Glucagon-Like Peptide-1 and Its Class B G Protein–Coupled Receptors: A Long March to Therapeutic Successes

1. Effect on Glucose Homeostasis.

The insulinotropic activity of GLP-1 is strictly glucose-dependent mediated through its receptor at the membrane of pancreatic β cells (Kreymann et al., 1987; Mojsov et al., 1987; Holst, 2007). GLP-1 could not stimulate insulin secretion at low levels of glucose in humans (Kreymann et al., 1987). GLP-1R belongs to the class B GPCR subfamily whose members include receptors for peptidic hormones such as glucagon, secretin, GIP, etc. The binding of GLP-1 to its receptor activates heterotrimeric Gs protein, which subsequently elicits adenylate cyclase activity, resulting in cAMP formation (Gromada et al., 1998, 2004). The increased level of cAMP in turn leads to activation of protein kinase A (PKA) and the cAMP-regulated guanine nucleotide exchange factor II (Ozaki et al., 2000). In the presence of high levels of glucose, GLP-1 has an effect on ATP-sensitive (K_ATP) or voltage-gated potassium and calcium channels, resulting in membrane depolarization and Ca²⁺ release from both internal and extracellular stores. Increased Ca²⁺ together with cAMP will then promote exocytosis of vesicles containing insulin (Prentki and Matschinsky, 1987; Renstrom et al., 1997). This glucose-dependent insulinotropic action of GLP-1 involves the glucose transporter, metabolic ADP/ATP ratio, K_ATP inhibition, Ca²⁺ channel opening, and, ultimately, insulin secretion (Gromada et al., 2004; Dyachok et al., 2006).

Another major activity of GLP-1 to reduce blood glucose relates to the suppression of glucagon secretion from α cells of the endocrine pancreas (Gromada and Rorsman, 2004). Glucagon is a major hyperglycemic hormone, in addition to epinephrine, and its release is reciprocally correlated with insulin secretion in glucose oscillation to mobilize hepatic glucose in the fasting state, thereby helping to ensure the maintenance of normoglycemia. In T2DM, both fasting hyperglucagonemia and exaggerated glucagon responses, which most likely contribute to the hyperglycemia of patients, were observed (Shah et al., 2000; Toft-Nielsen et al., 2001). Interestingly, only GLP-1, but not GIP, inhibits glucagon secretion (Nauck et al., 1993b). However, the detailed mechanism(s) by which GLP-1 suppresses glucagon secretion remains unclear. Because the levels of GLP-1 mRNA detected in α cells varied between none and 20% of a cell population, it is generally thought that local elevated insulin and somatostatin in response to GLP-1 stimulation are capable of suppressing glucagon secretion in α cells (Orskov et al., 1988; Heller et al., 1997; de Heer et al., 2008; Godoy-Matos, 2014). Nonetheless, in type 1 diabetic patients with absent β cell activity who lack insulin and somatostatin, GLP-1 could still reduce glucose concentrations, suggesting a direct suppression of glucagon secretion (Creutzfeldt et al., 1996; Gromada and Rorsman, 2004).

2. Effect on Gastric Emptying.

GLP-1 also has an important inhibitory activity on gut motility and gastrointestinal secretion (Wettergren et al., 1993; Nauck et al., 1997). It not only inhibits meal-induced pancreatic secretion, but also the gastric emptying process in humans (Wettergren et al., 1993). Its suppression of gastrin-induced acid secretion was demonstrated by injection of GLP-1 and/or peptide YY. Both peptides are released from L cells in the ileal mucosa of healthy people (Wettergren et al., 1997) and can exhibit additive effects on gastrin-stimulated acid secretion, a function of unabsorbed nutrients in the ileum (Holst, 1997). It was subsequently shown that this inhibitory action of GLP-1 is mediated via a vagal pathway (Wettergren et al., 1994). This ileal-brake activity of GLP-1 was further demonstrated using the GLP-1R antagonist exendin_9–39, and therefore, is believed to have physiologic relevance (Schirra et al., 2006; Maljaars et al., 2008).

3. Effect on Food Intake.

Another physiologic function of GLP-1 concerns inhibition of food intake that may have therapeutic value for body weight reduction. Whether this is related to its ileal brake effect is still debated. At least two neural mechanisms, central and peripheral, are involved in GLP-1 suppression of appetite and food intake. GLP-1 is expressed in the neurons of the brain stem, and GLP-1R is present in the hypothalamic areas that control energy homeostasis and food intake, including the arcuate nucleus, paraventricular nucleus, and dorsomedial nucleus (Jin et al., 1988; Kanse et al., 1988; Zheng et al., 2015). Intracerebroventricular injection of GLP-1 inhibits food intake in rats, and this activity is blocked by exendin_9–39 (Tang-Christensen et al., 1996; Turton et al., 1996), or by the arcuate nucleus-damaging reagent, monosodium glutamate (Tang-Christensen et al., 1998). In contrast, GLP-1 released by L cells of the intestine after a meal inhibits gut mobility and gastric emptying, allowing nutrients in the ileum to reduce food intake (Read et al., 1994). Indeed, infusion of GLP-1 into normal human subjects significantly enhances satiety and decreases food intake (Flint et al., 1998). Consistent findings have shown that GLP-1R agonism promotes weight loss and improves glucose homeostasis in rodents, monkeys, and humans (Verdich et al., 2001; Barrera et al., 2011), and such weight-reducing properties have also been well-documented for two marketed GLP-1 mimetics, exenatide (exendin-4) and liraglutide (Moretto et al., 2008; Astrup et al., 2009; Norris et al., 2009; Lean et al., 2014). Although the exact mechanism mediating reduced food intake by peripherally administered GLP-1 has yet to be elucidated, it may involve signals generated by GLP-1 binding to its receptors on neurons in the gastrointestinal tract or hepatoportal bed (Burcelin et al., 2001; Holst, 2007).

4. Effect on Cardiovascular Activity.

Recently, there is increasing evidence suggesting that GLP-1 may play a crucial role in the cardiovascular system (Grieve et al., 2009; Ussher and Drucker, 2014). GLP-1R is widely expressed in the heart and blood vessels, such as vascular smooth muscle, cardiomyocytes, endocardium, and coronary endothelium/smooth muscle, in both rodents and humans (Campos et al., 1994; Wei and Mojsov, 1995; Bullock et al., 1996). In an early study, treatment of adult rat cardiac myocytes with GLP-1 increased cAMP levels, but did not lead to increased cardiomyocyte contractility, as would be anticipated in the heart (Vila Petroff et al., 2001). Interestingly, treatment of mouse cardiomyocytes with GLP-1_{9–36 amide} (a GLP-1 degradation product in vivo) resulted in Akt activation, extracellular regulated kinase (ERK) phosphorylation, and reduced apoptosis induced by hypoxia or hydrogen peroxide stress, implying an unconventional action of GLP-1 (Ban et al., 2010). In a rat ischemia-reperfusion model using isolated perfused heart and whole animal, GLP-1 significantly decreased infarction size, and this protection was abolished by exendin_9–39, as well as inhibitors of adenylyl cyclase, phosphatidylinositol 3-kinase (PI3K), and p42/44 mitogen-activated protein kinase (MAPK), suggesting that these pathways were involved in GLP-1–mediated cardio-protection (Bose et al., 2005a). A direct action of GLP-1 on myocardial contractility and glucose uptake in normal and postischemic isolated rat hearts was also observed (Zhao et al., 2006). GLP-1 treatment significantly increased glucose uptake and decreased left ventricular end-diastolic and developed pressures. Infusion of GLP-1 into live dogs increased myocardial glucose uptake, which could be blocked by p38 MAPK kinase or endothelial nitric oxide synthase inhibitors, pointing to a direct effect of GLP-1 in myocardium (Nikolaidis et al., 2004a; Bhashyam et al., 2010). Meanwhile, GLP-1R was detected in human coronary artery endothelial cells and umbilical vein endothelial cells, and GLP-1 or exendin-4 treatment led to nitric oxide production in both cell types, indicative of GLP-1 involvement in the vasculature (Nystrom et al., 2004; Erdogdu et al., 2010; Ishii et al., 2014). GLP-1 infusion into rats also increased heart rate and blood pressure, thereby reflecting its direct role in the heart (Barragan et al., 1996). In recent clinical trials with either GLP-1R agonists or DPP-4 inhibitors, general beneficial effects of GLP-1 on the cardiovascular system in both normal and diabetic subjects have been revealed, an interesting finding that may lead to potential new treatment of cardiovascular diseases, although these effects require a long-term validation (van Genugten et al., 2013; Avogaro et al., 2014; Ussher and Drucker, 2014; see V. Pharmaceutical Development and Therapeutics).

5. Effect on Immune Response.

GLP-1 can also regulate immune responses. Its receptor mRNA was discovered in multiple immune cell types from mice, including thymoyctes, splenocytes, bone marrow–derived cells, regulatory T cells, macrophages, and invariant natural killer T cells (Hadjiyanni et al., 2010; Hogan et al., 2011; Panjwani et al., 2013). Liraglutide treatment of patients with psoriasis, an inflammatory condition associated with metabolic diseases such as obesity, diabetes, and dyslipidemia, led to improvement of psoriasis area and severity index as well as decreased cytokine secretion from invariant natural killer T cells in a glycemic control-independent manner (Hogan et al., 2011). High-fat diet–fed mice treated with exendin-4 displayed decreased mRNA levels of the proinflammatory cytokines monocyte chemoattractant protein 1, tumor necrosis factor-α, and signal transducer and activator of transcription 3 (Koehler et al., 2009), and, in a type 1 diabetes animal model in which islets were transplanted into nonobese diabetic mice, GLP-1/gastrin treatment increased the number of transforming growth factor-β1–secreting lymphocytes and decreased IFN-γ–secreting lymphocytes with delayed onset of diabetes (Suarez-Pinzon et al., 2008).

6. Effect on Kidney Function.

Some experimental data point to the participation of GLP-1 in kidney function. Infusion of GLP-1 into healthy and obese human subjects enhanced sodium excretion, urinary secretion, and glomerular filtration rate, suggesting a renal protective effect of this peptide (Gutzwiller et al., 2004). In rats, GLP-1 was able to downregulate Na⁺/H⁺ exchanger isoform 3 (NHE3) in the renal proximal tubule, implying a potential therapeutic value for hypertension and disorders of sodium retention (Crajoinas et al., 2011). Rats administered with exendin-4 significantly improved renal function and reduced inflammation, proteinuria, and fibrosis in the kidney via a mechanism that was independent of glucose lowering (Kodera et al., 2011).

7. Effect on Nervous System.

In addition to the metabolic function in the brain, GLP-1 may also exert neuroprotective and neurotropic effects (Heppner and Perez-Tilve, 2015). The aggregation of amyloid β protein (Aβ) and the microtubule-associated protein Tau cause senile plaques and neurofibrillary tangles, resulting in loss of long-term potentiation, one of the major characteristics of Alzheimer’s disease (AD). Infusion of GLP-1 or exendin-4 into the lateral ventricles of mice decreased endogenous level of Aβ, and treatment of rat hippocampus neurons with GLP-1 and exendin-4 prevented Aβ-induced cell death (Perry et al., 2003). Intracerebroventricular dosing of GLP-1 enhanced synaptic plasticity in the hippocampus and completely reversed impairment in long-term potentiation caused by subsequent injection of Aβ (Gault and Holscher, 2008). GLP-1R was found in nigrostriatal neurons, and loss of these neurons is a feature of Parkinson’s disease (PD). GLP-1 or exendin-4 supported cell viability during hypoxic injury in primary neurons from rat cerebral cortical tissues; this activity was blocked by GLP-1_9–36 (a GLP-1R antagonist) and not observed in neurons from GLP-1^−/− mice (Li et al., 2009). Exendin-4 also maintained cell viability and reduced apoptosis caused by H₂O₂-induced oxidative stress in NSC19 neuronal cells, a spinal cord cell line with similarities to cells in the central nervous system (CNS) (Li et al., 2012b).

Obviously, GLP-1 and GLP-1R are widely expressed in many tissues, and their physiologic actions have been incrementally elucidated, especially after the extensive clinical application of GLP-1 mimetics, including GLP-1R agonists and DPP-4 inhibitors, in the last decade. Although its roles in some tissues, such as adipocytes or skeletal muscles, are still illusive, GLP-1–related therapies have clearly provided multiple benefits to millions of patients suffering from metabolic disorders, and such effects are predominantly, but perhaps not always, based on signaling pathways mediated by its receptor.

1. α-Helical C-Terminal Region.

NMR, circular dichroism (CD) spectroscopy, and X-ray crystallography studies show that GLP-1 is unstructured in aqueous solution, but adopts a helical structure in a membrane-like environment and by binding to its receptor (Thornton and Gorenstein, 1994; Neidigh et al., 2001; Runge et al., 2007; Underwood et al., 2010). This conformational change upon receptor binding is also proposed for other class B GPCR peptide ligands, including glucagon (Braun et al., 1983; Siu et al., 2013), GIP (Alana et al., 2006; Parthier et al., 2007), and exendin-4 (Runge et al., 2007, 2008) (Fig. 2), implying a conserved receptor-ligand–binding mechanism (Parthier et al., 2009). Two-dimensional NMR experiments indicate that the C-terminal region of GLP-1 (Thr¹³-Lys³⁴) is in an α-helical conformation in trifluorethanol and dodecylphosphocholine (DPC) micelles, with a less well-defined α-helical region around Gly²² (Thornton and Gorenstein, 1994; Neidigh et al., 2001; Underwood et al., 2010). NMR studies suggest that exendin-4 adopts a more well-defined single helix (Ser¹¹-Lys²⁷) than GLP-1 in DPC micelles (Neidigh et al., 2001). These NMR data are in line with ligand-bound GLP-1R ECD crystal structures showing that the GLP-1 is a kinked but continuous α-helix (Thr¹³–Val³³), whereas the truncated exendin_9–39 is a straighter helix (Leu¹⁰-Lys²⁷) than GLP-1 when bound to GLP-1R (Runge et al., 2008; Underwood et al., 2010). In the ECD-bound crystal structure, the C-terminal α-helix of GLP-1 (Ala²⁴-Val³³) is stabilized by interaction with the ECD of GLP-1R, whereas the N-terminal helix (Thr¹³-Glu²¹) does not interact with the ECD. Although it cannot be excluded that the kink in GLP-1 observed in the crystal structure is a result of crystal packing between the N-terminal part of GLP-1 (Gly¹⁰-Glu²¹) and symmetry-related ECDs, the presence of two helical segments is consistent with the SAR studies with conformationally constrained GLP-1 analogs (Miranda et al., 2008; Murage et al., 2008). Cyclization of (mutated) residues Lys¹⁶-Glu²⁰ and Met¹⁸-Ala²² by a lactam bridge (constraining these regions in an α-helical conformation) does not affect binding affinity and activity, whereas cyclization of residues Thr¹¹-Asp¹⁵ even improved potency for GLP-1R compared with the corresponding linear analogs (Miranda et al., 2008; Murage et al., 2008).

