Excitation of Diverse Classes of Cholecystokinin Interneurons in the Basal Amygdala Facilitates Fear Extinction

Abstract There is growing evidence that interneurons (INs) orchestrate neural activity and plasticity in corticoamygdala circuits to regulate fear behaviors. However, defining the precise role of cholecystokinin-expressing INs (CCK INs) remains elusive due to the technical challenge of parsing this population from CCK-expressing principal neurons (CCK PNs). Here, we used an intersectional genetic strategy in CCK-Cre;Dlx5/6-Flpe double-transgenic mice to study the anatomical, molecular and electrophysiological properties of CCK INs in the basal amygdala (BA) and optogenetically manipulate these cells during fear extinction. Electrophysiological recordings confirmed that this strategy targeted GABAergic cells and that a significant proportion expressed functional cannabinoid CB1 receptors; a defining characteristic of CCK-expressing basket cells. However, immunostaining showed that subsets of the genetically-targeted cells expressed either neuropeptide Y (NPY; 29%) or parvalbumin (PV; 17%), but not somatostatin (SOM) or Ca2+/calmodulin-dependent protein kinase II (CaMKII)-α. Further morphological and electrophysiological analyses showed that four IN types could be identified among the EYFP-expressing cells: CCK/cannabinoid receptor type 1 (CB1R)-expressing basket cells, neurogliaform cells, PV+ basket cells, and PV+ axo-axonic cells. At the behavioral level, in vivo optogenetic photostimulation of the targeted population during extinction acquisition led to reduced freezing on a light-free extinction retrieval test, indicating extinction memory facilitation; whereas photosilencing was without effect. Conversely, non-selective (i.e., inclusive of INs and PNs) photostimulation or photosilencing of CCK-targeted cells, using CCK-Cre single-transgenic mice, impaired extinction. These data reveal an unexpectedly high degree of phenotypic complexity in a unique population of extinction-modulating BA INs.


Introduction
The basolateral amygdala complex is a neural structure subserving a range of behavioral functions and neural processes, including emotional regulation, and is implicated in the pathophysiology of anxiety and traumarelated disorders (Bukalo et al., 2014;Tovote et al., 2015). The amygdala is comprised of an assortment of cells which differ in their neurochemical identity and efferent and afferent connectivity, but the functional contribution of specific neuronal subpopulations to fear and extinction remains to be fully elucidated (Janak and Tye, 2015). Within the basal nucleus of the amygdala (BA), as in other cortical structures, the activity of principal neurons (PNs) is tightly regulated by local inhibitory GABAergic interneurons (INs) and there is growing evidence that local INs provide a critical regulatory component of the circuits mediating fear and extinction (Ehrlich et al., 2009;Mc-Cullough et al., 2016;Krabbe et al., 2018;Lucas and Clem, 2018).
Although INs represent a highly diverse set of cell types, there are evolving efforts to classify them based on their morphologic features, physiologic characteristics and molecular phenotype (DeFelipe et al., 2013). One commonly adopted method segregates IN subpopulations based on neurochemical content, including expression of Ca 2ϩ -binding proteins [e.g., parvalbumin (PV); McDonald and Betette, 2001;McDonald and Mascagni, 2001b] and neuropeptides such as somatostatin (SOM), neuropeptide Y (NPY), and cholecystokinin (CCK; Mascagni and Mc-Donald, 2003;Kepecs and Fishell, 2014). In turn, there is emerging evidence for critical contributions of some of these subpopulations to fear behaviors. For example, an elegant series of studies has shown that PV-containing and SOM-containing BA INs act in concert to gate the responses of PNs to conditioned stimulus (CS) and un-conditioned stimulus (US) during fear memory acquisition . However, the possible roles of other subpopulations of BA INs, including NPY-expressing (Ma Ј nko et al., 2012) and CCK-expressing (Rovira-Esteban et al., 2017), in regulating fear and extinction remain unclear.
Taken together, the current literature support the hypothesis that CCK INs affect extinction by dynamically adjusting the balance of inhibitory control over opposing BA output-pathways (Vogel et al., 2016). In this context, the goal of the current study was to further characterize the electrophysiological, morphological and molecular properties of CCK INs in the BA and assess their contribution to fear extinction. To this end, we employed an intersectional approach (Fenno et al., 2014) to genetically access and manipulate BA CCK INs (Taniguchi et al., 2011;Senn et al., 2014;Whissell et al., 2019).

