Discovery, design and synthesis of a selective S1P₃ receptor allosteric agonist

Abstract

Potent and selective S1P₃ receptor (S1P₃-R) agonists may represent important proof-of-principle tools used to clarify the receptor biological function and assess the therapeutic potential of the S1P₃-R in cardiovascular, inflammatory and pulmonary diseases. N,N-Dicyclohexyl-5-propylisoxazole-3-carboxamide was identified by a high-throughput screening of MLSMR library as a promising S1P₃-R agonist. Rational chemical modifications of the hit allowed the identification of N,N-dicyclohexyl-5-cyclopropylisoxazole-3-carboxamide, a S1P₃-R agonist endowed with submicromolar activity and exquisite selectivity over the remaining S1P_1,2,4,5-R family members. A combination of ligand competition, site-directed mutagenesis and molecular modeling studies showed that the N,N-dicyclohexyl-5-cyclopropylisoxazole-3-carboxamide is an allosteric agonist and binds to the S1P₃-R in a manner that does not disrupt the S1P₃-R–S1P binding. The lead molecule herein disclosed constitutes a valuable pharmacological tool to explore the molecular basis of the receptor function, and provides the bases for further rational design of more potent and drug-like S1P₃-R allosteric agonists.

Sphingosine-1-phosphate (S1P) is a naturally occurring lysophospholipid produced in most cell types, regulates fundamental biological processes and functions including cell proliferation, cell growth, angiogenesis, and inhibits apoptosis and lymphocyte trafficking. ¹ The generation of S1P is mediated by two cytosolic sphingosine kinase isoforms (SPK1 and SPK2) and occurs preferentially in the plasma membrane.² S1P has two distinct signaling roles, as an intracellular second messenger and as an extracellular ligand for a specific family of G-protein coupled receptors (GPCRs) named S1P_1–5-Rs.³ S1P can also affect cell function by either binding or modifying putative intracellular targets or by affecting the relative levels of other lipid products, particularly sphingomyelin and ceramide whose biological effects oppose those of S1P.^4,2a

Increasing evidence supports that different subfamilies of ATP-binding cassette (ABC) transporters contribute to, or are involved in, secreting the S1P lipids across the plasma membranes.⁵ The differential temporal and spatial pattern and intracellular signaling pathways of each S1P-R enable S1P to exert its diverse biological functions. S1P_1–3-Rs are widely expressed in almost all organs in mice and humans whereas S1P₄-R and S1P₅-R expression is restricted to specific organs and cell types.⁶ In humans S1P₃-R is highly expressed in heart, lungs, spleen, kidney and pancreas. S1P₃-R couples with G_i/o, G_q, and G_12/13. Interestingly, S1P₃-R deficient mice showed no obvious phenotype.⁷ However, clear association of S1P₃-R has been demonstrated in several cardiovascular functions including regulation of heart rate and blood pressure in rodents,⁸ vasorelaxation⁹ and cardiac fibrosis.¹⁰ Additionally, regulation of myocardial perfusion and cardiomyocytes protection in ischemia perfusion injury was dependent on the S1P₃-R.¹¹ Furthermore, studies with S1P₃-R-null mice and PAR1-deficient dendritic cells (DC) have shown that S1P₃-R acts as a downstream component in PAR₁-mediated septic lethality.¹² S1P₃-R is also shown to regulate the endothelial cells in splenic marginal sinus organization¹³ and the stimulation of endothelial progenitor cells.¹⁴ S1P₃-R has also been implicated in mediating and amplifying inflammatory responses in various CNS disorders in autocrine and paracrine fashion.¹⁵ Recently, it has been shown that S1P₃-R is markedly up-regulated in a subset of adenocarcinoma cells, and knock down of this receptor subtype inhibits proliferation and growth of lung adenocarcinoma cells.¹⁶

