Molecular, Morphological, and Functional Characterization of Corticotropin-Releasing Factor Receptor 1-Expressing Neurons in the Central Nucleus of the Amygdala

Abstract The central nucleus of the amygdala (CeA) is a brain region implicated in anxiety, stress-related disorders and the reinforcing effects of drugs of abuse. Corticotropin-releasing factor (CRF, Crh) acting at cognate type 1 receptors (CRF1, Crhr1) modulates inhibitory and excitatory synaptic transmission in the CeA. Here, we used CRF1:GFP reporter mice to characterize the morphological, neurochemical and electrophysiological properties of CRF1-expressing (CRF1+) and CRF1-non-expressing (CRF1–) neurons in the CeA. We assessed these two neuronal populations for distinctions in the expression of GABAergic subpopulation markers and neuropeptides, dendritic spine density and morphology, and excitatory transmission. We observed that CeA CRF1+ neurons are GABAergic but do not segregate with calbindin (CB), calretinin (CR), parvalbumin (PV), or protein kinase C-δ (PKCδ). Among the neuropeptides analyzed, Penk and Sst had the highest percentage of co-expression with Crhr1 in both the medial and lateral CeA subdivisions. Additionally, CeA CRF1+ neurons had a lower density of dendritic spines, which was offset by a higher proportion of mature spines compared to neighboring CRF1– neurons. Accordingly, there was no difference in basal spontaneous glutamatergic transmission between the two populations. Application of CRF increased overall vesicular glutamate release onto both CRF1+ and CRF1– neurons and does not affect amplitude or kinetics of EPSCs in either population. These novel data highlight important differences in the neurochemical make-up and morphology of CRF1+ compared to CRF1– neurons, which may have important implications for the transduction of CRF signaling in the CeA.

CRF and CRF 1 (Crhr1) expressing neurons are located in several brain regions including the central nucleus of the amygdala (CeA), which functions as the main output nucleus for amygdala functions (Gilpin et al., 2015). The CeA is comprised of medial (CeM) and lateral (CeL) subdivisions (Sun et al., 1991;Petrovich et al., 1996;Dong et al., 2001;Ciocchi et al., 2010;Davis et al., 2010;Haubensak et al., 2010;McCullough et al., 2018) and contains mostly GABAergic projection neurons and interneurons (Sun and Cassell, 1993;Veinante and Freund-Mercier, 1998). The CeM sends inhibitory projections to various effector regions [e.g., hypothalamus, periaqueductal gray, locus coeruleus, bed nucleus of the stria terminalis (BNST), and pedunculopontine tegmental nu-cleus; Pitkänen and Amaral, 1994]. The CeL sends inhibitory inputs to CeM, thereby gating the output activity of the CeA, but also to more distant brain regions such as the periaqueductal gray and paraventricular nucleus of the thalamus (Penzo et al., 2014).
Accumulating evidence implicates the CRF/CRF 1 system in the CeA in many animal models of physiological and pathological conditions (Chu et al., 2007;Funk et al., 2007;Roberto et al., 2010;Bruijnzeel et al., 2012;Lowery-Gionta et al., 2012;Iemolo et al., 2013;Ji et al., 2013;Baiamonte et al., 2014). However, cellular heterogeneity in this region has limited the identification and full functional characterization of the CRF 1 -expressing (CRF 1 ϩ) subset. Thus, we used a bacterial artificial chromosome (BAC) transgenic mouse line expressing the green fluorescent protein (GFP) under the Crhr1 promoter (CRF 1 :GFP) to readily identify neurons expressing CRF 1 (Justice et al., 2008). Previously CRF 1 ϩ neurons were observed to be mainly located in the CeM and exhibit an ongoing tonic GABAergic conductance driven by action potential-dependent GABA release. In contrast, CRF 1 non-expressing (CRF 1 -) neurons display no ongoing tonic inhibition (Herman et al., 2013). Although this functional analysis has yielded significant insight into cell type-specific properties in CeA microcircuits, a precise molecular characterization of the CRF 1 ϩ neuronal population is lacking.
Given the critical role of CRF and CRF 1 in the CeA, we sought to identify unique molecular, morphologic and functional properties that distinguish CeA CRF 1 ϩ neurons from their CRF 1 -neighbors. Utilizing CRF 1 :GFP mice we determined the following characteristics of CeA CRF 1 ϩ and CRF 1 -neurons: (1) co-expression patterns of CRF 1 (via GFP reporter) with CBPs parvalbumin (PV), calretinin (CR), and calbindin (CB); (2) co-expression patterns of Crhr1 with Penk, Pdyn, Sst, Npy, Crh, and Prkcd; (3) dendritic spine morphology and density; and (4) basal and CRF-modulated glutamatergic transmission. We found that Penk and Sst have the highest percentage of coexpression with Crhr1 in both the CeM and CeL, and that Penk is enriched in CeM Crhr1ϩ neurons compared to their Crhr1-neighbors. We also show that CeM CRF 1 ϩ neurons exhibit an overall lower spine density and differential fractions of mature versus immature spines compared to CRF 1 -neurons. Consistent with a comparable density of mature spines in these two populations, we found no difference in spontaneous EPSCs (sEPSCs) or miniature EPSCs (mEPSCs). Importantly, acute CRF application increased overall CeM glutamatergic transmission and does not affect amplitude or kinetics of EPSCs in either population. These data provide important information about the neurobiology of CeA CRF 1 ϩ neurons that may have critical implications for their functional role under physiological and pathological conditions.

