Article

Sensory Inputs to Intercalated Cells Provide FearLearning Modulated Inhibition to the Basolateral Amygdala Highlights

Authors

d

Amygdala medial paracapsular cells (mpITCs) receive sensory stimulus-related inputs

Douglas Asede, Daniel Bosch, ..., Francesco Ferraguti, Ingrid Ehrlich

d

mpITC sensory inputs are modified by fear learning

Correspondence

d

mpITCs are reciprocally connected with and control BLA projection neuron activity

[email protected]

mpITCs provide sensory-driven feedforward and feedback inhibition to the BLA

Asede et al. show that medial paracapsular intercalated cells (mpITCs) in the amygdala receive fear learningmodulated sensory inputs and provide feedforward and feedback inhibition to basolateral projection neurons. This places mpITCs in a key position to gate acquired fear behavior.

d

Asede et al., 2015, Neuron 86, 1–14 April 22, 2015 ª2015 Elsevier Inc. http://dx.doi.org/10.1016/j.neuron.2015.03.008

In Brief

Please cite this article in press as: Asede et al., Sensory Inputs to Intercalated Cells Provide Fear-Learning Modulated Inhibition to the Basolateral Amygdala, Neuron (2015), http://dx.doi.org/10.1016/j.neuron.2015.03.008

Neuron

Article Sensory Inputs to Intercalated Cells Provide Fear-Learning Modulated Inhibition to the Basolateral Amygdala Douglas Asede,1,2,3 Daniel Bosch,1,2 Andreas Lu¨thi,4 Francesco Ferraguti,5 and Ingrid Ehrlich1,2,* 1Hertie

Institute for Clinical Brain Research Reichardt Centre for Integrative Neuroscience University of Tu¨bingen, Otfried-Mu¨ller-Straße 25, 72076 Tu¨bingen, Germany 3Graduate School of Neural and Behavioral Sciences, International Max Planck Research School, University of Tu ¨ bingen, O¨sterbergstrasse 3, 72074 Tu¨bingen, Germany 4Friedrich Miescher Institute for Biomedical Research, Maulbeerstrasse 66, 4058 Basel, Switzerland 5Department of Pharmacology, Innsbruck Medical University, Peter Mayr Straße 1a, 6020 Innsbruck, Austria *Correspondence: [email protected] http://dx.doi.org/10.1016/j.neuron.2015.03.008 2Werner

SUMMARY

Increasing evidence suggests that parallel plastic processes in the amygdala involve inhibitory elements to control fear and extinction memory. GABAergic medial paracapsular intercalated cells (mpITCs) are thought to relay activity from basolateral nucleus (BLA) and prefrontal cortex to inhibit central amygdala output during suppression of fear. Recently, projection diversity and differential behavioral activation of mpITCs in distinct fear states suggest additional functions. Here, we show that mpITCs receive convergent sensory thalamic and cortical inputs that undergo fear learning-related changes and are dynamically modulated via presynaptic GABAB receptors recruited by GABA released from the mpITC network. Among mpITCs, we identify cells that inhibit but are also mutually activated by BLA principal neurons. Thus, mpITCs take part in fear learning-modulated feedforward and feedback inhibitory circuits to simultaneously control amygdala input and output nuclei. Our findings place mpITCs in a unique position to gate acquired amygdala-dependent behaviors via their direct sensory inputs. INTRODUCTION The amygdala plays a critical role in mediating fear and anxietyrelated behaviors and is a key site for acquisition and storage of fear memory (Davis, 2000; LeDoux, 2000). Plasticity of sensory inputs from thalamic and cortical areas to projection neurons in the lateral part (LA) of the basolateral amygdala (BLA) is an important mechanism underlying Pavlovian fear conditioning and contributes to formation of the association between conditioned (CS) and unconditioned stimuli (US) (Pape and Pare,

2010). According to the serial circuit model of amygdala function, associative information is conveyed from the LA, following internuclear excitatory projections via the basal and basomedial nuclei to the medial division of the central amygdala (CeM) to generate fear output (Maren, 2005; Moscarello and LeDoux, 2013). Such a model of passive information transfer is challenged by growing evidence that BLA and central amygdala (CeA) can mediate either independent parallel or additional associative functions in aversive conditioning (Balleine and Killcross, 2006; Jimenez and Maren, 2009; Moscarello and LeDoux, 2013). For example, fear-related synaptic and cellular plasticity is observed in the lateral part of the CeA (CeL), but the role of specific afferent pathways is incompletely understood (Ciocchi et al., 2010; Haubensak et al., 2010; Li et al., 2013; Wilensky et al., 2006). Another emerging principle is that plasticity of inhibitory synapses and interneurons in the amygdala controls acquired fear (Ehrlich et al., 2009). On one hand, regulation of inhibition within the BLA and of specific inhibitory synapses onto BLA principal neurons (PNs) has been implicated in the behavioral suppression of fear following extinction learning (Chhatwal et al., 2005; Trouche et al., 2013). On the other hand, activity of local GABAergic neurons may be regulated by plasticity of their excitatory inputs, both in BLA (Bauer and LeDoux, 2004; Mahanty and Sah, 1998; Polepalli et al., 2010) and CeL (Fu and Shinnick-Gallagher, 2005; Li et al., 2013; Samson and Pare´, 2005). Furthermore, a recent study showed that specific local interneurons within the BLA are differentially recruited during CS-US associations, mediating inhibition and disinhibition of distinct subcellular domains to control fear learning (Wolff et al., 2014). Thus, multiple plastic systems, and especially those integrating GABAergic neurons, could act in concert to encode aspects of stimulus associations in the amygdala. An amygdala inhibitory system with unique properties are the intercalated cells (ITCs), a network of interconnected GABAergic neurons organized in distinct clusters surrounding the BLA (Geracitano et al., 2007; Millhouse, 1986). Recent evidence suggests that ITCs located within the intermediate Neuron 86, 1–14, April 22, 2015 ª2015 Elsevier Inc. 1

Please cite this article in press as: Asede et al., Sensory Inputs to Intercalated Cells Provide Fear-Learning Modulated Inhibition to the Basolateral Amygdala, Neuron (2015), http://dx.doi.org/10.1016/j.neuron.2015.03.008

capsule between BLA and CeA (medial paracapsular ITCs [mpITCs]) may participate in fear and extinction learning and processing of emotional stimuli (Busti et al., 2011; PalomaresCastillo et al., 2012). For example, immediate early gene mapping demonstrated that mpITCs become activated and are plastic during high and low fear states, but inputs that drive this remain to be identified (Busti et al., 2011; Hefner et al., 2008; Knapska and Maren, 2009). The mpITCs receive excitatory inputs from BLA and mainly target more ventrally located ITCs and the CeA, thus, functioning as an inhibitory gate for the information flow between BLA and CeA (Busti et al., 2011; Duvarci and Pare, 2014; Ju¨ngling et al., 2008; Royer et al., 1999, 2000). During fear extinction, increased activity in the infralimbic medial prefrontal cortex (IL-mPFC) is thought to excite mpITCs, which in turn would reduce the fear response by decreasing amygdala output via inhibition of CeM output neurons (Amano et al., 2010; Amir et al., 2011; Sierra-Mercado et al., 2011). However, dorsal and ventral mpITCs receive moderate to few projections from the IL-mPFC (Dobi et al., 2013; Mcdonald et al., 1996; Pinard et al., 2012), which are not plastic upon extinction learning (Cho et al., 2013). An important role for mpITCs in fear extinction is supported by several findings: first, their ablation attenuated extinction retrieval (Likhtik et al., 2008). Second, increasing their BLA input by neuropeptide S (NPS) enhanced extinction learning and retrieval (Ju¨ngling et al., 2008). Third, BLA inputs to ventral mpITCs were enhanced by extinction training (Amano et al., 2010). In contrast, a recent study demonstrated that BLA inputs to mpITCs are potentiated upon fear learning and reduced to control levels upon extinction training (Huang et al., 2014). These apparently conflicting results can be reconciled taking into account a preferential topographic connectivity of dorsal versus ventral mpITCs with LA and BA as their major input and CeL and CeM as their major output (Duvarci and Pare, 2014). However, this cannot completely resolve all observations on plasticity and activation of mpITCs nor explain differential effects on amygdala output and fear behavior (Busti et al., 2011; Hefner et al., 2008). It also does not fully account for other inputs to mpITCs, the impact of mpITCs on local targets, or their diverse projections and targets outside of CeM (Amir et al., 2011; Busti et al., 2011). In this context, it has been proposed that dorsal mpITCs may drive fear expression by disinhibiting CeM via CeL (Busti et al., 2011). Based on these emerging complex functions, we speculated that mpITCs are integrated in parallel plastic pathways within fear circuits. Here, we aimed to identify and characterize novel inputs and outputs of the dorsal mpITC network. We show that mpITCs receive direct and convergent excitatory inputs from sensory thalamic and temporal cortical regions conveying CS and US information. These inputs are modified upon fear learning and retrieval and are dynamically regulated by heterosynaptic GABAB receptors recruited by GABA released from the mpITC network. Interestingly, we find a novel inhibitory projection of mpITCs that effectively controls the output of BLA PNs and demonstrate a reciprocal connectivity of mpITCs with BLA PNs. This indicates that a subgroup of mpITCs provides both sensory-driven feedforward and feedback inhibitory control to the BLA. 2 Neuron 86, 1–14, April 22, 2015 ª2015 Elsevier Inc.

