Nonconventional glucagon and GLP-1 receptor agonist and antagonist interplay at the GLP-1 receptor revealed in high-throughput FRET assays for cAMP

G protein–coupled receptors (GPCRs) for glucagon (GluR) and glucagon-like peptide-1 (GLP-1R) are normally considered to be highly selective for glucagon and GLP-1, respectively. However, glucagon secreted from pancreatic α-cells may accumulate at high concentrations to exert promiscuous effects at the β-cell GLP-1R, as may occur in the volume-restricted microenvironment of the islets of Langerhans. Furthermore, systemic administration of GluR or GLP-1R agonists and antagonists at high doses may lead to off-target effects at other receptors. Here, we used molecular modeling to evaluate data derived from FRET assays that detect cAMP as a read-out for GluR and GLP-1R activation. This analysis established that glucagon is a nonconventional GLP-1R agonist, an effect inhibited by the GLP-1R orthosteric antagonist exendin(9–39) (Ex(9–39)). The GluR allosteric inhibitors LY2409021 and MK 0893 antagonized glucagon and GLP-1 action at the GLP-1R, whereas des-His1-[Glu9]glucagon antagonized glucagon action at the GluR, while having minimal inhibitory action versus glucagon or GLP-1 at the GLP-1R. When testing Ex(9–39) in combination with des-His1-[Glu9]glucagon in INS-1 832/13 cells, we validated a dual agonist action of glucagon at the GluR and GLP-1R. Hybrid peptide GGP817 containing glucagon fused to a fragment of peptide YY (PYY) acted as a triagonist at the GluR, GLP-1R, and neuropeptide Y2 receptor (NPY2R). Collectively, these findings provide a new triagonist strategy with which to target the GluR, GLP-1R, and NPY2R. They also provide an impetus to reevaluate prior studies in which GluR and GLP-1R agonists and antagonists were assumed not to exert promiscuous actions at other GPCRs.

To optimize dual agonist or triagonist peptides for therapeutic purposes, it is necessary to achieve "balanced agonism" in which simultaneous stimulation of multiple GPCRs is achieved across the desired concentration range. It is also necessary to identify the selectivity with which such peptides activate GPCRs, and in this regard it is necessary to identify potential nonconventional actions that allow them to exert off-target effects. Furthermore, the optimization of dual or triagonists requires detailed knowledge of their molecular mechanisms of action. In this regard, modeling of GPCR structure/conformation is facilitated by the availability of GPCR agonists and antagonists that aid in the identification of orthosteric and allosteric mechanisms of ligand binding and receptor activation. One such agonist for the GLP-1R is exendin-4 (Ex-4) (20,21). Four antagonists of family B GPCRs are exendin-(9 -39) (Ex (9 -39)) acting at an orthosteric site on the GLP-1R (20 -22), des-His 1 -[Glu 9 ]glucagon acting at an orthosteric site on the GluR (23), and LY2409021 (24) and MK 0893 (25) acting at an allosteric site on the GluR (26).
Recently, we developed a microplate high-throughput fluorescence resonance energy transfer (FRET) assay that monitors the kinetics with which GPCR agonists stimulate an increase or decrease in levels of the second messenger adenosine-3Ј,5Ј-cyclic monophosphate (cAMP) in cell monolayers (27). This assay uses the H188 biosensor designed by Klarenbeek et al. (28) in which cAMP binds directly to a modified Epac1 protein that is flanked by mTurquoise2⌬ FRET donor and tandem cp173 Venus-Venus FRET acceptor chromophores. Here, we used this FRET assay to discover nonconventional actions of family B GPCR agonists and antagonists, while also investigating the pharmacological properties of a hybrid peptide (GGP817) that incorporates amino acid sequences present within glucagon and PYY. Our analysis reveals unexpected features of GPCR agonist and antagonist action, while also establishing GGP817 to be a prototype triagonist at the GluR, GLP-1R, and NPY2R.

FRET-based assays for GPCR agonist and antagonist action
FRET assays were performed in a 96-well format so that it was possible to monitor the kinetics and dose dependence with which GPCR agonists stimulated or inhibited cAMP production. For this purpose, we used HEK293 cells that stably express recombinant GPCRs and that were virally transduced with H188, thereby allowing FRET to be monitored in real time using confluent cell monolayers. This approach was complemented by our use of a new clone of HEK293-H188-C24 cells that stably express H188 (27) and that were transiently transfected with select recombinant GPCRs. The GPCR agonists tested included the NPY2R agonist PYY(3-36) that inhibits cAMP production or the GPCR agonists GLP-1, exendin-4, glucagon, and GIP that stimulate cAMP production (Fig. 1). Also tested were GPCR antagonists previously reported to be selective for the GluR (des-His 1 -[Glu 9 ]glucagon, LY2409021, and MK 0893) or the GLP-1R (exendin (9 -39)) or NPY2R (BIIE0246) (Fig. 1) (29). Finally, we tested for novel dual or triagonist properties of GGP817, a synthetic hybrid peptide that contains full-length glucagon to which a 12-amino acid C-terminal fragment of PYY is fused at the C terminus of glucagon (Fig. 1). Importantly, control experiments verified that GLP-1, exendin-4, glucagon, and GGP817 failed to alter levels of cAMP in HEK293 cells that expressed H188 but that were not transfected with recombinant GPCRs (Fig. S1, a-f).

