1. Journal of Pathology
J Pathol 2009; 219: 41–51
Published online 8 April 2009 in Wiley InterScience
(www.interscience.wiley.com) DOI: 10.1002/path.2565
Original Paper
Morphological alterations to neurons of the amygdala
and impaired fear conditioning in a transgenic mouse
model of Alzheimer’s disease
Shira Knafo,1 * † Cesar Venero,2† Paula Merino-Serrais,1 Isabel Fernaud-Espinosa,1 Juncal Gonzalez-Soriano,3
Isidro Ferrer,4 Gabriel Santpere4 and Javier DeFelipe1 *
1 Instituto Cajal (CSIC), Madrid, Spain
2 Department of Psychobiology, Universidad Nacional de Educaci´ n a Distancia, Madrid, Spain
o
3 Department of Anatomy, Faculty of Veterinary Medicine, Complutense University, Madrid, Spain
4 Institut Neuropatolog´a, IDIBELL-Hospital Universitari de Bellvitge, Universitat de Barcelona, Hospitalet
ı de LLobregat, Barcelona, Spain
*Correspondence to: Abstract
Shira Knafo or Javier DeFelipe,
Instituto Cajal (CSIC), Patients with Alzheimer’s disease (AD) suffer from impaired memory and emotional
Madrid, Spain. disturbances, the pathogenesis of which is not entirely clear. In APP/PS1 transgenic mice,
E-mail: defelipe@cajal.csic.es; a model of AD in which amyloid β (Aβ) accumulates in the brain, we have examined
sknafo@cbm.uam.es neurons in the lateral nucleus of the amygdala (LA), a brain region crucial to establish
† These
cued fear conditioning. We found that although there was no neuronal loss in this region
authors contributed and Aβ plaques only occupy less than 1% of its volume, these mice froze for shorter times
equally to this work.
after auditory fear conditioning when compared to their non-transgenic littermates. We
The authors have no conflicts of performed a three-dimensional analysis of projection neurons and of thousands of dendritic
interest to disclose. spines in the LA. We found changes in dendritic tree morphology and a substantial decrease
in the frequency of large spines in plaque-free neurons of APP/PS1 mice. We suggest that
these morphological changes in the neurons of the LA may contribute to the impaired
auditory fear conditioning seen in this AD model.
Copyright  2009 Pathological Society of Great Britain and Ireland. Published by John
Received: 8 January 2009 Wiley & Sons, Ltd.
Revised: 25 March 2009
Keywords: Alzheimer’s disease; unbiased stereology; morphology; confocal microscopy;
Accepted: 27 March 2009
APP; PS1; amyloid; plaques; cognition; dementia; learning; dendritic spines
Introduction For example, the lateral nucleus of the amygdala (LA)
is a key site of plasticity that underlies fear learn-
Alzheimer’s disease (AD) is a progressive neurode- ing [7,8]. The LA receives sensory input from the
generative disease that causes dementia and emotional thalamus and cerebral cortex, and it generates emo-
disturbances [1]. AD is neuropathologically charac- tional responses by activating different subcortical
terized by the accumulation of extracellular fibril- regions [9]. The outputs of the LA arise from pro-
lar amyloid beta peptide (Aβ) in amyloid plaques jection neurons [10], which were the subject of this
(plaques) and of intraneuronal neurofibrillary tangles study.
consisting of aggregated hyperphosphorylated tau, and In mice, the expression of proteins that are impli-
by elevated brain levels of soluble Aβ oligomers. cated in familial AD — a chimeric mouse/human
Plaques and neurofibrillary tangles are distributed in amyloid precursor protein (Mo/HuAPP695swe) and a
the hippocampus, neocortex, and in subcortical regions mutant human presenilin 1 (PS1-dE9) — leads to the
such as the amygdala, nucleus basalis, thalamus, locus early appearance of amyloid plaques [11]. We used
coeruleus, and raphe nuclei [2]. The amygdala plays these double transgenic (APP/PS1) mice to investi-
a major role in the processing and memorizing of gate the effects of Aβ overproduction and deposi-
emotional reactions [3]. The amygdala of AD patients tion on auditory fear conditioning, and to examine
undergoes significant shrinkage, distortion, and loss whether Aβ deposition provokes neuronal loss and
of neurons, as well as extensive gliosis [4,5]. More- changes in the morphology of projection neurons in
over, the extent of amygdaloid atrophy correlates pos- the LA. Previous studies performed in transgenic mice
itively with the degree of emotional memory impair- carrying a single APP transgene, or different com-
ment [6]. The different nuclei of the amygdala have binations of the APP and PS1 transgenes, reported
unique connections and they fulfil specific functions. contrasting results after training in the auditory cued
Copyright  2009 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.
www.pathsoc.org.uk

2. 42 S Knafo et al
fear conditioning paradigm [12–14]. Hence, we eval- conditioning was performed by pairing a tone to a
uated whether cued fear memory was impaired in foot shock and evaluating freezing during tone pre-
12- to 14-month-old APP/PS1 mice. We show that sentation. Pain threshold was evaluated by applying
auditory fear conditioning is severely dampened in an electric current at increasing intensities and deter-
APP/PS1 mice and that this cognitive impairment is mining the intensity that provoked discomfort. Details
not attributable to neuronal loss in the LA. of the behavioural procedures are described in the Sup-
As projection neurons in the LA receive inputs porting information, Supporting material.
