Neuron
Article
Fear Extinction Causes Target-Specific Remodeling
of Perisomatic Inhibitory Synapses
Ste´ phanie Trouche,1,* Jennifer M. Sasaki,1,2 Tiffany Tu,1 and Leon G. Reijmers1,2,*
1Department of Neuroscience, School of Medicine
2Graduate Program in Neuroscience, Sackler School of Graduate Biomedical Sciences
Tufts University, Boston, MA 02111, USA
*Correspondence: trouche.stephanie@gmail.com (S.T.), leon.reijmers@tufts.edu (L.G.R.)
http://dx.doi.org/10.1016/j.neuron.2013.07.047
SUMMARY
A more complete understanding of how fear extinction
alters neuronal activity and connectivity within
fear circuits may aid in the development of strategies
to treat human fear disorders. Using a c-fos-based
transgenic mouse, we found that contextual fear
extinction silenced basal amygdala (BA) excitatory
neurons that had been previously activated during
fear conditioning. We hypothesized that the silencing
of BA fear neurons was caused by an action of
extinction on BA inhibitory synapses. In support
of this hypothesis, we found extinction-induced
target-specific remodeling of BA perisomatic inhibitory
synapses originating from parvalbumin and
cholecystokinin-positive interneurons. Interestingly,
the predicted changes in the balance of perisomatic
inhibition matched the silent and active states of the
target BA fear neurons. These observations suggest
that target-specific changes in perisomatic inhibitory
synapses represent a mechanism through which
experience can sculpt the activation patterns within
a neural circuit.
INTRODUCTION
Exposure therapy is widely used to treat fear disorders, but it
rarely leads to a complete and permanent loss of maladaptive
fear. A deeper understanding of the neurobiological mechanisms
that underlie exposure therapy can be achieved by studying fear
extinction in animal models (Graham et al., 2011) and may be
useful for the development of more effective therapies. Over
the past decades, studies on the neurobiological basis of fear
extinction have discovered that multiple brain regions are recruited
by fear extinction (Corcoran and Maren, 2001; Falls
et al., 1992; Morgan et al., 1993; Vianna et al., 2001). These brain
regions include both cortical and subcortical areas that are
reciprocally connected, thereby forming a distributed extinction
circuit that can be recruited by behavioral extinction training and
that, upon its recruitment, can lead to the loss or suppression of
fear (Orsini and Maren, 2012). In addition to the extinction circuit,
a fear circuit has been characterized that is responsible for the
storage and expression of fear memories and that is also distributed
over multiple brain regions (Orsini and Maren, 2012). Important
for using rodents as model organisms, both the extinction
and fear circuits are highly conserved between rodents and
humans (Hartley and Phelps, 2010). In this study, we address
the question of the precise anatomical and functional connection
between the extinction circuit and the fear circuit toward the aim
of gaining a greater understanding of how they interact during
fear extinction.
One potential strategy for identifying the interface between the
extinction circuit and the fear circuit is to identify neurons within
the fear circuit that are silenced by extinction and then use these
neurons as a starting point for determining which upstream
events within the extinction circuit cause their silencing. The first
step toward applying this strategy was made using electrophysiological
recordings of neurons in the amygdala, a brain region
known as a central hub within the fear circuit (Orsini and Maren,
2012). Electrophysiological recordings revealed that neurons in
the lateral amygdala and the basal amygdala can increase their
firing in response to fear conditioning and, subsequently, can
be silenced in response to fear extinction (Amano et al., 2011;
Herry et al., 2008; Hobin et al., 2003; Livneh and Paz, 2012;
Repa et al., 2001). However, the precise mechanisms through
which the extinction circuit achieves the extinction-induced
silencing of amygdala fear neurons are not fully understood.
Modification of synaptic input, by either decreasing excitatory
input or increasing inhibitory input, is a candidate mechanism.
The importance of inhibitory synaptic plasticity is increasingly
being appreciated (Kullmann et al., 2012), and inhibitory plasticity
has been implicated in fear extinction (Ehrlich et al., 2009;
Makkar et al., 2010).
In this study, we employed an imaging approach to identify the
precise location of basal amygdala (BA) fear neurons that are
silenced by contextual fear extinction and determine how these
fear neurons are silenced. We previously imaged BA fear neurons
with a transgenic mouse that uses tetracycline-controlled
tagging (TetTag) of neurons activated during fear conditioning
(Tayler et al., 2013; Reijmers et al., 2007). Here, we utilize the TetTag
mouse to image BA fear neurons that are silenced by extinction.
We find evidence for structural plasticity of perisomatic
inhibitory synapses originating from parvalbumin-positive interneurons
after silencing of BA fear neurons by fear extinction.
Importantly, these parvalbumin-positive synapses were located
immediately around the soma of the silenced BA fear neurons,
revealing an anatomical and functional connection between the
1054 Neuron 80, 1054–1065, November 20, 2013 ª2013 Elsevier Inc.
extinction circuit and the fear circuit. In addition, fear extinction
altered the presence of perisomatic endocannabinoid receptors
around the soma of BA fear neurons that remained active after
fear extinction. Our findings provide insight into the mechanisms
underlying the silencing of fear circuits and, more generally, add
to our knowledge of how behavior can sculpt activity and connectivity
within a neural circuit.
RESULTS
Fear Extinction Silences the Basal Amygdala Fear
Memory Circuit
Prior electrophysiological studies have revealed that fear extinction
can decrease the firing of BA fear neurons (Amano et al.,
2011; Herry et al., 2008; Livneh and Paz, 2012), but the underlying
mechanisms are not fully understood. We took advantage of
a c-fos-based reporter mouse, the TetTag mouse (Reijmers
et al., 2007), to image the effect of contextual fear extinction
on BA fear neuron activation. The TetTag mouse expresses
long-lasting nuclear GFP under control of the c-fos promoter
(Figure 1A), which enabled us to tag excitatory neurons activated
during fear conditioning (i.e., fear neurons, Figures 1B, S1A, and
S1B available online). The expression of the immediate-early
gene zif268/egr1 (Zif) served as a marker for neurons activated
during a later retrieval test (Okuno, 2011; Reijmers et al., 2007)
(Figures 1C and S1C). Two groups of TetTag mice (fear conditioned
[FC], n = 15; FC followed by extinction [FC+EXT], n =
17) were subjected to contextual fear conditioning (Figures 1C
and 1D). As expected, similar numbers of BA fear neurons
were tagged with GFP in both groups (Figure 1E). The next
2 days, only one group (FC+EXT) was subjected to extinction
trials, while the other group (FC) remained in the home cage.
Extinction caused the loss of fear expression as indicated by
the total suppression of freezing at the end of the extinction
procedure (Figure 1F) and during retrieval on day 4 (Figure 1G).
Extinction had no significant effect on the activation of BA neurons
that were not tagged during fear conditioning (GFPÀZif+,
Figure 1H). We analyzed the BA fear memory circuit by defining
two types of fear neurons (i.e., tagged during fear conditioning):
silent fear neurons (GFP+ZifÀ, fear neurons not reactivated during
retrieval) and active fear neurons (GFP+Zif+, fear neurons
reactivated during retrieval) (Figure 1I). We found that fear
A
TetTag mouse
0
10
20
30
***
cfos-P
tetO
FC
FC+EXT
0.0
0.4
0.8
1.2
1.6
BA
All GFP+
G
tTA
Histone2B-GFP
Retrieval
C
E
Frngeezi(%)
50
40
FC
FC+EXT
+DOX x - DOX
Extinction learningFear conditioningD F
0
20
40
60
***
°°°
Frngeezi(%)
E1 E2 E8E3 E4 E5 E6 E7S1
0
10
20
30
40
50
60
70
***
***
Frngeezi(%)
S2 S3
FC
FC+EXT
°°°
°°° FC+EXT
FC
FC+EXT
Fear conditioning
Extinction Extinction
Day 1 Day 2 Day 3 Day 4
- DOX
GFP expression
(GFP+)
FC groupRetrieval
Retrieval FC+EXT group
GFP+Zif+
Zif expression
( )Zif+
GFP
CamKII
Overlay
B
0
1
2
3
4
5
0
5
10
15
20
***
I
BABA
**
***
Chance
Chance
GFP-Zif+
GFP+Zif+
JH
Chance
FC
FC+EXT FC
FC+EXT
FC FC+EXT
BA
+ DOX + DOX
GFP+Zif+DAPI+
%Zif+amongGFP+neurons
%Zif+amongGFP-neurons
%GFP+amongDAPIcells
E1 E2 E3 E4
S1 S2 S3
E5 E6 E7 E8
Figure 1. Fear Extinction Silences the Basal
Amygdala Fear Memory Circuit
(A) A double transgenic TetTag mouse line was
used that expresses the tetracycline transcriptional
activator (tTA) under control of the activityregulated
c-fos promoter. In the absence of
doxycycline (ÀDOX), tTA binds to the tet operator
(tetO) in the second transgene and induces
expression of a long-lasting histone2B-GFP
fusion protein.
(B) Image of an excitatory basal amygdala (BA)
neuron tagged with long-lasting histone2B/GFP
(green) and immunolabeled with a calcium/
calmodulin-dependent kinase II (CamKIIa) antibody
(red). Scale bar, 10 mm.
