Non-peptidic glucose-like peptide-1 receptor agonists: aftermath of a serendipitous discovery

Abstract

Glucagon-like peptide-1 (GLP-1) receptor is an ideal target in the development of incretin-based therapies for diabetes and obesity. Two approaches have been adopted: GLP-1 receptor agonists that mimic the effects of native GLP-1 and dipeptidyl peptidase-4 inhibitors that increase endogenous GLP-1 levels. During the past two decades, search for orally active, non-peptidic GLP-1 receptor agonists has been the focal point of research and development activities in many multinational pharmaceutical companies. Such efforts have not resulted in any success thus far. Serendipitous discovery of substituted cyclobutanes represented by Boc5 as a new class of GLP-1 receptor agonists led us to believe that a small molecule approach to class B G-protein coupled receptor agonism is no longer a fantasy but a reality. However, major obstacles still pose great challenges, and the reasons of which are discussed in this perspectives.

Type 2 diabetes mellitus (T2DM) is a progressive, complex metabolic disorder mainly characterized by insulin resistance and hyperglycemia¹. The prevalence of diabetes, of which T2DM accounts for more than 90% of patients, is rising at an alarming rate in certain parts of the world². Although the exact incidence in China is arguable, the age-adjusted prevalence of diabetes and pre–diabetes was reported to be 9.7% and 15.5%, respectively^{3, 4}. T2DM has thus become a global heath concern due to its complications, such as cardiovascular conditions that result in increased morbidity and mortality. While the currently available anti-diabetic agents are effective in lowering glucose, some of them, including insulin, sulfonylureas and thiazolidinediones, are often limited by weight gain and/or hypoglycemia. However, recent emergence of incretin-based therapies, exemplified by glucagon-like peptide-1 (GLP-1) mimetics and dipeptidyl peptidase-4 (DPP-4) inhibitors, will drastically transform the landscape of diabetes care⁵. GLP-1 maintains glucose homeostasis through several mechanisms that include, but not limited to, stimulating insulin secretion, suppressing glucagon production, improving β-cell mass and function, inhibiting food intake and slowing gastric emptying⁶. Therefore, GLP-1 based incretin therapy is designed to target the fundamental defects of T2DM, capable of reducing both glycosylated hemoglobin (HbA1c) and body weight, and has potential benefits on blood pressure, lipids, and other surrogate markers, leading to decreased cardiovascular risk^{7, 8}.

The native GLP-1 has a very short half-life (<2 min) mainly because of the degradation by DPP-4 and neutral endopeptidase (NEP)⁹. Two approaches have been employed: GLP-1 receptor agonists that mimic the effects of native GLP-1 and DPP-4 inhibitors that increase endogenous GLP-1 levels⁷. The former is represented by two peptidic mimetics, Exenatide^{10, 11} and Liraglutide¹², approved by the US Food and Drug Administration (FDA) for the treatment of T2DM. Other peptides that are under development include long-acting human GLP-1 analog CJC-1131¹³, GLP-1 analog-Fc fusion protein LY2189265¹⁴ and pegylated GLP-1 analog LY2428757¹⁵, recombinant human serum albumin exendin–4 conjugated protein CJC-1134-PC¹⁶, Taspoglutide¹⁷, as well as AVE0010¹⁸.

Unlike peptidic mimetics that require injections, DPP-4 inhibitors (such as Sitagliptin, Vildagliptin and Saxagliptin)^{19, 20, 21} are small in molecular nature, orally active, and transported and stored at ease. Administration of DPP-4 inhibitor has led to improved plasma concentrations of endogenous GLP-1 and marked reduction of HbA1c, but its effect on body weight has been neutral²¹. Thus, oral formulation of GLP-1 analogs remains a development option²².

It is conceivable that an orally active, non-peptidic GLP-1 receptor agonist could resolve the issues mentioned above. Indeed, this has been the focal point of drug discovery efforts in many multinational pharmaceutical companies. Such efforts have not resulted in any success thus far.

The GLP-1 receptor belongs to the glucagon-secretin B family of the G protein-coupled receptors (GPCRs). A characteristic structural feature of this family is a long and structurally complex extracellular amino-terminal domain (N-domain) containing six conserved cysteine residues that form disulfide bonds important for stabilizing the folded protein. The N-domain is connected to a juxtamembrane domain (J-domain) of the seven membrane-spanning α-helices with intervening loops and a C-terminal tail²³. From a biological perspective, class B GPCRs are highly attractive therapeutic targets, but the search for non-peptidic modulators (agonists in particular) has been met with great difficulties. Very few 'druggable' small molecule ligands have been identified for this class of GPCRs. With reference to GLP-1 receptors, only a very small number of non-peptidic agonists have been reported over the years as summarized below in Table 1.

Table 1. The structures of small molecule GLP-1 receptor agonists and their claimed biological activities in vitro and in vivo.