Subjects
C57BL/6J (JAX stock #000664), CCK-IRES-Cre (Cck tm1.1(cre)Zjh /J, JAX stock #012706) and Dlx5/6-Flpe (Tg(mI56i-flpe)39Fsh/J, JAX stock #010815) mice were obtained from The Jackson Laboratory. Homozygous Cck-Cre and homozygous Dlx5/6-Flpe mice were bred to produce Cck Cre ;Dlx5/6 Flp (CCK IN) mice. Males and females were used for electrophysiological recordings and male mice were used for behavioral testing. Mice were housed in same-sex groupings (two to four per cage); mice with chronic fiber implantations for in vivo optogenetics were single-housed after surgery to prevent cagemates damaging the cranial implants. Housing was in a temperature-and humidity-controlled vivarium under a 12/12 h light/dark cycle (lights on 6 A.M.). Experiments were conducted during the light phase.
All experimental procedures were performed in accordance with the Institutional Ethical Codex, Hungarian Act of Animal Care and Experimentation (1998. XXVIII. section 243/1998, renewed in 40/2013, the European Union guidelines (directive 2010/63/EU), the National Institute of Health (NIH) Guide for the Care and Use of Laboratory Animals and approved by the Institutional Animal Care and Use Committee of the Institute of Experimental Medicine and the local National Institute on Alcohol Abuse and Alcoholism (NIAAA) and Vanderbilt Animal Care and Use Committees.

Stereotaxic surgery
Mice were placed in a stereotaxic frame (David Kopf Instruments) to bilaterally inject viral constructs into the BA (coordinates: anterior-posterior -1.4 to 1.5 mm, medial-lateral Ϯ3.22 to 3.3 mm, dorsal-ventral -4.4 to 4.85 mm to bregma). Virus was injected in a volume of 0.2 l per hemisphere at a rate of 3 nl/s (for ex vivo optogenetics) or in a volume of 0.4 -0.5 l per hemisphere over 10 min (for in vivo optogenetics), according to each laboratory's local practices and pilot work. Injections were done using a 1-l syringe (Neuros model 7001 KH, Hamilton Robotics) connected to a UMP3 UltraMicroPump and SYS-Micro4 Controller or Nanoliter NL2010MC4 injector (World Precision Instruments, LLC). The syringe was left in place for an additional 5 min to ensure constructs diffused into the tissue. For in vivo optogenetics, during the same surgery as viral injections, ferrules and 200-m diameter fiber optics (numerical aperture, 0.37) were bilaterally inserted into the BA and affixed to the skull with dental cement. The ferrule-fiber assembly was constructed according to previously published methods (Bukalo et al., 2015;Bergstrom et al., 2018;Radke et al., 2019).

Fluorescence in situ hybridization
At least five weeks after delivery of AAVdj-hSyn-Con/ Fon-hChR2(H134R)-EYFP-WPRE, CCK IN mice were killed by cervical dislocation, then brains were immediately removed and frozen in 2-methyl butane on dry ice and stored at -80°C. Coronal sections, 16-m-thick, were cut using a cryostat (model HM500 OM, Microm International GmbH) and mounted directly onto Super Frost Plus slides (Fisher Scientific) and maintained at -20°C before transfer to a staining jar containing 4°C 10% buffered formalin solution. After a 20-min fixation, slides were rinsed twice in PBS for 1 min each, then dehydrated in an ethanol dilution series (50%, 70%, and 100% ϫ2) and stored at -20°C overnight in 100% ethanol.
The following day, the sections were processed using an RNAscope Fluorescent Multiplex Assay kit (Advanced Cell Diagnostics USA; Gunduz-Cinar et al., 2019) according to the manufacturer's instructions. Slides were first air dried for 10 min, then a hydrophobic barrier drawn around each section with an ImmEdge barrier pen (Vector Laboratories) to limit the spread of solutions. The sections were treated with Pretreat-4 protease solution for 20 min at room temperature and slides washed twice with distilled water. In one set of samples, target probes for Mm-Cck (gene ID 12424, catalog #402271) and Mm-Gad1 (glutamate decarboxylase 1, GAD-67; gene ID: 14415, catalog #400951-C2) were spread evenly with a pipette tip on the sections, and the slides were incubated at 40°C for 2 h in a hybridization oven (model HybEZ, Advanced Cell Diagnostics). A second set of samples were treated with Mm-Cck and Mm-Slc17a7 (VGluT1; gene ID 72961, catalog #416631-C2). These probes and those described below were obtained from Advanced Cell Diagnostics.
Sections were next treated with amplifier and fluorescent probes to separately label each gene (AMP1 at 40°C for 30 min; AMP2 at 40°C for 15 min, AMP3 at 40°C for 30 min, and AMP4 AltB at 40°C for 15 min). Slides were washed twice with 1ϫ wash buffer according to the manufacturer's guidelines between incubation steps. Finally, the sections were incubated for 20 s with DAPI at room temperature, and then coverslipped with Vectashield Hard-Set fluorescent mounting medium (Vector Laboratories). For each brain, separate sections were processed as either (1) a negative control to confirm the absence of background labeling on each channel using a bacterial mRNA (DapB of Bacillus subtilis strain; 3-plex Negative Control Probe, catalog #320871), or (2) a positive control to confirm the ability to detect the presence of labeling on each channel, using three housekeeping genes [RNA polymerase II subunit A (Polr2a), peptidylprolyl isomerase B (Ppib), ubiquitin C (Ubc); 3-plex Positive Control Probe, catalog #320881]. Images of the BA were obtained on 405-, 488-, 550-nm fluorescent channels using a confocal microscope (model LSM 700, Carl Zeiss Microscopy LLC) under a Plan-Apochromat 20ϫ/0.8 M27 objective.

