Structural basis for GLP-1 receptor activation by LY3502970, an orally active nonpeptide agonist

Abstract

Glucagon-like peptide-1 receptor (GLP-1R) agonists are efficacious antidiabetic medications that work by enhancing glucose-dependent insulin secretion and improving energy balance. Currently approved GLP-1R agonists are peptide based, and it has proven difficult to obtain small-molecule activators possessing optimal pharmaceutical properties. We report the discovery and mechanism of action of LY3502970 (OWL833), a nonpeptide GLP-1R agonist. LY3502970 is a partial agonist, biased toward G protein activation over β-arrestin recruitment at the GLP-1R. The molecule is highly potent and selective against other class B G protein-coupled receptors (GPCRs) with a pharmacokinetic profile favorable for oral administration. A high-resolution structure of LY3502970 in complex with active-state GLP-1R revealed a unique binding pocket in the upper helical bundle where the compound is bound by the extracellular domain (ECD), extracellular loop 2, and transmembrane helices 1, 2, 3, and 7. This mechanism creates a distinct receptor conformation that may explain the partial agonism and biased signaling of the compound. Further, interaction between LY3502970 and the primate-specific Trp33 of the ECD informs species selective activity for the molecule. In efficacy studies, oral administration of LY3502970 resulted in glucose lowering in humanized GLP-1R transgenic mice and insulinotropic and hypophagic effects in nonhuman primates, demonstrating an effect size in both models comparable to injectable exenatide. Together, this work determined the molecular basis for the activity of an oral agent being developed for the treatment of type 2 diabetes mellitus, offering insights into the activation of class B GPCRs by nonpeptide ligands.

Keywords: LY3502970; OWL833; cryoelectron microscopy; glucagon-like peptide-1 receptor; type 2 diabetes mellitus.

Fig. 1.

Pharmacology and structural analysis of LY3502970 in complex with GLP-1R/G_siN18/Nb35/scFv16. (A) Chemical structure of LY3502970 (molecular weight: 882.96, formula: C₄₈H₄₈F₂N₁₀O₅). Moieties that are discussed in the cryo-EM structural analysis are highlighted in colored boxes. (B and C) The signal transduction pharmacology of LY3502970 and GLP-1(7-36) was determined. Human GLP-1R density-dependent pharmacology of ligands was quantified by measuring the potency and efficacy for cAMP accumulation at increasing levels of receptor density (high, medium, low). Functional potency and efficacy for β-arrestin recruitment using enzyme fragment complementation was determined. Representative concentration response curves are presented. Summarized data with statistics are presented in SI Appendix, Table S1. (D) LY3502970 does not stimulate cAMP accumulation in HEK293 cells expressing the mouse GLP-1R. Data are presented as the mean ± SD of three independent experiments. (E) Mice expressing the human GLP-1R (n = 5 mice/group) were fasted overnight and orally administered vehicle or LY3502970 (0.1 to 10 mg/kg). Five hours later, animals received an intraperitoneal (i.p.) injection of glucose (2 g/kg). As a control group, one cohort was dosed with a s.c. injection of exenatide (1 nmol/kg) 1 h prior to receiving the i.p. glucose. For all mice, the circulating concentration of glucose over various time points was measured using glucometers. Each dose of LY3502970 reduced the glucose excursion AUC versus vehicle (P < 0.05; one-way ANOVA followed by the Dunnett’s test). (F) Similar studies were perfumed in Glp1r null mice. In these experiments, vehicle or LY3502970 (10 mg/kg) was administered orally and exenatide (1 nmol/kg) or gastric inhibitory polypeptide, (GIP) (30 nmol/kg) was dosed via s.c. injection. Data are mean ± SEM (n = 4 to 5 mice/group). (G) Overall structure of GLP-1R/LY3502970/G_siN18/Nb35/scFv16. Each subunit or ligand is shown with a different color (GLP-1R: green for 7TM and blue green for ECD; LY3502970: magenta; G_sαiN18: bright blue; Gβ: light orange; Gγ: green cyan; scFv16: salmon; Nb35: violet). (H) Surface representation of LY3502970 binding pocket. The 7TM domain of GLP-1R is colored in green, while ECD is colored blue green. LY3502970 is shown by spheres. (I) LY3502970 interacts with residues from both the ECD and 7TM of GLP-1R. LY3502970, and its interaction residues are shown by sticks. Hydrogen bonds are indicated by dashed lines.

Fig. 2.