through asymmetric synapses located mainly on the
heads of dendritic spines [15], we examined the mor- Stereology and morphology
phology of the dendritic trees and of individual den-
dritic spines in this region. Dendritic spines represent Unbiased stereology was employed to evaluate the
the main postsynaptic elements of excitatory synapses volume occupied by Aβ plaques and to determine
in the cerebral cortex [16] and they are fundamen- the neuronal density using Stereo Investigator soft-
tal in memory, learning, and cognition [17]. Dendritic ware (MicroBrightfield, Inc, Williston, VT, USA). The
spines undergo significant activity-dependent struc- density of the plaques and neurons in the LA was
tural changes [18], which are also influenced by spine estimated using the optical fractionator method. To
head size [19]. Importantly, recent evidence indicates estimate the plaque volume, the edges of each amy-
that spine heads are affected by oligomeric Aβ [20] loid plaque were marked with the Nucleator probe
and therefore the morphology of spine heads may link [27]. Dendrites were traced with Neurolucida (Micro-
Aβ pathology with synaptic dysfunction. Brightfield) and spine morphology was measured with
Previous studies have only identified a weak corre- Imaris software [28]. Details of the histological and
lation between the presence of Aβ plaques and AD morphological procedures appear in the Supporting
dementia [21], questioning whether plaques contribute information, Supporting material.
to dementia [22]. In an attempt to find a structural
basis for amygdala-dependent cognitive impairment, Statistics
we used advanced imaging and measurement tech-
niques to examine the morphology of the dendritic For the behavioural study, the results were analysed
trees and of individual spines. We measured the head using a two-tailed unpaired t-test or repeated measure
volume and neck length of thousands of dendritic ANOVA with the percentage of freezing for each
spines in the LA, both within and outside plaques. minute of the test as the repeated measure. For
We found that the morphology of the dendritic tree the morphological study, the results were analysed
of projection neurons that do not interact directly using a two-tailed unpaired t-test to test for the
with plaques is modified in the amygdala of APP/PS1 overall effect. When more than two groups were
mice and that there is a significant decrease in large compared (for analyses of dendrites and spines), one-
spines on these neurons. This is the first morphological way ANOVA was used, followed by Newman–Keuls
description of dendrites and spines in the amygdala of multiple comparison post-hoc tests. Comparisons of
an AD model, although similar studies have been per- Sholl analysis plots were performed with two-way
formed previously in the hippocampus and neocortex ANOVA, followed by the Bonferroni post-hoc test.
[23,24]. Comparisons between cumulative distributions were
made according to two-sample Kolmogorov–Smirnov
tests [29]. The significance of the results was accepted
Materials and methods at p < 0.05 and the data are presented as mean ± SE.
All experimental procedures were carried out in accor-
dance with the guidelines set out in the European Results
Community Council Directives (86/609/EEC).
Normal anxiety-related behaviour
Mice We first determined, using an elevated plus-maze,
The APP/PS1 mouse line used in this study (age whether APP/PS1 mice differ in their levels of anx-
12–14 months, males) expressed a Mo/Hu APP695- iety, a factor that may influence the results of fear
swe construct in conjunction with the exon-9-deleted conditioning [30] (see the Supporting information,
variant of human presenilin 1 (PS1-dE9) [25]. The Supporting material). APP/PS1 mice displayed normal
specific strain was B6C3-Tg (APPswe, PSEN1dE9) anxiety-related behaviour (Figures 1a and 1b), imply-
85Dbo/J. Age-matched littermates without the trans- ing that any change in fear conditioning cannot be
gene (Tg− ) served as controls. explained by different levels of anxiety.
Behavioural procedures Impaired auditory fear conditioning
Anxiety-related behaviour was evaluated in an ele- Auditory fear conditioning is a model for emotional
vated plus-maze as described previously [26]. Fear learning in animals and it is a response that depends
J Pathol 2009; 219: 41–51 DOI: 10.1002/path
Copyright  2009 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

3. Impaired learning and structural alteration in APP/PS1 mice 43
Figure 1. Normal anxiety-like behaviour and impaired auditory fear conditioning in APP/PS1 mice. (a, b) Exploration in the
elevated plus-maze was considered a measure of anxiety. The total distance travelled in the maze (a) and the time spent in
the closed and open arms (b) were similar in APP/PS1 and Tg− mice. (c) Average percentage freezing before, during, and after
tone-shock pairings. (d) Freezing responses on the testing day before and during the presentation of the tone. Note that the
percentage of freezing is lower for APP/PS1 mice. ∗ p < 0.05; ∗∗∗ p < 0.001
on the amygdala [31,32]. In this learning paradigm, not show alterations in their sensitivity to pain [35]
an emotionally neutral auditory conditioned stimulus (see the Supporting information, Supporting material),
elicits fear after it is paired with an aversive uncon- implying that conditioning was not affected by pain
ditioned stimulus [33]. We tested the auditory fear perception.
conditioning in these mice using immobility (freez-
ing) as an index of fear learning [34]. During con-
ditioning training, baseline freezing behaviour was
extremely low in both groups of mice before the Amyloid plaques occupy a small fraction of the LA
presentation of the tone (Figure 1c). When freezing
across conditioning trials was analysed, it was evi- In an attempt to determine the extent to which audi-
dent that both groups acquired fear responses, as tory fear conditioning might be affected by the Aβ
implied from the increased freezing during the post- plaques in these animals, we first quantified the vol-
shock periods (F2,38 = 18.78, p = 0.0001, repeated ume fraction that they occupied in the LA. We chose
measures ANOVA). Statistical analyses of the freez- to examine this area because of the key role that it
ing behaviour between groups indicated no signifi- plays in the acquisition and expression of fear-related
cant differences when animals were presented with the behaviour [36], and thus we immunocytochemically
tone (t11 = 1.30, p = 0.218) or immediately after foot stained Aβ plaques in sections of APP/PS1 brains
shock (t11 = 1.14, p = 0.277). (Figure 2a). Since counting spherical plaques in two-
On the day of testing, both genotypes showed a dimensional cross-sections provides an imprecise mea-
low amount of freezing during the baseline period sure of the amount of Aβ, missing small and het-
on the test day (t11 = 1.77, p = 0.21, Figure 1d). erogeneous assemblies of Aβ [37], we used unbiased
However, following the presentation of the tone, stereology to count plaques and determine their vol-
the freezing behaviour increased immediately in the ume. The density of plaques in the LA was 4.10 ± 0.45
Tg− mice, while it remained relatively low in the plaques/mm3 (N = 4) and the average plaque volume
APP/PS1 mice (t11 = 4.70, p = 0.0007). Indeed, the was 1496 ± 317.7 µm3 . The estimated volume occu-
averaged freezing response over the 3 min of the pied by the Aβ plaques was only 0.60 ± 0.11%. These
memory test showed a 42% reduction in APP/PS1 results suggest that Aβ plaques occupy a small volume
mice compared with the Tg− animals, indicating fraction of the LA of aged APP/PS1 mice, in accor-
that auditory fear conditioning memory is impaired dance with a previous study in which relatively few
in APP/PS1 mice. Importantly, APP/PS1 mice did Aβ plaques were found in the LA of AD patients [38].