(C) Schema of the experimental procedure.
‘‘ÀDOX’’ opened a time window for tagging fear
conditioning-activated neurons (GFP+; i.e., fear
neurons) in two experimental groups (FC, n = 15;
FC+EXT, n = 17). Zif expression during retrieval
(Zif+) was used to detect the reactivation of tagged
fear neurons (GFP+Zif+; i.e., active fear neurons).
The absence of Zif in GFP+ neurons was used to
identify silent fear neurons (GFP+ZifÀ).
(D) As training progressed, FC and FC+EXT mice
showed an increase in their level of freezing
(between sessions: ***p < 0.001 for FC and 
p <
0.001 for FC+EXT).
(E) FC and FC+EXT groups had similar percentages
of GFP+ neurons (sum of GFP+ZifÀ and
GFP+Zif+) in the BA (p = 0.49).
(F) The FC+EXT group showed a significant decrease in freezing during the extinction sessions on days 2 and 3 (***p < 0.001 E1 versus others and 
p < 0.001 E2
versus others).
(G) Freezing during retrieval on day 4 in the FC+EXT group was absent and was significantly lower compared to the FC group (p = 0.00014).
(H) There was no significant difference in the percentage of Zif+ among GFPÀ neurons between the two groups (p = 0.064).
(I) Representative image of GFP-positive (GFP+, green) and Zif-positive (Zif+, red) nuclei in the basal amygdala (BA) of an FC (left) and an FC+EXT mouse (right).
Arrows indicate active fear neurons (GFP+Zif+) and arrowheads indicate silent fear neurons (GFP+ZifÀ). Scale bar, 50 mm.
(J) Extinction decreased the number of active BA fear neurons, as indicated by FC+EXT mice having less GFP+Zif+ neurons than FC mice (p = 0.00023). In both
groups, the percentage of GFP+ZIF+ neurons was higher than chance level (FC: p = 0.0000098; FC+EXT p = 0.0014). Chance level was determined by using the
percentage of Zif+ among GFPÀ neurons as shown in (H). Graphs show means ± SEM. **p < 0.01, ***p < 0.001.
See also Figure S1.
Neuron
Extinction Causes Perisomatic Synapse Remodeling
Neuron 80, 1054–1065, November 20, 2013 ª2013 Elsevier Inc. 1055
extinction caused a 2.3-fold decrease in the number of active BA
fear neurons, with a subgroup of fear neurons remaining active
after extinction (GFP+Zif+; Figures 1J and S1D). This extinction-induced
silencing of the BA fear memory circuit, caused
by contextual fear extinction, is similar to that previously found
in electrophysiological studies on tone fear extinction (Amano
et al., 2011; Herry et al., 2008; Livneh and Paz, 2012). In addition,
the observed concomitant reduction in active BA fear neurons
and freezing replicates previous studies (Herry et al., 2008;
Reijmers et al., 2007) and reflects the causative role of the basal
amygdala in the behavioral expression of contextual fear (Maren,
1998). Therefore, our results provide further support for a model
in which extinction decreases fear by silencing the fear memory
circuit within the BA.
Activation of Brain Regions Upstream of the Basal
Amygdala Is Not Altered after Fear Extinction
We next explored where extinction acted to cause a silencing of
the BA fear memory circuit. We first addressed the possibility
that contextual fear extinction might act on brain regions upstream
of the BA and thereby indirectly silence fear neurons in
the BA. The BA receives inputs from the CA1 area of the hippocampus
and from the infralimbic prefrontal cortex, brain regions
that have been implicated in fear extinction (Hartley and Phelps,
0
5
10
15
20
25
0
2
4
6
8
C
D
dCA1 vCA1
A
0
5
10
15
dCA1
0
1.0
vCA1
0.5
GFP-Zif+
FC
FC+EXT FC
FC+EXT
FC
FC+EXT FC
FC+EXT
0
2
4
6
IL
FC
FC+EXT
IL
FC
FC+EXT
0
5
10
All GFP+
B
dCA1
vCA1
FC FC+EXT
IL
%GFP+amongDAPIcells
%GFP+amongDAPIcells
%Zif+amongGFP-neurons%GFP+amongDAPIcells
%Zif+amongGFP-neurons
%Zif+amongGFP-neurons
GFP+Zif+DAPI+GFP+Zif+DAPI+GFP+Zif+DAPI+
0
10
20
30
40
dCA1
***
Chance
Chance
**
0
1
2
3
4
vCA1
**
Chance
*
Chance
n.s. n.s.
FC FC
FC+EXT
FC+EXT
IL
***
Chance
**
Chance
GFP+Zif+
Chance
n.s.
FC
FC+EXT
0
10
15
20
25
5
%Zif+amongGFP+neurons
%Zif+amongGFP+neurons
%Zif+amongGFP+neurons
Figure 2. Activation of Brain Regions Upstream
of the Basal Amygdala Is Not Altered
after Fear Extinction
(A) The FC and FC+EXT groups had similar percentages
of GFP+ neurons in the dorsal CA1
(dCA1), ventral CA1 (vCA1), and infralimbic cortex
(IL) (dCA1: p = 0.44; vCA1: p = 0.50; IL: p = 0.17).
(B) Representative image of GFP+ neurons (green)
and Zif+ neurons (red) in the dCA1 (top), vCA1
(middle),andIL(bottom)ofanFCmouse(left)andan
FC+EXT mouse (right). Blue, DAPI. Arrows indicate
active fear neurons (GFP+Zif+). Scale bar, 50 mm.
(C) No significant differences in the percentage of
Zif+ among GFPÀ neurons were found (dCA1: p =
0.25; vCA1: p = 0.21; IL: p = 0.056).
(D) Extinction had no effect on reactivation of dCA1,
vCA1, and IL. The numberof GFP+Zif+ neurons was
similar between the two groups (dCA1: p = 0.26;
vCA1: p = 0.61; IL: p = 0.27). The number of GFP+
Zif+ neurons was above chance level in the dCA1,
vCA1, and IL of both groups (dCA1: FC versus
chance, p = 0.00017; FC-EXT versus chance, p =
0.000017; vCA1: FC versus chance, p = 0.0027; FCEXT
versus chance, p = 0.0072; IL: FC versus
chance, p = 0.000062; FC-EXT versus chance, p =
0.0026). Chance level was determined by using the
percentage of Zif+ among GFPÀ neurons as shown
in (C). Graphs show means ± SEM. n.s., not significant;
*p < 0.05, **p < 0.01, ***p < 0.001.
2010; Orsini and Maren, 2012). We therefore
tested whether fear extinction altered
the activation of excitatory neurons in the
CA1 region of the hippocampus (both
dorsal and ventral: dCA1 and vCA1) and
in the infralimbic prefrontal cortex (IL).
We analyzed the same brains that were used for the BA analysis,
since the TetTag mouse enables the tagging of neurons recruited
by fear conditioning throughout the whole brain (Deng et al.,
2013; Garner et al., 2012; Liu et al., 2012; Matsuo et al., 2008;
Reijmers et al., 2007; Tayler et al., 2011, 2013). As expected,
the FC and FC+EXT groups had similar percentages of neurons
tagged in these brain regions during contextual fear conditioning
(Figures 2A and 2B). Contextual fear extinction did not alter the
activation of nontagged dCA1 and vCA1 neurons during retrieval
(Figure 2C) or the reactivation of tagged dCA1 and vCA1 neurons
during retrieval (Figures 2B and 2D). These results are consistent
with a previous study reporting that contextual fear extinction
acts on a population of CA1 neurons that is segregated from
the CA1 neurons recruited during fear conditioning (Tronson
et al., 2009) and indicate that the memory of the context is
retained after contextual fear extinction (Rudy et al., 2004).
Similar to the hippocampal CA1 neurons, the activation of nontagged
IL neurons (Figure 2C) and the reactivation of tagged IL
neurons (Figures 2B and 2D) were not affected by contextual
fear extinction. Overall, we did not detect extinction-induced
functional changes in two important brain structures upstream
of the BA. We therefore shifted our focus to potential local
changes within the BA that might have caused the silencing of
the BA fear memory circuit.
