Discovery of APD334: Design of a Clinical Stage Functional Antagonist of the Sphingosine-1-phosphate-1 Receptor

Abstract

APD334 was discovered as part of our internal effort to identify potent, centrally available, functional antagonists of the S1P₁ receptor for use as next generation therapeutics for treating multiple sclerosis (MS) and other autoimmune diseases. APD334 is a potent functional antagonist of S1P₁ and has a favorable PK/PD profile, producing robust lymphocyte lowering at relatively low plasma concentrations in several preclinical species. This new agent was efficacious in a mouse experimental autoimmune encephalomyelitis (EAE) model of MS and a rat collagen induced arthritis (CIA) model and was found to have appreciable central exposure.

Sphingosine-1-phosphate-1 (S1P₁) is a Class A G-protein coupled receptor (GPCR) expressed on lymphocytes, neural cells, and the endothelium.¹ This receptor belongs to a family of five GPCRs (S1P_1–5), which recognize the sphingolipid sphingosine-1-phosphate (S1P) and collectively perform a variety of functions, with S1P₁ regulating vascular development and lymphocyte trafficking.² Gilenya (1, fingolimod) (Figure 1) is an oral medication that targets the S1P receptors and is indicated for the treatment of relapsing forms of multiple sclerosis (MS).³In vivo, fingolimod is phosphorylated by sphingosine kinase to the (S)-monophosphate (2), which is an agonist of four of the five S1P receptors (S1P₁, S1P₃, S1P₄, and S1P₅).⁴ Sustained activation (agonism) of lymphocyte expressed S1P₁ results in internalization and proteasomal degradation of this receptor, thus resulting in “functional antagonism” of S1P₁.⁵ Reduced S1P₁ expression interrupts the normal trafficking of lymphocytes and retains autoreactive lymphocytes in the lymph nodes and thymus,⁶ thus preventing their accumulation in the brain and spinal cord of MS patients. Additionally, fingolimod readily passes through the blood–brain barrier and is believed to reduce neuroinflammation through direct action on S1P₁ expressed on astrocytes.^7,8 The unique biology associated with the S1P₁ receptor has encouraged others to identify alternative agents for treatment of autoimmune disease.⁹

Figure 1.

Structures of gilenya, monophosphate 2, 5-phenyl-1,2,4-oxadiazole 3, and APD334.

Monophosphate 2 is also a potent agonist of S1P_3, S1P₄, and S1P₅, but the relative contribution of these subtypes to efficacy in humans is unclear. S1P₅ is expressed on oligodendrocytes and its activation on human brain endothelial cells improves barrier integrity and reduces transendothelial monocyte migration.¹⁰ S1P₄ is expressed on hematopoietic, lymphocytic, and dendritic cells and the loss of this receptor has been shown to reduce T_H17 differentiation of T_H cells.¹¹ T_H17 cells are associated with MS pathogenesis, and populations of this cell type in MS patients are typically increased.¹² Reduced interaction with S1P₃ has been a consistent theme in the literature, as this receptor was believed to be responsible for the transient bradycardia observed in humans treated with fingolimod. It is now understood that in humans at least, this cardiovascular event more likely results from S1P₁-dependent G protein gated inwardly rectifying potassium (GIRK) channel activation on human atrial myocytes.^13,14 Regardless, S1P₃-sparing agonists may still be desirable, as nonselective S1P₁ agonists are associated with pro-fibrotic responses via S1P₂- and S1P₃-dependent signaling.^15,16

Identification of new agents with reduced elimination half-life has been sought after, as full lymphocyte recovery following fingolimod treatment requires >5 weeks and could be a complicating factor in the event of opportunistic infection.¹⁷ In addition, teratogenicity has been observed in rodents, and the long half-life in humans (8 days) requires contraception for two months after stopping treatment.¹⁸ A further challenge is identifying therapeutics with appreciable central exposure as this is believed to be necessary for success in treating MS.¹⁹ Lastly, new agents with reduced hepatotoxicity in humans would be highly beneficial. Herein we report the discovery of APD334 (4, Figure 1), a centrally available, selective, functional antagonist of the S1P₁ receptor currently in development for the treatment of autoimmune disease.

