Molecular Characterisation of Small Molecule Agonists Effect on the Human Glucagon Like Peptide-1 Receptor Internalisation

The glucagon-like peptide receptor (GLP-1R), which is a G-protein coupled receptor (GPCR), signals through both Gαs and Gαq coupled pathways and ERK phosphorylation to stimulate insulin secretion. The aim of this study was to determine molecular details of the effect of small molecule agonists, compounds 2 and B, on GLP-1R mediated cAMP production, intracellular Ca2+ accumulation, ERK phosphorylation and its internalisation. In human GLP-1R (hGLP-1R) expressing cells, compounds 2 and B induced cAMP production but caused no intracellular Ca2+ accumulation, ERK phosphorylation or hGLP-1R internalisation. GLP-1 antagonists Ex(9–39) and JANT-4 and the orthosteric binding site mutation (V36A) in hGLP-1R failed to inhibit compounds 2 and B induced cAMP production, confirming that their binding site distinct from the GLP-1 binding site on GLP-1R. However, K334A mutation of hGLP-1R, which affects Gαs coupling, inhibited GLP-1 as well as compounds 2 and B induced cAMP production, indicating that GLP-1, compounds 2 and B binding induce similar conformational changes in the GLP-1R for Gαs coupling. Additionally, compound 2 or B binding to the hGLP-1R had significantly reduced GLP-1 induced intracellular Ca2+ accumulation, ERK phosphorylation and hGLP-1R internalisation. This study illustrates pharmacology of differential activation of GLP-1R by GLP-1 and compounds 2 and B.


Enzyme linked immunosorbent assay (ELISA)
This assay was carried out as described previously with unpermeabilised cells to quantify cell surface expression of the hGLP-1R [20,28]. Briefly, HEK293 cells expressing the hGLP-1R were serum starved for 1h and then stimulated without or with agonist at 37°C/5% CO 2 . Where indicated, cells were incubated with antagonist for 30min prior and during stimulation with agonist at 37°C/5% CO 2 . Cells were then fixed with 4% paraformaldehyde (PFA,) for 10min, blocked with 1% BSA made in TBS (1% BSA/TBS) for 45min and probed with the anti-GLP-1R mouse antibody (diluted 1:15000) in 1% BSA/TBS for 1h. Cells were washed with TBS and then incubated with the HRP-conjugated anti-mouse IgG (diluted 1:5000) in 1% BSA/TBS for 1h. Cells were washed and developed using 1-step Ultra TMB-ELISA substrate (Bio-Rad, Herts, UK) for 15min and the reaction stopped by adding an equal volume of 2M sulphuric acid. The optical density was read at 450nm using a plate reader.

Live cell fluorescence imaging
For live cell fluorescence imaging, HEK293 cells transiently transfected with SP-VSVG-hGLP-1RΔN23-GFP plasmid for 24h were plated into an 8 chamber glass bottom slide (Thermo Scientific) pre-coated with 0.1mg/ml poly-L-lysine and incubated at 37°C/5% CO 2 in FSM. After 24h, cells were washed 3 times with and incubated in 250μl per well of SFM for 1h at 37°C/5% CO 2 for serum starvation. Live cells were then imaged at 37°C by using Zeiss LSM710 confocal microscope fitted with a 63x oil immersion lens. Cells were imaged twice (0 and 3 min) before adding agonist and for every 3min after stimulating with agonist for 60min, as described previously [21,29].

cAMP assay
Cells were serum starved for 1h and then stimulated, in the presence of 0.25mM phosphodiesterase inhibitor Ro201724, without or with 100nM GLP-1 for 1h at 37°C/5% CO 2 . Cells were then lysed and cAMP levels in the cell lysates were estimated using the cAMP direct immunoassay kit (Abcam, Cambridge, UK), as described previously [20,28].