Glucagon and LY2409021 both target glucagon and GLP-1 receptors
Using HEK293 cells that stably express the rat glucagon receptor (HEK293-GluR), and that were virally transduced with H188, it was confirmed that glucagon acts as a GluR agonist to raise levels of cAMP in a dose-dependent manner (Fig. 2, a 1 -a 3 ) (30). LY2409021 dose-dependently blocked this action of glucagon (Fig. 2, b 1 -b 3 ), but it exerted no effect when tested alone (Fig. S2a).
EC 50 and IC 50 values for glucagon agonist or LY2409021 antagonist action at the rat GluR were 400 pM and 1.8 M, respectively (Fig. 2, a 3 and b 3 ). Consistent with a prior report concerning glucagon action at the GLP-1R (5), high concentrations of glucagon (1-100 nM) raised levels of cAMP in HEK293-GLP-1R cells expressing the human GLP-1R (Fig. 2, c 1 -c 3 ). This action of glucagon was also blocked by LY2409021 (Fig. 2, d 1 -d 3 ), whereas LY2409021 exerted no effect alone at the GLP-1 R (Fig. S2b). EC 50 and IC 50 values for glucagon agonist and LY2409021 antagonist actions at the GLP-1R were 4.9 nM and 1.2 M, respectively (Fig. 2, c 3 and d 3 ). LY2409021 antagonism of glucagon action at the GLP-1R was also measured using HEK293-H188 C24 cells transiently transfected with the human GLP-1R (Fig. S3, a-d). Although glucagon acted with lower potency than GLP-1 at the GLP-1R, it exhibited full agonist properties because 100 nM glucagon stimulated a 60% maximal change of the H188 FRET ratio (⌬FRET) when it was tested at either the GluR or the GLP-1R (cf. Fig. 2, a 1 and c 1 ).

LY2409021 is a GIP-R antagonist but fails to block adenosine action at A2 B receptors
LY2409021 consistently blocked peptide hormone agonist actions at all family B GPCRs tested. This was the case not only for the GluR and GLP-1R, but also for the GIP-R when testing the action of GIP using HEK293-GIP-R cells that stably express GPCR agonist and antagonist actions the rat GIP-R (Fig. S4, a-c) (32). However, LY2409021 failed to block cAMP-elevating actions of adenosine that are mediated by its A2 B receptor that is a family A GPCR and that is an endogenous GPCR expressed in parental HEK293 cells (Fig. S5, a and b) (33). Instead, this action of adenosine was blocked by the A2 B receptor antagonist GS 6201 (Fig. S5, c and d). Thus, LY2409021 exhibited high selectivity for family B GPCRs, as expected if it acts as an allosteric inhibitor in which it specifically interacts with a conserved binding motif within the GluR, GLP-1R, and GIP-R (see below; Fig. 6).

LY2409021 blocks GGP817 agonist action at the GluR and GLP-1R
Here, we designed the synthetic hybrid peptide GGP817 with the expectation that it might act as a triagonist because it incorporates amino acid residues found within glucagon, GLP-1, and PYY. For example, because glucagon is an agonist at the GluR, and also an agonist at the GLP-1R, GGP817 might possess triagonist properties in which it activates the GluR and GLP-1R in addition to NPY2R. When tested using HEK293-GluR cells, GGP817 exhibited agonist action at the GluR (Fig. 4, a 1 -a 3 ), and this effect was blocked by LY2409021 (Fig. 4, b 1 -b 3 ). However, GGP817 also exhibited agonist action at the GLP-1R in HEK293-GLP-1R cells (Fig. 4, c 1 -c 3 ), and this effect was blocked by LY2409021 (Fig. 4, d 1 -d 3 ). Note that when it was tested using these GluR-and GLP-1R-expressing HEK293 cells, GGP817 was less potent compared with the naturally occurring receptor ligands glucagon and GLP-1. Also note that GGP817 failed to exhibit agonist action at the GIP-R (Fig. S4d).
The EC 50 value for GGP817 agonist action at the GluR was 183 nM (Fig. 4a 3 ), and it was 61 nM for the GLP-1R (Fig. 4c 3 ). Still, when GGP817 was tested at a saturating concentration (3,000 nM), it acted as a full agonist at both the GluR and the GLP-1R. This was established by monitoring its ability to stimulate a 60% maximal ⌬FRET in the cAMP assay. Of interest, the antagonist potency of LY2409021 at the GluR and GLP-1R in assays using GGP817 resembled that which was observed when testing LY2409021 in assays using glucagon as a stimulus for GluR or GLP-1R activation (cf. Figs. 2, b 3 and d 3 , and 4, b 3 and d 3 ). When comparing antagonist actions of LY2409021 to block GGP817 agonist action at the GluR (Fig. 4b 3 ) or GGP817 agonist action at the GLP-1R (Fig. 4d 3 ), the IC 50 values were 908 nM and 1.6 M, respectively.

Antagonist properties of MK 0893 at the GluR and GLP-1R
MK 0893 is a GluR antagonist that is structurally related to LY2409021 (26), and we confirmed that it acted as an antagonist of glucagon agonist action at the GluR in HEK293-

GPCR agonist and antagonist actions
GluR cells (Fig. S6a). However, for HEK293-GLP-1R cells, the agonist actions of glucagon and GLP-1 were also inhibited by MK 0893 (Fig. S6, b and c), thereby demonstrating that MK 0893 is not a specific GluR antagonist. Similar to LY2409021, MK 0893 exerted no effect on its own in these GluR-and GLP-1R-expressing HEK293 cell lines (Fig. S6, d  and e). When comparing the actions of LY2409021 and MK 0893, it was also established that both antagonists inhibited the agonist actions of glucagon and GLP-1 in rat INS-1 832/13 insulinoma cells (see below; Fig. 7) that are a model system for studies of pancreatic ␤-cell function (34) and that co-express endogenous GluR and GLP-1R. Collectively, these findings established the promiscuous nature with which LY2409021 and MK 0893 exerted antagonist effects at family B GPCRs.

