibody (148
The Ab42 antibody was providing for the fluorescent channel
with the highest signal-to-noise ratio and was therefore the most
reliable signal for the analysis of a potentially target-free region.
The quantification results with the Ab42 antibody were always
lower than 0.09% compared to neocortex of tgtg-young animals
which was larger than 3.09%. The average cerebellar binding of
the Ab42 antibody per group was 0.0260.01% (range: 0.01–0.03)
for tg-old, 0.0160.00% (range: 0.01–0.02) for tgtg-young,
0.0460.03% (range: 0.01–0.09) for tgtg-old and 0.0260.01%
(range: 0.01–0.02) for ctl-old. No significant differences could be
seen between the groups: ctl-old to tgtg-young (p=0.59), tgtg-
young to tg-old (p=0.25) and tgtg-young to tgtg-old (p=0.17).
Thioflavin S showed unspecific binding behavior in tissues
without Ab deposits (Figure 2D) as described above. Nevertheless,
the highest unspecific results in cerebellum were far below the
lowest specific results in neocortex (0.82 vs 3.53%). The average
cerebellar binding of Thioflavin S per group was 0.3660.05%
(range: 0.32–0.44) for tg-old, 0.2560.06% (range: 0.16–0.33) for
tgtg-young, 0.3060.04% (range: 0.25–0.36) for tgtg-old and
0.6660.13% (range: 0.45–0.82) for ctl-old. Differences between
groups were significant, here, but it was the control groups that
showed slightly higher binding than the transgenic animals.
These results for the cerebellum quantitatively confirmed that
this region can be used as a reference region.
Ab40 and Ab42 protein levels (ELISA)
Detailed differential Ab protein analyses were performed in this
animal model, previously, and tight correlations with relative Ab
plaque load were shown .
Figure 3. Extracerebral tracer retention. High [11C]PiB uptake in regions frontal to the brain were accurately validated to be extracerebral. (A)
Cranial [3H]PiB ex vivo autoradiography. 15 mm thick section of a complete mouse head showing exact anatomical locations of unspecific tracer
retention (male tgtg, 16 month old). Exposition time needed to be shortened to achieve good resolution of extracerebral tissues. For this reason, only
few plaques can be seen in the brain. Color table: Red Hot (ImageJ) (B) CNS removal during [11C]PiB PET. 9 month old male homozygous APP/PS1
mouse was scanned in vivo for 30 min before the complete brain was extracted and scanned for further 30 min together with the skull. The skull of
the ex vivo [11C]PiB PET scan is co-registered to a cranial CT for better orientation and shown on six horizontal slices which are 1 mm apart (top left
horizontal level at about 21.9 mm Bregma in correspondence to all other figures). Both parotid glands can be seen on bottom section. Color table is
UCLA (Pmod) (C) Ex vivo biodistribution of [11C]PiB relative to cerebellar uptake in (extracerebral) glandular tissues and eyebulbs in both homozygous
and both control study groups. Cerebral biodistribution data from the same animals as presented in Table 2 is included graphically as reference. Data
show that olfactory bulb does not contribute to high surrounding uptake in harderian glands and eyebulbs. Column heights represent means, error
bars represent standard deviation.
In Vivo PiB Imaging in AD Mice
PLoS ONE | www.plosone.org8March 2012 | Volume 7 | Issue 3 | e31310
To further validate our PET imaging results and to understand
how our in vitro tracer binding results with human and mouse
brain homogenates (see below) relate to Ab protein levels, brain
tissue of 14 animals that received a PET scan, brain tissue of 8
animals and three samples of post-mortem human brain tissue for
radioligand binding assay were biochemically quantified for
human Abx–40 and Abx–42 protein (Table 1). Differential
extraction procedures were applied in order to determine the
levels of either soluble or insoluble forms of Ab species for all
In general, our results confirm the previous report .