RESULTS mpITCs Receive Convergent Excitatory Inputs from Sensory Thalamic and Cortical Afferents Because mpITCs are situated aside the internal and within the intermediate capsule, we hypothesized that they receive primary sensory afferents from thalamic and cortical regions conveying CS and US information to the LA. Tracing experiments revealed that fibers from medial geniculate nucleus (MGm) and intralaminar nuclei (PIN) of the thalamus en route to LA also branch into the mpITC cluster avoiding the adjacent capsular and lateral part of CeA (Figures 1A and 1B). Similarly, fibers originating from temporal cortex (TeA) coursing through intermediate capsule to LA also appeared to innervate the cluster (Figures 1C and 1D). To confirm functional innervation of both mpITCs and LA PNs, we performed paired recordings. Electrical stimulation of internal (thalamic) and external capsules (cortical) elicited monosynaptic excitatory postsynaptic currents (EPSCs) with similar latencies and amplitudes (Figures 1E and S1A–S1C). Thalamic- and cortical-evoked EPSCs in mpITCs were mediated by AMPA- and NMDA-type glutamate receptors (Figures S1D–S1G). As previously observed for LA PNs, we found convergent sensory inputs in most mpITCs (Figure 1F). Experiments interleaving single pathway and co-stimulation of sensory pathways indicated that co-stimulation resulted in addition of response amplitudes, strongly suggesting that the two inputs were independent (Figure 1F). Sensory Inputs to mpITCs Undergo Fear-Learning Driven Changes Sensory inputs onto LA projection neurons undergo bidirectional plastic changes upon fear and extinction learning that can be detected in ex vivo recordings (Kim et al., 2007; McKernan and Shinnick-Gallagher, 1997; Tsvetkov et al., 2002). To examine whether sensory inputs onto mpITCs are also altered, in experiment 1, we subjected animals to fear conditioning (CS-US pairing) or to CS presentations only. On day 2, conditioned animals showed CS-evoked freezing, whereas animals in the CS only group did not (Figures 2A and 2B). Acute brain slices were prepared 30 min later by an experimenter blind to treatment of the animals to examine presynaptic and postsynaptic properties of sensory inputs onto mpITCs. When assessing paired pulse ratio (PPR) as an indirect measure of presynaptic strength, we found an increased PPR of the AMPA-EPSC in thalamic and cortical inputs upon fear memory retrieval compared with CS only and naive groups (Figures 2C and 2D). The increased PPR correlated with CS-evoked freezing in both pathways (Figures S2A and S2B), suggesting that fear learning decreased presynaptic release probability. To assess postsynaptic parameters that could be a signature of plastic changes, we measured the AMPA/NMDA ratio (A/N ratio) (Kessels and Malinow, 2009). In the thalamic pathway, we found a significant decrease in A/N ratio upon fear memory retrieval compared with CS only and naive groups. Surprisingly, for the cortical input, CS presentations alone significantly reduced the A/N ratio compared with naive, but not to the fear memory group (Figures 2E and 2F). Furthermore, postsynaptic alterations correlated with CS-evoked freezing only in the thalamic pathway, but not in the cortical one (Figures S2C and S2D).

Please cite this article in press as: Asede et al., Sensory Inputs to Intercalated Cells Provide Fear-Learning Modulated Inhibition to the Basolateral Amygdala, Neuron (2015), http://dx.doi.org/10.1016/j.neuron.2015.03.008

Figure 1. mpITCs Receive Convergent Inputs from Thalamic and Cortical Afferents

GFP Fluoro-ruby

A

B1

B2

mpITC MGm

LA LA

CeA PIN BA

C

D1

D2 mpITC

TeA CeA

LA

BA

EcT

Thalamic

Cortical

LA PN mpITC

8

8

6

6

4

4

2

2

0 0

2 4 6 Latency (ms) mpITC

8

0

F

EPSC (% [T+C])

LA PN Latency (ms)

E

LA

0

2 4 6 8 Latency (ms) mpITC

100

50

(A and B) mpITCs receive projections from sensory thalamus. (A) Fluoro-ruby injection site in MGm/ PIN. (B1) Labeled fibers enter the amygdala via internal capsule. (B2) Labeled fibers traverse mpITC cluster toward LA. (C and D) mpITCs receive projections from temporal cortex. (C) Fluoro-ruby injection site in temporal cortex (TeA). (D1) Labeled fibers enter the amygdala via external and intermediate capsule. (D2) Confocal image of labeled fibers in mpITC cluster. Scale bars represent 200 mm (A, B1, C, and D1), 50 mm (B2), 30 mm (D2). (E) EPSCs (Vh = 70 mV) evoked by electrical stimulation of internal (thalamic) and external (cortical) capsule recorded simultaneously in an mpITC and LA PN. Scale bars represent left, 200 pA, 15 ms; right, 400 pA, 15 ms. Similar EPSC latencies in mpITCs and LA PNs. Thalamic: mpITC 3.64 ± 0.21 ms, LA PN 3.31 ± 0.25 ms (n = 9 pairs, p = 0.30); cortical: mpITC 4.48 ± 0.05 ms, LA PN 4.72 ± 0.32 ms (n = 5 pairs, p = 0.48). (F) Additivity of thalamic (T) and cortical (C) inputs to mpITCs. EPSCs (Vh = 70 mV) and bar graph with quantification for each pathway, co-stimulation of the pathways (TC) or obtained by mathematical addition (T+C). T = 49.3% ± 5.9%, C = 56.1% ± 5.7%, TC = 97.6% ± 6.4%, T+C = 100% (n = 7 cells) normalized to T+C amplitude. TC versus T+C was not different (p = 1.00). Scale bars represent 50 pA, 50 ms.

0

Next, we asked whether sensory inputs to mpITCs undergo rapid changes following fear and extinction learning. Therefore, in experiment 2, we subjected animals to fear conditioning or to conditioning and subsequent extinction learning (Figures 2A and 2B) using an established 2-day extinction protocol (Herry et al., 2008). Brain slices were obtained 2 hr after training and in parallel from naive mice. In the thalamic input, the PPR of AMPA-EPSCs was increased immediately after fear conditioning, and this increase was still apparent after extinction training (Figures 2G and 2H). In the cortical input, there was a trend toward increased PPR upon fear and extinction learning (Figures 2G and 2H). When assessing the A/N ratio, we found a significant decrease following fear conditioning in the cortical but not the thalamic input (Figures 2I and 2J). This could be explained by the fact that CS presentations alone altered the A/N ratio during conditioning in the cortical input. For the thalamic input, this suggests that changes in A/N ratio occurred in a protracted fashion and were only observed upon fear memory retrieval. To directly compare synaptic changes over the course of conditioning, memory retrieval, and extinction learning, we normalized synaptic properties to the respective naive control group in the two experiments (Figures S2E–S2H). This revealed that the increase in PPR in thalamic and cortical inputs was equally prominent after fear conditioning (day 1), memory retrieval (day 2) and extinction learning (day 3, Figures S2E and S2F). In contrast, the A/N ratio increased when comparing fear memory retrieval and

T

C

TC

T+C

extinction groups (Figures S2G and S2H), suggesting additional postsynaptic changes upon extinction learning. In summary, in contrast to the LA, sensory inputs onto mpITCs exhibited a lasting decrease in presynaptic and postsynaptic strength upon fear conditioning and retrieval that was partially reversed following extinction learning. Thus, sensory inputs to mpITCs are altered specifically and differentially during opposing behavioral states. Presynaptic GABAB Receptors Depress Excitatory Sensory Responses from MGm/PIN and Temporal Cortex onto mpITCs MpITCs are interconnected by GABAergic synapses that signal via GABAA receptors and lack postsynaptic GABAB responses (Geracitano et al., 2007). We hypothesized that GABA released upon sensory activation of the mpITC network modulates sensory inputs via presynaptic GABAB and assessed whether functional GABAB receptors are present on sensory afferents. The GABAB receptor agonist baclofen induced a significant depression in EPSC amplitude evoked by electrical stimulation of thalamic and cortical afferents, which was completely reversed by co-application of the GABAB receptor antagonist CGP55845 (Figures S3A–S3C and S3E–S3G). Synaptic depression was accompanied by a significant increase in PPR at an interstimulus interval of 50 ms (Figures S3D and Figure S3H). Consistent with a presynaptic effect, the size of the synaptic depression correlated well with the concomitant Neuron 86, 1–14, April 22, 2015 ª2015 Elsevier Inc. 3

Please cite this article in press as: Asede et al., Sensory Inputs to Intercalated Cells Provide Fear-Learning Modulated Inhibition to the Basolateral Amygdala, Neuron (2015), http://dx.doi.org/10.1016/j.neuron.2015.03.008

A

B day 2

day 3

% Freezing

Experiment 1

day 1 Naive CS only 5x FM

5x

4x +

20

5x 5x

Ext

% Freezing

Experiment 2

60

FC

+ +

16x

16x

CS only

FM

D Paired pulse ratio

Naive

day 1

day 2

*****

***

40

*

20 0

C

***

40

0

4x

Naive

**

60

day 1

day 2

Thalamic EPSCs ** *** 1.2

FC

Ext

AMPA / NMDA ratio

FM

0.8 0.4

0.4 0.0

J Naive

FC

Ext

*

6

0.0

N CS only FM

*

**

**

4 4 2

2 0

0 N CS only FM

H Paired pulse ratio

Naive

CS only

AMPA / NMDA ratio

I

Cortical EPSCs * **

1.2

F Naive

G

day 3

0.8

N CS only FM

E

* **

*

N CS only FM

*

p=0.059

1.2

1.2

0.8

0.8

0.4

0.4

0.0

N

FC

Ext

0.0

4

4

2

2

0

N

FC

Ext

FC

Ext

*

0 N

FC

Ext

N

(legend on next page)

4 Neuron 86, 1–14, April 22, 2015 ª2015 Elsevier Inc.

Please cite this article in press as: Asede et al., Sensory Inputs to Intercalated Cells Provide Fear-Learning Modulated Inhibition to the Basolateral Amygdala, Neuron (2015), http://dx.doi.org/10.1016/j.neuron.2015.03.008

increase in PPR (Figures S3I and S3J). This indicates that both sensory inputs express functional presynaptic GABAB receptors in a similar fashion (compare Figures S3C and S3G). To unequivocally identify the source of sensory inputs, we turned to ex vivo optogenetic activation. Because our tracing experiments suggested that PIN/MGm and temporal cortex innervate mpITCs, we focused on these regions. First, we expressed ChR2 by rAAV in the PIN/MGm (Figures 3A–3C). Thalamic fibers were readily seen in acute brain slices, and at 3- to 4-weeks postinfection, light responses were evoked with properties resembling those of electrically stimulated EPSCs (Table S1). Importantly, specific PIN/MGm inputs were also presynaptically modulated by GABAB receptors (Figures 3D–3F). Revealing the morphology of recorded mpITCs, we confirmed that they were typical medium spiny ITCs that received PIN/MGm boutons in close apposition to dendritic spines (Figure 3C), corroborating that thalamic inputs target mpITCs. To identify the origin of cortical inputs, we stimulated ChR2expressing fibers after targeted expression in temporal cortex (TeA; Figure 3G). Labeled fibers from TeA coursed through external and intermediate capsules, giving off boutons in the mpITC cluster (Figures 3H and 3I). Light activation reliably resulted in short latency EPSCs with similar properties to electrically evoked (Table S1). Like thalamic afferents, specific temporal cortical inputs were presynaptically modulated by GABAB receptors (Figures 3J–3L). Optogenetic stimulation of either input pathway resulted in shorter latency responses than electrical stimulation, likely due to direct activation of fibers within the ITC cluster (Table S1). Taken together, our results demonstrate that defined and specific sensory thalamic and cortical afferents express functional GABAB heteroreceptors that can presynaptically depress these inputs. Interaction between Sensory Inputs Recruits Modulation by Presynaptic GABAB Receptors Sensory inputs onto LA PNs and local interneurons express presynaptic GABAB receptors, but their activation by physiological