GPCR agonist and antagonist actions
HEK293-GluR cells (Fig. S7, a-c). Because Ex(9 -39) failed to antagonize the agonist actions of glucagon and GGP817 at the GluR in HEK293-GluR cells (Fig. 7, d and e), Ex(9 -39) is a specific antagonist of GLP-1R activation and is without significant effect at the GluR.

Computational modeling of LY2409021 binding to the GluR and GLP-1R
Previous structural biology studies used small molecule allosteric inhibitors of the GluR (NNC0640 and MK-0893) (26,35) or the GLP-1R (NNC0640 and PF-06372222) (36) to identify their binding sites located at the interfaces between transmembrane helices (TMs) 5-7 of these receptors. Using computational receptor modeling, we sought to predict the potential binding site where LY2409021 interacts with the GLP-1R. In this regard, we made use of available data concerning binding of MK 0893 to the GluR (26). Our docking studies using homology models of family B receptors predict a conserved mode of binding for LY2409021 at the rat GLP-1R and also GluR ( Fig. 6, a-c). However, it should be noted that future receptor mutagenesis studies would be required to obtain full validation of these predicted interactions.
For both the GLP-1R and GluR, the carboxyl group of LY2409021 is predicted to form a salt bridge with the receptor's R2.47b amino acid residue located in TM2 (Fig. 6, b and c), as indicated using the Wooten numbering scheme (37). This residue in the GluR is already established to be a significant factor determining direct receptor interactions with downstream G s proteins (36). The hydrophobic surface of TM5 has several hydrophobic aliphatic residues that are conserved in family B GPCRs and that are likely to interact with the trimethyl group of LY2409021, as inferred from the structure of MK 0893. These are I5.58b, I/V6.52b, L5.65b, and V/L5.61b. The aromatic benzene group of LY2409021 is predicted to form hydrophobic

GPCR agonist and antagonist actions
interactions with L6.43b, present in both the GLP-1R and the GluR. L7.56b is also a crucial contact for the aromatic ring of LY2409021, and it is conserved among family B GPCRs. The two remaining aromatic groups of LY2409021 are predicted to sandwich around the aliphatic chain of K6.40b located in the GLP-1R (Fig. 6b) or R6.40b in the GluR (Fig. 6c). S6.41b forms a hydrogen bond with the only hydrogen donor of LY2409021, and the antagonist's trifluoro group, pointing outwards, probably interacts with the hydrophobic region of the membrane bilayer.

Computational modeling of GLP-1 and glucagon binding to the GLP-1R
Amino acid residues of the human GLP-1R that mediate binding of GLP-1 were previously identified by crystallography and cryo-EM (38,39). Furthermore, the detailed nature of glucagon binding to the extracellular domain (ECD) of the GluR was established by determination of the crystal structure of the GluR in complex with a glucagon analogue (40). The main recognition interface for GLP-1 within the ECD of the GLP-1R is governed by hydrophobic ligand-receptor interactions (Fig. 6d)
Site-directed receptor mutagenesis was previously used to investigate interactions of the agonist exendin-4 with the GLP-1R (38). Amino acid residues that were mutated correspond to Tyr-69, Tyr-88, Leu-89, Pro-90, and Leu-123 (Fig. 6f). In binding assays, the P90A and L123A mutations resulted in loss of receptor affinity for exendin-4 (38). This is significant because these same amino acids are predicted by our model to constitute a binding interface for Ex (9 -39) to the GLP-1R. In contrast, Y69A, Y88A, and L89A mutations were reported to reduce GLP-1R expression (38).
When comparing the amino acid sequence of Ex(9 -39) with that of GLP-1 and glucagon (Fig. 1), an important difference is the presence of Lys-19 in Ex(9 -39) instead of Val-36 in GLP-1 or Met-27 in glucagon. Lys-19 may form a salt bridge with Asp-67 of the GLP-1R ECD, thereby strengthening its binding so that it acts as a strong competitive antagonist (Fig. 6f). Additional interactions with Tyr-69 and Leu-123 of the ECD are predicted to be conferred by Phe-14, Ile-15, Leu-18, and Lys-19 of Ex(9 -39) (Fig. 6f). In this regard, Asp-67 was reported to be an essential residue for peptide binding to the GLP-1R (43,44).

des-His 1 -[Glu 9 ]glucagon blocks glucagon action at the GluR but is without effect at the GLP-1R
Seeking a more specific GluR antagonist, we investigated the possible usefulness of GluR antagonist des-His 1 -[Glu 9 ]glucagon (23) to differentiate between actions of glucagon that are mediated by the GluR or GLP-1R. In assays using HEK293-GluR cells, des-His 1 -[Glu 9 ]glucagon fully suppressed the agonist action of glucagon (Fig. 7, a 1 -a 3 ). Complete block was measured at 3.0 M, and the IC 50 value was 0.2 M. However, for HEK293-GLP-1R cells, des-His 1 -[Glu 9 ]glucagon did not significantly alter the agonist actions of GLP-1 or glucagon, although a small inhibitory trend was noted at high concentrations of the antagonist (Fig. 7, b 1 -b 3 and c 1 -c 3 ). Thus, compared with LY2409021 and MK 0893, the antagonist action of des-His 1 -[Glu 9 ]glucagon was highly selective for the GluR.