Characterization of the stability and conformational changes of full-length, truncated, and GLP-1/exendin-4 chimeric peptides upon binding to the isolated ECD of GLP-1R by far-UV CD, differential scanning calorimetry, and fluorescence spectroscopy measurements demonstrated that exendin-4 has a higher α-helical propensity than GLP-1 in solution (Runge et al., 2007). Combination of these biophysical data with pharmacological studies showed that there is a positive correlation between the stability and α-helical propensity of the ligand in solution and its affinity for the ECD of GLP-1R (Runge et al., 2007). Comparison of the crystal structures of GLP-1–bound and exendin_9–39-bound ECD provides possible explanations for the higher stability of the α-helix of exendin-4 as opposed to GLP-1 (Runge et al., 2008; Underwood et al., 2010). Although the α-helix of exendin-4 is stabilized by strong intramolecular ionic interactions between Glu¹⁶/Glu¹⁷ and Arg²⁰ and between Glu²⁴ and Lys²⁷, corresponding (Gln²³, Lys²⁶) or differentially positioned (Glu²⁷, Lys³⁴) polar residues in GLP-1 do not form such intramolecular interactions when bound to the isolated ECD of GLP1-R [Protein Data Bank (PDB): 3IOL]. The GIP-bound GIP receptor (GIPR) ECD crystal structure indicates that the α-helix of GIP is stabilized by similar (Asp¹⁵-Gln¹⁹) and alternative (Gln²⁰-Asn²⁴) intramolecular H-bond interactions (Parthier et al., 2007), whereas glucagon may also be able to form helix-stabilizing intramolecular H-bond networks (e.g., Asp²¹-Arg²⁴, Arg²³-Asp²⁷) (Siu et al., 2013). In addition to the higher α-helix propensity, the more pronounced amphiphilic nature of exendin-4 enables stronger polar and hydrophobic interactions with the ECD of GLP-1R via opposite sides of the α-helix than GLP-1 (vide infra). Moreover, exendin-4 is eight residues longer than GLP-1, and NMR studies have shown that this extended C-terminal region, comprising Ser³²-Ser³⁹, forms a stable tertiary structure that folds around Trp²⁵ in trifluorethanol and glycol (Neidigh et al., 2001). This Trp-cage conformation is, however, not observed in NMR studies in DPC micelles (Neidigh et al., 2001) and has weak electron density in the exendin_9–39-bound ECD GLP-1R crystal structure (Runge et al., 2008), suggesting that a stable Trp-cage conformation (which is absent in GLP-1) is only partially populated in the receptor-bound state of exendin-4.

2. Flexible N-Terminal Region.

NMR and X-ray crystallography studies indicate that the N-terminal regions preceding the conserved α-helix of GLP-1 (His⁷-Thr¹³), exendin-4 (His¹-Thr⁷), GIP (Tyr¹-Ile⁷), glucagon (His⁷-Thr¹³), and other class B GPCR peptide ligands are flexible in solution as well as in membrane-bound and ligand-bound states (Braun et al., 1983; Neidigh et al., 2001; Parthier et al., 2007; Runge et al., 2008; Underwood et al., 2010). The structure of the N-terminal region of GLP-1 (His⁷-Thr¹³) was not clearly elucidated in NMR studies in DPC micelles due to high conformational flexibility (Neidigh et al., 2001). In the ECD-bound GLP-1 crystal structure, this N-terminal region is unstructured, and no electron density was observed for His⁷-Glu⁹ (Underwood et al., 2010). Whereas the GLP-1R ECD crystal structures provide atomic details of the molecular interactions of the C-terminal regions of GLP-1 and exendin_9–39 with the ECD of GLP-1R (Runge et al., 2008; Underwood et al., 2010), these structures do not give information on the receptor-bound conformation of the flexible ligand N terminus, nor on its interactions with GLP-1R. Ligand and receptor mutation studies suggest that an extended flexible conformation of these first residues allows the peptide ligands of class B GPCRs to reach deep into the pocket (Hollenstein et al., 2014). This receptor-bound peptide conformation is proposed to be stabilized by an amino acid motif (Thr¹¹-Phe¹²-Thr¹³ in GLP) that is conserved in class B GPCR ligands (Neumann et al., 2008), which can induce an N-capping conformation similar to the one observed in the receptor-bound NMR structure of pituitary adenylate cyclase–activating polypeptide (PACAP) (Inooka et al., 2001). Substituting three residues in the flexible N terminus of GLP-1 by corresponding PACAP residues (Ala⁸Ser/Glu⁹Asp/Thr¹¹Ile) does not affect affinity and potency for GLP-1R (Xiao et al., 2001), implying that GLP-1 adopts a similar conformation to PACAP upon binding to its receptor. Recently, NMR structures of an 11-mer GLP-1 analog were solved in alterative conformations containing a C-terminal α-helix (PDB: 2N08) and an N-terminal β-turn (PDB: 2N09), and stabilization of these conformations by cyclization cross-links (PDB: 2N0N and 2N0I) differentially influenced GLP-1R–binding affinity and agonist potency (Hoang et al., 2015). The accumulated ligand and receptor SAR (see II. Glucagon-Like Peptide-1 and III. Glucagon-Like Peptide-1 Receptor) suggests that a flexible conformation of the first seven residues allows GLP-1 to interact with residues in the 7TM-binding pocket of GLP-1R, and that the ligand N terminus may adopt a more constrained conformation to activate the receptor.

3. Interaction between the C-Terminal Helix and ECD.

Comparison of the crystal structures of GLP-1 and exendin_9–39-bound GLP-1R ECD (Runge et al., 2008; Underwood et al., 2010), GIP-bound GIPR ECD (Parthier et al., 2007), and the antibody bound glucagon receptor (GCGR) ECD (Koth et al., 2012) provides information about similarities and differences in class B GPCR-bound peptide ligand conformations, and gives insights into the structural determinants of class B GPCR ligand recognition and selectivity (Fig. 2). The crystal structures of the ECDs of different class B GPCRs show that this domain has a conserved fold, including two central antiparallel β-sheets (β₁-β₄) and an N-terminal α-helix (α₁) interconnected by several loops and stabilized by three conserved disulfide bonds (Donnelly, 2012; Pal et al., 2012; Hollenstein et al., 2014) (for more detailed description of the ECD of GLP-1R, see III. Glucagon-Like Peptide-1 Receptor). Class B GPCR ECD peptide ligand complexes exhibit overall a similar binding mode in which the C terminus of the peptide ligand adopts an α-helical conformation that binds between α₁ and β₁–β₄ of the ECD (Parthier et al., 2009; Donnelly, 2012; Hollenstein et al., 2014). The conserved structural fold and binding orientation suggest that common mechanisms underlie ligand recognition of class B GPCRs and indicate that peptide ligand selectivity is in part determined by specific interactions of the C-terminal α-helix of the ligand with the ECD. The ligand-binding mode observed in the GLP-1R crystal structure is consistent with GLP-1 (vide infra) and GLP-1R (see III. Glucagon-Like Peptide-1 Receptor) mutation studies, as well as photo cross-linking experiments placing Gly³⁵ located in the α-helix of GLP-1 in close proximity to E125^ECD located in the linker region between the ECD and 7TMD of GLP-1R (Chen et al., 2009). The structural details of the ECD and 7TMD of GLP-1R are discussed in detail in III. Glucagon-Like Peptide-1 Receptor.

The α-helices of GLP-1, exendin-4, glucagon, and GIP are amphiphilic, containing a hydrophilic and hydrophobic region at opposite sides of the helix. Conserved apolar residues in the C-terminal part of GLP-1 (Phe²⁸, Ile²⁹, Leu³²), exendin-4 (Phe²², Ile²³, Leu²⁶), GIP (Phe²², Val²³, Leu²⁶), and glucagon (Phe²⁸, Val²⁹, Leu³²) share a similar hydrophobic interaction site with the ECD of the corresponding receptor (GLP-1R/GIPR/GCGR: L32^ECD/A32/M29, T35^ECD/L35/L32, V36^ECD/Y36/F33, and W39^ECD/39/36 in α₁, Y69^ECD/68/65 in β-turn 1 connecting β₁-β₂, Y88^ECD/87/84, L89^ECD/88/85, and P90^ECD/89/86 in β-turn 2 connecting β₃-β₄) (Parthier et al., 2007; Runge et al., 2008; Underwood et al., 2010; Koth et al., 2012). Differences in hydrophobic/polar interactions with the N-terminal part of the ligand α-helix as well as ligand-specific ionic interactions with the ECD can partially explain the different relative affinities of GLP-1, exendin-4, GIP, and glucagon for the ECD of the corresponding receptor. The positively ionizable Lys²⁶ of GLP-1 and homologous Arg²⁰ of exendin_9–39 both form ionic interactions with E128^ECD in the C-terminal region of the ECD of GLP-1R. The Lys²⁷ residue of exendin_9–39 forms an additional ionic interaction with E128^ECD in GLP-1R (Runge et al., 2008; Underwood et al., 2010), which in combination with the higher α-helical propensity of exendin-4 compared with GLP-1 (Runge et al., 2007) might explain the increased affinity of exendin-4/exendin_9–39 for the isolated ECD of GLP-1R compared with GLP-1. There is no ionic interaction between peptide ligand and ECD side chains observed in the GIP-bound GIPR ECD crystal structure and the ECD GCGR-glucagon docking model based on the antibody-bound GCGR and GLP-1–bound GLP-1R crystal structures. It should be noted, however, that the apolar residues of GLP-1 (Val³³), GIP (Leu²⁷), and glucagon (Met³³) that are aligned with Lys²⁷ of exendin_9–39 form an additional hydrophobic interaction site with the ECD of GLP-1 (Y69^ECD/L123^ECD), GIPR (Y68^ECD /H115^ECD), and GCGR (Y65^ECD/A118^ECD), respectively. Ala²⁴ and Ala²⁵ of GLP-1 and Val¹⁹ in exendin-4 form another hydrophobic interaction surface with α₁ of the ECD of GLP-1R. The structurally aligned Gln¹⁹ of GIP forms an H-bond network with Asp¹⁵ and the N-terminal backbone of α₁ of the GIPR ECD (Q30^ECD and A32^ECD), whereas corresponding residues in glucagon (Arg²⁴, Ala²⁵) may form a polar/apolar interaction site with the N-terminal region of the α₁ helix of GCGR (Fig. 2).

4. Interaction between Flexible N terminus and 7TMD.

The peptide ligand-bound crystal structures of the ECDs of class B GPCRs (including GLP-1– and exendin-4–bound GLP-1R, GIP-bound GIPR, and antibody-bound GCGR) do not provide information about the molecular interactions between the receptor and the flexible N-terminal region of peptide ligands. Receptor/ligand mutagenesis and photo cross-linking studies nevertheless indicate that the 7TMD of class B GPCRs determines binding (selectivity) of the flexible N-terminal region of peptide ligands (Donnelly, 2012; Pal et al., 2012; Hollenstein et al., 2014). A recently reported full-length GCGR–glucagon model based on the crystal structures of the GCGR 7TMD (Siu et al., 2013), the antibody-bound GCGR ECD (Koth et al., 2012), the GLP-1–bound GLP-1R ECD (Underwood et al., 2010), and the N-capped conformation of PACAP (Inooka et al., 2001) offers a template for full-length GLP-1–bound GLP-1R models (Fig. 2A) that is consistent with the results of mutation studies of GLP-1 (vide infra), GLP-1R (see III. Glucagon-Like Peptide-1 Receptor), and other class B GPCRs (Hollenstein et al., 2014). This full-length GLP-1R model furthermore satisfies spatial constraints defined by GLP-1R cross-linking studies connecting the following: 1) Ala²⁴ in the C-terminal part of the α-helix of GLP-1 to E133^ECD in the region linking the ECD and 7TMD of GLP-1R (not solved in the GLP-1R crystal structure); 2) Leu²⁰ in the N-terminal part of the α-helix of GLP-1 to W297^ECL2 in ECL2 between TM4 and TM5 in GLP-1R; 3) Phe¹² and Val¹⁶ in the flexible N-terminal region of GLP-1 that are positioned near Y145^1.40b and L141^1.36b in the TM1 of GLP-1R, respectively; and 4) a photolabile probe at position 6 (one position before His⁷) in GLP-1 to Y205^ECL1 in ECL1 between TM2 and TM3 in GLP-1R (Chen et al., 2009, 2010b; Miller et al., 2011) (Fig. 2A).

1. Truncated GLP-1 Analogs.

N-terminally truncated forms of GLP-1 (including des-[His⁷]-GLP-1_8–36 and des-[His⁷,Ala⁸]-GLP-1_9–36) are competitive antagonists of GLP-1R, whereas C-terminally truncated GLP-1 constructs (including des-[Gly³⁷,Arg³⁶]-GLP-1_7–37) remain agonists (Mojsov, 1992; Montrose-Rafizadeh et al., 1997). The binding affinities of N-terminally truncated GLP-1 constructs are not affected by mutations in the 7TMD of GLP-1R (Al-Sabah and Donnelly, 2003a,b; López de Maturana et al., 2004), indicating that interactions between the N-terminal region of GLP-1 and the 7TMD of GLP-1R are required for receptor activation. Most truncated GLP-1 constructs have a decreased affinity for GLP-1R (Mojsov, 1992; Montrose-Rafizadeh et al., 1997; Donnelly, 2012), suggesting that interactions with both the ECD and the 7TMD are important determinants of GLP-1 binding. Similar SARs have been observed for the endogenous ligands of related class B GPCRs, including glucagon/GCGR (Unson et al., 1989) and GIP/GIPR (Hinke et al., 2001), demonstrating that interactions of peptide ligands in the 7TMD of class B GPCRs are required for receptor activation. The N-terminally truncated des-[His⁷]-GLP-1_8–36 and des-[His⁷,Ala⁸]-GLP-1_9–36 variants of GLP-1 have a 100- and 1000-fold lower affinity than that of wild-type, whereas truncation of more than two N-terminal residues further diminishes GLP-1R binding (Mojsov, 1992; Montrose-Rafizadeh et al., 1997). Truncation of up to three C-terminal residues (Gly³⁵, Arg³⁶, and Gly³⁷) only has a moderate effect on GLP-1R–binding affinity (up to fivefold decrease), but further deletion of C-terminal residues decreases binding affinity significantly. Nevertheless, several undecapeptide GLP-1_7–15 analogs in which the C-terminal 21 residues are replaced with a biphenylalanine dipeptide have been reported to possess almost the same potency as wild-type GLP-1 (Mapelli et al., 2009).

