A Small-Molecule Oral Agonist of the Human Glucagon-like Peptide-1 Receptor

Abstract

Peptide agonists of the glucagon-like peptide-1 receptor (GLP-1R) have revolutionized diabetes therapy, but their use has been limited because they require injection. Herein, we describe the discovery of the orally bioavailable, small-molecule, GLP-1R agonist PF-06882961 (danuglipron). A sensitized high-throughput screen was used to identify 5-fluoropyrimidine-based GLP-1R agonists that were optimized to promote endogenous GLP-1R signaling with nanomolar potency. Incorporation of a carboxylic acid moiety provided considerable GLP-1R potency gains with improved off-target pharmacology and reduced metabolic clearance, ultimately resulting in the identification of danuglipron. Danuglipron increased insulin levels in primates but not rodents, which was explained by receptor mutagensis studies and a cryogenic electron microscope structure that revealed a binding pocket requiring a primate-specific tryptophan 33 residue. Oral administration of danuglipron to healthy humans produced dose-proportional increases in systemic exposure (NCT03309241). This opens an opportunity for oral small-molecule therapies that target the well-validated GLP-1R for metabolic health.

Figure 1

Identification of small-molecule GLP-1R agonists in a CHO-GLP-1R cellular assay in the absence or presence of the positive allosteric modulator BETP. (A) Assay concept. Covalent modification of Cys347 in the GLP-1R by BETP lowers the receptor activation barrier, enabling the identification of weak agonists. (B–D) Validation of the BETP-sensitized screening assay. (B) BETP potentiates agonist-induced cAMP production of a small molecule (Boc-5) (Figure 2). (C and D) BETP potentiates cAMP production and β-arrestin recruitment, respectively, by peptide 1 (Figure 2) at the human GLP-1R. (E) cAMP curves of a representative small-molecule HTS hit, compound 2 (Figure 2). cAMP data normalized using 100 nM (+BETP) or 1 μM (+DMSO) GLP-1 response (100%), β-arrestin recruitment data normalized using 1 μM GLP-1 (100%). Data represent the mean ± the standard error of the mean (SEM) from two (B, DMSO), three (C and D, BETP), four (C and D, DMSO), six (B, BETP), seven (E, DMSO), or eight (E, BETP) experiments, each performed in duplicate; error bars are within the displayed symbols. Abbreviations: BETP, 4-[3-(benzyloxy)phenyl]-2-ethylsulfinyl-6-(trifluoromethyl)pyrimidine; cAMP, cyclic adenosine monophosphate; CHO, Chinese hamster ovary; DMSO, dimethyl sulfoxide; GLP-1R, glucagon-like peptide-1 receptor; PAM, positive allosteric modulator.

Figure 2

Structures of reference compounds. (A) Structure of Boc-5. (B) Structure of peptide 1.

Figure 3

Optimization of small molecule 2 culminating in the identification of the clinical candidate danuglipron. (A) Structure of key compounds in the progression of small-molecule HTS hit 2 to clinical candidate danuglipron. Four structural regions were considered in our efforts to improve the GLP-1R agonist activity of 2: the piperidine, the benzyl ether, the 5-fluoro-pyrimidine, and the benzimidazole. (B) Early evidence of activity in the cAMP assay without BETP. An increased level of cAMP acumulation (percent effect relative to 1 μM GLP-1) at a test compound concentration of 20 μM was observed in the absence of BETP as potency (EC₅₀) improved in the presence of BETP. Data from 1882 analogues. (C) Small-molecule agonist activity independent of BETP sensitization. Increased cAMP potency (EC₅₀) was observed in the absence of BETP as potency improved in the presence of BETP (red circles, non-acids; green circles, acid-containing analogues). Dashed line at a 100:1 EC₅₀ ratio (−BETP/+BETP), solid line at a 1:1 ratio. Data from 1884 analogues. (D and E) Candidate selection CHO-GLP-1R cell line with a lower GLP-1R expression level confirms the efficacy-driven nature of small-molecule 5 agonism at the human GLP-1R. Data represent the mean ± SEM. (D) Saturation binding analysis in CHO cells expressing higher (purple squares, SA) and lower (blue circles, CS) human GLP-1R density. Data represent the mean ± SEM from two individual experiments performed in duplicate. (E) Small molecule 5 induced cAMP signaling in the CS cell line (blue circles), as well as in the SA in the presence (red triangles) or absence (purple squares) of BETP. Data represent the mean ± SEM from 81 (SA + BETP), 105 (SA + DMSO), or 87 (CS) independent experiments, each performed in duplicate; error bars are within the displayed symbols. cAMP data normalized using 100 nM (+BETP) or 1 μM (+DMSO) GLP-1 response (100%). Abbreviations: BETP, 4-[3-(benzyloxy)phenyl]-2-ethylsulfinyl-6-(trifluoromethyl)pyrimidine; cAMP, cyclic adenosine monophosphate; CHO, Chinese hamster ovary; CS, candidate selection; DMSO, dimethyl sulfoxide; EC₅₀, half-maximal effective concentration; GLP-1, glucagon-like peptide-1; GLP-1R, GLP-1 receptor; HTS, high-throughput screening; PAM, positive allosteric modulator; SA, screening assay; SEM, standard error of the mean.

