Following nutrient ingestion, glucagon-like peptide 1 (GLP-1) is secreted from intestinal L-cells and mediates anti-diabetic effects, most notably stimulating glucose-dependent insulin release from pancreatic β-cells but also inhibiting glucagon release, promoting satiety and weight reduction and potentially enhancing or preserving β-cell mass. These effects are mediated by the GLP-1 receptor (GLP-1R), which is a therapeutic target in type 2 diabetes. Although agonism at the GLP-1R has been well studied, desensitisation and resensitisation are perhaps less well explored. An understanding of these events is important, particularly in the design and use of novel receptor ligands. Here, using either HEK293 cells expressing the recombinant human GLP-1R or the pancreatic β-cell line, INS-1E with endogenous expressesion of the GLP-1R, we demonstrate GLP-1R desensitisation and subsequent resensitisation following removal of extracellular GLP-1 7-36 amide. Resensitisation is dependent on receptor internalisation, endosomal acidification and receptor recycling. Resensitisation is also regulated by endothelin-converting enzyme-1 (ECE-1) activity, most likely through proteolysis of GLP-1 in endosomes and the facilitation of GLP-1R dephosphorylation and recycling. Inhibition of ECE-1 activity also increases GLP-1-induced activation of extracellular signal-regulated kinase and generation of cAMP, suggesting processes dependent upon the lifetime of the internalised ligand–receptor complex.
Following nutrient ingestion, a number of incretin hormones are released from the gastrointestinal tract to mediate a range of effects that promote a reduction in blood glucose levels. Both glucose-dependent insulinotropic polypeptide (GIP) and glucagon-like peptide-1 (GLP-1) are important incretin hormones that have major effects on postprandial glucose levels through the stimulation of glucose-dependent insulin release from pancreatic β-cells. A range of other anti-diabetic effects have been demonstrated, particularly for GLP-1, including enhanced proinsulin transcription and insulin biosynthesis, anti-apoptotic and proliferative effects on β-cells, the inhibition of glucagon secretion from pancreatic α-cells and the reduction in gastric emptying and suppression of appetite [1–3]. As a consequence of this broad range of effects and the fact that in contrast with GIP, the action of GLP-1 is preserved in type 2 diabetes, the GLP-1 system has become an established therapeutic target in the disease. Current therapies targeting this system seek to either prevent the rapid degradation of plasma GLP-1 by the inhibition of the proteolytic enzyme, dipeptidyl-peptidase IV (DPP-IV), or alternatively target the GLP-1 receptor (GLP-1R) with injectable, degradation-resistant GLP-1 mimetic peptides. There is now also considerable interest in targeting allosteric sites on the GLP-1R to generate small-molecule, orally active GLP-1R agonists or positive allosteric modulators .
An understanding of signalling by the GLP-1R provides opportunities to exploit novel pharmacologies such as biased agonism and allosteric regulation for therapeutic purposes . Alongside this, it is also important to consider other regulatory features of the GLP-1R including desensitisation and resensitisation. In model cellular systems, the GLP-1R undergoes both homologous and heterologous desensitisation. For example, desensitisation of Ca2+ signalling has been demonstrated and related to agonist-mediated receptor phosphorylation . The phosphorylated GLP-1R subsequently internalises into intracellular compartments through the clathrin- or caveolin-1-dependent machinery, possibly dependent on the cell type [6–8]. In line with the recovery of signalling by many other G-protein-coupled receptors (GPCRs), GLP-1R-mediated Ca2+ signalling resensitises after removal of the extracellular ligand independent of de novo receptor synthesis in Chinese hamster lung fibroblast cells expressing recombinant GLP-1Rs . The currently accepted model for GPCR desensitisation implies a ligand-dependent receptor phosphorylation and subsequent arrestin recruitment, followed by receptor internalisation. For those receptors that are recycled to the plasma membrane rather than degraded within the lysosomal pathway, the acidic environment of the intracellular endosomal compartment is thought to promote dissociation of any receptor-bound ligand, allowing loss of arrestin binding, receptor dephosphorylation and recycling [9–12]. Similar to interactions between other signalling peptides and their receptors, the major circulating, active forms of GLP-1 (GLP-1 7-36 amide and GLP-1 7-37) bind to the GLP-1R with high affinity and are therefore likely to be internalised along with the receptor [6,13]. It is unclear whether endosomal acidification is sufficient (or indeed necessary) for loss of peptide binding to internalised GLP-1Rs. Furthermore, an important role for endosomal endothelin-converting enzyme-1 (ECE-1) in the processing of some internalised peptide ligands has been established [14–21]. Such processing impacts on the lifetime of the internalised ligand–receptor complex, affecting signalling and recycling events along with physiological outcomes. If such processing also regulates aspects such as the recycling and resenstitisation of the GLP-1R, this could have important implications for the design and development of alternative peptide and non-peptide GLP-1R ligands for the treatment of type 2 diabetes. The present study investigated the mechanism of GLP-1R resensitisation and examined the role of ECE-1 in regulating resensitisation and signalling in both HEK293 cells expressing human recombinant GLP-1R (HEK-GLP-1R) and the pancreatic β-cell line, INS-1E, that endogenously expresses the GLP-1R.