Materials and Methods
All procedures were approved by our Institutional Animal Care and Use Committee and were consistent with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Transgenic CRF 1 :GFP mice were generated on a mixed C57BL/6J x BALB/c background using BAC recombination (see Justice et al., 2008 for transgene design and immunohistochemical validation of reporter expression). A colony was established at our institute in 2008 and has been backcrossed to C57BL/6J mice every two to three years since. Mice were genotyped by PCR on tail snip lysates using the following transgenespecific primers: forward 5'-GGT CAC CCC AAA AAT AAT CTC T-3'; reverse 5'-AGG ATT GGG AAG ACA ATA GC-3'. We also amplified a positive control band using the following primers: forward 5'-TCC TCA AAG ATG CTC ATT AG-3'; reverse 5'-GTA ACT CAC TCA TGC AAA GT-3'. Adult male mice were used for all experiments.

Tissue preparation
Mice (n ϭ 4) were anesthetized (isoflurane) and perfused with cold PBS followed by 4% paraformaldehyde (PFA). Brains were dissected and immersion fixed in PFA for 24 h at 4°C, cryoprotected in sterile 30% sucrose in PBS for 24 -48 h at 4°C or until brains sank, flash frozen in pre-chilled isopentane on dry ice, and stored at -80°C. Free floating 35-m brain sections were obtained using a cryostat and kept at 4°C in PBS containing 0.01% sodium azide.

Imaging and analysis
Sections were imaged on a Zeiss LSM 780 laser scanning confocal microscope (10ϫ objective, tile scanning of CeA). All microscope settings were kept the same within experiments during image acquisition. Analyst was blind to the identity of the red fluorescent signal (CBP) when performing cell counts, and analysis was performed manually in an unbiased manner at four anterior-posterior levels (equidistant sections located -1.00 through -1.70 mm from bregma). Data are presented as mean Ϯ SEM.

In situ hybridization Tissue preparation
Mice (n ϭ 3-4) were anesthetized (isoflurane) and perfused with cold PBS/Z-fix (Fisher Scientific, NC9378601). Brains were dissected and immersion fixed in Z-fix for 24 h at 4°C, cryoprotected in sterile 30% sucrose in PBS for 24 -48 h at 4°C or until brains sank, flash frozen in prechilled isopentane on dry ice, and stored at -80°C. Brains were then sliced on a cryostat in 20-m-thick sections, mounted on SuperFrost Plus slides (Fisher Scientific, 1255015) and stored at -80°C.

In situ hybridization
In situ hybridization was performed using RNAscope fluorescent multiplex kit (ACD, 320850) in RNase-free conditions. To perform the RNAscope in situ hybridization, a target retrieval pretreatment protocol was performed as outlined in the manual (ACD, doc. no. 320535). Briefly, slides were submersed in target retrieval buffer (ACD, 322000) at 95-98°C for 10 min, immediately washed in distilled water, dehydrated with 100% ethanol (storage at -80°C if required), and lastly digested with Protease IV for 20 min at 40°C. Following this pretreatment the RNAScope Fluorescent Multiplex Reagent kit User Manual (ACD, doc.no. 320293) was followed and slides were mounted with DAPI-containing Vectashield (Fisher Scientific, NC9029229). A negative control (ACD, 320751) was run in tandem for each experiment. The probes used from ACD Biotechne were as follows: Crhr1

