CS-US pairings. By contrast, in unpaired con-
ditioned mice exposed to equal CSs and USs,
but in a temporally unpaired manner, CS re-
sponses of somas and dendrites decreased
uniformly (fig. S7, C to F and J to M). CS-US
pairing thus induces dendritic plasticity of
CS responses that is not fully reflected at the
level of the soma. To explore the relationship
between somatic and dendritic CS response
plasticity at the single-neuron level, we simul-
taneously examined conditioning-induced
changes in somas and in connected dendrites.
Consistent with the population-level analysis
and with previous one-photon imaging results
from populations of amygdala neurons in
freely moving animals ( 20 , 21 ), we found that
LA PN somas could be divided into three
functional plasticity types with either increased,
decreased, or unchanged CS responses. Only a
minority of CSup and CSdown somas exhib-
ited converging CS and US responses during
conditioning, indicating that both the up-
regulation and the down-regulation of somatic
responses may result from different induction
processes involving synaptic and/or dendritic
plasticity as well as network effects.
Our finding that individual dendritic branches
can exhibit US responses that do not result in
somatic transients and are uncoupled from
somatic activity suggests that conditioning
may lead to local, compartmentalized plastic-
ity in dendrites that might encode the CS-US
association memory in the BLA ( 27 ). CSup and
CSdown spines are preferentially located on
CSup and CSdown dendrites, respectively (fig.
S8). Further, consistent with the idea of an-
atomical clustering of potentiated synaptic in-
puts ( 31 ), the distance between CSup spines
located on a given dendritic branch is smaller
as compared with the distance between CSdown
spines (fig. S6L). Therefore, conclusions based
on somatic US responses may underestimate
the complexity of dendritic integration and
plasticity as a source of neuronal output adap-
tation and learning ( 32 ), as suggested by a re-
cent computational modeling study ( 33 ).
We observed that FC resulted in a decorre-
lation of CS responses between dendrites and
their parent soma, as well as between individ-
ual dendritic branches belonging to the same
neuron(Fig.5,B,C,andE).Thismayreflect
conditioning-induced and branch-specific
changes in action potential backpropaga-
tion, possibly resulting from changes in SST+
interneurons–mediated dendritic inhibition.
More likely, FC might lead to the induction of
localized synaptic and dendritic CS response
plasticity. The more pronounced decorrelation
observed in more distal, higher-order dendrites
(Fig. 5C) argues for local dendritic plasticityas the origin of soma-dendrite decorrelation
than for alterations in action potential back-
propagation that would increase soma-dendrite
correlations. Moreover, although the propor-
tion of dendrite-only CS responses increased
after FC (Fig. 5E), the overall proportion of
spontaneous dendrite-only Ca2+transients
did not change (fig. S9, Q to S). This suggests
that FC does not lead to a general increase
in dendritic excitability but specifically in-
creases dendritic responses to the conditioned
sensory inputs ( 34 ).
Whereas dendrites and their parent soma
of CSup neurons generally exhibited a corre-
lated increase in their CS responses after
learning, this was not the case for CSdown
neurons. Even though the somas of CSdown
neurons showed decreased CS responses upon
learning, dendrites belonging to the same
neurons did not but instead exhibited in-
creased CS responses. This strongly suggests
that in CSdown neurons, learning-induced
dendritic and somatic plasticity mechanisms
are distinct from CSup neurons.
