Distribution and characterisation of Glucagon-like peptide-1 receptor expressing cells in the mouse brain

Objective Although Glucagon-like peptide 1 is a key regulator of energy metabolism and food intake, the precise location of GLP-1 receptors and the physiological relevance of certain populations is debatable. This study investigated the novel GLP-1R-Cre mouse as a functional tool to address this question. Methods Mice expressing Cre-recombinase under the Glp1r promoter were crossed with either a ROSA26 eYFP or tdRFP reporter strain to identify GLP-1R expressing cells. Patch-clamp recordings were performed on tdRFP-positive neurons in acute coronal brain slices from adult mice and selective targeting of GLP-1R cells in vivo was achieved using viral gene delivery. Results Large numbers of eYFP or tdRFP immunoreactive cells were found in the circumventricular organs, amygdala, hypothalamic nuclei and the ventrolateral medulla. Smaller numbers were observed in the nucleus of the solitary tract and the thalamic paraventricular nucleus. However, tdRFP positive neurons were also found in areas without preproglucagon-neuronal projections like hippocampus and cortex. GLP-1R cells were not immunoreactive for GFAP or parvalbumin although some were catecholaminergic. GLP-1R expression was confirmed in whole-cell recordings from BNST, hippocampus and PVN, where 100 nM GLP-1 elicited a reversible inward current or depolarisation. Additionally, a unilateral stereotaxic injection of a cre-dependent AAV into the PVN demonstrated that tdRFP-positive cells express cre-recombinase facilitating virally-mediated eYFP expression. Conclusions This study is a comprehensive description and phenotypic analysis of GLP-1R expression in the mouse CNS. We demonstrate the power of combining the GLP-1R-CRE mouse with a virus to generate a selective molecular handle enabling future in vivo investigation as to their physiological importance.


INTRODUCTION
Glucagon-like peptide 1 (GLP-1) is traditionally recognised as a peripheral incretin hormone. Released postprandially from intestinal Lcells, it binds to GLP-1 receptors (GLP-1R) on pancreatic b-cells to increase insulin secretion. Additionally, GLP-1 acts as a neuropeptide and is produced by preproglucagon (PPG) neurons located in the lower brainstem, primarily in the caudal nucleus tractus solitarii (NTS) and the intermediate reticular nucleus [1,2]. Numerous rodent studies have addressed the question of GLP-1's action within the brain (for review see: [3,4]). Most studies have examined the effects of cerebroventricular injection of GLP-1 or its stable analogue exendin-4 (Ex-4) on food intake [5], but other reported effects include changes in heart rate and blood pressure, as well as effects on learning, memory and thermogenesis [6e9]. Furthermore, central GLP-1 appears to be involved in mediating the rewarding effect of food. Activation of GLP-1R within the ventral tegmental area (VTA) and nucleus accumbens (NAc) results in a significant decrease in the consumption of high calorie food and reduced body weight gain over 24 h [10]. Although the exact mechanism by which GLP-1 affects the brain is unclear, it has been demonstrated that within the VTA and NAc, GLP-1 appears to bind to presynaptic GLP-1R on glutamatergic terminals facilitating glutamate release, without triggering direct postsynaptic effects [11,12].
It has previously been shown that GLP-1 producing neurons in the NTS send ascending axons to forebrain and brainstem regions associated with metabolic and autonomic control in the mouse, including the arcuate nucleus (ARC), hypothalamic paraventricular nucleus (PVN), rostral ventrolateral medulla and dorsomedial hypothalamus (DMH) [1,13]. These projections correlate well with the pattern of GLP-1R expression throughout the brain of the rat [14]. Given the potent effects of central GLP-1 on food intake, it is important to fully elucidate the exact expression profile of GLP-1R throughout the brain. The distribution of GLP-1R mRNA has previously been assessed in the rat brain by in situ hybridisation [14]; however, little is known about the distribution in mouse. This gap in knowledge is notable, especially given the increased usage of transgenic mouse models in metabolic physiology.
Here we detail the distribution of GLP-1R expressing cells throughout the mouse brain using a novel transgenic model, in which crerecombinase is expressed under the control of the Glp1r gene. We show that GLP-1R expression correlates well with that observed in the rat [14], non-human primates [15] and with the projection pattern of mouse PPG neurons [1,13]. We also demonstrate that GLP-1R expressing cells do not exhibit immunoreactivity for glial fibrillary acid protein (GFAP) or parvalbumin (PARV), but do show tyrosine hydroxylase (TH) immunoreactivity in some areas. Furthermore, we establish that this mouse model is amenable to manipulation of specific subsets of GLP-1R expressing cells with cre-dependent viral gene transfer. Finally, we assess the feasibility of using this mouse model for in vitro analysis using brain slice patch-clamp recordings by recording electrical responses to exogenous GLP-1 from GLP-1R expressing cells in several brain regions.