stimuli was only detected at inputs to PNs, due to differential GABA accumulation and receptor localization (Pan et al., 2009; Shaban et al., 2006). In contrast to local interneurons, we hypothesized that heterosynaptic modulation of sensory afferents occurs within the mpITC GABAergic network. We adapted a previously published priming protocol (Pan et al., 2009) and applied tetanic stimulation to one of the pathways while assessing the effect on the other pathway (Figures 4A and 4B). Priming the thalamic input with cortical tetanic stimulation (CT), or vice versa (TC), resulted in a significant depression of the EPSC in the primed input (Figure 4C) that was accompanied by a correlated increase in the PPR (Figures 4D, S4A, and S4B). This activityinduced suppression was completely reversible (Figures 4C and 4D). Next, we sought to confirm that this crosstalk between inputs was indeed mediated by GABAB receptors. In keeping, application of the GABAB antagonist CGP55845 during priming significantly reduced the activity-dependent heterosynaptic depression and prevented the concomitant change in PPR in both pathways (Figures 4E–4H). Our data suggest that sensory inputs onto mpITCs, unlike those onto local BLA interneurons, are under presynaptic modulatory control by GABAB receptors recruited via GABA released from the mpITC network. This mechanism can reversibly modulate excitatory input strength in an activity-dependent manner. Activation of Sensory Inputs Drives Spike Output of mpITCs Since mpITCs are interconnected via inhibitory synapses and inputs are depressed by GABAB receptors, a key question is whether sensory inputs can drive mpITC spike output. Thalamic and cortical stimulation in the presence of intact inhibition elicited a biphasic excitatory/inhibitory (EPSC/IPSC) response in mpITCs (Figures 5A–5C). Consistent with the observed high excitation to inhibition (E/I) ratio, both inputs were able to evoke spikes when stimulus trains were applied (Figures 5C and 5D). To assess spike output noninvasively, we turned to cell-attached recordings. In keeping with and extending our previous findings,

Figure 2. Modification of Sensory Inputs to mpITCs upon Fear and Extinction Learning (A) Behavioral training and experimental groups: fear memory (FM), fear conditioning (FC), and extinction (Ext). (B) Freezing responses in all groups. Open bars, baseline (bl); closed bars, CS-evoked. Day 1: average freezing to last two CS-presentations; days 2 and 3: average freezing to blocks of 4 CSs. (Top) FM group (red) but not the CS only group (black) shows fear responses. Day 1: CS only (n = 9): bl 2.1% ± 1.5% and CS 3.8% ± 3.8%; FM (n = 10, data from one animal lost): bl 4.2% ± 2.1% and CS: 56.1% ± 9.8% (p = 0.001). Day 2: Cs only (n = 9): bl 4.2% ± 1.4% and CS1-4: 4.1% ± 1.3%; FM (n = 11): bl 7.7% ± 2.9% and CS1-4 55.5% ± 6.0% (p < 0.001). (Bottom) FC (orange) and Ext groups (blue) acquire and extinguish fear responses. Day 1: FC (n = 13) bl 5.4% ± 4.2% and CS, 54.2% ± 6.1% (p < 0.001); Ext (n = 9, data from two animals lost): bl 2.1% ± 1.2% and CS 51.2% ± 9.2% (p = 0.001). Ext day 2 (n = 11): bl 9.4% ± 2.5% and CS1-4 40.4% ± 4.3% (p < 0.001 versus bl), CS5-8 28.7% ± 4.3%, CS9-12 20.3% ± 2.8%, CS13-16 26.1% ± 5.7%. Ext day 3 (n = 11): bl 16.7% ± 2.0%, CS1-4 40.9% ± 4.7%, CS5-8 23.2% ± 4.3%, CS9-12 16.6% ± 2.6%, CS13-16 13.8% ± 2.0%. Stars denote reduction in freezing versus CS1-4 on day 2. (C–F) Changes in sensory inputs upon fear memory retrieval. (C) Normalized EPSCs (Vh = 70 mV) after thalamic (black) and cortical (gray) paired pulse stimulation. Scale bar represents 50 ms. (D) PPR increased upon FM retrieval (thalamic, ANOVA p < 0.001; cortical, ANOVA p = 0.006). Thalamic: naive 0.79 ± 0.06 (n = 27), Cs only 0.70 ± 0.05 (n = 24), FM 1.13 ± 0.09 (n = 27). Cortical: naive 0.75 ± 0.05 (n = 24), Cs only 0.70 ± 0.05 (n = 25), FM 0.99 ± 0.09 (n = 22). (E) Normalized AMPA-EPSCs (Vh = 70 mV) and combined AMPA-NMDA-EPSCs (Vh = +40 mV) after thalamic stimulation. Scale bar represents 50 ms. (F) AMPA/NMDA ratio decreased upon CS presentation and FM retrieval (thalamic, ANOVA p = 0.002; cortical, ANOVA p = 0.002). Thalamic: naive 4.34 ± 0.69 (n = 15), Cs only 4.81 ± 0.66 (n = 24), FM 2.36 ± 0.25 (n = 25). Cortical: naive 3.63 ± 0.61 (n = 15), Cs only 2.22 ± 0.29 (n = 17), FM 1.77 ± 0.23 (n = 19). (G–J) Changes in sensory inputs upon fear conditioning, but not extinction. (G) Normalized EPSCs (Vh = 70 mV) after thalamic (black) and cortical (gray) paired pulse stimulation. Scale bar represents 50 ms. (H) PPR increased upon FC and Ext in the thalamic (ANOVA p = 0.006) and showed a trend in the cortical pathway (ANOVA p = 0.06). Thalamic: naive 0.97 ± 0.06 (n = 35), FC 1.26 ± 0.08 (n = 37), Ext 1.25 ± 0.06 (n = 36). Cortical: naive 0.97 ± 0.08 (n = 27), FC 1.25 ± 0.10 (n = 30), Ext 1.30 ± 0.12 (n = 32). (I) Normalized AMPA-EPSCs (Vh = 70 mV) and combined AMPA-NMDA-EPSCs (Vh = +40 mV) after cortical stimulation. Scale bar represents 50 ms. (J) AMPA/NMDA ratio decreased in the cortical pathway upon FC (thalamic: ANOVA p = 0.41, cortical: ANOVA p = 0.02). Thalamic: naive 3.71 ± 0.36 (n = 28), FC 3.06 ± 0.39 (n = 27), Ext 3.55 ± 0.32 (n = 27). Cortical: naive 3.26 ± 0.41 (n = 19), FC 1.85 ± 0.26 (n = 20), Ext 2.55 ± 0.31 (n = 20). Stars denote significant post hoc tests in (D, H, F, and J).

Neuron 86, 1–14, April 22, 2015 ª2015 Elsevier Inc. 5

Please cite this article in press as: Asede et al., Sensory Inputs to Intercalated Cells Provide Fear-Learning Modulated Inhibition to the Basolateral Amygdala, Neuron (2015), http://dx.doi.org/10.1016/j.neuron.2015.03.008

GFP Biocytin ChR2

A

PIN/MGm inputs

B

Biocytin ChR2 C

BLA

-2.92

-3.08

-3.28 E

G

Bac

Bac+Cgp

F

*

** *** PPR (% Bsl)

Bsl

EPSC (% Bsl)

D

100 50 0

Bsl Bac Bac+Cgp

*

200 100 0

Bsl Bac Bac+Cgp

GFP ChR2

Temporal Cortex inputs

H

I

CeA BLA

-4.04

-4.24 K

Bsl

Bac

Bac+Cgp

EPSC (% Bsl)

J

*

*

100 50 0

Bsl Bac Bac+Cgp

L

PPR (% Bsl)

-3.80

200

*

*

100 0

Bsl Bac Bac+Cgp

Figure 3. Thalamic and Cortical Inputs from MGm/PIN and Temporal Cortex Are Presynaptically Modulated by GABAB Receptors (A–F) Optogenetically evoked MGm/PIN inputs are modulated by GABAB receptors. (A) Injection sites in MGm/PIN, bregma level is indicated (n = 5 animals). (B) ChR2-tDimer labeled axons and mpITC labeled during recording. Scale bar represents 50 mm. (C) Confocal image of a dendrite of filled mpITC with labeled axonal bouton in close apposition to a spine. Scale bar represents 10 mm. (D) Optogenetically evoked thalamic EPSCs (Vh = 70 mV) under Bsl, during application of Bac, and Bac+Cgp. Scale bars represent 40 pA, 50 ms. (E) Relative EPSC amplitude changes (n = 9, ANOVA p < 0.001): EPSCs were depressed in Bac (55.5% ± 8.1%, p = 0.002 versus Bsl) and recovered in Bac+Cgp (107.6% ± 13.5%, p < 0.001 versus Bac). (G–L) Relative PPR changes (J) (n = 9, ANOVA p < 0.001): PPR increased in Bac (210.8 ± 29.7, p = 0.03 versus Bsl) and recovered in Bac+Cgp (129.1% ± 26.9%, p = 0.01 versus Bac). Optogenetically evoked temporal cortex inputs are modulated by GABAB receptors. (G) Injection sites in temporal cortex. Bregma level is indicated (n = 6 animals). (H) ChR2-tDimer labeled axons in a slice used for recording. Scale bar represents 200 mm. (I) Confocal image stack with boutons of temporal cortical labeled axons in the mpITC cluster. Scale bar represents 30 mm. (J) Optogenetically evoked cortical EPSCs (Vh = 70 mV) under Bsl and during application of Bac and Bac+Cgp. Scale bars represent 40 pA, 50 ms. (K) Relative EPSC amplitude changes (n = 9, ANOVA p < 0.001): EPSCs were depressed in Bac (39.4% ± 7.8%, p = 0.02 versus Bsl) and recovered in Bac+Cgp (107.0% ± 11.0%, p = 0.01 versus Bac). (L) Relative PPR changes (n = 9, ANOVA p = 0.001): PPR was enhanced in Bac (184.0% ± 22.5%, p = 0.03 versus Bsl) and reversed in Bac+Cgp (89.4% ± 12.2%, p = 0.02 versus Bac).

