Identification and characterisation of glucagon-like peptide-1 receptor expressing cells using a new transgenic mouse model

Abstract

Glucagon-like peptide-1 (GLP-1) is an intestinal hormone with widespread actions on metabolism. Therapies based on GLP-1 are highly effective because they increase glucose-dependent insulin secretion in people with type 2 diabetes, but many reports suggest that GLP-1 has additional beneficial, or in some cases potentially dangerous, actions on other tissues, including the heart, vasculature, exocrine pancreas, liver and central nervous system. Identifying which tissues express the GLP-1 receptor (GLP1R) is critical for the development of GLP-1 based therapies. Our objective was to identify and characterise the targets of GLP-1 in mice, using a method independent of GLP1R antibodies. Using newly-generated glp1r-cre mice crossed with fluorescent reporter strains, we show that major sites of glp1r expression include pancreatic β and δ-cells, vascular smooth muscle, cardiac atrium, gastric antrum/pylorus, enteric neurones and vagal and dorsal root ganglia. In the central nervous sytem, glp1r-fluorescent cells were abundant in the area postrema, arcuate nucleus, paraventricular nucleus and ventromedial hypothalamus. Sporadic glp1r-fluorescent cells were found in pancreatic ducts. No glp1r-fluorescence was observed in ventricular cardiomyocytes. Glp1r-positive enteric and vagal neurons were activated by GLP-1, and may contribute to intestinal and central responses to locally-released GLP-1, such as regulation of intestinal secretomotor activity and appetite.

Introduction

GLP-1 is one of several metabolically active peptides that are released from enteroendocrine cells in the gut epithelium following food intake. Together with glucose dependent insulinotropic polypeptide (GIP), it accounts for up to 70% of meal-stimulated insulin secretion in a phenomenon known as the incretin effect. GLP-1 based therapies were developed originally on the basis that they enhance insulin secretion in people with type 2 diabetes, but are now known also to lower glucagon levels, slow gastric emptying and reduce appetite (1). More surprising are reports that GLP-1 improves memory and learning (2), is cardioprotective during myocardial ischaemia (3), promotes hepatocyte function (4) and increases pancreatic exocrine hyperplasia (5). A critical obstacle to distinguishing direct from indirect effects of GLP-1, however, is our incomplete understanding of which cell types express receptors for GLP-1 (6).

The GLP-1 receptor (GLP1R), was cloned originally from pancreatic β-cells where it is coupled to cAMP production and enhanced glucose-dependent insulin release (7). It is a high affinity receptor for “active” GLP-1, a term encompassing GLP-1(7-36)amide and GLP-1(7-37) that are released post-prandially from enteroendocrine L-cells in the epithelium of the small and large intestine (8). Although believed to target β-cells via the bloodstream, GLP-1 is rapidly cleaved and “inactivated” by dipeptidyl peptidase 4 (DPP4) (9) once it enters the circulation. This has led to idea that receptors located close to L-cells may act as local sensors of endogenous GLP-1 before it is inactivated. One proposed signalling route involves GLP1R located on branches of the afferent vagus nerve innervating the portal vein (10). Some literature suggests that DPP4-cleaved GLP-1 may also be a weak partial agonist or antagonist of GLP1R (11), eliciting physiological responses such as vasodilation (12). Elucidating the actions of GLP-1 and its metabolites is critical for our understanding of post-prandial physiology and the pharmacology of clinically approved GLP-1 mimetics and DPP4 inhibitors.

Identifying which tissues and cell types express GLP1R faces the obstacle that many existing antibodies to GLP1R lack specificity (6). The objective of this study was to identify and characterise targets of GLP-1 by a method independent of the use of antibodies, and to investigate whether glp1r is expressed in cells located close to enteroendocrine cells. For this purpose, we generated a transgenic mouse model in which the glp1r promoter drives Cre-recombinase. By cloning cre into the coding sequence of glp1r in a bacterial artificial chromosome (BAC), we aimed to retain the receptor promoter together with long 5′ and 3′ sequences, and achieve cell specific expression mirroring that of native glp1r. When cross-bred with floxed reporter strains, these mice produce offspring expressing fluorescent markers in target tissues that currently, or during development, express glp1r.