Detailed individual results of ELISA analyses are shown in
Table 3 and Table 4 for the corresponding experiments and are
described there. The individual results and correlation of insoluble
Ab protein levels to results from PET imaging, autoradiography
and histological plaque quantification are shown in Figure S10.
In vitro [3H]PiB binding assay
In vitro tracer binding to brain homogenates provides a
sensitive reference to the other experimental modalities .
Therefore, seven samples of brain tissue were used for assessing in
vitro [3H]PiB binding to mouse brain tissue and postmortem
human brain tissue at distinct disease stages (mouse: tg-old, tgtg-
young, tgtg-old; human: Cerad-C/Braak V and Cerad-0/Braak
Figure 4. Histological Ab plaque burden and plaque size distribution in neocortex. Ab plaque burden and size of individual plaques were
analyzed on histological sections stained with Thioflavin S and double immunofluorescence against Ab40 and Ab42 by applying a semi-automatic
imaging algorithm. All animals were analyzed in PET, before. Shown here, are the results for neocortex of the transgenic study groups: tg-old (orange),
tgtg-young (yellow) and tgtg-old (red). (A) Ab plaque burden of each transgenic group as measured by Thioflavin S, compound anti-Ab40/42, anti-
Ab42 and anti-Ab40. Compound anti-Ab40/42 result shows co-localization of both Ab species. (B) Plaque size distribution in each transgenic study
group. Here, the anti-Ab42 signal was used for its highest signal-to-noise. Its strong association with the Thioflavin S signal is shown in Figure S7.
In Vivo PiB Imaging in AD Mice
PLoS ONE | www.plosone.org9 March 2012 | Volume 7 | Issue 3 | e31310
II). Binding to postmortem human brain tissue was performed to
provide a reference for the results from mouse brain tissue. A
binding assay to synthetic Ab1-40fibrils was also performed and
used as positive control (Figure S8). The huAD-C (Cerad-C/Braak
V) tissue sample was retested twice (Figure S9). All brain tissues
used for in vitro binding were also analyzed for Ab protein levels
with ELISA (Table 3).
Some tissues were lacking sufficient [3H]PiB binding saturation
behavior (Figure 5). Hence, in vitro binding potential (BP) was
chosen as the target value for tissue comparisons after global
nonlinear regression with the single binding site model to total and
nonspecific binding data (Table 3). As Kdand Bmaxare correlated,
BP can be measured very accurately either as ratio from the
estimates, as done here, or from the initial slopes to the specific
binding curves  even though independent estimates for Kdand
Bmaxmay not reasonably be possible.
The severely affected human AD tissue
(huAD-C) homogenate provided an estimate for BP of 38.264.7
Figure 5 shows how the huAD-C tissue homogenate reached
saturation binding at comparably low concentrations of the tracer
(around 8 nM).
While the huAD-C sample yielded binding values correspond-
ing to previous reports for severely affected human AD brain tissue
[8,50,51], the total binding data of the mildly affected human AD
sample (huAD-0) and the human control brain (huCTL) was
showing no difference to nonspecific binding data (data not shown)
indicating that a binding component which could compete with
3 mM PiB was not present (BP estimates close to 0). This
corresponded to the Ab protein quantification results (Table 3) in
spite of repeated positive neuropathological staging according to
BrainNet Europe standards.
In ELISA, human Ab-free matched control tissue was included
to provide reference values. The huAD-0 sample showed soluble
and insoluble Abx–40levels comparable to control tissue, while
Abx–42 levels were increased and at about one third of the
huAD-C sample for the insoluble fraction. The Abx–42levels of
huAD-C were more than fourfold to the Abx–40levels of the same
The initial steepness of the binding curves for
mouse brain tissue homogenates was comparable to and even
higher (tgtg-old) than for the severely affected human AD tissue
(Figure 5). This binding behavior at low tracer concentrations is
considered a prerequisite for successful PET imaging and confirms
our positive PET imaging outcomes described above.
BP in the transgenic mouse brain tissues were estimated to
11.960.9 (tg-old), 14.360.9 (tgtg-young) and 69.262.7 (tgtg-old).
