Signalling, trafficking and glucoregulatory properties of glucagon‐like peptide‐1 receptor agonists exendin‐4 and lixisenatide

Background and Purpose Amino acid substitutions at the N‐termini of glucagon‐like peptide‐1 (GLP‐1) receptor agonist peptides result in distinct patterns of intracellular signalling, sub‐cellular trafficking and efficacy in vivo. Here, we to determine whether sequence differences at the ligand C‐termini of clinically approved GLP‐1 receptor agonists exendin‐4 and lixisenatide lead to similar phenomena. Experimental Approach Exendin‐4, lixisenatide and N‐terminally substituted analogues with biased signalling characteristics were compared across a range of in vitro trafficking and signalling assays in different cell types. Fluorescent ligands and new time‐resolved FRET approaches were developed to study agonist behaviours at the cellular and sub‐cellular level. Anti‐hyperglycaemic and anorectic effects of each parent ligand and their biased derivatives were assessed in mice. Key Results Lixisenatide and exendin‐4 showed equal binding affinity, but lixisenatide was fivefold less potent for cAMP signalling. Both peptides induced extensive GLP‐1 receptor clustering in the plasma membrane and were rapidly endocytosed, but the GLP‐1 receptor recycled more slowly to the cell surface after lixisenatide treatment. These combined deficits resulted in reduced maximal sustained insulin secretion and reduced anti‐hyperglycaemic and anorectic effects in mice with lixisenatide. N‐terminal substitution of His1 by Phe1 to both ligands had favourable effects on their pharmacology, resulting in improved insulin release and lowering of blood glucose. Conclusion and Implications Changes to the C‐terminus of exendin‐4 affect signalling potency and GLP‐1 receptor trafficking via mechanisms unrelated to GLP‐1 receptor occupancy. These differences were associated with changes in their ability to control blood glucose and therefore may be therapeutically relevant.


| INTRODUCTION
The glucagon-like peptide-1 (GLP-1) receptor is a well-established pharmacological target for the treatment of both type 2 diabetes and obesity due to its beneficial effects on weight loss and pancreatic beta cell function (Andersen, Lund, Knop, & Vilsbøll, 2018). The main endogenous ligand for GLP-1 receptor, the 29 amino acid peptide GLP-1(7-36)NH 2 , is highly susceptible to degradation by proteolytic enzymes that rapidly destroy it in the circulation, making it unsuitable as a therapeutic agent (Deacon et al., 1998). Therefore, a number of synthetic GLP-1 agonists with longer circulatory half-lives have been developed and subsequently approved for human use (de Graaf et al., 2016). One example is the GLP-1 homologue peptide exendin-4 (Eng, Kleinman, Singh, Singh, & Raufman, 1992), in clinical use for type 2 diabetes treatment as exenatide. This molecule features an extended, proline-rich Cterminal extension (sequence GAPPPS-NH 2 ), which is absent in GLP-1 itself. The precise role of this feature is not clear, but various possibilities have been suggested, including stabilisation of the peptide helical structure (Neidigh, Fesinmeyer, Prickett, & Andersen, 2001), facilitation of inter-protomer coupling within receptor oligomers (Koole et al., 2017) and protection against enzymatic degradation (Lee et al., 2018). A further approved type 2 diabetes GLP-1 mimetic peptide, lixisenatide, shares the first 37 amino acids with exendin-4, including most of the GAPPPS sequence but includes an additional six lysine residues at the Cterminus prior to the terminal amidation (Andersen et al., 2018).
Due to putative importance of the exendin-4 C-terminus, it is conceivable that the lixisenatide-specific changes could affect its pharmacology.
Biased signalling has emerged as a promising strategy to improve the therapeutic efficacy of drugs through selective activation of "beneficial" intracellular pathways, while minimising those thought to be responsible for adverse effects (Kenakin, 2018).
Recent work has highlighted how GLP-1 receptor signal bias and related membrane trafficking effects regulate insulin release from beta cells (Zhang et al., 2015;Buenaventura et al., 2018;. Following agonist binding the GLP-1 receptor is rapidly endocytosed and while active GPCRs can continue to generate intracellular signals within the endosomal compartments (Eichel & von Zastrow, 2018), the availability of surface GLP-1 receptors to extracellular ligand appears to be an important determinant of sustained insulinotropic efficacy in a pharmacological setting . The GLP-1 receptor ligand N-terminus interacts with the receptor core to instigate conformational rearrangements needed for stable engagement with intracellular signalling and trafficking effectors, while its C-terminus facilitates this process by establishing the

What is already known
• Glucagon-like peptide-1 receptor agonists are used to treat type 2 diabetes and obesity.
• Recently described biased GLP-1 receptor agonists show distinct patterns of intracellular signalling and membrane trafficking.