Immunostaining for IN subtype markers
At least five weeks after delivery of AAVdj-hSyn-Con/ Fon-hChR2(H134R)-EYFP-WPRE, mice were terminally overdosed with ketamine/xylazine and transcardially perfused first with saline followed by ice cold 4% PFA in phosphate buffer. After removing the brain from the skull postfixation took place at 4°C for 3 h. Coronal sections (50-m-thick) were cut using a vibratome (model VT100S, Leica Biosystems).

Ex vivo optogenetics Electrophysiological slice recordings
Ten to 12 weeks after injection of the INTRSECT-ChR2 virus, mice were deeply anesthetized with isoflurane, the brain was quickly removed and placed into ice-cold solution containing: 252 mM sucrose, 2.5 mM KCl, 26 mM NaHCO 3 , 0.5 mM CaCl 2 , 5 mM MgCl 2 , 1.25 mM NaH 2 PO 4 , 10 mM glucose, bubbled with 95% O 2 /5% CO 2 (carbogen gas). Horizontal 200-m thick brain sections containing the BA were prepared with a vibratome (model VT1200S, Leica Biosystems) and kept in an interface-type holding chamber containing ACSF at 36°C that gradually cooled down to room temperature. ACSF contained the following: 126 mM NaCl, 2.5 mM KCl, 1.25 mM NaH 2 PO 4 , 2 mM MgCl 2 , 2 mM CaCl 2 , 26 mM NaHCO 3 , and 10 mM glucose, bubbled with carbogen gas. After at least a 60-min-long incubation, slices were transferred to a submerged-type recording chamber and perfused with 32-34°C ACSF with a flow rate of 2-2.5 ml/min.
Recordings were performed under visual guidance using differential interference contrast microscopy (via a model FN-1 Nikon upright microscope) using 40ϫ water dipping objective. EYFP expression was visualized with the aid of a mercury arc lamp and a CCD camera (Andor Technology). Neurons that did not express EYFP were recorded where EYFPϩ fibers were densest. Patch pipettes (5-7 M⍀) for whole-cell recordings were pulled from borosilicate capillaries with inner filament (thin walled, OD 1.5) using a P1000 pipette puller (Sutter Instrument). In whole-cell recordings the patch pipette contained a K-gluconate based intra-pipette solution containing the following: 115 mM K-gluconate, 4 mM NaCl, 2 mM Mg-ATP, 20 mM HEPES, 0.1 mM EGTA, 0.3 mM GTP (sodium salt), and 10 mM phosphocreatine adjusted to pH 7.3 using KOH, with an osmolarity of 290 mOsm/l. The pipette also contained 0.2% biocytin. Recordings were performed with a Multiclamp 700B amplifier (Molecular Devices), low-pass filtered at 3 kHz, digitized at 10 kHz, recorded with Clampex 10.4 (Molecular Devices), and were analyzed with Clampfit 10.4 (Molecular Devices) and OriginPro 2015 (OriginLab Corp).
Whole-field blue light (447 nm) laser illumination (Roithner Laser Technik, Vienna, Austria) was applied for 100 ms using a Digital Mirror Device based pattern illuminator (Mightex Polygon 400, Mightex Systems) to activate even those neurons expressing ChR2 that membrane time constants are rather slow. The recorded neurons were clamped at a holding potential of -50 mV. Series resistance was in the range of 15-25 M⍀. For peak and area analysis, five consecutive traces were averaged. Drug effects were evaluated after a 10-min wash-in of bathapplied gabazine (5 M), CGP 5699A (1 M), and CP 55,940 (2 M), or after 20-min wash-in of bath-applied AM251 (2 M). The peak amplitude for the fast and slow components was determined at distinct time points (n ϭ 23). The area of these two components was calculated only for those events in which the fast component was blocked by gabazine (n ϭ 10): the remaining slow component was subtracted from the original trace resulting in the area for the fast component. For firing pattern analyses, EYFPϩ neurons were recorded in current clamp mode at a holding potential of -65 mV. Voltage responses were tested to a series of hyperpolarizing and depolarizing square current pulses of 800-ms duration and amplitudes between -100 and 100 pA at 10-pA step intervals, then up to 300 pA at 50-pA step intervals and finally up to 600 pA at 100-pA step intervals.