In contrast to GLP-1, which requires its N-terminal region for high-affinity binding to GLP-1R, the homologous agonist exendin-4 and the N-terminally truncated antagonist exendin_9–39 have similar affinities for GLP-1R (Montrose-Rafizadeh et al., 1997). Although full-length GLP-1R binds exendin-4 and GLP-1 with similar high affinity, the isolated ECD maintains high affinity for exendin-4 and exendin_9–39, but has decreased affinity for GLP-1 (Al-Sabah and Donnelly, 2003a; López de Maturana et al., 2003). In addition, GLP-1 binding is more sensitive to site-directed mutagenesis of the 7TMD of GLP-1R than exendin-4 (Al-Sabah and Donnelly, 2003a,b; López de Maturana et al., 2003; see III. Glucagon-Like Peptide-1 Receptor). Consistently, radioligand competition studies combining isolated C-terminal (ECD) and N-terminal (7TMD) GLP-1R constructs with native and N-terminally or C-terminally truncated GLP-1 and exendin-4 showed that: 1) GLP-1 binding is primarily determined by interactions with the ECD, but also requires interactions with the 7TMD of GLP-1R; and 2) exendin-4–binding affinity is mainly determined by interactions with the ECD of GLP-1R and does not heavily depend on interactions with the 7TMD (López de Maturana et al., 2003). Based on pharmacological studies with truncated GLP-1 and exendin-4, it was further postulated that the eight-residue C-terminal extension of exendin-4 (Trp-cage, see above) may play a role in its superior affinity for the ECD of GLP-1R (Al-Sabah and Donnelly, 2003a).

2. Chimeric GLP-1 Analogs.

Chimeric constructs of GLP-1 in combination with other class B GPCR peptide ligands (including exendin-4, glucagon, GIP, and PACAP) have been used to identify structural determinants of ligand selectivity for GLP-1R (and other class B GPCRs). Radioligand competition studies with truncated and chimeric GLP-1 and exendin-4 peptides suggested that divergent residues in the central part of GLP-1 (Val¹⁶-Arg³⁶) and exendin-4 (Leu¹⁰-Gly³⁰) determine GLP-1/exendin-4 selectivity for the ECD of GLP-1R, and indicated that the Trp-cage only plays a minor role in the increased affinity of exendin-4 for the ECD of GLP-1R compared with GLP-1 (Runge et al., 2003b). These pharmacological data are consistent with comparative CD and fluorescence spectroscopy, NMR, and X-ray crystallography analyses of (truncated and chimeric forms of) GLP-1 and exendin-4 (Runge et al., 2007), implying that the differential affinity for the ECD can be explained by the higher α-helical propensity of exendin-4 in solution and by stronger (ionic) interactions between GLP-1R and exendin-4, compared with GLP-1 (see above). GLP-1/glucagon chimera experiments showed that a peptide ligand consisting of the N-terminal part of glucagon combined with the C-terminal part of GLP-1 has high affinity for both GCGR and GLP-1R (Hjorth et al., 1994). Substitution of N-terminal GLP-1 residues with corresponding glucagon residues [e.g., (Ala⁸Ser/Glu⁹Gln)-GLP-1 and (Val¹⁶Tyr/Ser¹²Lys)-GLP-1] maintains high GLP-1R–binding and low GCGR-binding affinities, whereas substitution of N-terminal glucagon residues with corresponding GLP-1 residues decreases GCGR binding and maintains low GLP-1R–binding affinities. In contrast, substitution of C-terminal GLP-1 residues with corresponding glucagon residues decreases GLP-1R binding and maintains low GCGR-binding affinities, whereas replacement of, for example, the last three C-terminal residues of glucagon (Met³³, Asn³⁴, and Thr³⁵) with the corresponding GLP-1 residues (Val³³, Lys³⁴, and Gly³⁵) and an additional C-terminal GLP-1 residue (Arg³⁶) increases GLP-1R binding with affinity for GCGR remaining moderate. Glucagon/GLP-1 chimeras were combined with GCGR/GLP-1R chimeras to identify the receptor domains that interact with N-terminal and C-terminal parts of glucagon and GLP-1 (Runge et al., 2003b). The chimeric GLP-1-(His⁷-Leu²⁰)/glucagon-(Asp²¹-Thr³⁵) is unable to bind and activate GCGR, but its binding affinity and potency are rescued by substituting the 7TMD of GCGR with the 7TMD of GLP-1R, suggesting that the N-terminal region of GLP-1 (His⁷-Leu²⁰) interacts with the 7TMD of GLP-1R (Runge et al., 2003b). The GCGR(ECD)/GLP-1R(7TM) chimera has equal binding affinity and potency for GLP-1 and the chimeric glucagon-(His⁷-Leu²⁰)/GLP-1-(Asp²¹-Gly³⁷), although these are decreased compared with wild-type GLP-1R, indicating that the ECD of GLP-1R is the major determinant of GLP-1/glucagon selectivity by interacting with the C-terminal region of GLP-1. Chimeras combining different N-terminal and C-terminal regions of GLP-1 (His⁷-Leu²⁰, Asp²¹-Arg³⁶) and GIP (His¹-Leu¹⁴, Asp¹⁴-Lys³⁰), or substituting the middle parts of GIP with corresponding residues of GLP-1 (Ser¹⁸-Ala²⁴, Glu²¹-Ala²⁴), all had more than 100-fold lower binding affinity for GLP1-R compared with GLP-1-(His⁷-Arg³⁶) (Hareter et al., 1997). Substituting three to five residues in the N-terminal part of GLP-1 by corresponding residues from GIP (His⁷Tyr/Thr¹³Ile/Val¹⁶Tyr), secretin (Ala⁸Ser/Glu⁹Asp/Asp¹⁵Glu/Val¹⁶Leu), vasoactive intestinal peptide (Ala⁸Ser/Glu⁹Asp/Gly¹⁰Ala/Thr¹¹Val), or PACAP (Ala⁸Ser/Glu⁹Asp/Thr¹¹Ile/Ser¹⁴Asp/Asp¹⁵Ser) diminished ligand potency and decreased binding affinity by more than 10-fold (Hareter et al., 1997). Substituting two to three residues in the N-terminal part of GLP-1 by corresponding residues from glucagon (Ala⁸Ser/Glu⁹Gln), peptide histidine isoleucine (Glu⁹Asp/Thr¹¹Ile), or PACAP (Ala⁸Ser/Glu⁹Asp/Thr¹¹Ile), in contrast, had only small effects on affinity for GLP-1R.

3. Substituted GLP-1 Analogs.

In addition to ligand (and receptor) truncation and chimera studies, several site-directed substitution experiments were performed to provide more detailed information regarding the molecular determinants of GLP-1/GLP-1R binding and selectivity.

1. Receptor Cloning.

To better understand the action of GLP-1, Thorens isolated and characterized the first cDNA of the rat pancreatic β cell GLP-1 receptor in 1992 (Thorens, 1992), followed by cloning of the human receptor in 1993 (Dillon et al., 1993; Graziano et al., 1993; Thorens et al., 1993), revealing a sequence of 463 residues. This deduced primary sequence resembled that of the receptors for secretin, parathyroid hormone, and calcitonin, and hence enabled it to be classified within what was then a new branch of the GPCR superfamily (family B, class B, or secretin receptor-like) that now includes 15 members (Segre and Goldring, 1993; Hoare, 2005).

2. Receptor Expression.

Using RNA enzyme protection techniques, GLP-1R was found widely expressed in the human and mouse pancreas, lung, brain, stomach, heart, and kidney, but was not seen in tissues involved in glucose metabolism, such as the liver, skeletal muscle, and fat (Dunphy et al., 1998). DNA hybridization experiments showed that the human GLP-1R (hGLP-1R) gene is localized to chromosome 6p21, and the size is 40 kb, containing about 14 exons. Both human and rat receptors contain 463 amino acids, whereas that of the mouse has 489 amino acids, displaying a homology with hGLP-1R of 91% and 84%, respectively. Although the expression levels of GLP-1R vary among different tissues and cell types, Northern hybridization studies demonstrated that its expression in pancreatic islets, heart, and lungs is significantly higher (Dunphy et al., 1998). In addition, GLP-1R was also found in the lateral septum, thalamus, and hippocampus of the rat brain, and the cDNA fragment cloned from the brain, heart, and pancreas encodes the same amino acids of the receptor, suggesting its physiologic relevance in the cardiovascular and central nervous systems (Satoh et al., 2000).

It is known that the expression of GLP-1R in the islets of Langerhans is regulated by glucose and dexamethasone, but not by the PKA-dependent signaling pathway (Abrahamsen et al., 1995). Starvation and refeeding could influence its expression in the rat hypothalamus and other sites of the CNS (MacLusky et al., 2000). Although our knowledge about the transcriptional regulation mechanisms of GLP-1R is still limited, promoter analysis indicate that 1) transcription factors Sp1 and Sp3 may play important regulatory roles and 2) homologous desensitization and internalization are closely related to the intracellular sites of 441/442, 444/445, and 451/452 (Wildhage et al., 1999).

3. Receptor Biosynthesis.

About 5–10% of GPCRs contain an N-terminal cleavable signal peptide (Schulein et al., 2012). Most GPCRs mediate the integration of receptor into cell membrane by a signal anchor sequence that is located at the first TMD (Audigier et al., 1987; Wallin and von Heijne, 1995). Both types of signal sequences regulate the endoplasmic reticulum (ER) signal process at the beginning of the secretory pathway (Belin et al., 1996; Kochl et al., 2002).

The signal peptide is a sequence of approximately 20 amino acids in the N terminus of the receptor that includes a polar and charged N-terminal (n) region, a central hydrophobic (h) region, and a polar C-terminal (c) region (Schneider and Fechner, 2004; Clerico et al., 2008). The C-terminal side regularly contains helix-breaking proline and glycine residues and small uncharged residues at positions 1 and 3 of the cleavage site. After synthesis of the receptor's N-tail in the cytoplasm, the translation ceases by its binding to the signal recognition particle (SRP) (Halic and Beckmann, 2005; Shan and Walter, 2005). A translocon complex is attached to the membranes of ER that includes a GTP-dependent interaction between SRP and SRP receptor, translocon gating, protein synthesis, and integration of nascent chains with the bilayer (Kochl et al., 2002). The signal peptide is finally cleaved off by signal peptidases on the ER membrane. In contrast, signal anchor sequences are not cleaved and form a part of the mature protein (Audigier et al., 1987; Wallin and von Heijne, 1995; Higy et al., 2004; Schulein et al., 2012). In addition to recognition and binding of SRP and translocation to ER, the N-terminal signal sequence may play key roles in protein folding and trafficking. This is highly dependent upon whether the signal sequence is a cleaved signal peptide or an uncleaved signal anchor.

For GLP-1R, studies on the signal peptides of corticotropin-releasing factor-1 receptor (CRF₁R) and corticotropin-releasing factor-2 receptor give a good model to compare. Both CRF receptors exhibit a high probability of N-terminal cleavable signal peptide by computational prediction. In reality, however, only the signal peptide of CRF₁R is cleaved, whereas that of corticotropin-releasing factor-2 receptor is translocated and embedded to the plasma membrane by taking TM1 as a signal anchor sequence. It is thus integrated to be a part of the N-terminal ECD: not mediating ER targeting, but contributing to receptor activation (Alken et al., 2005). Interestingly, a mutant of rat CRF₁R without the signal peptide sequence was able to translocate to the cell surface and display similar biologic function to the wild-type. The rat CRF_2aR contains an uncleaved N-terminal signal peptide, and deletion of this resulted in a significant reduction in its membrane expression, suggesting that receptor transportation was affected (Rutz et al., 2006). The complexity of these signal sequences highlights the need for experimental verification of the role of the signal sequence in GLP-1R.

According to the signal peptide prediction program “SignalP 4.0” (Nielsen et al., 1997; Bendtsen et al., 2004), the sequence of the first 23 amino acids in GLP-1R fits all the criteria of an N-terminal signal peptide. The existence of a functional signal peptide was demonstrated experimentally by Huang et al. (2010), who showed that it was required for GLP-1R synthesis and was cleaved thereafter. Mutation of the signal peptide (A21-R) resulted in retention of the receptor within ER, whereas mutation of E34-R augmented its cell surface expression when the signal peptide was deleted. The amino acid sequence following the signal peptide in the GLP-1R, G27^ECD-W39^ECD, is relatively hydrophobic, and this region may be recognized by SRP (Huang et al., 2010). Ge et al. (2014) went deep into the function of this signal peptide by use of constructs containing epitope tags at the N and/or C terminus. A mutant GLP-1R without the signal peptide sequence was expressed in human embryonic kidney (HEK) 293 cells and displayed normal functionality with respect to ligand binding and cAMP activation, suggesting that the putative signal peptide may not be required for receptor synthesis. Immunoblotting analysis showed that the amount of GLP-1R synthesized in HEK293 cells was low without the signal peptide, indicating its role in facilitating receptor expression (Ge et al., 2014). Epitope tags at the N terminus of GLP-1R were detectable by immunofluorescence and immunoblotting, an observation that is inconsistent with another report that studied both signal peptide and the hydrophobic region after the signal peptide. It was found that the signal peptide was cleaved in the mature hGLP-1R, and cell surface expression was almost abolished by the mutation A21-R that prevented the cleavage, demonstrating that hydrophobic region after the signal peptide is necessary for efficient hGLP-1R trafficking to the cell surface. Because glycosylation is vital to cell surface expression, cleavage of the signal peptide will affect this process. In addition, mutating W39^ECD, Y69^ECD, and Y88^ECD of hGLP-1R to alanine eliminated its cell surface expression without influencing N-linked glycosylation and cleavage of the signal peptide (Thompson and Kanamarlapudi, 2014).

Clearly, such discrepancies may not only reflect differences in the methods employed among these three studies, but also the cellular background and the complexity of the underlying mechanism(s).

1. N-Terminal Domain.

b. Site-directed mutagenesis of the ECD.

The N-terminal domain has been the subject of a number of mutagenesis studies summarized in Table 1 and Fig. 3 (Wilmen et al., 1997; Tibaduiza et al., 2001; Mann et al., 2010b; Underwood et al., 2010; Day et al., 2011; Koole et al., 2011; Patterson et al., 2013). Residues that have been mutated but do not greatly affect GLP-1 affinity/efficacy include P7^ECDL, R20^ECDK, L32^ECDA, T35^ECDA, V36^ECDA, R44^ECDH, E68^ECDA, L123^ECDA, E127^ECDA, E127^ECDQ, E128^ECDA, E128^ECDQ, and E128^ECDM. To study a species-selective small-molecule antagonist, Tibaduiza et al. (2001) mutated W33^ECD to Ser (the human to rat GLP-1R substitution), as well as to various other residue types, showing that there was no effect on GLP-1 albeit that the affinity and potency values were not given for the peptide ligands used. Wilmen et al. (1997) substituted six Trp residues in the ECD with Ala and found that, whereas W87A behaved like wild-type GLP-1R in binding and cAMP assays, the remaining substitutions at 39, 72, 91, 110, and 120 resulted in the abolition of detectable radioligand binding. Modest five- to eightfold reductions in GLP-1 potency have been observed for E68^ECDK, P90^ECDA, and R121^ECDA (Underwood et al., 2010; Day et al., 2011). The ECD structures suggest that E68^ECD is close to the C-terminal region of the peptide ligand, but, although it appears to have little significant interaction with exendin_9–39 in the human receptor, the equivalent residue in rat GLP-1R, D68^ECD, enhances exendin_9–39 affinity by forming a hydrogen bond with Ser³² of the peptide (Mann et al., 2010b). Other mutations, Y69^ECDA, Y88^ECDA, and L89^ECDA, have caused catastrophic effects on the ability of the receptor to be detected in binding or signaling assays (Underwood et al., 2010).