Tissue culture plasticware and the cover slips were from Nunc (VWR International, Lutterworth, U.K.). Geiner ELISA strip plates (96-well format) were purchased from Scientific Laboratory Supplies (Willford Industrial Estate, Nottingham, U.K.). Media, foetal bovine serum, fluo-4-acetoxymethyl ester (fluo-4-AM), Lipofectamine, RNAmaxi transfection reagent and anti-mouse Alexa fluor 568 antibody were from Invitrogen (Paisley, U.K.). GLP-1 7-36 amide was purchased from Bachem (Weil am Rhein, Germany). 125I-Exendin 9-39 amide was from PerkinElmer Life and Analytical Sciences Ltd. (Buckinghamshire, U.K.). Rhodamine–GLP-1 7-36 amide was from Phoenix Pharmaceuticals, Inc. (Burlingame, U.S.A.). Scrambled and ECE-1 short-interfering RNA (siRNA) were purchased from Dharmacon (Thermo Scientific, U.K.). Antibody against ECE-1 was from GeneTex (Irvine, U.S.A). Antibodies against phospho-ERK1/2 (extracellular signal-regulated kinase 1/2), ERK1/2 and ribosomal S6 were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA, U.S.A.). The antibody against early endosome antigen 1 (EEA1) was from BD Biosciences (Oxford, U.K.), and anti-rabbit IgG antibody was from Cell Signalling (Herts, U.K.). Inhibitors and other chemicals were from Sigma–Aldrich (Gillingham, U.K.).
Human ECE-1 isoforms a, b, c and d (ECE-1a-d) were cloned from HEK293 cells by reverse transcription PCR and subcloned into pcDNA3.1(+) at Spe1 and EcoR1 restriction sites. Constructs of ECE-1a-d with a C-terminal mCherry tag were generated by cloning ECE-1a-d (Spe1 and Kpn1) without their stop codons and then ligating into pcDNA3.1(+) containing mCherry. All constructs were confirmed by sequencing.
Transfection of siRNA
Cells were grown on 8-well strips or 24-well plates coated with poly-d-lysine [0.1% (w/v)] to ∼50% confluence. The ECE-1 siRNA consisted of four distinct siRNA duplexes targeted to knockdown all isoforms of human ECE-1 mRNA. Cells were transfected with the appropriate amount of scrambled or ECE-1 siRNA (5 pmol/well for 8-well strips and 10 pmol/well for 24-well plates) in serum-free Opti-MEM for 16 h using Lipofectamine RNAiMAX reagent (1.5 and 3 µl for 8-well strips and 24-well plates, respectively) following the manufacturer's instructions. Opti-MEM was then replaced with standard culture media. After 36–48 h, the cells were used for experiments.
Transfection of plasmid cDNA constructs
Cells in poly-d-lysine (0.1% w/v) coated 8-well strips, 24-well plates or, alternatively, 6-well plates with 25 mm coverslips (∼50% confluence) were transfected with plasmid cDNA constructs using a standard calcium phosphate method. A mixture of 250 mM Ca2+ and DNA (0.2 µg/well for 8-well strips and 0.4 µg/cm2 for the plates) were mixed with an equal volume of 1.5 mM phosphate. The mixture of Ca2+–DNA–phosphate was then added into each well. After 36–48 h incubation, the cells were used for experiments.
Intracellular Ca2+ signalling
HEK-GLP-1R cells at ∼90% confluence in ELISA strip plates (96-well format) that had been precoated with poly-d-lysine (0.1% w/v) were loaded with 2 µM fluo-4-AM in Krebs–HEPES buffer with 0.1% (w/v) bovine serum albumin [KHB-BSA; 10 mM HEPES, 4.2 mM NaHCO3, 11.7 mM d-glucose, 1.18 mM MgSO4·7H2O, 1.18 mM KH2PO4, 4.69 mM KCl, 118 mM NaCl and 1.3 mM CaCl2·2H2O, 0.1% (w/v) BSA, pH 7.4] at 37°C for 40 min. Cells were then washed and equilibrated for 5 min at 37°C in KHB-BSA. Using a microplate reader (NOVOstar; BMG LABTECH, Aylesbury, U.K.), 20 µl of KHB-BSA or ligand (prepared in KHB-BSA) was added into wells (at 200 µl/s), and the fluorescence was determined at 0.5 s intervals by excitation at 485 nm and collection of emitted light at 520 nm. The fluorescence changes were recorded as an index of intracellular [Ca2+] ([Ca2+]i). To assess receptor desensitisation and resensitisation, cells were prechallenged with buffer (control) or ligand for the required time followed by washing with standard KHB-BSA unless indicated otherwise and allowed to recover in ligand-free buffer for the indicated times at 37°C. Cells were then challenged or rechallenged with ligand and the responses were measured. Fluorescence data were calibrated to [Ca2+]i as described previously .
A radioligand-binding assay was used to measure cell-surface GLP-1R binding. To limit ligand depletion, binding of the radioligand (125I-exendin 9-39 amide) to cells at various densities and using different volumes were tested in preliminary studies. In the final assays, the proportion of total 125I-exendin 9-39 amide added that was bound was ≤20% of the total added in both HEK-GLP-1R and INS-1E cells. HEK-GLP-1R cells were used at 80% confluence in ELISA strip plates (96-well format) precoated with poly-d-lysine, along with 200 µl/well of 0.05 nM 125I-exendin 9-39 amide. INS-1E cells were used at 80% confluence in 24-well plates precoated with poly-d-lysine with 350 µl/well of 0.05 nM 125I-exendin 9-39 amide.
HEK-GLP-1R cells or INS-1E cells were treated as required. At the end of treatment, cells were washed with acidified buffer (KHB-BSA, pH 4.0) to remove ligand bound to cell-surface receptors, followed by washing with standard KHB-BSA and then incubated with 200 µl (HEK-GLP-1R) or 350 µl (INS-1E) of 0.05 nM 125I-exendin 9-39 amide either with or without 1 µM exendin 9-39 amide (to determine non-specific binding) for 16 h at 4°C. Cells were then washed twice with 200 µl (HEK-GLP-1R) or 500 µl (INS-1E) of ice-cold KHB-BSA. NaOH (0.1 M) was then added to each well (100 µl for HEK-GLP-1R or 200 µl for INS-1E). The plates were then washed with 100 µl (HEK-GLP-1R) or 200 µl (INS-1E) of 0.1 M HCl, which was added to the NaOH. Samples were counted by liquid scintillation spectrometry in 2 ml (HEK-GLP-1R) or 4 ml (INS-1E) Safefluor scintillant.