Neuron
Extinction Causes Perisomatic Synapse Remodeling
1056 Neuron 80, 1054–1065, November 20, 2013 ª2013 Elsevier Inc.
Fear Extinction Causes Target-Specific Remodeling of
Perisomatic Inhibitory Synapses in the Basal Amygdala
Around 85% of the neuronal cell population within the BA consists
of excitatory projection neurons, whereas the remaining
15% are local interneurons that make inhibitory synapses onto
the projection neurons (McDonald, 1992). Because BA inhibitory
interneurons have been implicated in fear extinction (Ehrlich
et al., 2009; Heldt and Ressler, 2007), we addressed the possibility
that structural changes involving inhibitory circuits in the BA
might have caused the extinction-induced silencing of BA fear
neurons by increasing local inhibition. We first examined the
expression of 67 kDa glutamic acid decarboxylase (GAD67), a
key enzyme in GABA synthesis. Both GAD67 and the smaller isoform
GAD65 have been implicated in fear extinction, but a specific
role within the amygdala has so far only been established
for GAD67 (Heldt et al., 2012; Sangha et al., 2009). We did not
find evidence for increased GAD67 expression in either the
complete BA or in the soma of BA interneurons (Figures 3A,
3B, and 3C), consistent with a recent study (Sangha et al.,
2012). We hypothesized that fear extinction might act on a
synaptic site where local interneurons interface with the BA
fear neurons. We tested this hypothesis by imaging a special
type of inhibitory synapse called perisomatic synapse. Periso-
E
0.0
2.0
*
GAD67pixels/cellperimeter
D
0.0
2.0
FC
FC+EXT FC
FC+EXT
0.5
1.0
1.5
0.5
1.0
1.5
GAD67pixels/cellperimeter
FC FC+EXT FC FC+EXT
Zif-GAD67+MaskofGAD67GFP+GAD67+
GFP+Zif+GFP+Zif-
Zif+GAD67+MaskofGAD67GFP+GAD67+
0.87 1.37 1.22 1.23
BA
FC+EXT
Fluorescenceintensity
(grayscalevalue)
0
50
100
150
FC
Global GAD67
expression in BA
GAD67 in soma of
BA interneurons
Fluorescenceintensity
(grayscalevalue)
0
50
100
150
FC
FC+EXT
A B
C
Gad67 soma
GAD67
Figure 3. Fear Extinction Causes TargetSpecific
Remodeling of Perisomatic Inhibitory
Synapses in the Basal Amygdala
(A) Image of glutamic acid decarboxylase-67
(GAD67) immunolabeling of the basal amygdala
(BA; scale bar, 100 mm). Arrows indicate GAD67positive
(GAD67+) soma in the BA. The inset
shows a representative image of a GAD67+ interneuron
soma (scale bar, 10 mm).
(B) Extinction had no effect on GAD67 expression
in the total BA (p = 0.30).
(C) Extinction had no effect on somatic expression
of GAD67 in the BA (p = 0.16).
(D and E) Left: representative images of perisomatic
GAD67 immunolabeling around silent BA
fear neurons (D, GFP+ZifÀ) and active BA fear
neurons (E, GFP+Zif+). Three images are shown
for each fear neuron (top: GFP in green, GAD67 in
gray; middle: Zif in red, GAD67 in gray; bottom:
GAD67 mask in black, value for that neuron
obtained by dividing black pixels by perimeter of
yellow outline). Scale bars, 10 mm. Right: extinction
increased perisomatic GAD67 around silent fear
neurons (D, p = 0.047) but not around active fear
neurons (E, p = 0.46). Graphs show means ± SEM.
*p < 0.05.
See also Figure S2.
matic inhibitory synapses are a plausible
candidate for silencing BA fear neurons,
since they are well positioned to modulate
the functional activation of excitatory
neurons (Miles et al., 1996). Consistent
with our hypothesis, we found that silent
fear neurons had increased GAD67
around their soma after extinction (Figure
3D). Interestingly, this increase in perisomatic GAD67 was
not observed around active fear neurons (Figure 3E). The selective
increase in perisomatic GAD67 around silent fear neurons
seemed to be caused by a selective increase in the number of
inhibitory synapses (Figures S2A and S2B). Thus, our data reveal
that extinction can cause the target-specific remodeling of perisomatic
inhibitory synapses in the BA, with extinction-induced
changes in perisomatic GAD67 matching the activation states
of the postsynaptic fear neurons.
Fear Extinction Increases Perisomatic Parvalbumin
around Silent BA Fear Neurons
We decided to further investigate the nature of the extinctioninduced
remodeling of perisomatic inhibitory synapses in the
BA. Parvalbumin-positive interneurons (PV+) are the predominant
interneuron type in the BA and form perisomatic synapses
around BA excitatory projection neurons (McDonald and
Betette, 2001). Extinction did not change the expression of
PV in the soma of BA interneurons (Figures 4A and 4B). Next,
we analyzed the presence of PV around the soma of BA fear
neurons. We verified that our perisomatic PV immunolabeling
represented perisomatic synapses (Figure S3A). Consistent
with the extinction-induced increase in perisomatic GAD67,
Neuron
Extinction Causes Perisomatic Synapse Remodeling
Neuron 80, 1054–1065, November 20, 2013 ª2013 Elsevier Inc. 1057
extinction also increased perisomatic PV around the silent fear
neurons (Figure 4C). Again, there was no significant increase
around the active fear neurons (Figure 4D). The effects of
extinction on perisomatic PV seemed to reflect changes in synapse
numbers (Figures S3B and S3C). Importantly, the increase
in perisomatic PV that we detected with image analysis is
similar to that reported to increase perisomatic inhibition
using electrophysiological analysis (Gittis et al., 2011; Kohara
et al., 2007). Thus, our data suggest an extinction-induced
increase in perisomatic inhibition underlies the decreased
number of active BA fear neurons and the resulting silencing
of the fear memory circuit. This reveals a direct connection
between extinction-induced structural and functional changes
in the BA.
The Extinction-Induced Increase in Perisomatic
Parvalbumin Reflects New Learning
We asked whether the extinction-induced increase in perisomatic
PV might have reversed any fear conditioning-induced
changes in those synapses, which would indicate that BA
perisomatic inhibitory synapses were part of the original fear
circuit. To address this question, we performed a separate
experiment in which we compared a fear conditioned group
(FC) with a home cage group (HC) (Figures 5A and 5B). Consistent
with our previous study (Reijmers et al., 2007), BA neurons
activated during fear conditioning were tagged with long-lasting
expression of GFP (Figure 5C). During retrieval on day 4,
the FC group showed significant freezing (Figure 5D). The
retrieval of contextual fear caused activation of both nontagged
(GFPÀZif+; Figure 5E) and tagged (GFP+Zif+; Figure 5F) neuPV
in soma of
BA interneurons
0
50
100
150
FC
FC+EXT
Fluorescenceintensity
(grayscalevalue)
0.0
0.5
1.0
1.5
**
PVpixels/cellperimeter
0.0
0.5
1.0
1.5
PVpixels/cellperimeter
FC
FC+EXT FC
FC+EXT
FC FC+EXT FC FC+EXT
0.58 0.51
GFP+Zif+GFP+Zif-
Zif-PV+MaskofPVGFP+PV+
Zif+PV+MaskofPVGFP+PV+
1.09 0.69
C D
A B
PV
Figure 4. Fear Extinction Increases Perisomatic
Parvalbumin around Silent BA Fear
Neurons
(A) Representative image of a parvalbumin-positive
(PV+) interneuron soma in the basal amygdala
(BA; scale bar, 10 mm).
(B) Extinction had no effect on somatic expression
of PV in the BA (p = 0.064).
(C and D) Left: representative images of perisomatic
PV immunolabeling around silent BA fear
neurons (C, GFP+ZifÀ) and active BA fear neurons
(D, GFP+Zif+). Three images are shown for each
fear neuron (top: GFP in green, PV in gray; middle:
Zif in red, PV in gray; bottom: PV mask in black,
value for that neuron obtained by dividing black
pixels by perimeter of yellow outline). Scale bars,
10 mm. Right: extinction increased perisomatic PV
around silent fear neurons (C, p = 0.0061) but not
around active fear neurons (D, p = 0.18). Graphs
show means ± SEM. **p < 0.01.
See also Figure S3.
rons in the BA, with a preferential reactivation
of the tagged BA fear neurons
(Figures 5E and 5F). Importantly, we
did not find fear conditioning-induced
changes in perisomatic PV around silent
or active fear neurons (Figures 5G and
5H). These data strongly suggest that
the extinction-induced changes in PV+ perisomatic synapses
constituted a new form of learning that occurred within the
extinction circuit.
Fear Extinction Increases Perisomatic CB1R around
Active BA Fear Neurons
In addition to PV+ perisomatic synapses, the BA also contains
perisomatic inhibitory synapses that originate from cholecystokinin
(CCK) interneurons (Yoshida et al., 2011). We therefore
examined whether fear extinction also affected perisomatic
CCK+ synapses. Extinction did not change the expression of
CCK in the soma of BA interneurons (Figures 6A and 6B). In
addition, perisomatic CCK around fear neurons, either silent
or active, was not altered by fear extinction (Figures 6C, 6D,
S4A, and S4B). Fear conditioning itself also did not change
perisomatic CCK in the BA (Figures S4C and S4D). We next
examined whether extinction might have affected CCK+ perisomatic
synapses without changing CCK levels. Recently, it was
discovered that BA perisomatic synapses positive for CCK,
but not PV, contain a unique enrichment of proteins involved
in endocannabinoid signaling, including cannabinoid receptor
type 1 (CB1R) (Yoshida et al., 2011). Since CB1R in the BA
have been implicated in fear extinction (Marsicano et al.,
2002), we examined whether perisomatic CB1R presence was
modulated by extinction and whether this modulation was
target-specific. We first confirmed that in the BA our CB1R immunolabeling
colocalized with our CCK, but not PV, immunolabeling
(Figure 6E) and that CB1R and CCK colocalized
directly adjacent to the soma (Figure S4E). We did not observe
labeling of CB1R in the soma of BA interneurons, so somatic
Neuron
Extinction Causes Perisomatic Synapse Remodeling
1058 Neuron 80, 1054–1065, November 20, 2013 ª2013 Elsevier Inc.
expression of CB1R was not quantified. Perisomatic CB1R
presence around the silent fear neurons was similar in the FC
and FC+EXT groups (Figures 6F and S4F). Intriguingly, extinction
increased perisomatic CB1R around the active fear neurons
(Figures 6G and S4G). This effect of extinction appeared
to constitute a new form of learning, as fear conditioning itself
did not change perisomatic CB1R in the BA (Figures S4H and
S4I). Since CB1R inhibit the release of g-aminobutyric acid
(GABA) (Katona et al., 2001), these results suggest that the
extinction-induced upregulation of perisomatic CB1R facilitated
the persistence of a small subset of active BA fear neurons in
the extinction group (Figure 1J).