A wealth of S1P-Rs agonists has been described in the literature. However, the development of subtype selective S1P₃-R agonists as useful pharmacological tools has been limited as a consequence of targeting the orthosteric binding site for receptor activation. The orthosteric binding site displays a high level sequence homology among the S1P-R family subtypes. S1P₁-R and S1P₃-R are the most closely related by sequence, particularly at their agonist binding pockets which consist of a lower hydrophobic and an upper polar region where Leu-276 in S1P₁-R and Phe-263 in S1P₃-R are the main difference. Receptor structure modeling and ligand docking studies revealed that the S1P₃-R binding pocket is contracted between the lower lipophilic area and the upper polar section by 1.5–1.8 Å compared to the S1P₁-R due to the presence of Phe-263. These differences in steric and space constrains in the S1P₃-R orthosteric binding site may explain the difficulty in designing S1P₃-R agonists devoid of S1P₁-R agonist activity.¹⁷

Topologically distinct from the conserved orthosteric binding site, an allosteric site provides a means to overcome important selectivity issues associated with the orthosteric ligands, in particular within GPCRs in which the orthosteric site is highly conserved between subtypes. In addition to offering a potential subtype-selectivity, allosteric ligands may stabilize different conformations and functional states, thus activating a distinct repertoire of receptor signaling and regulatory properties that orthosteric ligands are unable to initiate.¹⁸ The S1P binding site is highly conserved among the S1P-R family thus selective allosteric ligands may be useful pharmacological tools to decipher individual receptor biological functions.

A high-throughput screening (HTS) of the Molecular Libraries-Small Molecule Repository (MLSMR) library identified the N,N-dicyclohexyl-5-propylisoxazole-3-carboxamide 1a (Fig. 1) as a S1P₃-R agonist with acceptable in vitro potency/selectivity profile. ^17,19 The structural integrity of the hit was corroborated by the re-synthesis (Scheme 1) of the title compound that showed confirmed EC₅₀’s of 434 nM at S1P₃-R, 7.87 µM at S1P₁-R and no agonist activity at S1P_2,4-Rs at concentrations up to 50 µM (Fig. 1).

Figure 1.

HTS S1P₃-R agonist hit 1a.

Scheme 1.

Synthesis of 1a–d. Reagents and conditions: (i) (a) 2 (1 equiv), SOCl₂, benzene, reflux, 3 h; (b) 3a–d (1.5 equiv), DIPEA (1.5 equiv), CH₂Cl₂, 0 °C–rt, 3 h, 70– 98% (over two steps).

Our SAR studies commenced varying the amide region C. The synthesis and biological results of 1b–d are outlined in Scheme 1. The carboxylic acid 2 was transformed into the corresponding acid chloride and coupled with amines 3b–d to furnish the amide products 1b and 1d. Surprisingly, exchanging a cyclohexyl from 1a for a cyclopentyl group (1b) led to loss in potency of 18-fold for the S1P₃-R but only of three-fold for the S1P₁-R. Furthermore, changing a cyclohexyl for a phenyl (1d) or hydrogen (1c) led to complete loss of potency at both receptors. These results underscore the key role played by the N,N-dicyclohexyl amide for binding to the S1P₃-R.

Next we explored region A of the HTS-hit while keeping regions B and C constant. Compounds 5a–i (Table 1) were synthesized from a series of isoxazole carboxylic acids 4a–i commercially available (4a–c and 4f–i) or readily obtained (4d and 4e) according to literature procedures (Scheme 2).²⁰

Table 1.

S1P₃-R agonist activity of compounds 5a–i

Compd	Carboxylic acid	R	EC50α (nM)
			S1P3-R	S1P1-R
5a	4a	Isobutyl	1375	6420
5b	4b	Isopropyl	959	22,250
5c	4c	Methyl	6595	>50,000
5d	4d	Cyclohexyl	559	>50,000
5e	4e	Cyclopentyl	339	>50,000
5f	4f	Cyclopropyl	105	33,300
5g	4g	Phenyl	103	662
5h	4h	4-Methoxyphenyl	323	859
5i	4i	3,4-Diethoxyphenyl	8070	1104

^αData are reported as mean of n = 3 determinations.

Scheme 2.

Synthesis of 5a–i. Reagents and conditions: (i) (a) 4a–h (1 equiv), SOCl₂, benzene, reflux, 3 h; (b) 3a (1.5 equiv), DIPEA (1.5 equiv), CH₂Cl₂, 0 °C–rt, 3 h, 85– 98% (over two steps).