Postrecording IN subtype identification
Biocytin content was visualized using Cy3-conjugated streptavidin (1:10,000, catalog #S6402, Sigma-Aldrich) and confocal images of the filled cells were obtained using a confocal microscope (Nikon model C2) under a Plan-Apochromat VC 20ϫ objective (N.A. 0.75, z step size: 1 m, xy: 0.40 m/pixel). Cells were immunostained with antibodies selected on the bases of a combination of their firing characteristics and the features of their dendritic and axonal arbors. Putative CCK ϩ basket cells were immunostained with goat anti-CB1R antibody (1:1000, catalog #CB1-Go-Af450, Frontier Institute) and visualized using DyL405-conjugated donkey anti-goat antibody (1: 500, catalog #705-475-003, Jackson ImmunoResearch). Only those cells (five of seven) expressing CB1R in their axonal terminals were categorized as CCK ϩ basket cells.
To further confirm the latter classification, in vitro slices containing both PVϩ basket cells and axo-axonic cells were re-sectioned into 30 m-thick sections, pepsin digested as described previously (Veres et al., 2014) and immunostained using a mouse anti-ankyrin G antibody (1:100, catalog #75-146, NeuroMab) visualized with an Alexa Fluor 488-conjugated donkey anti-mouse antibody (1:500, catalog #A21202, Millipore). Those cells which showed cartridges of axonal terminals in close apposition to ankyrin Gϩ profiles were confirmed as axo-axonic cells. Because the cells selected for recording and subsequent biocytin filling based on being EYFPϩ, there was a low level of EYFP fluorescence in their axonal terminals. In these images, the red channel containing the signal of the Cy3-conjugated streptavidin used to visualize the biocytin was subtracted from the green channel, and the axonal terminal apposition to ankyrin Gϩ profiles was evaluated in these modified images merged with the original image taken in red channel. All images were obtained using a confocal microscope (Nikon model C2) under a Plan Apo VC 60ϫ objective (N.A. 1.4, z step size: 0.15-0.2 m, xy: 0.08 -0.10 m/pixel).

In vivo optogenetics In vivo photostimulation and photosilencing
For photostimulation, blue light ( ϭ 473 nm) was bilaterally shone on the BA at 20 Hz, in 5-ms pulses, consistent with commonly used ChR2-excitation parameters and mimicking the activity of highly active neurons. For photosilencing, green light ( ϭ 532 nm) was bilaterally shone on the BA continuously, which is also in line with commonly used Arch-excitation parameters. For IN-TRSECT experiments in CCK IN mice, laser power was set to 7 mW both for blue and green laser, measured at the tip of optic fiber. For the ChR2-photostimulation experiment in CCK-Cre mice, the laser power was reduced to 3 mW after pilot data showed 7-mW produced indications of seizure activity. Laser power was calibrated before each experiment by measuring the power at the tip of the patch cord with a PM100D optical power meter with an S120C sensor (Thorlabs) and multiplying that power by the transmittance of the ferrule connection on each optic fiber.

Behavioral testing
Fear conditioning and extinction testing was conducted (Whittle et al., 2010) at least four weeks after virus delivery. Before testing, each mouse was handled for 2 min/d for 6 d and habituated to the connected optic-fiber cables in the home cage for 40 min/d for 3 d.

Fear conditioning
Fear conditioning was conducted in context A: a 30 ϫ 25 ϫ 25 cm operant chamber (Med Associates, Inc.) with metal walls and a metal rod floor. To provide an additional olfactory cue, the chamber was cleaned between subjects with a 79.5% water: 19.5% ethanol: 1% vanilla extract solution. Beginning after a 120-to 180-s stimulusfree period, there were 3ϫ pairings (60-to 90-s interpairing interval) between a 30-s, 80-dB white noise cue (CS) and a 2-s 0.6-mA scrambled co-terminating footshock (US) followed by a 120-s stimulus-free period. The Med Associates Video Freeze Monitor System controlled CS and US presentation.

Fear extinction training
Fear extinction training occurred the following day in context B: a 27 ϫ 27 ϫ 14 cm operant chamber with transparent walls and a floor covered with wood chips, cleaned between subjects with a 99% water:1% acetic acid solution and housed in a different room from training. After a 180-s stimulus-free baseline, there were 50 ϫ 30-s CS presentations (5-s inter-CS interval). During each CS, light (blue for photostimulation, green for photosilencing) was shone on the BA.

Extinction retrieval
Extinction retrieval was tested the day after extinction training in context B, via 5 ϫ 30-s no-light CS presentations (5-s inter-CS interval) beginning after a 180-s stimulus-free baseline. Mice were connected to the fiber cables during the retrieval test, to control for the influence of a potential extinction-context cue.
Freezing, the absence of any visible movement except respiration, was scored from video every 5 s throughout testing by an experienced observer blind to genotype. The mean number of freezing observations per baseline period and 5ϫ CS block was converted to a percentage [(number of freezing observations/total number of observations per period) ϫ 100] for analysis. Mice with freezing scores on any CS-block that were Ͼ3 SDs from the group mean were excluded from the analysis.

Novel open field test
Mice were tested in the novel open field test for exploratory locomotion and anxiety-like behaviors (Fitzgerald et al., 2010) 7 d after the completion of fear and extinction testing. The apparatus was a square arena (39 ϫ 39 ϫ 35 cm) with opaque white Plexiglas walls and floor (ϳ95 lux). Mice were connected to the fiber optic cables and placed in the center for a 10-min test session, during which light (blue for photostimulation, green for photosilencing, as above) was shone on the BA for 2 ϫ 2-min periods, interspersed with 2-min light-off periods (i.e., off-on-off-on-off). Total distance traveled, average movement speed and time spent in a 25-cm 2 center zone was measured by the EthovisionXT videotracking system (Noldus Information Technology).