Subcellular localisation of early endosomes
Cells grown on coverslips (12 mm) precoated with poly-d-lysine were washed with phosphate-buffered saline (PBS) and fixed in 4% paraformaldehyde (10 min at room temperature) followed by incubation for 2 min at −20°C with methanol (precooled to −20°C). Cells were then blocked for 90 min with 2% BSA in PBS at room temperature. Cells were incubated with primary antibody against EEA1 (1 : 500 in PBS with 2% BSA) overnight at 4°C. Cells were then washed with PBS × 3 and incubated with anti-mouse Alexa fluor 568 antibody (1 : 500 in PBS with 2% BSA) for 2 h at room temperature followed by three washes with PBS. The cells were incubated in Vectashield mounting medium and kept in the fridge until confocal imaging (see below).
HEK293 cells with either stable or transient expression of EGFP- or mCherry-tagged proteins were grown in monolayers on poly-d-lysine-coated coverslips (25 mm) for 24–48 h. Cells were imaged in KHB-BSA (37°C) using an UltraVIEW confocal microscope (PerkinElmer LAS, Beaconsfield, Bucks., U.K.) with a 60× oil-immersion objective lens and a Kr/Ar laser line at either 488 nm for EGFP or 568 nm for mCherry, rhodamine and Alexa fluor 568. Emitted light was collected above 510 nm for the fluorescent emission of EGFP or above 560 nm for mCherry, rhodamine and Alexa fluor 568, and images were captured using a CCD camera. When required, cells were challenged with ligands at 37°C either on the bench or within the perfusion chamber of the microscope stage. Confocal images to assess receptor internalisation were quantified as previously described .
Immunoblotting: determination of ERK activation and ECE-1 expression
The activation of ERK was determined by immunoblotting of phospho-ERK (pERK). Monolayers of cells (70–90% confluent) cultured on poly-d-lysine-coated 24-well plates were serum-starved for 20–24 h, washed and equilibrated with KHB-BSA at 37°C. Agonist was then added and cells were incubated for the indicated times. In cases where SM-19712 was used, this was added 30 min before agonist challenge. Reactions were terminated by placing the plates on ice and rapid aspiration followed by the addition of 100 µl of Laemmli sample buffer (62.5 mM Tris–HCl, pH 6.8, 0.1% (w/v) bromophenol blue, 2% (w/v) SDS, 10% (v/v) glycerol and 50 mM DTT). Samples were collected and centrifuged (16 100g, 1 min, 4°C). Supernatants were boiled for 5 min at 100°C, centrifuged again and either used or stored at −20°C before use. Immunoblotting of ECE-1 expression was performed on solubilised monolayers of cells processed as above following the addition of Laemmli buffer.
Proteins were separated by SDS–PAGE, transferred onto polyvinylidene fluoride membranes, blocked for 1 h in 5% (w/v) skimmed milk powder in TTBS (150 mM NaCl, 50 mM Tris, pH 7.5, 0.1% Tween-20) and incubated overnight at 4°C with antibodies to ECE-1 (1 : 1000), phospho-ERK1/2 (pERK1/2; 1 : 2000), ERK1/2 (1 : 2000) or ribosomal S6 (1 : 20 000) in 5% (w/v) BSA in TTBS. Blots were then washed (3 × 8 min) in TTBS and incubated for 1 h with anti-rabbit IgG antibody (1 : 3000 in blocking buffer) or anti-mouse IgG antibody (1 : 1000 in blocking buffer) as appropriate. After washing in TTBS (3 × 8 min), blots were exposed to enhanced chemiluminescence detection reagents (Uptima-Interchim, Montlucon, France) according to the manufacturer's guidelines and bands were visualised using Kodak Medical X-ray film (Wolf Laboratories Ltd, Pocklington, U.K.). The intensities of the immunoblot bands were determined using ImageJ. Equivalent loading across lanes was assessed by comparison of ribosomal S6 band intensities. Where direct comparisons were to be made (e.g. between cells in the presence and absence of inhibitor), experiments were performed alongside each other, including exposure of the immunoblot to the same piece of film for the same time.
Generation and measurement of cAMP
The generation of intracellular cAMP was performed at 37°C on monolayers of confluent cells in 24-well plates. Where SM19712 was used (10 µM), cells were preincubated for 30 min and the inhibitor was then included throughout the experiment. Reactions were terminated at the appropriate point by aspiration of the KHB-BSA and the addition of 400 µl of 0.5 M ice-cold trichloroacetic acid. Following incubation on ice (20 min), 500 µl of a freshly prepared oil mixture [1 : 1 (v/v) tri-n-octyl-amine and 1,1,2-trichlorotrifluoroethane] and 50 µl of 10 mM EDTA (pH 7.0) were added. Samples were vortexed and left at room temperature for 12 min. After further vortexing, samples were microfuged (16 100g, 4 min) and 200 µl of the upper aqueous phase was transferred to 50 µl of 60 mM NaHCO3, which was used for the subsequent determination of cAMP immediately or following storage at 4°C for up to 7 days. The concentration of cAMP in samples was determined by a competitive radioreceptor assay and calculated by comparison to a standard curve exactly as previously described .
Where required, data were analysed using Prism 6.0 (GraphPad Software, Inc., San Diego, CA, U.S.A.). Concentration–response curves were analysed using a four parameter non-linear regression analysis. Data are representative of n ≥ 3 or are presented as mean ± SEM. Statistical analysis was performed by unpaired Student's t-test (two-tailed) for comparisons between two datasets. Where more than two datasets were compared, analysis was by either one-way or two-way analysis of variance (ANOVA) and, where P < 0.05, with Bonferroni's post-test for multiple comparisons. Statistical significance was accepted at P < 0.05.