Figure 4

Molecular pharmacology of small-molecule GLP-1R agonist danuglipron. (A) cAMP concentration–response curves for exenatide (black inverted triangle), liraglutide (green squares), and danuglipron (blue circles) in the candidate selection cell line. Data represent the mean ± SEM from 14, 12, and 22 individual experiments, respectively, each performed in duplicate. (B) β-Arrestin recruitment concentration–response curves for exenatide (black inverted triangle), liraglutide (green squares), and danuglipron (blue circles). Data represent the mean ± SEM from four, four, and three individual experiments, respectively, each performed in duplicate. (C) GLP-1R agonist-driven receptor internalization assessed using the FAP-tagged human GLP-1R stably expressed in HEK 293 cells. Data represent the mean ± SEM from three independent experiments, each performed in triplicate. (D) Assessment of danuglipron-induced internalization and recycling of a GFP-tagged human GLP-1R (green) in a HEK 293 cell construct (blue nuclear staining) using confocal microscopy (representative images). (E) Competition binding curve for danuglipron using the [³H]PF-06883365 probe. Data represent the mean ± SEM from four individual experiments, each performed in quadruplicate. cAMP, β-arrestin recruitment, and internalization data normalized to a 1 μM GLP-1 response (100%). In the curves, error bars are within the displayed symbols.

Figure 5

Binding affinity of danuglipron, as evaluated using a competition binding assay based on radiolabeled small-molecule agonist [³H]PF-06883365. (A) Structure of the small-molecule agonist radioligand [³H]PF-06883365. (B) Saturation binding analysis for [³H]PF-06883365 showed that [³H]PF-06883365 binds plasma membranes from CHO cells stably expressing a high density of the hGLP-1R (“binding cell line”) with an averaged K_d of 38 nM and a B_max of 5470 fmol/mg. Data represent the mean ± SEM from two independent experiments, each performed in quadruplicate; error bars are within the displayed symbols.

Figure 6

Functional activity of danuglipron at the GLP-1R stably expressed by CHO cells, as assessed using cAMP accumulation assays. (A) cAMP accumulation in CHO cells expressing either the human, cynomolgus monkey, or mouse GLP-1. Data represent the mean ± SEM. (B) EC₅₀ values of danuglipron at the cynomolgus monkey, rat, rabbit, and mouse GLP-1R stably expressed in CHO cells. The EC₅₀ value is expressed as the geometric mean with a 95% confidence interval, while the E_max value is reported as the arithmetic mean with the standard deviation for the number of replicates indicated; error bars are within the displayed symbols. Data normalized to the 1 μM GLP-1 response (100%).

Figure 7

Tryptophan 33 is critical for the function of small-molecule GLP-1R agonists. (A) In contrast to liraglutide, danuglipron does not reduce glucose AUC during an intraperitoneal glucose tolerance test in C57BL/6 mice. (B and C) In contrast to GLP-1, danuglipron promotes cAMP production in GLP-1R-expressing cells only when residue 33 is tryptophan (W), not serine (S). (B) Danuglipron signals in CHO cells expressing human GLP-1R (filled green squares), but not the mouse GLP-1R (filled pink triangles). Danuglipron signaling is restored in CHO cells expressing mouse GLP-1R S33W (empty pink triangles) and is negated in human GLP-1R W33S (empty green squares). (C) GLP-1 promotes signaling in mouse and human wild-type and mutant constructs. (D) Cryogenic electron microscope (cryo-EM) structure of PF-06883365 (green) bound to the human GLP-1R illustrating the map density around the ligand (Protein Data Bank entry 7S15). W33 closes the top of the small-molecule binding pocket. Arginine 380 (R380) interacts with the carboxylic substituent of the small-molecule agonist. Helix 4 was omitted for the sake of clarity (panel D prepared with Chimera). Data represent the mean ± SEM, and error bars are within the displayed symbols. cAMP data are from three independent experiments, each performed in duplicate, and normalized to the 1 μM GLP-1 response (100%).

Figure 8

Danuglipron potentiates glucose-stimulated insulin release and reduces the rate of food intake in monkeys. (A) Profile of danuglipron plasma concentration vs time after intravenous (iv) or oral (po) dosing in male cynomolgus monkeys (n = 2 each). (B–E) Danuglipron increased the rate of glucose disappearance and enhanced insulin secretion during an iv glucose tolerance test (IVGTT) (250 mg of 50% dextrose/kg) in cynomolgus monkeys (n = 8 each). (B) Serum glucose, (C) K value, and (D) serum insulin during the IVGTT when danuglipron was iv infused to 3.0 μM (55 nM unbound) serum levels; liraglutide was administered by subcutaneous injection to achieve 58 nM (0.31 nM unbound) serum levels. (E) iv and po (100 mg/kg) administration of danuglipron potentiated glucose-stimulated insulin release (AUC_0–30 min) in an exposure-proportional manner during an IVGTT. (F) Food intake in monkeys treated with either vehicle or subcutaneously administered danuglipron (2.9 mg/kg) once daily for 2 days (n = 6 each) (orange band). All values are presented as mean ± SEM. *p < 0.05 and **p < 0.01 vs vehicle (mixed model comparisons between LS-Means estimates).

Figure 9

Median plasma danuglipron concentration–time profiles after single-dose oral administration of danuglipron (3–300 mg) to humans (n = 6/dose, except n = 12 in the 300 mg group) in the fasted state. Plasma exposure increased in an approximately dose-proportional manner, as assessed by dose-normalized geometric mean C_max and AUC_inf, with a mean t_1/2 ranging from 4.3 to 6.1 h. The median time to C_max (T_max) values ranged from 2.0 to 6.0 h post-dose. The human pharmacokinetic parameters of danuglipron with associated statistics are listed in Table S7. IR, immediate release.

Scheme 1. Synthesis of Piperidine 12

Scheme 2. Synthesis of Benzimidazole 18

Scheme 3. Synthesis of Danuglipron