Animals
All animal experimentation was carried out in accordance with the UK Animals (Scientific Procedures) Act, 1986, and had the required ethical approval. The entire study was performed on GLP-1R-CRE mice that express cre-recombinase under the Glp1r promoter. Generation of these mice has been described in detail before [16]. These mice were crossed with ROSA26-tdRFP or ROSA26-eYFP reporter strains [17] to enable fluorescent detection of cells expressing GLP-1R by cytosolic tdRFP or eYFP expression, respectively. A total of 3 GLP-1R-CRE-eYFP and 8 GLP-1R-CRE-tdRFP mice of both sexes were used for the immunofluorescence study, whereas the electrophysiological study was performed entirely with the GLP-1R-CRE-tdRFP strain. Mice were kept on a 12 h light e 12 h dark cycle, had access to food and water ad libitum, and were 8e16 weeks old (weighing between 25 and 35 g) when used for experiments. Females were consistently lighter than males.

Immunofluorescence
Mice were anaesthetised with urethane (1.5 g/kg; i.p.) and perfusefixed transcardially with heparinised 0.1M phosphate buffer (PB) followed by 4% paraformaldehyde in 0.1M PB, pH 7.4. Brains were removed from skulls, post-fixed overnight in the same fixative before being cryoprotected in 30% sucrose in 0.1M PB. Brains were sectioned coronally at a thickness of 30 mm (1:5 series) from the olfactory bulbs to the spinomedullary junction.
2.2.1. Immunohistochemical identification of eYFP or tdRFP Tissues were blocked with 0.1% Triton X-100 and 10% normal goat serum diluted in 0.1M PB for 1 h at room temperature. GLP-1R-expressing cells were detected with antibodies raised against either green fluorescent protein (GFP; Catalogue #AB13970, lot #623923; Abcam, Cambridge MA, USA) or dsRED (Clontech #632496). These were added to the blocking solution in a 1:1000 dilution and incubated overnight at 4 C. Subsequently, sections were washed 3 times for 5 min in 0.1M PB at room temperature, followed by incubation with a fluorescently labelled secondary antibody (1:500 dilution) in blocking solution for 2 h. Either an anti-chicken Alexa Fluor 488 (Catalogue# A-11039, Invitrogen) or an anti-rabbit Cy3 (Catalogue# C2306, Sigma Aldrich, St Louis MO, USA) was used. Sections were washed again as previously described, then mounted onto polylysine-coated slides (Catalogue# 631-1349, VWR), air dried and coverslipped with Vectashield mounting medium (Catalogue# H-1200, Vector Labs).

Double immunofluorescence
Staining was performed as described above with the exception of using sheep serum rather than goat for TH staining. Sections were then incubated with anti-GFP and anti-TH (1:1000 dilution, catalogue# sc-14007, Santa Cruz) or anti-GFP and anti-GFAP conjugated to Cy3 (catalogue# C9205, Sigma Aldrich, St Louis MO, USA), or anti-dsRED and anti-parvalbumin (1:5000 dilution, catalogue# 235, Swant, Switzerland) overnight at 4 C. Following primary antibody incubation sections were washed as previously described and incubated with the appropriate secondary antibody: for TH the anti-rabbit conjugated to Cy3, and for parvalbumin the Alexa Fluor 488 goat anti-mouse (catalogue# A10680, Invitrogen), respectively. Sections were then washed and mounted as described above.