View this table:

• View inline
• View popup

TABLE 1

Summary of effects on GLP-1 pharmacology (affinity and ability to activate cAMP pathway) in published site-directed mutagenesis studies of GLP-1R

WT (wild-type) refers to mutations that resulted in either <fivefold or no statistically significant change from wild-type GLP-1R. ND (not determinable) refers to a property that was measured, but for which a value was not determinable. Blank cells mean that the assays used to estimate that particular pharmacological property were not carried out in the cited work. Residues with symbol ^† refer to data from rat GLP-1R (if different, the equivalent human residue number is displayed in the table to aid comparison). GLP-1 affinity or potency fold-change values with suffix ^M are from membrane preparations, whereas suffix ^C is from whole-cell assays. Cell surface expression values below 75% of WT are shown (>75% are shown as WT): a suffix ^E represents estimation from ELISAs; suffix ^mic was evaluated from immunofluorescent microscopy; suffix ^cyt was evaluated by flow cytometry with an anti-Flag antibody; suffix ^Ag refers to affinity or cell surface expression levels determined from agonist radioligand-binding assays, whereas suffix ^Ant was from antagonist radioligand-binding assays. ΔLog τ_c values relative to WT are shown where >0.5 and were calculated from data where the expression-corrected efficacy term τ_c had been calculated using the operational model of agonism, as defined in Wootten et al. (2013c). Residues with transmembrane helices are numbered according to Wootten et al. (2013c).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Summary of GLP-1R mutagenesis. A snake plot of GLP-1R from http://www.gpcrdb.org that has been colored (see Key) to highlight the location of signal peptide, glycosylation, and phosphorylation sites, as well as the mutated residues in Table 1. It should be noted that the color coding of 74 of the 195 mutated residues (W39, W72, W87, W91, W110, F169-C174, R176, N177, H180, N182, A200-Q213, W214-G225, R227-F230, L232, M233, E262-L268, F321-I332, K334-K336, and R348-K351) reflects the effects of rat GLP-1R mutations projected on the hGLP-1R amino acid sequence. The color coding of 27 of the 185 residues (K202, W203, S206-Q211, Q213, W214, G216-Q221, S223-G225, R227-F230, L332, M233, I325, and F326) reflects the effect of double mutations, not single-point mutations. Information on the fold change in ligand affinity and potency as well as expression levels of the GLP-1R mutants is reported in Table 1.

2. Seven-Transmembrane Domain.

1. Signaling.

b. Pancreatic β cells.

Research has largely focused on characterizing the GLP-1R signaling in β cells, where it mediates increased insulin secretion, storage, and synthesis as well as increased β cell mass. Consistent with the findings in recombinant cells, elevated cAMP is vital in β cells for glucose-dependent insulin secretion mediated by GLP-1R, and, although calcium mobilization, phosphorylation of ERK, and β-arrestin are important, Gαq/11 does not seem to play a major role. A summary of known pathways involved in pancreatic β cell function is illustrated in Fig. 4.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Summary of the main characterized pathways of glucose and GLP-1 signaling in the pancreatic β cell. Glucose enters the cell through glucose transporter 2 and undergoes glycolysis to produce pyruvate that enters the mitochondria for oxidative metabolism and ATP production. This increase in cytosolic ATP closes the K_ATP channels, depolarizing the membrane and opening the voltage-dependent calcium channels, increasing calcium influx into the cell, causing insulin exocytosis. GLP-1 increases insulin exocytosis through a number of mechanisms. GLP-1R couples to Gαs, activates adenylate cyclase that converts ATP to cAMP, and mobilizes two downstream effectors, PKA and Epac. These have a range of effects, including closing K_ATP channels, enhancing fusion of insulin secretory granules with the membrane, whereas PKA also closes Kv channels, inhibiting membrane repolarization. PKA and Epac also increase intracellular calcium by facilitating CICR through the opening of inositol 1,4,5-trisphosphate receptor and ryanodine receptor calcium channels, respectively. This increase in calcium has also been proposed to upregulate mitochondrial ATP production and activate calcineurin, so nuclear factor of activated T cells promotes insulin gene transcription to increase insulin stores. Alongside increasing insulin synthesis and exocytosis, GLP-1 signals through a number of pathways to increase β cell mass. PKA reduces ER stress through ATF-Gadd34 signaling, increases β cell neogenesis by activating cyclin D, and elevates the expression of insulin receptor substrate 2 (IRS2), a β cell survival factor, as well as anti-apoptotic proteins Bcl-2 and Bcl-x_L through CREB. PI3K is activated by IRS2 and transactivation of epidermal growth factor receptor, and this further promotes increased β cell mass through upregulation of PDX-1 and nuclear factor κB, which upregulates anti-apoptotic Bcl-2 /Bcl-x_L and inhibitor of apoptosis protein-2.

i. Insulin Secretion.

β cells undergo glucose-stimulated insulin secretion. Glucose enters the β cell through glucose transporter 2 and is converted through glycolysis to pyruvate, which enters the mitochondria for oxidative phosphorylation, increasing the cytosolic ATP/ADP ratio (MacDonald et al., 2005). The increased ATP closes K_ATP channels, depolarizing the plasma membrane and increasing calcium influx through L-type voltage-dependent calcium channels, causing release of calcium from intracellular stores through calcium-induced calcium release (CICR) (MacDonald et al., 2005). Increased cytoplasmic calcium stimulates exocytosis of the insulin secretory granules.

GLP-1R signaling enhances this glucose-dependent insulin secretion through activation of Gαs, upregulation of cAMP, and subsequent activation of PKA and exchange protein activated by cAMP (Epac). PKA is a holoenzyme composed of both regulatory and catalytic subunits; the latter are released upon binding of cAMP to the regulatory subunits, leading to phosphorylation of downstream substrates, whereas Epac2 primarily functions as a guanine nucleotide exchange factor for small Ras-like G proteins, thereby regulating their activity. PKA phosphorylates the sulfonylurea receptor (SUR1) subunit of the K_ATP channels, closing them and further depolarizing the membrane (Light et al., 2002). Epac also inhibits K_ATP by increasing its sensitivity to ATP (Kang et al., 2008). The cAMP/PKA pathway, in concert with the PI3K/protein kinase C (PKC)ζ pathway (discussed in more detail below), also inhibits voltage-gated potassium channels, which in response to depolarization, opens and allows K⁺ efflux, repolarizing the cell (MacDonald et al., 2003). This delays repolarization, allowing increased calcium influx via voltage-dependent calcium channels (MacDonald et al., 2003). PKA and Epac1/2 are also involved in CICR from the ER, which is mediated by PKA via the inositol 1,4,5-trisphosphate receptor and by Epac1/2 through ryanodine receptors, increasing intracellular calcium (Kang et al., 2003; Tsuboi et al., 2003; Dyachok and Gylfe, 2004). These mechanisms, therefore, enhance the ability of β cells to exocytose insulin secretory granules. There is evidence that CICR also contributes to increased mitochondrial ATP, in conjunction with GLP-1–induced increases in ATP (Tsuboi et al., 2003). It was proposed that the intracellular calcium may enter the mitochondria to activate mitochondrial dehydrogenases to increase ATP production (Tsuboi et al., 2003). PKA and Epac also seem to have a role in the exocytosis of insulin secretory granules, with PKA phosphorylating snapin, a protein that is vital to the regulation of vesicle assembly, and Rab-3–interacting molecule (Rim) and Munc-131, which are involved in vesicle fusion (Kwan et al., 2007; Song et al., 2011). Epac2 also interacts with Rim2 and Piccolo, a calcium sensor, to form a complex that is important for insulin secretion (Fujimoto et al., 2002). Studies in islets of phospholipase C and Epac knockout mice also suggest that Epac2 acts via Rap1-regulated phospholipase Cε to cause calcium-induced insulin exocytosis (Shibasaki et al., 2007; Dzhura et al., 2011). GLP-1 enhances glucokinase activity through Epac2 in a Rim2/Rab3A-dependent manner, sensitizing β cells to glucose. Knockdown of β-arrestin 1 in cultured pancreatic cells resulted in a reduction of both GLP1-mediated cAMP production and insulin secretion (Sonoda et al., 2008). Evidence suggests that this occurs via an increase in β-arrestin 1–mediated ERK1/2 phosphorylation that enhances cAMP response element-binding protein (CREB) activation and cAMP levels, subsequently contributing to insulin secretion (Quoyer et al., 2010).

ii. Insulin Synthesis and Storage.

GLP-1 also acts to increase insulin stores in β cells by promoting insulin gene transcription, its mRNA stability, and biosynthesis. This occurs in both cAMP/PKA-dependent and independent manners (Baggio and Drucker, 2007). Studies with the rat insulin I promoter have implicated basic region leucine zipper proteins, similar in structure to CREB, which bind to the cAMP-responsive element site to upregulate insulin transcription in a cAMP/PKA-independent mechanism (Skoglund et al., 2000; Chepurny et al., 2002). The upregulation of the expression and activity of transcription factor pancreatic-duodenum homeobox-1 (PDX-1) and its increased activity also promote insulin transcription and biosynthesis via a PKA-mediated mechanism (Wang et al., 1999, 2001). Glucose and GLP-1 potentiate insulin gene transcription, by increasing calcium levels, activating calcineurin that dephosphorylates nuclear factor of activated T cells, resulting in its nuclear localization and the promotion of insulin gene transcription (Lawrence et al., 2002).

iii. β Cell Survival, Proliferation, and Neogenesis.

GLP-1 promotes β cell proliferation, neogenesis, and inhibition of apoptosis in rat models of T2DM (Buteau, 2011). GLP-1R activation facilitates the release of β-cellulin by membrane-bound metalloproteinases, inducing transactivation of epidermal growth factor receptor, which then signals through PI3K and activates Akt/protein kinase B (PKB), PKCζ, and p38 MAPK and ERK downstream (Buteau, 2011).

PDX-1 is also critical in increasing β cell mass due to its proliferative and anti-apoptotic effects (Li et al., 2005). As well as the increased activity mentioned above, GLP-1R enhances its expression. Akt inhibits forkhead transcription factor (Foxo1) action through its nuclear exclusion, relieving its inhibition on expression of forkhead box protein A2, which controls PDX-1 expression (Kitamura et al., 2002; Buteau et al., 2006). In addition, GLP-1–mediated activation of PKA upregulates CREB, increasing the expression of IRS2, a β cell survival factor, leading to activation of PI3K/PKB and the anti-apoptotic protein Bcl-2 (Wilson et al., 1996; Rhodes and White, 2002; Jhala et al., 2003). Bcl-2 and Bcl-x_L are also upregulated through the activation of nuclear factor κB downstream of PKB (Buteau et al., 2004). Nuclear factor κB increases expression of inhibitor of apoptosis protein-2, thereby preventing apoptosis (Buteau et al., 2004). GLP-1 signaling has also been implicated in downregulation of proapoptotic caspase-3 and decrease in the cleavage of poly-ADP-ribose (Hui et al., 2003; D'Amico et al., 2005). ERK1/2 signaling via β-arrestin 1 also activates p90RSK, which phosphorylates proapoptotic Bad and inactivates it (Quoyer et al., 2010).

PKA activation of MAPK and cyclin D1 is important in the transition of the G1/Gs phase essential to cell cycle progression: they promote β cell neogenesis (Friedrichsen et al., 2006). GLP-1R agonists also improve β cell function and survival upon ER stress. This is believed to occur through PKA activation of C/EBP homologous protein and growth arrest and DNA damage-inducible protein (Gadd34), which prevents the dephosphorylation of translation initiator elF2α, allowing the ER to recover from stress (Yusta et al., 2006).

2. Ligand-Directed Signal Bias.

As described above, GLP-1R is pleiotropically coupled to a range of signaling effectors, each of which can impact on the physiologic response elicited by receptor activation. This allows the potential for individual ligands to evoke different patterns of response upon interaction with the receptor in what is termed ligand-directed biased signaling (Shonberg et al., 2014). Such biased signaling and regulation are thought to arise through the stabilization of distinct ensembles of receptor conformations that occur through the varying chemical contacts between ligands and the receptor. Although gross changes in signal bias can be recognized as a reversal of potency or efficacy (Kenakin and Miller, 2010; Koole et al., 2013b; Shonberg et al., 2014), less dramatic effects require a quantitative framework to identify significant differences in ligand response. The most robust method to quantify efficacy is the operational model of Black and Leff (Kenakin and Christopoulos, 2013), with the transduction ratio of τ/K_A used to define strength of signaling of an individual ligand for a specific pathway, and this ratio can be determined from classic concentration-response experiments (Kenakin et al., 2012).

a. Peptide-mediated signal bias.

There are multiple endogenous peptides that can interact with and activate GLP-1R, including the fully processed GLP-1_{7–36 amide} and GLP-1_7–37 peptides, the extended 1–36 and 1–37 forms of these peptides, as well as oxyntomodulin. The search for more stable forms of GLP-1 has also provided a range of mimetic peptides, including exendin and modified forms of GLP-1 that have been developed for therapeutic use (V. Pharmaceutical Development and Therapeutics). Evidence for biased signaling requires measurement of multiple signaling endpoints. Early work characterizing ligands for GLP-1R was largely limited to measurement of cAMP, as the latter is the most well-coupled pathway and critical for the incretin effect on β cells. Therefore, it is only relatively recently that evidence for ligand-directed signaling has emerged.