What this study adds
• Two commonly prescribed GLP-1 agonists, exendin-4 and lixisenatide, perform differently in vitro and in vivo.
• These differences may be linked to their distinct effects on GLP-1 receptor recycling.

What is the clinical significance
• Signal bias and trafficking should be considered in the development of novel GLP-1 agonists. correct orientation of the peptide through interactions with the receptor extracellular domain (de Graaf et al., 2016). Suggesting that the C-terminal sequence differences between exendin-4 and lixisenatide might impact on these cellular processes, a limited evaluation in our earlier study implied that lixisenatide displays reduced signalling potency and insulinotropism compared to exendin-4 .
In the present study, we extended our earlier evaluation to include formal comparison of bias between cAMP signalling and endocytosis in different cell types, measurements of ligand-induced clustering at the plasma membrane, post-endocytic targeting to recycling and degradative pathways aided by a novel cleavable time-resolved FRET probe and assessment of the impact of these changes on exendin-4 versus lixisenatide metabolic responses in vivo. penicillin/streptomycin and G418 (1 mgÁml −1 ). HEK293T (RRID: CVCL_0063) cells were maintained similarly but without G418.

| Saturation binding experiments with fluorescein isothiocyanate (FITC)-ligands
Cells were treated with FITC-ligands over a range of concentrations for 24 h at 4 C before measurement. Equilibrium binding constants were calculated using the "one site-specific binding" algorithm in Prism 8 (GraphPad Software).

| Competition binding experiments at equilibrium
Cells were treated with a fixed concentration (10 nM) of exendin (9-39)-FITC in competition with a range of concentrations of unlabelled exendin-4 or lixisenatide for 24 h at 4 C before measurement. Binding constants were calculated using the "one site-fit K i " algorithm in Prism 8, using the equilibrium dissociation constant for exendin(9-39)-FITC measured in the same experiment by saturation binding.

| Competition kinetic binding experiments
TR-FRET signals were measured at regular intervals before and after addition of different concentrations of exendin(9-39)-FITC, or different concentrations of unlabelled agonist in combination with a fixed concentration (10 nM) of exendin(9-39)-FITC at 37 C. Rate constants for association and dissociation of the unlabelled ligands were calculated using the "kinetics of competition binding" algorithm in Prism

| Measurement of mini-G and β-arrestin recruitment by NanoBiT complementation
The plasmids for mini-G s , -G i and -G q , each tagged at the N-terminus with the LgBiT tag and the SmBiT-tagged endothelin A receptor (ETAR) plasmid (Wan et al., 2018) were a gift from Prof Nevin Lambert, Medical College of Georgia. The plasmid for β-arrestin-2 fused at the N-terminus to LgBiT was obtained from Promega (plasmid no. CS1603B118); this configuration was used due to previous success with another class B GPCR (Shintani et al., 2018). TR-FRET was monitored before and after agonist addition in a Flexstation 3 plate reader at 37 C using the following settings: λ ex = 335 nm, λ em = 520 and 620 nm, delay 400 μs, integration time 1,500 μs. TR-FRET was quantified as the ratio of fluorescent signal at 520 nm to that at 620 nm after subtraction of background signal at each wavelength (simultaneously recorded from wells containing 24-μM fluorescein in HBSS but no labelled cells).

| Measurement of GLP-1 receptor clustering by TR-FRET
The assay was performed similarly to a previous description (Buenaventura et al., 2019). HEK293-SNAP-GLP-1 receptor cells were labelled in suspension with SNAP-Lumi4-Tb (40 nM) and SNAP-Surface 649 (1 mM) for 1 h at room temperature in complete medium.
After washing, cells were resuspended in HBSS, and TR-FRET was monitored before and after agonist addition at 37 C in a Spectramax i3x plate reader in HTRF mode. TR-FRET was quantified as the ratio of fluorescent signal at 665 nm to that at 616 nm, after subtraction of background signal at each wavelength. Lumi4-Tb is a cleavable SNAP-tag probe that allows release of the lanthanide moiety following reduction of its disulfide bond when exposed to reducing agents. After washing, BG-SS-Lumi4-Tb labelled

| Electron microscopy
Resin-embedded ultrathin 70 nm sections on copper grids were imaged on a FEI Tecnai T12 Spirit TEM. Images were acquired in a CCD camera (Eagle).