J Pathol 2009; 219: 41–51 DOI: 10.1002/path
Copyright  2009 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

4. 44 S Knafo et al
Figure 2. Coronal sections through the amygdala and adjacent regions showing the pattern of distribution of plaques. (a) An
example of the amygdaloid region as it appears in a section stained with the anti-Aβ antibody and Nissl, as used for plaque
quantification. (b, c) Coronal sections from Tg− and APP/PS1 mice stained for anti-NeuN. (d) Double staining for anti-β amyloid
and NeuN. (e) Neuron density for Tg− and APP/PS1 mice. BLA = basolateral nucleus of the amygdala; EC = external capsule;
LA = lateral nucleus of the amygdala; PIR = piriform cortex. Scale bars: (a) 350 µm; (b–d) 120 µm
Neuron density is conserved in LA significantly different between these three groups of
neurons (Figures 3d and 3f, and Supporting informa-
We determined whether neuronal loss in the LA tion, Supporting Table 1). However, a detailed anal-
of APP/PS1 mice might explain the dampening of
ysis of the length as a function of the distance
the auditory fear conditioning by assessing neuronal
from the soma (Sholl analysis analysed with two-way
density using unbiased stereology in sections stained
ANOVA, p < 0.0001) revealed that neurons belonging
with antibodies for NeuN, a neuron-specific nuclear
to APP/PS1 mice had a significantly smaller dendritic
protein [39] (Figures 2b–2d). No neuronal loss was
found in the LA of APP/PS1 mice (Figure 2e) in line length 30–40 µm from the soma (p < 0.05, Bonfer-
with previous studies performed in aged transgenic roni post-hoc test, Figure 3e). Also, there were fewer
mice with mutant amyloid precursor protein (APP), intersections at 30–40 µm from the soma (p < 0.05,
where there was no neuronal loss in cortical areas [11]. Bonferroni post-hoc test, Figure 3g).
These findings raised the possibility that the ram-
ification of the dendritic tree differed in APP/PS1
Altered dendritic structure neurons. Indeed, we found that the three neuronal
Since auditory fear conditioning is thought to be medi- types differed significantly in the total number of
ated by synaptic changes in the LA [40], and given that dendritic branches per neuron (p = 0.016, one-way
most excitatory synaptic connections occur on den- ANOVA, Figure 3h). Moreover, a post-hoc analysis
dritic spines [41], we examined whether dendrites and revealed a significant (p < 0.05, Newman–Keuls mul-
dendritic spines were altered in the LA of APP/PS1 tiple comparison test) decrease in this parameter for
mice. We traced projection neurons in the LA and then PFNs when compared with Tg− neurons (Supporting
divided them into three categories, according to their information, Supporting Table 1). Quantification of the
location with respect to Aβ plaques (Figures 3a–3c): numbers of each dendritic branch order per neuron
(1) neurons from control mice (Tg− ); (2) neurons revealed a change in the number of second-, third-,
with no dendrite that enters a plaque (plaque-free and fourth-order dendrites per neuron in PFNs (p <
neurons, PFNs); and (3) neurons with at least one 0.001, p < 0.05, and p < 0.01, respectively, two-
branch of a dendrite passing through or entering a way ANOVA, Figure 3i, Supporting information, Sup-
plaque (plaque neurons, PNs). The total dendritic porting Table 1). Moreover, a post-hoc analysis also
length and the total number of intersections were not revealed a significant difference between the PNs and
J Pathol 2009; 219: 41–51 DOI: 10.1002/path
Copyright  2009 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

5. Impaired learning and structural alteration in APP/PS1 mice 45
Figure 3. Intracellular injections and morphometric analysis. (a) Panoramic confocal (10×) views of the LA showing
Alexa594-injected neurons and thioflavin-s-positive plaques in a Tg− mouse (left) and an APP/PS1 mouse (right). (b) Representative
images of projection neurons from a Tg− mouse (left) and an APP/PS1 mouse (right). (c) The method used to distinguish dendrites
and spines within and outside plaques. Left: a plaque suspected of containing a dendrite due to the rotation of its three-dimensional
image. Centre: the plaque surface is marked with the aid of the IsoSurface tool of Imaris software. Right: the voxels outside the
surface are set to zero, leaving only the dendritic segment within the plaque. This process was performed after blind morphological
measurements were made. (d) The total dendritic length of the different neuronal categories. (e) Sholl analysis showing the
dendritic length as a function of the distance from the soma. Red asterisks represent a significant difference between Tg− and PFN;
blue asterisks represent a significant difference between Tg− and PN; black asterisks represent a significant difference between PN
and PFN. Inset: diagram showing a traced LA neuron and a series of concentric circles representing the Sholl analysis. This analysis
is performed with concentric spheres around the soma rather than circles, in order to give a three-dimensional result. (f) Total
dendritic number of intersections for the different neuronal categories. (g) Sholl analysis showing the number of intersections as
a function of the distance from the soma. The colour of the asterisks follows the scheme indicated in e. (h) Total number of
branches per neuron for the different neuronal categories. (i) Quantity of branches per order per neuron. Scale bars: (a) 50 µm;
(b) 25 µm; (c) 5 µm. ∗ p < 0.05; ∗∗ p < 0.01; ∗∗∗ p < 0.001
PFNs for the fourth branch order (p < 0.05, Bon- described for the neocortex [43] and hippocampus
ferroni post-hoc test, Figure 3i). We did not detect [28], typical thioflavin-s-positive plaques consisted of
any significant differences in the average tortuosity a core surrounded by a diffuse ring of decreasing
of the different branch orders (Supporting informa- density. We encountered 20 dendrites that entered
tion, Supporting Table 1). Thus, we concluded that plaques (Figure 3c) and they were always located
the PFNs of APP/PS1 mice had a less complex den- in the diffuse peripheral ring (Figure 3c). Thus, like
dritic tree. This finding is in line with a previous the classification of neurons, the dendrites were cat-
study in which somatosensory cortical neurons were egorized according to their location with respect to
less branched in TG2576 mice, another AD model Aβ plaques (Figure 4a): (1) dendrites from Tg− mice;
[42]. (2) dendrites belonging to neurons with no dendrites
entering a plaque (PFN); (3) segments of dendrites
Decreased dendrite diameter and spine density within a plaque (Plaque); and (4) dendrites arising
within plaques from neurons of which one of their branches passed
into or entered a plaque, in segments outside plaques
We examined 259 amyloid plaques and 143 APP/PS1 (PN).