Experimental procedures

Animals

Animal procedures were approved by the local ethics committee and conformed to United Kingdom Home Office regulations. Mice were killed by cervical dislocation and the various tissue types were collected into ice-cold Leibovitz-15 (L-15) medium (Sigma, UK, all chemicals were supplied by the same manufacturer unless otherwise stated). Ganglia were collected directly into 4% paraformaldehyde (PFA) for immunohistochemistry.

Generation of mice

To express Cre-recombinase under the control of the glp1r promoter we replaced the sequence between the start codon in exon 1 and the stop codon in exon 13 in the murine based bacterial artificial chromosome (BAC) RP23-408N20 (Children’s Hospital Oakland Research Institute, Oakland, CA, USA) initially by a counter-selection cassette rpsL-neo (Genebridges, Heidelberg, Germany) and subsequently by the improved Cre (iCre) sequence (13) using Red/ET recombination technology (Genebridges) (Fig. 1A). Briefly, the rpsL-neo or iCre sequences were amplified by PCR adding glp1r-gene specific 3′ and 5′ sequences (see oligonucleotides in supplementary Table 1) and homologous recombination was achieved upon co-transforming the BAC containing E.coli DH10B clone with the PCR product and the plasmid pSC101-BAD-gbaA, which provides the recombination enzymes (Genebridges). Positive recombinants were isolated using appropriate antibiotic selection and characterised by PCR and restriction analysis. Identity and correct positioning of the introduced iCre sequence was confirmed by direct sequencing using oligonucleotides in supplementary Table 1. BAC-DNA for microinjection was purified using the large-construct Maxi-Prep kit (Qiagen) and dissolved at ~1-2 ng/μl in injection buffer containing (mM): 10 Tris-HCl pH 7.5, 0.1 EDTA, 100 NaCl, 0.03 spermine, 0.07 spermidine. Pronuclear injection into ova derived from C57B6/CBA F1 parents and reimplantation of embryos into pseudopregnant females was performed by the Central Biomedical Services at Cambridge University. DNA of pups was isolated from ear clips by proteinase K digestion and screened for the transgene by PCR using the following primer pairs (Supplementary Table 1): GLP1R-001/Cre001 (462 bp), Cre002/003 (537 bp) duplexed with RM41/42 (220 bp), which amplifies β-catenin sequence used as a DNA quality control. The founder strain was backcrossed for >8 generations onto a C57B6 background. The Cre-reporter transgenes used in the study, Rosa26-tdRFP or –EYFP, also on a C57B6 background, were screened for with the PCR primers tdRFPsense/anti (726 bp) and GFP002/003 (470 bp), respectively.

Fig. 1. Glp1r-cre transgenic mice label pancreatic β and δ-cells.

A. Schematic depicting the insert of the bacterial artificial chromosome (BAC) RP23-408N20, and replacement of the glp1r coding region from exons (ex) 1-13 with iCre. Lengths (in kb) of the murine genomic sequence 3′ and 5′ of the glp1r coding region in the BAC are indicated. B. Dispersed pancreatic islet cells from glp1r-cre/ROSA26-YFP mice, analysed by FACS. Green (yellow) fluorescence was measured by excitation at 488 nm and emission at 530/20 nm and is plotted against forward scatter in arbitrary units (AU). C. Cells with high red fluorescence from glp1r-cre/ROSA26-tdRFP mice were FACS separated into populations with low and high side scatter (SSC). These were analysed by qRT-PCR for expression of hormones (D), or glp1r (E). n=3 each, *** p<0.001 by Student’s t-test. F,G. Fixed pancreatic slices were co-immunostained RFP or YFP (representing glp1r-fluorescence) together with glucagon and insulin (F) or somatostatin (G). Nuclei are visualised with Hoechst stain. In F, a cell triple positive for glucagon, insulin and glp1r-fluorescence is marked by the asterisk.

Flow cytometry

Pancreatic islets were isolated and dispersed as described previously (8). Cell suspensions were separated by fluorescence assisted cell sorting (FACS) using a MoFlo Beckman Coulter Cytomation sorter (Coulter Corp., Hialeah, FL) or analysed using a BD LSRFortessa analyser (BD Biosciences, San Jose, CA, USA). Side scatter, forward scatter, and pulse-width gates were used to exclude debris and aggregates. Islet populations were sorted into RNAse-free collection buffer for mRNA analysis, or into protein lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 1 % IGEPAL-CA 630, 0.5 % deoxycholic acid and 1 tablet of complete EDTA-free protease inhibitor cocktail (Roche)) for measurement of GLP-1 content by MSD total and active GLP-1 assays (Mesoscale Discovery, Rockville, MD, USA). FACS analysed samples were processed using FlowJo 7.6 software.