Stimulation with GLP-1 7-36 amide evoked desensitisation of GLP-1R-mediated intracellular Ca2+ signalling in HEK-GLP-1R cells
In fluo-4-loaded cells, GLP-1 7-36 amide produced a concentration-dependent increase in [Ca2+]i with higher concentrations evoking a greater and more rapid increase in [Ca2+]i compared with lower concentrations (Figure 1A). The maximum increases in [Ca2+]i were determined and used to generate concentration–response curves that showed a pEC50 of 9.63 ± 0.24 (Figure 1B). To explore the desensitisation of Ca2+ signalling, cells were challenged (1–30 min) with 10 nM GLP-1 7-36 amide (a concentration that evoked a maximal Ca2+ response in naive cells) followed by washing with acidified buffer (pH 4.0, 40 s) and a 5 min recovery period in standard KHB-BSA (for protocol, see Figure 1C). Acidified buffer was used for the wash to remove cell-surface-bound ligand and enable desensitisation to be examined initially in the absence of continued receptor-bound ligand. In preliminary studies comparing buffers with a variety of pH values with washes for different times, this protocol was demonstrated to result in a maximum response in cells that had been pre-treated with GLP-1 7-36 amide and did not affect the response in agonist-naive cells (data not shown). Pre-treatment markedly reduced the Ca2+ response to the subsequent rechallenge with 10 nM GLP-1 7-36 amide (Figure 1D,E). The extent of reduction in signal (i.e. desensitisation) was dependent upon the period of pre-exposure to GLP-1 7-36 amide and was maximum (within a 1–30 min window) following 10 min of agonist pre-exposure (Figure 1E).
Desensitisation of GLP-1 7-36 amide-mediated increases in [Ca2+]i.
Resensitisation of the GLP-1R following ligand removal is independent of protein synthesis but dependent on both receptor internalisation and endosomal acidification
To investigate receptor resensitisation, experiments were performed in which the cells were washed with standard (pH 7.4) KHB-BSA following initial challenge with GLP-1 7-36 amide (10 min, 10 nM) and allowed to recover for between 5 min and 3 h before rechallenge with GLP-1 7-36 amide (10 nM) (for protocol; see Figure 2A). The wash within this protocol was designed to remove free extracellular ligand, but may not fully remove receptor-bound ligand. This will more closely mimic physiological circumstances where ligands are released in a pulsatile manner and then subject to removal by, for example, dilution and proteolytic degradation. The protocol was designed to enable investigation of the consequence of the presence of receptor-bound ligand on the resensitisation process. After desensitisation that was induced by pre-treatment with GLP-1 7-36 amide, subsequent washing with KHB-BSA resulted in a time-dependent recovery of GLP-1R-mediated Ca2+ signalling over a 3 h period (Figure 2B). The response following 5 min recovery was 11 ± 1% (n = 3) of the response in naive (control) cells. This compares to a response of 38 ± 2% (n = 3) of control 5 min following ligand removal with the acid wash (Figure 1E).
GLP-1R desensitisation and resensitisation.
The current model for GPCR trafficking indicates that following agonist-mediated activation, the phosphorylation of receptor residues within intracellular domains, particularly the C-terminal and third intracellular loop, results in the recruitment of arrestin. This leads to the assembly of protein complexes that promote receptor internalisation within endocytotic vesicles . Here, the assessment of GLP-1R binding in intact cells indicated a loss of cell-surface receptors following challenge with 10 nM GLP-1 7-36 amide (Figure 3A). Given that Ca2+ signalling represents a potentially amplified response with spare receptors and that receptor phosphorylation leading to internalisation often depends directly on receptor occupancy , we also examined the impact of a higher concentration of GLP-1 7-36 amide on cell-surface expression. Indeed, a higher concentration of GLP-1 7-36 amide (100 nM) evoked a more pronounced loss of cell-surface binding than 10 nM (Figure 3A). The loss of plasma membrane binding induced by 100 nM GLP-1 7-36 amide recovered over a 90 min period following removal of free extracellular ligand (e.g. Figure 3B). To detect potential mechanisms of the loss and recovery of plasma membrane GLP-1Rs, a range of inhibitors were used at concentrations effective in HEK293 cells [27–29]. This recovery of cell-surface GLP-1R binding was not affected by inhibition of de novo protein synthesis by cycloheximide (17.5 µM) (Figure 3B), but was significantly reduced by inhibition of endosomal acidification with monensin (50 µM), although notably, monensin reduced cell-surface binding in naive (unstimulated) cells (Figure 3C). We have previously demonstrated that GLP-1 7-36 amide causes internalisation of a C-terminally EGFP-tagged GLP-1R (GLP-1R-EGFP) in HEK293 cells . Here, we demonstrate that dynasore blocks GLP-1 7-36 amide-mediated internalisation of the GLP-1R-EGFP (Figure 3D).
Cell-surface GLP-1R expression.
Treatment with cycloheximide did not affect resensitisation of the GLP-1R-mediated Ca2+ response (Figure 4). However, inhibition of either endosomal acidification with monensin or receptor internalisation with dynasore significantly reduced resensitisation without affecting the Ca2+ responses of naive cells (Figure 4).
GLP-1R resensitisation is independent of de novo protein synthesis, but dependent on receptor internalisation and endosomal acidification.
