Optimization of GPR40 Agonists for Type 2 Diabetes

Abstract

GPR40 (FFA1 and FFAR1) has gained significant interest as a target for the treatment of type 2 diabetes. TAK-875 (1), a GPR40 agonist, lowered hemoglobin A1c (HbA1c) and lowered both postprandial and fasting blood glucose levels in type 2 diabetic patients in phase II clinical trials. We optimized phenylpropanoic acid derivatives as GPR40 agonists and identified AMG 837 (2) as a clinical candidate. Here we report our efforts in searching for structurally distinct back-ups for AMG 837. These efforts led to the identification of more polar GPR40 agonists, such as AM-4668 (10), that have improved potency, excellent pharmacokinetic properties across species, and minimum central nervous system (CNS) penetration.

Type 2 diabetes mellitus (T2DM) is a metabolic dysfunction characterized by increased insulin resistance and impaired insulin secretion, resulting in higher blood glucose levels. Insulin secretagogues, such as sulfonylureas and glinides, are commonly used to increase insulin levels in diabetic patients.¹ However, these drugs promote insulin secretion independent of blood glucose levels, thereby leading to the risk of hypoglycemia.^1,2 Novel insulin secretagogues with a low risk of hypoglycemia are thus desirable.

G-protein coupled receptor 40 (GPR40, IUPHAR recommended name: FFA1) is highly expressed in pancreatic β cells and responds to endogenous fatty acids resulting in amplification of insulin secretion only in the presence of elevated glucose levels.³⁻⁶ Therefore, GPR40 agonists may lower blood glucose levels with decreased risk of hypoglycemia compared to classic insulin secretagogues. On the basis of these findings, GPR40 has drawn considerable attention as a potential therapeutic target for T2DM.⁷⁻¹³ Robust antiglycemic effects were demonstrated using GPR40 agonists in rodent diabetic models.^7−9,14,15 Positive proof of concept data was obtained in humans using a GPR40 agonist TAK-875 (1, Figure 1) in phase II clinical trials: TAK-875 lowered hemoglobin A1c (HbA_1c) by an amount similar to that of glimepiride in type 2 diabetic patients at week 12, and hypoglycemic incidents for all doses of TAK-875 were similar to placebo and significantly lower compared to glimepiride treatment groups.¹⁶ TAK-875 is now being evaluated in phase III clinical trials.¹⁷

Figure 1.

TAK-875 (1), AMG 837 (2), and a close analogue (3).

We reported the identification of AMG 837 (2, Figure 1), which is a potent and selective GPR40 agonist selected for evaluation in human clinical trials.^14,15 AMG 837 shows similar PK properties, potency and efficacy in rats to those of 1.^7,15 After the discovery of compound 2, we initiated an effort in searching for structurally distinct GPR40 agonists. We emphasized achieving structural diversity through introducing greater polarity into the molecule. Because compound 2 is a lipophilic acid with low polar surface area (46.5, Table 1), it is likely that 2 has significant central nervous system (CNS) exposure. A close analogue of 2 showed a rat brain to plasma ratio of 0.6 three hours after an oral dose of 5 mg/kg. While GPR40 is expressed in the brain, its function there is unclear. However, its role in modulating insulin secretion is through direct activity on the pancreatic beta cell.^6,13,18 In order to avoid both potential target mediated and off-target effects in the CNS, we aimed to design non-CNS penetrant GPR40 agonists. Since the range of physicochemical properties that allows ready access to the CNS is rather limited,¹⁹ we took the approach of introducing greater polar surface area (PSA) into the molecules to decrease CNS penetration.²⁰

Table 1. GPR40 Activity and Rat Pharmacokinetic Properties of Compounds 4–6.

^aGPR40 aequorin assay in CHO cells stably transfected with human GPR40. Assay was run in buffer containing 0.01% human serum albumin. Values are reported as the average of at least two determinations.

^bGPR40 IP₃ assay using A9 cells stably transfected with human GPR40 in 0.3% human serum. See refs (14) and (15) for assay protocol. V values are reported as the average of at least two determinations.

^cTopological polar surface area was calculated using Daylight Toolkit v4.8.1.

^dIV dose at 0.5 mg/kg.

^eOral dose at 2 mg/kg.

^fValues are from one experiment.

^gNot determined.

One successful way to introduce greater PSA was by replacing the propynyl group of AMG 837 (2) with polar heterocycles. Among the initial heterocycles studied were imidazole, triazole, and tetrazole (Table 1). We initially set the biphenyl moiety to that of compound 3 (Figure 1) because it maximized potency. GPR40 is a G_αq-coupled GPCR, so activation of GPR40 by the compounds was measured in an aequorin assay using CHO cells stably transfected with human GPR40 and in an inositol phosphate accumulation assay using A9 cells also stably transfected with human GPR40.^14,15 As shown in Table 1, imidazole 4 and triazole 5 showed similar EC₅₀ to those of compounds 2 and 3. Tetrazole 6 was more potent, but its rat clearance was high (CL = 2.8 L/h/kg). It was interesting to notice that the clearance values trended higher as the number of nitrogen in the heterocycles increased.

The observations revealed in Table 1 were confirmed in compounds that have different biaryl tail pieces. The phenylthiazole tail (Table 2) has a good degree of structural diversity compared to the biphenyl tails and retains potency on the receptor. With the phenylthiazole tail, tetrazole 9 was also more potent than imidazole 7 and triazole 8 (Table 2). Once again, in vivo rat clearance increased with increasing nitrogen count in the heterocyclic ring β to the carboxylic acid (compounds 7–9).