Initial studies used heterologously expressed GLP-1R to more broadly examine three canonical pathways that have been linked to physiologic signaling in β cells, specifically cAMP accumulation, ERK phosphorylation, and intracellular calcium (iCa²⁺) mobilization. Even with this relatively limited assessment of signaling, quantitative evidence for peptide-mediated signal bias was observed, with oxyntomodulin exhibiting a relative bias toward pERK over cAMP and iCa²⁺ signaling compared with either GLP-1 or exendin (Koole et al., 2010), and this supported earlier observations of differences in cAMP signaling and β-arrestin recruitment between oxyntomodulin and GLP-1_{7–36 amide} (Jorgensen et al., 2007). Likewise, the extended form of GLP-1_{1–36 amide} had a relative bias toward pERK, but a loss of iCa²⁺ signaling. Thus, this work demonstrated GLP-1R was subject to the potential for peptide-mediated biased signaling, although no significant bias was observed between amidated and nonamidated forms of GLP-1_7–36 or between GLP-1 and exendin for these pathways (Koole et al., 2010). More extensive analysis of signaling and regulatory pathways that included recruitment of arrestins revealed additional tiers of biased signaling with both exendin and oxyntomdulin exhibiting stronger arrestin recruitment relative to GLP-1_{7–36 amide}, whereas GLP-1_1–36 did not recruit arrestins (Wootten et al., 2013b) (Fig. 5, left panel), at least in this recombinant system. The synthetic 11-mer peptide, BMS21, although much less potent than GLP-1_{7–36 amide}, had a relatively similar profile of activation of the canonical pathways (cAMP, pERK, and iCa²⁺), but did not elicit recruitment of arrestins (Wootten et al., 2013b). Intriguingly, the principal metabolite of GLP-1, GLP-1_{9–36 amide}, displays a relative preservation of pERK signaling (Li et al., 2012a; Wootten et al., 2012), despite a marked attenuation of cAMP production and ability to induce insulin secretion (Wootten et al., 2012), indicating that this peptide is strongly biased toward pERK and may be capable of signaling in a physiologically relevant manner. Additional evidence for peptide-mediated signaling bias is seen with a yeast assay of chimeric G protein activation (Weston et al., 2014), consisting of the yeast G protein GPA1 substituted by five amino acids from the human Gα sequence. In this system, exendin, oxyntomodulin, and glucagon exhibited relative bias toward the GPA1/Gαi chimera over the GPA1/Gαs chimera when compared with signaling by GLP-1_7–36, whereas liraglutide had a similar profile of signaling to GLP-1 across these two pathways (Weston et al., 2014). This indicates that biased signaling occurs even at the level of G protein recruitment and is consistent with stabilization of different ensembles of receptor conformations by different peptide ligands.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Web of bias illustrating distinctions in the pattern of signaling of different peptide agonists (left panel) or nonpeptidic modulators (right panel) at GLP-1R. The web of bias plots ΔΔτ/K_A values on a logarithmic scale for each ligand and for every signaling pathway tested. Formation of these values included normalization to the reference ligand GLP-1_{7–36 amide} and the reference pathway, cAMP accumulation. The plots do not provide information on absolute potency, but on relative efficacy for signaling of individual pathways in comparison with that for cAMP. Data are from Koole et al., 2010; Willard et al., 2012a; and Wootten et al., 2013b.

3. Molecular Basis for Signal Bias.

There is increasing evidence demonstrating that the conformational change that drives activation transition involves reordering of hydrogen bond networks and intrahelical packing. Such a change promotes reorganization of the intracellular face of the receptor, thereby enabling interaction with intracellular effectors (Zhou et al., 2000; Curran and Engelman, 2003; Angel et al., 2009; Illergard et al., 2011). These networks are relatively well conserved within GPCR subfamilies and involve conserved polar residues (Venkatakrishnan et al., 2013). This is best studied for class A GPCRs, and suggests that there is evolutionary conservation of the mechanisms underpinning activation transition. Although there is limited direct homology in the amino acid sequence of the TM domains between the major receptor families, it is expected that such conserved polar residues within class B GPCRs play a similarly important structural role and contribute to the mechanism of ligand-directed biased signaling. This view is supported by experiments in which conserved intramembranous polar residues in GLP-1R were mutated to alanine. A cluster of residues, including R190^2.60b, N240^3.43b, H363^6.52b, and Q394^7.49b, when mutated, was shown to change the signaling profile of the receptor for cAMP, pERK, and iCa²⁺ in both a residue-specific and ligand-specific manner (Wootten et al., 2013c). Although differences were particularly marked between GLP-1 and oxyntomodulin, selective differences in the effect of mutation were also seen between GLP-1 and exendin, providing molecular evidence that the mechanism of receptor activation differs between these two peptides, despite relatively similar overall efficacy for the peptides across these pathways at the wild-type receptor (Wootten et al., 2013c). In addition to these residues that displayed ligand-specific effects, there was a series of serine residues (S155^1.50b, S186^2.56b, and S392^7.47b) that played a role in the control of receptor-dependent signal bias; however, these had global effects on all peptides (Wootten et al., 2013c). These residues sit at the interface between either TM1 and TM7 (S155^1.50b and S392^7.47b) or TM2 and TM3 (S186^2.56b) and are most likely involved in tight packing of these helices. With the solution of crystal structures of the TM domain of the related CRF₁ and glucagon receptors (Hollenstein et al., 2013; Siu et al., 2013; Jazayeri et al., 2016), it is now possible to model the location of these residues with improved precision. Intriguingly, such modeling indicates that these residues reside at a fulcrum position of the receptor TM bundle, where the splayed helices of the open extracellular face of the receptor converge (Fig. 6), with the residues that contribute to ligand-dependent signaling forming a central interaction network and the smaller polar residues that are globally important for signaling external to this core.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 6.

Homology model of GLP-1R illustrating the relative position of key residues involved in the receptor-signaling bias. The modeling indicates that these residues reside at a fulcrum position of the receptor transmembrane bundle, where the splayed helices of the open extracellular face of the receptor converge, with the residues that contribute to ligand-dependent signaling forming a central interaction network (space fill, red) and the smaller polar residues that are globally important for signaling external to this core (space fill, blue). The receptor is displayed in three views at different horizontal rotation. Transmembrane helices are numbered with Roman numerals.

Although not broadly conserved across the class B subfamily, there is an additional polar threonine in TM1, at position 149^1.44b, that is the site of a naturally occurring polymorphism leading to incorporation of a methionine in this position (Beinborn et al., 2005). Methionine at this position leads to a marked loss in coupling of the receptor to cAMP accumulation (Fortin et al., 2010; Koole et al., 2011) and iCa²⁺ mobilization, whereas ERK phosphorylation is relatively preserved (Koole et al., 2011). This is a global effect for peptide activators of the receptor (Koole et al., 2011). Exploration of the tolerance of this position to different amino acids revealed that serine substitution had relatively minimal detrimental effect, suggesting that maintenance of the polar nature of this residue is important. However, amino acids of similar size, although leading to attenuation of response in a pathway-specific manner, were the least affected, indicating that side chain packing was also important (Koole et al., 2015). Thus, this mutation changes the physiologic bias of the receptor in addition to markedly attenuating peptide-mediated signaling.

Although these experiments have been principally interpreted in the context of in cis signaling of receptor to effectors, GLP-1R, like most, if not all class B GPCRs, undergoes functionally important dimerization (Harikumar et al., 2006, 2010, 2012; Gao et al., 2009; Schelshorn et al., 2012). Intriguingly, disruption of the TM4 dimer interface of GLP-1R differentially impacted the signaling elicited by agonist peptides with a greater abrogation of iCa²⁺ mobilization relative to that of either cAMP accumulation or ERK phosphorylation (Harikumar et al., 2012). This indicates that differential ligand-mediated signaling could also involve the allosteric interaction of protomers within a receptor dimer.

In addition to an emerging appreciation of how ligands can selectively alter key hydrogen-bonding networks, there is some limited information on how extracellular loop residues contribute to peptide-mediated bias, in particular for ECL2 (Koole et al., 2012a,b). K288^ECL2, C296^ECL2, W297^ECL2, and N300^ECL2 had critical roles in governing signal bias of the receptor, but were also globally important for peptide signaling. Nonetheless, peptide-specific effects on relative efficacy and signal bias were most frequently observed for residues 301–305, although R299^ECL2A mutation also exhibited different effects for individual peptides. M303^ECL2 appeared to play a greater role for exendin and oxyntomodulin actions than those of GLP-1 peptides. Interestingly, ECL2 mutation was generally more detrimental to exendin-mediated iCa²⁺ mobilization than GLP-1_{7–36 amide}, providing additional support for subtle variances in receptor activation by these two peptides.

4. Allosteric Modulation.

As described above (II. Glucagon-Like Peptide-1 and III. Glucagon-Like Peptide 1 Receptor), peptide ligands engage GLP-1R via a diffuse pharmacophore that includes key interactions for affinity with the N-terminal extracellular domain, as well as poorly defined interactions with the extracellular loops and TM domain core that drive receptor activation. Nonpeptidic ligands have a distinct mode of binding, and in many cases can do so via topographically distinct, allosteric sites from those of the native peptides. Such ligands can bind simultaneously with peptide ligands and interact in a cooperative manner to alter the binding and/or efficacy of peptide signaling (and vice versa) (Leach et al., 2007; May et al., 2007; Keov et al., 2011; Wootten et al., 2013a; Gentry et al., 2015). Allosteric modulation of GPCRs is now a well-established paradigm that provides both advantages and challenges for drug discovery when compared with classic orthosteric ligands (Wootten et al., 2013a). The first characterized allosteric modulators of GLP-1R were identified by Novo Nordisk, and are exemplified by compound 2, which, as noted above, is an allosteric agonist of the receptor. In addition to its intrinsic efficacy, this series of compounds augmented the binding of radiolabeled GLP-1, indicating that it could act as a positive allosteric modulator (Knudsen et al., 2007). Nonetheless, there was only limited impact on the efficacy of GLP-1R for cAMP signaling. A hallmark of allosteric interactions is the phenomenon of probe dependence that describes the capacity for different effects, depending upon the orthosteric and allosteric ligand combination. This is also observed for ligands of GLP-1R. Although compound 2 has only limited effect on GLP-1_{7–36 amide}–induced cAMP production, it yields a ∼30-fold augmentation of oxyntomodulin signaling via this pathway (Koole et al., 2010; Willard et al., 2012b; Wootten et al., 2013b). Probe dependence is also seen for effects on the extended GLP-1 peptides and exendin, where very limited augmentation of cAMP signaling is observed. A similar pattern of effect is observed for BETP in that this compound also augments oxyntomodulin signaling, but has minimal effect on GLP-1_{7–36 amide}–, exendin-, or GLP-1_{1–36 amide}–mediated cAMP production (Willard et al., 2012b; Wootten et al., 2013b). A weak potentiation of compound 2–mediated cAMP production (or surrogate readouts such as cAMP response element–driven luciferase reporter) has also been reported for interaction of this ligand with truncated forms of exendin, including exendin_5–39, exendin_7–39, and exendin_9–39, even in the absence of any measurable intrinsic activity (Coopman et al., 2010; Cheong et al., 2012).

Intriguingly, if the cooperative effect of BETP or compound 2 is studied over a broader array of signaling endpoints, then marked differences in effect are observed (Koole et al., 2012a; Li et al., 2012a; Wootten et al., 2012, 2013b). In contrast to the selective augmentation of cAMP response for oxyntomodulin, the allosteric ligands confer a similar, modest enhancement of arrestin recruitment for both GLP-1 and oxyntomodulin (Willard et al., 2012b; Wootten et al., 2013b), although a greater degree of negative cooperativity is seen on pERK responses of GLP-1_{7–36 amide} versus oxyntomodulin. Thus, the allosteric ligands also alter the signaling bias mediated by endogenous peptides. This is perhaps not surprising as both signaling bias and the cooperative effects of cobound ligands are driven through changes to the ensemble of conformations that the receptor samples (Wootten et al., 2013a). Indeed, the allosteric ligand-bound receptor can effectively be considered a distinct receptor for the interaction of the natural ligand(s).

Remarkably, both BETP and compound 2 very strikingly enhanced the cAMP response of the principal GLP-1 metabolite, GLP-1_{9–36 amide}, by up to ∼400-fold in the case of compound 2 (Li et al., 2012a; Wootten et al., 2012). Parallel augmentation of insulin secretion was observed in isolated rat islets and in vivo when pharmacological levels of the metabolite were coadministered with subthreshold levels of BETP (Wootten et al., 2012). The augmentation of signaling was principally limited to cAMP production, although weak potentiation of pERK and iCa²⁺ signaling was seen in HEK293 cells recombinantly expressing the receptor and INS-1E cells (Li et al., 2012a), but not in CHO cells recombinantly expressing the receptor (Wootten et al., 2012), providing further evidence that allosteric modulators can alter the signal bias of the receptor. Nonetheless, the lack of effect on the extended GLP-1 peptide indicates that the magnitude of the cooperative effect of BETP and compound 2 is not driven solely by the intrinsic efficacy of the activating peptide.

Ligands such as BETP and compound 2 are highly electrophilic and can form adducts with free cysteine residues (Eng et al., 2013). This is a feature of the action of these molecules to form, in particular, the covalent interaction with C347^6.36b in ICL3/TM6 that is critical for both the intrinsic efficacy and the cooperative allosteric effect (Nolte et al., 2014). Nevertheless, the covalent modification of the cysteine is insufficient to explain the full extent of activity of these compounds as BETP and compound 2 have differences in their profile of intrinsic efficacy (see above) and also in their cooperative effect (Wootten et al., 2013b). Interestingly, there is a naturally occurring polymorphism that can appear at position 333 of the human receptor, being either a serine or a cysteine with the latter occurring only rarely (Koole et al., 2011, 2015). However, the C333^ICL3 variant selectively abrogates compound 2–mediated cAMP production and cooperativity between compound 2 and oxyntomodulin for this pathway (Koole et al., 2011) while maintaining peptide-mediated signaling, suggesting that the environment surrounding C347^6.36b is important for the action of compound 2. In contrast to the effect of the C333^ICL3 polymorphism, the M149^1.44b polymorphic variant has no effect on the intrinsic efficacy of compound 2, but, as noted above, causes a very marked attenuation of peptide-mediated cAMP signaling with over 100-fold loss of GLP-1_{7–36 amide} or exendin potency. At this mutant, compound 2 could restore the potency of both GLP-1_{7–36 amide} and exendin to that of the wild-type receptor (Koole et al., 2011). Thus, allosteric modulators have the capacity to alleviate genetic disease arising from the loss of function of GLP-1R.

Disruption of the TM4 dimer interface of GLP-1R also abolishes the intrinsic efficacy of BETP and compound 2, but the cooperative augmentation of cAMP signaling by oxyntomodulin is maintained, indicating that the allosteric effect occurs in cis at least for this class of ligands (Harikumar et al., 2012).

Although compounds such as BETP and compound 2 were among the first recognized and the most widely studied allosteric modulators of GLP-1R, a number of additional compounds have been reported to act as modulators of peptide response. This includes the flavonol, quercetin, which lacks intrinsic activity, but could selectively augment calcium signaling of efficacious peptides such as GLP-1_{7–36 amide}, GLP-1_7–37, and exendin, albeit that an inhibition of response was seen with high concentrations of quercetin (Koole et al., 2010; Wootten et al., 2011). This augmentation of iCa²⁺ was dependent upon the presence of a 3-hydroxyl group on the flavone backbone and was improved when a 3′4′-dihydroxyl modification was present (Wootten et al., 2011). An additional series of compounds that could selectively enhance iCa²⁺ signaling was recently reported following high-throughput screening (Morris et al., 2014). Unlike the flavonols, this series of compounds had significant intrinsic efficacy for mobilization of iCa²⁺. An exemplar from this series, termed (S)-9b, was examined in further detail, in which it was reported to also augment liraglutide-mediated GLP-1R internalization in recombinant cells and exendin-mediated insulin secretion in primary mouse islets (Morris et al., 2014). This latter effect occurred in both high and low glucose, which would be problematic from a therapeutic standpoint, but may also require further investigation to determine whether this is mediated by GLP-1R. Likewise, compound (S)-9b reduced haloperidol-induced catalepsy in rats (Morris et al., 2014), but, again, it needs to be confirmed whether this effect is mediated via GLP-1R. In other work, using virtual screening against a molecular model of GCGR, de Graaf et al. (2011) identified a number of compounds that had activity at both GCGR and GLP-1R, including weak antagonists and one compound that acted as an inhibitor of glucagon-mediated cAMP production at GCGR, but as a weak positive modulator of GLP-1R–mediated cAMP response.