Data collection and analysis
Stained sections were analysed on an upright epi-fluorescence microscope (Leica DMRB, Leica Microsystems, Milton Keynes, UK) equipped with a Retiga 3000 colour digital camera (QImaging, Surrey, Canada). Images were captured using QCapture software (QImaging) and subsequently exported to Corel Photo-Paint X3 (Corel Corporation, Ottawa, Canada) where brightness and contrast were adjusted. Microsoft Image Composite Editor (Microsoft Corporation, Redmond, USA) was used to produce montages from several frames.

Viral targeting
Adeno-associated virus (AAV) particles were produced as described by Murray et al. [18]. Briefly, to produce rAAV virions containing a 1:1 ratio of type 1 and type 2 capsid proteins human embryonic kidney (HEK) 293 cells were co-transfected with the AAV plasmid pAAV-EF1adouble floxed-hChR2(H134R)-EYFP-WPRE-HGHpA (Addgene plasmid # 20298) and AAV1 (pH21), AAV2 (pRV1) helper plasmids as well as the adenovirus helper plasmid pFD6 using calcium phosphate. 64 h post-transfection, cells were harvested, digested with sodium deoxycholate allowing AAVs to be purified from lysates using 1 ml HiTrap heparin columns (Sigma). Eluted virions were concentrated using Amicon Ultra-4 centrifugal filter devices (Millipore U FC810024). Adult GLP-1R-CREetdRFP mice were anaesthetised with ketamine (75 mg/kg; i.m.) and medetomidine (0.5 mg/kg; i.m.). Deep anaesthesia was confirmed by the absence of a pedal withdrawal reflex to a toe pinch. The head was fixed in a stereotaxic frame and the skull was exposed to allow a cranial window to be drilled at the correct location for a unilateral microinjection (150 nl; 50 nl/min) of a solution containing viral particles of EF1a-double floxed-hChR2(H134R)-EYFP AAV serotype 1/2 to target GLP-1R neurons. The coordinates used to target the PVN were 0.82 mm caudal and 0.1 mm lateral from Bregma and 4.75 mm ventral from the surface of the skull at Bregma. After removal of the injection needle the skin was sutured and anaesthesia reversed with atipamezole (1 mg/kg; i.m.). Post-operative analgesia (Buprenorphine, 0.05 mg/kg per day, s.c) was given for two days. Viruses were incubated for 3 weeks before injected animals were transcardially perfused-fixed as detailed above.

Electrophysiology
Coronal (250 mm thick) brain slices were obtained from adult transgenic mice of either sex as described previously [2]. Slices were kept at 34 C in artificial cerebrospinal fluid (ACSF) of the following composition (in mmol/l): 118 NaCl, 3 KCl, 25 NaHCO 3 , 1 MgCl 2 , 2 CaCl 2 , 10 glucose; pH 7.4. Experimental solutions were constantly bubbled with 95% O 2 /5% CO 2 . Glucagon-like peptide 1 (7e36) amide (#2082) was obtained from R&D systems and prepared as a 100 mM stock solution in MilliQ H 2 O. Whole-cell recording were carried out in ACSF at 28e32 C using a 'low Cl' pipette solution of the following composition (in mM): 120 K-gluconate, 5 HEPES, 5 BAPTA, 1 NaOH, 1 MgCl 2 , 1 CaCl 2 , and 5 K 2 ATP, 0.25 NaGTP; pH 7.2. The junction potential of À10 mV for this pipette solution was added offline. All figures and values in the results are corrected for the junction potential. Patch-clamp recordings in the voltage-clamp or current-clamp mode were performed on visually identified red fluorescent cells within the bed nucleus of the stria terminalis (BNST), the hypothalamic PVN and the hippocampus using an EPC-9 amplifier and Pulse/Pulsefit software (HEKA-Electronics, Germany). In voltage-clamp membrane resistance was determined by 500 ms hyperpolarising voltage pulses of 40 mV amplitude from the holding potential every 20s. Data are given as mean AE one S.E.M. Statistical significance was tested using a paired or unpaired t-test as appropriate, p-values < 0.05 (*) or < 0.01 (**) were considered significant.