cAMP, Ca 2+ and ERK luciferase assay
This assay was carried out as described [20]. HEK293 cells cotransfected with the plasmids of the hGLP-1R and luciferase reporter for cAMP (pGL4.29-Luc-CRE) or intracellular Ca 2+ (pGL4.30-Luc-NFAT) or ERK phosphorylation (pGL4.33-Luc-SRE) were treated with increasing concentrations of agonist for 4h (cAMP and ERK) or 8h (Ca 2+ ) at 37°C/5% CO 2 . After the stimulation, an equal volume of ONE-Glo™ lysis buffer containing luciferase substrate (Promega, Southampton, UK) was added to each well and luminescence measured using a plate reader in accordance with the manufacturer's instructions.

Data analysis
Data were analysed using the GraphPad Prism programme. The data presented as mean ± SEM of three independent experiments. Statistical comparisons between a control and test value was made by a one-tailed paired student t-test. Statistical analysis between multiple groups was determined by the Bonferroni's post test after one-way or two-way analysis of variance (ANOVA), where p>0.05 was considered as statistically not significant (n.s.), and p<0.05, p<0.01, and p<0.001 shown as Ã , ÃÃ and ÃÃÃ respectively. Concentration response curves were also fitted using Prism, according to a standard logistic equation.

Pharmacological analysis of two small molecule agonists of the hGLP-1R
We have recently shown that two small molecule agonists (compounds 2 and B) of GLP-1R stimulate cAMP production but have no effect on GLP-1R internalisation, intracellular Ca 2+ accumulation or ERK phosphorylation [20]. Here, compounds 2 and B agonistic effect on the hGLP-1R was pharmacologically assessed (using the cAMP production, intracellular Ca 2+ accumulation, ERK phosphorylation and receptor internalisation as readouts (Fig 1)) and compared to that of GLP-1. GLP-1 stimulated a concentration dependent increase in cAMP production in HEK293 cells expressing the hGLP-1R ( Fig 1A). Compound 2 and compound B also induced the same levels of cAMP production in hGLP-1R expressing HEK293 cells, demonstrating both compounds 2 and B stimulate cAMP production with E max values similar to that of GLP-1 [25]. However, the EC 50 of GLP-1 for inducing cAMP production is 3 orders of magnitude lower than that of compounds 2 and B ( Fig 1D) [25]. GLP-1 also potentiated intracellular Ca 2+ accumulation ( Fig 1B) and ERK phosphorylation ( Fig 1C) in a concentration dependent manner in hGLP-1R expressing cells. However, compounds 2 and B had no effect on intracellular Ca 2+ accumulation and ERK phosphorylation. Taken together, these results demonstrate compounds 2 and B induce cAMP production with similar E max to GLP-1 but do not potentiate intracellular Ca 2+ accumulation or ERK phosphorylation in hGLP-1R expressing cells.