The BP of homozygous old mouse brain tissue was clearly above
the value for severely affected human AD tissue while the tg-old
and tgtg-young samples were at about one third of the result for
huAD-C. The total binding data of mouse control brain (msCTL),
like the huAD-0 and huCTL samples was not different to
nonspecific binding (data not shown) indicating that a binding
component which could compete with 3 mM PiB was not present
(BP estimates close to 0).
Independent estimates for Kd and Bmax of the high-affinity
component were yielded for the tgtg-old tissue as it reached a
sufficient degree of tracer saturation binding. Fitting these data to
Bmax=200612 fmol/mg) for[3H]PiB.
Table 3. In vitro binding potential and Ab40/42 protein
study groupsoluble protein
huAD-0 0.81.6 15.8121.90
tg-old517.8186.2 169915.9 145010.5 11.9
tgtg-young450.5 304.8172152.3 178821.114.3
tgtg-old 912.4485.4657818.2 572182.5 69.2
In vitro binding potential (BP) as yielded with [3H]PiB radioligand saturation
binding assay and corresponding soluble and insoluble Abx–40and Abx–42
protein fractions (picogram protein per milligram wet tissue) for the same
human and mouse tissue samples. Binding curves of the severe human AD and
all transgenic mouse brain samples are shown in Figure 5.
Table 4. Multi-modal combined experiment with [3H]PiB/[11C]PiB cocktail.
modality AD1*AD2CTL1* CTL2
PET [BPNDneocortex] 0.06 0.06
Biodistribution [11C]PiB [region-to-cerebellum ratio]olfactory system 1.001.071.05 0.98
diencephalon and midbrain 1.261.47 1.22 1.10
Autoradiography [neocortex-to-cerebellum ratio][11C]PiB 1.882.1 0.72 0.95
Histology [% plaque area neocortex]Thioflavin S 4.004.10 0.54 0.41
5.88 6.140.02 0.01
Ab protein levels (forebrain) [pg protein/mg tissue wet weight] solubleAbx–40
An all-in-one experiment was performed for four animals of the young study groups (2 tgtg-young (AD1 and AD2) and 2 ctl-young (CTL1 and CTL2)) to retrieve a large
spectrum of multi-modal information from a single animal. Asterisk (*) marks animals that are shown in Figure 6.
In Vivo PiB Imaging in AD Mice
PLoS ONE | www.plosone.org 10March 2012 | Volume 7 | Issue 3 | e31310
the two site binding model revealed a Kdof 5.764.1 nM and a
Bmaxof 289.36123.9 fmol/mg while the huAD-C data provided a
Kdof 3.461.4 nM and a Bmaxof 156.767.3 fmol/mg as estimates
for the same fit.
Scatchard graphs corresponding to the specific binding curves
are displayed additionally in Figure 5, together with a semiloga-
rithmic representation of the specific binding curves. The
semilogarithmic plot shows the absence of infliction points in the
Figure 5. In vitro binding assay with [3H]PiB. Specific [3H]PiB binding to mouse and human brain homogenates and nonlinear modeling of data.
(A) Binding isotherms for [3H]PiB with transgenic mouse brain tissues and human AD tissue containing Ab deposits. Solid black curves show
nonlinear fits with a single site model. Dashed lines describe 95% confidence bands around the fit. Resulting in vitro binding potential (BP) values and
Ab protein levels of these tissues are described in Table 3. (B) Semilogarithmic representation of the specific binding data as seen in panel (A) to
delineate possible infliction points. (C) Scatchard graphs showing the same data as panel (A). Each data point is derived from the mean value of the
original data octuples. Data show representative samples from tgtg-old (red), tgtg-young (yellow) and tg-old (orange) transgenic study group and
severely affected human AD (huAD-C) tissue (green).