6 Neuron 86, 1–14, April 22, 2015 ª2015 Elsevier Inc.

Please cite this article in press as: Asede et al., Sensory Inputs to Intercalated Cells Provide Fear-Learning Modulated Inhibition to the Basolateral Amygdala, Neuron (2015), http://dx.doi.org/10.1016/j.neuron.2015.03.008

A

T

C

T

T

B

C

T

C

C

C

**

50 0

T

C

CT TC

Con

T

G EPSC (% Bsl)

100

T

+Cgp

CT

T

*

CT

CT TC Con

*

100

C

F

50 0

T C

CT TC

Con

C

T C

+Cgp TC

H

CT TC Cgp

*

*

150

150

***

50

0

*

D

***

PPR (% Bsl)

E

100

***

PPR (% Bsl)

EPSC (% Bsl)

***

C

*

TC

*

100 50 0

CT TC Con

CT TC Cgp

Figure 4. Input Crosstalk Recruits Presynaptic GABAB Receptors on Sensory Afferents (A and B) Priming experiment (200 Hz, 10 pulses) to examine crosstalk between thalamic and cortical inputs. Scale bars represent 70 pA, 50 ms; Vh = 70 mV. (A) Thalamic input before, depressed thalamic input with increased PPR after cortical priming, recovered thalamic input. (B) Cortical input before depressed cortical input with increased PPR after thalamic priming, recovered cortical input. (C) Relative EPSC amplitudes for cortical priming of thalamic (CT, n = 19, ANOVA p < 0.001) and thalamic priming of cortical input (TC, n = 19, ANOVA p < 0.001). Thalamic EPSCs were reduced by CT (59.8% ± 6.8%, p < 0.001 versus T) and cortical EPSCs were reduced by TC (45.6% ± 7.7%, p < 0.001 versus C). Both recovered after priming (T, 95.0% ± 4.2%, p < 0.001 versus CT; C, 90.7% ± 5.2%, p = 0.001 versus TC). (D) Relative PPR for cortical priming of thalamic (CT, n = 19, ANOVA p = 0.001) and thalamic priming of cortical input (TC, n = 19, ANOVA p = 0.001). PPR was increased in both cases (CT, 146.4% ± 14.5%, p = 0.02 versus T; TC, 156.2% ± 18.1%, p = 0.02 versus C) and recovered (T, 111.2% ± 4.4%, p = 0.03 versus CT; C, 102.3% ± 5.6%, p = 0.03 versus TC). (E–H) Priming-induced presynaptic crosstalk is mediated by GABAB receptors. (E and F) EPSCs in priming experiments in absence and presence of Cgp55845 (Cgp). (E) Thalamic pathway, scale bars represent 35 pA, 50 ms. (F) Cortical pathway, scale bars represent 70 pA, 50 ms. (G) Relative EPSC amplitudes for priming in the absence and presence of Cgp. CT: 62.1% ± 8.5% in Con versus 82.6% ± 5.3% in Cgp (n = 12, p < 0.05) and TC: 48.2% ± 9.2% in Con versus 80.8% ± 5.8% in Cgp (n = 12, p < 0.001). (H) Relative PPR for

sensory afferent stimulation reliably produced action potential output that depended on glutamatergic synaptic transmission (Figure 5E), but spike probability decreased during train stimulation (Figure 5F). To test our hypothesis that GABAB receptors participate in this process, we determined steady-state spike probability to the last five stimuli in a train in the absence and presence of GABAA and GABAB antagonists (Figure 5G). We found that coapplication of CGP55845 and PTX, versus PTX alone or control, significantly increased spike probability (Figure 5H), suggesting that GABAB receptors depressed spike output during train stimulation. Together, our data show a functional role for GABAB receptors in modulating the efficacy of mpITC output in response to sensory input. mpITCs Innervate the BLA To address which neurons are targeted by mpITCs, we analyzed projection patterns of recorded cells. As recent data from juvenile animals identified diverse projections to local target regions (Busti et al., 2011; Geracitano et al., 2007), we asked whether in older animals mpITCs are similarly characterized by complex efferents. The vast majority of cells showed axonal patterns consistent with previously published cell types of capsular-, sublenticular-, and CeA-projecting cells (data not shown). Remarkably, 15% of filled mpITCs (n = 9 of 60 cells) gave off prominent axon collaterals into the BLA (Figures 6A–6C and S5A). Reconstruction of BLA-projecting neurons (n = 6) revealed that a large fraction of their axonal arbor (about 55%) was located within the BLA (Figures 6B and 6C). In some cases, axons traversed the entire BLA and reached the external capsule (Figure S5A). Within the BLA, mpITC axonal boutons were typically found in close apposition with CaMKII-positive, but not GFP-positive, profiles in the neuropil (Figure 6D), suggesting they synapse onto dendrites of BLA PNs. Occasionally, axons were also seen in close apposition to the soma of BLA PNs (Figure S5B). To confirm that mpITCs make functional synapses in BLA and to identify subcellular target structures, we employed preembedding electron microscopy. We found that biocytin-labeled boutons of mpITCs formed symmetrical synapses, suggesting that they are GABAergic (Figures 6E and S5C). The vast majority of synapses were onto small caliber dendrites (18 of 19 synapses, n = 3 cells; Figure 6F), sometimes in close proximity to spines, whereas one synapse was found on a somatic spine (Figure S5C). Together, this suggests that mpITCs primarily innervate dendrites of BLA PNs. To further corroborate that mpITCs innervate the BLA, we injected retrograde tracer (retrobeads) into the BLA (Figure S5D) and determined the fraction of retrobead-labeled ITCs by confocal imaging and cluster reconstruction (Figures 6G and 6H). Indeed, we found BLA-projecting ITCs within the medial paracapsular clusters, but as expected also in the lateral (Marowsky et al., 2005) and the ventrally located main ITC clusters (Figure 6I). Our results indicate that a comparable fraction of ITCs in each of these clusters projects to a similar target region within BLA. priming in the absence and presence of Cgp. CT: 137.0% ± 14.4% in Con versus 93.9% ± 3.2% in Cgp (n = 12, p = 0.01) and TC: 128.1% ± 16.6% in Con versus 97.6% ± 4.3% in Cgp (n = 12, p < 0.05).

Neuron 86, 1–14, April 22, 2015 ª2015 Elsevier Inc. 7

Please cite this article in press as: Asede et al., Sensory Inputs to Intercalated Cells Provide Fear-Learning Modulated Inhibition to the Basolateral Amygdala, Neuron (2015), http://dx.doi.org/10.1016/j.neuron.2015.03.008

B

C

T

Erev IPSC (mV)

C BLA

mpITC

-70 -50 -30 -10

Control

NBQX

F

Spike probability (%)

T

E

Figure 5. Sensory Inputs Drive Spike Output of mpITCs Modulated by GABAB Receptors

D EPSC/IPSC ratio

A

8 6 4 2 0

C

T

C

Thalamic stim

120

Cortical stim 10 Hz 20 Hz 40 Hz

100 80 60 40 20 0 0

2

4

6

8

10

0

2

4

6

8

10

Spike number (in trains)

G

H Control

PTX+CGP

Spike probability (%)

PTX

10 Hz

20 Hz

**

80

**

*

50

*

40

60

30

40

20 20 0

10 Con

PTX

PTX +CGP

0

mpITCs Provide Feedforward Inhibition to BLA PNs and Control Their Output Activity To address whether activation of mpITCs functionally inhibits BLA PNs, we first used restricted aperture stimulation of mpITCs expressing ChR2-tDimer (Figures S6A and S6B). In the absence of blockers, this consistently evoked pure inhibitory postsynaptic currents (IPSCs) in BLA PNs that were monosynaptic based on latency (Figures S6C and S6D). Because we could not completely rule out recruitment of other adjacent neurons, we turned to conditional expression of ChR2-YFP in Tac2-Cre mice (Figure S7A). As expected, ChR2-YFP expression was restricted to ITCs and a population of Tac2-positive cells in the CeA (Figures S7B and S7C; compare Mar et al., 2012). Lightinduced spikes were evoked in both infected CeA neurons and mpITCs (Figures S7D and S7E). Only light activation of the mpITC cluster, but not of the CeA, consistently resulted in synaptic responses in BLA PNs (Figures S7F and S7G). Further analysis confirmed that synaptic inputs from mpITCs were monosynaptic IPSCs (Figures 7A–7C). To assess functional consequences of this inhibitory input, we used several approaches (Figure 7D). First, we recruited the mpITC cluster optogenetically in Tac2-Cre mice just prior to subthreshold stimulation of excitatory thalamic inputs such that the offset of mpITC activation coincided with input stimulation. This resulted in a small, but highly significant reduction of EPSPs in LA 8 Neuron 86, 1–14, April 22, 2015 ª2015 Elsevier Inc.

Con

PTX

PTX +CGP

(A–F) Sensory inputs evoked biphasic EPSCs/ IPSCs and spike output. (A) Experimental scheme. (B) Thalamic and cortical EPSC/IPSC sequence recorded at different holding potentials ( 90, 70, 50 mV). Dotted line indicates where IPSC reversal potential was measured. Scale bars represent 200 pA, 50 ms. (C) Reversal potentials of IPSCs and EPSC/IPSC ratios were identical in both inputs (n = 6 each). IPSC reversal potentials: thalamic 63.6 ± 1.0 mV; cortical 64.5 ± 2.5 mV (p = 0.76). EPSC/IPSC ratios: thalamic 4.4 ± 1.0; cortical, 4.6 ± 1.7 (p = 0.91). (D) Spike activity in whole-cell mode evoked by thalamic and cortical stimulation (10 Hz). Scale bars represent 20 mV, 100 ms. (E and F) Stimulation of sensory inputs generated spike output in cell-attached recordings. (E) (Left) Cell-attached spikes (thalamic stimulation, 10 Hz) were blocked by 10 mM NBQX. (Right) Expanded spike. Scale bars represent left, 15 pA, 50 ms; right, 15 pA, 1 ms. (F) Average spike probability for stimulus trains at different frequencies (10, 20, and 40 Hz). Thalamic: n = 13, 14, and 13; cortical: n = 8, 9, and 9. (G and H) GABAB receptors modulated spike output. (H) Thalamic-evoked spikes to last 5 stimuli (10 Hz) under control conditions, in the presence of PTX, and PTX+CGP. Scale bars represent 20 pA, 100 ms. (G) Average spike probability was significantly reduced by application of PTX+CGP (n = 10). 10 Hz: Con 19.8% ± 5.4%, PTX 42.2% ± 10.1%, PTX+CGP 63.8% ± 11.7% (Con versus PTX+CGP, p = 0.009; PTX versus PTX+CGP, p = 0.004). 20 Hz: Con 4.4% ± 1.5%, PTX 12.4% ± 3.3%, PTX+CGP 36.0.8% ± 9.7% (Con versus PTX+CGP, p = 0.02; PTX versus PTX+CGP, p = 0.02).