Structure	Receptor binding	Cellular effect	Animal study
24	Exhibited 68% binding activity vs [125I]GLP-1	EC50: 145 nmol/L Cell activation: 99.6%	Not reported
Boc5, full agonist25	Ki: 1.47 μmol/L	EC50: 2.73 μmol/L Caused the release of insulin from normal rat islets22	Diabetic db/db mice: reduction in food intake, body weight, gastric emptying, blood glucose, HbA1c, insulin resistance, etc25, 26
S4P, partial agonist25	Ki: 287 nmol/L	EC50: 1.08 μmol/L At concentrations up to 10 μmol/L evoked maximally only 37% of the GLP-1 response	Diabetic db/db mice: HbA1c was robustly lowered25
T0632, inverse agonist27	IC50: 0.62 μmol/L (wild type receptor) (C347-QQRY mutant receptor)	Decreased the basal GLP-1R activity during a short incubation while increased the surface distribution of the C347-QQRY GLP-1R mutant after a long incubation; right-shifted the dose-response curve of GLP-128	Not reported
Compound A27	IC50: 5.94 μmol/L (wild type receptor) IC50: 5.10 μmol/L (mutant receptor)	Increased cAMP production	Not reported
Compound B27	IC50: 25 μmol/L (wild type receptor) IC50: 20 μmol/L (mutant receptor)	Increased cAMP production	Not reported
29	Not reported	Not reported	Enhanced blood concentrations of GLP-1 30 min after dosing
Ago-allosteric modulator30, 31, 32	Enhanced GLP-1 binding EC50: 0.032 μmol/L)32	EC50: 0.16 μmol/L Emax (%): 85% Caused the release of insulin from normal mouse islets and a perfused rat pancreas32	Not reported
6Cu33	IC50: 3.52 nmol/L Exhibited 36% binding vs [125I] GLP-1	EC50: 0.40 nmol/L (in the presence of GLP-1)	Dose-dependently inhibited acute food intake in normal mice (ED50: 0.90 mg)

Of the compounds listed, diallylmethylamine derivative is more like a DPP-4 inhibitor than a GLP-1 receptor agonist, based on the limited information available in the public domain²⁹. The remaining eight molecules possess fairly diverse structural features. Despite what were described by the inventors for these compounds in terms of orthosteric, allosteric, ago-allosteric, inverse, or partial/full agonists, their functional varieties are manifested in the following manners: (1) inducing both biochemical and cellular responses without in vivo activities^{24, 27}; (2) behaving like an antagonist in cell-based assays (eg, T0632)²⁷; (3) promoting native ligand binding to the receptor in vitro³²; (4) demonstrating a full range of GLP-1 properties including in vivo efficacy^{25, 26}; and (5) enhancing GLP-1 activities via binding to the receptor³³. Since no distinct commonality could be found in the structures of these five types of compounds, their binding to the GLP-1 receptor must be realized via different sites. Among them, Boc5 is the only one that demonstrated therapeutic benefits in vivo^{25, 26}. While peptidic GLP-1 mimetics displayed some side effects such as nausea and vomiting in the clinic⁹, both normal and diabetic db/db mice are highly tolerable to Boc5²⁵ such that a similar reaction on the conditioned taste aversion required a dose well beyond the therapeutic window²⁶. Long-term toxicity (≥3 month) has yet to be determined. It appears that Boc5 is neither metabolized by nor interacts with the cytochrome P450 in vivo exhibiting a half-life ranging from 12.1 to 35.4 h in mice and rats, respectively when injected intraperitoneally. However, its oral bioavailability is extremely poor due to metabolism by esterase in the gut (unpublished data).

The current binding model of GLP-1 receptor is a two-step mechanism where initially the C-terminal part of the peptide ligand interacts with the N-domain of the receptor, thereby conferring high affinity. In the second step, the N-terminal part of the ligand interacts with the core domain of the receptor (transmembrane helices and connecting loops), leading to activation and signal transduction²³. Both the N-domain and J-domain of the receptor are needed for this interaction, and the former is also critical for ligand selectivity between glucagon and GLP-1. Molecular elucidation of such an interaction is nearly impossible due to the inherent complexity of GPCR structures. However, attempts were made using both the crystal structure of human GLP-1 receptor N-terminal extracellular domain³⁴ and computational simulation³⁵. These studies have implicated that activation of GLP-1 receptor by an agonist is related to some intrinsic conformation changes. There may be two major states of the GLP-1 receptor structure: (1) the inactive state, in which the orthosteric agonist-binding site is partially blocked as the consequence of the relative motions between the N-domain and the transmembrane domain; and (2) the active state, in which the orthosteric agonist-binding site is fully accessible³⁵. Binding of an agonist makes a GLP-1 receptor stay in the latter state, as in the case of Boc5²⁵, while interaction with an inverse agonist such as T0632 favors the former state^{23, 27}. Compound 2 (quinoxaline derivative), which binds the GLP-1 receptor at an allosteric site, rigidifies its structure with an open binding site to improve the activity of a full agonist³⁵.

Obviously, the exact requirements needed for a small molecule to fully mimic GLP-1 have yet to be understood. Compounds that confer the conventional wisdom³⁶ such as 'Rule of 5' do not display any in vivo bioactivities, while molecules that demonstrate therapeutic benefits in animal models of T2DM and obesity such as Boc5 are considered not druggable because of poor oral bioavailability. Serendipitous discovery of substituted cyclobutanes represented by Boc5 as a new class of GLP-1 receptor agonists led us to believe that a small molecule approach to class B GPCR agonism is no longer a fantasy but a reality³⁷.

Nonetheless, major obstacles still pose great challenges in terms of developing an orally active non-peptidic GLP-1 receptor agonist. To what extent should such a molecule look like from a medicinal chemistry point of view? Can it possess a suitable molecular size that meets the requisite ADME/T profile while maintaining a full range of incretin-like efficacy? Can it be produced economically, so that it can sufficiently compete against its peptidic counterparts? These are the questions that must be addressed in order to sustain the vitality of a research program in this particular area.