Interestingly, the isobutyl 5a was approximately three-fold less potent than the hit at the S1P₃-R and four-fold less selective against the S1P₁-R. The isopropyl derivative 5b was approximately two-fold less active than 1a for the S1P₃-R, but the selectivity against the S1P₁-R remained similar. The methyl derivative 5c was 15-fold less potent than 1a. Remarkably, the cyclohexyl (5d) and cyclopentyl (5e) analogs were slightly less and more active at the S1P₃-R and inactive at the S1P₁-R. Of note, the cyclopropyl derivative 5f (CYM5541) was found four-fold more potent than the hit and nearly 18-fold more selective against the S1P₁-R. The phenyl derivative 5g (CYM5544) was found equipotent to 5f for the S1P₃-R but significantly less selective against the S1P₁-R. Interestingly, the 4-methoxyphenyl 5h was slightly more potent than the hit compound at the S1P₃-R, but its selectivity against the S1P₁-R was less than three-fold. Intriguingly, 3,4-diethoxyphenyl 5i was 18-fold less potent than the hit at the S1P₃-R but more active (~7-fold) at the S1P₁-R. All this information together indicates that an aromatic ring at position five of the isoxazole hit (portion A), although favorable for the S1P₃-R activity, is also detrimental for the selectivity against the S1P₁-R.

Based on the obtained results, we focused our attention on the SAR studies of 5f, particularly on the amide region C. The synthesis of 7a–l is depicted in Scheme 3. The biological results are listed in Table 2.

Scheme 3.

Synthesis of 7a–l. Reagents and conditions: (i) (a) 4d (1 equiv), SOCl₂, benzene, reflux, 3 h; (b) 3a, 3c, 6a–j (1.5 equiv), DIPEA (1.5 equiv), CH₂Cl₂, 0 °C–rt, 3 h, 65–98% (over two steps).

Table 2.

S1P₃-R agonist activity of compounds 7a–l

Compd	Amine	R1	R2	EC50α (nM)
				S1P3-R	S1P1-R
7a	3c	H	Cyclohexyl	>50,000	>50,000
7b	3b	Cyclopentyl	Cyclohexyl	1683	>50,000
7c	6a	4-Tetrahydropyranyl	Cyclohexyl	3536	>50,000
7d	6b	Cyclohexylmethyl	Cyclohexyl	3430	>50,000
7e	6c	2-Methylcyclohexyl	Cyclohexyl	337	33,100
7f	6d	2-Methylcyclohexyl	2-Methylcyclohexyl	202	>50,000
7g	6e	2-Methylcyclohexyl	Cycloheptyl	535	4700
7h	6f	2-Methylcyclohexyl	4-Tetrahydropyranyl	2820	3260
7i	6g	2-Methylcyclohexyl	Cyclopentyl	1740	>50,000
7j	6h	Cyclopentyl	Cycloheptyl	2753	34,150
7k	6i	Phenyl	Benzyl	7000	44,900
7l	6j	2-Pyridinyl	Benzyl	6563	>50,000

^αData are reported as mean of n = 3 determinations.

As previously observed, removal of a cyclohexyl group (7a) led to complete loss of activity. Similarly, exchanging a cyclohexyl for a cyclopentyl (7b), 4-tetrahydropyranyl (7c) or cyclohexylmethyl (7d) led to 16-, 33- and 32-fold loss of potency, respectively, confirming the importance of the bicyclohexyl amide system. Interestingly, the 2-methylcyclohexyl derivative 7e (CYM5558) was three-fold less potent and selective than 5f, probably due to conformational changes caused by the methyl group. Unexpectedly, when both cyclohexyls were substituted with a methyl group (7f, CYM5556) the potency decreased by only two-fold compared to 5f and the selectivity remained nearly similar. The N-(2-methylcyclohexyl)- N-cycloheptyl derivative 7g was almost two-fold less potent than 7e while the 4-tetrahydropyranyl 7h was slightly more potent than the cyclohexyl derivative 7c, although the selectivity of both compounds decreased considerably. Surprisingly, the cyclopentyl 7i was nearly equipotent to 7b. Furthermore, the N-cyclopentyl-N-cycloheptyl derivative 7j was 26-fold less potent than 5f. Not surprisingly, the installation of an aromatic substituent (e.g., 7k and 7l) led to drastic loss in potency.