RESULTS
GLP-1R-tdRFP and -eYFP expressing mice showed similar intensity of fluorescent reporter expression throughout the murine brain, with both strains exhibiting good somatic and proximal neurite staining. Distal neurites were only observed in certain brain regions, assumed to be a result of lower cell density. For example, cells in the PVN showed somatic and proximal neurite staining only, whereas cells in the BNST showed both proximal and more distal neurite staining. Both tdRFPand eYFP-expressing mice were used in this study and both tdRFPand eYFP-immunoreactive cells will hereby be referred to as GLP-1R cells.

Distribution of GLP-1R cells
GLP-1R expressing neurons were examined from the spinomedullary junction to the olfactory bulb of the mouse brain. Densely grouped GLP-1R neurons were observed both in the area postrema (AP; Figure 1A, B) and in the subfornical organ (SFO; Figure 1C), indicating sites where gut-released GLP-1 may directly influence neuronal activity. Interestingly, only few GLP-1R cells were found in the organum vasculosum lamina terminalis, the most rostral circumventricular organ. Many GLP-1R neurons were also found throughout hypothalamic nuclei (Figure 1DeE), particularly in the PVN, DMH and ARC, in-line with in situ hybridisation and immunocytochemical studies from rat and primate (Table 1). Furthermore, considerable numbers of GLP-1R neurons were observed in the periaqueductal grey, the mammillary recess, the central nucleus of the amygdala and the caudal hippocampus ( Figure 1F). A schematic representation of the distribution of GLP-1R neurons throughout the mouse brain is given in Figure 2 and the density of GLP-1R neurons in individual brain nuclei is quantified in Table 1. Within the Table 1 e Location of GLP-1R expressing cells in this study, compared to PPG neuron projections [1] in mouse, detected by transgene expression, rat GLP-1R expression, analysed by in situ hybridisation [14], and non-human primate GLP-1R expression, determined by in situ hybridisation [15].

Region
Mouse TG

Mouse PPG projections
Rat ISH

Monkey ISH
Telencephalon Olfactory Granule cell layer    In the diencephalon, GLP-1R neurons were located at a high density throughout the hypothalamus, particularly in the magnocellular and parvocellular subdivisions of the PVN, DMH and ARC, whereas smaller numbers of neurons were scattered throughout the lateral and posterior hypothalamus. In the thalamus, GLP-1R neurons were observed in the ventral thalamic nuclei, the paraventricular nucleus, the zona incerta, the nucleus reuniens, parasubthalamic nucleus, reticular thalamic nucleus, posterior thalamus, parafascicular thalamus and precommissural area. In the hippocampus, few neurons were observed in the rostral aspect of the CA3 region and the dentate gyrus. Many more neurons, however, were observed in the caudal CA3 and CA1 regions where the hippocampus extends ventrally ( Figure 1F). In the telencephalon, expression was observed in the plexiform cell layer of the olfactory bulb, posterior and ventral subnuclei of the BNST ( Figures 1C and 3C), lateral septum ( Figure 3E, ventral tenia tecta, claustrum ( Figure 3B), preoptic area, SFO, central, medial and basolateral amygdaloid nuclei ( Figure 3D). Few scattered neurons were located in both the nucleus accumbens shell and core ( Figure 3B). Within the neocortex two regions were observed that showed populations of GLP-1R neurons, the piriform and cingulate cortices (Figures 2, 3A, 5B, C). The remainder of the cortex was virtually devoid of GLP-1R neurons with very occasional rare scattered neurons found, as indicated in Figure 2.
The mesolimbic system exhibited high densities of GLP-1R neurons in the BNST (Figure 3C), the amygdala ( Figure 3D) and the lateral septum (LS; Figure 3E). Interestingly, only few somata were identified in the ventral tegmental area (VTA) ( Figure 4E) and NAc ( Figure 3B), although both areas have previously been investigated in relation to the effects of GLP-1R activation on food intake in rats.