Fear Extinction Might Alter Perisomatic Inhibition
Outside of the Fear Circuit
The results of our analysis of the effects of extinction on silent
(GFP+ZifÀ) versus active (GFP+Zif+) fear neurons are summarized
in Figure 7. To explore whether fear extinction might also
alter perisomatic inhibition outside of the fear circuit, we quantified
perisomatic markers around GFPÀZif+ neurons in the BA.
We found that extinction increased PV and CB1R around
GFPÀZif+ cells but had no significant effect on perisomatic
GAD67 and CCK (Figure S5A). This result has two possible explanations
that are not mutually exclusive. First, it suggests
that fear extinction might alter perisomatic inhibition around BA
neurons that are not part of the fear circuit. One possibility is
that fear extinction also changes perisomatic inhibition of BA
neurons that are part of the extinction circuit (Herry et al.,
2008). These extinction neurons were reported to be silent during
fear conditioning and subsequently activated during extinction,
so they would not be tagged with GFP in our experimental
design. Second, some GFPÀZif+ neurons might actually be
fear neurons that were not tagged with GFP. For example, neurons
with a relatively low level of c-fos promoter activation during
fear conditioning might express c-fos protein and tTA protein,
but the tTA protein level might be too low to activate the tetO promoter
and trigger GFP expression. We tried to find support in
favor of one of these two explanations by correlating the number
of GFPÀZif+ neurons with freezing levels during the extinction
trials. If a significant portion of the GFPÀZif+ neurons were
extinction neurons, then a negative correlation with freezing during
extinction might be observed. On the other hand, if a significant
portion of the GFPÀZif+ neurons were nontagged active
fear neurons, then a positive correlation with freezing during
extinction might be observed similar to the positive correlation
found for GFP+Zif+ neurons (Figure S1D). We did not find a significant
correlation, either positive or negative (Figure S5B). This
suggests that GFPÀZif+ neurons might consist of a mix of
neurons with varying functions. Table S1 summarizes the extinction-induced
perisomatic changes observed around the three
types of BA neurons, showing that the changes around
GFPÀZif+ neurons differ from the changes around fear neurons,
either silent (GFP+ZifÀ) or active (GFP+Zif+). The different
perisomatic profiles around the three BA cell types illustrate
the target-specific nature of fear extinction-induced perisomatic
synapse remodeling.
DISCUSSION
Our findings reveal that remodeling of perisomatic inhibitory synapses
located immediately around fear neurons in the basal
amygdala occurs during fear extinction. These perisomatic synapses
represent a site where the circuits for fear extinction and
fear storage connect. The direct anatomical and functional relationship
between the perisomatic synapses and the fear neurons
suggests a straightforward mechanism for the silencing of fear
circuits. Perisomatic inhibitory synapses therefore provide an
attractive therapeutic target for improving the efficacy of fear
extinction in humans treated with exposure therapy. In addition,
we found that extinction might alter perisomatic inhibition
outside of the fear circuit, possibly contributing to the behavioral
B Fear conditioning
0
BA
S1
0
10
20
30
40
50
60
70
***
***
Freezing(%)
S2 S3
FC
Retrieval
Freezing(%)
0
FC
10
20
30
40
5
10
15
20
Day 1
+ DOX + DOX- DOX
Home cage
Retrieval
HC group
FC group
HC FC
**
0.0
0.2
0.4
0.6
0.0
0.2
0.4
0.6
Day 2 Day 3 Day 4
%GFP+amongDAPIcells
0.0
0.5
1.0
1.5
HC FC
**
**
0
1
2
3
BA
GFP-Zif+
*
Chance
Chance
n.s.
GFP+Zif+
Chance
GFP+Zif+GFP+Zif-
*
All GFP+
A
G H
D EC
F
BA
HC FC
HC FC HC FC
%Zif+amongGFP-neurons
%Zif+amongGFP+neurons
PVpixels/cellperimeter
PVpixels/cellperimeter
Fear conditioning
S1 S2 S3
Figure 5. The Extinction-Induced Increase in Perisomatic Parvalbumin
Reflects New Learning
(A) Design of an experiment with a home cage (HC, n = 8) and a fear-conditioned
group (FC, n = 8).
(B) Freezing during contextual fear conditioning on day 1. As training
progressed, FC mice showed an increase in their level of freezing (p =
0.0006).
(C) The FC group had a larger number of GFP+ neurons than the HC group in
the basal amygdala (BA, p = 0.0049).
(D) FC mice showed significant freezing during retrieval on day 4 (p = 0.0077).
(E) The FC group had a larger number of GFPÀZif+ neurons than the HC group
(p = 0.036).
(F) The FC group had a larger number of GFP+Zif+ neurons than the HC group
(p = 0.0043). Only in the FC group was the percentage of GFP+ZIF+ neurons
higher than chance level, confirming that a subset of BA fear neurons
was reactivated during retrieval (HC: p = 0.23; FC: p = 0.016). Chance level
was determined by using the percentage of Zif+ among GFPÀ neurons as
shown in (E).
(G and H) Fear conditioning had no effect on perisomatic PV around either type
of tagged BA neuron. (G) GFP+ZifÀ, p = 0.10. (H) GFP+Zif+, p = 0.49. Graphs
show means ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001.
Neuron
Extinction Causes Perisomatic Synapse Remodeling
Neuron 80, 1054–1065, November 20, 2013 ª2013 Elsevier Inc. 1059
effects of extinction by altering perisomatic inhibition of extinction
neurons (Herry et al., 2008). The fine-tuned nature of the
observed perisomatic synapse remodeling provides an important
insight into how behavior can sculpt the flow of information
in the brain.
Notably, the extinction-induced remodeling of perisomatic
synapses was interneuron and target-neuron specific, and the
predicted changes in the balance of perisomatic inhibition
matched the state of the target fear neurons in two ways (Figure
7). First, the silent state of fear neurons (GFP+ZifÀ) corresponded
to an extinction-induced increase in perisomatic PV,
which is predicted to increase perisomatic inhibition (Gittis
et al., 2011; Kohara et al., 2007). Second, the active state of
fear neurons (GFP+Zif+) corresponded to an extinction-induced
increase in perisomatic CB1R. We propose that the CB1R increase
prevented a subset of active fear neurons from switching
into silent fear neurons by decreasing GABA release from CCK
terminals (Katona et al., 2001). The CB1R-mediated inhibition
of GABA release around active fear neurons could strengthen
their future activation by facilitating long-term potentiation
of excitatory synapses (Carlson et al., 2002). Therefore, an
0.0
0.5
CB1Rpixels/cellperimeter
1.0
1.5
F
0.0
0.5
CB1Rpixels/cellperimeter
1.0
1.5
*
G
FC
FC+EXTFC
FC+EXT
FC FC+EXT FC+EXT
0.66 0.72
CB1R CCK Overlay
CB1R PV Overlay
E
FC
GFP+Zif- GFP+Zif+
Zif-CB1R+MaskofCB1RGFP+CB1R+
Zif-CB1R+MaskofCB1RGFP+CB1R+
0.0
0.2
0.4
0.6
0.8
CCKpixels/cellperimeter
0.0
0.2
0.4
0.6
0.8
CCKpixels/cellperimeter
FC
FC+EXTFC
FC+EXT
C D
GFP+Zif- GFP+Zif+
0.67 1.10
A CCK in soma of
BA interneurons
0
50
100
150
FC
FC+EXT
Fluorescenceintensity
(grayscalevalue)
B
CCK
Figure 6. Fear Extinction Increases Perisomatic
CB1R around Active BA Fear Neurons
(A) Representative image of cholecystokinin-positive
(CCK+) interneuron soma in the basal amygdala
(BA; scale bar, 10 mm).
(B) Extinction had no effect on somatic expression
of CCK in the BA (p = 0.25).
(C and D) Extinction had no effect on perisomatic
CCK around silent (C, GFP+ZifÀ, p = 0.25) or
active BA fear neurons (D, GFP+Zif+, p = 0.42).
(E) Top: colocalization of perisomatic CB1R (red)
and perisomatic CCK (gray) in the BA. Bottom: no
colocalization of perisomatic CB1R (red) and perisomatic
PV (gray) was detected in the BA. Scale
bars, 10 mm.