In Vivo PiB Imaging in AD Mice
PLoS ONE | www.plosone.org 11 March 2012 | Volume 7 | Issue 3 | e31310
data of tg-old and tgtg-young which would be necessary for the
distinction of different binding sites and for the ability to fit one or
two lines (one- or two-site model) to the Scatchard data .
In accordance to previous reports , the Ab levels of mouse
brain tissue were higher than those found in human AD brain by a
factor of around 1000 (Table 3). Furthermore, the mouse Ab levels
also seem to indicate a correspondence to measured BP values.
Combined multi-modal experiment
To bring together a large set of experimental modalities applied
to an individual animal, and to cross-validate and address their
relationship within a single animal, a combined experiment was
performed (Figure 6 and Table 4) consisting of PET, regional
brain biodistribution, dual-label digital autoradiography, histolog-
ical Ab plaque quantification with Thioflavin S and anti-Ab40/42
and Ab40/42 ELISA.
Four animals from the young study groups (2 tgtg-young, 2 ctl-
young) were given a bolus cocktail of [11C]PiB/[3H]PiB in the
PET scanner and their brain tissue processed immediately after
PET imaging. The individual results for all four animals are shown
in Table 4. The good correspondence of ex vivo [11C]PiB and ex
vivo [3H]PiB autoradiography together with double anti-Ab40/42
fluorescent stains of a neighboring section are shown in Figure 6.
This combination experiment showed how excellent the results
from different experimental modalities correlate on the level of
individual animals and how activity uptake in PET is real tracer
uptake. Also, these individual results confirmed that young
homozygous animals could clearly be distinguished from control
animals in all modalities. Furthermore, it showed the consistency
Figure 6. Multi-modal combined experiment. Single multi-modal in vivo/ex vivo combination experiment with 4 animals from the young study
groups (2 tgtg-young and 2 ctl-young) showing the whole spectrum of results on an individual level. After a bolus injection of a [11C]PiB/[3H]PiB
cocktail, the animals passed a 30 min CT/PET scan, were then killed for immediate [11C]PiB regional brain biodistribution and dual-label digital
autoradiography. Brain halves used for biodistribution were analyzed for Ab protein levels. The other brain halves were stained with Thioflavin S and
anti-Ab40/42 and used for histological plaque quantification. Columns (1) to (3): 9 month old female homozygous APP/PS1 mouse (‘‘AD1’’) and
Column (4) to (6): 9 month old female C57BL/6J control mouse (‘‘CTL1’’), presented in a mirror fashion. Ex vivo [11C]PiB (red)/[3H]PiB (green) dual-label
digital autoradiographs with underlying optical scans of horizontal 12 mm half brain sections of AD1 (right brain) (A and B) and CTL1 (left brain) (D
and E) (marked with asterisk (*) in Table 3) and corresponding magnified views of double immunofluorescence stains for Ab40 (green) and Ab42 (red)
of neighboring sections for the same region (C and F). All four modalities are shown individually (outer two columns) and co-localized (central
columns). Limits of green and red color look-up-tables represent minimum and maximum of measured signal. The analytical results of all experiments
are shown in Table 3 below this figure.
In Vivo PiB Imaging in AD Mice
PLoS ONE | www.plosone.org 12 March 2012 | Volume 7 | Issue 3 | e31310
and robustness of the groupwise results described above on an
individual analysis level.
Relationship of in vivo PET to other experimental
The study was designed to provide as many validation
experiments to PET imaging in every animal as possible, in order
to analyze the relationship of in vivo radioligand binding in PET
to relative ex vivo tracer uptake in brain biodistribution and
autoradiography, to histological Ab plaque burden and to Ab
protein levels (Table 1).
