GLP‐1 neurons form a local synaptic circuit within the rodent nucleus of the solitary tract

Abstract Glutamatergic neurons that express pre‐proglucagon (PPG) and are immunopositive (+) for glucagon‐like peptide‐1 (i.e., GLP‐1+ neurons) are located within the caudal nucleus of the solitary tract (cNTS) and medullary reticular formation in rats and mice. GLP‐1 neurons give rise to an extensive central network in which GLP‐1 receptor (GLP‐1R) signaling suppresses food intake, attenuates rewarding, increases avoidance, and stimulates stress responses, partly via GLP‐1R signaling within the cNTS. In mice, noradrenergic (A2) cNTS neurons express GLP‐1R, whereas PPG neurons do not. In this study, confocal microscopy in rats confirmed that prolactin‐releasing peptide (PrRP)+ A2 neurons are closely apposed by GLP‐1+ axonal varicosities. Surprisingly, GLP‐1+ appositions were also observed on dendrites of PPG/GLP‐1+ neurons in both species, and electron microscopy in rats revealed that GLP‐1+ boutons form asymmetric synaptic contacts with GLP‐1+ dendrites. However, RNAscope confirmed that rat GLP‐1 neurons do not express GLP‐1R mRNA. Similarly, Ca2+ imaging of somatic and dendritic responses in mouse ex vivo slices confirmed that PPG neurons do not respond directly to GLP‐1, and a mouse crossbreeding strategy revealed that <1% of PPG neurons co‐express GLP‐1R. Collectively, these data suggest that GLP‐1R signaling pathways modulate the activity of PrRP+ A2 neurons, and also reveal a local “feed‐forward” synaptic network among GLP‐1 neurons that apparently does not use GLP‐1R signaling. This local GLP‐1 network may instead use glutamatergic signaling to facilitate dynamic and potentially selective recruitment of GLP‐1 neural populations that shape behavioral and physiological responses to internal and external challenges.

GLP-1R also is expressed within the cNTS in rats and mice Merchenthaler et al., 1999;Reiner et al., 2016), where GLP-1 may act locally to mediate or support at least some of the behavioral and physiological effects attributed to central GLP-1 signaling pathways (Hayes, Bradley, & Grill, 2009;Hayes, Skibicka, & Grill, 2008;Richard et al., 2015;Skibicka, 2013). Indeed, intra-NTS administration of GLP-1R agonists in rats reduces food intake and operant responding for food reward, and also reduces place preference conditioned by palatable food access (Hayes et al., 2011;Kanoski, Rupprecht, Fortin, De Jonghe, & Hayes, 2012); conversely, food intake is increased in rats after local NTS blockade of GLP-1Rs (Hayes et al., 2009). The local circuitry mediating these pharmacological effects is unclear. In mice genetically engineered to express yellow fluorescent protein (YFP) under control of the glucagon promoter, light microscopy revealed YFP+ fibers and varicosities (presumably arising from GLP-1 neurons) in close apposition with catecholaminergic NTS neurons, including those that comprise the A2 noradrenergic cell group (Llewellyn-Smith et al., 2013;Richard et al., 2015). Further, at least a subset of catecholaminergic NTS neurons express mRNA for GLP-1R in mice . While similar localization studies have not been conducted in rats, central (i.c.v.) administration of a GLP-1R antagonist reduced the ability of restraint stress to activate cFos expression by A2 neurons that co-express prolactin-releasing peptide (PrRP) (Maniscalco, Zheng, Gordon, & Rinaman, 2015). Although central GLP-1R antagonism likely alters neural responses to stress across many brain regions, one potential explanation is that PrRP+ A2 neurons express GLP-1R and receive axonal input from GLP-1+ neurons in rats, which would be consistent with reports that a subset of catecholaminergic NTS neurons receive close appositions from the axons of PPG neurons and express GLP-1R in mice Llewellyn-Smith et al., 2013;Richard et al., 2015).