Co-localisation with catecholamine neurons
Catecholamine neurons were revealed through immunofluorescent labelling of tyrosine hydroxylase (TH) on brains of GLP-1R-eYFP mice.
In the medulla, TH-immunoreactive neurons were identified throughout the AP, NTS and ventrolateral medulla (VLM). Within these areas, a subset of TH-immunoreactive neurons co-localised with eYFP expression (Figure 4AeC). Within the AP the majority of THimmunoreactive neurons were localised at the dorsal surface, and in agreement with previously published work demonstrating GLP-1 receptors on these cells [19], almost all TH neurons co-expressed eYFP. Within the NTS, TH-immunoreactive neurons were most commonly located towards the midline, whereas eYFP expressing cells were more commonly expressed towards the lateral edge, however, there were a few instances where colocalisation occurred (Figure 4 A, B). TH-immunoreactive neurons were found scattered throughout the VLM, and around half were found to co-localise with eYFP positive neurons ( Figure 4C, G). However the number of eYFP positive neurons observed was much greater when compared with TH-immunoreactive neurons. Altogether, only a small proportion of GLP-1R neurons in the VLM were found to be TH-immunoreactive.
The midbrain VTA has a high density of TH-immunoreactive neurons throughout, but a low level of eYFP-positive somata ( Figure 4E, H, I). Of these eYFP-positive neurons, a small number, predominantly located on the ventral border, were found to contain TH-immunoreactivity.
Within the hypothalamus, TH-immunoreactive neurons were found scattered throughout the dorsomedial hypothalamus (DMH), a small proportion of these were found to co-stain for eYFP ( Figure 4D, F). However the majority of GLP-1R neurons in the DMH were not THimmunoreactive. TH-immunoreactive neurons were also identified in the PVN, with the highest densities located in the caudal PVN. Here, no eYFP neurons were identified as being TH immunoreactive, however eYFP neurons were interspersed amongst TH-immunoreactive neurons.
3.3. Co-localisation with parvalbumin neurons Parvalbumin (PARV)-immunoreactive neurons were predominately found in the hippocampus, cortex and thalamus (Figure 5AeC). In the thalamus, PARV staining was restricted to the reticular nucleus and superior colliculus. Within these regions, no cells were found to be immunoreactive for both RFP and PARV. Similarly, in the hippocampus, cortex and reticular nucleus, populations of both RFP-and PARVimmunoreactive neurons intermingled, but the two antigens were not found to co-localise.

Co-localisation with GFAP
Astrocytes were detected by immunofluorescent labelling of glial fibrilliary acidic protein (GFAP) in GLP-1R-eYFP brain sections (Figure 5DeG). Within the dorsal vagal complex (DVC), GFAPimmunoreactive cells were observed throughout the dorsal motor nucleus of the vagus (DMNX), the midline of the NTS and along the border of the AP. No co-localisation between eYFP immunofluorescence and GFAP immunofluorescence was observed in the NTS or the DMNX. GFAP staining was predominately observed along the midline of the DVC, whereas eYFP-positive cells were generally observed more laterally. Within the AP, although both GFAP-positive and eYFP-positive cells were found, no co-localisation was observed ( Figure 5D). Similarly in the hypothalamic PVN, GFAP was observed predominately around the border of the 3rd ventricle, with scattered processes observed throughout the PVN. No eYFP-positive cell was observed to contain GFAP immunoreactivity ( Figure 5E). The ARC contained relatively large numbers of GFAP positive cells, but, again, there was no evidence of co-expression of eYFP-immunofluorescence and GFAP in individual cells ( Figure 5F). A similar observation was also made in the rostral CA3 region of the hippocampus, that exhibited a number of GLP-1R cells, but none of these were immunoreactive for GFAP ( Figure 5G).

Selective targeting of GLP-1R neurons in vivo Expression of cre-recombinase in cells where the GLP-1R promoter is active enabled the functional manipulation of GLP-1R neurons in vivo.
We injected EF1a-double floxed-hChR2(H134R)-EYFP AAV particles into the PVN to express ChR2-eYFP selectively in PVN GLP-1R neurons only ( Figure 6A). Axons expressing ChR2-eYFP were primarily found in the caudal NTS ( Figure 6B, C) and to a much smaller extent in the VLM (not shown).