(F and G) Left: representative images of perisomatic
CB1R immunolabeling around silent BA
fear neurons (F, GFP+ZifÀ) and active BA fear
neurons (G, GFP+Zif+). Three images are shown
for each fear neuron (top: GFP in green, CB1R in
gray; middle: Zif in red, CB1R in gray; bottom:
CB1R mask in black, value for that neuron
obtained by dividing black pixels by perimeter of
yellow outline). Scale bars, 10 mm. Right: extinction
increased perisomatic CB1R expressed in CCK
terminals around active fear neurons (G, p = 0.036)
but not around silent fear neurons (F, p = 0.19).
Graphs show means ± SEM. *p < 0.05.
See also Figure S4.
intriguing possibility is that BA fear neurons
that remain active after contextual
fear extinction might, over time, reawaken
the fear circuit and limit the effectiveness
of exposure therapy by triggering spontaneous
recovery (Myers and Davis, 2007).
The subset of fear neurons that remained
active after extinction did not
trigger freezing during the retrieval test
(Figures 1G and 1J). This suggests that, in addition to BA perisomatic
synapses, an additional downstream site exists where the
extinction circuit inhibits the fear circuit. This downstream site
might be located in the central amygdala, which contains neurons
that mediate the effects of BA fear neurons on freezing.
Previous studies support a model in which central amygdala
neurons are inhibited by intercalated interneurons, which
become more active as a result of infralimbic prefrontal cortex
activation during extinction (Amano et al., 2010; Likhtik et al.,
2008; Milad and Quirk, 2002). Though we did not observe extinction-induced
changes in the activation of brain regions upstream
of the basal amygdala (Figure 2), a role for these upstream brain
regions in fear extinction remains probable. For example, recent
studies have found that projections from the prefrontal cortex
and the hippocampus to the basal and lateral amygdala can
regulate to what extent an extinguished fear memory is retrieved
(Knapska et al., 2012; Orsini et al., 2011). It needs to be determined
how the numerous neural circuits involved in fear
extinction, located in various brain regions such as the prefrontal
cortex, hippocampus, and amygdala, work together to silence
the fear circuit. We propose that the approach used in this study
Neuron
Extinction Causes Perisomatic Synapse Remodeling
1060 Neuron 80, 1054–1065, November 20, 2013 ª2013 Elsevier Inc.
can be more widely applied for this purpose. Identifying additional
elements of the fear circuit that are silenced by extinction
might enable the reconstruction of multiple functional extinction
circuits, each responsible for silencing a specific element of the
fear circuit. The discovery of structural changes in BA perisomatic
synapses lays the foundation for reconstructing at least
one coherent functional extinction circuit, with future studies
determining which neural circuits need to be recruited during
extinction to enable the target-specific remodeling of perisomatic
synapses around BA fear neurons.
Does fear extinction decrease fear by suppressing or erasing
the fear memory circuit? If extinction-induced changes in perisomatic
inhibitory synapses constitute a form of erasure, then they
should reverse changes induced by fear conditioning at these
synapses. We did not find evidence for this, as fear conditioning
itself did not change the perisomatic presence of PV, CCK, and
CB1R. This strongly suggests that the observed changes in perisomatic
synapses constituted a new form of learning that led to
suppression of the fear memory circuit. This leaves open the
possibility that in our study the fear memory circuit remained
completely intact, for example, by retaining a pattern of strengthened
and weakened excitatory synapses induced by fear conditioning.
In contrast with our findings, two recent papers reported
examples of possible erasure of components of the fear memory
circuit. One study using mice found that extinction reversed
changes in dendritic spines that were induced by fear conditioning
(Lai et al., 2012). It should be noted that the reported spine
dynamics occurred in the frontal association cortex, a brain region
that has not been firmly established yet as an essential
component of the fear memory circuit. Nevertheless, this study
provides an important first step toward identifying a mechanism
by which fear memory circuits can be erased. Another recent
study using human subjects reported that a certain behavioral
extinction protocol, in which extinction follows a retrieval trial,
can erase a memory trace in the amygdala (Agren et al., 2012).
However, in this study, the erasure of the memory trace was inferred
from changes in the activation state of the complete basolateral
amygdala. Our data illustrate how extinction-induced
changes in local inhibition within the basal amygdala might alter
the activation state of the complete brain region without erasing
the fear memory circuit, in which case it should be considered
suppression. The question of suppression versus erasure has
important implications for the treatment of fear disorders, as a
treatment based on a form of erasure might make the return of
debilitating fear less likely. Future studies using animal models
will be invaluable to address the suppression versus erasure
distinction, because validating a true mechanism for fear memory
erasure will require more data collected at the cellular, subcellular,
and ultimately the molecular level.
Our findings shed light on two proposed molecular mechanisms
of extinction. Studies in humans and rodents have found
that both CB1R (Gunduz-Cinar et al., 2013; Heitland et al., 2012;
Marsicano et al., 2002; Rabinak et al., 2013) and brain-derived
neurotrophic factor (BDNF) (Andero et al., 2011; Chhatwal et al.,
2006; Soliman et al., 2010) signaling in the BA support fear
extinction. CB1R and BDNF signaling can both occur at inhibitory
and excitatory synapses, and it is unclear which synapse
type mediates their effects on fear extinction. In the case of
CB1R signaling, the perisomatic CCK+ inhibitory synapses provide
a plausible site of action, since the major components of
CB1R signaling in the BA are highly enriched and colocalized
in these synapses (Yoshida et al., 2011). However, the increase
in perisomatic CB1R around the remaining active fear neurons
seems in contradiction with a potential role for perisomatic
CB1R signaling in the reduction of fear. We found that extinction
might also increase CB1R outside of the fear circuit. If
this increase occurred around extinction neurons (Herry et al.,
2008), then it might have contributed to the increased activation
of extinction neurons. A fear extinction role for CB1R at perisomatic
inhibitory synapses would be in agreement with a recent
finding that expression of CB1R at excitatory synapses is not
sufficient to support fear extinction (Ruehle et al., 2013). In the
case of BDNF, it is interesting to note that postsynaptic release
of BDNF promotes the formation of perisomatic PV+ synapses
in the cortex (Hong et al., 2008; Huang et al., 1999; Jiao et al.,
2011; Kohara et al., 2007). We therefore propose that BDNF
signaling in the BA supports fear extinction by increasing the
number of perisomatic PV+ synapses around BA fear neurons,
which is predicted to increase perisomatic inhibition (Gittis
et al., 2011; Kohara et al., 2007). A better understanding
of the molecular mechanisms used by BDNF to increase PV+
perisomatic synapse numbers could lead to new therapeutic
targets for the treatment of fear disorders. Though BDNF acts
on many types of synapses, both inhibitory and excitatory, it
seems to use different signaling pathways within each type of
synapse (Gottmann et al., 2009; Matsumoto et al., 2006). It
is therefore feasible that targets will be identified that specifically
modulate the effect of BDNF on perisomatic inhibitory
synapses.
Silent Active
PV
PV PV
--
-
-
Active
ActiveSilent Silent
-
-
Extinction
Freezing
CB1R
CB1R
-
-
-
CB1R
CB1R
CB1R CB1R
--
-
-
-
-
-
-
-
-
PV
-
PV
-
PV
No
freezing
-
-
- -
- -
Figure
7. Model Connecting Extinction-Induced Functional and
Structural Changes in the Basal Amygdala
Extinction increases the ratio of silent (GFP+ZifÀ) versus active (GFP+Zif+) fear
neurons in the basal amygdala (BA), thereby causing the elimination of
behavioral expression of fear as indicated by the absence of freezing during
retrieval. In this model, the increased number of silent fear neurons is caused
by increased perisomatic inhibition from PV neurons. The increase in CB1R
occurs around a selective subset of BA fear neurons that as a result remain
active because of increased CB1R-mediated disinhibition. Inhibition is represented
by the ‘‘À’’ sign.
See also Figure S5 and Table S1.
Neuron
Extinction Causes Perisomatic Synapse Remodeling
Neuron 80, 1054–1065, November 20, 2013 ª2013 Elsevier Inc. 1061
A potential role for inhibitory synapse plasticity in shaping patterns
of neural circuit activation has recently become more
appreciated (Kullmann et al., 2012). Inhibitory interneurons can
be highly interconnected, resulting in synchronized firing (Bartos
et al., 2007), and are in many brain regions outnumbered by
excitatory neurons, with a single interneuron contacting as
many as a 1,000 excitatory neurons (Miles et al., 1996). These
traits make inhibitory interneurons seem ill-suited to exert finely
targeted effects on individual excitatory neurons. The discovery
of various forms of inhibitory synapse plasticity has made clear
how inhibitory interneurons can specifically modulate the activation
of individual target neurons (Kullmann et al., 2012). Perisomatic
inhibitory synapses are especially well-positioned to
enable this ‘‘personalized inhibition’’ by using their ability to suppress
action potentials in the target neuron (Miles et al., 1996),
thereby functioning as a brake that keeps the excitatory ‘‘gas
pedal’’ in check. If perisomatic synapses indeed participate in
the fine-tuned sculpting of patterns of neural circuit activation,
then they should be subjected to forms of target-specific plasticity
so that two excitatory neurons receiving perisomatic
synapses from the same cluster of interneurons can be differentially
inhibited. Recently, target-specific properties have been reported
for perisomatic PV+ synapses in the striatum (Gittis et al.,
2011) and for perisomatic CCK+ synapses in the entorhinal cortex
(Varga et al., 2010). Our study adds to the understanding of
perisomatic synapse dynamics in three ways. First, our study
shows that target-specific changes can differentially affect PV
and CCK perisomatic synapses around the same type of neuron.
