Building the Glucagon-Like Peptide-1 Receptor Brick by Brick: Revisiting a 1993 Diabetes Classic by Thorens et al.

Following food intake, transit of bile acids and nutrients through the small intestine stimulates release of glucagon-like peptide-1 (GLP-1) from sparsely distributed enteroendocrine L cells (1). GLP-1 rapidly pervades the neighboring pancreas, where it binds to the glucagon-like peptide-1 receptor (GLP-1R), a seven-transmembrane receptor belonging to the secretin/class B G protein–coupled receptor (GPCR) family, which is highly expressed in pancreatic β-cells (2). In its nonstimulated state, GLP-1R undergoes diffusion at the cell surface (3), constrained into specific zones by the cell cytoskeleton (4). Following ligand stimulation, GLP-1R rapidly (milliseconds) arrests at the membrane (5), during which time it engages the G_s subunit to activate adenylate cyclase and generate cAMP, a potent second messenger (6). In pancreatic β-cells, cAMP binds protein kinase A (PKA) and exchange factor directly activated by cAMP to amplify Ca²⁺-dependent and glucose-dependent exocytosis of insulin granules (7,8), as well as activate prosurvival cell pathways (with less well-characterized parallel inputs from mitogen-activated protein kinase-extracellular signal–regulated kinase, phospholipase C/inositol triphosphate/diacylglycerol, etc. [9]). Continued ligand stimulation leads to GLP-1R phosphorylation by GPCR kinases, recruitment of β-arrestin proteins, clathrin-mediated internalization, and endocytic sorting, either back to the membrane for reactivation or to the lysosome for degradation (10–13). Some GLP-1R continues to signal inside the cell through endosomal cAMP production (11,14), while degraded GLP-1R is replaced by nascent GLP-1R, which traffics from the Golgi apparatus to the cell membrane (10). Given the knowledge about signaling obtained from experiments in cell lines, preclinical models, and even human, GLP-1R is now regarded as an exemplar class B GPCR.

Since GLP-1 is rapidly cleaved and inactivated by dipeptidyl peptidase 4, it was clear from an early stage that stabilized GLP-1R agonists needed to be developed if GLP-1R was to become a therapeutic target. Through modifications such as amino acid substitutions, Fc conjugation, and acylation, degradation-resistant GLP-1R agonists were produced with long circulating half-lives. GLP-1R agonists were thus cemented as a major type 2 diabetes treatment in 2005 due to profound insulin secretory and glucose-lowering effects (historical narrative provided by Campbell et al. [15]). However, preclinical studies were beginning to piece together the central effects of GLP-1R agonists on food intake as early as the 1990s (16,–,18), a decade prior to the first GLP-1R agonist approval for type 2 diabetes. Following randomized placebo-controlled trials, the GLP-1R agonist drug class became a nonsurgical treatment for obesity when liraglutide was approved for use in adults in 2014 (19), with semaglutide following in 2017 and 2018 (20).

In this Classics in Diabetes article, we revisit a 1993 Diabetes article by Thorens et al. (21), in which the cloning of the human GLP-1R was described. The article also showed that exendin4(1-39) (termed exendin4 here) is a full agonist at the human GLP-1R, whereas its N-terminally truncated form, exendin4(9-39) (termed exendin9 here), is a full antagonist. Exendin4 subsequently became the first GLP-1R agonist used in the therapy of type 2 diabetes under the name Byetta.

Following the discovery of GLP-1R agonists as potential type 2 diabetes therapies, the race was on to understand the structure of GLP-1R to allow functional studies as well as drug development. Bernard Thorens had single-handedly cloned the rat GLP-1R the year before (22), providing first insights into structure-function. However, homology was not guaranteed between species, and basic science and drug development programs needed details of the human GLP-1R to speed up discovery and patient benefit. Unlike today, where DNA can be sequenced with high fidelity in a matter of days, 30 years ago cDNA libraries needed to be meticulously constructed and screened to isolate single clones, a time-consuming task. In the case of human GLP-1R, Mike Mueckler’s laboratory at Washington University School of Medicine in St. Louis used a lambda gt11 expression vector to construct a human islet cDNA library, which was generously gifted for cloning efforts. The library was screened using rat GLP-1R cDNA (21), the sequence of which was known courtesy of work the year before (22), leading to isolation of a single clone, hGLP1R-20, with ∼83% homology to rat GLP-1R. A second human islet cDNA library was then constructed using a γ ZAP II expression vector and forward screened with hGLP1R-20, revealing a clone with cDNA encoding the full-length GLP-1R. The isolated human GLP-1R cDNA was ∼96% homologous to the rat GLP-1R and corresponded to an ∼2.6 kb transcript on Northern blot that translated to a 463 amino acid protein. Following subcloning and overexpression in Chinese hamster lung (CHL) immortalized cells, human GLP-1R bound radiolabeled GLP-1 with high affinity (half-maximal inhibitory concentration 0.5 nM). CHL cells overexpressing human GLP-1R displayed cAMP increases in response to GLP-1 (half-maximal effective concentration 93 pM) and exendin4 (33 pM), values that still resonate with molecular pharmacologists today. Of note, the N-terminally truncated exendin9 was found to act as a full antagonist at the human GLP-1R. Together, these data provided the amino acid sequence and size of the human GLP-1R, showed its responses to GLP-1 and stabilized agonist, and demonstrated that most of the binding affinity of GLP-1 and other GLP-1R agonists is dictated by their C-terminal regions (21) (Fig. 1).