injected projection neurons in the LA by laser scan- Dendritic shaft diameter differed significantly
ning confocal microscopy (Figures 3a and 3b). As among these four categories of dendrites (p = 0.0001,
J Pathol 2009; 219: 41–51 DOI: 10.1002/path
Copyright  2009 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

6. 46 S Knafo et al
Figure 4. The decrease in dendritic diameter and dendritic spine density is limited to plaques. (a) Representative projection
images of dendrites from Tg− and APP/PS1 mice (63×, glycerol). Tg− , a dendrite from a control mouse; PFN, a dendrite belonging
to plaque-free neurons of APP/PS1 mouse; PN, a dendrite arising from a neuron contacting a plaque in a plaque-free segment;
Plaque, a dendrite that enters a plaque, with and without the green channel that contains the plaque image. (b) Dendrite diameter
was significantly decreased within plaques and (c) the spine density was also significantly lower inside plaques. (d) Spine density
as a function of the distance from the soma (Sholl analysis) is similar in Tg− mice and APP/PS1 mice in plaque-free regions (PN
and PFN). (e) For each plaque-related dendritic segment, the distance of the plaque from the soma was measured, and the ratio
between the spine density for the segment and the average spine density for the same distance in Tg− mice was calculated.
(f) Spine density at increasing distances from the plaque edges. Note that outside the plaques, the spine density is conserved. Scale
bar = 5 µm. ∗∗ p < 0.01; ∗∗∗ p < 0.001
J Pathol 2009; 219: 41–51 DOI: 10.1002/path
Copyright  2009 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

7. Impaired learning and structural alteration in APP/PS1 mice 47
one-way ANOVA, Figure 4b and Supporting informa- Discussion
tion, Supporting Table 2). Accordingly, the dendritic
shaft diameter inside plaques was significantly smaller In this study, aged APP/PS1 mice showed a damp-
(20 dendrites, p < 0.001, Newman–Keuls multiple ening of auditory fear conditioning, a learning task
comparison test) than in dendrites from PNs (45 that depends on the LA nucleus of the amygdala [36].
dendrites), dendrites from PFNs (58 dendrites), and Using a series of experiments, we found that this inhi-
Tg− dendrites (66 dendrites). Conversely, no sig- bition did not arise from a change in anxiety or the
nificant differences were found between Tg− and animals’ sensitivity to shock. In addition, we show that
PFN, or PN dendrites in segments that did not pass the volume occupied by plaques in LA is less than 1%
through a plaque. The spine density was also signif- and that the neuronal density in this nucleus was con-
icantly different among the four categories of den- served. However, we did find that plaque-free neurons
drites (p = 0.002, one-way ANOVA, Figure 4c) and in the LA of APP/PS1 mice have altered dendritic ram-
it was significantly lower within plaques (p < 0.05, ifications and fewer large spines. Although previous
Newman–Keuls multiple comparison test) than in the studies have shown changes in dendrites and spines in
other categories of dendrites (Supporting information, transgenic mouse models of AD, these studies focused
Supporting Table 2). There were no significant dif- on the neocortex [23,42,44] and hippocampus [28].
ferences in the spine density between PFN, PN and The results demonstrate for the first time that morpho-
Tg− dendrites. logical changes in dendrites also occur in the amygdala
Spine density normally changes as a function of the and that these changes may account for the dampening
distance from the soma and these changes can be esti- in amygdala-dependent learning. APP/PS1 mice serve
mated using Sholl analysis. This analysis revealed that as a model of Alzheimer’s disease (AD) since they
the spine density for Tg− dendrites was not signifi- express two of the mutations that exist in patients of
cantly different from the density in PNs and PFNs over familial AD. AD patients show a marked impairment
their entire length (Figure 4d and Supporting informa- of fear conditioning [45], implying that the inhibition
tion, Supporting Table 2). The spine density for each of non-declarative memory is common to both AD
segment within a plaque was compared with the aver- patients and APP/PS1 mice. Thus, our results represent
age spine density at the same distance from the soma an additional demonstration that behavioural features
in Tg− mice; there was a decrease of 33.43 ± 11.05% of this AD model resemble those found in AD [46].
in the spine density within plaques (p = 0.0059, We have shown here that the performance in
t-test, Figure 4e). We then examined the spine den- an amygdala-dependent task is severely impaired in
sity as a function of the distance from plaques and we APP/PS1 male mice. Our results are in accordance
found that the spine density in the surrounding den- with the recent demonstration that overexpression of
drites was not significantly different from the values APP in rodents, leading to elevated levels of Aβ damp-
in Tg− mice. Thus, in accordance with our previous ens auditory fear conditioning [12,47]. Importantly,
study of the dentate gyrus [28], our data show that the other authors have reported reduced contextual, but
decrease in spine density in the LA was restricted to not cued, fear conditioning memory in 5- or 9-month-
plaques. old APP/PS1 mice [13]. The discrepancy between our
results and these earlier findings may reflect the dif-
Decreased frequency of large spines in PFNs ferent combination of APP and PS1 transgenes in the
Spine head volume and neck length were measured in mice used in these two studies, which might affect
three dimensions in confocal image stacks (Figures 5a the amount of Aβ in the brains of these mice. In
and 5b) and for each group, the number of spines addition, different strains were employed in these stud-
measured was Tg− , 3949; PFN, 3268; PN, 3577; and ies (C57B6/C3 versus C57B6/SJL), which may be
Plaque, 63. No significant differences were found in relevant given that strain differences have already
the average spine neck length among these dendritic been reported to influence cued fear conditioning [48].