| Raster image correlation spectroscopy (RICS)
HEK293 NaN 3 -free Alexa Fluor 488 Streptavidin, 10-nm colloidal gold (Molecular Probes) and stimulated with 100 nM of the indicated agonist for 1 h. Sample preparation for conventional EM was performed as previously described (Tomas, Futter, & Moss, 2004). Ultrathin 70 nm sections were cut en face with a diamond knife (DiATOME) and collected on 100 mesh hexagonal copper grids prior to imaging.
Electron micrographs were individually thresholded to create binary images displaying only gold particles. All images were systematically processed to quantify gold-labelled receptors so that multi-particle aggregates would be registered as single large complexes as follows: First, the ImageJ "dilate" algorithm was applied three times so that adjacent gold particles within a complex would coalesce and second, the ImageJ "particle analysis" algorithm was run on the processed image to quantify the area of all particles and particle aggregates. The number of particles per aggregate was estimated using the determined size of clearly identified single gold particles after image processing. Forty images per treatment were quantified.

| Insulin secretion assays
INS-1832/3 cells were stimulated with agonist for 16-18 h in complete medium at 11-mM glucose. At the end of the incubation period, a sample of supernatant was removed, diluted and analysed for secreted insulin by HTRF (Insulin High Range Assay, Cisbio). Insulin secretion was expressed relative to that from cells stimulated with 11-mM glucose alone in the same experiment.

| Animal studies
Animal studies are reported in compliance with the ARRIVE guidelines (Kilkenny et al., 2010) and with the recommendations made by the

| Intraperitoneal glucose tolerance tests
Mice were lightly fasted (2-3 h) before the test; 2 gÁkg −1 of 20% glucose was injected via the i.p. route, either concurrently with agonist or after a specified delay. Blood glucose was recorded at the indicated time-points from the tail vein using a handheld glucometer (GlucoRx Nexus).

| Food intake assay
Individually caged mice were fasted overnight. Diet was returned immediately after a 100 μl i.p. injection of agonist or vehicle (0.9% NaCl) and intake monitored by measuring food weight at the indicated time-points. Food spillage was not accounted for but would be expected to apply equally to each treatment.

| Pharmacokinetic study
Mice were injected i.p. with 100 μl agonist and a 25 μl blood sample was subsequently taken from the tail vein into lithium-heparin capillary tubes onto ice. The high ligand dose was required due to assay sensi- was not possible to perform in vitro treatments in a blinded manner.

| In vivo experiments
All in vivo experiments included at least five mice per group. Group sizes were determined on the basis of previously established n numbers required to demonstrate the size of the effect expected from the in vitro results, without formal power calculations. Treatment order was randomly assigned and the average body weight within each group calculated to ensure this did not differ by more than 1 g. The investigator performing the experiment was blinded to treatment allocations.

| Data and statistical analysis
TThe data and statistical analysis comply with the recommendations of the British Journal of Pharmacology on experimental design and analysis in pharmacology (Curtis et al., 2018). Quantitative data were analysed using Prism 8.0 (GraphPad Software). Mean ± standard error of mean (SEM), or individual replicates, are displayed throughout.  we also used the FRET biosensor T Epac VV (Klarenbeek et al., 2011) to obtain real-time readouts of cAMP signalling ( Figure S1E). Comparison of T Epac VV and DERET potencies at 10-min intervals indicated the selective loss of cAMP potency with lixisenatide was preserved throughout the stimulation period ( Figure S1F). Downstream coupling to PKA activation was similarly reduced with lixisenatide ( Figure S1G).
These studies indicate that coupling of lixisenatide to cAMP signalling is reduced compared to with exendin-4, in a manner unrelated to receptor occupancy.