Glucagon dually regulates the GluR and GLP-1R in a pancreatic ␤-cell line
Moens et al. (45) originally reported that glucagon secreted from islet ␣-cells exerted an intra-islet paracrine hormone effect to stimulate the GluR and GLP-1R on ␤-cells. More recently, it was reported that under conditions of stress, ␣-cells secrete GLP-1 in addition to glucagon (46 -53). Thus, there is new interest to identify intra-islet actions of glucagon and GLP-1, and in this regard, it is necessary to validate the receptor specificity of commonly used GluR and GLP-1R agonists and antagonists.
Here, we used the INS-1 832/13 ␤-cell line as a model system with which to investigate the promiscuous nature of GPCR agonist and antagonist action, while also evaluating the relative importance of glucagon agonist action at the ␤-cell GluR versus GLP-1R. Consistent with findings obtained using HEK293 cells, LY2409021 and MK 0893 each reduced the dose-dependent actions of glucagon (Fig. 8, a 1 -a 3 ) and GLP-1 (Fig. 8, b 1 -b 3 ) to raise levels of cAMP in INS-1 832/13 cells. However, these antagonists had little or no effect on their own (Figs. S2c and S6f). Such findings are expected because INS-1 832/13 cells co-express the GluR and GLP-1R. However, because of their promiscuous GPCR antagonist actions, LY2409021 and MK 0893 are not suitable tools with which to define the relative importance of glucagon action at the GluR versus GLP-1R in ␤-cells.
The GLP-1R antagonist Ex(9 -39) reduced but did not fully block the agonist action of glucagon to raise levels of cAMP in INS-1 832/13 cells (Fig. 8c 1 ). In contrast, Ex(9 -39) fully blocked the action of GLP-1 (Fig. 8c 2 ), while having no effect on its own  (Fig. 8c 3 ). These findings are expected if the action of glucagon is dually mediated by the GluR and the GLP-1R, whereas the action of GLP-1 is mediated solely by the GLP-1R. The key to interpretation of these findings is the fact that Ex(9 -39) failed to alter the agonist action of glucagon at the GluR in HEK293-GluR cells (Fig. S7d). However, Ex(9 -39) did block GLP-1 and glucagon agonist action at the GLP-1R in HEK293-GLP-1R cells (Fig. 5, a 1-3 and c 1-3 ). Thus, for INS-1 832/13 cells, we interpret the Ex(9 -39)-insensitive agonist action of glucagon to be mediated by the GluR, whereas glucagon acts at the GLP-1R in an Ex(9 -39)-sensitive manner.

GPCR agonist and antagonist actions
A dual agonist action of glucagon at the ␤-cell GluR and GLP-1R is further supported by studies using INS-1 832/13 cells treated with the GluR antagonist des-His 1 -[Glu 9 ]glucagon. These studies

GPCR agonist and antagonist actions
revealed that des-His 1 -[Glu 9 ]glucagon reduced but did not fully block the agonist action of glucagon (Fig. 9a). In contrast, it failed to alter the agonist action of GLP-1 (Fig. 9b), nor did it exert any effect on its own (Fig. 9c). Such outcomes are interpretable based on findings obtained using HEK293 cells in which it was established that des-His 1 -[Glu 9 ]glucagon acted as a specific GluR antagonist with no detectable action at the GLP-1R (Fig. 7). Thus, for INS-1 832/13 cells, treatment with des-His 1 -[Glu 9 ]glucagon revealed a residual agonist action of glucagon that was most likely mediated by the GLP-1R (Fig. 9a).
Findings summarized above for INS-1 832/13 cells led us to predict that the des-His 1 -[Glu 9 ]glucagon-insensitive action of glucagon should be eliminated by the GLP-1R-specific antagonist Ex (9 -39). To test this prediction, INS-1 832/13 cells were treated with des-His 1 -[Glu 9 ]glucagon in the absence or presence of Ex (9 -39). Simultaneously, the agonist action of glucagon was monitored. For cells treated only with des-His 1 -[Glu 9 ]glucagon, there existed a residual-stimulatory action of glucagon mediated by the GLP-1R, and this residual action of glucagon was nearly eliminated when cells were co-treated with . c 1 -c 3 , Ex(9 -39) (3 or 30 nM) reduced but did not block the action of 10 nM glucagon (c 1 ), whereas it fully inhibited the action of 3 nM GLP-1 (c 2 ), while having no effect on its own when tested at 1000 nM (c 3 ).

GPCR agonist and antagonist actions
des-His 1 -[Glu 9 ]glucagon and Ex(9 -39) (Fig. 9d). An additional example from an identical replicate experiment is provided in which the relative contributions of the GluR and GLP-1R to glucagon agonist action are illustrated (Fig. 9e). In summary, this analysis using specific GluR and GLP-1R antagonists indicated that the Ex(9 -39)-sensitive component of glucagon agonist action in INS-1 832/13 cells resulted from the stimulatory action of glucagon at the GLP-1R.