PNs (Figures 7E and 7F). Second, we assessed the effect of mpITC activation on suprathreshold stimulation of thalamic inputs and found a significant reduction in spike probability of LA PNs (Figures 7F and 7G). Third, because mpITCs can also innervate the soma of BLA PNs, we tested whether mpITC activation reduces spike probability by somatic current injection and again found a significant reduction (Figures 7F and 7H). Together, this provides strong evidence that activation of mpITCs results in functional inhibition of BLA PNs to effectively control their output. Because sensory input activity can drive mpITCs spiking and this in turn inhibits BLA PN output, our data suggest that mpITCs are part of a sensory-driven feedforward inhibitory network controlling the BLA. mpITCs Are Also Part of Feedback Inhibitory Circuits Controlling BLA There is ample evidence from electrical stimulation experiments suggesting that BLA PNs innervate and activate mpITCs directly (Amano et al., 2010; Ju¨ngling et al., 2008; Royer et al., 1999). Thus, we hypothesized that BLA-projecting mpITCs could also provide feedback inhibition to BLA. To circumvent confounds of electrical stimulation, such as activation of fibers en passage or backfiring, we used specific optogenetic activation of BLA neurons to ask which mpITCs receive BLA inputs (Figure 8A). We locally co-injected retrobeads and

Please cite this article in press as: Asede et al., Sensory Inputs to Intercalated Cells Provide Fear-Learning Modulated Inhibition to the Basolateral Amygdala, Neuron (2015), http://dx.doi.org/10.1016/j.neuron.2015.03.008

Figure 6. mpITCs Project to and Make Symmetric Synapses in the BLA

GFP Biocytin

A

B

C 60

CeA

Axon length (%)

CeA

BLA BLA

40 20 0

*** ** *** *** IMC BLA Astr CeA EXC EPC Target Area

GFP Biocytin CaMKII

D

E

F dendrite diameter (µm)

t d

a

Target dendrites

1.0 0.8 0.6 0.4 0.2 0.0

cell #1 cell #2 cell #3 Analyzed mpITC GFP Retrobeads

G

I

H

CeA

m

l v

BLA

30 labeled ITCs (%)

d

GFP Retrobeads

Retrobead-labeled ITCs

25 20 15 10 5 0

ChR2-expressing virus in the BLA and recorded from retrobead-labeled and unlabeled mpITCs (Figures 8B–8E and S8A). Indeed, light stimulation evoked monosynaptic EPSCs in both retrobead-labeled BLA-projecting and unlabeled mpITCs (Figure 8E). When revealing the morphology of recorded mpITCs, we observed axonal boutons from BLA projection neurons in close apposition to mpITC spine heads (Figure 8C). Furthermore, axons of mpITCs that were driven by BLA-inputs projected back to the BLA (Figures S8B–S8D). In summary, our results demonstrate that BLA PNs, and not other fibers en passage, functionally innervate mpITCs and demonstrate that BLA-projecting mpITCs provide BLA-driven feedback inhibition to the BLA. DISCUSSION Here, we reveal a novel parallel circuit in the amygdala that directly relays sensory information to neurons in the mpITC cluster, which in turn send inhibitory projections to BLA. Sensory inputs to mpITCs arise from thalamic and cortical areas that convey CS- and US-related excitation to LA neurons during fear conditioning. These inputs undergo fear-related changes and are dynamically modulated by presynaptic GABAB heteroreceptors. MpITCs projecting to BLA efficiently control BLA PN

mp

lp main ITC cluster

(A) Confocal image of a biocytin-filled mpITC. Scale bar represents 30 mm. (B) Reconstruction of cell in (A) revealing axonal branches in BLA. Scale bar represents 200 mm, soma and dendrites in red, axon in blue. (C) Axonal quantification of six BLA-projecting mpITCs. IMC (intermediate capsule): 29.4% ± 10.1%, BLA: 54.9% ± 12.4%, Astr (amygdala striatal transition zone): 5.6% ± 2.7%, CeA (central amygdala): 8.6% ± 5.0%, EXC (external capsule): 0.8% ± 0.8%, EPC (endopiriform cortex): 0.4% ± 0.4%. Axonal length in BLA was larger than in all other regions (p % 0.001) except for IMC. (D) Confocal image of axonal branch (red) in the BLA. BLA PNs are stained for CaMKIIa (blue), interneurons shown in green (GFP). Scale bar represents 10 mm. (E) Electron micrograph of a filled mpITC axon (a) with a bouton (t) making a symmetrical synapse (arrow) onto a small caliber dendrite (d) in BLA. Scale bar represents 0.5 mm. (F) Target dendrite diameter for symmetrical synapses made by three mpITCs in BLA. (G) Confocal section through the medial paracapsular cluster with mpITCs (green, GFP) and BLA-projecting cells labeled with red retrobeads (arrows). Scale bar represents 30 mm. (H) 3D reconstruction of the cluster with distribution of BLA-projecting cells, retrobead-labeled cells in red. Scale bar represents 30 mm. (I) Quantification of BLA-projecting cells in different ITC clusters (three hemispheres, two animals, ANOVA p = 0.82). mpITC, medial paracapsular clusters 13.9% ± 4.3% (n = 7); lpITCs, lateral paracapsular clusters 10.8% ± 3.1% (n = 4); mITCs: main clusters 11.1% ± 1.6% (n = 4).

output and are reciprocally connected with them. In conclusion, mpITCs are part of sensory-driven feedforward and feedback inhibitory circuits to BLA PNs (Figure 8F). The LA receives CS and US inputs via specific thalamic and cortical regions, required for CS-US association during fear conditioning (Romanski et al., 1993; Romanski and LeDoux, 1992; Shi and Davis, 1999). We demonstrate that the same inputs also innervate mpITCs providing them with direct sensory information. Especially BLA-projecting mpITCs constitute a parallel feed-forward inhibitory pathway with distinct properties compared with local interneurons. Transmission onto mpITCs is mediated by AMPA and NMDA receptors and is modulated by presynaptic GABAB receptors, whereas a fraction of local LA interneurons express functional NMDA receptors (Polepalli et al., 2010; Szinyei et al., 2003), and presynaptic control by GABAB receptors is not recruited by physiological stimuli (Pan et al., 2009). Thus, sensory inputs onto mpITCs are endowed with unique properties to dynamically adjust their strength during stimulus processing or associations. Furthermore, input modulation rather than GABAergic interconnectivity (Geracitano et al., 2007; Royer et al., 2000) determined mpITC output efficacy. During conditioning, this could reduce mpITC-driven feedforward inhibition in a state-dependent manner to gate plasticity in target cells (Figure 8G, left). Neuron 86, 1–14, April 22, 2015 ª2015 Elsevier Inc. 9

Please cite this article in press as: Asede et al., Sensory Inputs to Intercalated Cells Provide Fear-Learning Modulated Inhibition to the Basolateral Amygdala, Neuron (2015), http://dx.doi.org/10.1016/j.neuron.2015.03.008

B

Tac2-Cre DIO-ChR-YFP

ChR-YFP FOXP2

BLA

C

Erev IPSC (mV)

A

mpITC stim 0 mV -20 mV -50 mV -70 mV

+ PTX

E T

T-EPSP T-EPSP + mpITC

BLA mpITC

-70 mV -50 mV

stim # 1 2 : : :

ISI 10 s

100

2 1 0

0 50 40 30 20 10 0

*** **

80

*

60 40 20 0

H

ISI 5 s

9 10 T-stim T-stim + mpITC

-20

T-EPSP T-spike Som. spike prob. size prob.

IPSP

G

F

Fraction of control (%)

D

4 3

-40

IPSC size (pA)

Latency (ms)

mpITC

-60

Somatic Somatic current current + mpITC

Synaptic plasticity of sensory inputs onto local GABAergic neurons in LA (Bauer and LeDoux, 2004; Mahanty and Sah, 1998; Polepalli et al., 2010) and CeA (Samson and Pare´, 2005) has been demonstrated in slices, but has not been linked to behaviorally driven plasticity. On the other hand, inputs along the serial processing stream, i.e., intra-amygdala inputs from LA onto a population of somatostatin-positive CeL cells (Li et al., 2013; Penzo et al., 2014) or BLA inputs onto ventral and dorsal mpITCs (Amano et al., 2010; Huang et al., 2014) undergo plasticity related to fear or extinction learning. Here, we report ex vivo changes pointing to an association-specific concerted decrease in sensory input strength onto mpITCs by presynaptic and postsynaptic alterations in thalamic AMPA/NMDA ratio upon fear retrieval. Initial changes in cortical AMPA/NMDA ratio may be driven by the CS and related to lasting changes in ITC firing in response to auditory stimuli in vivo (Collins and Pare´, 1999). Alterations in AMPA/NMDA ratio can most easily be explained by removal and addition of synaptic AMPA-Rs (Kessels and Malinow, 2009). While the decrease in release probability outlasts extinction training, postsynaptic changes are partially reversed. Although still correlative and calling for further mechanistic investigations, our findings suggest a role of sensory input plasticity onto mpITCs particularly in fear learning. Given the recent 10 Neuron 86, 1–14, April 22, 2015 ª2015 Elsevier Inc.