Around the same time, the laboratory of Aubrey E. Boyd III at Tufts University School of Medicine, and the Merck Group at Merck Research Laboratories in New Jersey separately published details of the human GLP-1R sequence using human or rat cDNA libraries. Many of the key details were cross-confirmed between the three competing studies, including 463 amino acid size, ∼2.7 kb transcript size, and 90–91% homology with the rat GLP-1R (23,24).

Given the complexity of cloning in the era before next-generation sequencing, it is remarkable that three independent laboratories published side-by-side sequences, a testament to the rigor of science at the time.

Three decades have passed since the human GLP-1R was cloned, and the impact on the diabetes and metabolism field is wide-ranging. Below are specific examples of work directly informed by knowledge gained from studies reporting the human GLP-1R sequence as well as functional expression (summarized in Fig. 1).

Before the rat and human GLP-1R were cloned, most pharmacology studies relied on tissue expressing high levels of endogenous GLP-1R. While accurate and important, use of primary tissue or slow-growing cell lines was not conducive to the high-throughput pharmacology needed for drug target identification and validation. Thus, insertion of the human GLP-1R into an expression vector allowed overexpression in rapidly growing and robust mammalian cell lines, such as CHL, CHO, and HEK293, which otherwise do not express the receptor. In this way, GLP-1R (ant)agonist efficacy could be (more) rapidly determined using standard pharmacology assays, including cAMP, Ca²⁺, and β-arrestin screens versus GLP-1 or exendin4. From these or similar screens, allosteric modulators (25,26) and small-molecule agonists (27) were also developed with activity against human GLP-1R. In addition, the concept of biased agonism was established, where some ligands have preference over certain signaling pathways, endowing them with advantageous properties such as GLP-1R surface expression and more persistent activity in the case of β-arrestin bias (28). Lastly, human GLP-1R–based screens showed that the superior efficacy of the dual agonist tirzepatide is largely because of its unique signaling bias with reduced β-arrestin recruitment and imbalance toward gastric inhibitory peptide (GIP) receptor over GLP-1R activation (29).

From mutational screens (and cryoelectron microscopy [cryo-EM]; see below), it was predicted that alterations to the GLP-1R N terminus would be well tolerated, whereas those to the transmembrane or C-terminal regions would be more likely to influence signaling and trafficking. From this, the human GLP-1R was modified with fluorophores (GFP), epitope tags (His and Myc), luminescent tags (NanoLuc), and enzyme self-labels, allowing prestimulation and poststimulation receptor fate to be precisely determined. Using these approaches, it was found that GLP-1R displays little constitutive activity but undergoes rapid and extensive internalization in response to exendin4 (10–13). After internalization, GLP-1R signals in the endosomal compartment, is recycled to the membrane for repeat stimulation, or is degraded in lysosomes (10–14). Notably, ligands biased away from β-arrestin reduce internalization and facilitate surface retention of GLP-1R, which prevents desensitization and improves signaling and insulin secretory responses over the long term (28).

Single nucleotide polymorphisms in human GLP-1R are associated with loss of function, including decreased cAMP, Ca²⁺, and ERK1/2 signaling (30). Since most GLP-1R single nucleotide polymorphisms are in linkage disequilibrium with other genes, it has been more difficult to causally associate GLP-1R variants with metabolic disease traits in humans. A recent large-scale functional genetic study identified ∼60 rare GLP-1R variants whose effects range from changes in cell surface expression to either complete or pathway-specific loss or gain in function (31). Notably, GLP-1R variants for which exome sequencing is available are associated with increased HbA_1c, BMI, and diastolic blood pressure, an effect further strengthened when the analysis excluded loss-of-function variants for β-arrestin 2 (i.e., those that limit GLP-1R internalization/trafficking) (31). Large-scale genome-wide association studies also showed that common and rare GLP-1R variants are associated with random glucose in humans, with strength of G_s coupling being predictive for variant effect-size (32). Thus, reduced GLP-1R cell surface expression or GLP-1R pathway coupling are risk factors for metabolic disease. In all these studies, variant human GLP-1R constructs were used to establish the links between genetics and signaling.

