Allosteric Modulation of the Activity of the Glucagon-like Peptide-1 (GLP-1) Metabolite GLP-1 9–36 Amide at the GLP-1 Receptor

Glucagon-like peptide-1 (GLP-1) released from intestinal L cells in response to nutrients has many physiological effects but particularly enhances glucose-dependent insulin release through the GLP-1 receptor (GLP-1R). GLP-1 7–36 amide, the predominant circulating active form of GLP-1, is rapidly truncated by dipeptidyl peptidase-4 to GLP-1 9–36 amide, which is generally considered inactive. Given its physiological roles, the GLP-1R is targeted for treatment of type 2 diabetes. Recently ‘compound 2’ has been described as both an agonist and positive allosteric modulator of GLP-1 7–36 amide affinity, but not potency, at the GLP-1R. Importantly, we demonstrated previously that exendin 9–39, generally considered a GLP-1R antagonist, enhances compound 2 efficacy (or vice versa) at the GLP-1R. Given that GLP-1 9–36 amide is the major circulating form of GLP-1 post-prandially and is a low affinity weak partial agonist or antagonist at the GLP-1R, we investigated interaction between this metabolite and compound 2 in a cell line with recombinant expression of the human GLP-1R and the rat insulinoma cell line, INS-1E, with native expression of the GLP-1R. We show compound 2 markedly enhances efficacy and potency of GLP-1 9–36 amide for key cellular responses including AMP generation, Ca2+ signaling and extracellular signal-regulated kinase. Thus, metabolites of peptide hormones including GLP-1 that are often considered inactive may provide a means of manipulating key aspects of receptor function and a novel therapeutic strategy.


Introduction
Glucagon-like peptide-1 (GLP-1) is released from intestinal Lcells in response to nutrient ingestion and is a key incretin hormone, not only potentiating glucose-dependent insulin release but contributing to glucose homeostasis by enhancing insulin biosynthesis, suppressing glucagon secretion, stimulating b-cell mass and suppressing appetite [1][2][3]. These effects are mediated by the GLP-1 receptor (GLP-1R), a Family B G-protein-coupled receptor that is coupled preferentially to Ga s but which may couple to other G-proteins [4][5][6][7]. GLP-1 is generated within intestinal L-cells by the action of prohormone convertase 1/3 on the proglucagon precursor molecule to generate GLP-1 7-37. At the time of synthesis, the C-terminal glycine of a proportion of GLP-1 7-37 is removed by a peptide amidating monooxygenase to generate GLP-1 7-36 amide [8]. GLP-1 7-37 and GLP-1 7-36 amide are essentially equipotent at the GLP-1R [9] and although present at approximately equivalent concentrations in plasma during fasting, the amidated version is the major circulating form post-prandially [10]. GLP-1 secretion may be impaired in type 2 diabetes and its insulinotropic and glucagon suppressing potency may be reduced [11]. However, exogenous GLP-1 potentiates glucose-dependent insulin secretion and can normalize hyperglycaemia in type 2 diabetes, while the impaired action of GLP-1 may be improved by good glycaemic control [11]. Thus strategies focusing on GLP-1 and its receptor have become targets for the treatment of this condition.
GLP-1 has a plasma half-life of a few minutes due to proteolytic degradation by the serine protease, dipeptidyl peptidase-IV (DPP-IV), which cleaves the N-terminal histidine and alanine residues from GLP-1 to generate GLP-1 9-37 and GLP-1 9-36 amide. Such proteolysis is thought to remove biological activity [12]. This degradation mitigates against the therapeutic use of GLP-1 itself and a range of DPP-IV-resistant peptide analogues have been developed and licensed for clinical use [11,13,14]. Alongside this, there has been a drive for the development of small-molecule, orally active agonists of the GLP-1R that would provide alternative and potentially improved treatment regimes. The Novo Nordisk compound, 6,7-dichloro-2-methylsulfonyl-3-N-tertbutylaminoquinoxaline or 'compound 2', is an ago-allosteric modulator of the GLP-1R, not only enhancing affinity of the GLP-1R for GLP-1 but providing effective direct agonism [15]. Evidence that compound 2 mediates its effects through binding to an allosteric site includes the inability of the orthosteric antagonist, exendin 9-36 to inhibit activity [15]. A number of studies have investigated interactions between compound 2 and established agonists of the GLP-1R [7,15,16]. In one such study using HEK-293 cells with stable expression of the human GLP-1R, we also surprisingly observed that compound 2 efficacy for cAMP generation (as assessed on the basis of E max ) was enhanced by exendin 9-36, or alternatively that compound 2 engendered agonist properties to this orthosteric ligand [7]. Given that DPP-IV cleavage of GLP-1 7-36 amide generates GLP-1 9-36 amide, which has been described as either a low affinity weak partial agonist or antagonist of the GLP-1R [17,18] and that GLP-1 9-36 amide can be present at five-to ten-fold higher concentrations than GLP-1 7-36 amide [10], the present study examined potential interactions between compound 2 and GLP-1 9-36 amide at the GLP-1R.