GPCR agonist and antagonist actions
Finally, we investigated a potential inverse agonist action of Ex(9 -39) originally reported for a mouse ␤-cell line (54). Inverse agonism should be detectable as a decrease of the H188 FRET ratio that signifies reduced levels of cAMP, yet no such ⌬FRET was measured in response to Ex(9 -39) using HEK293-GLP-1R cells (Fig. S7f). Studies with INS-1 832/13 cells transduced with H188 also provided no evidence for inverse agonism (Fig. 9f). However, weak inverse agonism was measured using an AKAR3 FRET reporter that is highly sensitive to cAMP signaling in this assay (55) and that responds to GLP-1 (Fig. 9, g  and h).

GGP817 is an NPY2R agonist that reduces levels of cAMP
Because GGP817 contains a C-terminal fragment of PYY  fused to glucagon, we sought to determine whether it might bind to NPY2R to activate G i proteins, thereby reducing levels of cAMP. To test this, HEK293-H188-C24 cells were transfected with human NPY2R (56) so that cAMP-lowering actions of GGP817 and NPY2R agonist PYY(3-36) could be compared. In view of the fact that HEK293-H188-C24 cells have low basal adenylyl cyclase activity and low basal levels of cAMP (27), we used adenosine (acting at endogenous A2 B receptors) or forskolin (acting at adenylyl cyclase) to initially enhance cyclase activity, thereby raising levels of cAMP prior to inhibitory agonist treatment. In this manner, an NPY2R-mediated counterregulatory action of GGP817 to lower levels of cAMP was evaluated.

FRET assays reveal novel features of family B GPCR agonist and antagonist action
The primary goal of this study was to apply new high-throughput FRET assay technology to real-time analysis of GPCR agonist and antagonist action in living cells so that a determination could be made concerning the specificity with which GluR and GLP-1R agonists and antagonists act to control intracellular levels of cAMP. To test for physiological significance, this analysis included not only an assessment of agonist or antagonist action at recombinant GPCRs expressed in HEK293 cells, but also endogenous GPCRs expressed in INS-1 832/13 cells.
These approaches revealed that the GluR antagonists LY2409021 and MK 0893 exerted significant off-target effects to block agonist action at the GLP-1R and GIP-R. This analysis also demonstrated that des-His 1 -[Glu 9 ]glucagon is a specific GluR antagonist in that it lacks any capacity to block glucagon or GLP-1 agonist action at the GLP-1R. Furthermore, Ex(9 -39) was shown to be a specific GLP-1R antagonist because it blocked the actions of glucagon and GLP-1 at the GLP-1R, whereas it failed to block the action of glucagon at the GluR. Findings reported here also established the capacity of glucagon to exert a dual agonist effect to simultaneously activate endogenous glucagon and GLP-1 receptors expressed in INS-1 832/13 cells.
We also report that the hybrid peptide GGP817 acted as a triagonist at the GluR, GLP-1R, and NPYR2. These findings are understandable because GGP817 contains full-length glucagon fused at its C terminus to a 12-amino acid C-terminal fragment of PPY. Triagonist properties of GGP817 were validated in assays that monitored its GluR-and GLP-R-mediated actions to raise levels of cAMP, or in assays that monitored its NPY2Rmediated action to lower levels of cAMP. Such findings establish GGP817 to be a prototype triagonist peptide with biological actions that are of potential medical relevance to the treatment of T2D and obesity.

Pharmacological properties of GluR, GLP-1R, and NPY2R agonists and antagonists
In FRET assays conducted in a 96-well format using the biosensor H188, the agonist action of glucagon at the GluR was monitored in real time over a time span of 0 -400 s. As expected, this action of glucagon was inhibited by the allosteric GluR inhibitors LY2409021 and MK 0893. The action of glucagon was also blocked by the orthosteric GluR antagonist des-His 1 -[Glu 9 ]glucagon, but not by the orthosteric GLP-1R antagonist Ex (9 -39). Neither GLP-1 nor Ex-4 exerted stimulatory effects at the GluR. However, GGP817 was a full agonist (EC 50 183 nM) at the GluR, as judged by the maximal ⌬FRET moni-  9 ]glucagon (1 M) and Ex(9 -39) (300 nM). f, no inverse agonist action of Ex(9 -39) (3,000 nM) to lower levels of cAMP was measurable for INS-1 832/13 cells transduced with the FRET reporter H188 that has less affinity for cAMP compared with PKA regulatory subunits. g, Ex(9 -39) exerted a weak inverse agonist effect to lower levels of cAMP for INS-1 832/13 cells transduced with the FRET reporter AKAR3 that is a substrate for PKA-mediated phosphorylation, and that indirectly monitors binding of cAMP to its high-affinity receptor on PKA. h, AKAR3 expressed in INS-1 832/13 cells reported GLP-1-stimulated cAMP production, but the dynamic range of AKAR3 was limited compared with H188. Note that when levels of cAMP rise, H188 reports a decrease of FRET that is displayed here as the ⌬485/538 ratio, whereas AKAR3 reports an increase of FRET displayed here as the ⌬535/485 ratio.

GPCR agonist and antagonist actions
tored in this assay. This agonist action of GGP817 at the GluR was blocked by LY2409021 but not by Ex (9 -39).
A nonconventional agonist action of glucagon to stimulate the GLP-1R was also measured when testing high concentrations of glucagon. However, the EC 50 value for glucagon action at the GLP-1R was 4.9 nM, whereas that of GLP-1 was 65 pM. Still, high concentrations of glucagon did produce a full agonist effect. This action of glucagon at the GLP1-R was blocked by LY2409021, MK 0893, and Ex(9 -39), whereas des-His 1 -[Glu 9 ]glucagon was without effect. Further analysis revealed that Ex-4 (EC 50 30 pM) and GGP817 (EC 50 61 nM) exerted full agonist actions at the GLP-1R, and that these effects were also inhibited by LY2409021, MK 0893, and Ex (9 -39). Evidence for an inverse agonist action of Ex(9 -39) was obtained only in studies of INS-1 832/13 cells transduced with AKAR3.