Trp33^ECD mediates key interaction with LY3502970 and the surrounding region. (A) Amino acid sequence comparison of the ECD regions for GLP-1Rs of different species. (B) LY3502970 increases cAMP accumulation in HEK293 cells expressing a chimeric form of the mouse GLP-1R (domains shown in green) where the N-terminal 142 residues have been replaced by the corresponding region of the human GLP-1R (domains shown in blue). (C) Mutation of Ser33^ECD to Trp33^ECD in the mouse GLP-1R enables LY3502970 to stimulate receptor-induced cAMP accumulation, while the reciprocal mutation (D) in the human GLP-1R abolishes compound function. The chimera and mutant data are presented as the mean ± SD of three independent experiments. (E) Trp33^ECD interacts with LY3502970 and residues on TM2, ECL1, and ECL2. Residues and ligands are shown by sticks except Trp33^ECD, which is also shown by spheres. Hydrogen bonds are indicated by dashed lines. (F) GLP-1R ECD orientation in the LY3502970 bound structure stabilized by aromatic interactions with ECL1. (G) The unique ECD orientation of ECD in the LY3502970 bound structure as a result of the unique Trp33^ECD position. The structures of GLP-1R bound to LY3502970 (7TM in green, ECD in blue green, LY3502970 in magenta), GLP-1 (yellow, PDB ID code 6VCB), and peptide 5 (gray, PDB ID code 5NX2) are aligned. Trp33^ECD is shown by a stick. Other peptide bound GLP-1R structures (GLP-1, PDB ID code 5VAI; ExP5, PDB ID code 6B3J) are not shown here, but their ECD orientations are very similar to GLP-1 bound structure 6VCB in yellow. (H and I) Competitive inhibition of [¹²⁵I]GLP-1(7-36) binding to membranes isolated from cells expressing the human GLP-1R (H) or human W33S GLP-1R (I). Data are represented as the mean ± SD of three independent experiments.

Fig. 3.

LY3502970 binding and the conformation of TM1, TM2, and TM7 of the GLP-1R. (A) LY3502970 (magenta) interactions with TM1, TM2, and TM7 of GLP-1R (green). Residues and ligands are shown by sticks. (B) TM1, TM7, and ECL3 of the inactive state structure of GLP-1R (salmon, PDB ID code 6LN2) need to shift to interact with LY3502970 (magenta sphere). (C) Unique conformation of TM1 and TM2 to accommodate LY3502970. The structures of GLP-1R bound to LY3502970, GLP-1 (yellow, PDB ID code 6VCB), and TT-OAD2 (cyan, PDB ID code 6ORV) are aligned. Other peptide bound GLP-1R structures (GLP-1, PDB ID code 5VAI; ExP5, PDB ID code 6B3J; peptide 5, PDB ID code 5NX2) are not shown here, but their TM1, TM2, and TM3 conformations are very similar to GLP-1 bound structure 6VCB in yellow.

Fig. 4.

Structures of GLP-1R bound to agonists with different signaling profiles shed light on the structural basis for partial agonism and biased signaling. (A) Structures of the GLP-1R bound to LY3502970 (7TM in green, LY3502970 shown by magenta spheres), native GLP-1 (receptor in yellow, GLP-1 in beige, PDB ID code 6VCB), ExP5 (receptor in orange, ExP5 in pink, PDB ID code 6B3J), or TT-OAD2 (receptor in cyan, TT-OAD2 in blue, PDB ID code 6ORV) are aligned and shown from an identical view. Critical residues are shown by sticks. Hydrogen bonds are indicated by dashed lines. (B) The structure of GLP-1R bound to LY3502970 is aligned with GLP-1R bound to ExP5 and TT-OAD2 (Upper) and GLP-1 (Lower).

Fig. 5.

LY3502970 enhances glucose-stimulated insulin secretion and reduces food consumption in cynomolgus monkeys. (B–E) Effects of LY3502970 in IVGTT experiments. (A) Schematic diagram of the IVGTT procedure in anesthetized cynomolgus monkeys. Dosing of LY3502970 or exenatide was performed by manual bolus injection, followed by continuous infusion for 80 min. Forty minutes after initiating dosing of vehicle or the drug, glucose was administered intravenously. (B and C) Effect of LY3502970 (B) or exenatide (C) on blood glucose levels. (D and E) Effect of LY3502970 (D) or exenatide (E) on serum insulin levels. Data are represented as the mean ± SEM (n = 7). Significantly different from vehicle 1 or vehicle 2 by Dunnett’s test at *P < 0.025 and **P < 0.005 (one tailed), respectively. Vehicle 1: PEG400/PG/Gly NaOH buffer (pH 9), vehicle 2: 0.05% Tween80 PBS. (F and G) Mean food consumption following administration of LY3502970 (F) or exenatide (G). Cynomolgus monkeys were administered LY3502970, exenatide, or vehicle once daily for 5 d with a 2 d recovery period using an 8 × 5 cross-over design. Food consumption during the 90 min period after feeding was measured. Data are represented as the mean ± SEM (n = 8). Significantly different from the vehicle group by Dunnett’s test at **P < 0.005 (one tailed). Vehicle: dimethyl sulfoxide (DMSO)/Cremophor/PEG400/Gly-NaOH buffer (pH 10) [p.o.] + 0.05% Tween80 PBS [s.c.], LY3502970: LY3502970 [p.o.] + 0.05% Tween 80 PBS [s.c.], exenatide: DMSO/Cremophor/PEG400/Gly-NaOH buffer (pH 10) [p.o.] + exenatide [s.c.].