Imaging and analysis
Slides were imaged on a Zeiss LSM 780 laser scanning confocal microscope (40ϫ oil immersion, 1024 ϫ 1024, tile scanning of CeA at approximately bregma -1.46 mm, 5-m z-stacks). All microscope settings were kept the same within experiments during image acquisition. To perform quantification, ImageJ (Schindelin et al., 2012) was used to outline individual nuclei as identified by DAPI staining and count all nuclei in an unbiased manner (all settings kept the same within experiments). The fluorescence intensity for each probe per DAPI counted nuclei was then measured and the background signal, as determined by the average intensity of the negative control, was subtracted for each channel. For each probe, signal intensity present after background subtraction identified positive nuclei. Experiment was performed in an unbiased manner as probe fluorescence was quantified blindly, independently, and after nuclei identification by computational means.
Next, the percentage of nuclei positive for one or both probes and the percentage of signal co-expression were calculated. Percentage of total nuclei was determined by dividing the total number of nuclei expressing that marker by the total number of DAPI positive nuclei per image. The percentage of Crhr1ϩ nuclei expressing a marker of interest was determined by dividing the number of colabeled nuclei by the total number of Crhr1ϩ nuclei. The percentage of Crhr1-nuclei co-expressing a marker of interest was determined by dividing the number of cells expressing the marker of interest but not Crhr1 by the total number of Crhr1-nuclei per image. Crhr1ϩ compared to Crhr1-patterns of co-expression was assessed and normalized to the Crhr1-cells.
For densitometry, signal intensity measured after background subtraction was quantified, log2 transformed, and normalized to the CeL for visualization of the difference in expression. Unpaired Student's t test was used to assess significance of the CeL versus CeM expression difference for each gene, as well as the Crhr1ϩ to Crhr1-percentage co-expression of each gene. Analysis and statistics were performed using R programming (R Core Team, 2018). All analyses were performed on raw images. Outliers were detected by Grubb's test. Brightness/contrast and pixel dilation are the same for all representative images shown per figures.

Dendritic spine analysis Tissue preparation
Mice (n ϭ 4) were anesthetized (isoflurane) and perfused with cold PBS/4% PFA. Brains were extracted and immersion fixed in 4% PFA at 4°C for 2 h before being sectioned coronally into 100-m slices on a vibrating microtome (Leica VT1000S, Leica Microsystems).

Imaging and analysis
Slides were imaged for CeA on a Zeiss LSM 710 laser scanning confocal microscope (Carl Zeiss MicroImaging; 63ϫ oil immersion, 1024 ϫ 1024, 1-m step z-stacks). All microscope settings were kept the same within experiments during image acquisition. ImageJ (Schindelin et al., 2012) was used to perform a quantification of filopodia, thin, stubby and mushroom-shaped dendritic spines in both CRF 1 ϩ (n ϭ 13 neurons, 37 dendritic segments) and CRF 1 -neurons (n ϭ 24 neurons, 41 dendritic segments). The experimenter was blind to cell type (CRF 1 ϩ vs CRF 1 -) when performing spine quantification. Data obtained in each mouse were averaged for each cell type, such that the number of mice was used as the n in statistical analyses. Spine densities and spine type proportions in CRF 1 -and CRF 1 ϩ neurons were compared using paired t tests. Data are presented as mean Ϯ standard error. In all cases, p Ͻ 0.05 was the criterion for statistical significance.

Electrophysiological recordings
We visualized neurons using infrared differential interference contrast (IR-DIC) optics and CCD camera (Infinity 3 s, Lumenera). A 60ϫ magnification water immersion objective (Olympus) was used for identifying and approaching neurons. To avoid photolytic damage, initial exposure to episcopic fluorescence illumination was brief (Ͻ2 s). We detected fluorescent neurons using the Lumen 300 LED system (Prior Scientific) and captured images using Luminera software (Lumenera Corp.). Whole-cell current-clamp and voltage-clamp recordings were obtained with patch pipettes (3-4 M⍀; Warner Instruments) using a Multiclamp 700B amplifier (Molecular Devices), sampled at 20 kHz, low pass filtered at 10 kHz, digitized (Digidata 1440A; Molecular Devices), and stored on a computer using pClamp 10 software (Molecular Devices). Series resistance was not compensated but monitored throughout the experiment, and cells with a series resistance Ͼ15 M⍀ or a Ͼ20% change were excluded.
The intracellular solution used for voltage and current clamp recordings was composed of 145 mM potassium gluconate, 0.5 mM EGTA, 2 mM MgCl 2 , 10 mM HEPES, 2 mM Na-ATP, and 0.2 mM Na-GTP. Drugs were dissolved in aCSF and applied by bath perfusion. To isolate sEPSC and mEPSC, recordings were performed in the presence of GABA A (30 M bicuculline) and GABA B (1 M CGP 55845A) receptor antagonists, and tetrodotoxin (TTX; 0.5 M) was included for mEPSCs. CeA neurons were held at -60 mV for voltage-clamp recordings and maintained at -70 mV for current-clamp recordings.

Statistical analysis
Frequency, amplitude, rise, and decay time of EPSCs were analyzed and visually confirmed using a semiautomated threshold based mini detection software (Mini Analysis, Synaptosoft Inc.). Average EPSC characteristics were determined during baseline and experimental drug conditions in 3 min bins containing a minimum of 60 events. All detected events were used for event frequency analysis, but superimposed events were eliminated for amplitude and kinetic analysis. Data are presented as mean Ϯ standard error, and statistical significance was assessed using a two-tailed t test or one-sample t test using Prism 6 (GraphPad Prism). In all cases, p Ͻ 0.05 was the criterion for statistical significance, and n represents the number of cells or mice as indicated.