Electrical responses of GLP-1R neurons to GLP-1 in vitro
In order to verify that tdRFP-or eYFP-immunoreactive neurons did indeed express GLP-1R and that the receptor is functional, whole-cell patch-clamp recordings were performed on tdRFP-neurons in three different brain regions in an in vitro brain slice preparation.
3.6.1. PVN Seven red-fluorescent PVN neurons were recorded in voltage clamp at a holding potential of À50 mV, and five cells were recorded in current-clamp mode. The mean resting potential in current-clamp mode was À57 AE 2 mV (n ¼ 5). Bath-application of 100 nM GLP-1 caused a depolarisation by 14 AE 1 mV in all five cells that reversed upon washout of the drug. In voltage clamp GLP-1 application induced a reversible inward current of 16 AE 3 pA (n ¼ 7) that was accompanied by a decrease in membrane resistance from 1.1 AE 0.2 GU to 0.8 AE 0.1 GU (Figure 7AeC, F).

BNST
Six BNST neurons were recorded in current clamp. Their resting potential was À56 AE 3 mV and bath application of 100 nM GLP-1 elicited a depolarisation by 8 AE 1 mV that was reversible upon washout of the drug. An additional five recordings were performed in voltage clamp at a holding potential of À50 mV. In these cells GLP-1 caused an inward current of 10 AE 2 pA ( Figure 7D, F).

DISCUSSION
This study has described the anatomical distribution of tdRFP-or eYFPimmunoreactive cells throughout the brain, from the spinomedullary junction through to the prefrontal cortex of the GLP-1R-Cre mouse. This mouse uses the Glp1r promoter to drive expression of the enzyme cre-recombinase, which in turn allows the expression of tdRFP or eYFP in the ROSA26 locus. These fluorescent markers were used to visually identify cells, either directly in acute brain slices used for electrophysiological recordings, or by immunofluorescence in fixed tissue sections. The extent of cellular labelling was comparable between the tdRFP and eYFP strains. In both cases, mostly somata and proximal dendrites were visualised and immunoreactivity did not reveal more distal dendrites or axons. In contrast to our earlier studies on PPG neurons [1,13,20], axon projections of GLP-1R neurons could not be visualised, due to insufficient amounts of RFP or eYFP in these structures. However, targeting the GLP-1R with AAV constructs as discussed below led to sufficient expression of the virally-encoded fluorescent marker in the axons of these cells to allow visualisation of projection targets. Given that cre-recombinase is expressed under the Glp1r promoter, tdRFP or eYFP should identify cells that express the GLP-1 receptor. Clear proof of whether this is indeed the case would be successful labelling with a GLP-1R antibody. However, the paucity of reliable GLP-1R antibodies is well established [21,22], and, consequently, a functional approach was taken in this study. We screened tdRFPpositive cells in a few brain regions with established (e.g. PVN), and with questionable (e.g. hippocampus), GLP-1R expression for electrophysiological responses to GLP-1. The results demonstrated that tdRFP fluorescence indeed signified functional responsiveness to GLP-1, and thus most likely the presence of GLP-1R on these cells.

Original article
726 Inevitably numbers of cells tested in this manner are low, and we cannot exclude with certainty that some subpopulations of RFPimmunoreactive cells described in this study might reflect either ectopic expression or lineage tracing [16]. However, within these potential constraints, the GLP-1R-Cre/ROSA26-tdRFP/eYFP mouse is an invaluable tool for examining GLP-1 function within the CNS. Firstly, this mouse obviates the need for notoriously unreliable antibodies for the detection of GLP-1R and therefore offers a greater level of reliability as antibodies raised against eYFP and tdRFP are generally less susceptible to non-specific binding (and their specificity can be easily tested in brain tissue from wildtype mice). Secondly, it allows the identification of GLP-1R expressing cells in live slice preparations and thus permits functional characterisation of defined GLP-1R expressing cells at cellular level in vitro. Finally, it enables the use of viral gene transfer to manipulate specific sets of GLP-1R expressing cells in vivo.