Unlike catecholaminergic cNTS neurons, PPG neurons in mice do not express mRNA for GLP-1R (as determined by single-cell reverse transcription polymerase chain reaction (RT-PCR)), and whole-cell patch recording data indicate that PPG neurons do not respond to  agonists (Hisadome, Reimann, Gribble, & Trapp, 2010). However, as GLP-1 neurons are glutamatergic (Zheng, Stornetta, et al., 2014), a lack of GLP-1R expression does not preclude the possibility that GLP-1 neurons are coupled via glutamatergic synapses. There is no published evidence regarding the presence or absence of GLP-1R mRNA expression by identified GLP-1 or noradrenergic/PrRP neurons in rats, and the ultrastructural features of GLP-1 neurons and their synaptic associations within the cNTS has not been reported in either rats or mice.
This study used light and confocal microscopy to examine potential anatomical interactions (1) between GLP-1+ and PrRP+ A2 neurons within the cNTS of adult male rats, and (2) between GLP-1+ profiles in rats and between PPG/YFP+ profiles in mice. Using electron microscopy, we further examined the ultrastructural features of GLP-1+ profiles within the rat cNTS. After synaptic contacts between GLP-1+ profiles were confirmed, we used RNAscope to examine whether cNTS noradrenergic/PrRP+, GABAergic, and/or GLP-1+ neurons express mRNA for GLP-1R in rats. Finally, we revisited the possibility that PPG neurons in mice express functional GLP-1R. For this purpose, Ca 2+ imaging was performed in ex vivo slices from Glu-Cre/ Rosa26-GCaMP3 mice to assess both somatic and dendritic responses of PPG neurons to GLP-1, and a new transgenic mouse crossbreeding strategy was used to investigate potential colocalization of fluorescent reporter genes identifying PPG-and GLP-1R-expressing neurons.

| Rats
Adult male Sprague-Dawley rats weighing 250-350 g were used for the present report. Rats were pair-housed in a temperature-controlled (22-25 C) colony room with a 12:12 hr light-dark cycle (lights on at 0700). Food and water were available ad libitum. All experimental procedures using rats were conducted at the University of Pittsburgh or at Florida State University, conformed to NIH regulations detailed in the Guide for the Care and Use of Laboratory Animals, and were approved by the relevant Institutional Animal Care and Use Committees.

| Immunocytochemical reagents
Glucagon-like peptide-1 was localized in rat tissue sections using a rabbit polyclonal antiserum and a mouse monoclonal antibody. The rabbit polyclonal (T-4363, Peninsula Laboratories, San Carlos, CA; RRID: AB_518978) was generated against a synthetic peptide corresponding to amino acids 7-37 of the human GLP-1 sequence. Specificity of this antiserum as applied to rat and human brain tissue has been documented , including an absence of immunolabeling after pre-absorbing the primary antiserum overnight at 4 C with a 10-fold higher concentration of synthetic GLP-1 (7-37) acetate salt (Bachem, Torrance, CA, USA, H-9560) before tissue incubation. The mouse monoclonal antibody was generated against Nterminal residues 7-19 of the human GLP1 peptide, and was provided by Dr. Guibao Gu (RRID: AB_2314562). Specificity of this antibody for GLP-1 has been documented (Gu et al., 2013), including (in our laboratory) an absence of immunolabeling after pre-absorbing the primary antiserum overnight at 4 C with a 10-fold higher concentration of synthetic GLP-1 (7-37) acetate salt (Bachem, H-9560) before tissue incubation. Its selectivity also was tested against other cleavage products of the PPG gene, and was reported to weakly immunolabel GLP-2, which is expressed along with GLP-1 in the same neurons (Gu et al., 2013). PrRP was localized using a rabbit polyclonal antiserum (H-008-52, Phoenix Pharmaceuticals, Inc., Burlingame, CA, USA) raised against the synthetic 31 amino acid sequence of rat PrRP peptide.