Our early efforts to identify a functional antagonist with a competitive pharmacodynamic profile focused on optimization of 5-phenyl-1,2,4-oxadiazoles (e.g., 3, Figure 1).^20,21 Despite discovery of several potent and selective analogues, we failed to identify a compound with good activity in the mouse lymphocyte lowering (LL) experiment at an acceptable plasma concentration (e.g., 3 LL IC₅₀ = 1.62 μM). Furthermore, 3 was undetectable in rat brain and CSF after 24 h intravenous infusion. To improve these parameters, we investigated a related series of 7-benzyloxy analogues (e.g., 7–21). These racemic mixtures were prepared from 5 utilizing the synthetic sequence outlined in Scheme 1.

Scheme 1.

Internalization of S1P₁ is preceded by activation of the receptor (agonism); thus, EC₅₀ values were generated in a homogeneous time-resolved fluorescence (HTRF) cyclase assay to assess engagement of the receptor (Table 1). Compound 7 was the first analogue prepared and possesses the CN/OCF₃ substitution pattern previously identified as a potency driver in the 1,2,4-oxadiazole structure–activity relationship (SAR) (e.g., 3, S1P₁ EC₅₀ = 1.1 nM). This aryl motif, however, was less well tolerated in the benzyloxy series (7, S1P₁ EC₅₀ = 16 nM). Instead, the analogous R¹ and R² disubstituted benzene was preferred (9, S1P₁ EC₅₀ = 1.7 nM). Subsequent SAR focused on the R² group and revealed a preference for larger alkyl groups. In particular, cyclopentyl and cyclohexyl were optimal in terms of potency (13–15). Several of these analogues were examined in an acute mouse lymphocyte lowering experiment measuring lymphocyte count and plasma concentration at 24 h following a single 1.0 mg/kg oral dose. This abbreviated pharmacokinetic/pharmacodynamics (PK/PD) analysis suggested 15 may offer a significant improvement in in vivo potency relative to 3 and prompted characterization of each enantiomer independently.

Table 1. Racemic 7-Benzyloxy SAR (human S1P₁, HTRF cAMP).

compd	R1	R2	R3	EC50, nM (%)a
7	CN		OCF3	16.0 (90%)
8	CF3		CF3	26.0 (98%)
9	CN	OCF3		1.7 (109%)
10	CN	OCH3		120 (136%)
11	CF3	CN		141 (104%)
12	CF3	Cl		159 (117%)
13	CN	cyclohexyl		0.087 (98%)
14	CF3	cyclohexyl		0.077 (96%)
15	CF3	cyclopentyl		0.073(104%)
16	CF3	cyclobutyl		0.45 (109%)
17	CF3	cyclopropyl		3.68 (118%)
18	CF3	n-propyl		0.49 (107%)
19	CF3	neopentyl		55.8 (112%)
20	CF3	iso-butyl		0.44 (103%)
21	CF3	methylcyclohexyl		4.58 (111%)

^aEC₅₀ values are the mean of three or more replicates. Value in parentheses is the % agonism relative to S1P. S1P EC₅₀ = 0.062 nM; 2 EC₅₀ = 0.006 nM (101%).