categories (p = 0.48, one-way ANOVA, Figure 5c Indeed, SJL mice are more sensitive to cued fear con-
and Supporting information, Supporting Table 2) and ditioning than C3 mice [48] and this sensitivity would
likewise, the average head volume was not signifi- facilitate the increase in freezing behaviour after cued
cantly different between the four categories of den- fear conditioning observed by Dineley et al [13]. This
drites (p = 0.42, one-way ANOVA, Figure 5d and enhanced freezing behaviour may mask the differences
Supporting information, Supporting Table 2). How- between the transgenic and the wild-type mice. More-
ever, a cumulative frequency plot analysis revealed over, the training conditions used in our study (three
a statistically significant shift in the head volume of foot shocks of 0.75 mA with an inter-trial interval
spines from PFNs when compared with that of spines of 30 s) were different from those used by Dineley
belonging to the other categories (p = 0.012–0.048, et al(two or five foot shocks of 0.5 mA with a 5 min
Kolmogorov–Smirnov test, Figure 5e). A decrease of or a 40 s interval between each CS–US pairing). Vari-
74% was found in the frequency of large spines (head ation in the intensity of the foot shock, as well as in
volume > 0.18 µm3 ) when compared with Tg− neu- the inter-trial interval and total time spent condition-
rons. Thus, PFNs have a lower frequency of large ing, affects the conditioning in response to explicit
spines. and contextual stimuli [31,49,50]. Taken together, the
J Pathol 2009; 219: 41–51 DOI: 10.1002/path
Copyright  2009 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

8. 48 S Knafo et al
Figure 5. Changes in the spine head volume of plaque-free neurons. (a) Maximum projection confocal images (63×, glycerol)
of dendrites representative of each different category. (b) An amplified example of the head volume measurement obtained by
determining the surface of the spine head (blue). (c, d) Bar graphs of the neck length and head volume indicating that no significant
differences were found for the average values. (e) Left: a cumulative frequency plot showing the distribution of spine head volumes.
Note the significant left shift in head volume in PFN (Kolmogorov–Smirnov test). Right: histograms for head volumes. Tg− , a
dendrite from a no transgene (control) mouse; PFN, a dendrite from a plaque-free neuron; PN, a dendrite from a plaque neuron;
Plaque, a dendrite entering a plaque. See text for further details. Scale bar = 3 µm
setting in which training was performed may favour reasonable to hypothesize that the neuronal loss in the
the differences in fear learning between transgenic and amygdala is a late neuropathological feature of AD and
control mice. The alterations to conditioned fear are that other, more subtle synaptic changes may occur in
unlikely to result from distinct perception and/or pro- early stages of the disease that cause the impaired fear
cessing of the foot shock, or from different levels of memory [52]. We therefore examined dendrites and
anxiety, because both Tg− and APP/PS1 mice showed dendritic spines, which are the major sites of synaptic
the same pain sensitivity to a foot shock of rising inten- contacts in the brain.
sity and similar anxiety-related behaviour. Three different observations from our study support
We have shown here that in contrast to AD patients, the notion that Aβ plaques are not responsible for
who suffer marked neuronal loss in the amygdala [51], the dampened auditory fear conditioning in APP/PS1
there is no neuronal loss in the LA of aged APP/PS1 mice. (1) Aβ plaques in the LA occupy less than
mice. Given the impaired auditory fear learning in 1% of its volume, implying that the changes within
these mice, this finding raises doubts as to whether the plaques, the decrease in dendritic diameter and
the fear memory disturbances seen in AD patients arise spine density, are restricted to a small fraction of
from neuronal loss in the amygdala. Instead, it seems the neuropil. (2) The decreased dendritic ramification
J Pathol 2009; 219: 41–51 DOI: 10.1002/path
Copyright  2009 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

9. Impaired learning and structural alteration in APP/PS1 mice 49
in APP/PS1 mice is limited to PFNs, neurons that and that they are therefore less able to generate fear
do not contact Aβ plaques. Branching patterns are memories.
related to the degree of compartmentalization of the
inputs to the cell. Indeed, it has been proposed that Morphological changes within plaques may have
a stronger potential for compartmentalization results a minor effect on cognition
in a significant increase in the representational power We have also shown here that dendrites that pass
and greater learning and memory capacity [53]. Since through a plaque are thinner and suffer from decreased
PFNs are less complex but they have a similar spine spine density. Aβ plaques have previously been asso-
density, the fewer branches probably means the loss of ciated with local synaptic abnormalities and with a
synapses, in accordance with previous findings in AD smaller diameter of the neuronal processes [23,44].
brains [22,54]. Thus, the reduced complexity of the Since the correlation between the plaque load and the
PFNs of APP/PS1 mice may contribute to their limited degree of memory impairment in transgenic mice is
ability to learn the fear response. Nevertheless, the relatively weak [70,71], the relevance of these plaque-
possibility that impaired fear learning and the changes related morphological changes to AD pathogenesis is
in dendritic ramification are two unrelated findings unclear and even questioned by many investigators
cannot be discarded. (3) The decrease in the frequency [72]. We suggest that the morphological changes inside
of large spines in APP/PS1 mice is limited to the PFNs, plaques might affect local synaptic circuits. Neverthe-
implying that the decreased frequency of large spines less, plaques only occupy a small volume of the LA,
is not related to plaques. Importantly, we recently less than 1%, and thus the alterations to these local
reported that plaque-free regions in the dentate gyrus circuits are restricted to only a small portion of the
of APP/PS1 mice also have fewer large spines [28]. neuropil. Therefore, it is more likely that the changes
Since in both studies the density of spines remained in dendritic complexity and in the frequency of large
unchanged outside the plaques, a decrease of large spines not directly related to the plaques contribute to
spines appears to be a general feature of APP/PS1 the cognitive impairment seen in APP/PS1 mice.
mice. Further studies in other brain regions will be
necessary to confirm this possibility.