Neocortex, i.e. complete cortex without hippocampus, was used
as the primary target region. It was defined in the same way on
horizontal PET slices (Figure S4) and horizontal autoradiographi-
cal and histological sections. The brain region that was used for
ELISA analysis was defined as previously reported  and hence
contained telencephalon and the largest part of diencephalon and
midbrain without the olfactory system. This presumably provided
for slightly weaker correlations of Ab protein levels with the other
Figure 7 summarizes these relationships for the animals of the
three transgenic study groups and the old control group. In
general, [11C]PiB binding in PET was correlating strongly with ex
vivo tracer uptake and in vitro Ab load. In addition, each study
group was clearly separate from each other in all experimental
modalities (shown by different color for each study group). The
summary of data demonstrates the robustness of the small-animal
PET results and their consistency with the validation experiments.
A complete overview of the cross-validation approach is shown in
The need for more specific imaging markers of Ab, tau and
alpha-synuclein  requires robust and feasible translational
imaging tools to enable the evaluation and ranking of new tracers.
Previous PET imaging studies with [11C]PiB in different AD
mouse models were not successful, despite high Ab plaque loads
[33,38,39] and the only successful study relied on very high
specific activities of the tracer . This led researchers to
principally question the feasibility and potential of this imaging
method for translational AD research [32,33,40,54].
For these reasons, we addressed the development of a feasible,
reproducible and robust preclinical Ab plaque PET imaging setup
in transgenic AD mice that reliably detects a specific signal in
animals young enough to allow for longitudinal follow-up studies.
We were able to show that the measured uptake of [11C]PiB with
PET in individual transgenic animals at different disease stages
was robust and strongly correlated with several independent
experimental methods in the same animals.
The old hemizygous AD group was selected to correspond to
animal ages used in previous imaging studies [33,38,41]. In
general, the results of this group tended to be relatively close to the
young homozygous animals. These results show the value of the
homozygous animals of this APP/PS1 mouse model for imaging:
the reliable specific PET signal in young animals in combination
with a virtually normal life span and low premature death rate of
homozygous mice allows for sufficiently long follow-up studies.
The PET results in control animals are notable as they indicate
a volume of distribution ratio ,1 (cerebellum as reference) and
hence a larger volume of distribution for the cerebellum than for
neocortex. This finding is consistent with Maeda et al.  and is
probably related to white matter binding of the tracer in the
cerebellum in contrast to the target region which does not contain
white matter. It is further supported by our regional brain
biodistribution results which also yielded ratios ,1 for telenceph-
alon-to-cerebellum ratios of injected tracer doses normalized to
The considerable amount of unspecific PiB retention in tissues
outside of the brain (like the salivary and harderian glands) is likely
to be model-independent. Previous studies [33,38,41] do not
provide information on whether the applied PET technologies
have been able to resolve the uptake in extracerebral regions
neighboring the olfactory bulb and the frontal cortex. In our study,
we have identified these issues as a potential error source and it
highlights the importance of precise PET image co-registration to
MRI such that volumes-of-interest can be defined reliably. In our
experience, manual image co-registration of well pre-processed
small-animal data yields excellent and very reliable results. A
similar manual method has been reported by Pfluger et al. for
human MRI-SPECT data .
In a different APP/PS1 model, Klunk et al. detected an uptake
of [11C]PiB of 100–120% in the entire cerebrum relative to PS1
mice . Their results were not statistically significant, which
may have been due to the small sample size (1 transgenic versus 1
control animal per age group) and the global VOI-based approach
employed (large VOI encompassing the entire brain, no reference
region). In another study, Toyama et al. included a reference
tissue-based analysis in their study with six Tg2576 mice at a mean
age of 22 months, using the cerebellum as a reference region .
Although they calculated significantly higher binding ratios in the
transgenic mice, Toyama et al. concluded that their study could
not prove specific binding of [11C]PiB to Ab plaques due to the
overall small difference in absolute tracer uptake between
transgenic and control animals which may have been due to the
presence of Ab plaques in the cerebellum of this animal model
. The sole report on successful in vivo [11C]PiB imaging with
PET in a single transgenic mouse model (APP23) claimed
extraordinarily high specific activities of their PiB preparation
(max. 291 GBq/mmol) to be required for imaging of Ab plaques in
their animals . Specific activities of this magnitude are not
obtainable at most PET centers which may explain why these
results have not been reproduced by others. Furthermore, the
proportionality between tracer uptake and Ab plaque load as
derived from a small-animal PET study may not be transferable to
humans, if 10 to 20-fold higher specific activities are applied in the
animal model. Another relevant aspect of the study of Maeda et al.