Antibody specificity has been documented by the manufacturer and also within our laboratory, based on positive and specific immunolabeling of a subset of dopamine beta hydroxylase-positive noradrenergic neurons in the rat caudal brainstem (Maniscalco et al., 2015;Maniscalco, Kreisler, & Rinaman, 2013), and fully consistent with the reported distribution of PrRP mRNA expression and immunolabeling in mice and rats (Ellacott, Lawrence, Rothwell, & Luckman, 2002;Lawrence, Celsi, Brennand, & Luckman, 2000). The chicken anti-GFP antibody (Catalogue #AB13970, lot #623923; Abcam, Cambridge, MA; RRID: AB_300798) used to localize YFP in mice was raised against a recombinant full length protein corresponding to green fluorescent protein (GFP). This antibody yields a single band on western blot (manufacturer's technical information), and specificity was demonstrated in this study by a lack of immunolabeling in tissue sections from wildtype rats and mice. Affinity purified secondary antibodies conjugated to fluorescent markers, biotin, and horseradish peroxidase (HRP) were purchased from Jackson ImmunoResearch Laboratories  (McLean & Nakane, 1974). Fixed brains were removed from the skull, postfixed for 24 hr, and then transferred to 20% sucrose solution for at least 24 hr. Brains were blocked and then sectioned serially at 35 μm using a freezing stage sliding microtome (Leica SM2000 R; Leica Biosystems, Wetzlar, Germany). Sections were collected sequentially in six vials of cryopreservant solution (Watson, Wiegand, Clough, & Hoffman, 1986) such that each vial contained sections at a frequency of 210 μm throughout the caudal brainstem. Vials of tissue were stored at −20 C prior to immunocytochemical processing.
For transmission electron microscopic (TEM) analyses, one rat was perfused with 150 mL of 1.5% acrolein (Polysciences, Inc., Warrington, PA) and 2.0% PF in PBS followed by 100 mL of 2.0% PF in the same buffer. A second rat was perfused with 300 mL of a mixture of 2% PF and 0.5% TEM grade glutaraldehyde (Sigma-Aldrich) in 0.2 M sodium cacodylate buffer. Each brain was removed from the skull, blocked to separate the forebrain from the brainstem, and postfixed in 2.0% buffered PF for 6 hr at 4 C.
Rats used for RNAscope in situ hybridization (n = 4) were perfused with 100 mL physiological saline solution followed by 300 mL ice-cold 4% PF in 0.1 M sodium phosphate buffer (PB). Brains were removed from the skull, postfixed in the same fixative overnight at 4 C, incubated in 20% sucrose solution for 16-18 hr, and cut into six series of 25-μm thick sections using a freezing stage sliding microtome. Tissue sections were collected in cryopreservant solution and stored at −20 C until use.

| Light microscopic immunocytochemical localizations
2.4.1 | Immunoperoxidase labeling of rat tissue Cryopreserved sections were brought to room temperature and washed in multiple changes of 10 mM sodium PB on a rocker table.
One vial of tissue sections from each brain was processed for immunoperoxidase localization of GLP-1 using the rabbit polyclonal GLP-1 antiserum. Sections were pretreated by incubation in 1.0% sodium borohydride followed by 0.15% hydrogen peroxide to increase antibody penetration and reduce nonspecific background. After PB rinsing, sections were then incubated in a 1:10 K dilution of the primary antibody in PB containing 0.3% Triton-X and 1% normal donkey serum for 24-48 hr at 4 C. Following multiple washes in PB the tissue was transferred to a 1:200 dilution of biotin-conjugated affinity purified donkey anti-rabbit secondary antibody and incubated for 2 hr at room temperature with agitation. Thereafter the sections were subjected to multiple PB washes prior to immunoperoxidase localization of GLP-1 using Vectastain reagents (Vector Laboratories, Burlingame, CA) and diaminobenzidine (DAB) catalyzed with hydrogen peroxide.

| Immunofluorescent labeling of rat tissue
Dual immunofluorescence localization of GLP-1 and PrRP in rat tissue sections was achieved using the mouse monoclonal GLP-1 antibody and the rabbit PrRP polyclonal antiserum. Sections were pretreated by incubation in 1.0% sodium borohydride to increase antibody penetration and reduce nonspecific background. Following PB washes the tissue was incubated in mouse anti-GLP-1 overnight at room temperature, followed by multiple washes in PB prior to and following a 3-hr room temperature incubation in affinity purified donkey anti-mouse secondary antibody conjugated to HRP. Sections were then incubated in Tyramide conjugated to CY5 or FITC for 10 min at room temperature. After GLP-1 labeling, sections were treated with 0.5% H2O2 for 15 min to remove residual peroxidase activity, followed by three 10min rinses with PB. The immunofluorescence processing was then repeated using the rabbit polyclonal PrRP antiserum at a dilution of 1:5 K, affinity purified donkey anti-rabbit secondary antibody conjugated to HRP, and tyramide conjugated to CY3. Sections were mounted on gelatin-coated slides and coverslipped using Vectashield Hard Set Mounting Medium (Vector Laboratories).