Since intracellular Ca 2+ accumulation and ERK phosphorylation are required for GLP-1 stimulated hGLP-1R internalisation [20], the effect of compounds 2 and B on hGLP-1R internalisation was pharmacologically assessed next. HEK293 cells expressing the hGLP-1R were challenged with increasing concentrations of GLP-1, compound 2 or compound B for 60min and then the cell surface expression of the receptor was analysed by ELISA using the anti-GLP-1R antibody (Fig 2A) and the anti-VSVG antibody ( Fig 2B). The orthosteric agonist GLP-1 induced a dose dependent increase in hGLP-1R internalisation and had a maximal effect of 76.03 ± 4.4% at 100nM. The EC 50 of GLP-1 for hGLP-1R internalisation is shown in Fig 2C. Interestingly, compound 2 showed no induction of hGLP-1R internalisation up to 3.3μM and at its highest concentration (100μM) only 16.63 ± 6.96% of cell surface receptors were  internalised. Additionally, compound B showed no effect on internalisation of the receptor up to a concentration of 100μM. We were unable to use higher than 0.1mM of compounds 2 and B since they found to be toxic to HEK293 cells above that concentration (data not shown) [4]. When hGLP-1R internalisation was assessed by ELISA using the anti-VSVG antibody, the results obtained were similar to that obtained with the anti-GLP-1R antibody (Fig 2B and 2C). This indicated the anti-GLP-1R antibody does not interfere with compound 2 and compound B binding to the receptor and therefore only the anti-GLP-1R antibody was used in further experiments. These results were confirmed by immunofluorescence analysis (Fig 2D) where intracellular punctate structures, indicative of hGLP-1R internalisation, were observed in cells treated with GLP-1, but were absent in cells treated with compound 2 or B. Additionally, the time dependent effect of GLP-1, compound 2 and compound B on hGLP-1R internalisation was determined by using ELISA (Fig 3A) and live cell imaging (Fig 3B). GLP-1 induced hGLP-1R internalisation in a time dependent manner, reaching maximum internalisation of the receptor at approximately 60min of stimulation (73.57 ± 5.81%). In contrast, no internalisation of the receptor was observed for compounds 2 and B. Live cell fluorescence imaging showed the appearance of internalised cell surface GFP-tagged receptor as intracellular punctate structures when challenged with GLP-1 but not with compound 2 or compound B, supporting the ELISA results. Together, these results demonstrate that, unlike GLP-1, the small molecule agonists do not internalise the hGLP-1R most likely because they are unable to induce intracellular Ca 2+ accumulation or ERK phosphorylation.
Ex(9-39) and JANT-4 act as antagonists for GLP-1 but not compounds 2 and B Ex(9-39) and JANT-4 are known antagonist of the GLP-1R that work by binding to the orthosteric binding site [9,10,26,32], which therefore competitively inhibit GLP-1 binding to the receptor. Compounds 2 and B have been described as ago-allosteric agonists [4,17,18,33]. To confirm this, the effect of antagonists Ex(9-39) and JANT-4 on these small molecule agonists was determined. For this purpose, hGLP-1R expressing cells pre-incubated with Ex(9-39) or JANT-4 were stimulated with GLP-1 (Fig 4A), compound 2 ( Fig 4B) or compound B (Fig 4C) and then determined cAMP production. The EC 50 of GLP-1, compound 2 and compound B in the presence and absence of Ex(9-39) or JANT-4 for cAMP production is shown in Fig 4D. GLP-1 stimulated a dose dependent increase in cAMP production. In the presence of Ex  or JANT-4, GLP-1 induced cAMP production was reduced. In contrast, Ex(9-39) and JANT-4 had no effect on compound 2 stimulated cAMP production. Similarly, antagonists Ex(9-39) and JANT-4 had no effect on the cAMP production stimulated by compound B. These results confirmed compound 2 and compound B do not bind to the orthosteric agonist binding site.
The idea that compound 2 and compound B act through a binding site distinct from the orthosteric binding site was further assessed by using two hGLP-1R mutants (V36A and K334A). The V36A mutant of hGLP-1R prevents agonists binding to the orthosteric binding site [35] whereas the K334A mutant reduces cAMP production by inhibiting the receptor coupling to Gα s subunit [36,37]. The V36A and K334A mutants were assessed for their expression at protein level (determined by immunoblotting (Fig 6A)), cell surface expression and agonist induced internalisation (determined by ELISA (Fig 6B and 6C) and immunofluorescence (Fig 6D)). The    V36A and K344A total protein expression and cell surface expression was similar to that of the wild type (WT) hGLP-1R (103.2 ± 9.55% and 108.9 ± 2.17%, p>0.05, respectively). As expected, agonist induced hGLP-1R internalisation was almost abolished by the V36A mutation (12.4 ± 7.27%, p<0.001) but unaffected by the K334A mutation (97.54 ± 3.67%, p>0.05) [35][36][37]. HEK293 cells expressing either the hGLP-1R WT, V36A mutant or K334A mutant were treated with increasing concentrations of GLP-1 (Fig 7A), compound 2 ( Fig 7B) and compound B (Fig 7C), assessed for cAMP production and calculated EC 50 values (Fig 7D). GLP-1 increased cAMP production in a concentration dependent manner in the WT expressing cells but not in the V36A mutant (p<0.001) expressing cells. Compound 2 stimulated cAMP production in a  These results confirmed that the V36A mutation affects the orthosteric binding site of the hGLP-1R. Stimulation of cAMP production in the K334A mutant expressing cells was significantly reduced with GLP-1, compound 2 or compound B treatment. These results suggest that, although the small molecule agonists bind at a different site on the hGLP-1R, GLP-1, compound 2 and compound B may alter conformation of the receptor in a similar way so that the receptor couples to Gα s and induces cAMP production.