The spine head volume reflects the size of the post- Acknowledgements
synaptic density [55–57], which correlates with the This work was supported by the following grants: CIBERNED,
number of presynaptic vesicles and with the number RETICEF, Fundaci´ n Caixa (BM05-47-0), The EU 6th Frame-
o
of docked vesicles [55]. The postsynaptic density area work Programme (PROMEMORIA LSHM-CT-2005-512012),
is proportional to the number of postsynaptic receptors and the Spanish Ministry of Science and Technology (grants
[58], whereas the number of docked vesicles is propor- BFU2006-13395 to JD and BFU2006-01050 to CV). SK was
the recipient of postdoctoral fellowships from the Ram´ n y
o
tional to the ready releasable pool of transmitter [59].
Cajal Programme of the Ministry of Science and Technology.
Therefore, the volume of the spine head is likely to We thank Ana Martinez and Gerard Muntane for genotyping
be directly proportional to the average reliability and the mice, and Luis Carrillo for technical assistance.
strength of its synapse, and thus it is an important mor-
phometric parameter reflecting its activity [55–59]. It
has been reported that large spines in the LA are post- Supporting information
synaptic to thalamic afferents and that they harbour
R-type voltage-dependent Ca2+ channels (VDCCs), Supporting information may be found in the online
thereby providing the synapses on their heads with version of this article.
the capacity to express associative long-term modifi-
cations of synaptic strength [60]. Given that long-term
synaptic plasticity in the LA is required for memory References
consolidation of fear conditioning [3], a decrease in
the frequency of large spines may contribute to the 1. Copeland MP, Daly E, Hines V, Mastromauro C, Zaitchik D,
loss of fear memory seen in APP/PS1 mice. Gunther J, et al. Psychiatric symptomatology and prodromal
Alzheimer’s disease. Alzheimer Dis Assoc Disord 2003;17:1–8.
Large spines are thought to act as physical traces 2. Mirra SS, Hyman BT. Aging and dementia. In Greenfield’s
of long-term memory [61,62], particularly since they Neuropathology (7th edn), Graham DI, Lantos PI (eds). Arnold:
are formed after synaptic potentiation [18,62] but New York, 2002; 195–272.
also because they bear exceptionally stable synapses 3. Schafe GE, Doyere V, LeDoux JE. Tracking the fear engram: the
[63,64]. A recent study demonstrated a unique role for lateral amygdala is an essential locus of fear memory storage.
J Neurosci 2005;25:10010–10014.
large spines in learning-associated changes of receptor 4. Vereecken TH, Vogels OJ, Nieuwenhuys R. Neuron loss and
trafficking and it suggested that they are critical for shrinkage in the amygdala in Alzheimer’s disease. Neurobiol Aging
behaviourally relevant plasticity [65]. Large spines are 1994;15:45–54.
also enriched with AMPA receptors [66], which make 5. Scott SA, DeKosky ST, Scheff SW. Volumetric atrophy of
their synapses functionally stronger [66–68], and they the amygdala in Alzheimer’s disease: quantitative serial
reconstruction. Neurology 1991;41:351.
are also needed for synaptic plasticity [69]. Thus, a 6. Mori E, Ikeda M, Hirono N, Kitagaki H, Imamura T, Shimo-
decrease in the frequency of large spines may imply mura T. Amygdalar volume and emotional memory in Alzheimer’s
that synaptic plasticity is dampened in APP/PS1 mice disease. Am J Psychiatry 1999;156:216–222.
J Pathol 2009; 219: 41–51 DOI: 10.1002/path
Copyright  2009 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

10. 50 S Knafo et al
7. Blair HT, Schafe GE, Bauer EP, Rodrigues SM, LeDoux JE. 28. Knafo S, Alonso-Nanclares L, Gonzalez-Soriano J, Merino-Serrais
Synaptic plasticity in the lateral amygdala: a cellular hypothesis P, Fernaud-Espinosa I, Ferrer I, et al. Widespread changes in den-
of fear conditioning. Learn Mem 2001;8:229–242. dritic spines in a model of Alzheimer’s disease. Cereb Cortex
8. Rodrigues SM, Schafe GE, LeDoux JE. Molecular mechanisms 2009;19:586–592.
underlying emotional learning and memory in the lateral amygdala. 29. Conover WJ. Statistics of the Kolmogorov–Smirnov type. In
Neuron 2004;44:75–91. Practical Nonparametric Statistics. John Wiley: New York, 1999;
9. McDonald AJ. Cortical pathways to the mammalian amygdala. 428–456.
Prog Neurobiol 1998;55:257–332. 30. Lister RG. The use of a plus-maze to measure anxiety in the
10. McDonald AJ, Culberson JL. Neurons of the basolateral amyg- mouse. Psychopharmacology (Berlin) 1987;92:180–185.
dala: a Golgi study in the opossum (Didelphis virginiana). Am J 31. Phillips RG, LeDoux JE. Differential contribution of amygdala
Anat 1981;162:327–342. and hippocampus to cued and contextual fear conditioning. Behav
11. Borchelt DR, Ratovitski T, van Lare J, Lee MK, Gonzales V, Neurosci 1992;106:274–285.
Jenkins NA, et al. Accelerated amyloid deposition in the brains 32. Kim M, Davis M. Electrolytic lesions of the amygdala block
of transgenic mice coexpressing mutant presenilin 1 and amyloid acquisition and expression of fear-potentiated startle even with
precursor proteins. Neuron 1997;19:939–945. extensive training but do not prevent reacquisition. Behav Neurosci
1993;107:580–595.