| Immunoperoxidase labeling of mouse tissue
Yellow fluorescent protein immunoperoxidase-labeled brainstem sections from six male and two female PPG-YFP mice were generated in previously published projects that focused on other research questions (Llewellyn-Smith et al., 2013;Llewellyn-Smith, Reimann, Gribble, & Trapp, 2011). Thus, while no new immunoperoxidase-labeled mouse material was generated for this study, the cNTS-focused microscopic analysis of this material has not been previously reported.
Close appositions by PPG-YFP axons onto PPG-YFP dendrites were identified in sections labeled using chicken anti-GFP (Abcam; 1:20 K; AB_13970) and a Ni-intensified DAB reaction, as previously reported (Llewellyn-Smith et al., 2011, 2013. Close appositions by PPG-YFP axons onto PPG-YFP cell bodies were identified in sections labeled using the same chicken anti-GFP antibody diluted 1:400 K, followed by an imidazole-intensified DAB reaction.

| Electron microscopic immunoperoxidase localizations
The fixed caudal brainstem from each rat perfused for ultrastructural analysis was sectioned at 100 μm using a Vibratome 1000 Plus (MyNeuroLab, Richmond, IL). Sections were collected sequentially in two wells of PBS such that each well contained sections at a frequency of 200 μm. Both wells of tissue were rinsed in PB for 30 min prior to and following a 10 min incubation in 1.0% sodium borohydride. The sections were then incubated in 1.0% hydrogen peroxide for 10 min, washed in PB for 30 min, transferred to cryoprotectant (Watson et al., 1986) and placed in a −80 C freezer for 1 hr.
Sections were returned to room temperature over 30 min, washed in 0.1 M Tris-buffered saline (TBS) for 30 min, and then treated with a blocking solution containing 1.0% bovine serum albumin, 3.0% normal donkey serum, and 0.4% Triton X-100 in TBS for 30 min. One well of sections was incubated overnight at room temperature with agitation in the GLP-1 rabbit polyclonal antiserum (1:7 K) and the other well in the GLP-1 mouse monoclonal antibody (1:15 K), both diluted with the blocking serum. Sections were subsequently washed in TBS for 30 min prior to and following a 1-hr incubation in species-specific biotinylated secondary antibodies at a dilution of 1:250. Vectastain reagents (Vector Laboratories) and DAB catalyzed with hydrogen peroxide were used to generate the immunoperoxidase reaction product.
Following the DAB reaction the sections were washed in multiple changes of PB and prepared for plastic resin infiltration and embedding. This involved postfixation of sections in a 1.0% osmium tetroxide solution (Electron Microscopy Sciences, Hatfield, PA) containing 1.5% potassium ferricyanide (Sigma-Aldrich) for 30 min followed by multiple PB washes to clear unbound osmium from the tissue. Sections were then dehydrated using 15-min changes in increasing concentrations of ethanol (50, 50, 70, 70, 95, 95, 100, 100%) and passed through four 15-min changes of acetone.
The last acetone wash was followed by sequential hour-long incubations of sections in 50, 70, and 95% Epon-Araldite plastic resin diluted with acetone followed by three consecutive hour-long incubations in 100% resin. Each section was then flat embedded in 100% resin between acrylic sheets, followed by resin polymerization in a 60 C oven.
2.6 | RNAscope in situ hybridization 2.6.1 | Co-localization of mRNA for GLP1R and PPG or GAD1 , and air-dried at room temperature for 2 hr. Slides were then dipped in 100% ethanol for 10 sec and air-dried before creating a hydrophobic barrier around each tissue section using a Pap Pen (195506; ThermoFisher Scientific). The barrier was allowed to completely dry at room temperature overnight before proceeding to the next step. Unless otherwise noted, incubations were performed at 40 C within a HybEZTM Oven, using the HybEZTM Humidity Control Tray. For each section, two to four drops of each of the following reagent solutions were used to cover tissue sections, followed by three 3-min 1× washing buffer rinses at room temperature.