4.1.
Distribution of GLP-1R throughout the murine brain GLP-1R expressing cells were observed throughout the rostrocaudal extent of the mouse brain, and generally correlated with the projection targets of brainstem PPG neurons [1,13,23e25] and the expression of GLP-1R described in rat [14], although some differences were observed. Interestingly, the expression pattern is also very similar to that of a non-human primate (Table 1; [15]), which suggests that the GLP-1 system is largely conserved throughout the brain between mammalian species. This might not be surprising, given that the distribution of GLP-1 producing PPG neurons is also quite conserved between species [1,23,26,27].  [31] and the BNST is a major projection target for PPG neurons in both rat and mouse [1,32]. The AP and SFO, which form part of the brains' circumventricular organs, also contained large numbers of GLP-1R expressing cells. Moderate numbers of GLP-1R expressing cells were observed in the ventrolateral medulla, which like the PVN harbours presympathetic neurons, and thus both regions might contribute to the effects of GLP-1 on cardiovascular control as well as thermogenesis [8,9,33].
Of particular interest were brain areas such as the VTA and NAc, which have recently received attention for their ability to reduce food intake and reward behaviour following GLP-1R activation [10,34,35]. In these regions, there was little GLP-1R expression, in contrast to observations in rat [14], but in agreement with recent in situ hybridisation studies in mouse [31]. Indeed, tracing studies in rats have observed that around 30% of NTS PPG neurons project to the VTA and around 40% project to the NAc [25] suggesting that they are sites of GLP-1 release from brainstem PPG neurons. In vitro electrophysiological studies in rats have reported increased glutamatergic signalling onto dopaminergic neurons by GLP-1 acting via presynaptic receptors [11,12]. It is therefore possible to speculate that GLP-1Rs in these regions are present on terminals arising from other sites and the somata of these neurons would most likely be located in distal nuclei, thus providing an explanation for the absence of GLP-1R positive cell bodies in the VTA and NAc. Conversely, this discrepancy could be the result of species variation, since most in vivo behavioural studies involving the actions of GLP-1 in these brain regions were performed in rats. This notion may be supported by the fact that very few PPG axons can be found in the NAc of mice. Previous studies have reported the putative existence of GLP-1R on glial cells throughout the CNS [36,37]. Here, using double immunofluorescence for eYFP and the astroglial marker, GFAP, we found no co-localisation throughout the brain. This is in agreement with a recent study that located GLP-1R on microglia, but not astroglia, in rat spinal cord [38]. It should be noted, however, that immunofluorescent identification of GFAP predominately labelled astrocytic processes,  whereas eYFP predominately labelled somata. It is therefore possible that a more specific marker for astrocytic somata may reveal the existence of GLP-1R on glial cells. However, transcriptome analysis of astrocytes failed to demonstrate significant expression of GLP-1R [39] (see also  [43], like the PPG neurons, this seems a distinct possibility.

GLP-1R in hippocampus
Various studies have reported effects of GLP-1 in hippocampus, both at the cellular and systemic levels [6,44]. A recent study even reported a reduction in food intake upon injection of Ex-4 into the hippocampus [45]. However, in the absence of any clear innervation with GLP-1 containing axons from PPG neurons in rodents or primates [1,23,26], it is difficult to imagine the source of GLP-1 in hippocampus. Nevertheless, whilst non-human primates do not seem to express GLP-1R in hippocampus [15], both in situ hybridisation and the current study identify GLP-1R in parts of the hippocampus of rat and mouse, respectively [14]. The study by Merchenthaler and colleagues reported GLP-1R mRNA expressing cells in the ventral hippocampus of rat, particularly in the CA3 region [14]. The current study confirmed this in mouse, but also identified a small, succinct group of pyramidal cells in the rostral CA3. In contrast, GLP-1 binding studies failed to identify the hippocampus as a target [46]; however, since cell numbers in this region were low, it could be below the detection limit. Additionally, here we show electrical responses of individual identified red fluorescent hippocampal neurons to GLP-1. The recordings indicated two types of electrical responses, either the activation of an inward current, presumably a non-selective cation channel, or the activation of a K þ current. The former is an excitatory and the latter an inhibitory response. The combination of the fluorescence identification and the electrical responsiveness of these cells makes it highly likely that these hippocampal neurons indeed express GLP-1 receptors. This raises the question of whether these cells would be exposed to GLP-1 in vivo, and, if so, where would this GLP-1 originate? It has been proposed that GLP-1 could reach the hippocampus from the cerebrospinal fluid (CSF), and indeed detectable levels of GLP-1 were found in CSF samples [45]. However, whether CSF GLP-1 originates from gut-derived GLP-1 or is released into the CSF by PPG neurons is currently unclear.