12. Gerlai R, Fitch T, Bales KR, Gitter BD. Behavioral impairment of
33. Fanselow MS, LeDoux JE. Why we think plasticity underlying
APP(V717F) mice in fear conditioning: is it only cognition? Behav
Pavlovian fear conditioning occurs in the basolateral amygdala.
Brain Res 2002;136:503–509.
Neuron 1999;23:229–232.
13. Dineley KT, Xia X, Bui D, Sweatt JD, Zheng H. Acceler-
34. Nader K, Schafe GE, Le Doux JE. Fear memories require protein
ated plaque accumulation, associative learning deficits, and up-
synthesis in the amygdala for reconsolidation after retrieval. Nature
regulation of alpha 7 nicotinic receptor protein in transgenic mice
2000;406:722–726.
co-expressing mutant human presenilin 1 and amyloid precursor 35. Marsch R, Foeller E, Rammes G, Bunck M, Kossl M, Holsboer F,
proteins. J Biol Chem 2002;277:22768–22780. et al. Reduced anxiety, conditioned fear, and hippocampal long-
14. Trinchese F, Fa’ M, Liu S, Zhang H, Hidalgo A, Schmidt SD, term potentiation in transient receptor potential vanilloid type 1
et al. Inhibition of calpains improves memory and synaptic receptor-deficient mice. J Neurosci 2007;27:832–839.
transmission in a mouse model of Alzheimer disease. J Clin Invest 36. Koo JW, Han JS, Kim JJ. Selective neurotoxic lesions of
2008;118:2796–2807. basolateral and central nuclei of the amygdala produce differential
15. Sah S, Faber ES, Lopez De AM, POWER J. The amygdaloid effects on fear conditioning. J Neurosci 2004;24:7654–7662.
complex: anatomy and physiology. Physiol Rev 2003;83: 37. Stark AK, Pelvig DP, Jorgensen AM, Andersen BB, Pakken-
803–834. berg B. Measuring morphological and cellular changes in
16. Gray EG. Electron microscopy of synaptic contacts on dendrite Alzheimer’s dementia: a review emphasizing stereology. Curr
spines of the cerebral cortex. Nature 1959;183:1592–1593. Alzheimer Res 2005;2:449–481.
17. Lamprecht R, LeDoux J. Structural plasticity and memory. Nat 38. Kromer Vogt LJ, Hyman BT, Van Hoesen GW, Damasio AR.
Rev Neurosci 2004;5:45–54. Pathological alterations in the amygdala in Alzheimer’s disease.
18. Lang C, Barco A, Zablow L, Kandel ER, Siegelbaum SA, Neuroscience 1990;37:377–385.
Zakharenko SS. Transient expansion of synaptically connected 39. Mullen RJ, Buck CR, Smith AM. NeuN, a neuronal specific
dendritic spines upon induction of hippocampal long-term poten- nuclear protein in vertebrates. Development 1992;116:201–211.
tiation. Proc Natl Acad Sci U S A 2004;101:16665–16670. 40. Rogan MT, Staubli UV, LeDoux JE. Fear conditioning induces
19. Majewska A, Tashiro A, Yuste R. Regulation of spine associative long-term potentiation in the amygdala. Nature
calcium dynamics by rapid spine motility. J Neurosci 1997;390:604–607.
2000;20:8262–8268. 41. Yuste R, Denk W. Dendritic spines as basic functional units of
20. Lacor PN, Buniel MC, Furlow PW, Sanz Clemente A, Velasco neuronal integration. Nature 1995;375:682–684.
PT, Wood M, et al. A{beta} oligomer-induced aberrations in 42. Alpar A, Ueberham U, Bruckner MK, Seeger G, Arendt T,
synapse composition, shape, and density provide a molecular Gartner U. Different dendrite and dendritic spine alterations in
basis for loss of connectivity in Alzheimer’s disease. J Neurosci basal and apical arbors in mutant human amyloid precursor protein
2007;27:796–807. transgenic mice. Brain Res 2006;1099:189–198.
21. Terry RD, Masliah E, Salmon DP, Butters N, DeTeresa R, Hill R, 43. Cruz L, Urbanc B, Buldyrev SV, Christie R, Gomez-Isla T,
et al. Physical basis of cognitive alterations in Alzheimer’s disease: Havlin S, et al. Aggregation and disaggregation of senile
synapse loss is the major correlate of cognitive impairment. Ann plaques in Alzheimer disease. Proc Natl Acad Sci U S A
1997;94:7612–7616.
Neurol 1991;30:572–580.
44. Spires TL, Meyer-Luehmann M, Stern EA, McLean PJ, Skoch J,
22. Terry RD. Cell death or synaptic loss in Alzheimer disease.
Nguyen PT, et al. Dendritic spine abnormalities in amyloid
J Neuropathol Exp Neurol 2000;59:1118–1119.
precursor protein transgenic mice demonstrated by gene
23. Tsai J, Grutzendler J, Duff K, Gan WB. Fibrillar amyloid
transfer and intravital multiphoton microscopy. J Neurosci
deposition leads to local synaptic abnormalities and breakage of
2005;25:7278–7287.
neuronal branches. Nature Neurosci 2004;7:1181–1183.