Sections were incubated with Protease III (322337) for 30 min and washed in deionized H 2 O four times for 1 min each. Sections were then incubated with a cocktail of the two probes (Rn-Glp1r plus either Rn-Gcg-C2 or Rn-GAD1-C2) for 2 hr, followed sequentially by amplification steps with Amp1 for 30 min, Amp2 for 15 min, Apm3 for 30 min, and Amp4-AltC (#332857) for 15 min. Sections were incubated with Amp5 for 30 s at room temperature and immediately covered with Fluoromount-G ® (0100-01; Southern Biotech, Birmingham, AL, USA) mounting medium, without rinsing.
2.6.2 | Co-localization of GLP1R mRNA and PrRP immunolabeling GLP1R mRNA was first identified using the Rn-Glp1r probe 2.7 | Microscopy data collection and analysis 2.7.1 | Light microscopy in rat tissue The distribution and density of GLP-1+ cells and processes within the cNTS was characterized in rats using both immunoperoxidase-and immunofluorescence-labeled material. An Olympus BX51 photomicroscope equipped with a Hamamatsu camera (Hamamatsu Photonics, Hamamatsu, Japan) was used to collect brightfield and epifluorescence images, and a Leica TCS SP8 confocal microscope was used to collect optical slice images and z-stacks. Leica LAS 4.0 image collection software was used to generate maximum intensity z-projections and 3D maximum intensity projections (MIPs), and the 3D visualization module was used for rendering/reconstructing tissue volumes.
Region of interests (ROIs) from each 3D MIP were extracted to reveal details of associations between fluorescently labeled profiles.

| Light microscopy in mouse tissue
An Olympus BH-2 brightfield microscope was used to examine immunoperoxidase-labeled sections revealing YFP+ neurons and processes. Close appositions between YFP+ axon terminals and YFP+ cell bodies and dendrites were identified using a 100× oil immersion objective. YFP+ terminals lying adjacent to YFP+ somata or dendrites were classed as closely apposed when there was no space visible between the profiles. In Ni-DAB-reacted tissue sections, the dense black reaction product prevented detection of appositions in which YFP+ axons overlay YFP+ cell bodies. This problem was overcome by further diluting the anti-GFP antibody to 1:400,000, and using a lighter brown immunoperoxidase reaction product to detect YFP+ profiles. In these sections, YFP+ terminals were considered to closely appose an underlying YFP+ cell body when the terminal and the cell body were in the same focal plane. Digital images of appositions were collected as TIFF files using a SPOT Insight Model 18.2 firewire color camera running SPOT RT software v5.2 (Diagnostic Instruments, Inc., Sterling Heights, MI). Adobe PhotoShop was used to adjust image sharpness, brightness, and contrast, and to create montages.
Immunofluorescence was visualized on an upright widefield microscope (Leica Wetzlar, Germany) equipped with GFP (Chroma 49002 Chroma Technology, Bellows Falls, Vermont, USA) and red fluorescent protein (RFP) (Chroma 49008) filter sets. Images were captured using a color camera (Retiga 3000; QImaging Surrey, BC, Canada) and Q-Capture Pro 7 software (QImaging). Each ultrathin section was systematically sampled using a previously described approach (see Figure 4 in Rinaman, Card, Schwaber, & Miselis (1989)). This involved serial horizontal passages across the full extent of the ultrathin section, taking note of landmarks and immunopositive profiles identified in the thick sections. GLP-1+ profiles were photographed at both low and high magnification in one or more ultrathin sections. Digital electron micrographs of immunopositive cells were captured to document the subcellular distribution of immunoperoxidase reaction product. Immunopositive axon terminals were examined to determine the distribution of immunoperoxidase reaction product, the cellular elements that they contacted and, when visible, the symmetry of synaptic contacts.