Expression of CBPs in CeA CRF 1 ؉ neurons
The CBPs CB, CR and PV label distinct subsets of neurons in the brain and have been widely used as markers of neuronal populations in the amygdala (Andressen et al., 1993;Kemppainen and Pitkänen, 2000). We therefore examined the expression profile of CBPs in CeA GFPϩ/CRF 1 ϩ neurons (Fig. 1). Using immunohistochemistry in CRF 1 :GFP mouse brain sections, the coexpression of PV (PV; Fig. 1A), CR (Fig. 1B), and CB (Fig.  1C,D) with GFP was analyzed in the CeA. Of the CBPs labeled, CB was the most widely expressed, while CR was expressed in few cells and PV was virtually absent from the CeA (Fig. 1E). The majority of CeA GFPϩ cells were located in the CeM, as described previously (Justice et al., 2008;Herman et al., 2013). The percentage of CRF 1 ϩ cells co-expressing each CBP followed the same pattern as the overall counts of CBPϩ cells, wherein CB was present in 19%, CR was present in 2%, and PV was present in 0.1% of CRF 1 ϩ cells (Fig. 1F). Based on GFP labeling in CBPϩ cells, CRF 1 was present in 58% of CBϩ cells and 63% of CRϩ cells, and no PVϩ cells (Fig. 1G).

Expression of neuropeptides and subpopulation markers in CeA Crhr1؉ neurons
We next examined the expression of subpopulation markers, including neuropeptides, in Crhr1ϩ neurons in both the CeL and CeM ( Fig. 2A). First, the validity of the CRF1:GFP transgenic mice was assessed through the co-expression of GFP mRNA and Crhr1 mRNA in the CeL and CeM where 58% and 86% co-expression in Crhr1ϩ nuclei and 72% and 71% co-expression in GFP mRNAϩ nuclei was observed, respectively (Fig. 2B). Additionally, the Crhr1 probe was validated in the septum, a region of low CRF1 expression (Van Pett et al., 2000; Fig. 2C). Glutamic acid decarboxylase 2 (Gad2) and vesicular glutamate transporter 1 (Slc17a7) were used as GABAergic and glutamatergic neuronal markers, respectively. In situ hybridization in both the CeL (Fig. 2D-F) and CeM (Fig.  2G-I) identified a high proportion of nuclei expressing Gad2, 89% and 98%, respectively, and a low proportion expressing Slc17a7, 3% and 4%, respectively. This is consistent with the majority of neurons being GABAergic in this region (Pitkänen and Amaral, 1994;Ehrlich et al., 2009). Almost all Crhr1ϩ cells expressed Gad2 at 98% in the CeL and 100% in the CeM (Fig. 2E,H), whereas very few expressed Slc17a7 at 6% in the CeL and 5% in the CeM (Fig. 2F,I). Additionally, a negative control was run in each series for background subtraction (Fig. 2J). Crhr1 expression was observed in ϳ 23% and 34% of all nuclei on average in CeL and CeM, respectively. These data indicate that CeA Crhr1ϩ neurons, as well as the vast majority of CeA neurons, are GABAergic and are located in both the CeL and CeM.
Co-expression of Crhr1 with the neuropeptides Penk, Sst, Pdyn, Crh, and Npy, as well as with Prkcd, was also examined (CeL, Fig. 3A-F; CeM, Fig. 4A-F). In the CeL, transcripts for all these genes, except Npy, were found in high abundance, where Penk was expressed in 86% of cells, Sst in 59%, Pdyn in 58%, Prkcd in 71%, Crh in 36%, and Npy in 5% of cells (Fig. 3G). The proportion of total nuclei expressing each of these transcripts was similar to the proportion observed in Crhr1ϩ nuclei only, where Penk was expressed in 92%, Sst in 75%, Pdyn in 70%, Prkcd in 53%, Crh in 42%, and Npy in 10% of Crhr1ϩ cells (Fig. 3H). Accordingly, when comparing the expression of the above peptides in CeL Crhr1ϩ versus Crhr1neurons no significant differences were observed between the two populations (Fig. 3I).

Glutamatergic transmission in CRF 1 ؉ and CRF 1neurons in the CeM
We next examined glutamatergic transmission in CRF 1 ϩ and CRF 1 -CeM neurons. Before recording gluta-    activity, whole-cell current clamp recordings with a step protocol consisting of hyperpolarizing to depolarizing current injections were obtained from each cell to determine cell type based on spiking characteristics. As previously described, CeA neurons are composed of three principal cell types: regular spiking, low threshold bursting, and late spiking neurons (Dumont et al., 2002;Chieng et al., 2006;Herman et al., 2013; Fig. 6A). CRF 1 ϩ neurons consisted of only regular spiking and low threshold bursting cell-types as previously reported (Herman et al., 2013(Herman et al., , 2016, while the majority of CRF 1 -neurons were regular spiking and late spiking neurons (Fig. 6B). No significant differences were observed in the membrane properties of CRF 1 ϩ and CRF 1 -CeA neurons (Table 1).