Quantitative RT-PCR

RNA from homogenised tissues was extracted using TRI Reagent or RNeasy Micro Kit (Qiagen, UK). RNA was treated with DNAse I (Ambion), and reverse transcribed according to standard protocols. Quantitative RT-PCR was performed as described previously (8) using primers/probes as indicated in Supplementary Table 2. Expression of the test gene was measured relative to that of actb measured in parallel in the same sample using the ΔCt method. Data are presented as 2^ΔCt, with the upper SEM calculated from the SEM of the ΔCt data.

Immunohistochemistry

Tissues were fixed in 4% PFA, dehydrated in 15% and 30% sucrose and frozen in OCT embedding media (CellPath, UK). Cryostat-cut sections (6-10 μm) were mounted directly onto poly-lysine covered glass slides (VWR, Belgium). Slides were incubated for 1 h in blocking solution containing 10% goat or donkey serum, and overnight in phosphate buffered saline (PBS)/0.05% Triton X-100/10% serum with primary antisera of interest (Supplementary Table 3). Sections were washed with PBS, and incubated with appropriate secondary antisera (donkey or goat AlexaFluors 488, 546, 555 or 633; Invitrogen) diluted 1/300. Control sections stained with secondary antisera alone. Sections were mounted with Prolong Gold (Life Technologies) prior to confocal microscopy (Zeiss LSM 510, Carl Zeiss, UK).

Immunofluorescence of brain sections

Mice under ketamine and medetomidine anaesthesia were perfused transcardially with 4% formaldehyde in PBS, pH 7.4. Brains were removed, post-fixed overnight in the same fixative and cryoprotected in 30% sucrose. Coronal sections were cut to 30μm using a cryostat and washed in 0.1M phosphate buffer (PB; pH 7.4). After incubation in blocking buffer containing 10% sheep serum, 0.1% Triton X-100 diluted in 0.1M PB for 30 minutes at room temperature, sections were transferred to 1:1000 anti-DsRed (Clontech #632496) in blocking buffer and incubated overnight on a shaker at 4°C. Sections were washed 3 times in PB and incubated with 1:500 Cy3-conjugated anti-rabbit secondary antibody (Sigma #C2306) in blocking buffer for 2 hours on a shaker at room temperature. Sections were washed in PB, mounted onto slides and air dried before being coverslipped and viewed using epifluorescence (Nikon Eclipse 80; Kingston upon Thames, UK). Photomicrographs were taken with a Micropublisher 3.3 RTV camera and QCapture Pro software (Qimaging Inc., Surrey, BC, Canada).

Nodose and Enteric Ganglia culture

Nodose ganglia were transferred into Dullbecco’s modified Eagle’s medium (DMEM) containing 1 mg/ml collagenase type 1 and trypsin (Worthington, USA). For enteric ganglia culture, the small intestinal muscle layer was minced and digested in DMEM containing 1.5mg/ml collagenase I and 1.3mg/ml trypsin. All ganglia were digested for 1 h at 37°C and dissociated by trituration. Cells were plated on poly-lysine coated dishes, and after attachment, dishes were flooded with DMEM supplemented with 10 % (v/v) FBS, 100 units/ml penicillin, 0.1 mg/ml streptomycin and nerve growth factor (50 ng/ml). Calcium imaging and electrophysiology were performed within 48 h of dissociation.

Calcium imaging

Nodose ganglia cultures were loaded in 2 μM fura-2 AM (Invitrogen, UK) for 30 min in standard saline solution containing (mM) 4.5 KCl, 138 NaCl, 4.2 NaHCO₃, 1.2 NaH₂PO₄, 2.6 CaCl₂, 1.2 MgCl₂, 5 glucose and 10 HEPES (pH 7.4, NaOH). Imaging was performed as described previously (8). Fura2 was excited at 340 and 380 nm, and RFP at 555 nm. Fura2 fluorescence measurements were taken every 3 s, background corrected and expressed as the 340/380 nm ratio. Average fluorescence ratios were determined over 20 s and responses were expressed as the maximum ratio achieved during stimulation divided by the mean of the ratios measured before and after washout. Cells were included in the analysis if they responded to 30 mM KCl.