| Recruitment responses measured using NanoBiT complementation
To attempt to understand why lixisenatide shows reduced cAMP signalling despite similar binding affinity to exendin-4, we used a NanoBiT complementation approach (Dixon et al., 2016) to measure recruitment of "mini-G proteins" (Wan et al., 2018) and β-arrestin-2 to GLP-1 receptor (Figure 2a,b). The GLP-1 receptor construct developed in house for these assays, which bears a FLAG-tag at the N-terminus and a SmBiT tag at the C-terminus, was validated by showed the same pattern ( Figure S2C). As the ligand concentration used was supramaximal, the implication of these studies is that lixisenatide is a modestly less efficacious ligand than exendin-4 for both G s and β-arrestin-2 recruitment.

| GLP-1 receptor clustering responses with exendin-4 and lixisenatide
We previously demonstrated that GLP-1 receptor segregates into plasma membrane nanodomains after stimulation with exendin-4 and that this process is required for functional responses such as endocytosis (Buenaventura et al., 2019). As recent evidence raises the possibility that exendin-4 simultaneously binds two GLP-1 receptor protomers through a non-canonical interaction made by its C-terminus and the receptor extracellular domain (Koole et al., 2017), which might influence the receptor oligomeric state, we investigated whether the distinct C-termini of exendin-4 and lixisenatide could differentially modulate the GLP-1 receptor clustering response. We employed raster image correlation spectroscopy (RICS), which has previously been used to demonstrate how ligand binding leads to receptor clusters with restricted diffusion in the plasma membrane (Compte et al., 2018).

| Differences in GLP-1 receptor recycling following exendin-4 and lixisenatide treatment
After initial endocytosis, differences in intracellular receptor trafficking can modulate GLP-1 receptor-induced insulin release . To measure the rate of GLP-1 receptor plasma membrane recycling, we treated HEK293-SNAP-GLP-1 receptor cells with exendin-4 or lixisenatide, followed by a variable recycling period, after which we applied a SNAP-Surface fluorescent probe to label total surface GLP-1 receptor. Treatment with lixisenatide versus exendin-4 was associated with a slower rate of GLP-1 receptor recycling (Figure 4a). To corroborate this finding, we developed a TR-FRET assay to measure SNAP-GLP-1 receptor recycling in real-time with a plate reader, using a cleavable form of the SNAP-labelling TR-FRET donor SNAP-Lumi4-Tb to allow reversible labelling by release of the fluorophore under mild reducing conditions (see Figure 4b for a graphical description of the assay principle and Figure S4A-F for further validation of the probe). Using CHO-K1 cells stably expressing SNAP-GLP-1 receptor, which remain adherent during the multiple wash steps, we again found a reduced rate of GLP-1 receptor recycling after lixisenatide compared to exendin-4 treatment ( Figure 4c).
As pH-dependent dissociation of ligand-receptor complexes within acidic endosomes is known to influence post-endocytic receptor sorting (Borden, Einstein, Gabel, & Maxfield, 1990), we wondered if the impact of low pH might specifically affect interactions made by the positively charged lixisenatide C-terminus, modulating intraendosomal binding and thereby explaining its different recycling rate.
We therefore developed FITC-conjugates of each agonist ( Moreover, the progressive loss of TR-FRET signal from agonist prebound to surface receptors at low temperature and subsequently endocytosed by return to 37 C was equal for both conjugates (Figures 4i and S4H, I), also suggesting intra-endosomal dissociation of agonist-receptor complexes does not differ between agonists. Overall, these results suggest the difference in recycling rate between exendin-4 and lixisenatide is not related to differences in persistence of receptor binding within endosomes.
As GPCR recycling can be controlled by PKA (Vistein & Puthenveedu, 2013), we also wondered whether our observation of reduced PKA signalling despite similar occupancy for lixisenatide versus exendin-4 ( Figure S1G) might explain their different recycling rates. However, treatment with the PKA inhibitor H89 did not affect GLP-1 receptor recycling after exendin-4 pretreatment (Figure 4j).