Receptor and ligand co-internalisation
The data above demonstrate a loss of cell-surface binding and ligand-dependent internalisation of the GLP-1R. To confirm the co-internalisation of ligand and receptor, cells expressing GLP-1R-EGFP were challenged with rhodamine-labeled GLP-1 7-36 amide (Rho-GLP-1 7-36 amide, 100 nM). Despite a lower binding affinity to cell-surface GLP-1Rs compared with non-labelled GLP-1 7-36 amide (Figure 5C) and a reduced potency in cAMP generation (Figure 5D), Rho-GLP-1 7-36 amide was predominantly intracellular after 60 min stimulation with some co-localisation with the GLP-1R-EGFP (Figure 5Ai–iv). These data indicate internalisation of both receptor and ligand. Co-incubation of Rho-GLP-1 7-36 amide with non-labelled GLP-1 7-36 amide blocked the appearance of both cell-surface and intracellular rhodamine fluorescence thereby demonstrating the specificity of this interaction (Figure 5B). Internalisation of both GLP-1R-EGFP and Rho-GLP-1 7-36 amide was blocked by the dynamin inhibitor, dynasore (Figure 5Av–viii). GLP-1R-EGFP co-localised with EEA1 under basal (non-stimulated) conditions (Figure 5Eiii). Stimulation with 100 nM GLP-1 7-36 amide resulted in the internalisation of GLP-1R-EGFP and increased co-localisation with EEA1 (Figure 5Dvi). Taken together, these data indicate the co-localisation of the GLP-1R and GLP-1 7-36 amide in an early endosomal compartment.
GLP-1R-EGFP internalises with rhodamine-labelled GLP-1 7-36 amide and is associated with early endosomes.
ECE-1 activity regulates resensitisation of GLP-1R-mediated signalling following desensitisation with GLP-1 7-36 amide but not exendin-4
As GLP-1 7-36 amide internalises with the receptor and ECE-1 has been shown to degrade a number of peptide ligands within the endosomal compartment to regulate resensitisation and signalling [14–21], we examined the role of ECE-1 activity in regulating GLP-1R resensitisation. Pre-incubation of HEK-GLP-1R cells with the selective ECE-1 inhibitor, SM19712 at 10 µM , before the period of GLP-1 7-36 amide pre-treatment (10 nM, 10 min), markedly inhibited the Ca2+ response to rechallenge after a 90 min period of recovery (Figure 6A,C). To assess whether such inhibition could arise from the protection of extracellular, free peptide during the initial challenge (and the potential, for example, of enhanced desensitisation), in separate experiments, SM19712 was added only after removal of the free GLP-1 7-36 amide used for the pre-treatment. Under these circumstances, SM19712 also inhibited recovery of the Ca2+ response following an initial challenge (Figure 6B,C), highlighting that protection of the degradation of extracellular peptide ligand does not fully account for the effects. SM19712 did not affect the response of naive cells to GLP-1 7-36 amide (Figure 6A–C). Treatment with SM19712 also modestly inhibited the recovery in GLP-1R binding at the plasma membrane following the challenge of cells with GLP-1 7-36 amide (100 nM, 10 min), but had no influence on receptor binding in naive cells (without GLP-1 7-36 amide treatment) (Figure 6D).
The inhibition of ECE-1 blocks resensitisation of GLP-1 7-36 amide-mediated Ca2+ signalling.
The GLP-1R agonist, exendin-4, also produced a concentration-dependent increase in [Ca2+]i in HEK-GLP-1R cells with a pEC50 of 10.41 ± 0.03 (Figure 7A,B). Similar to the observations with GLP-1 7-36 amide, pre-treatment with 1 nM (Emax) of exendin-4 for 10 min followed by removal of extracellular ligand and a 5 min recovery period resulted in a markedly reduced Ca2+ response to the rechallenge with 1 nM exendin-4 compared with that in naive cells (i.e. without pre-treatment) (Figure 7C,D). After a 6 h period of recovery, the response to rechallenge had not fully recovered (Figure 7C,D) in contrast with the situation when cells had been desensitised with GLP-1 7-36 amide where recovery was essentially complete after 3 h (Figure 2B). Treatment with SM19712 did not influence the recovery of the Ca2+ response using exendin-4 as the agonist (Figure 7C,D), which is also in contrast with the inhibition of recovery when GLP-1 7-36 was used (Figure 6). The GLP-1R small-molecule ago-allosteric compound, compound 2  (1 μM, 10 min), desensitised subsequent responses to GLP-1 7-36 amide that fully recovered (≤3 h) but were unaffected by pre-treatment with SM19712 (Figure 7E).
The inhibition of ECE-1 does not affect resensitisation of exendin-4- or compound 2-mediated Ca2+ signalling.
Transfection of cells with siRNA against ECE-1 markedly reduced the expression of ECE-1 as determined by immunoblotting (Figure 8A). Although transfection of cells with siRNA against ECE-1 did not affect the initial Ca2+ response to 10 nM GLP-1 7-36 amide, knockdown inhibited resensitisation as shown by the inhibition of the response to a rechallenge with GLP-1 7-36 amide after 90 min recovery in cells prestimulated with GLP-1 7-36 amide (10 nM, 10 min) (Figure 8B). Transfection of cells with scrambled siRNA (control) had no effect on either the expression of ECE-1 or GLP-1 7-36 amide-mediated Ca2+ signalling (Figure 8A,B).
Knockdown of ECE-1 inhibited GLP-1R resensitisation.
In experiments to assess the consequences of overexpression of ECE-1, HEK-GLP-1R cells were transfected with either vector alone (control) or plasmids containing the cDNA of one of the four ECE-1 isoforms. The expression of the ECE-1 isoforms after transfection was assessed by immunoblotting with an ECE-1 antibody, although this was unable to distinguish between the different isoforms (Figure 9A). The overexpression of individual isoforms of ECE-1 did not affect Ca2+ responses upon an initial exposure to GLP-1 7-36 amide (Figure 9B–E). However, overexpression of any of the isoforms significantly enhanced the Ca2+ response to rechallenge with 10 nM GLP-1 7-36 amide after 60 min recovery (Figure 9B–E). A 60 min recovery period was selected to provide a suitable window to detect any promotion of resensitisation. In each case, this enhanced recovery was sensitive to SM19712, which abolished the enhanced resensitisation giving levels of recovery that were numerically but not significantly lower than that seen in untransfected cells (Figure 9B–E).