is that reasonable tracer uptake has been found in animals .21
months of age, despite the high specific activity preparations. This
further limits the applicability of this imaging protocol for follow-
up studies as the average life span of APP23 mice is around two
In our study, significant tracer uptake in regions with Ab
plaques was demonstrated in transgenic mice as young as 9
months injected with 28.967.93 MBq of [11C]PiB in a specific
activity of 11 GBq/mmol. Hence, our specific activities were about
20-fold lower and better comparable to that routinely applied in
studies of AD patients (range: 11.1–14.8 GBq/mmol). An even
higher molar amount of [3H]PiB (2.5 nmol) was used for the ex
vivo autoradiography studies compared to our in vivo PET studies
(1.6 nmol of PiB) which both demonstrated specific binding of the
tracer to Ab plaques. Several factors may be responsible for this
discrepancy of our findings compared to previous work. The
choice of animal model may be a key explanation.
In contrast to the abundance of available transgenic AD models
with high content of cortical Ab plaques , good preclinical
models for imaging Ab deposits have still been lacking [32,40].
However, various ex vivo analyses with Ab ligands in AD mice
In Vivo PiB Imaging in AD Mice
PLoS ONE | www.plosone.org 13 March 2012 | Volume 7 | Issue 3 | e31310
[31,33,58,59] and in vivo fluorescent labeling [36,60] would
suggest that the in vivo measurement of Ab plaque load in mice
with PET should be possible. In humans, negative PiB imaging in
severely amyloid-positive patients with the arctic APP mutation
has been observed  and may indicate a parallel phenomenon
to negative PiB-PET results in animal models. In which way
different Ab isoform patterns  and their degree of fibrillarity
 contribute to PET imaging results remains to be examined.
A number of previously reported favorable characteristics of the
APP/PS1 mouse model employed in our study  probably
contribute to the observed positive findings for several reasons.
The potential advantages are: a) an early-onset and rapid
progression of plaque load, b) plaques showing similar morphology
to those in human AD, c) low inter-animal variability and no
gender effects (in contrast to other transgenic animal models which
show high variability of Ab plaque expression ), d) co-inherited
transgenes and a C57BL/6 background leading to good breeding
capabilities of a homozygous line and a low rate of premature
death of hemizygous and homozygous mice up to normal old age
. The homozygous animals show earlier onset and more rapid
progression of Ab plaque deposition compared to hemizygous
animals and are therefore good candidates for at least one and a
half years of longitudinal imaging.
While some other models develop Ab deposits in the cerebellum
over time [24,30], we could show that the cerebellum of our model
stays free until old age and can therefore be used for reference
tissue approaches. This is an important feasibility advantage as
alternative methods for analysis require arterial input information
and calibration to injected dose both of which remain method-
The plaque quantification results in our study differ somewhat
from what has been reported, previously . Willuweit et al.
measured relative plaque burden in 19 to 20 months old
transgenic mice of 10.5% (hemizygous) and 35.2% (homozygous).
Here, the measured relative plaque burden in even older animals
were lower and the results lay closer together (around 5% and
12%). There may be several methodological reasons for this.
Firstly, we have used frozen brain material for this study while
Willuweit et al. took paraffin sections. Secondly, we have used
other primary antibodies for Ab40 and Ab42 detection with a
different staining protocol. Thirdly, the parameters for the
automated plaque detection algorithm needed to be adjusted to
histological material, stain and exposure times.