Figure 7. Activation of mpITCs Elicits Inhibition that Controls Spike Output of BLA PNs (A) Strategy for specific light activation of mpITCs by conditional expression of ChR2-YFP in Tac2Cre mice. (B) Confocal image of a slice used for recording, confirming ChR2-YFP expression in mpITCs colabeled with the marker FOXP2 (red). Scale bar represents 10 mm. (C) (Top) IPSCs recorded at different holding potentials in a BLA PN. Scale bars represent 30 pA, 50 ms. Average IPSC reversal potential (Erev = 45.6 ± 4.5 mV, n = 7). (Bottom) IPSC (Vh = 0 mV) in a BLA PN was blocked by Picrotoxin (100 mM). Bar graphs of IPSC data (n = 13), latency: 2.69 ± 0.49 ms, amplitude (Vh = 0 mV): 40.2 ± 6.1 pA. (D) Strategy to assess the effect of light activation of mpITCs on thalamic EPSPs and spike output in BLA PNs. (E) (Bottom) mpITC-evoked IPSP (light blue). (Top) Thalamic-evoked EPSP (black) and its reduction when preceded by 5 ms light activation of mpITCs (dark blue). Scale bars represent 1 mV, 60 ms. (F) mpITC activation reduced EPSP size and spike probability (percentage of control): EPSP amplitude (91.2% ± 2.4%, p < 0.001), thalamicevoked spike probability (60.2% ± 11.2%, p = 0.005), and somatic-evoked spike probability (64.0% ± 11.2%, p = 0.01). (G and H) Spike raster plots and example traces showing reduction in spike number when thalamic stimulation (G) or somatic current injection (H) was given without (black) or preceded by 5 ms light activation of mpITCs (dark blue). Scale bars represent 20 mV, 10 ms. Neurons were held at 70 mV, interstimulus intervals (ISIs) were 10 and 5 s for synaptic and somatic stimulation, respectively. Scale bars represent 20 mV, 5 ms.

finding that BLA inputs to mpITCs are enhanced following fear retrieval, and reversed upon extinction (Huang et al., 2014), a key question is how the temporal sequence and balance between these inputs dynamically shapes mpITC output to control fear. Based on our data, we propose a model in which during fear learning, mpITC-mediated sensory-driven feedforward inhibition is decreased to facilitate local plasticity, possibly in BLA, CeL, and mpITC cells at the same time. This in turn results in subsequent recruitment of mpITCs via enhanced inputs from activated LA neurons and regulates other downstream target cells that promote fear expression (Figure 8G, left). During fear retrieval, CS presentations would result first via decreased sensory inputs and then via increased LA inputs in disinhibition of fear-promoting and inhibition of fear-suppressing mpITC targets (Figure 8G, right). Together, opposing plasticity of sensory and LA inputs may alter the balance of target cell recruitment via feedforward and feedback inhibition. The finding of BLA-projecting cells further expands the complexity of mpITC projections (Amir et al., 2011; Busti et al., 2011). Despite the limitation of incomplete axonal sampling in brain slices, the observed innervation pattern implies that the BLA is a principal target. Only full reconstructions of axons from in vivo filled mpITCs (compare Bienvenu et al., 2012) will

Please cite this article in press as: Asede et al., Sensory Inputs to Intercalated Cells Provide Fear-Learning Modulated Inhibition to the Basolateral Amygdala, Neuron (2015), http://dx.doi.org/10.1016/j.neuron.2015.03.008

GFP ChR2 Biocytin

A

B

C

BLA mpITC ChR2-tDim BLA

Retrobeads Retrobeads GFP ChR2 BLA stim

E

D

NBQX

F PIN/MGm

4 3 2 1 0

-80

PPR EPSC

EPSC size (pA)

Latency (ms)

BLA

-60 -40 -20

nRB RB

0

nRB RB

G Fear Conditioning

TeA

LA

LA

LA 1 CeL 1

CeL

* *

mpITC

BA

nRB RB

Fear Memory

* * GABAB * *

1.0 0.8 0.6 0.4 0.2 0.0

2

2

2

main ITC

BA

main ITC

ultimately resolve this issue. Using complimentary approaches, we demonstrate that this subset of mpITCs provides functional inhibition to BLA PNs, a role previously ascribed only to lateral paracapsular ITCs and local interneurons (Ehrlich et al., 2009; Marowsky et al., 2005; Morozov et al., 2011). Most importantly, specific optogenetic activation of mpITCs reliably evoked inhibitory responses in BLA PNs. The prominent innervation of small to medium caliber dendrites and the ability to shape excitatory thalamic inputs is in line with a role for mpITCs in excitatory input compartmentalization and/or regulation of plasticity in dendrites (Bar-Ilan et al., 2012; Bienvenu et al., 2012; Chiu et al., 2013). The sparser somatic innervation likely contributed to suppression of somatically evoked spikes and suggests that mpITCs control BLA PN output by a combined action on distal dendrites and somata. Thus, mpITCs could act together with dendrite targeting and other interneurons that target dendritic and somatic domains (Capogna, 2014; Wolff et al., 2014).

BA

1

Figure 8. mpITCs Provide Feedback Inhibition to BLA PNs (A) Strategy for assessing inputs from ChR2expressing BLA neurons onto retrobead-labeled and unlabeled mpITCs. (B) Viral injection site in BLA with a recorded mpITC. Scale bar represents 250 mm. (C) ChR2-tDimer-expressing BLA axons in mpITC cluster and adjacent CeA and filled mpITC from (B). Scale bar represents 30 mm. (Insert) Labeled BLA axons in close apposition to dendrite of filled mpITC. (D) Co-injection of ChR2-tDimer and green retrobeads in the BLA. Scale bar represents 250 mm. (E) Optogenetic activation of BLA neurons elicited EPSCs in BLA-projecting mpITCs. (Top) EPSC (Vh = 70 mV) that was blocked by NBQX (10 mM). Scale bars represent 25 pA, 25 ms. (Bottom) Nonlabeled (nRB, n = 7) and retrobead-labeled (RB, n = 6) mpITCs received EPSCs with similar properties. Latency: nRB 2.77 ± 0.28 ms, RB 2.93 ± 0.34 ms (p = 0.52). EPSC amplitude: nRB, 50.0 ± 11.1 pA, RB 49.5 ± 10.4 pA (p = 0.94). PPR (at 50 ms): nRB 0.55 ± 0.22, RB 0.55 ± 0.10 (p = 0.91). (F and G) Summary and model of circuit changes. (F) Novel circuit elements (*) revealed in this study. Excitatory sensory inputs target mpITC dendrites and are regulated by GABAB heteroreceptors. A subset of mpITCs innervates LA and BA PNs with inhibitory synapses mainly onto dendrites. BLAprojecting mpITCs are reciprocally connected to BLA projection neurons. (G) Stimulus association during conditioning decreases sensory inputs to mpITCs (black and gray downward arrows), which results in reduced feedforward inhibition onto mpITC targets (green downward arrows). This disinhibition could promote plasticity (i.e., at sensory inputs to LA PNs, upward arrows), thereby increasing output of fear-promoting cells (upward arrows). During fear retrieval, CS presentations drive fear output via decreased sensory input strength (thin lines) and disinhibition of fearpromoting (1) and then via increased LA inputs and inhibition of fear-suppressing mpITC targets (2). Sensory inputs to LA cells are potentiated (thick lines).

In current models of amygdala function, mpITCs were thought to mainly act as inhibitory interface between BLA and CeA and a relay of feedforward inhibition from the IL-mPFC to CeA to suppress fear during extinction retrieval (Cho et al., 2013; Duvarci and Pare, 2014; Pape and Pare, 2010; Sierra-Mercado et al., 2011). Our findings define an additional role of mpITCs in fear circuits and fear learning. We propose that these pathways function in parallel and conjunction with known connectivity and plasticity mechanisms. Our observations on sensory input plasticity and our model are in line with an upregulation of the activity and plasticity marker zif in dorsal mpITCs upon fear learning (Busti et al., 2011). While mPFC inputs onto dorsal and ventral mpITCs were not altered by fear or extinction learning, our data suggest that mPFC activity could drive coordinated inhibition onto neurons in BLA and CeA to suppress fear output, especially when mPFC excitatory inputs to BLA are depressed (Cho et al., 2013). Furthermore, sensory and other excitatory inputs, Neuron 86, 1–14, April 22, 2015 ª2015 Elsevier Inc. 11

Please cite this article in press as: Asede et al., Sensory Inputs to Intercalated Cells Provide Fear-Learning Modulated Inhibition to the Basolateral Amygdala, Neuron (2015), http://dx.doi.org/10.1016/j.neuron.2015.03.008

including those from mPFC, could cooperatively control the level of feedforward inhibition. Our data are in line with the emerging view of parallel processing of sensory information and distributed plasticity within the amygdala (Balleine and Killcross, 2006; Ehrlich et al., 2009; Maren, 2005; Moscarello and LeDoux, 2013). It is intriguing to speculate that fear-related plasticity in connections between LA and CeL cells (Li et al., 2013; Penzo et al., 2014) is promoted by mpITC disinhibition when their sensory inputs are decreased. In contrast, increased LA input to mpITCs might drive fear expression by disinhibiting CeM via CeLOFF cells (Busti et al., 2011; Ciocchi et al., 2010; Haubensak et al., 2010; Huang et al., 2014). The same holds for the BLA, where cells with different response profiles and functions in high and low fear states exist (Amano et al., 2011; Herry et al., 2008; Repa et al., 2001; Senn et al., 2014). Thus, changes in stimulus-evoked mpITC inhibition could contribute to behaviorally state-dependent changes in activity of fear or extinction cells, or specific projection neurons in the BA. Given the potential parallels, a crucial question relates to the distinct function of the mpITC inhibitory circuit versus that of specific local interneurons in amygdala-dependent behaviors. To address this, tools allowing for in vivo manipulation of defined clusters and ITC projection types need to be developed. Even for the more intensely investigated BLA interneurons, only recently, a first study started to shed light on subtype-specific functions in awake behaving animals (Wolff et al., 2014). In conclusion, our findings reveal novel wiring and function of mpITCs that significantly extend our view on parallel inhibitory and disinhibitory processes in amygdala circuits. They also support the idea that a reciprocal interaction between mpITCs and BLA projection neurons may be part of a sensory feedback circuit engaged in time- and stimulus strength-dependent regulation of fear expression. EXPERIMENTAL PROCEDURES More details are available in Supplemental Information. Animals and Behavioral Procedures We used young adult male GAD67–GFP (Tamamaki et al., 2003) or Tac2-Cre transgenic mice (B6.129-Tac2/J; Jackson #18938). All procedures were approved by the Veterinary Department of the Canton of BaselStadt or the Regierungspraesidium Tuebingen. Mice were conditioned by five pairings of CS (30-s tone) and US (1-s mild footshock). The conditioning context differed from the extinction context. Extinction training was performed over 2 days with 16 CSs presented each day. Slice Recordings Coronal brain slices (320 mm) were prepared according to Hu¨bner et al. (2014). Patch-clamp recordings were performed in artificial cerebrospinal fluid (ACSF) at 30 C. For recording postsynaptic currents, we used Cs-methylsulphonatebased pipette solution, and for postsynaptic potentials and spikes, we used a K-gluconate-based solution. In most whole-cell recordings, biocytin was included in the pipette. For cell-attached recordings, pipettes were filled with ACSF. Data were acquired with a Multiclamp 700B amplifier, Digidata 1440, and Clampex software. Signals were filtered at 2 and digitized at 5 kHz for synaptic currents and filtered at 6 and digitized at 20 kHz for spike recordings. Data were excluded if series resistance changed >20%. EPSCs were evoked by bipolar tungsten electrodes (Science Products) or with 470nm light pulses (0.2–4 ms, 0.51–1.06 mW/mm2) from a light-emitting diode (CoolLed) through the 603 1.0 NA objective of an upright microscope

