Discovery of a GPR40 Superagonist: The Impact of Aryl Propionic Acid α-Fluorination

Abstract

GPR40 is a G-protein-coupled receptor which mediates fatty acid-induced glucose-stimulated insulin secretion from pancreatic beta cells and incretion release from enteroendocrine cells of the small intestine. GPR40 full agonists exhibit superior glucose lowering compared to partial agonists in preclinical species due to increased insulin and GLP-1 secretion, with the added benefit of promoting weight loss. In our search for potent GPR40 full agonists, we discovered a superagonist which displayed excellent in vitro potency and superior efficacy in the Gα_s-mediated signaling pathway. Most synthetic GPR40 agonists have a carboxylic acid headgroup, which may cause idiosyncratic toxicities, including drug-induced-liver-injury (DILI). With a methyl group and a fluorine atom substituted at the α-C of the carboxylic acid group, 19 is not only highly efficacious in lowering glucose and body weight in rodent models but also has a low DILI risk due to its stable acylglucuronide metabolite.

Type II diabetes (T2D) is a serious chronic disease that is characterized by hyperglycemia, insulin resistance, and insufficient insulin secretion.¹ While existing therapies help patients approach normoglycemia by increasing insulin sensitivity by enhancing insulin secretion or reducing renal glucose reabsorption, they are associated with various adverse effects such as hypoglycemia, pancreatic effects, genital/urinary tract infections, and weight gain.² GPR40 is an emerging new target for T2D therapeutics with a unique mechanism of action. It is a G-protein-coupled receptor which mediates fatty acid-induced glucose-stimulated insulin secretion (GSIS) from pancreatic beta cells and incretion release from enteroendocrine cells of the small intestine. GPR40 agonist TAK-875 (1, Figure 1), developed by Takeda, in phase III studies decreased HbA1c as efficiently as a sulfonylurea without signs of hypoglycemia, which provided proof-of-concept for GPR40 agonism as a drug target.³

Figure 1.

Structures of GPR40 partial and full agonists.

To date, two types of synthetic GPR40 agonists have been reported: partial agonists and full agonists (Figure 1).⁴ TAK-875 was among several GPR40 partial agonists that have been assessed in clinical settings.⁵ Discovered by Amgen researchers, the full agonists exhibit superior glucose lowering to partial agonists in preclinical species due to increased insulin and GLP-1 secretion.^6,7 Another advantage of the full agonists over partial agonists is their ability to lower body weight in chronic settings.⁸

Synthetic GPR40 agonists are constrained fatty acid mimetics with a carboxylic acid headgroup and a hydrophobic tail (Figure 1). The partial and full agonists have different substitution patterns on the aromatic rings, which leads to binding at distinct sites on the receptor.^9,10 Our efforts on discovery of novel GPR40 full agonists began with an examination of the structure–activity relationships of the Amgen full agonist, AM-1638 (4). By incorporating N atoms to the B and C phenyl rings and simplifying the bulky alkyl substitution on the B ring, we identified our first-generation GPR40 full agonists, as represented by 7 (Figure 2).¹¹ The dimethylcyclopentenyl group on AM-1638 is important for its potency,¹² however, we found that it may be replaced with small alkyl groups (7) and may even be removed when the middle aromatic ring is changed to a saturated cyclohexyl ring, like the trans-cyclohexyl compound 8.

Figure 2.

Medicinal chemistry efforts led to a novel series of GPR40 full agonists with a trans-cyclohexyl central ring.

A FLIPR-based intracellular calcium mobilization assay with HEK293 cells overexpressing GPR40 receptors was used as the primary cell-based assay. Because of high receptor reserve, all agonists showed maximal efficacy in this assay. Compound activity was then confirmed in an inositol monophosphate (IP1) HTRF assay and cAMP assay, with CHO-K1 cells expressing relatively low levels of human GPR40 to allow differentiation between partial and full agonists.¹³ As shown in Figure 3A, in the Gα_q-mediated IP1 turnover experiments, both partial agonist (TAK-875) and full agonist (AM-1638) stimulated IP1 accumulation, however, the full agonist showed higher potency and efficacy (E_max).

Figure 3.

GPR40 partial and full agonists are differentiated in IP1 and cAMP assays.

Elucidation of the different signaling pathways for partial and full agonists were further demonstrated using the Gα_s-mediated cAMP assay: only the full agonist (AM-1638) stimulated weak cAMP production, and the partial agonist (TAK-875) showed no signal at the cAMP pathway (Figure 3B). In these assays, 8 behaved similarly to AM-1638 regarding both potency and efficacy (Figure 3). Therefore, 8 is a full agonist. Interestingly, its cis-isomer 9 is a partial agonist like TAK-875 as shown in the same assays. The geometry of the central cyclohexyl ring appears to play an important role in determining the site of GPR40 binding.