Thioflavin S is an easy to use staining agent and was applied for
histological Ab plaque quantification, before . Here, we
applied our semi-automatic imaging algorithm  to Thioflavin
Figure 7. Relationship of in vivo PET to other experimental
modalities. Association of in vivo [11C]PiB binding potential in mouse
neocortex with relative neocortical [3H]PiB uptake in autoradiography
(A), with relative neocortical Ab plaque burden as stained by Thioflavin
S (B) and with insoluble Ab40 and Ab42 protein levels in forebrain (C
and D). Data across the modalities was acquired from tissue of the same
animals (as shown in Table 1). Individual animals are identified by their
unique number code within their study group. The coloring of study
groups in the scatter plots shows how each group is fully separated
from each other. Color code: tg-old (orange), tgtg-young (yellow), tgtg-
old (red) and ctl-old (blue). Pairwise correlation coefficients (r) for each
pair of modalities are noted in each scatter plot. Histological
quantification with Thioflavin S is used representatively for all
histological quantification results because of its tight correlation with
anti-Ab40/42 as described in Figure S7. Here, the animals presented in
Figures 1 and 2 are coded with #5 (tg-old), #5 (tgtg-young), #1 (tg-
old) and #1 (ctl-old). The full scatter matrix for the cross-validation of
experimental results is shown in Figure S10.
In Vivo PiB Imaging in AD Mice
PLoS ONE | www.plosone.org 14March 2012 | Volume 7 | Issue 3 | e31310
S stained sections for the first time and validated this approach in
relation to Ab40 and Ab42 antibodies.
The dual-label autoradiographs (Figure 6) show different
nonspecific binding of [11C]PiB and [3H]PiB to mouse brain tissue.
While [11C]PiB was also taken up by white matter (seen best in
panel D2), nonspecific [3H]PiB binding was mostly observed in
vessels (seen best in the choroid plexus). Both versions of PiB (i.e.
labeled with either
plaques. The specific activities of both versions of PiB were the same
as they were applied in a cocktail bolus. However, the measurement
of uptake ratios relative to cerebellum in autoradiography is less
reliable with [11C]PiB, as slight deviations in thickness within a
section has direct influence on the ratios. This effect is negligible
when using [3H]PiB ratios due to less energy of tritium.
In vitro PiB binding was first studied in transgenic mice by
Klunk et al. who found only a very small high-affinity component
for [3H]PiB in very old hemizygous APP/PS1 animals (BP=78)
 using classical Scatchard analysis. The authors reasoned that
the low concentration of high-affinity binding sites compared to
humans (BP=636) might be the reason for unsuccessful PET
imaging in mice.
Our results agree with some aspects of what has been reported
by Klunk et al. (e.g. different binding kinetics for rodent AD tissue
and correlation of Bmaxwith insoluble Ab), while it deviates in
other points (e.g. existence of high-affinity binding component and
comparable BPs). Using a state-of-the-art global nonlinear
regression approach for analyzing total and nonspecific binding
data for a larger range of tracer concentrations, we revealed a
higher BP (=12) in young hemizygous animals relative to severely
affected human AD tissue (=38). Furthermore, the BP (=69) of
our old homozygous mouse brain sample was nearly double
compared to human.
The slow tracer saturation in AD mouse brain indicates a
different binding behavior to rodent Ab plaques in the presence of
considerably higher levels of Ab in mouse than in human . It
may be difficult to identify a high-affinity component in
hemizygous old mouse brain tissue. Homozygous old mouse brain
tissue, however, provides a definite high-affinity binding compo-
nent comparable to human AD tissue.
A direct correlation between Bmaxand insoluble Ab content was
reported by Klunk et al.. In our study, an association between BP
and insoluble Ab in the transgenic mice was observed. As one may
assume that the affinity of [3H]PiB to the binding sites of various
APP/PS1 mouse brain tissues of the same genetic strain is similar
and that, hence, BP is mostly related to Bmax  these
observations seem to be similar.