Nested PCR

After digestion and trituration, small enteric ganglia were picked using a glass capillary pipette filled with RNAse-free PBS. The pipette tip was crushed into RNAse-free collection buffer (4 μl First Strand Buffer (Invitrogen, UK), 4 μl RNAse-free water, 1 μl RNAseOUT). Samples were reverse transcribed using SuperScriptIII (Invitrogen, UK) using standard protocols. Actb and glp1r were amplified by multiplex nested PCR using 25 cycles each and primers given in Supplementary Table 4.

Electrophysiology

Experiments were performed on identified RFP positive cells in 12-48 hour old cultures containing partially dissociated enteric ganglia from glp1r-Cre/ROSA26-tdRFP mice. Microelectrodes were pulled from borosilicate glass (GC150 TF-15, Harvard Apparatus, UK) and the tips coated with refined yellow beeswax. Pipette solution contained (mM) 107 KCl, 1 CaCl₂, 7 MgCl₂, 11 EGTA, 10 HEPES and 5 K₂ATP (pH 7.2 KOH). Membrane potential was recorded using an Axopatch 200B and pCLAMP software (Molecular Devices, UK), in standard whole cell patch configuration at room temperature. Cells were perfused with standard saline solution (see above), to which GLP-1 was added as indicated.

Data analysis

Data are presented as mean ± SEM. Statistical analysis was performed using Microsoft Excel and GraphPad Prism 5.0. Data that were normally distributed were analysed by Student’s t-test or ANOVA with post-hoc Bonferroni test, as indicated in the figure legends. Calcium imaging and electrophysiology data were analysed with the non parametric tests, Mann Whitney or Wilcoxon Rank Sign test, as indicated, with a threshold for significance of p<0.05.

Results

Two glp1r-cre founder strains containing the transgenic constructs depicted in Fig. 1A were generated and crossed with a ROSA26-tdRFP reporter (14). Tissues/cells in which the fluorescent reporter is visible either directly or by immunostaining are described here as exhibiting glp1r-fluorescence. One founder strain showed sparse glp1r-fluorescence in pancreatic islets and was discontinued. The other showed marked islet fluorescence consistent with the known expression of glp1r in β-cells, and was used for all analyses. It was crossed with a ROSA26-YFP reporter strain to generate mice producing YFP in glp1r expressing cells, and with GLU-Venus (8) to enable the co-labelling of pancreatic α-cells and enteroendocrine L-cells that make glucagon/GLP-1.

Glp1r in the pancreas

Fluorescence assisted cytometric (FACS) analysis of pancreatic islet cell suspensions revealed a glp1r-fluorescent population (Fig. 1B) that could be divided into two groups with distinct side scatter (SSC) (Fig. 1C). These were separated by FACS, analysed for expression of ins, gcg, sst and ppy by qRT-PCR, and found to correspond to β-cells (high SSC) and δ-cells (low SSC) (Fig. 1D). Expression of glp1r in both populations was confirmed by qRT-PCR (Fig. 1E), and was ~10-fold lower in δ than β-cells (p=0.006). Colocalisation of glp1r-fluorescence with insulin or somatostatin immunostaining was also evident in pancreatic slices (Fig. 1F,G).

To investigate expression of glp1r in α-cells, we examined pancreata from glp1r-Cre/ROSA26-tdRFP/GLU-Venus mice, in which α-cells are additionally labelled with Venus. By FACS analysis, only 9.5 ± 1.4 % of α-cells (n=11) were glp1r-fluorescent, consistent with the low expression of glp1r in the total α-cell population (Fig. 2A), and our previous detection of GLP1R immunofluorescence in a small proportion of α-cells (15). The population of α-cells exhibiting glp1r-fluorescence was found by qRT-PCR to express gcg, as predicted, but also had significantly more ins than their non-glp1r-fluorescent α-cell counterparts (Fig. 2B). Rare cells immunopositive for insulin, glucagon and glp1r-RFP were correspondingly identified in tissue slices (see cell marked with asterisk in Fig. 1F). Both single and double fluorescent α-cells, however, contained very little active GLP-1 (Fig. 2C): ~1.5 pg/1000 α-cells compared with ~650 pg/1000 colonic L-cells (8).

Fig. 2. Glp1r in pancreatic α-cells.