The second-generation GPR40 full agonist 8 was profiled extensively in in vitro and in vivo pharmacology. It showed desirable ADME properties and was efficacious in various animal models such as acute and repeat-dose oral glucose tolerance tests (oGTT) in ZDF rats and diabetic cynomolgus monkeys.¹⁴ Chronic dosing of 8 for 38 days in ZDF rats at 10 mg/kg daily dose caused 6% weight loss, accompanied by 10% food intake reduction and a 6.2-fold increase of plasma total GLP-1. In the same study, the partial agonist, TAK-875 did not impact weight, food intake, and GLP-1 secretion.¹⁴

One of the unique features of 8 is its metabolite profile. The major metabolite in human hepatocytes is O-desmethyl-sulfate and not the acyl glucuronide (AG), which is the major metabolite of many other carboxylic acid-bearing drugs. AG metabolites are reactive and could cause idiosyncratic drug toxicity.¹⁵ It was postulated that reactive AG contributed to the DILI, which led to the termination of phase III trials of the GPR40 partial agonist, TAK-875.¹⁶

The synthesis of 8 is illustrated in Scheme 1. In the first synthesis of 8 and 9, Pd/C was used as the catalyst in the hydrogenation of the ester 12. Approximately a 1:1 mixture of the cis- and trans-cyclohexyl isomers were obtained and separated by supercritical fluid chromatography (SFC) at the final step. Because a large quantity of the trans-compound 8 was needed for in vivo studies, we used an organoiridium compound (Crabtree’s catalyst) on the hydrogenation of the alcohol 13 to obtain the trans-isomer 14 almost exclusively (>98%).

Scheme 1.

Reagents and conditions: (a) perfluorobutanesulfonyl fluoride, DBU, THF, rt, 95%; (b) (2-fluoro-5-methoxyphenyl)boronic acid, Pd(dppf)Cl₂, K₃PO₄, 60 °C; (c) LAH, THF, 0 °C, 87%; (d) Crabtree’s catalyst, H₂, DCM, 40 °C, 79%; (e) methyl (S)-3-cyclopropyl-3-(3-hydroxyphenyl)propanoate, DEAD, n-Bu₃P, ACN, 80 °C, 63%; (f) NaOH, THF, MeOH, 30 °C, 79%.

Compound 8 was not advanced due to its high predicted human daily dose (>600 mg/QD based on 5% weight loss efficacy in a DIO mouse model). Because potency and plasma protein binding (PPB) are two important factors in prediction of human dose, we elected to search for a compound which was more potent and less protein bound than 8. We systematically modified the three ring systems of 8, e.g., incorporation of a N atom on the A and C rings and exchanged the central cyclohexyl ring with other heterocyclic rings. However, none of these changes was successful. We became aware that researchers at Merck had pursued a series of biaryl chroman GPR40 full agonists and observed a potency boost when a methyl group was installed at the α-C of the carboxylic acid.¹⁷ The same strategy was applied to 8. As predicted, the S-configured methyl substitution (15) greatly improved potency of 8 in both Ca²⁺ and IP1 assays (Table 1). Its R-diastereomer 16 was slightly less potent than 8 in the Ca²⁺ assay, but about 10-fold more potent in the IP1 assay. Encouraged by the potency improvement seen with α-methyl substitution, we next examined the effect of other small substitutions, like an electron withdrawing fluorine atom at the same position. For the fluorinated analogues, the S,S-configured 17 was much less potent than its S,R-diastereomer 18, which retained human GPR40 activity in the Ca²⁺ assay. Compound 18 was equally potent as 8 in the Ca²⁺ assay but 10-fold more potent in the IP1 assay.

Table 1. Effect of α Methyl and Fluorine Substitutions on in Vitro GPR40 Activity.

compd	R1	R2	human Ca2+a EC50 (nM)c	rat Ca2+a EC50 (nM)c	human IP1b EC50 (nM)c
1			31 ± 18	39 ± 18
8	H	H	0.94 ± 0.37	4.1 ± 2.4	199 ± 102
15	Me	H	0.2	0.4	2.0
16	H	Me	2.2	4.9	22
17	F	H	54	163	NA
18	H	F	1.1	28	20
19	Me	F	0.25 ± 0.13	0.34 ± 0.18	5.3 ± 4.3
20	F	Me	6.1	6.9	89
21	Me	Me	0.11	0.31	220
22	F	F	16	34	NA

^aFLIPR based intracellular calcium mobilization assay.