Next, we explored the middle region B holding constant the bicyclohexyl amide region C as well as the cyclopropyl and phenyl groups from region A. The synthesis and biological activity of representative compounds 9a–c is depicted in Scheme 4. Interestingly the oxazole 9a and the furan 9b were inactive at the S1P₃-R; interestingly, 9b showed micromolar activity for the S1P₁-R. Fusing the phenyl ring of 5g into the isoxazole moiety led to 9c, which was inactive at the S1P₃-R. This modification highlights the isoxazole ring as an important binding motif of this chemotype.

Scheme 4.

Synthesis of 9a–c. Reagents and conditions: (i) (a) 8a–c (1 equiv), SOCl₂, benzene, reflux, 3 h; (b) 2a (1.5 equiv), DIPEA (1.5 equiv), CH₂Cl₂, 0 °C–rt, 3 h, 70– 98% (over two steps).

The functional activity of selected compounds was tested against the S1P_2,4-Rs subtypes (Table 3).¹⁷ Remarkably, all the compounds were highly selective against the S1P_2,4-Rs. Due to its selectivity profile, 5f was also tested against the S1P₅-R. Remarkably 5f was found exquisitely selective against S1P_1,2,4,5-R subtypes.

Table 3.

S1P_2,4,5-Rs selectivity counter screen of selected compounds

Compd	EC50α (nM)
	S1P3-R	S1P1-R	S1P2-R	S1P4-R	S1P5-R
3a	434	7870	>50,000	>50,000	NT
5b	959	22,250	>50,000	>50,000	NT
5f	105	33,300	>50,000	>50,000	>25,000
5g	103	662	>50,000	>50,000	NT
5h	323	859	>50,000	>50,000	NT
7f	202	>50,000	>50,000	>50,000	NT
7e	337	33,100	>50,000	>50,000	NT
7g	535	4703	>50,000	>50,000	NT

NT = not tested.

^αData are reported as mean of n = 3 determinations.

The selectivity profile of 5f was further investigated against the Ricerca panel of off target proteins including GPCRs, enzymes and ion channels at a concentration of 30 µM. The compound was found to be highly selective against the pool of therapeutically relevant targets tested (stimulation/inhibition <50% at 30 µM). A borderline inhibition (≥50%) was found for CYP450 2C19, CYP450 3A4, norepinephrine transporter, cannabinoid CB₁-R, histamine H₁-R, and sodium channel site two. Moderate activation (50%) was found only for nicotinic acetylcholine receptor.

The properties of 5f were further examined through a competitive binding experiment using [³³P]S1P as the orthosteric binding ligand and S1P as the competitive agonist in the control experiment. We found that S1P competes with [³³P]S1P for the S1P₃-R binding pocket. Remarkably, 5f does not compete with [³³P]S1P for the binding pocket of the S1P₃-R and binds in a manner that does not disturb the [³³P]S1P–S1P₃-R binding.²¹ Site-directed mutagenesis of S1P₁-R/S1P₃-R and docking studies performed on an S1P₃-homology model showed that 5f binds in an allosteric binding pocket.

Even though 5f is a highly selective S1P₃-R agonist with appropriate potency for further biological studies, its physicochemical properties remain suboptimal (tPSA = 41.9, ClogP = 3.9, solubility in PBS = 1.24 µM).

Based on the characterized ligand/binding pocket interactions, our research group will embark on a medicinal chemistry program aiming to further improve the potency and physicochemical properties while maintaining selectivity of 5f.

In summary, rational chemical modifications of the MLSMR hit 1a led to the development of S1P₃-R agonists 5f, 7e and 7f (CYM5541, CYM5558, CYM5556) exquisitely selective against the remaining S1P_1,2,4,5-R subtypes. Noteworthy, we have reported the discovery, design and synthesis of the first small molecule S1P₃-R allosteric agonist based on a N,N-dicyclohexyl-5-alkylisoxazole- 3-carboxamide chemotype, 5f, that is highly selective against a panel of therapeutically relevant targets. The studies herein described provide novel pharmacological tools to decipher the biological function and assess the therapeutic utility of the S1P₃-R. Further studies of our research program aiming to improve physicochemical properties and potency while maintaining selectivity will be communicated in due course.