Cell culture
HEK-Flp-In cells with stable recombinant expression of the human GLP-1R (HEK-GLP-1R) were cultured in DMEM with high glucose supplemented with 10% FBS, 100 mg/ml streptomycin and 100 units/ml penicillin sulphate. These cells were originally generated by transfection of HEK-Flp-In cells (Invitrogen: Paisley, U.K.) with pcDNA5/FRT containing the human GLP-1R. They have been characterized previously and express the GLP-1R at ,1 pmol/mg total cellular protein with a K d for GLP-1 7-36 amide of ,1 nM [7]. INS-1E cells were kindly provided by Prof. C.B. Wollheim of the University of Geneva, Switzerland [19]. These cells were used between passages 65-90 and cultured in RPMI media containing 11.1 mM glucose, 5% heat-inactivated FBS, 100 mg/ml streptomycin, 100 units/ml penicillin sulphate, 100 units/ml neomycin, 50 mM b-mercaptoethanol, 10 mM HEPES and 1 mM sodium pyruvate. All cells were cultured at 37uC in a 5% CO 2 humidified atmosphere and passaged at confluence.  (10 min, 37uC) in 360 ml KHB-BSA containing 500 mM isobutylmethylxanthine (IBMX). Ligands or vehicle control (40 ml) were added and reactions terminated after the appropriate times by replacement of the aqueous phase with ice-cold 0.5 M trichloroacetic acid. INS-1E cells grown to confluence in 24-well plates were treated similarly with the exception that after washing in KHB, cells were incubated in 200 ml KHB-BSA containing 1.4 mM glucose without IBMX for 2 h followed by challenge with ligands (or vehicle) for 15 min at 37uC in 400 ml KHB-BSA, in the presence of 10 mM glucose and 500 mM IBMX. Reactions were terminated as above.
The cAMP was extracted from either intact cell or membrane preparations using a method identical to that for extraction of Ins(1,4,5)P 3 [22]. Levels of cAMP were then determined by a competitive radioreceptor assay using binding protein purified from bovine adrenal glands [23] and related to cellular protein assessed by Bradford assay.

Data analysis
Concentration-response curves were fitted using GraphPad Prism (GraphPad Software Inc., CA) using a standard four parameter logistic equation. All data are representative of n$3 or are presented as mean+/6s.e.m., where n = 3 unless otherwise stated. Statistical analysis was by oneway ANOVA and where P,0.05, followed by Bonferroni's or Dunnett's multiple range test as indicated. Alternatively analysis was performed by Student's ttest. Where potency estimates are given these are pEC 50 values (2log 10 of the molar concentration giving 50% of the maximal response).
Over 60 min, GLP-1 9-36 amide (1 mM) evoked a minor increase in cAMP ( Figure 2). Compound 2 (1 mM) evoked a more robust increase, which peaked at 30 min and then declined. At all of the time points studied, co-stimulation with GLP-1 9-36 amide and compound 2 evoked cAMP responses significantly greater than the numerical addition of responses to the two ligands alone ( Figure 2).