A. Islet cells from glp1r-cre/ROSA26-tdRFP/GLU-Venus mice, FACS analysed to show red fluorescence (representing expression of glp1r) or green fluorescence (representing expression of gcg). B. Single or double positive cells were purified from quadrants Q1-Q3 marked in A, and analysed for expression of ins and gcg by qRT-PCR. *** p<0.001 by ANOVA with post hoc Bonferroni test, n=5-7. C. Active GLP-1 hormone concentrations in cells collected from different quadrants of A as indicated. Q1-4 all islet cells, Q1 + high SSC β-cells, Q2 + Q3 all α-cells, Q2 double fluorescent cells. (mean and SEM of n=4 FACS sorts, each of pooled islets from 1-2 mice). D-F. Pancreatic sections from glp1r-cre/ROSA26-YFP mice co-immunostained for α-smooth muscle actin (αSMA) or cytokeratin (CK) together with YFP. The marked area in E is shown at higher magnification in F.

In pancreatic slices, glp1r-fluorescence was also evident in non-islet structures. Predominant labelling was observed in blood vessels, as similarly observed in other tissues (Fig. 2D and see below). Co-staining with an antibody against cytokeratin revealed a few scattered glp1r-fluorescent cells in pancreatic ducts (Fig. 2E,F).

Glp1r in the cardiovascular system

Prominent glp1r-fluorescent labelling was observed in the aorta (Fig. 3A), and in arteries and arterioles in a range of tissues including the heart, kidney, pancreas and intestine (Fig. 3). Glp1r mRNA was corresponding detected in aorta by qRT-PCR (Fig. 3B). In the kidney (Fig. 3C), glp1r-fluorescence was found in arterioles and colocalised with smooth muscle α-actin (αSMA) and the pericyte marker NG2, suggesting the labelling of smooth muscle cells. Overlap of glp1r-fluorescence with immuno-staining for renin was also observed. In the intestine, many small blood vessels were glp1r-fluorescent and exhibited colocalisation with αSMA and NG2 (Fig. 3D,E). In tissue preparations where small veins could be seen lying adjacent to arterioles, glp1r-fluorescence appeared restricted to the arterial circulation (Fig. 3F).

Fig. 3. Glp1r in the cardiovascular system.

Sections of aorta (A), kidney (C), intestine (D,E), and cardiac ventricle (G) or atrium (H) from glp1r-cre/ROSA26-YFP mice were co-immunostained for YFP (representing glp1r-fluorescence) and α-smooth muscle actin (αSMA), NG2 or renin, as indicated. In kidney sections, G depicts the position of glomeruli. B. qRT-PCR analysis of glp1r expression in aorta, atrium and ventricle (geometric mean + 1SEM, of n=5-6 each). F. Blood vessels in intestine from a glp1r-cre/ROSA26-tdRFP mouse: upper panel light microscopy (A arteriole, V venule), lower panel tdRFP fluorescence.

Ventricular myocardium was devoid of glp1r-fluorescence with the exception of vascular structures immuno-positive for αSMA (Fig. 3G). In line with a recent report (16) that GLP-1 stimulates secretion of atrial natriuretic factor (ANF), scattered glp1r-fluorescent cells were found throughout the atrial myocardium (Fig. 3H). Expression of glp1r mRNA mirrored these findings, being ~6-fold higher in atrial than ventricular extracts (Fig. 3B).

Glp1r in the central nervous system (CNS) and afferent neuronal ganglia

Sections through the hypothalamus and brainstem of glp1r-cre/ROSA26-tdRFP mice revealed RFP positive cells in the area postrema, arcuate nucleus, ventromedial hypothalamus and paraventricular nucleus (Fig. 4A), consistent with previous reports (17). Glp1r-fluorescent cell bodies and glp1r mRNA were found in ganglia of the afferent vagus nerve (nodose ganglia, NG, Fig. 4B) and dorsal root ganglia (DRG, Fig. 4C), although by qRT-PCR we detected less glp1r mRNA in DRG than in NG (Fig. 4D). Consistent with the previous detection of calcitonin gene related peptide (CGRP) in extrinsic sensory neurons innervating the GI tract, a small population of glp1r-fluorescent neurons in DRGs were immunopositive for CGRP by immunostaining (Fig. 4C). In primary NG cultures, GLP-1 did not directly elevate intracellular Ca²⁺ concentrations but enhanced Ca²⁺ responses to ATP in glp1r-fluorescent cells, without affecting ATP responses in non-fluorescent cells (Fig. 4E,F). CCK, leptin and serotonin, by contrast, triggered larger Ca²⁺ responses in non-fluorescent than glp1r-fluorescent neurons (Fig. 4G).