Overexpression of ECE-1 isoforms enhance GLP-1R resensitisation.
ECE-1-mCherry isoforms co-localize with the GLP-1R-EGFP
To examine the distribution of ECE-1 and, in particular, the possible co-localisation with the GLP-1R, C-terminal mCherry-tagged versions of each ECE-1 isoform were generated. Overexpression of these ECE-1-mCherry isoforms enhanced GLP-1R-mediated resensitisation (data not shown) in a manner consistent with that of the untagged versions (see Figure 9). HEK-GLP-1R-EGFP cells were transfected with the individual ECE-1-mCherry isoforms and live cells imaged by confocal microscopy. Each of the ECE-1-mCherry isoforms displayed an intracellular localisation with some evidence of a plasma membrane localisation, particularly for ECE-1c-mCherry and ECE-1d-mCherry (Figure 10; middle panels). Consistent with this plasma membrane localisation, there was evidence of some co-localisation of both ECE-1c-mCherry and ECE-1d-mCherry with the GLP-1R-EGFP at the plasma membrane in unstimulated cells (Figure 10C,D; right, upper panels). Challenge of HEK-GLP-1R-EGFP cells with GLP-1 7-36 amide (100 nM, 60 min) resulted in the loss of plasma membrane GLP-1R-EGFP fluorescence and an increase in intracellular fluorescence (Figure 10A–D; left panels). In addition, this challenge enhanced the extent of co-localisation of GLP-1R-EGFP and each of the ECE-1-mCherry isoforms (Figure 10A–D, right panels). There was some suggestion that stimulation with GLP-1 7-36 amide enhanced the cytoplasmic localisation of the ECE-1 isoforms, particularly ECE-1c-mCherry (Figure 10C), although this was difficult to establish given the considerable cytoplasmic localisation in unstimulated cells (Figure 10A–D).
GLP-1R co-localizes with ECE-1-mCherry isoforms.
ECE-1 activity regulates both the resensitisation of cAMP production and the recovery of cell-surface GLP-1R binding in INS-1E cells
Pre-treatment with 100 nM GLP-1 7-36 amide for 10 min [in the absence of the phosphodiesterase inhibitor, 3-isobutyl-1-methylxanthine (IBMX)] followed by 60 min recovery induced a reduction in the magnitude (Emax) of the cAMP in response to GLP-1 7-36 amide rechallenge in the presence of IBMX (Figure 11A). Pre-incubation of cells with SM19712 further reduced the Emax following the 60 min recovery period (Figure 11A). There were, however, no differences in agonist potency (pEC50 values) between cells that had no pre-treatment and those that had been pre-treated in the absence or presence of SM19712 (Figure 11A). Radioligand binding demonstrated that challenge with 100 nM GLP-1 7-36 amide for 10 min followed by 5 min recovery evoked ∼50% reduction in GLP-1R binding at the cell surface that then recovered further during the recovery period (Figure 11B). This recovery was inhibited to a small extent by inhibition of ECE-1 with SM10712 (Figure 11B).
ECE-1 activity regulates the resensitisation of GLP-1R-mediated cAMP generation and recovery of cell-surface GLP-1R binding in INS-1E cells.
ECE-1 activity regulates GLP-1 7-36 amide-induced ERK activation and cAMP signalling in both HEK-GLP-1R cells and INS-1E cells
The challenge of either HEK-GLP-1R (Figure 12A) or INS-1E cells (Figure 12B) with GLP-1 7-36 amide (10 nM, 10 min) resulted in a marked activation of ERK as judged from the increased density of the immunoblot for pERK. Following removal of free extracellular ligand by a standard KHB-BSA wash, the level of pERK declined toward basal levels. In HEK-GLP-1R cells, levels declined over the subsequent 90 min recovery period (Figure 12A), whereas in INS-1E cells the basal level had been attained after 30 min recovery (Figure 12B). In HEK-GLP-1R cells, inhibition of ECE-1 with SM19712 resulted in a more sustained activation of ERK (Figure 12A). In INS-1E cells, treatment with SM19712 had no effect on levels of pERK at the early times following ligand removal, but resulted in an increase from 60 min following ligand removal suggesting a second phase of activation (Figure 12B).
ECE-1 activity regulates the extent and duration of signalling following removal of the extracellular ligand.
Following removal of the extracellular GLP-1 7-36 amide after 10 min stimulation of HEK-GLP-1R cells in the absence of IBMX, cAMP levels continued to increase during the next 30 min and then decreased such that at 60 min after washing levels had returned to basal (Figure 12C). Treatment with SM19712 significantly enhanced the cAMP levels at later times following removal of the extracellular ligand (Figure 12C). Similarly, in INS-1E cells, treatment with SM19712 did not influence basal levels of cAMP, but significantly enhanced the cAMP level at 60 min following removal of the extracellular ligand (Figure 12D).
GLP-1R agonists are an increasingly common treatment option for patients with type 2 diabetes. These ligands regulate blood glucose levels through mechanisms including enhanced insulin release, reduced glucagon release, increased satiety and a slowing of gastric emptying . Reduced gastric emptying is particularly important in limiting postprandial blood glucose excursion, but the impact of different agonists varies. For example, long-acting agonists show tachyphylaxis, whereas short-acting agonists provide more sustained impact [32,33]. Although different plasma half-lives may underlie such variation, differences in receptor desensitisation and resensitisation profiles may also contribute, thereby highlighting the importance of understanding GLP-1R desensitisation and resensitisation mechanisms to appropriately inform both clinical use and drug development. The present study confirms rapid GLP-1R desensitisation followed by resensitisation that is dependent upon internalisation, endosomal acidification and recycling but independent of de novo protein synthesis. Importantly, this study highlights that ECE-1 activity regulates not only GLP-1R resensitisation, but can also influence downstream signalling, particularly in experimental paradigms in which the receptor is subject to brief periods of agonist exposure.