Postbehavior confirmation of fiber placement and virus expression
At the completion of behavioral testing, mice were terminally overdosed with ketamine/xylazine and transcardially perfused with PBS, then 4% PFA. After overnight suspension in 4% PFA, followed by 0.1 M phosphate buffer for 1-2 d at 4°C, 50-m coronal sections were cut with a vibratome (Leica Biosystems). Sections were mounted and coverslipped with Vectashield HardSet mounting medium with 4=,6-diamidino-2-phenylindole (Vector Laboratories) and imaged with a fluorescent (model BX41, Olympus America, Inc.) or confocal microscope (model LSM 700, Carl Zeiss Microscopy) under a Plan-Apochromat 20ϫ/0.8 M27 objective.

Statistical analysis
Group effects in electrophysiological experiments were analyzed using paired t test. For scatterplots, each symbol represents the mean average of five consecutive events. Group effects in behavioral experiments were analyzed using Student's t test or ANOVA, depending on the number of groups, followed by Fisher's LSD post hoc tests paired Student's t test. The threshold for statistical significance was set at p Ͻ 0.05. Values given in the Results represent the mean Ϯ SEM.

Intersectional strategy for neuronal targeting
To gain selective genetic access to CCK INs, a doubletransgenic mouse line was generated by intercrossing a CCK-Cre driver line with a Dlx5/6-Flpe driver line, which expresses Flp recombinase in most cortical GABAergic neurons Taniguchi et al., 2011;Vogel et al., 2016;Fig. 1A). We then used an intersectional optogenetic viral strategy to visualize and control the targeted cells, involving the bilateral injection of CCK IN double-transgenic mice with INTRSECT viruses transfecting cells with EYFP-tagged ChR2 in a manner conditional on the presence of both Cre and Flp recombinases (Fenno et al., 2014;Fig. 1B,C). Initial, non-quantitative immunocytochemical staining for the GABAergic neuronal marker, GAD67, indicated co-localization with EYFP cells in the BA (Fig. 1D). Similarly, fluorescence in situ hybridization showed that EYFP expression appeared concordant with the labeling of BA cells that were also positive for Cck and Gad1 mRNA probes (Fig. 1E).

Targeted neurons are GABAergic
To further assess the selectivity of the intersectional targeting approach employed, we performed ex vivo electrophysiological recordings in CCK IN mice transfected with the INTRSECT-ChR2 virus ( Fig. 2A-C). Whole-cell recordings from EYFP-negative, likely PNs (n ϭ 23), which were postsynaptic to EYFP-expressing neurons in BAcontaining brain slices, revealed that blue light illumination evoked outward currents, but with substantial variance both in the peak amplitude and decaying phase (Fig. 2D,G). Further inspection of these responses indicated that in some cases, the light-evoked currents clearly had fast and slow components, recognized by distinct peaks (Fig. 2E). The two outward components had significantly different peak amplitude (fast component: 201.6 Ϯ 27.8 pA, slow component: 63.6 Ϯ 10.7 pA, n ϭ 23, t (43) ϭ 4.55, p Ͻ 0.001; Fig. 2G), but they carried similar charge (fast component: 18.4 Ϯ 6.2 pC, slow component: 8.4 Ϯ 2.2 pC, n ϭ 10, p ϭ 0.1; Fig. 2J).

Targeted neurons express functional CB1R
Based on earlier reports that CCK-containing basket cells in the BA express CB1R, we next asked whether the targeted BA INs were CB1R-positive (Katona et al., 2001;McDonald and Mascagni, 2001a;Vereczki et al., 2016;Veres et al., 2017). To test the responsivity of the targeted cells to CB1R activation, we generated light-evoked postsynaptic currents (PSCs) in EYFP-negative cells followed by bath application of the CB1R agonist, CP 55,940. The application of the agonist reduced the peak amplitude of light-evoked events by 50% (control: 209.8 Ϯ 88.4 pA, in CP: 112.6 Ϯ 47.6 pA, peak ratio: CP/ctr: 51.3 Ϯ 5.4%, n ϭ 5, t (4) ϭ 2.35, p ϭ 0.039; Fig. 2K,L). We verified that the CP 55,940-induced reduction was CB1R mediated by abolishing the effect via preincubation with the CB1R antagonist, AM251 (peak amplitude in AM251: 261.8 Ϯ 70.2 pA, in AM251 ϩ CP 55,940: 259.5 Ϯ 78.4 pA, peak ratio: AM251 ϩ CP 55,940/AM251: 96.7 Ϯ 8.9%, n ϭ 4, p ϭ 0.89; Fig. 2M,N). These data demonstrate that a significant component of synaptic currents in BA neurons evoked by light illumination in slices stems from signaling through the CB1R. Consistent with these findings, immu-nostaining BA sections containing transfected processes revealed that ϳ40% of EYFP-expressing axonal varicosities were positive for CB1R (Fig. 2O-R). These results indicate that a significant proportion of the targeted cells exhibit a defining feature of CCK-expressing basket cells.