45. Hamann S, Monarch ES, Goldstein FC. Impaired fear conditioning
24. Dong H, Martin MV, Chambers S, Csernansky JG. Spatial in Alzheimer’s disease. Neuropsychologia 2002;40:1187–1195.
relationship between synapse loss and beta-amyloid deposition in 46. Hsiao K, Chapman P, Nilsen S, Eckman C, Harigaya
Tg2576 mice. J Comp Neurol 2007;500:311–321. Y, Younkin S, et al. Correlative memory deficits, abeta
25. Scheuner D, Eckman C, Jensen M, Song X, Citron M, Suzuki N, elevation, and amyloid plaques in transgenic mice. Science
et al. Secreted amyloid beta-protein similar to that in the 1996;274:99–103.
senile plaques of Alzheimer’s disease is increased in vivo by 47. Clarke J, Thornell A, Corbett D, Soininen H, Hiltunen M, Jolkko-
the presenilin 1 and 2 and APP mutations linked to familial nen J. Overexpression of APP provides neuroprotection in the
Alzheimer’s disease. Nature Med 1996;2:864–870. absence of functional benefit following middle cerebral artery
26. Herrero AI, Sandi C, Venero C. Individual differences in anxiety occlusion in rats. Eur J Neurosci 2007;26:1845–1852.
trait are related to spatial learning abilities and hippocampal 48. Balogh SA, Wehner JM. Inbred mouse strain differences in
expression of mineralocorticoid receptors. Neurobiol Learn Mem the establishment of long-term fear memory. Behav Brain Res
2006;86:150–159. 2003;140:97–106.
27. Moller A, Strange P, Gundersen HJ. Efficient estimation of cell 49. Kaplan PS, Hearst E. Bridging temporal gaps between CS and US
volume and number using the nucleator and the disector. J Microsc in autoshaping: insertion of other stimuli before, during, and after
1990;159:61–71. CS. J Exp Psychol Anim Behav Process 1982;8:187–203.
J Pathol 2009; 219: 41–51 DOI: 10.1002/path
Copyright  2009 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

11. Impaired learning and structural alteration in APP/PS1 mice 51
50. Rescorla RA. ‘Configural’ conditioning in discrete-trial bar 63. Trachtenberg JT, Chen BE, Knott GW, Feng G, Sanes JR, Welker
pressing. J Comp Physiol Psychol 1972;79:307–317. E, et al. Long-term in vivo imaging of experience-dependent
51. Herzog AG, Kemper TL. Amygdaloid changes in aging and synaptic plasticity in adult cortex. Nature 2002;420:788–794.
dementia. Arch Neurol 1980;37:625–629. 64. Grutzendler J, Kasthuri N, Gan WB. Long-term dendritic spine
52. Selkoe DJ. Alzheimer’s disease is a synaptic failure. Science stability in the adult cortex. Nature 2002;420:812–816.
2002;298:789–791. 65. Matsuo N, Reijmers L, Mayford M. Spine-type-specific recruit-
53. Poirazi P, Mel BW. Impact of active dendrites and structural ment of newly synthesized AMPA receptors with learning. Science
plasticity on the memory capacity of neural tissue. Neuron 2008;319:1104–1107.
2001;29:779–796. 66. Matsuzaki M, Ellis-Davies GC, Nemoto T, Miyashita Y, Iino M,
54. Masliah E, Crews L, Hansen L. Synaptic remodeling during aging Kasai H. Dendritic spine geometry is critical for AMPA receptor
and in Alzheimer’s disease. J Alzheimer’s Dis 2006;9:91–99. expression in hippocampal CA1 pyramidal neurons. Nature
55. Harris KM, Stevens JK. Dendritic spines of CA 1 pyramidal cells Neurosci 2001;4:1086–1092.
in the rat hippocampus: serial electron microscopy with reference 67. Ashby MC, Maier SR, Nishimune A, Henley JM. Lateral diffu-
to their biophysical characteristics. J Neurosci 1989;9:2982–2997. sion drives constitutive exchange of AMPA receptors at den-
56. Schikorski T, Stevens CF. Quantitative ultrastructural analysis of dritic spines and is regulated by spine morphology. J Neurosci
hippocampal excitatory synapses. J Neurosci 1997;17:5858–5867. 2006;26:7046–7055.
57. Arellano JI, Espinosa A, Fairen A, Yuste R, Defelipe 68. Nimchinsky EA, Yasuda R, Oertner TG, Svoboda K. The number
J. Non-synaptic dendritic spines in neocortex. Neuroscience of glutamate receptors opened by synaptic stimulation in single
2007;145:464–469. hippocampal spines. J Neurosci 2004;24:2054–2064.
58. Nusser Z, Lujan R, Laube G, Roberts JD, Molnar E, Somogyi P. 69. Lee HK, Takamiya K, Han JS, Man H, Kim CH, Rumbaugh G,
Cell type and pathway dependence of synaptic AMPA et al. Phosphorylation of the AMPA receptor GluR1 subunit is
receptor number and variability in the hippocampus. Neuron required for synaptic plasticity and retention of spatial memory.
1998;21:545–559. Cell 2003;112:631–643.
59. Dobrunz LE, Stevens CF. Heterogeneity of release probabil- 70. Moechars D, Lorent K, de SB, Dewachter I, Van LF.
ity, facilitation, and depletion at central synapses. Neuron Expression in brain of amyloid precursor protein mutated in
1997;18:995–1008. the alpha-secretase site causes disturbed behavior, neuronal
60. Humeau Y, Herry C, Kemp N, Shaban H, Fourcaudot E, degeneration and premature death in transgenic mice. EMBO J
Bissiere S, et al. Dendritic spine heterogeneity determines 1996;15:1265–1274.
afferent-specific Hebbian plasticity in the amygdala. Neuron 71. Westerman MA, Cooper-Blacketer D, Mariash A, Kotilinek L,
2005;45:119–131. Kawarabayashi T, Younkin LH, et al. The relationship between
61. Kasai H, Matsuzaki M, Noguchi J, Yasumatsu N, Nakahara H. Abeta and memory in the Tg2576 mouse model of Alzheimer’s
Structure–stability–function relationships of dendritic spines. disease. J Neurosci 2002;22:1858–1867.
Trends Neurosci 2003;26:360–368. 72. Haass C, Selkoe DJ. Soluble protein oligomers in neurodegener-
62. Bourne J, Harris KM. Do thin spines learn to be mushroom spines ation: lessons from the Alzheimer’s amyloid beta-peptide. Nature
that remember? Curr Opin Neurobiol 2007;17:381–386. Rev Mol Cell Biol 2007;8:101–112.
J Pathol 2009; 219: 41–51 DOI: 10.1002/path
Copyright  2009 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.