^bIP1 assay on cells expressing low levels of hGPR40.

^cData are represented as the average of duplicates. When the assay was run more than 3 times for the same compound, data are represented as averages ± SD.

We observed that 18, with fluorine substituted at α-carbon, showed a higher E_max than the other full agonists in the cAMP assay (data not shown). We postulated that the electronic effect of the fluorine atom on the carboxylic acid might have amplified the Gα_s signaling, thus producing additional cAMP. Next, we disubstituted the α-C with a methyl group and a fluorine atom. Compound 19, with the same configuration of the methyl group as the monomethyl substituted 15, was equally potent as 15 in both Ca²⁺ and IP1 assays (Table 1). Its dose–response curve in the IP1 assay was left-shifted significantly compared to the curves of other full agonists, 7 and 8 (Figure 4A), while its efficacy remained the same as the full agonists. Thus, in the Gα_q-mediated signaling pathway, 19 acted as a full agonist with superior potency. In the cAMP assay, the curve of 19 was also left-shifted indicating enhanced potency. More remarkably, its E_max was almost doubled compared with the other full agonists, which had not been seen previously (Figure 4B). On the basis of its superior potency and greater efficacy than the other full agonists in the cAMP assay, 19 was deemed to be a superagonist in the Gα_s-mediated signaling pathway. There are reports of GPCR superagonists that display superagonism in one of all possible signaling pathways,¹⁸ and this is the first identification of a GPR40 superagonist.

Figure 4.

Compound 19 showed superior potency in the IP1 assay and superior efficacy in the cAMP assay.

Compound 20, the epimer of 19, was about 30-fold less potent than 19 in the IP1 assay, however, it had the same E_max as 19 in the cAMP assay (data not shown). The superior efficacy in the cAMP assay has been seen with three compounds (17, 19, and 20), all of which have a fluorine atom at the α-position. Dimethyl substituted compound 21 was very potent in the Ca²⁺ assays but lost its potency in the IP1 assay, and its E_max level was the same as 8. This SAR suggested that the methyl substitution (steric effect) drove the potency, and the fluorine substitution (electronic effect) contributed to the superior efficay in stimulating cAMP accumulation. Difluoro substitution at the α-C (22) resulted in loss of potency in the Ca²⁺ assay, and its activity in the IP1 assay was not determined.

In the ex vivo GSIS assay using human islets, 8 was efficacious at potentiating insulin secretion in the presence of high glucose (12 mM) with an EC₅₀ value of 1.37 μM (Figure 5A). In the same assay, the partial agonist TAK-875 showed about 40% efficacy of 8 at 10 μM, suggesting that the full agonist may have greater therapeutic utility as an anti-diabetic agent than the partial agonist. The markedly enhanced potency of 19 demonstrated in this assay that its EC₅₀ was 0.0037 μM, about 370-fold more potent than 8 (Figure 5B).

Figure 5.

Compound 19 demonstrated superior potency in the GSIS assay in human islets.

The full agonist 8 and superagonist 19 bind to hGPR40 in the same binding site as AM-1638 according to the competition binding experiments performed in membranes from stably transfected CHO-K1 cells. Both 8 and 19 fully competed off radio-labeled AM-1638 while enhancing the binding of a radio-labeled partial agonist, showing the characteristics of agoPAMs (positive allosteric modulators). Even though 8 and 19 are both full agonists, they have shown differences in the binding kinetics to the hGPR40 membrane. Compound 8 was fast-on and fast-off, with association T_1/2 of 0.8 min and dissociation T_1/2 of 0.14 min. Compound 19 was much slower on and off the membrane than 8, with association T_1/2 of 3.3 min and dissociation T_1/2 of 20 min. Even though slow binding kinetics may play a role for sustained GPCR agonist responses,¹⁹ the correlation between the slower binding kinetics of 19, and its enhanced efficacy remains to be explored.

Enantioselective synthesis of α-methylated 15 utilized an asymmetric hydrogenation condition first described by Christensen et al.²⁰ Deprotonation of 15 with a strong base followed by quenching with a fluorination agent afforded the mixture of 19 and 20, which was separated by SFC. After 19 was identified as the compound of interest, an enantioselective synthesis was developed. We used a chiral auxiliary to synthesize 19 with >99% dr (Scheme 2). In this synthesis, with the bulky chiral auxiliary in place, the order of fluorination step and methylation step was crucial. Unlike in the original method, the fluorination step had to precede the methylation step, otherwise the fluorination would not occur.

Scheme 2.