Subsets of targeted INs express NPY or PV
While our findings indicate that a substantial proportion of the targeted cells had properties of CCK/CB1Rexpressing basket cells, two observations led us to wonder whether other IN populations were also targeted. First, light-evoked responses were not fully blocked by CB1R agonism, in contrast to earlier recordings obtained in BA (Vogel et al., 2016) and second, they displayed a prominent GABA B receptor-mediated component in the light-evoked outward current that is uncharacteristic of basket cells (Vogel et al., 2016;Rovira-Esteban et al., 2017;Veres et al., 2017).
This led us to perform immunolabeling for presence of various IN neurochemical markers in EYFP-expressing cells. We found that ϳ29% of    Vereczki et al., 2016;Andrási et al., 2017). Finally, in line with our in situ hybridization and electrophysiological data, none of the EYFP-expressing cells examined (n ϭ 51) were immunopositive for a glutamatergic neuronal maker, CaMKII (Fig. 3D).

Targeted INs are morphologically diverse
Our immunolabeling results suggest that, in addition to CCK/CB1R-expressing basket cells, three IN subtypes: PVϩ basket cells, PVϩ axo-axonic cells and NPYϩ, likely neurogliaform cells (NGFCs), were targeted. To substantiate this, we intracellularly-labeled and immunostained EYFPϩ neurons obtained from our slice preparations to allow for a direct comparison between the firing properties and neurochemical phenotype of each cell (Fig. 4A). Of 33 EYFPϩ cells labeled, all had dendritic and axonal morphologic features consistent with INs and not PNs. Examination of these cells based on the action potential width at half maximum, 50% decay time of the afterhyperpolarization (AHP) and maximum firing rate led us to classify three main subcategories which, in turn, corresponded well to differences in their respective immunocytochemically features (Fig. 4B).
One group was characterized by a fast-spiking phenotype (i.e., narrow action potential and high, Ͼ100-Hz firing-rate; Fig. 4C). Within this group, both PVϩ basket cells (n ϭ 3; Fig. 4D) and PVϩ axo-axonic cells (n ϭ 5; Fig.  4E) were identifiable; the former showed immunoreactivity for Calb and avoided the axon initial segments visualized by Ankyrin G staining, while the latter lacked Calb and their axonal varicosities formed close appositions with continued (D) and decaying phase (D, E) is typical for events evoked in different neurons. The traces in E are peak scaled. Dashed lines show where the peak amplitude for fast and slow components of evoked currents was measured. F, Traces from an experiment measuring the antagonist-sensitivity of light-evoked responses. Gabazine (5 M) wash-in eliminated the fast GABA A -mediated component, while CGP 5699A (1 M) blocked the remaining slow GABA B -mediated component. Importantly, no inward, i.e., EPSC, could be observed in the presence of the GABA receptor antagonists, indicating that the applied intersectional strategy allowed us to excite selective GABAergic cells. G, Peak amplitude of the fast components in evoked responses measured in the same neurons was significantly larger than the peak amplitude of the slow components. H, The fast components were blocked by bath application of gabazine (‫ء‬paired t test). I, The slow components were eliminated by CGP 5699A wash-in. J, The area, i.e., the charge of the fast and slow components evoked in the same neurons, was not different. GABA A receptor-mediated fast responses were isolated by subtracting the responses recorded in the presence of gabazine from the control traces and their area was measured. Example traces (  axon initial segments. A second group discharged action potentials with an intermediate width and at the lowest rate. The cells in this group had axonal varicosities immu-nopositive for CB1R, identifying them as CCK/CB1expressing basket cells (n ϭ 5; Fig. 4F). The third last group had the widest action potentials and longest AHP In each case, a maximal intensity projection of a 3D confocal image of the labeled INs is shown together with its firing pattern and the EYFP expression at the soma level. D, An example for a PVϩ basket cell (PVBC) identified based on its firing pattern, Calb and PV positivity in its axonal boutons (white arrows in insets) and forming no close appositions (red arrows) with ankyrin G (AnkG)-labeled axon initial segments (delimited by green arrowheads). E, An example for a PVϩ axo-axonic cell (AAC) identified based on its firing pattern, PV positivity and Calb negativity in its axonal boutons (white arrows in insets) and forming close appositions by its axonal boutons (red arrows) with an AnkG-labeled axon initial segment (delimited by green arrowheads). F, An example for a CCKBC identified based on its firing pattern and on the CB1 content in its axonal boutons (white arrows in insets). G, An example of a NGFC based on its dendritic and axonal morphology and characteristic firing pattern. H, Pie chart showing the ratio of identified IN types in a group of EYFP-expressing neurons in the BA that were randomly sampled in slice preparations. For D-G depictions of maximal intensity projections of intracellularly filled cells, scale bar ϭ 40 m, insets ϭ 5 m; firing pattern scale bar x-axis ϭ 100 ms, y-axis ϭ 10 mV. and were identifiable as NGFCs based on previously published results (Tamás et al., 2003;Armstrong et al., 2012;Ma Ј nko et al., 2012; n ϭ 20; Fig. 4G). Indeed, quantification revealed the majority (ϳ60%) of the in vitro recorded and labeled cells fell into the latter, NGFC, subclass, with roughly equal (ϳ9-15%) proportions in the other classes (Fig. 4H).