| RNAscope imaging
Low magnification images were acquired with a KEYENCE epifluorescent microscope (BZ-X700). Confocal images were acquired using a Leica TCS SP8 confocal microscope, 100× oil objective and 3× digital zoom. The Atto550 fluorophore (Rn-Glp1r) was excited using a 552 nm OPSL laser, and Atto647 (Rn-Gcg or Rn-GAD1) was excited using a 638 nm Diode laser Images were collected sequentially using Leica LAS 4.0 image collection software, and using the 3D Visualization module for rendering/reconstructing tissue volumes. A total of 21 0.28 μm-thick optical sections were collected to generate a maximum intensity z-projection and a 3D MIP. A ROI from the 3D MIP was optically sectioned and extracted to reveal details of the labeling profiles.
Imaging was performed using a Zeiss Axioskop 2FS widefield microscope (Zeiss, Oberkochen, Germany) with a 40× water immersion lens. Slices were continuously superfused with 32 C standard ACSF at a flow rate of 3-4 mL/min. Recombinant human GLP-1 (7-37) was obtained from Peprotech (London, UK) and glutamate from Sigma-Aldrich (Gillingham, Dorset, UK). Stock solutions were made up in H 2 O and diluted at least 1,000× when added to the ACSF.
Images were captured at 12-bit on a charge-coupled device camera (Q-Click; QImaging). Recordings were imported into FIJI image analysis software. XY-drift was adjusted for using the StackReg plugin.
ROIs and an area for determining background fluorescent intensity were outlined and the mean pixel intensity calculated for each ROI.
Background intensity was subtracted from each ROI and recordings were adjusted for bleaching using a cubic polynomial function. Fluorescence intensity data are presented as ΔF/F 0 with F 0 being the average fluorescence intensity 5 min prior to the first stimulus and ΔF being the fluorescence intensity, F, minus F 0 . "N" numbers indicate number of analyzed ROIs. Responses were quantified by calculating area under the curve for more than 5 min during the stimulus, starting at the beginning of the stimulus. Because they were found not to be normally distributed, summary data are presented as box plots with whiskers, with the median indicated and whiskers marking the 10th and 90th percentile, respectively. Statistical significance of differences was determined using nonparametric statistical testing as described in the legend of Figure 11.

| RNAscope results
Dual in situ hybridization using RNAscope revealed both PPG and GLP-1R mRNA transcripts within the rat cNTS and medullary reticular formation (Figure 8). GLP1R mRNA transcripts also were present in additional regions of the caudal medulla, including the area postrema, inferior olive, midline raphe nuclei, and around the central canal.
Within the cNTS, GLP1R mRNA transcripts were sometimes present within the vicinity of PPG mRNA transcripts, but generally were not located in close proximity. In a few cases, however, scattered GLP1R mRNA transcripts appeared to overlap with PPG mRNA transcripts ( Figure 8a). In those cases, closer inspection using high-magnification confocal microscopy and rotation of 3D projection images failed to demonstrate intracellular colocalization of GLP-1R and PPG mRNA transcripts (Figure 8b). Conversely, GLP1R mRNA was localized within some GABAergic neurons (i.e., expressing GAD1 mRNA) within the cNTS and adjacent reticular formation (Figure 8c), whereas other GABAergic neurons did not co-express GLP1R mRNA (Figure 8d).
When RNAscope localization of GLP-1R mRNA was combined with immunofluorescent localization of PrRP, confocal microscopy revealed that GLP-1R mRNA transcripts were co-localized within a subset of PrRP+ noradrenergic A2 neurons (Figure 9).

| Transgenic mouse cross-breeding strategy
Pre-proglucagon (PPG)-YFP mice were crossed with GLP-1R-Cre/ tdRFP mice to obtain mice that expressed the YFP reporter in PPG neurons, and the tdRFP reporter in GLP-1R-expressing cells. Both tdRFP and YFP are expressed as soluble proteins, and labeled cell bodies were detected by immunofluorescence using antibodies raised against dsRed (to label RFP) and against GFP (to label YFP), as illustrated in Figure 10. In five mice (three male, two female), the incidence of co-localization of tdRFP in YFP+ cells was quantified within the cNTS, IRT, and midline raphe ventral to the hypoglossal nucleus.
Of 736 NTS PPG neurons, only seven were RFP+. Similarly, only 1 out of 322 IRT PPG neurons and 2 out of 65 midline PPG neurons were RFP+. Thus, the proportion of mouse PPG neurons that were doublelabeled for GLP-1R fluorescent reporter was very small (i.e., 0.95% in the NTS, 0.31% in the IRT, and 3.08% along the midline).