Discussion
Allosteric small molecule agonists have the potential benefit of binding to a site on the receptor that is distinct from the site used by the orthosteric agonist. Therefore, allosteric agonists can act upon the receptor at the same time as the endogenous orthosteric agonist and alter affinity and/or efficiency of the orthosteric agonist, potentially providing more 'physiological'  regulations [38]. Recently, two small molecule agonists, compound 2 and compound B, have been described, which act not only as allosteric modulators but also as agonists for the GLP-1R [17,18]. However, compounds 2 and B shown to have no effect on GLP-1R internalisation, which is mediated by the Gα q pathway [20]. In this study, we pharmacologically assessed their effect on GLP-1R activity and internalisation.
In agreement with our previous study [20], both the small molecule agonists of the hGLP-1R induced cAMP production but not intracellular Ca 2+ accumulation or ERK phosphorylation and as a result they did not induce hGLP-1R internalisation. Compounds B and 2 have previously been reported to induce a small or no increase in intracellular Ca 2+ accumulation [4,25,39], which effect, however, on GLP-1R internalisation is unknown. Studying the internalisation of GLP-1R induced by compounds 2 and B is useful in assessing the effectivity of these compounds with longer half-life. This is because internalisation of the receptor can lead to dampening of its biological response [32]. The μ-opioid receptor agonist, herkinorin, induces ERK1/2 phosphorylation but not internalisation of the receptor [40]. Additionally, the allosteric agonist AC-42 binds to the M 1 muscarinic acetylcholine receptor, which results in ERK phosphorylation and intracellular Ca 2+ mobilisation but not internalisation of the receptor [41,42]. This suggests orthosteric and allosteric agonists can cause subtle differences in the conformation of the receptors, activating separate signalling pathways. Additionally, this further supports the idea that the GLP-1R does not require cAMP for internalisation of the receptor, but instead intracellular Ca 2+ accumulation and ERK phosphorylation are essential [20].
In this study, antagonists Ex(9-39) [9,10] and JANT-4 [26] inhibited GLP-1 induced GLP-1R internalisation and signalling but not compound 2 or compound B induced signalling, suggesting a second agonist binding site on the hGLP-1R that is distinct from the orthosteric binding site. These findings are consistent with the results obtained in previous studies for compound 2 [17] and compound B [17,18], which showed antagonist Ex(9-39) had no effect on cAMP signalling. This is further confirmed by using two mutants of the hGLP-1R (V36A and K334A). The V36A mutation in the GLP-1R has previously been shown to affect GLP-1 binding to the orthosteric binding site [35]. In this study, HEK293 cells expressing the V36A mutant did not show GLP-1 stimulated cAMP. In contrast, the V36A mutant expressing cells did show compounds 2 and B stimulated cAMP production to the same levels produced in the hGLP-1R expressing cells upon exposure to these compounds. These results demonstrated that the V36A mutation in the hGLP-1R only affects the orthosteric binding site and, compounds 2 and B interact with the hGLP-1R at a site different to the orthosteric binding site. Additionally, the K334A mutation in the GLP-1R, which has previously been shown to prevent coupling of the receptor to the Gα s subunit [36,37], reduced cAMP production stimulated by GLP-1, compound 2 and compound B. This demonstrates that these small molecule agonists and GLP-1 induce similar conformational changes in the hGLP-1R, which are required for the Gα s coupling, although they bind at different sites on the hGLP-1R. While this manuscript was being prepared, a recent study showed that compounds 2 and B covalently modifies Cys347 in intracellular loop 3 of GLP-1R [34].