At this point in time, the physiologic or therapeutic implications of biased signaling are largely unknown. Despite differences in the signaling/regulatory profile of compounds such as Boc5 or TT15, these compounds (or similar analogs) have demonstrated efficacy in modulation of ex vivo and in vivo insulin secretion (Chen et al., 2007; He et al., 2012). Allosteric compounds such as BETP or compound 2 likewise can augment insulin secretion, albeit that this is complicated by endogenous circulating peptides. Pharmacologically, both BETP and compound 2 can augment responses of oxyntomodulin and the GLP-1 metabolite (GLP-1_{9–36 amide}), and this is linked to increased in vivo insulin secretion, at least in the context of threshold doses of the allosteric and orthosteric peptides (Willard et al., 2012b; Wootten et al., 2012). In vitro assessment of the effect of the modulators on the signaling profile of oxyntomodulin or GLP-1_{9–36 amide} indicates that their most prominent effect is to augment cAMP production, suggesting that this alone may be sufficient to improve insulin secretion, at least in the context of the baseline signaling of the peptides/modulators. The significance for GLP-1R function outside of insulin secretion is even less clear, as there has been very limited assessment of these other functions and none controlled for the influence of biased signaling. A major current limitation is the range of tools available to probe the significance of signaling differences. Most nonpeptidic compounds have low potency, nonfavorable pharmacokinetic profiles, or unknown interaction with other targets that limits their utility. Exendin is biased relative to GLP-1_{7–36 amide}, but this difference is subtle in the context of canonical signaling, making interpretation of physiologic data difficult. Oxyntomodulin is the most biased of the characterized peptides, displays good affinity for GLP-1R, and has distinctions in its physiologic actions from GLP-1, but dissecting the importance of this bias is complicated by its significant interaction with GCGR (Pocai, 2013). Nonetheless, this may be a useful tool to study biased signaling in GCGR knockout animals. Additional approaches to infer the potential role of selective activation of signaling/regulatory pathways include the generation of knock-in mice that have receptors designed to selectively abrogate signaling via a specific pathway (e.g., G protein versus arrestin), although the development of potent biased peptide ligands could also provide novel scope for better understanding this phenomenon.

1. Receptor Desensitization.

The activity of GLP-1R, like all GPCRs, is regulated by a coordinated balance between molecular mechanisms governing receptor signaling, desensitization, and resensitization. Receptor desensitization, or the reduced responsiveness of GPCR signaling to an agonist with time, is an important physiologic feedback mechanism that protects against both acute and chronic receptor overstimulation (Ferguson, 2001). This has important biologic and therapeutic implications.

GLP-1R undergoes rapid and reversible homologous and heterologous desensitization that has been observed in recombinant cell systems, in primary islets of Langerhans, and in insulinoma β cell lines (Fehmann and Habener, 1991; Gromada et al., 1996; Widmann et al., 1996a,b). Homologous desensitization refers to the loss of response to subsequent agonist stimulation following direct stimulation of the receptor and can occur through controlling the number of receptors present at the cell surface or by regulating the efficacy of the receptors at the cell surface. Heterologous desensitization describes receptor activation-independent regulation of receptors as well as mechanisms that occur after receptor activation that do not discriminate between activated, and nonactivated, receptors (Ferguson, 2001).

Acute incubation of islet cells with native GLP-1 induces rapid homologous GLP-1R desensitization in vitro (Fehmann and Habener, 1991; Gromada et al., 1996). In addition, exendin-4 also produces GLP-1R desensitization in islet cells following both acute and chronic exposure; however, this ligand is associated with a greater degree of desensitization compared with comparable incubations with GLP-1 (Baggio et al., 2004). Heterologous desensitization has been demonstrated in islets in response to multiple stimuli. For instance, GLP-1R heterologous desensitization was observed via acute exposure to phorbol myristate acetate (PMA), an activator of PKC (Widmann et al., 1996a,b; Baggio et al., 2004). In addition, prolonged hyperglycemia can induce desensitization, followed by downregulation with diminished responses to GLP-1R agonists that include weakened insulin secretion, reduced phosphorylation of CREB, impaired cAMP responses and PKA activity, and a reduction in GLP-1R mRNA (Rajan et al., 2015).

There is also the possibility that heterologous desensitization may occur due to activation of other receptors expressed in β cells that are involved in tightly controlling insulin secretion. To date, this has only been assessed with the ligand for GIPR. GLP-1R can heterodimerize with GIPR in vitro, with evidence that this heterodimerization can alter cellular signaling and trafficking profiles of GLP-1R (Schelshorn et al., 2012; Roed et al., 2015). However, despite their overlapping functions for signaling and insulin secretion, GIP does not produce meaningful heterologous desensitization of GLP-1R in islet cell studies (Rajan et al., 2015).

2. Underlying Mechanisms.

Desensitization is a consequence of a number of different mechanisms. They include uncoupling the receptor from heterotrimeric G proteins in response to receptor phosphorylation, the internalization of receptors to intramembranous compartments, and the downregulation of the total complement of receptors (Ferguson, 2001).

3. In Vivo Evidence of Desensitization.

Given the clinical interest in the therapeutic benefits of achieving sustained chronic elevations of GLP-1R agonists in the plasma through repeated or continual administration, the scope of GLP-1R desensitization has direct therapeutic relevance. Up to now, only a few studies have addressed the relevance of GLP-1R desensitization in vivo. Chronic or intermittent intracerebroventricular GLP-1 administration inhibited food intake and reduced weight gain in rats (Davis et al., 1998). Similarly, chronic intermittent administration of GLP-1R agonists to diabetic rodents is associated with inhibition of food intake, improvement of glycemia, and reduction in hemoglobin A1c (HbA1c), demonstrating that repeated administration of GLP-1 is not associated with a diminished therapeutic response in vivo (Szayna et al., 2000; Rolin et al., 2002; Kim et al., 2003).

Considering the prolonged activity and stability of GLP-1 mimetics compared with native GLP-1 (Young et al., 1999), it seems likely that islet cells and extrapancreatic GLP-1R would be exposed for a greater period of time with these ligands than the endogenous peptides. In vitro studies showed that exendin-4 is more potent than GLP-1 in producing GLP-1R desensitization; however, chronic exposure to exendin-4 in normal or transgenic mice that express exendin-4 was not associated with significant downregulation of GLP-1R–dependent responses coupled to glucose homeostasis (Baggio et al., 2004). Furthermore, patients treated with twice-daily exendin-4 or once-weekly liraglutide continue to exhibit a decrease in HbA1c and marked reductions in postprandial glycemic excursion (Buse et al., 2004; de Wit et al., 2014). Therefore, although in vitro experiments clearly demonstrate that GLP-1R has the capacity for desensitization, there is little evidence that it undergoes clinically meaningful desensitization in vivo in terms of glucose regulation.

There is, however, some limited evidence for desensitization occurring in vivo (Nauck et al., 2011). Administering native GLP-1 continuously into healthy human subjects for 8.5 hours, followed by assessment of glucoregulatory responses to liquid test meals given 5 hours apart with ongoing continuous GLP-1 infusion, demonstrated a reduced ability of GLP-1 to suppress gastric emptying and glucagon levels by the second test meal. In addition, levels of pancreatic polypeptide, a marker of vagal activation, were not as inhibited during the second test meal compared with the first. However, C-peptide and insulin levels were preserved with only a small reduction in the second meal. These studies reveal that even short-term continuous GLP-1R stimulation may be associated with some degree of rapid tachyphylaxis, most evident in effects mediated through the vagus nerve and gastric emptying (Nauck et al., 2011). Despite this evidence for in vivo desensitization, the physiologic significance of this is still unclear.

4. Receptor Internalization.

GLP-1R rapidly internalizes as a complex associated with its bound ligand. This has been observed for GLP-1, exendin-4, liraglutide, and compound 2 (Widmann et al., 1995; Kuna et al., 2013; Roed et al., 2014). In BRIN-BD11 cells, antibody-labeled GLP-1R and fluorescently labeled GLP-1 were colocalized at the perinuclear space following internalization (Kuna et al., 2013). Furthermore, internalization of fluorescently labeled GLP-1, exendin-4, and liraglutide has been observed in both recombinant cell systems and primary mouse pancreatic islets (Roed et al., 2014). There is also some evidence of ligand-directed bias with GLP-1R internalization, although this has not been directly quantified. The potencies for internalization by GLP-1 and exendin-4 were 10-fold higher than liraglutide, although kinetics of internalization were similar for all three ligands (Roed et al., 2014).

Currently, the mechanism for GLP-1R internalization in vivo is unclear, as there is evidence that this mechanism may be cell-type dependent (Widmann et al., 1995; Vazquez et al., 2005; Syme et al., 2006; Thompson and Kanamarlapudi, 2015). As GLP-1R is expressed in multiple tissues throughout the body with distinct physiologic functions depending upon the location of the receptor, this implies that the mechanism and role of receptor internalization and desensitization may be tissue-dependent. In different cell backgrounds, two distinct mechanisms of internalization have been observed, with both mechanisms dependent on dynamin (Widmann et al., 1995; Syme et al., 2006; Kuna et al., 2013).

Studies performed in CHO and CHL cells recombinantly expressing GLP-1R revealed internalization via clathrin-coated pits, although these studies indicated that there might not be complete internalization via this method (Widmann et al., 1995; Vazquez et al., 2005). Studies in CHL fibroblasts demonstrated that the same three phosphorylation sites linked to homologous desensitization of GLP-1R–mediated cAMP responses (441/442, 444/445, and 451/452) are also important for internalization, supporting a correlation between phosphorylation at these sites and internalization (Widmann et al., 1997). However, the exact contribution of each phosphorylation site to desensitization of the cAMP response and internalization is different, indicating that the precise molecular basis for control of receptor desensitization and endocytosis is, at least in part, distinct. In addition to phosphorylation of these serine doublets, a region of the C-terminal tail close to the bottom of TMD7 is also of importance in mediating agonist-dependent GLP-1R internalization (Vazquez et al., 2005).

Traditionally for GPCRs, β-arrestin recruitment targets receptors for clathrin-mediated internalization (Ferguson, 2001). However, the literature indicates that this may not be the case for GLP-1R in some cell backgrounds, despite the ability of the receptor to interact with these proteins (Syme et al., 2006). Although there is little literature on the role that β-arrestin 2 plays in GLP-1R internalization and desensitization, there is some evidence that these processes can occur independently of β-arrestin recruitment, in a clathrin-independent mechanism. Studies performed in HEK293, COS-7, and MIN6 cells exhibited caveolin-1 regulation of internalization (Syme et al., 2006; Thompson and Kanamarlapudi, 2015). GLP-1R contains a classic caveolin-1–binding motif within the second intracellular loop such that it interacts and colocalizes with the protein intracellularly. Occurrence of this interaction was also supported by studies using inhibitors of caveolin-1. This mechanism of GLP-1R internalization has been proposed to occur through activation of Gαq proteins, followed by activation of PKC (Thompson and Kanamarlapudi, 2015). In support of β-arrestin–independent internalization, it was shown that knockdown of β-arrestin 1 in INS-1 cells had no effect on internalization or desensitization of GLP-1R (Sonoda et al., 2008).

Although internalization can contribute to desensitization of receptors, internalization and desensitization are two distinct phenomena. Internalization has the potential to contribute to termination of some GLP-1R–mediated signaling events; however, it does not terminate all components of this signaling (Kuna et al., 2013; Roed et al., 2015). A recent study revealed that inhibition of internalization using a dominant-negative form of dynamin (K44E) resulted in decreased cAMP formation, ERK1/2 phosphorylation, and calcium signaling, suggesting that at least some of the signaling mediated by these pathways occurs via internalized receptors. In addition, inhibition of GLP-1R internalization also attenuates insulin secretion, confirming that internalization is an integral part of the signaling process for this receptor (Kuna et al., 2013). Using fluorescently labeled ligands and receptors, studies in recombinant systems displayed colocalization of internalized GLP-1R with adenylate cyclase in endosomes (Kuna et al., 2013). Taken together, these findings all imply that GLP-1R continues to signal following internalization, and that internalization is not necessarily part of the mechanism for desensitization. This level of spatiotemporal control achieved by the compartmentalization of signaling through internalization may be vital to specific, fine-tuned responses from GLP-1R activation and may have substantial implications both physiologically and therapeutically for targeting this receptor in disease management.

5. Receptor Resensitization.

GLP-1R resensitization has not been extensively studied. Following desensitization, GLP-1R–mediated calcium signaling resensitizes after removal of extracellular ligand within 1 hour and does not require de novo protein synthesis (Gromada et al., 1996). After internalization, GLP-1R colocalizes with transferrin, a marker for recycling endosomes, suggesting that it is recycling GLP-1R that is returned back to the cell surface after internalization (Roed et al., 2014). Moreover, there is evidence in HEK293 cells that GLP-1R recycles back to the cell surface in response to GLP-1, exendin-4, and liraglutide, even after prolonged agonist treatment. However, recycling was two to three times slower for exendin-4 and liraglutide compared with GLP-1 (Roed et al., 2014). In the presence of GLP-1, receptors were found in transferrin-positive endosomes up to 45 minutes post ligand addition, whereas in the presence of exendin-4 and liraglutide colocalization of internalized GLP-1R and transferrin was observed for up to 60 minutes following ligand addition. Resensitization is important, as irreversible desensitization would leave a cell unable to respond to external stimuli, and therefore this mechanism protects against prolonged desensitization (Ferguson, 2001). Nonetheless, to date, the mechanism of GLP-1R resensitization remains largely unknown and requires further investigation.

There is also evidence of partial degradation of GLP-1R upon prolonged activation with exendin-4, liraglutide, and GLP-1, with colocalization observed between the receptor and lysosomes (Kuna et al., 2013; Roed et al., 2014). This mechanism leads to downregulation of receptor numbers.

1. Receptor Trafficking.

Endocytosis of GPCRs and postendocytic trafficking between recycling and endosomal degradation are fundamental mechanisms that control the signaling capacity of receptors. Syme et al. (2006) studied the trafficking of green fluorescent protein–fused GLP-1R stably expressed on HEK293 cells in response to a GLP-1 agonist. It was found that GLP-1R associates with caveolin-1 in the lipid rafts of the cell membrane of HEK293 cells transfected with the green fluorescent protein–GLP-1R plasmid and MIN6 cells endogenously expressing the GLP-1 receptor. With confocal fluorescence microscopy and immunoprecipitation techniques, it was observed that stimulation by an agonist induces rapid and extensive internalization of the receptor in association with caveolin-1, and nearly all of the receptor was internalized after 15 minutes.