Discussion
Here, we investigated the neurochemistry, morphology, and physiology of CRF 1 ϩ neurons in the CeA of adult male mice. We found that CB is the predominant CBP expressed in the CeA but overall there is no specific enrichment or exclusion of CBPs in Crhr1ϩ neurons. Co-expression analysis using in situ hybridization revealed Crhr1 is coexpressed mostly with Penk and Sst and least with Npy. In the CeM, Penk is significantly enriched in CRF 1 ϩ neurons compared to CRF 1 -neurons. Morphologically, CRF 1 ϩ neurons have a lower spine density but a higher proportion of mushroom spines compared to CRF 1 -neurons. Accordingly, basal excitatory transmission between CRF 1 ϩ and CRF 1 -neurons are similar. Additionally, acute CRF application increases glutamatergic transmission in both CRF 1 ϩ and CRF 1 -neurons. The lack of enrichment or exclusion of any particular CBP or neuropeptide in this population contributes to the difficulty in identifying and studying this subpopulation in this heterogenous region.

CBP and neuropeptide expression in the CRF 1 ؉ subpopulation
GABAergic interneurons may be subcategorized based on their expression of the CBPs PV, CB, and CR, and these distinct subpopulations exhibit unique differences in their physiology, synaptology, and morphology (Kemppainen and Pitkänen, 2000). We found that CB was the predominant CBP in both CeL and CeM nuclei, with only a limited number of neurons expressing PV and CR. This pattern of relative CBP expression is in general agreement with immunohistochemical studies conducted in the rat (Kemppainen and Pitkänen, 2000), and with our previous finding that at least 30% of CeA CRF 1 -expressing neurons project out to the dorsolateral BNST (Herman et al., continued (approximately bregma -1.46 mm) is shown on Nissl stained image (Franklin and Paxinos, 2008) to indicate the region of interest. B, Representative images indicate the co-expression of GFP and Crhr1 in the GFP:CRF 1 mice in CeL (58%; n ϭ 8 images) and CeM (86%; n ϭ 7 images; Crhr1 ϭ red, GFP ϭ green, DAPI ϭ blue). Scale bar ϭ 10 m. C, Crhr1 probe specificity was validated in the septum of wild-type mice, a region of low Crhr1/GFP expression (n ϭ 2 images). Negligible expression was observed in the septum as compared to the infralimbic prefrontal cortex in the same slice (Crhr1 ϭ red, DAPI ϭ blue). Scale bar ϭ 10 m. D, Representative images for Gad2 (upper; green ϭ Gad2, red ϭ Crhr1, blue ϭ DAPI; n ϭ 5 images) and Slc17a7 (lower; green ϭ Slc17a7, red ϭ Crhr1, blue ϭ DAPI; n ϭ 7 images) in the CeL. Scale bar ϭ 10 m. E, Heat map represents the percentage of nuclei expressing (from left to right) RNA of interest, Crhr1, and co-expression in the total nuclei counted in the CeL. F, Heat map represents the percentage of nuclei co-expressing in the Crhr1ϩ population of nuclei in the CeL. Color scale from 100% (yellow) to 0% (purple). G, Representative images for Gad2 (upper; green ϭ Gad2, red ϭ Crhr1, blue ϭ DAPI; n ϭ 5 images) and Slc17a7 (lower; green ϭ Slc17a7, red ϭ Crhr1, blue ϭ DAPI; n ϭ 7 images) in the CeM. Scale bar ϭ 10 m. H, Heat map represents the percentage of nuclei expressing (from left to right) RNA of interest, Crhr1, and co-expression in the total nuclei counted in the CeM. I, Heat map represents the percentage of nuclei co-expressing in the Crhr1ϩ population of nuclei in the CeM. Color scale from 100% (yellow) to 0% (purple). Data are shown as mean Ϯ SEM. J, Representative negative control images for the bacterial probe DapB in the CeM indicate negligible fluorescence intensity for both channels shown (DapB ϭ red, DapB ϭ green, DAPI ϭ blue). Scale bar ϭ 10 m.  Representative images from top to bottom for (A) Penk (green ϭ Penk, red ϭ Crhr1, blue ϭ DAPI; n ϭ 8 images), (B) Sst (green ϭ Sst, red ϭ Crhr1, blue ϭ DAPI; n ϭ 9 images), (C) Pdyn (green ϭ Pdyn, red ϭ Crhr1, blue ϭ DAPI; n ϭ 10 images), (D) Prkcd (green ϭ Prkcd, red ϭ Crhr1, blue ϭ DAPI; n ϭ 9 images), (E) Crh (green ϭ Crh, red ϭ Crhr1, blue ϭ DAPI; n ϭ 10 images), and (F) Npy (green ϭ Npy, red ϭ Crhr1, blue ϭ DAPI; n ϭ 9 images) in the CeL. Scale bar ϭ 10 m. G, Heat map represents 2013, 2016), one of several downstream brain regions innervated by the CeA.