12 Neuron 86, 1–14, April 22, 2015 ª2015 Elsevier Inc.

(BX51; Olympus). Local stimulation was performed with a restricted aperture (80 mm diameter illumination spot). Data were analyzed with NeuroMatic (http://www.neuromatic.thinkrandom.com) and custom-written macros in IgorPro (Wavemetrics). Staining, Axonal Reconstructions, and Electron Microscopy Slices were fixed in 4% paraformaldehyde (PFA), 15% saturated picric acid solution, and 0.05% glutaraldehyde in PBS for 24 hr at 4 C, resectioned at 70 mm, and processed as described (Busti et al., 2011) with minor modifications. Biocytin was revealed using fluorescently conjugated Streptavidin. Immunostainings were performed by standard procedures; antibodies are listed in the Supplemental Information. Fluorescent sections were imaged using a laser-scanning microscope (LSM 710; Carl Zeiss). For reconstructions, sections were converted using the ABC Elite kit (Vector Laboratories) and Nickel-DAB as chromogen. Camera lucida drawings were performed on a Diaplan Microscope (Leitz Wetzlar) with a 403 0.8 NA objective, digitized, imported, and analyzed in ImageJ (NIH). Sections of BLA-projecting cells were permeabilized by freeze-thaw cycles after cryoprotection with 20% sucrose. Biocytin-filled axons were revealed with the ABC Elite kit (Vector Laboratories) and nickel-DAB. Contrast was enhanced using 2% osmium tetroxide (Agar Scientific) and 1% uranyl acetate (Agar Scientific) in 50% ethanol. Serial ultrathin sections (70 nm) were made from resin-embedded regions of interest, collected on Formvar-coated copper slot grids, and imaged on a Philips CM 120 electron microscope equipped with a Morada CCD TEM camera (Soft Imaging System). Stereotactic Injections and Analysis of Injection Sites and Labeled Cells GAD67-GFP mice were stereotaxically injected at the following coordinates (in mm from bregma): MGm/PIN: posterior 3.0, lateral ±1.8, ventral 3.8; temporal cortex: posterior 4.3, lateral ±4.6, ventral 3.5. For tracing, we injected 0.5 ml of 10% Fluoro-ruby (Invitrogen). For light activation, we injected 0.5 ml of rAAVCAG-ChR2(ET/TC)-tDimer or rAAV-CAG-hChR2(H134R)-mCherry (Penn Vector Core) and prepared slices after 4 weeks. To reveal and light-activate connected mpITCs and BLA neurons, we injected 0.25 ml retrobeads (Lumafluor), co-injected 0.5 ml retrobeads and rAAV-CAG-hChR2(H134R)-mCherry, or injected only rAAV-syn-hChR2(H134R)-YFP in the BLA at posterior 1.65, lateral ±3.4, and ventral 4.8. To target mpITCs, Tac2-Cre mice were injected with 0.5 ml of rAAV-EF1a-DIO-hChR2(H134R)-YFP at posterior 1.45, lateral ±3.35, and ventral 4.75. Slices were obtained after 2 to 3 weeks. For injections in PIN/MGm and TeA, brain slices caudal to amygdala were obtained and fixed in 4% PFA. For BLA injections, acute slices were fixed after recording. For tracing with Fluoro-ruby or retrobeads, 70 mm sections from perfusion-fixed brains were used. Injection sites were imaged on a fluorescence (Axio Imager, Carl Zeiss) or a laser-scanning microscope. To identify retrobead-labeled ITCs, z stacks of ITC clusters were obtained with a 633 1.4 NA objective and the pinhole set to 1 airy unit. Single and double-labeled cells were identified using automatic spot detection routines in Imaris (Bitplane). Statistics Data are presented as mean ± SEM. All statistical analyses were performed using SPSS (IBM); tests and results are detailed in Table S2. Data were considered significant if p < 0.05. Significance levels are denoted as follows: *p < 0.05, **p < 0.01, ***p < 0.001. SUPPLEMENTAL INFORMATION Supplemental Information includes Supplemental Experimental Procedures, eight figures, and two tables and can be found with this article online at http://dx.doi.org/10.1016/j.neuron.2015.03.008. AUTHOR CONTRIBUTIONS Experiments were designed by I.E., D.A., and F.F. and performed and analyzed by D.A., D.B., and I.E. in the lab of I.E., except for experiments in Figures 2G–2J, which were designed by I.E. and A.L. and performed by I.E. in the

Please cite this article in press as: Asede et al., Sensory Inputs to Intercalated Cells Provide Fear-Learning Modulated Inhibition to the Basolateral Amygdala, Neuron (2015), http://dx.doi.org/10.1016/j.neuron.2015.03.008

lab of A.L., and some anatomical and all electron microscopy experiments in Figures 6C, 6E, and 6F, which were performed by F.F.; I.E. and D.A. prepared figures. I.E. wrote the paper. All authors gave input and approved of the final version.

Dobi, A., Sartori, S.B., Busti, D., Van der Putten, H., Singewald, N., Shigemoto, R., and Ferraguti, F. (2013). Neural substrates for the distinct effects of presynaptic group III metabotropic glutamate receptors on extinction of contextual fear conditioning in mice. Neuropharmacology 66, 274–289.

ACKNOWLEDGMENTS

Duvarci, S., and Pare, D. (2014). Amygdala microcircuits controlling learned fear. Neuron 82, 966–980.

We thank A. Gall, G. Schmid, H. Riemenschneider, and M. Salib for technical support. This project was supported by the Werner Reichardt Centre for Integrative Neuroscience (CIN) at the University of Tu¨bingen (I.E.), an Excellence Cluster funded by the Deutsche Forschungsgemeinschaft (DFG) within the framework of the Excellence Initiative (EXC 307). Funding was also provided by the Hertie Foundation and the Brain and Behavior Research Foundation (NARSAD) (I.E.), the Volkswagenstiftung and Novartis Research Foundation (A.L.), the Austrian Science Fund Fonds zur Fo¨rderung der wissenschaftlichen Forschung (FWF) grant P-22969-B11 (F.F.), and a Network of European Neuroscience Schools (NENS) fellowship (D.A.). Received: January 2, 2014 Revised: January 7, 2015 Accepted: February 25, 2015 Published: April 2, 2015 REFERENCES

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. Fu, Y., and Shinnick-Gallagher, P. (2005). Two intra-amygdaloid pathways to the central amygdala exhibit different mechanisms of long-term potentiation. J. Neurophysiol. 93, 3012–3015. Geracitano, R., Kaufmann, W.A., Szabo, G., Ferraguti, F., and Capogna, M. (2007). Synaptic heterogeneity between mouse paracapsular intercalated neurons of the amygdala. J. Physiol. 585, 117–134. Haubensak, W., Kunwar, P.S., Cai, H., Ciocchi, S., Wall, N.R., Ponnusamy, R., Biag, J., Dong, H.W., Deisseroth, K., Callaway, E.M., et al. (2010). Genetic dissection of an amygdala microcircuit that gates conditioned fear. Nature 468, 270–276. Hefner, K., Whittle, N., Juhasz, J., Norcross, M., Karlsson, R.M., Saksida, L.M., Bussey, T.J., Singewald, N., and Holmes, A. (2008). Impaired fear extinction learning and cortico-amygdala circuit abnormalities in a common genetic mouse strain. J. Neurosci. 28, 8074–8085.

Amano, T., Unal, C.T., and Pare´, D. (2010). Synaptic correlates of fear extinction in the amygdala. Nat. Neurosci. 13, 489–494.

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.

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.

Huang, C.C., Chen, C.C., Liang, Y.C., and Hsu, K.S. (2014). Long-term potentiation at excitatory synaptic inputs to the intercalated cell masses of the amygdala. Int. J. Neuropsychopharmacol. 17, 1233–1242.

Amir, A., Amano, T., and Pare, D. (2011). Physiological identification and infralimbic responsiveness of rat intercalated amygdala neurons. J. Neurophysiol. 105, 3054–3066. Balleine, B.W., and Killcross, S. (2006). Parallel incentive processing: an integrated view of amygdala function. Trends Neurosci. 29, 272–279. Bar-Ilan, L., Gidon, A., and Segev, I. (2012). The role of dendritic inhibition in shaping the plasticity of excitatory synapses. Front. Neural Circuits 6, 118. Bauer, E.P., and LeDoux, J.E. (2004). Heterosynaptic long-term potentiation of inhibitory interneurons in the lateral amygdala. J. Neurosci. 24, 9507–9512. Bienvenu, T.C., Busti, D., Magill, P.J., Ferraguti, F., and Capogna, M. (2012). Cell-type-specific recruitment of amygdala interneurons to hippocampal theta rhythm and noxious stimuli in vivo. Neuron 74, 1059–1074. ko, M., Kaufmann, W., Busti, D., Geracitano, R., Whittle, N., Dalezios, Y., Man Sa¨tzler, K., Singewald, N., Capogna, M., and Ferraguti, F. (2011). Different fear states engage distinct networks within the intercalated cell clusters of the amygdala. J. Neurosci. 31, 5131–5144. Capogna, M. (2014). GABAergic cell type diversity in the basolateral amygdala. Curr. Opin. Neurobiol. 26, 110–116. Chhatwal, J.P., Myers, K.M., Ressler, K.J., and Davis, M. (2005). Regulation of gephyrin and GABAA receptor binding within the amygdala after fear acquisition and extinction. J. Neurosci. 25, 502–506. Chiu, C.Q., Lur, G., Morse, T.M., Carnevale, N.T., Ellis-Davies, G.C., and Higley, M.J. (2013). Compartmentalization of GABAergic inhibition by dendritic spines. Science 340, 759–762. Cho, J.H., Deisseroth, K., and Bolshakov, V.Y. (2013). Synaptic encoding of fear extinction in mPFC-amygdala circuits. Neuron 80, 1491–1507. Ciocchi, S., Herry, C., Grenier, F., Wolff, S.B., Letzkus, J.J., Vlachos, I., Ehrlich, I., Sprengel, R., Deisseroth, K., Stadler, M.B., et al. (2010). Encoding of conditioned fear in central amygdala inhibitory circuits. Nature 468, 277–282.