Photostimulating targeted INs facilitates fear extinction
Our next experiments assessed the contribution of the targeted population of BA INs to fear extinction. BA INs were infected with INTRSECT-ChR2 (Fig. 5A), INTRSECT-Arch, or an INTRSECT-EYFP control virus, and tested for fear conditioning, extinction training and extinction retrieval over consecutive days, using our standard extinction training protocol (Bukalo et al., 2015). For fiber placement maps see Extended Data Figure 5-1. During extinction training (only), blue or green light, respectively, was shone concomitant with each CS presentation.

Non-selectively photostimulating or photosilencing BA CCK cells disrupts fear extinction
Given the observation of an extinction-facilitating effect of photostimulating BA INs, we wondered how this effect would compare with the effect of manipulating a combination of BA INs and BA CCK-expressing glutamatergic neurons. We transfected cells of single-transgenic CCK-Cre mice in the BA with Cre-dependent ChR2, eArch3.0 or an EYFP control virus (Fig. 6A,B). Given the majority of amygdalar glutamatergic neurons express CCK (Allen Brain Institute, experiment: 77869074), we expected the effect of non-selectively photostimulating BA CCK cells would largely reflect stimulation of glutamatergic cells and thereby differ from the effect of BA INs. Using fluorescence in situ hybridization we found that following virus injections into the BA, EYFP-labeled neurons were positive for Cck mRNA and that some of these cells showed labeling for Gad1 and other for Slc17a7, consistent with the transfection of both CCK INs and PNs (Fig. 6C,D;Andrási et al., 2017).
We next performed fear conditioning and extinction testing (as above). For fiber placement maps see Extended Data Figure 6-1.
Finally, we used a novel open field test to confirm that the differences observed in the extinction retrieval were due to the manipulation of the BA CCK ϩ cells participating in fear circuits, and not to a different locomotor or anxiety level within the different groups. In this test, groups did not differ in anxiety-related exploration of the aversive center zone (effect of group: F (2,43) ϭ 8.098, p ϭ 0.001) or total distance traveled in the arena (effect of group: F (2,43) ϭ 8.098, p ϭ 0.001; Extended Data Fig. 6-2).