The potency of GLP-1 9-36 amide-mediated cAMP generation was increased by compound 2 in a concentration-dependent manner ( Figure 3AB, Table 1). Thus, the pEC 50 of GLP-1 9-36 amide alone was 6.5160.02 but this was progressively increased by increasing concentrations of compound 2 to 8.4160.22 at 3 mM. Subtraction of the response to compound 2 alone from that to co-addition with GLP-1 9-36 amide clearly showed the increased potency of GLP-1 9-36 amide by compound 2 and highlighted the increased E max values ( Figure 3B). Indeed, at all concentrations of compound 2 (0.03 mM-1 mM), cAMP responses to co-stimulation with compound 2 and the maximal concentration of GLP-1 9-36 amide (1 mM) were significantly greater than the numerical sum of the ligands alone ( Figure 3C).
HEK-293 cells express b 2 -adrenoceptors [26] that also couple to Ga s , adenylyl cyclase and the generation of cAMP. Despite stimulation of cAMP accumulation in HEK-GLP-1R cells by the  adrenoceptor agonist, isoproterenol (100 mM), co-stimulation with isoproterenol (100 mM) and compound 2 (1 mM) evoked cAMP accumulation that was only equivalent to the numerical addition of responses to the two ligands when used alone ( Figure 3D).
In membranes from HEK-GLP-1R cells, basal (unstimulated) levels of cAMP were relatively high (13696351 pmol/mg protein) ( Figure 4A) and GLP-1 7-36 amide stimulated cAMP generation (pEC 50 9.8460.11, E max 80576140 pmol/mg protein) ( Figure 4A). In contrast, GLP-1 9-36 amide evoked a minor elevation of cAMP (E max 2887683 pmol/mg protein, equivalent to 23% of the GLP-1 7-36 amide response) with low potency. Compound 2 alone (3 mM) elevated cAMP (26236269 pmol/mg protein) and enhanced the ability of GLP-1 9-36 amide to increase cAMP ( Figure 4A). Thus, across all concentrations tested, the increases in cAMP mediated by GLP-1 9-36 amide were greater in the presence compared to the absence of compound 2 (3 mM) ( Figure 4A). Furthermore, when tested at different concentrations, both 1 mM and 3 mM compound 2 in combination with 1 mM GLP-1 9-36 amide resulted in cAMP accumulation that was significantly greater than the numerical sum of the two added independently ( Figure 4B). Direct activation of adenylyl cyclase with forskolin robustly increased cAMP generation in these membranes ( Figure 4B).
In the pancreatic b-cell line INS-1E, GLP-1 7-36 amide evoked modest but potent (pEC 50 10.560.11) elevations of cAMP ( Figure 5). Compound 2 alone elevated cAMP with low-potency and to an E max of only 34% of that evoked by GLP-1 7-36 amide although concentrations of compound 2 greater than 10 mM were not tested. The cAMP response to GLP-1 9-36 amide was minor but markedly potentiated by compound 2 (3 mM) ( Figure 5). For   Figure 6A). This response has been characterized previously [7].
Addition of 10 mM GLP-1 9-36 amide evoked an increase in [Ca 2+ ] i , although less than that caused by GLP-1 7-36 amide ( Figure 6B). This response was abolished following depletion of the intracellular Ca 2+ stores pretreatment with thapsigargin (2 mM, 5 min) ( Figure 6B). Compound 2 also evoked Ca 2+ responses, which were detected against a background of fluorescence changes mediated by compound 2 itself. Thus, addition of compound 2 (100 mM) to either wild-type HEK-293 cells in the presence of fluo-4 loading ( Figure 6C) or to HEK-GLP-1R cells in the absence of fluo-4 loading (data not shown) resulted in an initial rapid increase in fluorescence followed by a more slowly developing  Figure 6D), thereby identifying a receptor-dependent increase in fluo-4 fluorescence that required a thapsigargin-sensitive intracellular Ca 2+ store. Addition of a low concentration of GLP-1 9-36 amide (1 mM) evoked little or no Ca 2+ response ( Figure 6E). Furthermore, addition of a lower concentration of compound 2 (10 mM) did not cause an appreciable, thapsigargin-sensitive increase in fluorescence ( Figure 6E). In contrast, the co-addition of both GLP-1 9-36 amide (1 mM) and compound 2 (10 mM) resulted in an increase in thapsigargin-sensitive fluorescence ( Figure 6E). This effect was not apparent in either fluo-4-loaded wild-type HEH-293 cells or HEK-GLP-1R cells in the absence of fluo-4 loading (data not shown). Ca 2+ responses to GLP-1R ligands (including GLP-1 7-36 amide) in populations of INS-1E cells were difficult to distinguish and could not be adequately assessed.