Fig. 4. Glp1r in the brain and afferent neuronal ganglia.

A. Coronal brain sections from glp1r-cre/ROSA26-tdRFP mouse, immunostained for RFP. In caudal brainstem (top), a high density of fluorescent cells was observed in the area postrema (AP), whereas only few were evident in the nucleus of the solitary tract (NTS), and dorsal vagal nucleus (DMNX) and none were found in the hypoglossal nucleus (HN). In hypothalamus (middle and lower panels), large numbers of RFP fluorescent cells were observed in the arcuate nucleus (Arc), ventromedial hypothalamus (VMH) and paraventricular nucleus (PVN). Abbreviations: ME, median eminence; CC, central canal; 3V, 3rd ventricle. Scale bar = 150μm. B. Section of nodose ganglion (NG) from glp1r-cre/ROSA26-tdRFP mouse, showing RFP positive cells. C. Dorsal root ganglion (DRG) section from a glp1r-cre/ROSA26-tdRFP mouse, co-stained for RFP and calcitonin gene related peptide (CGRP). D. Whole NG or DRG were analysed for glp1r vs actb expression by qRT-PCR (geometric mean +1SEM of n=3 each). E. NG from glp1r-cre/ROSA26-tdRFP mice were studied by fura2 Ca²⁺ imaging in primary culture. Representative trace from an RFP positive NG neuron to ATP (0.5 μM) applied in the absence or presence of GLP-1 (1 nM), as indicated by the horizontal bars. F. Mean 340/380 nm fluorescence ratios from 13 RFP positive cells (RFP+, open bars) and 12 RFP negative cells (RFP−, filled bars), monitored as in E and normalised to baseline recorded in the absence of test agent. ** p<0.01 by Wilcoxon signed rank test G. Ca²⁺ responses of RFP+ (open bars) and RFP− (filled bars) cells to cholecystokinin (CCK, 10 nM), leptin (10 nM) or 5HT (10 μM), normalised to baseline in control solution. From left to right, columns represent data from n=18, 41, 9, 31, 14 and 35 cells, respectively. *p<0.05, *** p<0.001 by Mann Whitney test.

Glp1r in the liver and intestines

Hepatocytes were not glp1r-fluorescent, but fluorescent fibres were observed in the wall of the portal vein near the liver hilus (Fig. 5A), consistent with previous reports (10). Dense networks of glp1r-fluorescent fibres were found in the muscular and submucosal layers of the stomach in the vicinity of gastric pylorus (Fig. 5B,C), in some places exhibiting co-labelling with nNOS. Scattered glp1r-fluorescent neuronal fibres and cell bodies were identified in the small intestine and colon, but contrary to a recent report (18), the epithelial layer was largely devoid of glp1r-fluorescence. Glp1r mRNA was correspondingly detected by PCR in myenteric ganglia containing glp1r-fluorescent cells (Fig. 5D). Patch clamp recordings of small intestinal myenteric neurones in primary culture revealed that glp1r-fluorescent neurons were electrically active and that their action potential frequency was increased by GLP-1 (Fig. 5E,F). Two glp1r-fluorescent cell populations were distinguishable electrophysiologically by the action potential waveform (Fig. 5G): 14/22 (64%) of neurons were S (synaptic)-type and 8/22 (36%) were AH (after-hyperpolarising)-type.

Fig. 5. Glp1r in enteric neurons.