Table 2. GPR40 Activity and Rat Pharmacokinetic Properties of Compounds 7–9.

^aGPR40 aequorin assay in CHO cells stably transfected with human GPR40. Assay was run in buffer containing 0.01% human serum albumin. Values are reported as the average of at least two determinations.

^bGPR40 IP₃ assay using A9 cells stably transfected with human GPR40 in 0.3% human serum. See refs (14) and (15) for assay protocol. Values are from one experiment.

^cTopological polar surface area was calculated using Daylight Toolkit v4.8.1.

^dIV dose at 0.5 mg/kg.

^eNot determined.

^fOral dose at 2 mg/kg.

It was desired to retain the improved potency of the tetrazoles without the increased clearance. The microsomal stability of the tetrazoles was similar to that of the imidazoles and triazoles. It is possible that the tetrazole head piece is recognized by an efflux transporter, a phase II metabolic enzyme, or both. The recognition seems to be reinforced as the number of heteroatoms in the heterocycles rises. On the basis of this observation, we reduced the number of heteroatoms and evaluated positioning of the remaining ones to maintain the good potency of the tetrazoles. A number of heterocycles were evaluated, and isoxazole 10 (designated AM-4668) was as potent as the corresponding tetrazole (Table 3). Other oxazoles, such as 11 and 12, were less potent.

Table 3. GPR40 Activity of Compounds 10–12.

^aGPR40 aequorin assay in CHO cells stably transfected with human GPR40. Assay was run in buffer containing 0.01% human serum albumin. Values are reported as the average of at least two determinations.

^bGPR40 IP₃ assay using A9 cells stably transfected with human GPR40 in 0.3% human serum. See refs (14) and (15) for assay protocol. Values are from one experiment.

^cTopological polar surface area was calculated using Daylight Toolkit v4.8.1.

^dValues are from one experiment.

^eNot determined.

With the number of heteroatoms reduced, AM-4668 indeed displayed excellent pharmacokinetic properties, not only in rat but also in dog and cynomolgus monkey (Table 4). It had low clearance, moderate to long half-life, and very good oral bioavailability across species. Unlike some compounds with the propynyl headgroup as 2 and 3, CNS penetration of AM-4668 was low, as indicated by a rat brain to plasma ratio of 0.02 three hours after an oral dose of 5 mg/kg. In addition, this compound showed no inhibition of the hERG channel and demonstrated low potential to induce drug–drug interaction by various CYP inhibition and induction assays.

Table 4. Pharmacokinetic Properties of AM-4668.

species	clearance (L/h/kg)a	T1/2 (h)	Vdss (L/kg)	oral bioavailabilityb
rat	0.09	5.3	0.68	77%
dog	0.15	5.6	0.60	100%
cynoc	0.04	14	0.35	65%

^aIV dose at 0.5 mg/kg.

^bOral dose at 2 mg/kg.

^cCynomolgus monkey.

Since the activity of AM-4668 was weak against rodent GPR40, it was difficult to study its efficacy in standard rodent diabetes models. We therefore evaluated it in mice where the human GPR40 gene has been replaced with the mouse GPR40 gene but is under the control of the endogenous mouse GPR40 promoter.²¹ Oral administration of AM-4668 at 10 mg/kg dose 1 h before oral glucose challenge significantly reduced the blood glucose levels (Figure 2A). The glucose AUC in the compound-treated animals is 19% lower than that in the vehicle-treated animals (Figure 2B). In a separate mouse oral PK study, the plasma concentration of AM-4668 at 1 mg/kg dose was about 2 μM for the first 4 h, which corresponded to an 8 nM unbound concentration. In addition, AM-4668 induced insulin secretion in pancreatic islets isolated from human GPR40 knock-in mice with an EC₅₀ of 55 nM (Figure 3).²² Because AM-4668 does not effectively enter the CNS, these results demonstrate that agonism of GPR40 by this compound in the CNS is likely not necessary for its glucose lowering effects in this model.

Figure 2.

Oral glucose challenge study of AM-4668 (10) in hGPR40 knock-in mice.

Figure 3.

Insulin secretion of AM-4668 (10) in islets isolated from hGPR40 knock-in mice.

The synthesis of AM-4668 is shown in Scheme 1. Michael addition of (S)-3-acetyl-4-benzyl-2-oxazolidinone (14) to trans-β-nitrostyrene 13 in the presence of titanium chloride and Hunig’s base produced compound 15.²³ Generation of the nitrile oxide²⁴ from the nitro group in 15 and reaction with vinyl bromide generated the isoxazole compound 16. Deprotection of the benzyl group of 16 using boron trichloride methyl sulfide complex afforded phenol 17, which was alkylated using commercially available choloromethylthiazole 18 to yield compound 19. Finally the oxazolidinone in compound 19 was removed in the presence of lithium hydroxide and hydrogen peroxide to afford AM-4668 (10).

Scheme 1. Synthesis of Isoxazole AM-4668 (10).

In summary, we replaced the propynyl group of AMG 837 (2) with heterocycles and introduced greater PSA into the molecules. It is likely that the polar heterocycles in the head piece significantly reduced CNS penetration of the molecules. After optimization for potency and PK, we identified isoxazole AM-4668 (10), which is more potent than AMG 837 (2), possesses excellent pharmacokinetic properties across species, and reduces the plasma glucose levels in an OGTT study in human GPR40 knock-in mice.

Supporting Information Available

Detailed synthetic experimental procedures and characterization for all compounds. This material is available free of charge via the Internet at http://pubs.acs.org.

The authors declare no competing financial interest.