The GLP-1R is subject to both heterologous and homologous desensitisation with a key role played by phosphorylation within the C-terminus, at least of the rat receptor [5,34]. Protein kinase C (PKC)-mediated phosphorylation of at least one serine doublet may evoke desensitisation independently of internalisation [35,36], whereas PKC-independent phosphorylation of at least three serine doublets may underlie homologous desensitisation and internalisation  and is likely to account for the observed recruitment of β-arrestin [37,38]. Following receptor phosphorylation and internalisation, receptor trafficking will determine the timing and extent of resensitisation, affecting different responses in variable ways depending on, for example, the extent of receptor reserve. A recent study  has emphasised that the nature of the GLP-1R ligand influences receptor internalisation and recycling such that agonists that maintain the plasma membrane location of the receptor, through faster agonist dissociation rates and reduced β-arrestin recruitment, generate greater long-term insulin release. Furthermore, a variety of proteins involved in GPCR trafficking have been identified that regulate GLP-1R-mediated insulin secretion . In addition to controlling plasma membrane receptor number, the regulation of trafficking is likely to influence signalling by intracellular receptors that are able to signal in ways that are either similar or different from plasma membrane receptors [20,41–48].
Our data confirm previous observations of rapid GLP-1R desensitisation with a resensitisation process that requires receptor internalisation and recycling [6,34,49]. Consistent with internalisation of radiolabelled GLP-1 , we demonstrate co-internalisation of fluorescently tagged GLP-1 7-36 amide with the GLP-1R-EGFP and show movement of both to early endosomes. The current model of GPCR trafficking suggests that receptors in this compartment either sort to lysosomes for degradation or are dephosphorylated and recycled to the plasma membrane. Following internalisation, endosomal acidification to pH ∼5.5 over a period of ∼10 min  is likely to facilitate ligand dissociation, thereby promoting dissociation of β-arrestin, receptor dephosphorylation and recycling of a resensitised receptor to the plasma membrane [10,11,51–53]. Although it is difficult to define precisely the role of receptor location and phosphorylation status in shaping signalling events, the present and previous studies demonstrate that resensitisation of GLP-1R-mediated signalling is coincident with receptor recycling and the recovery of cell-surface receptors [6,34,49]. Furthermore, this resensitisation is dependent on endosomal acidification and, consistent with a previous study on the rat receptor , independent of de novo protein synthesis.
Many GPCRs internalise with ligand bound, particularly peptidergic receptors with high affinity ligands [17,54,55], suggesting that the ligand is either recycled or processed within the cell. While peptidases play a variety of roles in the activity of many GPCRs , models of receptor recycling and resensitisation often ignore the fate of the ligand. Although some ligands, for example somatostatin-14, recycle intact to the plasma membrane , there is some sense for ligands to be processed. This would limit any signalling occurring through continued binding or rebinding to resensitised receptors, particularly when recycled to the plasma membrane and more favourable pH conditions. Indeed, a number of peptide ligands are processed within the endosomal compartment by the zinc, membrane-bound metalloprotease, ECE-1. For example, both substance P and calcitonin gene-related peptide (CGRP) are cleaved, thereby promoting both dissociation of the receptor–β-arrestin complex and receptor recycling [16,19,58]. There are four isoforms of ECE-1 generated from a single gene through alternative promoters. All contain a conserved catalytic C-terminal domain and a variable N-terminus that may be responsible for subcellular localisation, possibly due to isoform-specific phosphorylation [59,60]. At the plasma membrane, the catalytic domain is extracellular, allowing access to extracellular substrates, most notably big-endothelin, which it cleaves to endothelin. Plasma membrane invagination and the formation of endocytotic vesicles places the catalytic site of ECE-1 within the endosomal compartment, and indeed, ECE-1 isoforms co-localise with EEA1 [16,18]. The substrate specificity of ECE-1 is pH-dependent such that big-endothelin is a good substrate at neutral pH and a very poor substrate at acidic pH. However, the converse is true for both substance P and CGRP, accounting for their endosomal degradation [16,19,61]. Given that GLP-1 and the GLP-1R co-internalise and target to early endosomes and also that 125I-GLP-1 is degraded in a receptor-dependent manner within intracellular compartments , it is possible that recycling and resensitisation of the GLP-1R might also require ligand processing. The present study demonstrates that ECE-1 activity plays a role in the regulation of the GLP-1R, facilitating both the recovery of cell-surface receptor and resensitisation of signalling. These effects were independent of the protection of extracellular ligand. In addition, as resensitisation required both receptor internalisation and endosomal acidification, this suggests that ECE-1 degrades GLP-1 7-36 amide following co-internalisation with the GLP-1R and that this processing facilitates GLP-1R recycling and resensitisation. This is consistent with the endosomal localisation of both the GLP-1R and ECE-1 following receptor internalisation, which has also been observed with ECE-1 and the activated CLR and CRF1 [15,16]. It is, however, unclear whether there is an agonist-dependent internalisation of ECE-1 in the present or previous studies . Constitutive internalisation of ECE-1 may occur as a consequence of phosphorylation by intracellular kinases such as casein kinase 1, protein kinase A (PKA) and PKC [59,62,63]. Regulation by second messenger-dependent kinases provides a potential link to receptor activation and indeed, forskolin-mediated activation of PKA drives ECE-1b internalisation in CHO cells [59,60]. Interestingly, another metalloendopeptidase, endopeptidase 24.15, interacts with the C-terminal tails of the angiotensin AT1 receptor and bradykinin B2 receptor . Although there is no evidence of such direct interactions of ECE-1 and GPCRs, at least for ECE-1c and the CLR , co-localisation of the receptor and ECE-1 at the plasma membrane also has the potential to allow for co-internalisation into endosomes.