A. Fixed tissue section of liver from a glp1r-cre/ROSA26-tdRFP mouse, showing glp1r-fluorescence in structures surrounding a portal vein. B,C. Antral area of the stomach from a glp1r-cre/ROSA26-tdRFP/ROSA26-YFP mouse, immunostained in B for YFP (green) and αSMA (red) and in C for RFP (green) and nNOS (red). D. Nested PCR amplification of glp1r and β-actin, performed on 3 hand-picked glp1r-fluorescent enteric ganglia from glp1r-cre/ROSA26-tdRFP mice. E. Representative recording of a partially dissociated small intestinal glp1r-fluorescent enteric neuron in primary culture, monitored in whole cell current clamp, following addition of GLP-1 (10 nM). F. Mean action potential (AP) frequency of 7 cells under basal conditions and following GLP-1 application. * p<0.05 by Wilcoxon signed rank test. G. Enteric neurons, recorded as in E, were divided into AH-type (with characteristic hump on the action potential downstroke and after-hyperpolarisation) and S-type. The dashed line represents the 0 mV line. Small intestinal (H) and colonic (I) tissue sections from a glp1r-cre/ROSA26-tdRFP mouse co-stained for RFP together with nNOS or calretinin, as indicated. As direct RFP fluorescence is lost on fixation, a green secondary antibody was used to visualise RFP in the upper panel of H. J. Intestinal tissue section from a glp1r-cre/ROSA26-tdRFP/GLU-Venus mouse showing proximity of an L-cell (green) and a red glp1r-fluorescent fibre.

Immunostaining for neuronal nitric oxide synthetase (nNOS), a marker largely restricted to inhibitory motor neurones, revealed that 143/227 (63%) of glp1r-fluorescent neurones in small intestinal cultures (Fig. 5H), and 10/54 (19%) in colonic cultures (Fig. 5I), were nNOS positive. Most nNOS positive cell bodies, however, were glp1r negative (small intestine, 3523/3666, 96 %; colon 163/173, 94%). A smaller proportion of glp1r-fluorescent cells in culture stained for calretinin (Fig. 5G,H: small intestine, 7/142, 5%; colon, 10/24, 42%), CGRP or calbindin (not shown), markers typically associated with intrinsic primary afferent neurones (IPANs). Glp1r-fluorescent fibres were observed in the mucosa, in some cases relatively close to L-cells (Fig. 5J).

Discussion

Despite wide academic and commercial interest in the actions of GLP-1, attempts to identify the cellular targets of GLP-1 are hampered by the lack of specificity of antibodies to GLP1R. Our development of a new transgenic mouse model expressing cre-recombinase driven by the glp1r promoter provides an antibody-independent method for the identification and characterisation of live cells expressing glp1r, using floxed fluorescent reporter strains. The results illuminate not only which tissues exhibited glp1r-fluorescence, but also those that did not.

Establishing definitively that GLP1R protein is produced by all glp1r-fluorescent cells will be important, because our use of cre-recombinase results in a permanent activation of the fluorescent reporters even in cells that no longer express the receptor, as well as in the progeny of cells that have once expressed glp1r. Where neurones were identified, we were able to confirm expression of GLP1R protein by demonstrating functional responsiveness to added GLP-1. In other cases, such as the kidney and heart, one would ideally demonstrate co-staining of glp1r-fluorescent structures with antibodies against GLP1R, but this will remain difficult until better antibodies become available. It is possible that cells weakly expressing glp1r may generate insufficient cre-recombinase to activate the fluorescent reporters. This may explain our inability to confirm a recent report that GLP1R is located in the intestinal epithelial layer (18), but studies relying predominantly on the use of GLP1R antibodies should be interpreted with caution (6).

Throughout the body we observed glp1r-fluorescence colocalising with αSMA in large and small arteries, suggesting that vascular smooth muscle is a potential target for circulating GLP-1. Consistent with these findings, chronic treatment with GLP-1 mimetics in patients with type 2 diabetes is associated with a small reduction in blood pressure and increase in heart rate (19). The underlying physiology appears complex, however, and in rats GLP-1 was found to act as a vasodilator in the periphery, but as a vasoconstrictor in the splanchnic bed (20). Reports that DPP4-inactivated GLP-1 triggers vasodilation and that not all responses to GLP-1 are lost in mice lacking glp1r (12), have even led to the speculation that there may exist an alternative receptor responsive to GLP-1 and/or its metabolites. The vasodilatory effect of GLP-1 in mice was recently reported to be an indirect consequence of enhanced atrial natriuretic factor release (16). In support of this idea, we found that the cardiac atrium expressed particularly high levels of glp1r mRNA, and had scattered glp1r-fluorescent cells in addition to the labelled vasculature. We also detected glp1r-fluorescence in renal arterioles and renin-producing cells, consistent with reports that GLP-1 has direct effects on the kidney (21). Despite literature suggesting direct cardioprotective effects of GLP-1, however, glp1r-fluorescence was not observed in ventricular cardiomyocytes. In view of the prominent labelling of the arterial system, it seems likely that the reported effects of GLP-1 on cardiac function may be mediated indirectly via the coronary vasculature.