In addition, CeA neurons can express several neuropeptides that play important roles in fear and anxiety behaviors such as CRF, enkephalins, dynorphins, somatostatin and NPY. The overlapping expression pattern of these neuropeptides differentiates distinct neuronal subpopulations in several brain regions in mice and rats (Petrovich et al., 1996;Herry et al., 2008;Haubensak et al., 2010;Li et al., 2013;Cai et al., 2014;Penzo et al., 2014;Yu et al., 2016;Kim et al., 2017). We hypothesized that CRF 1 expression is restricted to a subset of CeA neurons that co-express a unique combination of neuropeptides. Most of the neuropeptides we investigated had similar trends in the total population compared to the Crhr1ϩ population. However, Penk was present in a higher proportion of Crhr1ϩ cells than Crhr1-cells in the CeM. Npy also trended toward an enrichment in Crhr1ϩ cells over Crhr1-cells in the CeL but co-expression results obtained for this neuropeptide showed more variability than for other markers, most likely due to the small number of Npyϩ nuclei. Additionally, densitometry identified a higher expression of Crh, Pdyn, Prkcd, and Sst in the CeL compared to the CeM, which is in agreement with previous studies in mice and rats (Veinante et al., 1997;Day et al., 1999;Marchant et al., 2007;Li et al., 2013;Kim et al., 2017;McCullough et al., 2018). Conversely, Npy was present at higher levels in the CeM than in the CeL.
The enrichment of Penk in CeM Crhr1ϩ neurons compared to their Crhr1-neighbors may have functional relevance for the recruitment of CeA Penkϩ neurons in response to food, drugs and stress. CeA Penkϩ are activated by subchronic exposure to fat, ethanol, and nicotine, as well as by withdrawal from morphine in animal models (Criado and Morales, 2000;Veinante et al., 2003;Loughlin et al., 2006;Chang et al., 2014). This activation of CeA Penkϩ neurons may result from their preferential expression of CRF 1 as some of these stimuli are known to increase CRF 1 signaling in the CeA (Heinrichs et al., 1995;Nie et al., 2004). Furthermore, chemogenetic activation of CeA Penk neurons produces sustained analgesia, suggesting that CRF 1 -mediated activation of these neurons could mediate stress-induced analgesia (Wiedenmayer et al., 2002;Paretkar and Dimitrov, 2019).
It is important to consider here that our method for identification of expressing cells is semi-quantitative and that only signal co-localized with DAPI staining after stringent background subtraction was quantified to increase confidence in detection. However, by excluding cell bodies and processes, we lost data concerning dynamic somatic and neurite expression. This method is also limited in its ability to distinguish only neuronal cells which may introduce a consistent level of non-neuronal cells in our analysis. Additionally, co-expression may vary from the rostral to caudal CeA, and a more in-depth study of CeA neuropeptide localization is necessary to identify the role of these neuropeptides in the CRF 1 ϩ subpopulation throughout the CeA. Further studies to identify dynamic expression of RNA or protein at subcellular levels may yield additional information concerning the functional roles of these peptides and proteins.
Our in situ hybridization analysis of Crhr1 expression may yield differing results than GFP expression in the CRF 1 :GFP mice due to differences in RNA versus protein expression, sensitivity of the methods used, image acquisition, and identification of GFP saturated soma versus puncta quantification in DAPI-stained nuclei. However, the Crhr1 probe was verified to yield negligible signal in the septum, a region with no CRF 1 /GFP expression (Van Pett et al., 2000). Furthermore, our observation that Crhr1 mRNA density is markedly higher in the CeM than in the CeL is consistent with the patterns of GFP native fluorescence and immunohistochemical staining reported here (Fig. 2B) and previously (Justice et al., 2008;Herman et al., 2013), with GFP expression being substantially more prominent in the CeM than in the CeL. Additionally, probe sensitivity and mRNA abundance might affect the analysis of Crhr1 and GFP mRNA co-expression. A strong overlap was found between both populations in the CeM, as well as a strong co-expression in GFP mRNAϩ nuclei in the CeL. However, co-expression was less pronounced in Crhr1ϩ nuclei in the CeL, reflective of less GFP protein expression (Fig. 2B). The higher sensitivity of the RNAscope assay compared to native fluorescence and immunohistochemistry probably explains why ϳ23% CeL cells are detected as Crhr1ϩ, while a very limited number of CeL cells are identified as GFPϩ.