Dual GluR-and GLP-1R-mediated actions of glucagon in INS-1 832/13 cells
Our interest in the pharmacology of family B GPCRs was prompted by a new appreciation that glucagon and GLP-1 act as paracrine hormones within the islets of Langerhans to stimulate glucagon and GLP-1 receptors located on islet ␤-cells. Note that BIIE0246 blocked the cAMP-lowering action of GGP817 under conditions in which cells were also treated with adenosine (2 M). b 3 , negative control demonstrating that GGP817 was without effect in HEK293-H188-C24 cells transfected with the empty vector (EV) and treated with adenosine. c 1 -c 3 , FRET data (a 1 ), box-and-whisker plot (a 2 ), and dose-response plot (a 3 ) summarizing findings in which PYY(3-36) (IC 50 18 nM) counteracted the cAMP-elevating action of adenosine (2 M) in HEK293-H188-C24 cells transfected with NPY2R. d 1 and d 2 , GGP817 reduced levels of cAMP in HEK293-H188-C24 transfected with NPY2R and treated with forskolin (d 1 ), whereas it was without effect in cells transfected with the EV (d 2 ). d 3 , a cAMP-lowering action of PYY  was also measured in HEK293-H188-C24 transfected with NPY2R and treated with forskolin. For a 2 and c 2 , the * indicates a p value of Ͻ 0.01, one-way ANOVA with post hoc Tukey when comparing the action of adenosine in cells not treated (vehicle control) or treated with the indicated concentrations of GGP817 or PYY .
Normally, glucagon is secreted from islet ␣-cells that are adjacent to these ␤-cells. However, under conditions of stress, as occurs in T2D, ␣-cells undergo a phenotypic switch so that they are able to synthesize and secrete GLP-1 in addition to glucagon (46 -53). Thus, local intra-islet paracrine actions of glucagon and GLP-1 may serve as important determinants of ␤-cell function, and in fact, one recent study using glucagon and GLP-1 receptor knockout mice supports this concept (52).
Studies reported here indicate that high concentrations of glucagon released from ␣-cells might act within the volumerestricted microenvironment of the islet to stimulate the ␤-cell GLP-1R despite a substantially lower affinity of glucagon for the GLP-1R. Superimposed on this nonconventional action of glucagon is its conventional action to activate the GluR expressed on ␤-cells (45,53). These actions of glucagon are likely to be complemented by the action of intra-islet GLP-1 released from ␣-cells to stimulate the ␤-cell GLP-1R (52). The net effect is that glucagon and GLP-1 coordinately stimulate cAMP production so that ␤-cell functionality is enhanced in the healthy state or "repaired" under conditions of T2D (57)(58)(59)(60)(61)(62).
New findings using the INS-1 832/13 ␤-cell line are summarized as follows: 1) glucagon is a dual ␤-cell GluR and GLP-1R agonist when tested at high concentrations that may exist in islets; 2) LY2409021 and MK 0893 are promiscuous antagonists at the ␤-cell GluR and GLP-1R; 3) Ex(9 -39) is a specific GLP1-R antagonist, but this effect is not restricted to antagonism of GLP-1 action because it also blocks glucagon action at the GLP-1R; and 4) des-His 1 -[Gluc 9 ]glucagon is a specific GluR antagonist.
It is important to note that interpretation of prior in vivo studies using systemically administered GluR and GLP-1R antagonists are clouded by uncertainties concerning the specificity with which these antagonists target the GluR or GLP-1R. LY2409021 is considered to be a selective GluR antagonist in vivo (22,63,64), whereas Ex(9 -39) is considered to be a selective GLP-1R antagonist in vivo (65)(66)(67)(68). However, we find that LY2409021 acts with similar IC 50 values as an antagonist of glucagon, GLP-1, and GIP action at the GluR, GLP-1R, and GIP-R. Furthermore, we find that Ex(9 -39) acts not simply to block GLP-1 action at the GLP-1R but also glucagon action at the GLP-1R.