Functional consequences
GLP-1Rs are expressed in regions protected by the blood brain barrier (BBB) as well as those outside. This provides two possible mechanisms through which GLP-1 may exert actions within the brain. It is generally believed that the rapid degradation of GLP-1 in the blood by circulating DPP-IV prevents gut secreted GLP-1 from crossing the BBB [47].
Receptors expressed on cells within circumventricular organs provide a mechanism by which circulating GLP-1 may exert actions within the brain. Indeed, it has been shown that GLP-1 responsive catecholaminergic neurons in the AP project to areas associated with autonomic control, including the parabrachial nucleus and RVLM [19]. However, the existence of GLP-1R in regions protected by the BBB suggests that brain-derived GLP-1 is involved in regulating at least some of the responses attributed to GLP-1 action in the brain. This is further supported by the existence of receptors in regions, which receive dense innervation from brainstem PPG neurons [1,13,23e25]. On the other hand, two of the circumventricular organs, the AP and the SFO, express a high density of GLP-1 receptor cells, but show a rather limited amount of innervation from PPG axons [1,13]. This might point to a role for peripheral GLP-1 in these regions, or to a role for CSF GLP-1. Further work will be needed to fully elucidate the relative contribution of brain-derived vs peripherally-derived GLP-1 on the brain GLP-1 system. Previous in vitro electrophysiological work has focussed on the VTA and NAc. It has been suggested that GLP-1Rs exert their effect through a pre-synaptic enhancement of glutamate release in the VTA and NAc [11,12]. In contrast, in the regions investigated in our study, we consistently showed whole-cell currents elicited by GLP-1 in RFPpositive neurons, indicative of post-synaptic GLP-1 receptors, activation of which opens a non-selective cation channel in most cells or a K þ channel in some hippocampal neurons. This discrepancy could indicate why few GLP-1R containing cell bodies are observed in the VTA and NAc in our study and that of Harasta et al. [31]. If they are expressed on axons terminals, their cells bodies may be located in distal nuclei. This would explain why microinjection of GLP-1 or Ex-4 into the VTA and the NAc [10,34,35] elicits functional effects, but few GLP-1R-expressing cell bodies are found in these regions. Alternatively, a species-difference between rat and mouse might exist, leaving mouse insensitive to GLP-1 in the VTA and NAc.

Conclusions
In this study, we have demonstrated that this mouse model provides not only the means to describe the distribution of GLP-1R expressing cells throughout the brain facilitating identification and functional characterisation in vitro, but also that it is amenable to genetic targeting of GLP-1R cells with cre-dependent viral gene expression. These properties will allow future studies of functional interference with specific subsets of GLP-1R expressing cells in vivo, as well as circuit mapping, and development of bespoke mouse strains containing knock outs of desired proteins in GLP-1R expressing cells. The difficulty of identifying neurons expressing the GLP-1R is made all the more complex by the lack of reliable antibodies. Here we have circumnavigated this problem by using a transgenic mouse, which expresses Cre recombinase under the control of the Glp1r promoter. Although this negates the need for antibodies to detect the expression of the GLP-1R, it is noteworthy that the expression of floxed fluorescent proteins requires only Glp1r promoter activity at some point during development [48], and does not necessarily reflect current GLP-1R expression. Validation has been previously carried out on this model in peripheral cells [16], but full validation of the central expressing neurons will require future work. The current study has achieved much towards addressing this question by employing both electrophysiological recordings and viral gene transfer to address the question of whether GLP-1R and cre-recombinase are both expressed in fully developed adult animals at the time of experimentation; in all instances tested here this was the case.

AUTHOR CONTRIBUTIONS
ST and FR conceived the project. FR and FMG provided the transgenic mouse strains. SCC performed the immunocytochemistry and electrophysiology. JER performed the in vivo procedures. MKH produced the AAV virus particles. ST and SCC analysed the data, wrote the manuscript, and all authors contributed to editing and provided intellectual input.