Crystallization and X-ray diffraction studies provided early insights into the N-terminal human GLP-1R structure, including the presence of an α-helix, conserved residues, and a ligand-binding domain (33). However, the advent of cryo-EM allowed the full ligand-bound human GLP-1R structure to be resolved, including interaction with its G protein. For cryo-EM, the human GLP-1R signal peptide is removed and replaced with a multiple His or Myc tag for overexpression, before transfection into insect or mammalian cells, lysis, and complex formation with G_s, ligand, and/or stabilizing nanobody. The obtained cryo-EM structures have provided much-needed insight into the key interactions and confirmations required for binding and activation as well as key differences between (biased) agonists, allowing rational drug design (34–36). For example, it is now known that peptide agonists interact with the human GLP-1R ectodomain via their C terminus, allowing the ligand N terminus to engage the transmembrane domain, which is stabilized by extracellular loops (34), as predicted from the original cloning and functional expression studies (21). Following ligand binding, the human GLP-1R transmembrane domain undergoes a conformational change, resulting in transmembrane helix 6 flexing toward the α5 helix responsible for G protein activation (35,36). Biased agonists promote faster G_s conformational changes, faster G_s heterotrimer dissociation, and faster cAMP generation than native GLP-1 (35,36). Structurally, these differences in G_s turnover are correlated with more transient interactions between biased agonist and conserved polar residues within the GLP-1R binding pocket (35,36). Providing insight into the mechanisms underlying dual agonism, tirzepatide- and GLP-1–bound human GLP-1R are almost undistinguishable, whereas tirzepatide- and GIP-bound GIP receptors differ in the conformation of extracellular loop 1 (37).

Endogenous GPCRs, including GLP-1R, are notoriously difficult to localize, since detergents required to stabilize protein folding can mask epitopes, limiting antibody production and immunohistochemical visualization (38). Based on the observation that N-terminally truncated exendin9 is a potent antagonist, exendin4 and exendin9 have been functionalized with various moieties, including fluorophore, positron emission tomography/MRI molecules, and oligonucleotides, without major loss of (ant)agonist activity (reviewed in Ast et al. [39]). The ensuing probes have been widely used to delineate GLP-1R binding sites in the brain and pancreas. Using fluorescent ligands, peripherally administered exendin4 and semaglutide were shown to readily access the hypothalamus and brainstem (40,–,42). Along similar lines, the same probes showed that GLP-1R is abundant in pancreatic β-cells but poorly expressed in α-cells (42). By taking advantage of GLP-1R as a largely β-cell–specific molecular address as well as the fact that GLP-1R undergoes ligand-dependent internalization, cell-impermeable antisense oligonucleotides have been delivered as gene therapy into β-cells using exendin4 as a cargo carrier (43). Lastly, positron emission tomography/MRI probes conjugated to exendin4 showed the potential to noninvasively quantify β-cell mass in preclinical models (44), including stem cell–derived β-like cell survival posttransplantation (45).

The past two decades have seen the advent of major new GLP-1R agonist–based therapies for type 2 diabetes and obesity, which have benefited hundreds of millions of patients. The success of GLP-1R agonists as a therapy is largely because of multiple strands of rigorous and replicable discovery, preclinical science, and clinical science, which have all informed the drug discovery process and vice versa. Undoubtedly, cloning, sequencing, and functional expression of the human GLP-1R was vital to allow investigators to understand receptor signaling, agonist efficacy and affinity, structure-function, and localization. Going forward, the knowledge gained will help to open up GLP-1R agonists as a potential new drug class for the treatment of metabolic dysfunction-associated steatotic liver disease, neurodegenerative disease, addiction, and inflammation.

Acknowledgments. The authors thank Dr. Jonathan E. Campbell (Duke University) for providing feedback on the manuscript.

Funding. D.J.H. was supported by the Medical Research Council (MR/S025618/1), Diabetes UK (17/0005681 and 22/0006389), and UKRI European Research Council (ERC) Frontier Research Guarantee (EP/X026833/1) grants. This work was supported on behalf of the “Steve Morgan Foundation Type 1 Diabetes Grand Challenge” by Diabetes UK and the Steve Morgan Foundation (grant number 23/0006627). This project has received funding from the ERC under the European Union’s Horizon 2020 research and innovation program (starting grant 715884 to D.J.H.). The research was funded by the National Institute for Health and Care Research (NIHR) Oxford Biomedical Research Centre.

The views expressed are those of the author(s) and not necessarily those of the National Health Service, the NIHR, or the Department of Health.

Duality of Interest. D.J.H. receives licensing revenue from Celtarys Research for provision of incretin-based chemical probes. D.J.H. has filed patents related to incretin-based chemical probes as well as GLP-1R agonism. B.T. receives consulting fees from SUN Pharma. No other potential conflicts of interest relevant to this article were reported.