Second, we demonstrate target-specific modulation of perisomatic
CB1R. Last and most important, our study reveals that
behavior can trigger target-specific changes in perisomatic synapses.
Behavior-induced target-specific plasticity of perisomatic
synapses may be a central feature of neural circuits across
the brain.
In summary, we discovered that contextual fear extinction
causes the remodeling of perisomatic inhibitory synapses
located directly around fear neurons in the basal amygdala.
This discovery provides an anatomical and functional connection
between the extinction circuit and the fear circuit. Since
perisomatic synapses directly impinge on the fear circuit, they
provide an attractive target for modulating maladaptive fear. In
addition, our study reveals a mechanism by which behavior
can use inhibitory synapse plasticity to alter the flow of information
through the neural circuits. An important goal for future
studies will be to determine the extent to which silencing of BA
fear neurons is achieved by changes in perisomatic inhibitory
synapses versus changes in other inhibitory and excitatory synapses
and changes in neuronal excitability.
EXPERIMENTAL PROCEDURES
Animals
All animal procedures were performed in accordance with the National
Institutes of Health Guide for the Care and Use of Laboratory Animals and
were approved by the Tufts University Institutional Animal Care and Use Committee.
The TetTag mouse line used in this study was heterozygous for two
transgenes: c-fos promoter-driven tetracycline transactivator (cfosP-tTA;
Jackson Laboratory stock number 008344) and a tet operator-driven fusion
of histone2B and GFP (tetO-His2BGFP; Jackson Laboratory stock number
005104). TetTag mice were backcrossed to a C57Bl6/J background. Thy1YFP
mice were obtained from Jackson Laboratory (line H; stock number
003782). Mice had food and water ad libitum and were socially housed (three
to five animals per cage) until the start of the experiment, which was at an age
of at least 12 weeks. Mice were kept on a regular light-dark cycle, and all experimental
manipulations were done during the light phase. Mice were raised on
food with doxycycline (40 mg doxycycline/kg chow). One week before fear
conditioning, all mice were individually housed, and 4 days before fear conditioning,
doxycycline was removed from the food. After the last fear conditioning
trial on day 1, mice were put on food with a high dose of doxycycline (1 g/kg)
to rapidly block the tagging of neurons activated after fear conditioning. On
day 2 mice were put back on the regular dose of doxycycline (40 mg/kg).
Behavior
A total of 48 TetTag mice were used for the study. Experiment 1 consisted of a
fear conditioning group (FC, n = 15) and a fear conditioning followed by extinction
group (FC+EXT, n = 17). Experiment 2 consisted of a home cage group
(HC, n = 8) and a fear conditioning group (FC, n = 8).
Experiment 1
The design of experiment 1 is summarized in Figure 1C. On day 1, mice were
subjected to contextual fear conditioning consisting of three training trials (S1,
S2, and S3) with 3 hr between each trial. The total duration of each training trial
was 500 s. A training trial started with placing the mouse in a square chamber
with grid floor (Coulbourn Instruments; H10-11RTC, 120W 3 100D 3 120H).
At 198 s, 278 s, 358 s, and 438 s, a foot shock was delivered (2 s, 0.70 mA).
On days 2 and 3, FC+EXT mice were subjected to four extinction trials per
day (day 2: E1 to E4; day 3: E5 to E8). Each extinction trial lasted 1,800 s
with a trial interval of 2 hr. For each extinction trial, mice were placed in the
same box used for fear conditioning without receiving foot shocks. On day
4, FC and FC+EXT mice were tested over 500 s during a single retrieval test.
For the retrieval test, mice were placed in the same box used for fear conditioning
without receiving foot shocks.
Experiment 2
The design of experiment 2 is summarized in Figure 5A. The FC group in experiment
2 was subjected to the same protocol as the FC group in experiment 1.
The HC group consisted of mice that stayed in their home cage during the
entire experiment. HC mice were perfused at the same time as the FC mice.
Quantification of Freezing
Freezing behavior was measured using a digital camera connected to a computer
with Actimetrics FreezeFrame software. The bout length was 1 s and the
threshold for freezing behavior was determined by an experimenter blind to
experimental conditions and animal group. Freezing scores were obtained
by averaging freezing during minutes 2 and 3 of each trial.
Tissue Preparation and Immunohistochemistry
Ninety minutes after retrieval testing, mice were deeply anesthetized with
ketamine/xylazine and transcardially perfused with 0.1 M phosphate buffer
(PB) followed by 4% paraformaldehyde (PFA 4%) dissolved in 0.1 M PB. Brains
were extracted and postfixed in PFA 4% for 24 hr. Brains were transferred to
30% sucrose for 48–72 hr before slicing 20 mm coronal sections of the entire
brain using a cryostat. Sections were stored in cryoprotectant at À20
C until
use. Free-floating sections were rinsed extensively in PBS with 0.25% Triton
X-100 (PBS-T). Sections were blocked for 1 hr at room temperature in PBST
with 10% normal goat serum (or 3% donkey serum for CB1R). Sections
were incubated in rabbit anti-Zif268 (Santa-Cruz; polyclonal; 1:3,000) with
either mouse anti-GAD67 (Millipore; monoclonal; 1:10,000), mouse anti-PV
(Millipore; monoclonal; 1:2,000), mouse anti-CCK/Gastrin (Center for Ulcer
Research and Education UCLA; monoclonal; 1:1,000), or goat anti-CB1 (kind
gift of Dr. K. Mackie; polyclonal; 1:2,000) (Harkany et al., 2005). Additional primary
antibodies used were rabbit anti-CamKII (kind gift of Dr. M. Jacob; polyclonal;
1:2,000) and rabbit anti-Rab3b (kind gift of Dr. T. Su¨ dhof; polyclonal;
1:4,000). Primary antibodies were diluted in the blocking solution, incubated
at 4
C for 72 hr, and rinsed three times for 15 min in PBS-T. Secondary antibodies
(Jackson ImmunoResearch; goat anti-rabbit 549 1/1,500, goat antimouse
647 1/500, donkey anti-rabbit 649 1/500, donkey anti-goat 549
1/500) were diluted in the blocking solution and were then applied to the sections
for 2 hr at room temperature followed by three rinses for 15 min in PBS-T.
Neuron
Extinction Causes Perisomatic Synapse Remodeling
1062 Neuron 80, 1054–1065, November 20, 2013 ª2013 Elsevier Inc.
Sections were mounted on slides and coverslipped using DAPI mounting
media to label cell nuclei and stored at 4
C.
Confocal Microscopy
A confocal laser-scanning microscope was used for all image acquisition
(Nikon A1R). The settings for PMT, laser power, gain, and offset were identical
between experimental groups. Images of cells expressing GFP and/or Zif
were collected for the basal amygdala (BA, minimum of seven sections per
mouse), hippocampal CA1 (dCA1 and vCA1, four sections per mouse), and
the infralimbic prefrontal cortex (IL, four sections per mouse). For the detection
of perisomatic GAD67, a 203 objective was used and image stacks were
collected with a 2 mm step. For the detection of the other perisomatic markers
(PV, CCK and CB1), a 403 objective was used and image stacks were
collected with a 1 mm step. For the detection of PV, Rab3b, CCK, and CB1
around neurons from Thy1-YFP mice, a 603 objective was used and image
stacks were collected with a 1 mm step.
Image Analysis
All quantification was performed blind to experimental groups.
Quantification of Activated Cells
Selection of GFP-labeled cells was designed to only include excitatory neurons
(Figures S1A and S1B). ImageJ software was used to select and count
the total number of DAPI-, GFP-, and Zif-positive nuclei and nuclei double positive
for GFP and Zif (Figure S1C). In order to avoid bias, all three cell types
(GFP+ZifÀ, GFP+Zif+, GFPÀZif+) were selected from the same pictures,
and the threshold settings for GFP and Zif were identical across all mice.
Quantification of Soma Expression
To quantify expression of GAD67, PV, and CCK within the soma of basal amygdala
interneurons (Figures 3C, 4B, and 6B), we outlined approximately 20
soma for each mouse and calculated average pixel intensity using ImageJ.
Somatic expression of CB1R was not quantified since no labeling of CB1R
was observed in the soma of basal amygdala interneurons.