ECE-1-mediated regulation of GLP-1R resensitisation is dependent on the nature of the ligand. Thus, ECE-1 activity regulated resensitisation following GLP-1 7-36 amide (the major post-prandial circulating form of GLP-1 ) but not exendin-4 or the non-peptide allosteric ligand, compound-2. Compound 2 is a low-affinity ligand and may well be removed by the wash, including from cell-surface receptors. Resensitisation following exendin-4 was delayed compared with that following GLP-1 7-36 amide. This is consistent with slower receptor recycling following exendin 4 compared with GLP-1 . Although exendin-4 has ∼50% homology to GLP-1, it is used therapeutically in type 2 diabetes due to its insensitivity to degradation by DPP-IV [66,67]. The present data suggest that exendin-4 may not be a substrate for ECE-1, at least within the acidic endosomal compartment. Alternatively or additionally, the higher affinity of exendin-4, compared with GLP-1 7-36 amide , may limit exendin-4 dissociation and prevent degradation. Indeed, the rate of agonist-receptor dissociation within the endosomal compartment is a determinant of post-endocytic targeting to recycling or degradative pathways  and may account for the agonist-dependence of GLP-1R trafficking and the rate of receptor recycling . Furthermore, agonist residence time is an important influence on the duration of drug action and has been suggested as a driver of sustained signalling from internalised receptors . The present study emphasises that ligand degradation may also contribute to receptor trafficking and sustained signalling. Differential processing by ECE-1 has also been demonstrated for ligands of the somatostatin receptor 2A in enteric neurons where somatostatin-14 induced receptor internalisation and recycling that required ECE-1 activity, whereas somatostatin-28 and other ECE-1-resistant peptide analogues limited recycling . Thus, ligands with differential sensitivities to ECE-1 degradation may trigger alternate pathways of receptor trafficking, resulting not only in different rates of receptor resensitisation but different functional outcomes.
In addition to regulating GLP-1R recycling and resensitisation, inhibition of ECE-1 promotes sustained activity of ERK and increases cAMP levels following ligand removal. GLP-1R-dependent ERK activation depends upon a rapid, transient, PKA-mediated activation followed by a later, more sustained β-arrestin1-dependent phase [72–74]. This latter, G-protein-independent phase is likely a consequence of scaffolding and subsequent activation of MAPK pathway components as reported for a variety of GPCRs [75–77]. Importantly, β-arrestin-associated signalling pathways often display distinct cellular and physiological functions compared with the G-protein-dependent pathway [78,79]. The extended activation of ERK by the GLP-1R as a consequence of ECE-1 inhibition is consistent with those of substance P-mediated ERK activation by rat neurokinin 1 receptors in HEK293 cells . Importantly, those studies demonstrated other signalling and functional consequences of ECE-1-mediated degradation of substance P and highlighted that ECE-1 activity regulated the lifetime of the receptor–arrestin complex and its ability to scaffold signalling molecules. ECE-1 also regulates neurotensin-induced proinflammatory signalling and proliferation in human colonocytes by modulating neurotensin receptor-1 recycling and β-arrestin-dependent ERK1/2 and JNK phosphorylation . Furthermore, sustained signalling from endosomes by the neurokinin 1 receptor and calcitonin receptor-like receptor sustains neuronal activity and pain transmission [20,21]. Processing of internalised ligand by ECE-1 may also influence classic second messenger pathways such as cAMP as there is a growing appreciation that these can also arise from internalised receptors, most probably from the endosomal compartment [41–43,81–84]. Indeed, β-arrestin1-dependent, GLP-1R-mediated cAMP generation has been reported and receptor internalisation appears to be critical for the full functionality of the GLP-1R [43,73,85]. Given that persistent cAMP generation by the GLP-1R is responsible for insulin secretion in pancreatic β-cells  and that GLP-1R-mediated ERK signalling is anti-apoptotic and plays an important role in promoting insulin secretion, β-cell proliferation and survival [72,86–88], ligands with different susceptibilities to ECE-1 activity may provide therapeutic challenges or opportunities
The present study suggests that cleavage of GLP-1, possibly following dissociation from the GLP-1R in the acidic environment of the endosomal compartment, is important in the resensitisation of the GLP-1R. This ligand processing may facilitate dissociation of β-arrestin from the GLP-1R, limiting G-protein-independent MAPK signalling but promoting receptor dephosphorylation, recycling and thus resensitisation. The potential for ligands to behave differently in these respects is of relevance to future drug design, either small-molecule allosteric ligands or peptide analogues of GLP-1. Interestingly, ECE-1-dependent generation of potent vasoconstrictor, endothelin-1, is elevated by high glucose levels in vivo , implying that ECE-1 expression increases under hyperglycaemic conditions. Indeed, ECE-1 expression, particularly ECE-1c, is enhanced by high glucose concentrations in vitro . These observations raise the possibility that the signalling and resensitisation of the GLP-1R could be affected in patients with diabetes, thereby affecting disease progression and treatment opportunities.
intracellular Ca2+ concentration
analysis of variance
bovine serum albumin
calcitonin gene-related peptide
early endosome antigen 1
enhanced green fluorescent protein
extracellular signal-regulated kinase
glucose-dependent insulinotropic polypeptide
glucagon-like peptide 1
glucagon-like peptide 1 receptor
human embryonic kidney
HEK293 cells expressing the human recombinant glucagon-like peptide 1 receptor
protein kinase A
protein kinase C
Tris-buffered saline with Tween 20
G.B.W. initiated the project. J.L. performed experiments. G.B.W. and J.L. analysed the data and wrote the manuscript.
The Authors declare that there are no competing interests associated with the manuscript.
Present address: State Key Laboratory of Cell Biology, Institute of Biochemistry and Cell Biology, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai 200031, China.