In this study, compounds 2 and B have been shown to reduce GLP-1 induced hGLP-1R internalisation, intracellular Ca 2+ accumulation and ERK phosphorylation. Reducing hGLP-1R internalisation prevents dampening of the receptor activity [43]. Therefore, these small molecule agonists may strengthen GLP-1 potency by allowing the orthosteric agonist to act on the receptor for a prolonged period before it is desensitised. Furthermore, they have also been shown to allosterically modulate the function of GLP-1 -NH 2 (GLP-1 metabolite) by sensitising the GLP-1R for activation by this metabolite [18,34,44,45]. Based on this ability, compounds 2 and B may provide insight into the mechanisms of agonist directed GLP-1R regulation and may represent a step further in the development of effective insulinotropic agents with limited adverse effects. This result is similar to the effect of allosteric agonists of the cannabinoid CB 1 receptor, because their binding to the receptor results in a conformation change that increases the affinity of the orthosteric agonist to the receptor [46]. In contrast to compounds 2 and B, allosteric agonist alcuronium inhibits the actions of orthosteric agonist pilocarpine on the M 2 muscarinic acetylcholine receptor [47]. It has also been shown that GLP-1R Internalisation and Small Molecule Agonists compounds 2 and B binding cause a conformational change in GLP-1R that increases GLP-1 access to the orthosteric binding site [17,18]. Therefore, it is possible that they prevent GLP-1 induced GLP-1R internalisation by inhibiting GLP-1 bound hGLP-1R coupling to the Gα q pathway required for intracellular Ca 2+ accumulation and thereby ERK phosphorylation [20]. It has been shown that the intracellular C-terminal domain of GLP-1R is important for agonist-induced internalisation of the receptor [23]. Furthermore, the agonist induced GLP-1R internalisation has recently been shown to be important for insulin secretion [48]. However, compounds 2 and B not only induce insulin secretion but also inhibit GLP-1R internalisation [17,18,20], indicating that GLP-1R may use multiple and compensatory pathways to induce insulin secretion in pancreatic islets.
In summary, small molecule agonists, compound 2 and compound B, were pharmacologically analysed in this study for their effect on hGLP-1R internalisation, cAMP production, intracellular Ca 2+ accumulation and ERK phosphorylation. Although the small molecule agonists induced cAMP production with similar E max to GLP-1, they did not induce intracellular Ca 2+ accumulation and ERK phosphorylation (mediated by the Gα q pathway) and as a result these agonists did not induce hGLP-1R internalisation (Fig 10). We have recently shown that agonist induced internalisation of the GLP-1R is mediated by the Gα q pathway [20]. With the use of antagonists and the V36A mutant of the hGLP-1R, this study demonstrated that compounds 2 and B act on a region of the hGLP-1R independent to the orthosteric agonist site. However, the use of the K334A mutant of the hGLP-1R demonstrated that compounds 2 and B induce a conformational change in the GLP-1R, which is required for the Gα s coupling, similar to that induced by the orthosteric agonist binding to the receptor. Additionally, compounds 2 and B reduced, in a non-competitive manner, GLP-1 induced GLP-1R coupling to Gα q and internalisation (Fig 10). Therefore, this study suggests a potential advantage in the selective activation of specific signalling pathways (biased agonism) by ago-allosteric agonists compounds, which may cause changes in GLP-1R conformation that are less favourable to the internalisation of the receptor.