Computational modeling predicts conserved features of ligand binding to the GluR and GLP-1R
As is the case for all GPCRs, the GLP-1R consists of an extracellular N-terminal domain (N-domain), an intracellular C-terminal region, and seven TMs that are connected to each other by three extracellular loops (ELs) and also three intracellular loops (10,69). The ELs and TMs form the J-domain of receptor that binds GLP-1. GLP-1 interacts with the GLP-1R according to a two-step model. In the first step, the C-terminal region of GLP-1 interacts with the receptor's N-domain, thereby providing an "affinity-trap" that facilitates a second step in which the N-terminal region of GLP-1 binds to the J-domain so that the receptor is activated. Although exendin(9 -39) contains a C terminus that allows its selective binding to the N-domain, it is unable to interact with the J-domain, thereby explaining its ability to act as orthosteric antagonist of GLP-1R activation.
The mode of binding of the allosteric inhibitor LY2409021 to family B GPCRs is predicted here on the basis of computational receptor modeling, and it is understandable based on prior models that explain how conformational changes of the TMs located within the J-domain allow functional interactions of family B receptors with their downstream G proteins. Allosteric inhibitors of GPCR signaling reduce agonist binding and receptor activation by intercalating between the TMs or by binding outside the TMs. The latter mechanism is suggested from the crystal structure complex formed between the GluR and its allosteric antagonist MK 0893 that is structurally related to LY2409021 (26). It is also suggested by crystal structure complexes formed between the GLP1-R and its allosteric antagonists NNC0640 and PF-06372222 (36). Specifically, these allosteric inhibitors interact with a structurally conserved binding pocket located outside TM5, TM6, and TM7 of the receptor, thus restricting the outward helical movement of TM6, which is essential for receptor activation and G protein coupling.
Because we found that LY2409021 blocked glucagon action at both the GluR and the GLP-1R, while also blocking GLP-1 action at the GLP-1R, and even GIP action at the GIP-R, we propose that LY2409021 recapitulates the allosteric inhibitor action of MK 0893 by interacting with the conserved binding pocket that is present within each of these receptors and that is located outside TM5, TM6, and TM7 (Fig. 6). Such findings are of potential clinical relevance in view of the fact that they may explain a promiscuous antagonist action of LY2409021, and other allosteric inhibitors related in structure to it, at multiple GPCRs, rather than exerting pure GluR selectivity. To what extent these allosteric inhibitors have any therapeutic value remains to be determined. Weight gain, increased blood pressure, and hepatic abnormalities were reported for patients administered LY2409021 and MK 0893 (63,70), and both compounds have been withdrawn from clinical trials.
It is also notable that Ex(9 -39) acted not simply as an antagonist of GLP-1 action at the GLP-1R, but also as an antagonist of glucagon action at the GLP-1R. These in vitro findings are important to the interpretation of prior in vivo studies in which stimulatory actions of glucagon at the GLP-1R were not seriously considered. It seems that systemically administered Ex(9 -39) might block glucagon agonist action at the GLP-1R, not only in the islets, but also in other tissues that express the GLP-1R. This nonconventional action of Ex(9 -39) might be explained by its ability to competitively inhibit binding of glucagon to the extracellularly-oriented N-domain of the GLP-1R where a structurally conserved binding motif determines the selectivity of ligand-receptor interactions (Fig. 6).

Triagonist properties of GGP817 acting at the GluR, GLP-1R, and NPY2R
When considering the herein reported triagonist properties of hybrid peptide GGP817, it can be concluded that this peptide's glucagon amino acid sequence motif mediates agonist actions at the GluR and GLP-1R, whereas NPY2R agonism is conferred by the peptide's C terminus that itself contains a 12-amino acid C-terminal fragment of PYY. Although GGP817 has lower affinity for the GluR and GLP-1R compared with glucagon and GLP-1, it serves as a prototype that might be of

GPCR agonist and antagonist actions
use for future studies directed at the design of triagonists with "balanced" agonism at the GluR, GLP-1R, and NPY2R.
Because GGP817 exhibits triagonism at the GluR, GLP-1R, and NPY2R, it might exhibit desirable in vivo metabolic actions not previously reported for hybrid peptides (1)(2)(3)(4)(5)(6)(7)(8)(9). In this regard, a predicted profile of action for GGP817 would be as follows: 1) GluR agonism leading to energy expenditure and reduced fat mass; 2) GLP-1R agonism with improved ␤-cell insulin secretion and glucoregulation even in the context of increased hepatic glucose output because of GluR agonism; and 3) GLP-1R and NPY2R agonism leading to appetite suppression. Interestingly, it was recently reported that combined administration of Ex-4 and PYY  led to a synergistic suppression of appetite in mice (71). Thus, it will be important to determine whether a triagonist such as GGP817 emulates the synergistic co-agonist actions of Ex-4 and PYY  to suppress appetite through dual-stimulatory effects at hypothalamic NPY2R and GLP-1R. Just as important, GGP817 will still need to be profiled across other receptors before advancement as a therapeutic candidate.

Conclusion
Current efforts in drug discovery research for the treatment of obesity and diabetes include the testing of hybrid peptides with dual or triagonist actions at the NYP2R, GluR, GLP-1R, GIP-R, and other GPCRs of major relevance to metabolic homeostasis. Findings reported here emphasize the importance of validating receptor selectivities for dual or triagonists, especially with respect to potential actions of the glucagon moiety acting at the GLP-1R. Future research using pharmacological and genetic manipulation of family B GPCRs should take into account potential confounding off-target actions of conventional receptor antagonists, as reported here for inhibitory effects of LY2409021 at the GLP-1R.

Cell culture
The parental HEK293 cell line was obtained from the American Type Culture Collection (ATCC, Manassas, VA). HEK293 cells stably expressing the human GLP-1R at a density of 150,000 receptors/cell were obtained from Novo Nordisk A/S (Bagsvaerd, Denmark) (72). HEK293 cells stably expressing the rat GlucR at a density of 250,000 receptors/cell were obtained from T. P. Sakmar, A. M. Cypess, and C. G. Unson (The Rockefeller University) (73,74). HEK293 cells stably expressing the rat GIP-R at a receptor density that has yet to be determined were obtained from T. J. Kieffer (University of British Columbia) (75). HEK293 cells stably expressing H188 were generated by O. G. Chepurny in the Holz laboratory (27). All cell cultures were maintained in Dulbecco's modified Eagle's medium containing 25 mM glucose and supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin. Cell cultures equilibrated at 37°C in a humidified incubator that was gassed with 5% CO 2 were passaged once a week. Culture media and additives were obtained from ThermoFisher Scientific (Waltham, MA).