In HEK-GLP-1R cells, GLP-1 7-36 amide (10 nM) evoked a robust increase in ERK activation as assessed by the increase in phospho-ERK but this was not matched by the increase in response to GLP-1 9-36 amide when used up to 10 mM ( Figure 7A). Compound 2 also activated ERK with 10 mM evoking a response equivalent to the maximally effective concentration of GLP-1 7-36 amide (10 nM) ( Figure 7A,B and data not shown). When cells were stimulated with sub-maximal concentrations of both compound 2 (0.1 mM to 1 mM) and GLP-1 9-36 amide (1 mM), ERK activation was 45-69% greater than the sum of the responses to the ligands alone ( Figure 7A,B). Neither GLP-1 9-36 amide, GLP-1 7-36 amide nor compound 2 increased ERK activation in wild-type HEK293 cells (data not shown). In INS-1E cells, GLP-1 7-36 amide, GLP-1 9-36 amide and compound 2 evoked robust and approximately equivalent activation of ERK, although much lower concentrations of GLP-1 7-36 amide were required ( Figure 7C,D). There was clear evidence that costimulation of cells with both GLP-1 9-36 amide and compound 2 enhanced responses compared to that expected from a simple addition of responses to the ligands added individually. This was particularly apparent at 1 mM GLP-1 9-36 amide and 3 mM compound 2, where, despite minor responses to the agonists individually, together they evoked a response approaching that of GLP-1 7-36 amide ( Figure 7C,D). Furthermore, 1 mM GLP-1 9-36 amide and 10 mM compound 2 evoked a response in excess of the maximal response to GLP-1 7-36 amide (10 nM and 100 nM GLP-1 7-36 amide were equivalent; data not shown).

Discussion
Estimates indicate that more than 340 million people in the world currently suffer from diabetes mellitus with approximately 90% having type 2. Despite access to a variety of treatment regimes, diabetes is the leading cause of blindness, amputation and kidney failure while cardiovascular disease accounts for 50-80% of deaths amongst the diabetic population. This highlights the inadequacy of current therapeutics and the need for alternative and improved treatment regimes. The GLP-1/GLP-1R system is a validated target for treatment of type 2 diabetes and both DPP-IV inhibitors and GLP-1 peptide mimetics have emerged as alternative therapies. A number of potential problems associated with inhibition of DPP-IV, along with the inferior clinical efficacy of DPP-IV inhibitors compared to GLP-1R agonists as second line treatments to reduce HbA 1c and body weight [2,27,28], the relative metabolic instability of peptide ligands and the requirement for injection of peptides and associated issues with patient compliance have driven the search for small molecule, orally active GLP-1R agonists. A number of experimental compounds have emerged that act as agonists or indeed antagonists of the GLP-1R [29]. For compounds where information is available, binding is at one or more allosteric sites on the receptor. Given that GLP-1 makes multiple contacts with the GLP-1R including sites within the N-terminal domain, extracellular loops and trans-membrane domains [30][31][32], it is likely that exploiting such allosteric sites presents the best opportunity, at least for small molecule agonists or positive allosteric modulators. In addition, allosteric regulation of GPCRs has a number of potential therapeutic advantages including specificity amongst receptors with similar orthosteric binding sites. Furthermore, positive allosteric modulators may allow more physiological receptor regulation by influencing receptor activity only in the presence of the endogenous ligand. Here we demonstrate that not only can allosteric ligands influence the activity of the endogenous ligand but that it is possible to allosterically manipulate the action of peptide ligand metabolites, which are often considered inactive but which can be present at high concentrations, particularly at relevant cellular locations.