In pancreatic islets, glp1r expression was largely restricted to the β and δ-cell populations, with only ~10% of α-cells exhibiting glp1r-fluorescence. The finding that glp1r was expressed in δ-cells but largely excluded from α-cells supports the idea that the inhibition of glucagon secretion by GLP-1 is mediated indirectly via enhanced somatostatin release (22). The role of the ~10% of α-cells that were glp1r-fluorescent but apparently expressed ins is unclear. Although we speculated that proglucagon may be alternatively processed in α-cells producing insulin, we found no evidence for production of substantial amounts of active GLP-1 in either double-fluorescent or the total α-cell population. In the exocrine pancreas, glp1r-fluorescence was largely confined to the vasculature, with only sporadic glp1r-fluorescent cells identifiable in pancreatic ducts. Whether these sparse glp1r-positive α-cells and ductal cells are significant targets of GLP-1 mimetics will be of interest in the context of current concerns that incretin therapies may have adverse effects on the pancreas, triggering α-cell hyperplasia, pancreatitis and pancreatic cancer (23).

In the brainstem, glp1r-fluorescence was prominent in the area postrema. This is a region with a leaky blood brain barrier that has been postulated previously as an area potentially capable of responding to circulating GLP-1 concentrations (24). In the hypothalamus, we found glp1r-fluorescence in the arcuate and paraventricular nuclei, coinciding with the sites we found previously to receive projections from proglucagon-producing neurones in the nucleus of the tractus solitarius (25). These cells are therefore likely to form a component of the relay circuit linking the brainstem to the appetite-controlling networks of the hypothalamus.

The identification of glp1r-fluorescent neurons in the gastric antrum and pylorus, intestine, nodose ganglia and dorsal root ganglia may be relevant to understanding the local targets of intestinally-released GLP-1. Dense innervation of the pylorus by glp1r-fluorescent fibres is consistent with the well-recognised action of GLP-1 to inhibit gastric emptying – the so-called “ileal brake”, although further studies will be required to elucidated the identity and origin of these fibres. Our functional analysis suggests that vagal cell bodies expressing glp1r form a distinct neuronal population, responding to GLP-1 but not CCK, leptin or 5HT. That GLP-1 activates vagal afferent neurons terminating in the wall of the portal vein has been recognised for several years (10). Vagal innervation of the intestine, however, diminishes along the length of the GI tract, and inversely mirrors the frequency of L-cells which predominate in the distal gut. Our identification of glp1r-positive cell bodies in the enteric nervous system and DRGs suggests that GLP-1 may also signal to the central nervous system via local and spinal sensory nerves, potentially providing a signalling route recruited by GLP-1 released more distally. Whether the glp1r-fluorescent nerve fibres we observed in the intestinal mucosa, in some cases close to L-cells, represent the terminals of these “sensory” neurons or of enteric neurons remains to be elucidated.

Glp1r-fluorescent enteric neurons in primary culture exhibited increased firing frequency on application of GLP-1, but their functional role remains to be established. Based on the reported characteristics of guinea pig enteric neurones (26), our data suggest that the S-type neurons may correspond to the population of glp1r-fluorescent nNOS positive cells, and might thus play role in the local inhibition of circular and/or longitudinal muscle tone. The finding of AH-type electrophysiology in some cells, and of co-immunostaining of glp1r-fluorescence with calretinin, calbindin and CGRP, suggests that there is also a population of GLP-1-responsive intrinsic primary afferent neurones that may relay signals to other regions of the enteric nervous system or afferent neurons.

Glp1r-fluorescent afferent and enteric neuronal cell bodies and nerve fibres within the intestinal mucosa are of particular interest when considering the physiological consequences of food intake and bariatric surgery. Gastric bypass results in dramatically elevated GLP-1 levels, antagonism of which largely abolishes post-surgical improvements in glucose-stimulated insulin release (27). GLP-1 mimetics, however, are markedly less effective than surgery in treating type 2 diabetes. The physiological consequences of endogenous GLP-1, released locally at high levels after surgery and potentially reaching high concentrations at the site of local nerve endings, clearly deserve further attention. Current and future findings using our new mouse model promise to improve our understanding of the desirable and undesirable actions of this therapeutically fascinating hormone.