Cell transfection
Transient transfections were performed using Lipofectamine and Plus Reagent (from ThermoFisher Scientific) using the methodol-ogy described previously by our laboratory for HEK293 cells (76). A plasmid encoding human NPY2R (I.D. NPYR20TN00) in pcDNA3.1 was from the University of Missouri-Rolla cDNA Resource Center (Rolla, MO). HEK293 cells stably expressing H188 were obtained by G418 antibiotic resistance selection using methodology described previously (27). Adenovirus for transduction of HEK293 cells was generated by a commercial vendor (Vira-Quest, North Liberty, IA) using the shuttle vector pVQAd CMV K-NpA and the H188 plasmid provided by Prof. Kees Jalink (28) or the AKAR3 plasmid provided by Prof. Jin Zhang (77).

FRET reporter assay in a 96-well format
HEK293 cells stably expressing recombinant GPCRs were plated at 80% confluence on 96-well clear-bottom assay plates (Costar 3904, Corning, NY) coated with rat tail collagen (Collaborative Biomedical Products, Bedford, MA). Cells were then transduced for 16 h with H188 virus at a density of ϳ60,000 cells/well under conditions in which the multiplicity of infection was equivalent to 25 viral particles per cell. The culture media were removed and replaced by 200 l/well of a standard extracellular saline (SES) solution supplemented with 11 mM glucose and 0.1% BSA. The composition of the SES was (in mM): 138 NaCl, 5.6 KCl, 2.6 CaCl 2 , 1.2 MgCl 2 , 11.1 glucose, and 10 Hepes (295 mosmol, pH 7.4). Real-time kinetic assays of FRET were performed using a FlexStation 3 microplate reader equipped with excitation and emission light monochromators (Molecular Devices, Sunnyvale, CA). Excitation light was delivered at 435/9 nm (455 nm cutoff), and emitted light was detected at 485/15 nm (cyan fluorescent protein) or 535/15 nm (yellow fluorescent protein) (27,78,79). The emission intensities were the averages of 12 excitation flashes for each time point per well. Test solutions dissolved in SES were placed in V-bottom 96-well plates (Greiner Bio-One, Monroe, NC), and an automated pipetting procedure was used to transfer 50 l of each test solution to each well of the assay plate containing monolayers of these cells. The 485/535 emission ratio was calculated for each well, and the mean Ϯ S.D. values for 12 wells were averaged. These FRET ratio values were normalized using baseline subtraction so that a y axis value of 0 corresponds to the initial baseline FRET ratio, whereas a value of 100 corresponds to a 100% increase (i.e. doubling) of the FRET ratio. The time course of the ⌬FRET ratio was plotted after exporting data to Origin 8.0 (OriginLab, Northampton, MA). Origin 8.0 was also used for nonlinear regression analysis to quantify doseresponse relationships.

Statistical analyses
Raw FRET data from individual experiments are expressed as the mean Ϯ S.D. and are derived from n ϭ 12 wells for each concentration of test agent. Box and whisker plots derived from these raw data show the following: mean (solid square), 25-75% range (open box), median (line across open box), minimum and maximum values (whiskers). The repeatability of findings was confirmed by performing each experiment a minimum of three times. FRET ratio data were evaluated for statistical significance by a one-way ANOVA test with a post hoc Tukey HSD test. Comparisons of individual data sets are defined in the accompanying figure legends. A p value of Ͻ 0.05 was considered to be statistically significant.

Computational studies
Homology models of the receptors used in this study were made with Modeler 9.19. The rat GLP-1R and GluR sequences were obtained from Uniport (IDs P32301 and P30082, respectively) and aligned to the crystal structure sequences of the GLP-1R (PDB code 5VEW) (36) and GluR (PDB code 5EE7) (26), respectively. Structural models of the rat GLP-1R bound to GLP-1, glucagon, and Ex(9 -39) were made using as a template the GLP-1R ECD-GLP-1 crystal structure (PDB code 3IOL) (38) in the first case, and the full-length GLP-1R-exendin-P5 cryo-EM structure (PDB code 6B3J) (42) for the latter two, using the rat GLP-1R sequence (Uniport ID P32301). The conserved S1.50b in TM1, H2.50b in TM2, E3.50b in TM3, W4.50b in TM4, N5.50b in TM5, G6.50b in TM6, and G7.50b in TM7 were used as reference points in all the sequence alignments. Superscripts denote the single most conserved residue among family B GPCRs, which is designated as X.50b according to the Wootten numbering scheme (37). The overall stereochemical quality of the homology models was evaluated by the discrete optimized energy and visual inspection. For evaluation of LY2409021 and MK 0893 binding to GPCRs, ligands were docked in the extracellular allosteric binding site of the receptor using VinaSH (80), a version of AutoDock Vina (81), which also takes into account sulfur and halogen bonding. The AutoDock-Tools program was used to prepare the corresponding PDBQTprotein file for the docking. The receptor was considered rigid during the docking process, although the ligand was allowed to be flexible. Docking calculations in both cases were performed using a grid box with dimensions of 26 ϫ 26 ϫ 26 Å, located around the binding site of MK-0893 (PDB code 5EE7) and a search space of 20 binding modes, and the search parameter was set to 5. The best docking pose was chosen on the basis of the lowest energy docking score.

Data availability
The data that support the findings of this study are available from the corresponding authors upon reasonable request.