Reagents and conditions: (a) oxalyl chloride, DMF (cat.), DCM, 100%; (b) (S)-4-isopropyl-5,5-dimethyloxazolidin-2-one, n-BuLi, THF, −78 °C, 93%; (c) LDA, NFSI, THF, −78 °C, 92%; (d) CH₃I, LiHMDS, THF, −78 °C, 72%; (e) LiOH, H₂O₂, THF, rt, 94%.

The class of GPR40 full agonists, represented by AM-1638 and 8, are lipophilic acids with high cLogP. Adding one methyl group at the carboxylic acid α-carbon (15) increases its lipophilicity, while addition of a fluorine atom (18) makes it less lipophilic (Table 2). With both a methyl and fluorine, 19 has the same cLogP as 8. However, the pK_a values of the fluorine-bearing compounds, 18 and 19, are much lower than 8. The substituents at the carboxylic acid α-carbon also affected the metabolite profiles of the compound when incubated with human hepatocytes. The monomethyl substituted 15 has a similar pK_a value to 8, but unlike 8 its major metabolite was the AG. Compound 18 produced the least amount of AG, while the major metabolite of 19 was the AG. Because of the potential DILI risk caused by reactive AG metabolite, as in the case of TAK-875, we examined the stability of the AG metabolites of 8 and 19. The longer the half-life of the AG, the more stable it is. The half-life of AG of 19 was over 7 h and much longer than the half-life of AG of 8, which was about 3 h. Both 8 and 19 had more stable AG metabolites than TAK-875, whose AG had a half-life of 0.5 h. According to Sawamura et al.,²¹19 should have a lower risk for DILI due to its more stable AG metabolite.

Table 2. Metabolite Composition in Human Hepatocytes.

			metabolites (%)a
compd	log Pb	pKab	unchanged drug (UD)	UD-acyl glucuronide	O-desmethyl sulfate	other
8	6.9	4.6	56	17	23	4
15	7.2	4.7	45	44	6	5
18	6.6	2.9	63	8	22	7
19	6.9	3.0	73	22	4	1

^aIncubation of human hepatocytes (1 million cells/mL) in duplicate for 2 h; metabolite identification and profiling were obtained using LC-UV/MS.

^bCalculated.

In terms of ADME profile,²²19 has good in vitro metabolic stability (in microsomes and hepatocytes) in mutiple species including human, no CYP inhibition/induction, no gene toxicity (in vitro micronucleus test and Ames test), and no off-target activity (hERG, CEREP, and kinase panels).

The pharmacokinetic profile of the superagonist 19 in SD rats was not the best among the compounds studied in terms of oral exposure and clearance rate (Table 3). However, its volume of distribution was higher than the other carboxylic acids and it had good oral bioavailability. Given its superior in vitro potency, 19 was further profiled for its efficacy in rodent models.

Table 3. Pharmacokinetic Properties in SD Rats.

compd	poaCmax (ng/mL)	po T1/2 (h)	ivbT1/2 (h)	Cl (mL/min/kg)	Vdss (L/kg)	% F
8	18195	9.1	7.2	0.4	0.2	91.4
15	1356	4.4	4.8	8.1	0.9	45.5
18	3700	4.3	4.7	2.7	0.9	73.8
19	495	7.9	4.3	17	3.2	92.1

^a10 mg/kg po dose formulated in 0.5% methocel.

^b2 mg/kg iv dose formulated in 20% HPβCD with pH adjustment.

Oral administration of 19 in DIO mice 40 min prior to dextrose challenge in oGTT significantly reduced blood glucose excursion in a dose-dependent manner (Figure 6). At a dose of 1 mg/kg, 19 demonstrated comparable efficacy to 8 at 10 mg/kg, yet with a much lower drug exposure level than 8. The maximal efficacious dose for 19 in this model is 3 mg/kg with a drug plasma level of 0.87 μM.

Figure 6.

Compound 19 dose dependently lowered blood glucose in the DIO mice oGTT acute model.

GPR40 full agonists stimulate secretion of the gut hormone GLP-1,²³ and GLP-1 is known for suppression of appetite, which may lead to weight loss.²⁴ We have seen the weight loss efficacy of 8 in ZDF rats, and we next tested 8 and 19 for weight loss efficacy in DIO mice. In the 15-day study, DIO mice was dosed orally once a day with either vehicle, 19 (at 0.3, 1, 3, and 10 mg/kg) or 8 (at 30 mg/kg). Compound 19 dose dependently caused weight loss of these obese mice (Figure 7A), accompanied by reduced food intake in a dose-dependent manner (Figure 7B). At 10 mg/kg, 19 demonstrated more weight loss efficacy (>13% WL compared with vehicle animals) than 8 at 30 mg/kg (∼10% WL compared with vehicle animals) and with a 55-fold less plasma drug exposure (Figure 7C).