Discussion
There is growing appreciation of the critical contribution of INs to the regulation of network activity to support the ChR2 group during each CS presentation of extinction training increased freezing during training and light-free extinction retrieval the following day, relative to EYFP controls. Photosilencing in the eArch3.0 group during each CS presentation of extinction training reduced freezing on light-free extinction retrieval the following day, relative to EYFP controls; n ϭ 12-27 per group; ‫ء‬p Ͻ 0.05, #p Ͻ 0.05 T0 versus T1 extinction training. Results of similar interventions on locomotion and anxiety level are presented in Extended Data Figure 6-2. behavioral functions subserved by the BA, including Pavlovian fear and extinction. However, a description of the subclass of BA INs expressing the neuropeptide, CCK, has proved elusive. Here, we sought to genetically access BA CCK INs to define the neurochemical and physiologic phenotype of these cells and assess their possible contribution to fear extinction.
Using an intersectional approach, entailing transfecting Creϩ/Flpϩ cells in CCK-Cre;Dlx5/6-Flpe transgenic mice with opsin-containing INTRSECT viruses, we selectively targeted GABAergic INs in the BA and showed that almost half of these expressed functional CB1R on their axonal boutons. In line with this anatomic observation, the application of a CB1R agonist reduced the amplitude of light-evoked IPSCs by half. Taken together, this is strong evidence that a significant portion of the cells targeted by INTRSECT strategy are CCK-expressing basket cells, given prior studies have shown that CCK-expressing basket cells express CB1R on their boutons that, when activated, markedly suppress inhibitory transmission (Katona et al., 2001;McDonald and Mascagni, 2001a;Azad et al., 2004;Vogel et al., 2016;Barsy et al., 2017;Rovira-Esteban et al., 2017;Veres et al., 2017). This functional connection between CCK and CB1R has been of particular interest to the field given compelling evidence linking endocannabinoid signaling, in the BA and other regions, to extinction (Patel et al., 2017). One proposal is that endocannabinoids could inhibit CCK release in the BA (Beinfeld and Connolly, 2001;Chhatwal et al., 2009) and thereby oppose the peptide's pro-fear/anxiety effects to enable extinction (Frankland et al., 1997;Harro, 2006;Del Boca et al., 2012;Bowers and Ressler, 2015).
We found that the genetically targeted INs also displayed a prominent GABA B receptor-mediated component in the light-evoked outward current that likely does not originate from CCK-expressing basket cells. Indeed, further examination using a combination of immunostaining, electrophysiological recordings and morphologic inspection indicated that targeting CCK cells also included PVϩ basket cells, PVϩ axo-axonic cells, and NGFCs. A portion of the latter INs might express NPY (Ma Ј nko et al., 2012). Based on available data, we conclude that recombination in these cells reflects the genuine presence of CCK in adult BA cells at low levels that were not detected by prior studies using other approaches. This conclusion is based on several observations. First, other recent studies using genetic approaches also support a significant diversity of CCK-expressing INs in the BA, among which INs with NGFC morphology and firing characteristics were evident (Jasnow et al., 2009;Vogel et al., 2016;Veres et al., 2017). Second, CCK mRNA was unequivocally detected in axo-axonic cells both in the neocortex (Paul et al., 2017) and hippocampus (Harris et al., 2018) and even in a portion of hippocampal PV-expressing basket cells (Harris et al., 2018). Third, using a different approach to that used herein, involving crossing offspring of CCK-Cre and Nkx2.1-Flp mice with an Ai65 reporter line, another recent study reported labeling of axo-axonic cells in the neocortex (Paul et al., 2017). Fourth, dense core vesicles are readily observed in axon terminals of both PV-containing basket and axo-axonic cells (Takács et al., 2015), indicating the presence of neuropeptides in these GABAergic cell types that have not been labeled so far in SOM-Cre, NPY-Cre, or VIP-Cre mouse lines. Taken together with these prior observations, our results strongly speak to the importance of applying multiple phenotypic criteria when classifying CCK IN cells and underscore the limitations of demarcating this population based on a single, neurochemical marker (Harris et al., 2018). The difference in ratio of distinct IN types obtained by immunostaining in perfused tissue (Figs. 2, 3) and by randomly sampled neurons recorded in in vitro slices (Fig. 4) may reflect the fact that some cell types are better able to tolerate the procedure of slice preparation than others.
The finding that optogenetic photostimulation of the targeted IN population in the BA produced behavioral effects is indicative of a facilitation of fear extinction. This is reminiscent of the recent finding that brain-wide chemogenetic activation of a population of INs geneticallyaccessed using a CCK-Cre;Dlx5/6-Flpe transgenic strategy similar to ours, improved performance on memory tasks (contextual fear and discrimination, social and object recognition, puzzle box; Whissell et al., 2019). Together, these behavioral effects raise intriguing questions about the relative contribution of specific subsets of INs that are targeted by this strategy, given our data show that the population are not simply "CCK-expressing," at least assessed by immunostaining. Earlier studies have implicated BA INs including NPY-expressing and PVexpressing cells in extinction (Gutman et al., 2008;Herry et al., 2010;Verma et al., 2012;Tovote et al., 2015). For instance, a reduction in GABAergic input from PVexpressing axo-axonic cells onto BA PNs resulted in an impaired extinction learning (Saha et al., 2017). However, whether one specific subset disproportionally contributes to the extinction-facilitating effects of stimulation remains to be determined and will be technically challenging to address, given the lack of exclusive markers for each given subpopulation. Another important goal will be positioning the various subclasses into the micro and macro circuits mediating extinction. Prior work has already demonstrated important functional connections both locally between different IN subtypes in the BA  and distally, via CCK IN projections to the medial PFC regions (Senn et al., 2014). In summary, our data together with recent findings imply that excitation of GABAergic microcircuits in the BA via local or distal projections with cortical or subcortical origin could potentially augment extinction memory formation.
In experiments using single CCK-Cre mice, we targeted a substantial portion of PNs (in addition to CCK INs) located in the BA, reflecting the fact that most excitatory neurons express CCK in this nucleus ; see also Allen Brain Atlas, experiment: 77869074). In contrast to the extinction-facilitating effects of photostimulating the IN population, photostimulation or photosilencing of globally-targeted (i.e., INs and PNs) BA CCK cells, via Cre-dependent opsin transfection in CCK-Cre single-transgenic mice, during extinction training led to an impairment in long-term extinction memory, as evidenced by higher freezing on a light-free extinction retrieval test. Of note, photostimulation also reduced freezing during initial extinction training (i.e., fear retrieval), which suggests either an acute anxiolytic-like effect or a failure to retrieve the fear memory. Alterna-tively, this reduction in freezing could reflect a photostimulation-induced potentiation of CS-induced escape behaviors, which would be in line with panicogenic effects of CCK agonism (de Montigny, 1989;Bradwejn et al., 1991;Kramer et al., 1995). This interpretation awaits further testing although we did not detect effects of photostimulation on motor or anxiety-related behaviors in a novel open field test in which no CS was presented.
Our data obtained in single CCK-Cre mice showing that photoinhibition of BA neurons impaired fear retrieval are in line with recent findings using a distinct mouse line, Thy1-Cre. As in our case, photoinhibition of Thy1-expressing, mainly PNs in the BA during extinction training resulted in a weakening in extinction memory formation (McCullough et al., 2016). These results suggest that BA circuits contain neural populations able to control fear extinction memory (Herry et al., 2008). Impairments in fear extinction are evident in various neuropsychiatric conditions, including trauma-and stressor-related disorders and some anxiety disorders. This has encouraged basic researchers to define the neurobiological basis of impaired and intact fear extinction as a potentially tractable approach to developing new treatments for these disorders. The resultant research has defined the amygdala as a central node within a distributed neural system comprising cortical, hippocampal, and midbrain structures, among others. The current findings add to a growing literature by describing a unique population of INs in the BA that, when activated, exert strong modulatory effects on extinction.