CRF 1 ؉ neurons differ from CRF 1 -neurons in their dendritic morphology
We characterized the dendritic morphology of CeA CRF 1 ϩ and CRF 1 -neurons including spine density and spine shape. The morphology and density of spines plays a crucial role in synaptic and neuronal function. More mature and enlarged spines (mushroom/stubby) are associated with increased synaptic strength, whereas immature spines (thin/filopodia) may sustain limited synaptic signaling (De Roo et al., 2008;Sala and Segal, 2014). Abnormalities in spine morphology are associated with a variety of neurologic and psychiatric disorders, including addiction (Mulholland and Chandler, 2007;Spiga et al., 2014;Phillips and Pozzo-Miller, 2015;Herms and Dorostkar, 2016;Qiao et al., 2016;Varodayan et al., 2018). Previous morphologic characterization of CeA neurons has identified different cell types (McDonald, 1982;Chieng et al., 2006). Most neurons in the CeM have long continued the percentage of nuclei expressing (from left to right) RNA of interest, Crhr1, and co-expression in the total nuclei counted in the CeM. H, Heat map represents the percentage of nuclei co-expressing in the Crhr1ϩ population of nuclei in the CeM. Color scale from 100% (yellow) to 0% (purple). I, Quantification of the difference between the percentage nuclei expressing RNA of interest (Crh, Prkcd, Pdyn, Npy, Penk, and Sst) in Crhr1ϩ nuclei versus Crhr1-nuclei in the CeL (mean Ϯ SEM).  Penk (green ϭ Penk, red ϭ Crhr1, blue ϭ DAPI; n ϭ 11 images), (B) Sst (green ϭ Sst, red ϭ Crhr1, blue ϭ DAPI; n ϭ 10 images), (C) Crh (green ϭ Crh, red ϭ Crhr1, blue ϭ DAPI; n ϭ 11 images), (D) Pdyn (green ϭ Pdyn, red ϭ Crhr1, blue ϭ DAPI; n ϭ 8 images), (E) Prkcd (green ϭ Prkcd, red ϭ Crhr1, blue dendrites that branch sparingly and have a moderate number of dendritic spines, while a smaller number of neurons have thick dendrites with virtually no spines (Mc-Donald, 1982). Here, we observed that the spine density in CRF 1 ϩ neurons was lower than in CRF 1 -neurons. We further analyzed the spine type proportions in CRF 1 ϩ and CRF 1 -neurons and found a higher proportion of mushroom spines and lower proportion of thin spines in CRF 1 ϩ neurons. The lower overall spine density is therefore at least partially offset by a higher fraction of mushroom (i.e., most mature) spines, resulting in a comparable density of functional spines in CRF 1 ϩ and CRF 1 -neurons.

CRF modulation of glutamatergic transmission in CRF 1 ؉ versus CRF 1 -neurons
CRF modulates glutamate transmission and has been shown to increase vesicular glutamate release at rat CeA synapses, which can be enhanced by CRF 1 activity (Varodayan et al., 2017). Therefore, we examined glutamatergic transmission in CRF 1 ϩ and CRF 1 -cells as well as the effect of CRF on glutamatergic transmission in these neu-ronal populations. Both populations receive similar glutamatergic input, consistent with the similar density of functional spines in CRF 1 ϩ and CRF 1 -neurons. Additionally, CRF increases mEPSC frequency in both populations with no effect on mEPSC amplitude or kinetics in either population. These results are indicative of CRF-induced increased presynaptic GABA release with no postsynaptic effect on glutamatergic receptors.