Figure 7.

(A) Percentage change in body weight over 15 days on qd dosing of DIO mice with 8 and 19. (B) Cumulative food intake over 15 days. (C) The weight loss effect caused by GPR40 full agonists 8 and 19, with plasma exposure at day 14.

The superagonist 19 shows a remarkable potency improvement over 8 in both in vitro and in vivo experiments. Because both compounds are highly plasma protein bound (>99.9%), the Dianorm system (equilibrium dialysis) was used to obtain definitive PPB data.²⁵ Compound 19 was about 10-fold less protein bound than 8, with 0.015% free unbound fraction (Table 4). Allometric scaling derived from mouse, SD rat, ZDF rat, and monkey PK gave estimated human PK parameters for the compounds. For a 70 kg human, 19 has a clearance of 44 mL/h/kg and volume of 1190 mL/kg with a calculated elimination half-life of 18.8 h. It has a larger volume distribution and longer half-life than 8, even though its clearance is higher. Multiple human dose predictions were modeled based on efficacy data from different preclinical PD models. On the basis of the glucose lowering effect in rodents (ZDF rats and DIO mice), the projected human dose is around 1 mg for 19, while it is over 200 mg for 8 (data not shown). On the basis of the 5% weight loss effect in DIO mice, the projected human dose for 19 is 9.4 mg/QD, while it is 664 mg/QD for 8. With a <10 mg daily dose, 19 is considered to have a low DILI risk according to an analysis by Lammert et al.²⁶

Table 4. Plasma Protein Binding and Projected Human Dose.

			human PKb
compd	IP1 EC50 (nM)	% fua	Cl (ml/h/kg)	Vd (mL/kg)	T1/2 (h)	human dosec (mg/QD)
8	199	0.0015	24	301	8.7	664
19	3.1	0.015	44	1190	18.8	9.4

^aFree fraction of drug in plasma; Dialysis incubation for 6 h (equilibrium established) in 20% plasma with 10 μM compound (n = 3).

^bSpecies included in allometric scaling: mouse, rat, ZDF-rat, and monkey.

^cBased on 5% weight loss in DIO mice PD model.

In conclusion, herein we report the discovery of a GPR40 superagonist 19 with superior in vitro potency and efficacy to the other agonists. It is highly efficacious in reducing glucose and lowering body weight in rodent models. Unlike many other carboxylic acid-bearing drugs, it has a low risk toward DILI due to its stable AG metabolite and <10 mg/day projected human dose. We credit these improvements to the installation of a methyl group and a fluorine atom on the α-carbon of the carboxylic acid headgroup. Both steric and electronic effects play important roles in achieving improved profiles of 19. The remarkable efficacy of 19 on glycemic control and weight loss in preclinical studies illustrate the promises of GPR40 full agonism as a new mechanism to treat T2D.

Glossary

Abbreviations

GPR40

g-protein coupled receptor 40

GLP-1

glucagon-like peptide 1

FLIPR

fluorometric imaging plate reader

HEK293

human embryonic kidney cell line 293

CHO

Chinese hamster ovarian cell line

cAMP

cyclic adenosine monophosphate

ADME

absorption, distribution, metabolism, and excretion

SFC

supercritical fluid chromatography

LiHMDS

lithium bis(trimethylsilyl)amide

THF

tetrahydrofuran

LAH

lithium aluminum hydride

DCM

dichloromethane

ADDP

1,1′-(azodicarbonyl)dipiperidine

QD

every day

DMF

dimethylformamide

LDA

lithium diisopropylamide

NSFI

N-fluorobenzenesulfonimide

mpk

mg per kg

HPβCD

hydroxypropyl-β cyclodextrin

DIO

diet-induced obese

ZDF

Zucker diabetic Fatty

PD

pharmacodynamic;

Biography

Mark R. Player received his doctorate in medicinal chemistry from the University of South Carolina and his M.D. degree from the Medical University of South Carolina in 1986. During postdoctoral training with Paul Torrence at the National Institutes of Health he worked on synthetic approaches to catalytic RNA/DNA chimeras directed against viral targets. In 1998 he accepted a position at 3-Dimensional Pharmaceuticals, and after the acquisition and merger of 3DP into Janssen in 2003, he began leading a medicinal chemistry team at the Spring House research site. He has delivered NMEs in the anti-inflammatory, analgesics, and cardiovascular/metabolic therapeutic areas into development at Janssen.

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsmedchemlett.8b00444.

• Synthetic procedures, analytical data, and assay protocols (PDF)

The authors declare no competing financial interest.