The GLP-1R couples primarily to Ga s and although coupling to Ga i/o and Ga q/11 has been reported [4][5][6], we have shown previously that in the HEK-GLP-1R cell line used here, both GLP-1 7-36 amide and compound 2 couple the GLP-1R to cellular signaling pathways through Ga s [7]. In addition to direct agonism at the GLP-1R, compound 2 is a positive allosteric modulator of GLP-1 [15]. However, although compound 2 modestly increases the affinity for GLP-1 it has little impact on potency or intrinsic activity [7,15]. Indeed, predictions suggest that even at high concentrations of compound 2 only a very minor shift in agonist potency would be expected [16]. In our earlier studies, compound 2 actually reduced GLP-1 7-36 amide potency although interpretation may be compromised by adverse effects of compound 2 [7]. Such lack of effect or an inhibitory effect of compound 2 on GLP-1R-mediated cAMP responses to GLP-1 7-36 amide is in stark contrast to the present study in which compound 2 markedly increases both the potency (pEC 50 ) and intrinsic activity (E max ) of GLP-1 9-36 amide. For example, in HEK-GLP-1R cells, the pEC 50 of GLP-1 9-36 amide-mediated cAMP generation was enhanced approximately 100-fold by 3 mM compound 2, along with marked increases in E max values ( Figure 3, Table 1). In INS-1E cells, only in the presence of compound 2 did GLP-1 9-36 amide evoke cAMP signaling with an E max approximately 50% of that evoked by the full agonist, GLP-1 7-36 amide. Although compound 2 can inhibit GLP-1 7-36 amidemediated internalization of the GLP-1R [7], this was not involved in this enhanced signaling as similar effects were seen in membrane preparations from HEK-GLP-1R cells (Figure 4). This, coupled with the inability of compound 2 to influence cAMP generation by either the b-adrenoceptor ( Figure 3D) or forskolin (to stimulate adenylyl cyclase directly; data not shown) demonstrate an effect of compound 2 on GLP-1 9-36 amide-mediated activation of the GLP-1R. A previous report has also suggested that the small molecule allosteric ligands of the GLP-1R, compound 2 and compound B, may increase the activity of GLP-1 9-36 amide (as assessed by a cAMP response elementluciferase reporter) [33], although it was unclear if this represented simple additivity or true potentiation.
The cAMP responses are clearly a critical component of GLP-1R-mediated events. However, at least in pancreatic b-cells, there is a complex network of subsequent signaling events that evoke Ca 2+ responses required for insulin release. This is largely dependent on Ca 2+ entry through voltage-operated Ca 2+ channels but there is also a role for Ca 2+ release from intracellular stores [2,3]. Although not a consistent finding [16], we and others have shown compound-2-mediated Ca 2+ signaling by the GLP-1R albeit with different kinetics to peptide agonists [7,33]. In our HEK-GLP-1R cells this is a consequence of release from intracellular stores [7]. Here we demonstrate that GLP-1 9-36 amide elevates [Ca 2+ ] i by release from an intracellular store although with low potency (requiring.1 mM) ( Figure 6B). Compound 2 also evoked Ca 2+ responses. Thus, only in fluo-4-loaded HEK-GLP-1R cells did high concentrations (100 mM) generate a thapsigargin-sensitive increase in fluorescence, thereby identifying a receptor-mediated increase in [Ca 2+ ] i that was dependent upon a replete intracellular Ca 2+ store. Importantly, concentrations of GLP-1 9-36 amide (1 mM) and compound 2 (10 mM) that evoked little or no increase in [Ca 2+ ] i alone (thapsigargin-sensitive increase in fluorescence) produced a marked increase when added in combination ( Figure 6E) indicating that compound 2 potentiates Ca 2+ signaling by GLP-1 9-36 amide.
In addition to elevation of cAMP and [Ca 2+ ] i , the GLP-1R couples to ERK activation although mechanisms may be cell-type dependent. The precise role of ERK in GLP-1R-mediated signaling is unclear but evidence suggests a critical role, particularly in pancreatic b-cells and b-cell precursors for proliferation and differentiation [34,35]. Such activity may well underlie aspects of the non-insulinotropic anti-diabetic effects of GLP-1R activation that potentially enhance the utility of this system in the treatment of type 2 diabetes. Here we confirm ERK activation by GLP-1 7-36 amide and demonstrate that both GLP-1 9-36 amide and compound 2 activate ERK in HEK-GLP-1R and INS-1E cells. Further, the data highlight that compound 2 has the ability to potentiate GLP-1 9-36 amide-mediated responses. This is in contrast to the lack of interaction between compound 2 and GLP-1 7-36 amide on ERK activation [16].