Although increased dendritic responses in
CSdown neurons likely result from potentia-
tion of synaptic inputs, decreased somatic re-
sponses may reflect soma-specific changes of
neuronal excitability ( 35 – 37 ) and/or plasticity of
local soma-targeting inhibitory circuits ( 28 , 38 ).d’Aquinet al.,Science 376 , eabf7052 (2022) 15 April 2022 8 of 13
HabituationDay 1
10 x CSTe s tDay 3
10 x CSConditioningDay 2
10 x CS-USCNOCSup neuronsTime (s) from CS onsetHabituation Test
F/F + CNO048 0 48CSdown neuronsTime (s) from CS onsetHabituation-0.2 04804800.20.431%
19%0246SalineCNOTransients per min*SST+hM4D(Gi)-mCherryVclampeIPSC amplitude (nA)
BL CNO00.511.5(^2)
19 %
31 %
15 %
35 %
PV HM4D
32 %
37 %
13 %
18 %
Conditioned
CS unresp. ()
CS stable (n.s)
CS down (***)
CS up (n.s)
proportion of
functional cell types
Norm. eIPSC amplitude
Baseline CNO
-1 0 1 2
0.0
0.4
0.8
1.2
Time (min)
hM4D(Gi)-mCherry
50 μm
GCaMP6s
GCaMP
hM4D
CS down somas
-4
-2
0
-6
Conditioned
CS responsePV HM4D
amplitude (
F/F
)
fold change to habituation
- -4
-2
0
-6
CS response
integral (
F/F
)
fold change to habituation
Conditioned
PV HM4D
CS up somas
0
5
10
15
20
CS response
integral (
F/F
)
fold change to habituation
ConditionedPV HM4D
n.s
0
5
10
15
CS response
amplitude (
F/F
)
fold change to habituation
ConditionedPV HM4D
n.s
Test CS up somas CS down somas
- CNO
PV-Cre
H
A B C
G
F
I J
D E
KLMN
PN PV+ PN
PV+
Fig. 6. Uncoupling of dendritic and somatic plasticity in CSdown neurons
is mediated by an increase in perisomatic inhibition.(A) Auditory FC
paradigm with simultaneous two-photon imaging. (B) Experiment scheme.
(C) Confocal microscopy image of hM4D(Gi)-mCherry in PV+ INs with sparse
GCaMP6s in LA PNs. (D) Whole-cell recording of evoked inhibitory postsynaptic
current (eIPSC) in BLA PNs. (E) Reduction of PV+ IN output effect on eIPSC
amplitude in BLA PNs (n=14 cells, three mice). Median (black) and interquartile
range (gray) are normalized to median baseline period. Green shaded area
indicates bath application of CNO. Top black line indicates the CNO analysis
window. (F) Statistical analysis of (E) (P= 0.017, Wilcoxon matched-pairs signed-
rank test). (G) Suppression of PV+ IN output increases the rate of somatic
transients (P= 0.012, one-sample Wilcoxon signed-rank test,n=66 neurons).
(H) A large proportion of somas up-regulate and others (I) down-regulate their
CS response after FC (mean ± SEM). (J) Similar proportion of CSup and smaller
proportion of CSdown somas in mice with suppressed PV+ IN activity in the
amygdala (CSup,P= 0.618; CSdown,P< 0.001; both FisherÕs exact test,
Bonferroni corrected; conditioned,n=177 neurons from nine mice; PV CNO,n=
66 neurons from six mice). (K) Reduction of PV+ IN output does not affect CS
response amplitude change in CSup neurons during auditory FC (P= 0.552
Wilcoxon rank-sum test; conditioned,n=57 neurons; PV CNO,n=39 neurons).
(L) Reduction of PV+ IN output does not affect CS response integral change in
CSup neurons during auditory FC (P= 0.622 Wilcoxon rank-sum test;
conditioned,n=57 neurons; PV CNO,n=39 neurons). (M) Reduction of PV+ IN
output decreases the CS response amplitude change of CSdown neurons after
auditory FC (P= 0.039 Wilcoxon rank-sum test; conditioned,n=66 neurons; PV
CNO,n=27 neurons). (N) Reduction of PV+ IN output decreases the CS
response integral change of CSdown neurons after auditory FC (P= 0.032,
Wilcoxon rank-sum test; conditioned,n=66 neurons; PV CNO,n=27 neurons).
RESEARCH | RESEARCH ARTICLE