| Divergent effects of exendin-4 and lixisenatide in pancreatic beta cells
Potentiation of glucose-stimulated insulin secretion is a major therapeutic goal of GLP-1 antagonists treatment. We used rat insulinoma- lixisenatide-FITC was observed by confocal microscopy (Figure 5d).
To measure GLP-1 receptor recycling after pretreatment with exendin-4 or lixisenatide, we applied fluorescent exendin-4-TMR (- Figure S5B, C) at the beginning of the recycling period following extensive wash of unlabelled agonist. As exendin-4-TMR is rapidly endocytosed by GLP-1 receptors that reappear at the cell surface, its intracellular accumulation is indicative of the recycling rate. Again, recycling of GLP-1 receptor after pretreatment with lixisenatide was less extensive than with exendin-4 ( Figure 5e).
Agonist-internalised receptors which do not follow a recycling pathway can be sorted towards lysosomal degradation and in keeping with this, we found increased co-localisation of SNAP-GLP-1 receptor with a fluorescent lysosomal marker after prolonged treatment with lixisenatide compared to exendin-4 ( Figure 5f). Moreover, electron microscopy imaging showed that, following live cell labelling of surface SNAP-GLP-1 receptors with a 10-nm gold probe prior to agonist stimulation, the distribution of gold particles favoured larger size aggregates with lixisenatide versus exendin-4 treatment (Figure 5g).
This pattern of gold aggregation is indicative of probe target lysosomal degradation, as previously demonstrated for the EGF receptor (EGFR) (Futter & Hopkins, 1989;Futter, Pearse, Hewlett, & Hopkins, 1996). As excessive loss of surface GLP-1 receptors without compensatory increases in recycling can limit insulinotropic efficacy , we measured cumulative insulin secretion after overnight treatment with each agonist ( Figure 5H).
Consistent with this paradigm, maximal insulin release with lixisenatide was reduced. Thus, the distinct pharmacological properties of lixisenatide and exendin-4 translate to functional differences in beta cells.

| Lixisenatide is less effective in vivo
GLP-1 agonists are primarily used for the treatment of type 2 diabetes to reduce glycaemia and promote weight loss through appetite reduction. We assessed the glucoregulatory effects of each ligand at varying doses in mice via i.p. glucose tolerance tests performed immediately and 6 h after agonist treatment, to identify acute and delayed effects. We found that the anti-hyperglycaemic effect of exendin-4 was greater than equimolar lixisenatide (Figure 6a-c). Measurements of food intake in overnight-fasted mice also showed that the anorectic effect of lixisenatide is reduced compared to exendin-4 (Figure 6d-f).