Quantification of Pixels and Clusters Positive for Perisomatic
Markers
One mouse from the FC+EXT group was excluded from the perisomatic
marker analysis, since no active fear neurons (GFP+Zif+) were found in the
basal amygdala of this mouse. For each marker (GAD67, PV, CCK, and
CB1R), tagged fear neurons (GFP+ZifÀ and GFP+Zif+) and nontagged neurons
(GFPÀZif+) were randomly selected in the basal amygdala, and confocal
images were analyzed at the z plane where the diameter of the nucleus was
largest. The average number of analyzed cells per mouse for each perisomatic
marker is summarized in Table S2. A mask for each perisomatic marker was
generated by thresholding the image of the perisomatic marker. For each perisomatic
marker, we used the same threshold settings for all the counted cells
in order to avoid any bias. This threshold was the same for both the pixel and
cluster countings. Threshold settings only differed between the different perisomatic
markers, since the signal intensity varied across different antibodies.
The mask of the perisomatic marker was used to draw an oval-shaped outline
that included all the pixels positive for the perisomatic marker around a single
neuron. The number of positive pixels and positive clusters (groups of adjacent
positive pixels) within the outline was counted using ImageJ. To normalize for
variation in size of neurons, we divided the numbers of pixels and clusters by
the outline perimeter.
Statistics
Data are presented as means ± SEM and were analyzed using ANOVAs
repeated-measures and two-tailed t test (unpaired or paired) for normally
distributed variables to evaluate statistical significance with p < 0.05 as level
of statistical significance. See Table S2 for the average number of analyzed
cells per mouse for each perisomatic marker and Table S3 for detailed statistical
results.
SUPPLEMENTAL INFORMATION
Supplemental Information includes five figures and three tables and can
be found with this article online at http://dx.doi.org/10.1016/j.neuron.2013.
07.047.
ACKNOWLEDGMENTS
We thank K. Kan and M. Parakala for technical assistance and M. Mayford for
providing TetTag mice. We thank J. Aggleton, J. Ainsley, L. Drane, L. Feig, M.
Jacob, K. Mackie, E. Perisse, and S. Waddell for critical reading of the manuscript.
This work was supported by an NIH Director’s New Innovator Award
(L.G.R.; DP2 OD006446), a Fyssen Foundation Postdoctoral Fellowship, a Bettencourt-Schueller
award, and a Philippe Foundation Award (S.T.), a Sackler
Dean’s Graduate Fellowship (J.S.), the Synapse Neurobiology Training Program
(J.S.; T32 NS061764; PI: K. Dunlap and M. Jacob), the Tufts Center for
Neuroscience Research (P30 NS047243; PI: R. Jackson), and DA011322 (PI:
K. Mackie).
Accepted: July 28, 2013
Published: October 31, 2013
REFERENCES
Agren, T., Engman, J., Frick, A., Bjo¨ rkstrand, J., Larsson, E.-M., Furmark, T.,
and Fredrikson, M. (2012). Disruption of reconsolidation erases a fear memory
trace in the human amygdala. Science 337, 1550–1552.
Amano, T., Unal, C.T., and Pare´ , D. (2010). Synaptic correlates of fear extinction
in the amygdala. Nat. Neurosci. 13, 489–494.
Amano, T., Duvarci, S., Popa, D., and Pare´ , D. (2011). The fear circuit revisited:
contributions of the basal amygdala nuclei to conditioned fear. J. Neurosci. 31,
15481–15489.
Andero, R., Heldt, S.A., Ye, K., Liu, X., Armario, A., and Ressler, K.J. (2011).
Effect of 7,8-dihydroxyflavone, a small-molecule TrkB agonist, on emotional
learning. Am. J. Psychiatry 168, 163–172.
Bartos, M., Vida, I., and Jonas, P. (2007). Synaptic mechanisms of synchronized
gamma oscillations in inhibitory interneuron networks. Nat. Rev.
Neurosci. 8, 45–56.
Carlson, G., Wang, Y., and Alger, B.E. (2002). Endocannabinoids facilitate the
induction of LTP in the hippocampus. Nat. Neurosci. 5, 723–724.
Chhatwal, J.P., Stanek-Rattiner, L., Davis, M., and Ressler, K.J. (2006).
Amygdala BDNF signaling is required for consolidation but not encoding of
extinction. Nat. Neurosci. 9, 870–872.
Corcoran, K.A., and Maren, S. (2001). Hippocampal inactivation disrupts
contextual retrieval of fear memory after extinction. J. Neurosci. 21, 1720–
1726.
Deng, W., Mayford, M., and Gage, F.H. (2013). Selection of distinct populations
of dentate granule cells in response to inputs as a mechanism for pattern
separation in mice. Elife 2, e00312, http://dx.doi.org/10.7554/eLife.00312.
Ehrlich, I., Humeau, Y., Grenier, F., Ciocchi, S., Herry, C., and Lu¨ thi, A. (2009).
Amygdala inhibitory circuits and the control of fear memory. Neuron 62,
757–771.
Falls, W.A., Miserendino, M.J., and Davis, M. (1992). Extinction of fear-potentiated
startle: blockade by infusion of an NMDA antagonist into the amygdala.
J. Neurosci. 12, 854–863.
Garner, A.R., Rowland, D.C., Hwang, S.Y., Baumgaertel, K., Roth, B.L.,
Kentros, C., and Mayford, M. (2012). Generation of a synthetic memory trace.
Science 335, 1513–1516.
Gittis, A.H., Hang, G.B., LaDow, E.S., Shoenfeld, L.R., Atallah, B.V.,
Finkbeiner, S., and Kreitzer, A.C. (2011). Rapid target-specific remodeling of
fast-spiking inhibitory circuits after loss of dopamine. Neuron 71, 858–868.
Gottmann, K., Mittmann, T., and Lessmann, V. (2009). BDNF signaling in the
formation, maturation and plasticity of glutamatergic and GABAergic synapses.
Exp. Brain Res. 199, 203–234.
Graham, B.M., Langton, J.M., and Richardson, R. (2011). Pharmacological
enhancement of fear reduction: preclinical models. Br. J. Pharmacol. 164,
1230–1247.
Gunduz-Cinar, O., Macpherson, K.P., Cinar, R., Gamble-George, J., Sugden,
K., Williams, B., Godlewski, G., Ramikie, T.S., Gorka, A.X., Alapafuja, S.O.,
et al. (2013). Convergent translational evidence of a role for anandamide in
Neuron
Extinction Causes Perisomatic Synapse Remodeling
Neuron 80, 1054–1065, November 20, 2013 ª2013 Elsevier Inc. 1063
amygdala-mediated fear extinction, threat processing and stress-reactivity.
Mol. Psychiatry 18, 813–823.
Harkany, T., Dobszay, M.B., Cayetanot, F., Ha¨ rtig, W., Siegemund, T., Aujard,
F., and Mackie, K. (2005). Redistribution of CB1 cannabinoid receptors during
evolution of cholinergic basal forebrain territories and their cortical projection
areas: a comparison between the gray mouse lemur (Microcebus murinus, primates)
and rat. Neuroscience 135, 595–609.
Hartley, C.A., and Phelps, E.A. (2010). Changing fear: the neurocircuitry of
emotion regulation. Neuropsychopharmacology 35, 136–146.
Heitland, I., Klumpers, F., Oosting, R.S., Evers, D.J., Leon Kenemans, J., and
Baas, J.M. (2012). Failure to extinguish fear and genetic variability in the human
cannabinoid receptor 1. Transl. Psychiatry 2, e162, http://dx.doi.org/10.1038/
tp.2012.90.
Heldt, S.A., and Ressler, K.J. (2007). Training-induced changes in the expression
of GABAA-associated genes in the amygdala after the acquisition and
extinction of Pavlovian fear. Eur. J. Neurosci. 26, 3631–3644.
Heldt, S.A., Mou, L., and Ressler, K.J. (2012). In vivo knockdown of GAD67 in
the amygdala disrupts fear extinction and the anxiolytic-like effect of diazepam
in mice. Transl. Psychiatry 2, e181, http://dx.doi.org/10.1038/tp.2012.101.
Herry, C., Ciocchi, S., Senn, V., Demmou, L., Mu¨ ller, C., and Lu¨ thi, A. (2008).
Switching on and off fear by distinct neuronal circuits. Nature 454, 600–606.
Hobin, J.A., Goosens, K.A., and Maren, S. (2003). Context-dependent
neuronal activity in the lateral amygdala represents fear memories after extinction.
J. Neurosci. 23, 8410–8416.
Hong, E.J., McCord, A.E., and Greenberg, M.E. (2008). A biological function for
the neuronal activity-dependent component of Bdnf transcription in the development
of cortical inhibition. Neuron 60, 610–624.
Huang, Z.J., Kirkwood, A., Pizzorusso, T., Porciatti, V., Morales, B., Bear, M.F.,
Maffei, L., and Tonegawa, S. (1999). BDNF regulates the maturation of inhibition
and the critical period of plasticity in mouse visual cortex. Cell 98,
739–755.
Jiao, Y., Zhang, Z., Zhang, C., Wang, X., Sakata, K., Lu, B., and Sun, Q.Q.
(2011). A key mechanism underlying sensory experience-dependent maturation
of neocortical GABAergic circuits in vivo. Proc. Natl. Acad. Sci. USA
108, 12131–12136.
Katona, I., Rancz, E.A., Acsa´ dy, L., Ledent, C., Mackie, K., Ha´ jos, N., and
Freund, T.F. (2001). Distribution of CB1 cannabinoid receptors in the amygdala
and their role in the control of GABAergic transmission. J. Neurosci. 21, 9506–
9518.