Fasting plasma concentrations of both GLP-1 7-37 and GLP-1 7-36 amide are ,10 pM. Following a meal, levels increase to ,10 pM and ,40 pM respectively, highlighting a more pronounced effect on the concentration of the amidated version, which may reflect higher levels in the secretory tissues [10]. Clearance of GLP-1 9-36 amide is slower than metabolism of GLP-1 7-36 amide by DPP-IV with the result that GLP-1 9-36 amide is the major circulating form of GLP-1 in the fed state [36][37][38]. This is supported by studies using oral glucose tolerance tests in which peak levels of ,60 pM GLP-1 9-36 amide were observed with very little increase above fasted levels of intact GLP-1 [39]. The GLP-1R has nanomolar to sub-nanomolar affinity for intact GLP-1 [16,18,30,40], whereas affinity for GLP-1 9-36 is approximately one hundred fold less [18], thereby supporting the notion that GLP-1 9-36 amide is an inert cleavage product. However, a number of studies have raised the possibility that GLP-1 9-36 amide is a weak partial agonist or an antagonist of the GLP-1R [17,18]. Interestingly, GLP-1 9-36 amide also activates Akt, eNOS and promotes proliferation in human coronary artery endothelial cells to a similar extent as GLP-1 7-36 amide, even at 100 nM [41]. Furthermore, studies have suggested that GLP-1 9-36 amide exerts effects on hepatocytes and the cardiovascular system independently of the known GLP-1R [8]. Here we show that, at least for cAMP generation, GLP-1 9-36 amide is a low potency, weak partial agonist of the GLP-1R in a system with high receptor expression (HEK-GLP-1R cells) but provides little or no agonism in a system with lower receptor expression (INS-1E cells). These properties provide the potential for GLP-1 9-36 amide to behave as an agonist at low concentrations of intact GLP-1 but as Enhancing GLP-1 Metabolite Activity at the GLP-1R PLOS ONE | www.plosone.org an antagonist at higher concentrations when GLP-1R occupancy by GLP-1 9-36 amide would provide low efficacy signaling and inhibit GLP-1 7-36 amide binding. This is consistent with studies in other cell lines [42]. Such dependency on both receptor expression levels and the concentration of competing ligand (intact GLP-1), coupled with the possibility of an additional receptor, highlight the difficulties in assessing physiological roles of GLP-1 9-36 amide and may account for some of the discrepancies in the literature.
The inability of compound 2 to influence GLP-1 7-36 amide potency [7,15] suggest that such compounds work through direct agonism in vivo. However, allosteric modulation of receptors toward other endogenous peptides, including metabolites of GLP-1, could be responsible and may be exploited therapeutically. Here we show that compound 2 markedly enhances agonism by GLP-1 9-36 amide, which is the main circulating form of GLP-1. Although the GLP-1R has low affinity for GLP-1 9-36 amide, compound 2 enhances its potency for cAMP generation by ,100fold in HEK-GLP-1R cells. The concentration of GLP-1 9-36 amide at relevant sites for GLP-1R function is unknown and may be considerably greater than circulating concentrations, particularly if the metabolite is generated locally by DPP-IV activity. Interestingly, patients with diabetes and associated chronic renal insufficiency are less able to clear incretin hormone metabolites including GLP-1 9-36 amide [39], which may also provide an area for therapeutic exploitation.
Irrespective of whether enhanced agonism by GLP-1 9-36 amide would contribute to any in vivo activity of compounds with activity similar to compound 2, these studies illustrate the significant principle that manipulation of the activity of endogenous metabolites could provide a therapeutic opportunity and should be considered in drug-screening strategies. This is true for not only GLP-1 but potentially for other peptide ligands where their activity is effectively reduced or terminated by metabolism to compounds considered to have little or no biological activity. The concept of probe-dependence [43] in which the outcome of allosteric modulation is determined by the nature of the orthosteric ligand may provide an additional area of exploitation if the signaling outcomes of allosteric modulation of metabolites differs from that of allosteric modulation of the primary, endogenous ligand as suggested by these studies.