| Ca 2+ imaging
The response of identified PPG neurons to bath application of GLP-1 and glutamate was investigated in ex vivo slices through the caudal medulla in Glu-Cre/Rosa26-GCaMP3 mice (n = 3 mice; Figure 11). When slices were exposed to 100 nM GLP-1 for 3 min, none of the identified PPG neurons (0 out of 27) exhibited an intracellular Ca 2+ response, whereas all 27 neurons responded to 100 μM glutamate (Figure 11d). This was also the case for the readily visible dendrites of PPG neurons; no intracellular Ca 2+ response was observed on application of GLP-1, but all sampled dendrites responded to 100 μM glutamate (n = 23 dendrites; Figure 11e).
Thus, bath application of GLP-1 did not elicit a functional response in either the somatic or dendritic compartments of PPG neurons.
However, the ability of restraint stress to activate cFos expression in PrRP+ neurons and also within the anterior bed nucleus of the stria terminalis (which receives axonal input from PrRP+ and GLP-1+ neurons) is significantly attenuated in rats after central (i.c.v.) administration of Exendin-9, a specific GLP-1R antagonist (Maniscalco et al., 2015). Our new anatomical evidence suggesting that GLP-1+ fibers provide direct input to PrRP+ neurons, coupled with RNAscope in situ hybridization evidence that PrRP neurons express GLP-1R mRNA, support the view that PrRP neurons are directly responsive to GLP-1, and that GLP-1R signaling at least partially underlies their stress sensitivity. Dual-labeling electron microscopy will be necessary to confirm the synaptic nature of close appositions between GLP-1+ and PrRP+ neurons in both mice and rats, although earlier work suggests that at least half of the close appositions observed at the light microscopic level represent actual synaptic contacts (Pilowsky, Llewellyn-Smith, Lipski, & Chalmers, 1992). However, unlike the point-to-point release of glutamate and GABA, which appears to require synaptic contacts, features of these GLP-1-to-GLP-1 synaptic contacts within the caudal NTS is consistent with evidence that GLP-1 neurons are glutamatergic, and provide asymmetric synaptic inputs to neural targets within the rat hypothalamus (Sarkar et al., 2003;. The presence or absence of synaptic contacts between GLP-1 neurons in mice has not been previously reported. Instead, in an earlier electrophysiological study using perforated patch recordings in ex vivo slices from PPG-YFP mice, no GLP-1R-mediated effects were observed on the resting membrane potential, membrane resistance, or firing frequency of identified PPG neurons and no GLP-1R mRNA was detected in single cell RT-PCR analysis (Hisadome et al., 2010). Given our new anatomical evidence for close appositions and synapses between PPG/GLP-1 axon terminals and dendrites, we revisted the lack of physiological effects using a different approach. For this, ex vivo slices were prepared from mice in which the Ca 2+ indicator GCaMP3 was selectively expressed in PPG neurons, followed by optical recordings of intracellular Ca 2+ concentration in PPG cell bodies and dendrites before and after exposure to GLP-1. This approach lacked the temporal resolution of electrophysiological recordings, but had higher spatial resolution and allowed analysis of several PPG neurons simultaneously. It also enabled us to specifically investigate the presence or absence of (distal) dendritic responses to GLP-1 that do not propagate to the cell soma. Such responses would have gone undetected in our earlier electrophysiological study, but have been reported recently for 5-HT acting on PPG neurons (Holt et al., 2017).
The observed absence of GLP-1-induced changes in somatic or dendritic intracellular Ca 2+ is consistent with the earlier electrophysiological reports and also with our inability to detect mRNA for GLP-1 in those cells Hisadome et al., 2010). Conversely, some catecholaminergic (i.e., TH+) cNTS neurons do express GLP-1R in mice , consistent with new evidence (discussed below) that noradrenergic PrRP+ neurons in rats express mRNA for GLP-1R.   (Maniscalco et al., 2015), consistent with the current report demonstrating a lack of GLP-1R mRNA expression by GLP-1 neurons in rats, or by PPG neurons in mice.