Hu¨bner, C., Bosch, D., Gall, A., Lu¨thi, A., and Ehrlich, I. (2014). Ex vivo dissection of optogenetically activated mPFC and hippocampal inputs to neurons in the basolateral amygdala: implications for fear and emotional memory. Front. Behav. Neurosci. 8, 64. Jimenez, S.A., and Maren, S. (2009). Nuclear disconnection within the amygdala reveals a direct pathway to fear. Learn. Mem. 16, 766–768. Ju¨ngling, K., Seidenbecher, T., Sosulina, L., Lesting, J., Sangha, S., Clark, S.D., Okamura, N., Duangdao, D.M., Xu, Y.L., Reinscheid, R.K., and Pape, H.C. (2008). Neuropeptide S-mediated control of fear expression and extinction: role of intercalated GABAergic neurons in the amygdala. Neuron 59, 298–310. Kessels, H.W., and Malinow, R. (2009). Synaptic AMPA receptor plasticity and behavior. Neuron 61, 340–350. Kim, J., Lee, S., Park, K., Hong, I., Song, B., Son, G., Park, H., Kim, W.R., Park, E., Choe, H.K., et al. (2007). Amygdala depotentiation and fear extinction. Proc. Natl. Acad. Sci. USA 104, 20955–20960. Knapska, E., and Maren, S. (2009). Reciprocal patterns of c-Fos expression in the medial prefrontal cortex and amygdala after extinction and renewal of conditioned fear. Learn. Mem. 16, 486–493. LeDoux, J.E. (2000). Emotion circuits in the brain. Annu. Rev. Neurosci. 23, 155–184. Li, H., Penzo, M.A., Taniguchi, H., Kopec, C.D., Huang, Z.J., and Li, B. (2013). Experience-dependent modification of a central amygdala fear circuit. Nat. Neurosci. 16, 332–339. 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.

Collins, D.R., and Pare´, D. (1999). Spontaneous and evoked activity of intercalated amygdala neurons. Eur. J. Neurosci. 11, 3441–3448.

Mahanty, N.K., and Sah, P. (1998). Calcium-permeable AMPA receptors mediate long-term potentiation in interneurons in the amygdala. Nature 394, 683–687.

Davis, M. (2000). The role of the amygdala in conditioned and unconditioned fear and anxiety. In The Amygdala, J.P. Aggleton, ed. (Oxford University Press), pp. 213–288.

Mar, L., Yang, F.C., and Ma, Q. (2012). Genetic marking and characterization of Tac2-expressing neurons in the central and peripheral nervous system. Mol. Brain 5, 3.

Neuron 86, 1–14, April 22, 2015 ª2015 Elsevier Inc. 13

Please cite this article in press as: Asede et al., Sensory Inputs to Intercalated Cells Provide Fear-Learning Modulated Inhibition to the Basolateral Amygdala, Neuron (2015), http://dx.doi.org/10.1016/j.neuron.2015.03.008

Maren, S. (2005). Synaptic mechanisms of associative memory in the amygdala. Neuron 47, 783–786. Marowsky, A., Yanagawa, Y., Obata, K., and Vogt, K.E. (2005). A specialized subclass of interneurons mediates dopaminergic facilitation of amygdala function. Neuron 48, 1025–1037. Mcdonald, A.J., Mascagni, F., and Guo, L. (1996). Projections of the medial and lateral prefrontal cortices to the amygdala: a Phaseolus vulgaris leucoagglutinin study in the rat. Neuroscience 71, 55–75. McKernan, M.G., and Shinnick-Gallagher, P. (1997). Fear conditioning induces a lasting potentiation of synaptic currents in vitro. Nature 390, 607–611. Millhouse, O.E. (1986). The intercalated cells of the amygdala. J. Comp. Neurol. 247, 246–271. Morozov, A., Sukato, D., and Ito, W. (2011). Selective suppression of plasticity in amygdala inputs from temporal association cortex by the external capsule. J. Neurosci. 31, 339–345. Moscarello, J.M., and LeDoux, J.E. (2013). The contribution of the Amygdala to aversive and appetitive pavlovian processes. Emot. Rev. 5, 248–253. Palomares-Castillo, E., Herna´ndez-Pe´rez, O.R., Pe´rez-Carrera, D., CrespoRamı´rez, M., Fuxe, K., and Pe´rez de la Mora, M. (2012). The intercalated paracapsular islands as a module for integration of signals regulating anxiety in the amygdala. Brain Res. 1476, 211–234. Pan, B.X., Dong, Y., Ito, W., Yanagawa, Y., Shigemoto, R., and Morozov, A. (2009). Selective gating of glutamatergic inputs to excitatory neurons of amygdala by presynaptic GABAb receptor. Neuron 61, 917–929. Pape, H.C., and Pare, D. (2010). Plastic synaptic networks of the amygdala for the acquisition, expression, and extinction of conditioned fear. Physiol. Rev. 90, 419–463.

Royer, S., Martina, M., and Pare´, D. (1999). An inhibitory interface gates impulse traffic between the input and output stations of the amygdala. J. Neurosci. 19, 10575–10583. Royer, S., Martina, M., and Pare´, D. (2000). Polarized synaptic interactions between intercalated neurons of the amygdala. J. Neurophysiol. 83, 3509– 3518. Samson, R.D., and Pare´, D. (2005). Activity-dependent synaptic plasticity in the central nucleus of the amygdala. J. Neurosci. 25, 1847–1855. Senn, V., Wolff, S.B.E., Herry, C., Grenier, F., Ehrlich, I., Gru¨ndemann, J., Fadok, J.P., Mu¨ller, C., Letzkus, J.J., and Lu¨thi, A. (2014). Long-range connectivity defines behavioral specificity of amygdala neurons. Neuron 81, 428–437. Shaban, H., Humeau, Y., Herry, C., Cassasus, G., Shigemoto, R., Ciocchi, S., Barbieri, S., van der Putten, H., Kaupmann, K., Bettler, B., and Lu¨thi, A. (2006). Generalization of amygdala LTP and conditioned fear in the absence of presynaptic inhibition. Nat. Neurosci. 9, 1028–1035. Shi, C., and Davis, M. (1999). Pain pathways involved in fear conditioning measured with fear-potentiated startle: lesion studies. J. Neurosci. 19, 420–430. Sierra-Mercado, D., Padilla-Coreano, N., and Quirk, G.J. (2011). Dissociable roles of prelimbic and infralimbic cortices, ventral hippocampus, and basolateral amygdala in the expression and extinction of conditioned fear. Neuropsychopharmacology 36, 529–538. Szinyei, C., Stork, O., and Pape, H.C. (2003). Contribution of NR2B subunits to synaptic transmission in amygdaloid interneurons. J. Neurosci. 23, 2549– 2556.

Penzo, M.A., Robert, V., and Li, B. (2014). Fear conditioning potentiates synaptic transmission onto long-range projection neurons in the lateral subdivision of central amygdala. J. Neurosci. 34, 2432–2437.

Tamamaki, N., Yanagawa, Y., Tomioka, R., Miyazaki, J., Obata, K., and Kaneko, T. (2003). Green fluorescent protein expression and colocalization with calretinin, parvalbumin, and somatostatin in the GAD67-GFP knock-in mouse. J. Comp. Neurol. 467, 60–79.

Pinard, C.R., Mascagni, F., and McDonald, A.J. (2012). Medial prefrontal cortical innervation of the intercalated nuclear region of the amygdala. Neuroscience 205, 112–124.

Trouche, S., Sasaki, J.M., Tu, T., and Reijmers, L.G. (2013). Fear extinction causes target-specific remodeling of perisomatic inhibitory synapses. Neuron 80, 1054–1065.

Polepalli, J.S., Sullivan, R.K., Yanagawa, Y., and Sah, P. (2010). A specific class of interneuron mediates inhibitory plasticity in the lateral amygdala. J. Neurosci. 30, 14619–14629.

Tsvetkov, E., Carlezon, W.A., Benes, F.M., Kandel, E.R., and Bolshakov, V.Y. (2002). Fear conditioning occludes LTP-induced presynaptic enhancement of synaptic transmission in the cortical pathway to the lateral amygdala. Neuron 34, 289–300.

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. Romanski, L.M., and LeDoux, J.E. (1992). Equipotentiality of thalamo-amygdala and thalamo-cortico-amygdala circuits in auditory fear conditioning. J. Neurosci. 12, 4501–4509. Romanski, L.M., Clugnet, M.C., Bordi, F., and LeDoux, J.E. (1993). Somatosensory and auditory convergence in the lateral nucleus of the amygdala. Behav. Neurosci. 107, 444–450.

14 Neuron 86, 1–14, April 22, 2015 ª2015 Elsevier Inc.

Wilensky, A.E., Schafe, G.E., Kristensen, M.P., and LeDoux, J.E. (2006). Rethinking the fear circuit: the central nucleus of the amygdala is required for the acquisition, consolidation, and expression of Pavlovian fear conditioning. J. Neurosci. 26, 12387–12396. Wolff, S.B., Gru¨ndemann, J., Tovote, P., Krabbe, S., Jacobson, G.A., Mu¨ller, C., Herry, C., Ehrlich, I., Friedrich, R.W., Letzkus, J.J., and Lu¨thi, A. (2014). Amygdala interneuron subtypes control fear learning through disinhibition. Nature 509, 453–458.