Conclusion
Our results reveal molecular, morphologic and functional characteristics of CRF 1 ϩ neurons highlighting the importance of identifying specific cell populations in the CeA. The CeA is a hub that integrates disparate inputs (stress) and drives appropriate behavioral outputs. Previous work has demonstrated a major role of the CRF system, and particularly the CRF 1 receptors, within the amygdala complex to influence cellular functions that produce maladaptive behavior in animal models. The amygdalar CRF 1 system represents a common pathway for the convergence of stress, addiction, pain, depres-continued ϭ DAPI; n ϭ 6 images), and (F) Npy (green ϭ Npy, red ϭ Crhr1, blue ϭ DAPI; n ϭ 8 images) in the CeM. G, Heat map represents the percentage of nuclei expressing (from left to right) RNA of interest, Crhr1, and co-expression in the total nuclei counted in the CeM. H, Heat map represents the percentage of nuclei co-expressing in the Crhr1ϩ population of nuclei in the CeM. Color scale from 60% (yellow) to 0% (purple). I, Quantification of the difference between the percentage nuclei expressing RNA of interest [Crh, Prkcd, Pdyn, Npy, Penk (p ϭ 0.004), and Sst] in Crhr1ϩ nuclei versus Crhr1-nuclei in the CeM (mean Ϯ SEM; ‫‪p‬ءء‬ Ͻ 0.01 unpaired t test). J, Bar graph showing the difference in log2 of the signal intensity between CeL and CeM of Crhr1 (n ϭ 3079 nuclei), Npy (n ϭ 90 nuclei), Penk (n ϭ 468 nuclei), Crh (n ϭ 273 nuclei), Prkcd (n ϭ 276 nuclei), Pdyn (n ϭ 141 nuclei), and Sst (n ϭ 145 nuclei) in Crhr1ϩ nuclei (mean Ϯ SEM; ‫‪p‬ءءء‬ Ͻ 0.001 unpaired t test from CeL).  . Spine density and morphology of CRF 1 ϩ and CRF 1 -neurons in the CeA. Representative images of dendritic segments in CRF 1 ϩ (A) and CRF 1 -(B) neurons from the CeM of CRF 1 :GFP male mice. Scale bar ϭ 10 m. C, Summary bar graph indicates spine density in CRF 1 ϩ and CRF 1 -neurons averaged over 10 m dendritic segments (‫ء‬p Ͻ 0.05 paired t test). D, Summary graph indicates proportion of mushroom spines, stubby spines, thin spines and filopodia in CRF 1 ϩ and CRF 1 -neurons in the CeA (‫ء‬p Ͻ 0.05, paired t test). Data are shown as mean Ϯ SEM (n ϭ 4 mice). sion, memory formation and anxiety-related signaling (Koob, 2003(Koob, , 2008Zorrilla and Koob, 2004). In particular, CRF 1 mechanisms and circuits have been implicated in the development of the negative emotional state associated with alcohol dependence, and it has been proposed that alleviation of this negative state drives the motivation to drink in mice and rats (Funk et al., 2006;Koob, 2010;Roberto et al., 2010;Lowery-Gionta et al., 2012). Our cellular studies revealed that acute and chronic alcohol induced significant alterations in GABAergic signaling in the CeA circuitry. In addition, this GABAergic signaling is cell type specific for the CRF 1 ϩ subpopulation and their connectivity and may contribute to the development of alcohol dependence (Herman et al., 2016).
Our new evidence on the CRF 1 ϩ cellular phenotype details the distinct neuronal CeA subpopulations in normal amygdalar function and highlights the need for their further characterization (including intracellular signaling etc) under pathologic conditions, such as alcohol dependence and anxiety disorders. The multiplicity of neuronal subpopulations and the complexity of local microcircuits provide numerous targets for potential dysregulation by Figure 6. Membrane properties and glutamatergic transmission of CRF 1 ϩ and CRF 1 -neurons in the CeA. A, Representative current-clamp recordings from three cell-types: regular spiking (left), low threshold bursting (middle) and late spiking (right) from a CeA neuron. B, Distribution of cell-types recorded from for CRF 1 -(n ϭ 15) and CRF 1 ϩ (n ϭ 16) CeA neurons. C, Representative traces of mEPSCs in CRF 1 ϩ (top) and CRF 1 -(bottom) CeA neurons. D, E, Frequency and amplitude of mEPSCs in CRF 1 ϩ (n ϭ 16) and CRF 1 -(n ϭ 12) CeA neurons. F, Representative traces of sEPSCs in CRF 1 ϩ (top) and CRF 1 -(bottom) CeA neurons. G, H, Frequency and amplitude of sEPSCs in CRF 1 ϩ (n ϭ 14) and CRF 1 -(n ϭ 14) CeA neurons. I, Representative traces of mEPCS during baseline (top) and 200 nM CRF application (bottom) from CRF 1 ϩ CeA neurons. J, K, CRF effects on frequency and amplitude of mEPSCs in CRF 1 ϩ (n ϭ 11) and CRF 1 -(n ϭ 9) CeA neurons, respectively. L, Average trace of mEPSC during baseline (black) and 200 nM CRF (green) application from a CRF 1 ϩ CeA neuron. M, N, CRF effects on rise and decay time of mEPSCs in CRF 1 ϩ and CRF 1 -CeA neurons, respectively (‫ء‬p Ͻ 0.05 by one-sample t test). Data are shown as mean Ϯ SEM.  The membrane capacitance, resistance, time constant, and resting potential are reported for CRF1ϩ and CRF1-neurons recorded from the CeA. Data are shown as mean Ϯ SEM. drugs of abuse (e.g., alcohol) and stress, and their potential clinical relevance. The frequency, amplitude, rise time, and decay time of mEPSCs and sEP-SCs are reported for CRF1ϩ and CRF1-neurons recorded from the CeA. Data are shown as mean Ϯ SEM.