GLP-1R is expressed by
Glutamatergic and GLP-1R signaling mechanisms interact in many brain regions (Gilman et al., 2003;Liu et al., 2017;Mietlicki-Baase et al., 2013, and GLP-1R signaling may modify excitatory synaptic transmission via cellular mechanisms involved in long-term synaptic plasticity . Speculatively, as GLP-1+ axonal varicosities in rats and YFP+ varicosities in transgenic PPG reporter mice are immunoreactive for vesicular glutamate transporter 2 , synaptic signaling between GLP-1 neurons may be mediated by glutamate. Light-evoked release of glutamate from PPG nerve terminals within the hypothalamic paraventricular nucleus has been demonstrated in mice ; similarly, clonal intestinal L-cells that secrete GLP-1 express VGLUT2 and co-release glutamate, hypothesized as important for intercellular signaling among intestinal L-cells (Uehara et al., 2006). The relatively low density of GLP-1+ fibers and terminals within the rat cNTS (see Electron micrographs illustrating GLP-1+ processes in the rat caudal nucleus of the solitary tract. (a) Immunopositive vesicles derived from the Golgi complexes of the cell soma extended into the dendritic arbor to produce discrete labeling (arrows). (b, c) Immunopositive axon terminals displayed a range of labeling density that was directly correlated with the proximity of the ultrathin section to the surface of the flat embedded tissue. Profiles within the top few microns of the tissue (b) contained dense concentrations of immunoperoxidase reaction product that filled the profile, while those deeper in the tissue (c) displayed more punctate labeling. GLP-1+ terminals were densely filled with 40 nm lucent spherical vesicles and occasional dense core vesicles. Immunopositive terminals formed appositions with both unlabeled (b-d) and labeled (e-h) dendrites. (f, g) When synapses (open arrowheads) were present within the plane of section they were asymmetric in character and, in many cases, formed multiple synapses with the same profile. Immunopositive dendrites also were recipient of synaptic input from immunonegative terminals that exhibited both asymmetric/excitatory (i) and symmetric/inhibitory characteristics. (j) GLP-1+ axon varicosities often were closely apposed to unlabeled axon terminals (ut) filled with 40 nm lucent spherical vesicles. Scale bars: (a) 2 μm; (b) 500 nm; (c, f ) 600 nm; (d) 800 nm; (g-i) 1 μm. d, dendrite; ut, unlabeled terminal neurons, with the local projections signaling primarily via glutamate. This remains to be determined using technical approaches with more sensitivity than immunolabeling. Given the lack of GLP-1R expression by GLP-1 neurons, GLP-1 released either locally or distally is unlikely to directly alter subsequent neurotransmitter release from these neurons. Conversely, locally or distally released GLP-1 may act at pre-or postsynaptic GLP-1R receptors to modulate neuropeptide and/or glutamate or GABA signaling from subsets of A2 PrRP+ cells and GABA cells within the cNTS, given our new evidence that subsets of these neurons express GLP-1R. Additional studies will be required to test these hypotheses, including studies to determine the axonal projections of PrRP+ neurons that either do or do not express GLP-1R.
Given the large number of CNS regions receiving axonal input from this relatively small population of hindbrain neurons (Gu et al., 2013), individual GLP-1/PPG neurons likely have axon collaterals that branch to innervate more than one target site; indeed, such collateralization has been demonstrated for GLP-1 neural inputs to the paraventricular and dorsomedial nuclei of the hypothalamus in rats (Vrang et al., 2007). If subsets of GLP-1/PPG neurons innervate different central target sites, then local glutamatergic synaptic connections between these neurons within the cNTS might provide moment-to-moment coupling that coordinates the activities of different target-specific Confocal imaging of PrRP immunolabeling (green) and GLP1R mRNA (magenta) within the rat caudal nucleus of the solitary tract. (a) "Side view" of a rotated image stack depicting a doublelabeled neuron. The same neuron is shown from a "top view" in a maximum projection z-stack image in panel (c), and from an "end-on" image stack view in panel (d). Within the same field of view (visible in each panel), other double-labeled PrRP+ neurons are located nearby, although some lack intracellular colocalization of GLP1R mRNA (e.g., the neuron located on the right in panels (a)