| DISCUSSION
In this study, we performed a side-by-side pharmacological evaluation of two closely related GLP-1 , receptor agonists both of which are in routine clinical usage and differ structurally only by the hexalysine Cterminal extension in lixisenatide. We found that, despite similar binding affinity, coupling of lixisenatide to cAMP signalling was reduced compared to exendin-4. After similar levels of GLP-1 receptor plasma membrane clustering and endocytosis induced by both peptides, recycling of GLP-1 receptor was slower after lixisenatide treatment, with apparent preferential targeting to a degradative lysosomal pathway. This resulted in a reduction in insulinotropic efficacy and the ability to control blood glucose 6-8 h after dosing in mice. Therefore, the structural differences between the C-termini of each peptide appear to have some functional importance. See Table 1 for a summary of the properties of each ligand.
Our finding here of reduced cAMP signalling potency with lixisenatide matches our earlier observation . We also found reduced cAMP potency in rat   , 2017), resulting in changes in its oligomeric state, a factor known to influence coupling to signalling intermediates (Milligan, Ward, & Marsango, 2019), including for GLP-1 receptor (Buenaventura et al., 2019;Harikumar et al., 2012). However, we observed no major differences between the GLP-1 receptor clustering responses induced by either ligand. Using NanoBiT complementation, we found subtle differences in GLP-1 receptor coupling to both G s and β-arrestin-2 between exendin-4 and lixisenatide, although it is not clear whether this is sufficient to explain the significant differences in potency for cAMP signalling. In our assays, we found that each ligand induced only minor levels of recruitment of mini-G q to GLP-1 receptor, raising questions about the importance of signalling via this G protein in GLP-1 receptor responses (Shigeto et al., 2015). However, we cannot exclude cell type-specific effects, or the possibility that the recruitment pattern of catalytically inactive mini-G q underestimates the potential for G q activation.
The other key finding here pertains to the post-endocytic trafficking of the two ligands, which was evaluated across different cell systems using a variety of complementary approaches. Despite similar internalisation profiles, GLP-1 receptor recycling after lixisenatide treatment was slower in both HEK293 and INS-1832/3 beta cells.
In contrast, a higher degree of lixisenatide-stimulated GLP-1 receptors tended to progressively co-localise with the lysosomotropic fluorescent probe Lysotracker, indicating preferential targeting of the receptor towards a degradative pathway. This was in agreement with the increased level of intracellular gold-conjugated SNAP-tag probe aggregation detected by electron microscopy, suggesting enhanced tendency for lysosomal degradation of the lixisenatidestimulated receptor. These phenotypes partly recapitulate the differences previously observed with biased exendin-4-derived GLP-1 agonists , which were linked to diminished insulin secretion efficacy. Indeed, we found in the present study that maximal insulin secretion was reduced when beta cells were exposed to lixisenatide versus exendin-4 over a sustained exposure period. However, despite these broadly similar sets of findings, the mechanism for slowing of GLP-1 receptor recycling with lixisenatide did not appear to depend on greater binding affinity, a factor that was previously demonstrated to influence the recycling of biased exendin-4-derived GLP-1 receptor agonists . We specifically aimed to address the possibility of protonation of the hexalysine scaffold of lixisenatide in acidic conditions leading to altered intra-endosomal agonist dissociation but could find no evidence for this. We did not examine post-translational modifications linked to target degradation such as ubiquitination (Clague & Urbé, 2010), but note that the GLP-1 receptor is not ubiquitinated by treatment with exendin-4, despite a considerable amount of receptor degradation measurable after continuous exposure to this agonist . The reason for the increased lysosomal post-endocytic targeting of the GLP-1 receptor with lixisenatide therefore remains unclear. It should also be emphasised that we only measured recycling after stimulation with a high ligand concentration so as to promote a large degree of initial GLP-1 receptor endocytosis and we cannot be certain that the same effects would be observed at the lower concentrations likely to be encountered in vivo.
Our observations of generally reduced biological effect of lixisenatide for physiologically important readouts suggest that these pharmacological differences are indeed translated to differences in downstream responses. In particular, we found reduced efficacy for sustained insulin secretion with lixisenatide using an in vitro beta cell system, as well as reduced anti-hyperglycaemic and anorectic effects in mice. This observation comes with the caveat that we cannot be certain that the effects observed in vivo represent the same phenomena as observed with our prolonged in vitro incubations.
Moreover, as our studies were performed only in male mice, we cannot exclude the possibility of sex-specific effects. As the "advantages" of exendin-4 in vivo were detectable acutely, in a dose-dependent manner, it is likely that they are partly attributable to agonist potency differences rather than the post-endocytic trafficking phenotypes.
However, the potential for GLP-1 receptor endocytosis to influence pharmacodynamics of exendin-4 has previously been modelled (Gao & Jusko, 2012) and while this model focused on receptor internalisation, differences in recycling and degradation rates could plausibly be linked via similar mechanisms.
Differentiation in the therapeutic profiles of GLP-1 antagonists in humans has been noted on many occasions (Aroda, 2018). A head-to-head comparison in patients with type 2 diabetes showed numerically greater HbA1c reduction and weight loss with exenatide compared to lixisenatide (Rosenstock et al., 2013).
Exenatide showed a beneficial effect on cardiovascular outcomes, albeit with borderline significance (Holman et al., 2017), whereas a separate trial of lixisenatide found no evidence of benefit (Pfeffer et al., 2015). However, understanding the link between the receptor pharmacology observed in our study and real-world performance of each agonist is hampered by the different dosing and administration schedules (10 μg twice daily for exenatide, or weekly as a sustained release preparation, compared to 20 μg once daily for lixisenatide).
Following the distinctive effects, we previously observed with biased GLP-1 antagonists derived from exendin-4 , we developed biased lixisenatide-derived compounds based on a similar design. While we did not compare these against their parent ligands, the -phe1 substitution in both exendin and lixisenatide configurations displayed favourable characteristics such as reduced internalisation and fast recycling compared to the -asp3 variants. This translated to improved insulin secretion in vitro and significantly better anti-hyperglycaemic effect in vivo.
These observations therefore add to the evidence that modifications to GLP-1 antagonist N-termini are capable of inducing functionally important signal bias (Zhang et al., 2015;Fremaux et al., 2019).
In summary, our study provides insights into specific signalling and trafficking differences of two GLP-1 antagonists in routine clinical use, linking these characteristics to their effects in vivo. The precise molecular mechanisms underpinning these differences remains to be elucidated.