Similarly, Roed et al. (2014) combined microscopy and a time-resolved fluorescence resonance energy transfer–based technique for monitoring both receptor trafficking and signaling in living cells. The GLP-1 receptor rapidly and repeatedly internalizes and recycles upon ligand activation. GLP-1R recycling was found to be two to three times slower when induced by exendin-4 and liraglutide, compared with GLP-1 at equipotent concentrations. Independent of the ligand, activated receptors demonstrated cycling for a prolonged period of time as well as sustained cAMP signaling.

2. Immunoassays.

Determination of incretins, especially endogenous ones, by immunologic assays is considered difficult because differential processing of precursor molecules gives rise to a number of different peptides that may crossreact with antisera raised against GLP-1 (Deacon and Holst, 2009). Subsequent degradation of the peptides by DPP-4 and NEP 24.11 further complicates the assay process (Heijboer et al., 2011).

When the full processing pattern of proglucagon was not fully elucidated, the assays were directed to the middle region of GLP-1 and became independent to the processing pattern. This caused peptides with a deletion or extension at either the N or C terminus to be detected in addition to GLP-1 itself. Antisera were then made toward either the C terminus (Hendrick et al., 1993; Holst et al., 1994; Ørskov et al., 1994) or the N terminus (Gutniak et al., 1996). However, it was found that C-terminally directed antisera reacted with peptides truncated or extended at the N terminus and could not distinguish the full-length GLP-1 from the metabolite resulting from DPP-4 degradation. Similarly, N-terminally directed antisera could not differentiate the intact peptide from truncated or extended peptides at the C terminus.

Kirk and coworkers developed a sandwich enzyme-linked immunosorbent assay for determining exogenous GLP-1_{7–36 amide} in plasma samples collected from pharmacokinetic studies (Pridal et al., 1995). It is referred to as active GLP-1 assay. This experiment employs an N-terminally directed antibody (polyclonal rabbit antibody directed to the residues at position 7–14) and a C-terminally directed antibody (monoclonal mouse antibody directed to the residues at position 26–33). When compared with a radioimmunoassay employing a polyclonal rabbit antibody directed to an epitope in the midregion of the peptide, it was found to be superior due to reduced crossreactivity with GLP-1 fragments. It has a working range from 10 to 500 pmol/L.

3. Pancreatic β Cell Regeneration.

Besides insulin secretion from the pancreatic β cells in a glucose-dependent manner, GLP-1 can induce regeneration of the cells evidenced by immunohistochemical staining studies (Edvell and Lindstrom, 1999; Xu et al., 2006). PDX-1 is essential for pancreogenesis, pancreatic cell differentiation, and maturation. Perfetti and coworkers demonstrated that continuous infusion of GLP-1 to young and old rats upregulates PDX-1 expression in islets and total pancreas (Hui and Perfetti, 2002). They also found induced pancreatic cell proliferation and β cell neogenesis (Perfetti et al., 2000).

1. GLP-1 Analogs.

Several GLP-1 analogs have successfully reached the market and become an important treatment of obesity and T2DM today (Table 3). Several reviews have been published during the past that describe the approved and emerging GLP-1R agonists and also summarize the most important clinical trials as well as toxicological considerations (Lund et al., 2011; Madsbad et al., 2011; Garber, 2012; Meier, 2012; Montanya, 2012; Lorenz et al., 2013; Trujillo and Nuffer, 2014; Trujillo et al., 2015; Madsbad, 2016; Zaccardi et al., 2016).

Taspoglutide (BIM-51077; Roche, Basel, Switzerland) is a potent GLP-1 analog with 93% homology to human GLP-1_{7–36 amide} as it has two amino acid substitutions in positions 8 and 35 with aminoisobutyric acid (Aib). Taspoglutide is protected against DPP-4 degradation by the Aib⁸/Aib³⁵ substitution but is still cleared rapidly by the kidney. A sustained-release formulation that contains zinc-chloride and precipitates as a sub cutis deposition was developed. This once-weekly dosing formulation of taspoglutide has shown promising antidiabetic efficacy in clinical trials. Unfortunately, the analog was discontinued in phase 3 clinical trials due to severe gastrointestinal side effects (Retterstol, 2009; Dong et al., 2011; Garber, 2012).

The GLP-1 human serum albumin fusion protein, albiglutide, also known as albugon, discovered by Genentech (South San Francisco, CA) and developed by GlaxoSmithKline (Brentford, UK), was approved by the Food and Drug Administration (FDA) in 2014. To have a free N-terminal, which is important for GLP-1 activity, the peptide was fused via the C terminus to albumin. The construct is composed of a tandem repeat of Ala⁸ to Gly⁸ GLP-1_7–36 that is fused to the N terminus of human albumin. Gly⁸ GLP-1 was used to protect for DPP-4 degradation of the N terminus. The plasma half-life was extended to about 6–8 days, which made this fusion construct applicable for once-weekly dosing. One drawback relates to a significantly reduced potency that was probably due to a combination of the Gly⁸ modification and the covalent attachment to the larger protein human serum albumin. The tandem repeat of GLP-1 was used to obtain longer distance between albumin and the distal GLP-1 peptide, but the affinity for GLP-1R is 0.61 nM compared with 0.02 nM of the native peptide. The approved initial dose to treat T2DM is 30 mg given s.c. once per week, which may be increased to 50 mg on the same day of each week if the glycemic response is inadequate (Kim et al., 2003; Tomkin, 2009; St Onge and Miller, 2010).

Dulaglutide (LY2189265; Eli Lilly, Indianapolis, IN) is a GLP-1R agonist fused to a modified Fc fragment of immunoglobulin that acts as a carrier to secure long residence time by reduced in vivo clearance. The sequence of the fusion protein Gly⁸Glu²²Gly³⁶-GLP-1_7–37-(Gly₄Ser)₃Ala-Ala^234,235Pro²²⁸-IgG₄-Fc is modified in the N terminus of GLP-1 to protect against DPP-4 cleavage (Gly⁸), and the IgG₄ Fc is modified to reduce interaction with high-affinity Fc receptors and avoid antibody-dependent cell-mediated cytotoxicity. The linker (Gly₄Ser)₃Ala was carefully selected to retain high GLP-1R affinity as direct fusion dramatically reduced in vitro activity by ∼95%. The half-life is above 1.5 days in rats and more than 2 days in monkeys. Dulaglutide was approved by the FDA in 2014 for once-weekly dosing of 0.75 mg s.c. (Glaesner et al., 2010).

There is still a need for optimization of once-weekly GLP-1 analogs because both albiglutide and dulaglutide are less efficacious than liraglutide with respect to weight loss (Dungan et al., 2014; Pratley et al., 2014).

The next generation of GLP-1 analogs using the reversible albumin affinity is represented by semaglutide (Novo Nordisk, Bagsværd, Denmark) currently in phase 3 clinical trials. This analog Aib⁸Lys²⁶Arg³⁴-GLP-1_7–37 has a more complex side chain composed of hexadecandinoyl attached to Lys²⁶ via an L-γ-glutamic acid linker and a small hydrophilic spacer. The Aib⁸ was introduced to improve the DPP-4 stability beyond that of liraglutide, and the new side chain increased the albumin affinity, resulting in a half-life in humans long enough for once-weekly dosing. In a phase 2 study over 12 weeks, semaglutide dose-dependently reduced HbA1c and body weight, with higher doses being more effective than liraglutide (Nauck and Sesti, 2012; Trujillo and Nuffer, 2014; Lau et al., 2015).

2. Exenatide Experience.

In 1993, exendin-4 was isolated from the salivary glands of the gila monster (Heloderma suspectum) and found to be highly active on GLP-1R (Goke et al., 1993a). Exendin-4 is a 39–amino-acid peptide that shares approximately 50% amino acid sequence identity with human GLP-1. The discovery of the antidiabetic effects of exendin-4 increased the interest in this peptide, and several pharmaceutical companies thus focused on its clinical development as well as on discovery efforts in exendin-based GLP-1 analogs.

In 2005, the first GLP-1R agonist exenatide (AstraZeneca, Cambridge, UK) was approved by the FDA for treatment of T2DM. Exenatide, as well as other GLP-1R agonists, reduces blood glucose through multiple mechanisms, including enhancement of glucose-dependent insulin secretion, suppression of excess glucagon secretion, reduction of food intake, and slowing of gastric emptying. It is administered s.c. twice daily with doses between 5 and 10 μg (Cvetkovic and Plosker, 2007).

The very low therapeutic dose of exenatide made it an excellent candidate for an extended release formulation. Amylin Pharmaceuticals (San Diego, CA), Eli Lilly, and Alkermes (Dublin, Ireland) developed a formulation to prolong exenatide release, using biodegradable polymeric microspheres that entrap exenatide. The microspheres consist of exenatide incorporated into a matrix of poly (D,L-lactide-co-glycolide), which is a common biodegradable medical polymer with a history of safe use in humans. The final formulation was approved in 2011 as Bydureon for once-weekly dosing (Kim et al., 2007; Malone et al., 2009).

Lixisenatide (Sanofi, Paris, France) is a once-daily short-acting GLP-1R agonist used in the treatment of T2DM. The discovery of lixisenatide was inspired by the development of structure-inducing probe technology, which improves the half-life of structure-inducing probe conjugates compared with the parent peptide by enhancing stability and, as a consequence, creating resistance to proteolytic degradation. Lixisenatide, des-Pro³⁶-exendin-4_1-39-Lys⁶ amide, comprises 44 amino acids and is based on the exendin-4 sequence with a proline deletion in position 36 and C-terminal extension with six lysines. Lixisenatide has an extended in vivo half-life of 91 minutes after i.v. administration (in rabbits) compared with 43 minutes for exendin-4. Lixisenatide was approved in 2013 and has a substantial sustained effect on gastric emptying and postprandial glucose excursions that may be related to the relatively shorter half-life. The combination with basal insulin to lower fasting plasma glucose is clinically valuable and could differentiate lixisenatide from other GLP-1R agonists, especially from those long-acting ones with little effect on gastric emptying and postprandial glucose (Werner et al., 2010; Christensen et al., 2014; Rosenstock et al., 2014).

3. GLP-1 Mimetics.

A novel class of 11–amino-acid GLP-1R agonists was discovered that consist of an analog of only the first 9 amino acids of GLP-1, in which the remaining 21 amino acids have been truncated and attached to an unnatural biphenyl diamino acid at the C terminus. The optimization of the sequence with several unnatural amino acids provided 11-mer GLP-1 mimetics with similar in vitro potency as the natural 30-mer peptide. The optimized 11-mers (e.g., BMS-686117) were reported to reduce plasma glucose and increase plasma insulin concentrations in diabetic ob/ob mice after a single relative high s.c. dose of the peptide (Mapelli et al., 2009; Haque et al., 2010). Based on NMR studies, the structures of these 11-mer peptides were determined and used to design lactam- and disulfide-based cyclic constrained peptides that retained high GLP-1R potency (Hoang et al., 2015).

BMS-686117 has a relatively shorter half-life of about 2 hours in dogs, and, accordingly, a sustained release formulation is required for clinical use. Formulation with Zn (II) forms a poorly water-soluble adduct such that the BMS-686117 content and reversible release of which can be tailored by varying the Zn:peptide ratio. An in vivo pharmacokinetics study in dogs of an optimized sustained release formulation concluded that a minimal initial burst and constant release are applicable to once-daily dosing (Qian et al., 2009).

1. Selection.

As the number of GLP-1R agonists in clinical use and trials expands, it will become clearer whether there are advantages of one drug over another in a specific clinical setting. We have already observed differences in the frequency of nausea, presumably reflecting the reduced gastric emptying, among these drugs. There is also variation in the pathway of inactivation and excretion of these drugs, making the selection of agent in patients with renal insufficiency important. Although biased agonism is now recognized for some of these agonists, there is still insufficient information on the broad pharmacodynamics properties of these drugs to differentiate the contribution of biased agonism from pharmacokinetic behavior to differential clinical responses, side effects, or toxicities, although this is likely to be an important consideration in future therapeutic development.

A very practical and useful categorization of GLP-1R agonists is based on their duration of action. The shortest-acting agents, like exenatide and lixisenatide with half-lives of 2–5 hours, seem to exert their beneficial effects predominantly by slowing gastric emptying and thereby reducing the rate of carbohydrate delivery, digestion, and absorption, leading to normal postprandial glucose levels. Of course, other nutrients also empty more slowly, and this has resulted in lower postprandial levels of free fatty acids and triglycerides, which has some metabolic benefits. For instance, decreased levels of chylomicrons are desired in patients with diabetes or metabolic syndrome. As the duration of GLP-1R agonist action is extended with peptides like albiglutide, dulaglutide, long-acting release exenatide, or liraglutide (half-lives of 12 hours to several days), the impact on blood glucose levels is enhanced as insulinotropic action and activity to reduce glucagon become more prominent than the effect on gastric emptying. Such prolonged activities have also been associated with increased side effects, such as diarrhea, tachycardia, immune responses, and injection site reactions. Of interest, nausea and vomiting appear to have greater incidence with the shorter-acting peptides. The longer-acting agents have typically provided more consistent effects on GLP-1R activation and blood glucose levels. Because the insulinotropic action of GLP-1 is dependent on the presence of hyperglycemia, long duration activity does not increase the risk of hypoglycemia. Therefore, long-acting release formulations have obvious advantages if the goal is to lower early morning glucose levels, because they retain efficacy throughout the night.

It is curious whether the induction of nausea, which is believed to be a result of slowed gastric emptying, is similar in the short-acting and longer-acting GLP-1R agonists, yet this seems to be attenuated quickly in the longer duration agonists and may take weeks to months to improve for the shorter duration agonists. Such a phenomenon most likely reflects differences in desensitization of GLP-1R, but the underlying mechanism has not been carefully studied. Although the weight loss effect seen with some of these GLP-1R agonists was thought to be related to slowed gastric emptying, the long-acting agents seem to result in greater weight reduction than the short-acting ones, suggesting certain other mechanism is dominant. This might relate to reduced appetite, potentially following GLP-1 action in the hypothalamus or other CNS regions.

2. Efficacy.

The first phase 2 trial to study a GLP-1R agonist in T2DM was started in 1999, comparing twice-daily administration of exenatide to saline (Kolterman et al., 2003). Not only was the exenatide effective in reducing glucose levels, but it did not induce hypoglycemia in patients who were on insulin therapy. Although exenatide was approved by the FDA for use in diabetes in 2005, concerns about hypoglycemia in insulin-requiring patients persisted, and this was finally studied in 2008 (Buse et al., 2011). The dramatic observations of enhanced reduction in HbA1c and no increase in hypoglycemic episodes, as well as weight loss, formally launched a new series of clinical studies to follow. A meta review published in Lancet in 2014 (Eng et al., 2014) chose 15 of 2905 studies that represented 4348 subjects to evaluate. HbA1c levels were decreased by an average of 0.44%, and weight was reduced by an average of 3.22 kg, without any increase in hypoglycemic episodes. Of note, the levels of HbA1c were not different using GLP-1R agonist with boluses of insulin, but the frequency of hypoglycemic episodes was significantly reduced with the former.