In summary, wehaveprovideda cross-validatedstudyforfeasible
small-animal PET imaging of Ab plaque deposition with [11C]PiB
in an APP/PS1 mouse model of Alzheimer’s disease. In vivo PET
imagingresults of three different transgenic mouse study groups and
matched control groups were validated with ex vivo and in vitro
methods. The transgenic study groups represented different disease
stages according to their Ab pathology and could well be
distinguished with PET. Group results were consistent in all
experimental modalities and individual results correlated tightly.
Thereported PETimaging protocoluses readilyachievablelevelsof
specific activity of the tracer and grants successful high-contrast
imaging down to ages of at least nine months. This provides the
opportunity for at least one and a half years of longitudinal studies
and, hence, truly translational Ab plaque imaging of Alzheimer’s
disease in a preclinical model. The established imaging setup and
multi-modal cross-validation protocol are applied to our tracer
development program for ranking and successful evaluation of new
imaging markers for Ab [18,19].
11C) had a similar specificity to Ab
Materials and Methods
The experiments were carried out with the approval of the
institutional animal care committee (Regierung von Oberbayern,
Munich, Germany) and in accordance with the German Animal
Welfare Act (Deutsches Tierschutzgesetz). Animal husbandry
followed the regulations of European Union (EU) guideline
All experiments were performed in hemizygous (tg) and
GmbH, Cologne, Germany) on a congenic C57BL/6J genetic
background and commercially available age- and gender-matched
controls (Harlan-Winkelmann, Borchen, Germany and Janvier, Le
Genest-St-Isle, France). The transgenic mouse model has been
characterized regarding onset, progression, distribution and extent
of Ab plaque deposition as well as behavioral features .
The animals were kept under temperature-controlled environ-
mental conditions (18–20uC, 50–60% relative humidity) on a
12:12 light-dark cycle (light from 6 am to 6 pm) and fed a standard
diet (Altromin 1326 mouse pellets, Altromin, Lage, Germany) with
free access to food and potable water until the start of the
experiments and after (no fasting). They were group-housed
(maximum of 5 individuals per group) in individually ventilated
type III cages (Ehret, Emmendingen, Germany) with dust-reduced
wood shavings as bedding. All animals underwent a minimum of
10 days acclimatization period.
Altogether 70 animals in five study groups were used in this
study such that groupwise and pairwise comparisons are possible.
Group age definition of animals was chosen to be ‘‘young’’ (9
months) and ‘‘old’’ (21 and 23 months). Three transgenic study
groups of hemizygous (tg-old) and homozygous (tgtg-young, tgtg-
old) animals were included to provide comparability with previous
reports and also to show how the imaging outcome can be
improved by using homozygous animals. In previous pilot studies,
young animals from seven to ten months were tested (unpublished
data). These preliminary results revealed reliable and satisfactory
Ab plaque detection and visualization with PET and we found an
age of nine months to be a feasible age definition for the young
study group. The two control study groups (ctl-young and ctl-old)
were designed to match gender and age and to additionally control
for any differences among the controls regarding gender (female
and male subgroups). Regarding body weight, female transgenic
animals tend to weigh less than female controls and female
controls weigh less than male controls. To our experience the
unavoidable differences in weight have no detectable influence on
the results presented, here.
Table 1 shows a detailed description of the study collective and
the combination of experiments performed for each subgroup.
Postmortem human brain tissue
Three samples of deep frozen human brain tissue were provided
by Neurobiobank Munich upon request to BrainNet Europe
(www.brainnet-europe.org) after approval of the ethics committee
at Technische Universita ¨t Mu ¨nchen. Neuropathological diagnos-
tics were performed according to BrainNet Europe standards.
All samples were taken from temporal cortex gray matter of
three female donors who died at an age of 79 to 85. Significantly
different amyloid-beta plaque load (as confirmed with 4G8
antibody stain) was a major selection criterium. Hence, one severe
AD brain staged as Cerad-C, Braak V (‘‘huAD-C’’), one mild AD
In Vivo PiB Imaging in AD Mice
PLoS ONE | www.plosone.org 15March 2012 | Volume 7 | Issue 3 | e31310