Knapska, E., Macias, M., Mikosz, M., Nowak, A., Owczarek, D., Wawrzyniak,
M., Pieprzyk, M., Cymerman, I.A., Werka, T., Sheng, M., et al. (2012).
Functional anatomy of neural circuits regulating fear and extinction. Proc.
Natl. Acad. Sci. USA 109, 17093–17098.
Kohara, K., Yasuda, H., Huang, Y., Adachi, N., Sohya, K., and Tsumoto, T.
(2007). A local reduction in cortical GABAergic synapses after a loss of endogenous
brain-derived neurotrophic factor, as revealed by single-cell gene
knock-out method. J. Neurosci. 27, 7234–7244.
Kullmann, D.M., Moreau, A.W., Bakiri, Y., and Nicholson, E. (2012). Plasticity of
inhibition. Neuron 75, 951–962.
Lai, C.S., Franke, T.F., and Gan, W.B. (2012). Opposite effects of fear conditioning
and extinction on dendritic spine remodelling. Nature 483, 87–91.
Likhtik, E., Popa, D., Apergis-Schoute, J., Fidacaro, G.A., and Pare´ , D. (2008).
Amygdala intercalated neurons are required for expression of fear extinction.
Nature 454, 642–645.
Liu, X., Ramirez, S., Pang, P.T., Puryear, C.B., Govindarajan, A., Deisseroth,
K., and Tonegawa, S. (2012). Optogenetic stimulation of a hippocampal
engram activates fear memory recall. Nature 484, 381–385.
Livneh, U., and Paz, R. (2012). Aversive-bias and stage-selectivity in neurons
of the primate amygdala during acquisition, extinction, and overnight retention.
J. Neurosci. 32, 8598–8610.
Makkar, S.R., Zhang, S.Q., and Cranney, J. (2010). Behavioral and neural analysis
of GABA in the acquisition, consolidation, reconsolidation, and extinction
of fear memory. Neuropsychopharmacology 35, 1625–1652.
Maren, S. (1998). Overtraining does not mitigate contextual fear conditioning
deficits produced by neurotoxic lesions of the basolateral amygdala.
J. Neurosci. 18, 3088–3097.
Marsicano, G., Wotjak, C.T., Azad, S.C., Bisogno, T., Rammes, G., Cascio,
M.G., Hermann, H., Tang, J., Hofmann, C., Zieglga¨ nsberger, W., et al.
(2002). The endogenous cannabinoid system controls extinction of aversive
memories. Nature 418, 530–534.
Matsumoto, T., Numakawa, T., Yokomaku, D., Adachi, N., Yamagishi, S.,
Numakawa, Y., Kunugi, H., and Taguchi, T. (2006). Brain-derived neurotrophic
factor-induced potentiation of glutamate and GABA release: different dependency
on signaling pathways and neuronal activity. Mol. Cell. Neurosci. 31,
70–84.
Matsuo, N., Reijmers, L., and Mayford, M. (2008). Spine-type-specific
recruitment of newly synthesized AMPA receptors with learning. Science
319, 1104–1107.
McDonald, A.J. (1992). Projection neurons of the basolateral amygdala: a
correlative Golgi and retrograde tract tracing study. Brain Res. Bull. 28,
179–185.
McDonald, A.J., and Betette, R.L. (2001). Parvalbumin-containing neurons in
the rat basolateral amygdala: morphology and co-localization of CalbindinD(28k).
Neuroscience 102, 413–425.
Milad, M.R., and Quirk, G.J. (2002). Neurons in medial prefrontal cortex signal
memory for fear extinction. Nature 420, 70–74.
Miles, R., To´ th, K., Gulya´ s, A.I., Ha´ jos, N., and Freund, T.F. (1996). Differences
between somatic and dendritic inhibition in the hippocampus. Neuron 16,
815–823.
Morgan, M.A., Romanski, L.M., and LeDoux, J.E. (1993). Extinction of
emotional learning: contribution of medial prefrontal cortex. Neurosci. Lett.
163, 109–113.
Myers, K.M., and Davis, M. (2007). Mechanisms of fear extinction. Mol.
Psychiatry 12, 120–150.
Okuno, H. (2011). Regulation and function of immediate-early genes in the
brain: beyond neuronal activity markers. Neurosci. Res. 69, 175–186.
Orsini, C.A., and Maren, S. (2012). Neural and cellular mechanisms of fear and
extinction memory formation. Neurosci. Biobehav. Rev. 36, 1773–1802.
Orsini, C.A., Kim, J.H., Knapska, E., and Maren, S. (2011). Hippocampal and
prefrontal projections to the basal amygdala mediate contextual regulation
of fear after extinction. J. Neurosci. 31, 17269–17277.
Rabinak, C.A., Angstadt, M., Sripada, C.S., Abelson, J.L., Liberzon, I., Milad,
M.R., and Phan, K.L. (2013). Cannabinoid facilitation of fear extinction memory
recall in humans. Neuropharmacology 64, 396–402.
Reijmers, L.G., Perkins, B.L., Matsuo, N., and Mayford, M. (2007). Localization
of a stable neural correlate of associative memory. Science 317, 1230–1233.
Repa, J.C., Muller, J., Apergis, J., Desrochers, T.M., Zhou, Y., and LeDoux,
J.E. (2001). Two different lateral amygdala cell populations contribute to the
initiation and storage of memory. Nat. Neurosci. 4, 724–731.
Rudy, J.W., Huff, N.C., and Matus-Amat, P. (2004). Understanding contextual
fear conditioning: insights from a two-process model. Neurosci. Biobehav.
Rev. 28, 675–685.
Ruehle, S., Remmers, F., Romo-Parra, H., Massa, F., Wickert, M., Wo¨ rtge, S.,
Ha¨ ring, M., Kaiser, N., Marsicano, G., Pape, H.C., and Lutz, B. (2013).
Cannabinoid CB1 receptor in dorsal telencephalic glutamatergic neurons:
distinctive sufficiency for hippocampus-dependent and amygdala-dependent
synaptic and behavioral functions. J. Neurosci. 33, 10264–10277.
Sangha, S., Narayanan, R.T., Bergado-Acosta, J.R., Stork, O., Seidenbecher,
T., and Pape, H.C. (2009). Deficiency of the 65 kDa isoform of glutamic acid
decarboxylase impairs extinction of cued but not contextual fear memory.
J. Neurosci. 29, 15713–15720.
Sangha, S., Ilenseer, J., Sosulina, L., Lesting, J., and Pape, H.C. (2012).
Differential regulation of glutamic acid decarboxylase gene expression after
extinction of a recent memory vs. intermediate memory. Learn. Mem. 19,
194–200.
Neuron
Extinction Causes Perisomatic Synapse Remodeling
1064 Neuron 80, 1054–1065, November 20, 2013 ª2013 Elsevier Inc.
Soliman, F., Glatt, C.E., Bath, K.G., Levita, L., Jones, R.M., Pattwell, S.S., Jing,
D., Tottenham, N., Amso, D., Somerville, L.H., et al. (2010). A genetic variant
BDNF polymorphism alters extinction learning in both mouse and human.
Science 327, 863–866.
Tayler, K.K., Lowry, E., Tanaka, K., Levy, B., Reijmers, L., Mayford, M., and
Wiltgen, B.J. (2011). Characterization of NMDAR-independent learning in the
hippocampus. Front. Behav. Neurosci. 5, 28.
Tayler, K.K., Tanaka, K.Z., Reijmers, L.G., and Wiltgen, B.J. (2013).
Reactivation of neural ensembles during the retrieval of recent and remote
memory. Curr. Biol. 23, 99–106.
Tronson, N.C., Schrick, C., Guzman, Y.F., Huh, K.H., Srivastava, D.P., Penzes,
P., Guedea, A.L., Gao, C., and Radulovic, J. (2009). Segregated populations of
hippocampal principal CA1 neurons mediating conditioning and extinction of
contextual fear. J. Neurosci. 29, 3387–3394.
Varga, C., Lee, S.Y., and Soltesz, I. (2010). Target-selective GABAergic control
of entorhinal cortex output. Nat. Neurosci. 13, 822–824.
Vianna, M.R., Szapiro, G., McGaugh, J.L., Medina, J.H., and Izquierdo, I.
(2001). Retrieval of memory for fear-motivated training initiates extinction
requiring protein synthesis in the rat hippocampus. Proc. Natl. Acad. Sci.
USA 98, 12251–12254.
Yoshida, T., Uchigashima, M., Yamasaki, M., Katona, I., Yamazaki, M.,
Sakimura, K., Kano, M., Yoshioka, M., and Watanabe, M. (2011). Unique inhibitory
synapse with particularly rich endocannabinoid signaling machinery on
pyramidal neurons in basal amygdaloid nucleus. Proc. Natl. Acad. Sci. USA
108, 3059–3064.
Neuron
Extinction Causes Perisomatic Synapse Remodeling
Neuron 80, 1054–1065, November 20, 2013 ª2013 Elsevier Inc. 1065