Short-acting GLP-1R agonists known to better control postprandial glucose levels have not been directly compared with that of long duration administered with basal insulin. Each has theoretical advantages and perhaps should be chosen based on the characteristics of the patient. Because long-acting agents can be given with short-acting insulin at meal time now, another important option exists, but needs to be evaluated.

The DURATION-3 trial (Diamant et al., 2014) compared long-acting release exenatide administered once per week to insulin glargine (a long-acting, man-made version of human insulin) in 456 patients with T2DM who had not achieved satisfactory glucose control with maximal tolerated doses of oral agents. Exenatide was able to better manage the blood glucose, with HbA1c levels being reduced by 1.01%, whereas glargine plus oral agents only reduced this by 0.81%. Again, the most notable observation was a much lower incidence of hypoglycemia in patients taking the GLP-1R agonist.

The AWARD-4 study did use dulaglutide injected only once per week and rapid-acting insulin with each meal to demonstrate improved levels of HbA1c than were observed in basal and bolus insulin administration (Jendle et al., 2014). Fixed-dose combinations of GLP-1R agonist and basal insulin have substantial appeal for convenience, although, at the current time, expense is a major concern. The DUAL-1 and DUAL-2 studies exhibited a 1.9% reduction in HbA1c in patients who followed a titration algorithm with a pen delivery device (Buse et al., 2011; Gough et al., 2014).

3. Benefits.

In addition to its action on blood glucose, mediated largely by impact on the pancreatic islet β and α cells, GLP-1R agonists have been shown to display favorable cardiovascular effects. This includes improvements in serum lipids, myocardial contractility, and endothelial function (Mudaliar and Henry, 2010). The effects on lipids are likely to reflect slowed gastric emptying and reduced delivery for absorption, as well as impact on hepatic and fat cell metabolism. Presumably, the direct cardiovascular benefit was achieved via abundant levels of GLP-1R expression in cardiomyocytes and blood vessels (Nolte et al., 2014). Experiments in rodents showed that GLP-1R agonists can reduce myocardial infarct size and improve left ventricular function, including better wall motility and ejection fractions (Nikolaidis et al., 2004b; Noyan-Ashraf et al., 2009). There also exists encouraging evidence in rodents of the neuroprotective role of GLP-1 for AD, PD, and even stroke (Harkavyi et al., 2008; McClean et al., 2011; Teramoto et al., 2011), although this has not been followed up by clinical studies.

A recent meta-analysis explored the impact of GLP-1R agonists on lipids in clinical trials (Sun et al., 2015). Thirty-five trials with 13 treatments lasting a minimum of 8 weeks were included, and GLP-1R agonists were associated with modest improvements in total cholesterol, low-density lipoprotein cholesterol, and triglyceride levels, whereas there was no change in high-density lipoprotein cholesterol. It is not yet clear whether this translates into an improvement in cardiac and vascular events. It follows that a latest retrospective review of macrovascular outcomes in patients with T2DM who had been treated with insulin alone or with GLP-1R agonists (Paul et al., 2015) suggested that heart failure, myocardial infarction, and stroke were less common in patients who were taking the drug. Although such epidemiologic studies need to be interpreted carefully, this is inspiring and clearly worthy of prospective evaluation. A nested case-control study was also recently performed (Gejl et al., 2015). T2DM patients who developed one or more of these cardiovascular problems were compared with control patients with diabetes who did not develop such cardiovascular events. More than 10,000 patients were included, and 1,947 had one or more of these events. Liraglutide was associated with significant risk reduction, with the degree of reduction being dose- and duration-dependent. Another retrospective cohort study also showed a reduction in heart failure in patients with diabetes (Velez et al., 2015). More than 19,000 adult patients with diabetes were studied and matched with twice as many controls, and 1,426 patients with diabetes used GLP-1R agonists. This was also encouraging, but only points toward the need for prospective studies.

There is increasing excitement about the potential role for fixed-dose combinations of a long-acting insulin preparation and a GLP-1R agonist as an adjunct to diet and exercise to improve glycemic control in adults with T2DM. Two such preparations have recently been recommended for FDA approval by advisory committees, based on portfolios of clinical trials, including double-blinded designs. These include IDegLira from Novo Nordisk, combining Tresiba (insulin degludec) and Victoza (liraglutide) into a once-daily injectable product (http://www.fda.gov/downloads/advisorycommittees/committeesmeetingmaterials/drugs/endocrinologicandmetabolicdrugsadvisorycommittee/ucm502074.pdf), and iGlarLixi from Sanofi, combining Lantus (insulin glargine) and Lyxumia (lixisenatide) into a similar product (http://www.fda.gov/downloads/advisorycommittees/committeesmeetingmaterials/drugs/endocrinologicandmetabolicdrugsadvisorycommittee/ucm502559.pdf). Both of the component drugs in IDegLira and the insulin glargine in iGlarLixi are already FDA approved, whereas lixisenatide is currently undergoing independent regulatory review. The efficacy of both combination products was superior to that of the components alone, leading to the successful outcomes. The safety of them appears to be consistent with that of the individual components, and there are indications that some side effects like nausea and weight gain may actually be less in the combination preparations than for some individual components. It seems likely that at least one of these products or something alike will achieve a prominent role in therapy soon.

1. Neurologic Disorders.

The expression of GLP-1R in multiple sites of the nervous system and the involvement of GLP-1 in satiety and food intake regulation are well-established (II. Glucagon-Like Peptide-1). Because this incretin was also shown to improve learning in rats (Oka et al., 2000) and exert neurotrophic or neuroprotective effects when given intracerebroventricularly, its potential therapeutic value for AD, PD, and pain has been investigated (During et al., 2003; McClean et al., 2011; Gong et al., 2014b).

AD is characterized by progressive cognitive decline with a defined neuropathology of Aβ production and tau hyperphosphorylation (Claeysen et al., 2012). T2DM has been identified as a risk factor for AD, and an epidemiologic study indicated that 65% of T2DM patients showed evidence of featured AD plaques in their autopsied brains (Bassil et al., 2014), indicating that insulin desensitization in the periphery may be a factor in initiating or accelerating the development of neurodegenerative processes. It has led to the notion that drugs developed for the treatment of T2DM may be beneficial in modifying the pathophysiology of AD. In preclinical models of AD, GLP-1 decreases Aβ toxicity in vitro (During et al., 2003; Perry et al., 2003). A follow-up study demonstrated that GLP-1 and exendin-4 could protect cultured rat hippocampal neurons against glutamate-induced apoptosis (Perry et al., 2003). Furthermore, Val⁸GLP-1 blocked synaptic degradation and rescued synaptic plasticity in the hippocampus (Gengler et al., 2012). The neuroprotection effect of GLP-1 was further displayed in studies in which treatment with liraglutide, lixisenatide, or geniposide in the transgenic amyloid precursor peptide/presenilin 1 mice produced improvements in synaptic plasticity and reduced neuronal damage, tau hyperphosphorylation, microglial activation, plaque, and oligomer formation in the CNS (Perry et al., 2002; McClean et al., 2011; Lv et al., 2014; McClean and Holscher, 2014). These GLP-1 analogs also attenuated memory deficits, restored impaired signaling within the Akt/GSK-3β pathway, and reversed elevated levels of reactive oxygen species in amyloid precursor peptide/presenilin 1 mice. The recent reported effect of liraglutide on an early-stage sporadic AD model supports the notion that liraglutide-induced GLP-1R activity might be a viable target in AD therapy (Hansen et al., 2015).

PD is typified by a loss of dopaminergic neurons and cellular degeneration in the striatum (Le et al., 2009). Epidemiologic data suggest an association between T2DM and some neurodegenerative disorders, such as PD (Pressley et al., 2003). In patients with PD, it was found that the levels of insulin receptors were markedly reduced in the basal ganglia and the substantia nigra (Moroo et al., 1994). Furthermore, increased IRS2 phosphorylation, a marker of insulin resistance, was found in the basal ganglia of the 6-hydroxydopamine lesion rat PD model (Morris et al., 2008). An earlier report demonstrated that in high-fat-diet–fed rats, insulin resistance was accompanied by attenuated dopamine release and diminished dopamine clearance in the basal ganglia (Morris et al., 2011). Neuropathological studies in patients with PD also revealed that loss of insulin receptor immunoreactivity and mRNA coincides with loss of tyrosine hydroxylase mRNA (the rate-limiting enzyme in dopamine synthesis) (Takahashi et al., 1996).

GLP-1 and exendin-4 have been shown to exert a neurotrophic effect on 6-hydroxydopamine–treated dopamine neurons (Li et al., 2010). A growing body of evidence now exists to support the neurotrophic properties of GLP-1: activating GLP-1R–signaling pathways in neurons leads to proliferation and differentiation of cells from precursors into neurons, thereby drawing a striking similarity between cellular responses in pancreatic β cells and neurons (During et al., 2003). Indeed, such a protective action was observed in rodent models of PD: GLP-1 and its mimetics were effective to protect tyrosine hydroxylase–positive dopaminergic neurons and to preserve dopamine levels in 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine–treated animals. Treatment with a GLP-1R agonist resulted in improved motor function, as demonstrated by rotarod and pole tests (Kim et al., 2009; Li et al., 2009). Recently, a single-blind clinical trial of exendin-4 in PD patients showed that exenatide was well tolerated, and the treated group exhibited clinically relevant improvements across motor and cognitive measures compared with the control group (Aviles-Olmos et al., 2013).

In a study to investigate glucoregulatory effects of exendins, a research group in Shanghai serendipitously discovered that GLP-1 produced a potent antinociception in the formalin test (Gong et al., 2011). They further examined the inhibitory role of the spinal GLP-1R signaling pathway in pain hypersensitivity states and its mechanism of action. Specific expression of GLP-1R on the spinal dorsal horn microglia was found, and the efficacy as well as potency of GLP-1, exenatide, and geniposide on antinociception in a variety of animal models of pain hypersensitivity versus acute nociceptive responses were evaluated (Gong et al., 2014b). The involvement of GLP-1 in antinociception was also verified by a nonpeptidic GLP-1R agonist (WB4-24) indicating that the effect is mediated by GLP-1R (Gong et al., 2014a; Fan et al., 2015).

2. Oncological Association.

As previously described above, GLP-1–based therapy is a long-term approach to the control of metabolic disorders exemplified by T2DM, whereby chronic adverse events become a major concern, including an increased risk of cancer (Esposito et al., 2012). Pancreas and thyroid are the main tissues of the concern.

Since the first case report of exenatide-induced pancreatitis in 2006 (Denker and Dimarco, 2006), increasing numbers of controversial observations were reported associated with the risk of pancreatitis when treating T2DM patients with GLP-1R agonists or DPP-4 inhibitors. Acute pancreatitis may potentially progress to chronic pancreatitis and ultimately develop pancreatic cancer (Yachida et al., 2010; Nauck and Friedrich, 2013). Safety alerts have been issued by the FDA for exenatide and sitagliptin in 2008 and 2009, respectively, for patients who have a history of pancreatitis or current symptoms suggestive of pancreatitis. In 2013, a workshop sponsored by the National Institute of Diabetes and Digestive and Kidney Diseases and the National Cancer Institute on Pancreatitis-Diabetes-Pancreatic Cancer was held in Bethesda, MD (Andersen et al., 2013). Based on current data, neither the GLP-1R agonist liraglutide nor the DPP-4 inhibitor sitagliptin established a solid link with the increase of risk for pancreatic ductal adenocarcinoma in the therapy of patients with T2DM. Ongoing surveillance will be important in future outcome and epidemiologic studies that will be performed as the indications for use of these drugs expand; however, existing studies have been reassuring to date.

The effects of GLP-1R activation by liraglutide on human pancreatic cancer cells, MIA PaCa-2 and PANC-1, were studied recently (Zhao et al., 2014). Liraglutide dose- and time-dependently increased GLP-1R expression, arrested cell cycles in the S phase, and induced apoptosis through stimulation of cAMP production and inhibition of Akt and ERK1/2 pathways. Pancreatic tumor growth was also attenuated by liraglutide in a mouse xenograft model in vivo. In patients with pancreatic cancer, 43.3% were GLP-1R–negative (13 of 30) and, in GLP-1R–positive tissues, tumor size was inversely correlated with GLP-1R expression. This is the only report of the cytoreductive effect of GLP-1R signaling activated by liraglutide on pancreatic cancer cells (Zhao et al., 2014).

Thyroid cancer is a relatively rare disease, although its incidence has increased at an alarming rate in both men and women in the United States (Aschebrook-Kilfoy et al., 2011; Udelsman and Zhang, 2014). C cells constitute a minor fraction of the thyroid mass (0.1%) and are the only cells that produce calcitonin (Huang et al., 2006). GLP-1R agonists are capable of inducing increased calcitonin gene expression and C cell hyperplasia in the thyroid of wild-type, but not the GLP-1R knockout (Glp-1r^−/−) mice (Yamada et al., 2008; Madsen et al., 2012). Long-term exposure to GLP-1R agonists resulted in C cell proliferation and the formation of C cell adenomas and (medullary thyroid) carcinomas in rodents (Elashoff et al., 2011; Bulchandani et al., 2012; Rosol, 2013). Data derived from the FDA Adverse Event Reporting System database also indicate a significantly elevated risk for thyroid cancer (histology not specified) with exenatide, but not sitagliptin treatment (Nauck and Friedrich, 2013). Several cases of thyroid cancer were reported during the liraglutide clinical development program (Neumiller et al., 2010; Elashoff et al., 2011). Thus, a Black Box warning regarding the risk of thyroid C cell cancer was labeled in all approved long-acting GLP-1 analogs (Parks and Rosebraugh, 2010; Andersen et al., 2013). However, a meta-analysis of serious adverse events reported with GLP-1R agonists indicates that neither liraglutide nor exenatide had an association with an increased risk of thyroid cancer (Alves et al., 2012). Such a discrepancy between humans and rodents may reside on the different expression level of GLP-1R in the thyroid because C cells are very sparse in primates when compared with rodents (Gallo, 2013; Pyke et al., 2014). Adequately powered long-term epidemiologic studies will be necessary to clarify the association between GLP-1–based therapies and the risk of thyroid cancer (Andersen et al., 2013).

Besides the pancreas and thyroid, GLP-1R is widely expressed in other tissues, such as the stomach, pituitary, heart, lung, kidney, and nervous system. Analysis of the FDA Adverse Event Reporting System database suggests that for several malignancies, excluding pancreatic and thyroid cancers, exenatide and sitagliptin apparently significantly reduced the odds ratios of some special cancers, such as lung cancer, prostate cancer, lymphoma/multiple myeloma for exenatide, and colon cancer and prostate cancer for sitagliptin (Vigneri et al., 2009; Koehler et al., 2011; Nauck and Friedrich, 2013). However, the data need to be interpreted with caution because the results are based on relatively small numbers of tumors and most likely are subject to reporting bias.