An enantiomeric separation was performed, and both isomers were tested in the S1P₁ HTRF assay. Each isomer was very active in vitro, with one offering a moderate potency advantage (i.e., hS1P₁ EC₅₀ = 0.04 nM, E_max = 102% versus 0.09 nM, E_max = 103%). A single crystal of the more potent enantiomer was grown from a saturated acetone/acetonitrile solution, and the absolute stereochemistry was determined by X-ray to be S. Correspondingly, an X-ray structure proof was also obtained for the R-isomer (Figure 2, 4, APD334). The R-isomer was ultimately selected as the development candidate in part due to superior stability in rat liver microsomes (t_1/2 > 60 min versus t_1/2 = 9 min for the S-isomer). Instability of the S-isomer was anticipated to complicate testing in rat disease models, PK experiments, and toxicity studies. Of note, both enantiomers were observed to be stable in human liver microsomes (t_1/2 > 60 min). In addition to improved rat microsomal stability, the R-isomer (APD334) forms a crystalline, high melting, nonhygroscopic, anhydrous l-arginine salt, which was regarded as a significant development advantage given the difficulty in identifying other suitable salt forms for either enantiomer. Interestingly, the crystalline free acid prepared for the X-ray determination was observed to possess similarly good physical properties; however, this crystal form was not orally bioavailable in rats when dosed as an aqueous suspension. This contrasts with the amorphous free acid form of 4, which demonstrated good bioavailability when dosed as a suspension (%F = 54). Importantly, the l-arginine salt achieved good exposure levels in SD rat when dosed orally as a solution in sterile water (C_max = 0.361 μg/mL, 1.0 mg/kg equivalent).

Figure 2.

Mercury 3.1 Development (Build RC5) ellipsoid rendition²² of X-ray crystal structure of (R)-2-(7-((4-cyclo-pentyl-3-(trifluoromethyl)benzyl)oxy)-1,2,3,4-tetrahydro-cyclopenta[b]indol-3-yl)acetic acid [4, APD334].

S1P₁ functional data for five species was obtained, as well as EC₅₀ values for the five human S1P subtypes (Table 2). S1P₁ activity was maintained in all preclinical species examined, and APD334 was devoid of any agonism or antagonism at human S1P₂ and S1P₃. Furthermore, moderate agonism at human S1P₄ and S1P₅ was observed but is reduced relative to S1P₁, both in terms of potency and efficacy. The extent to which 4 internalizes human S1P₁ was evaluated in CHO cells expressing HA tagged S1P₁, and 4 was found to have an IC₅₀ value of 1.88 nM.

Table 2. APD334 in Vitro S1P₁ Preclinical Species EC₅₀ Data and Human S1P_1-5 EC₅₀ Values.

preclinical species S1P1 EC50 cAMP, nM (Emax %)					human S1P1–5 EC50 β-arrestin, nM (Emax %)
human	mouse	rat	dog	monkey	S1P1	S1P2	S1P3	S1P4	S1P5
0.093 (103)	0.44 (95)	0.32 (99)	0.34 (87)	0.32 (95)	6.10 (110)	>10,000	>10,000	147 (63)	24.4 (73)

APD334 is stable in mouse, rat, dog, monkey, and human liver microsomes (t_1/2 > 60 min), and iv/po PK studies were performed in all preclinical species (Table 3). Consistent with the in vitro data, APD334 has a relatively low systemic clearance (<4% of hepatic blood flow) and high C_max across all species. In both dog and monkey a significant decrease in volume of distribution (V_ss) was observed relative to rodent. Oral bioavailability was in the range of 40–100%, and the terminal phase half-life varied from 6 h in monkey, to as long as 29 h in dog. Rat and monkey t_1/2 values for siponimod (another S1P₁ modulator currently in human trials) have been disclosed and are 6 and 19 h, respectively.²³

Table 3. APD334 Preclinical Species Pharmacokinetics and in Vivo Lymphocyte Lowering (LL) IC₅₀ Values^a.

species	dose iv/po (mg/kg)	Clsys (L/h/kg)b	%LBF	Vss(L/kg)b	Cmax (μg/mL)	tmax (h)	AUC0-inf (μg·h/mL)	t1/2 po (h)	%F	LL IC50 (μM)c
mouse	1/1	0.110	2.00	1.93	0.780	8.00	10.3	12.3	100	0.101
rat	1/1	0.145	3.45	1.61	0.218	7.33	3.79	9.64	54.0	0.051
dog	1/3	0.0138	0.746	0.609	3.71	7.33	164.0	28.8	73.2	0.058
monkey	1/3	0.0568	2.23	0.411	2.87	3.33	23.1	6.37	43.8	0.098

^aAmorphous APD334 (vide infra) was dosed as a suspension in 0.5% methyl cellulose.

^bCl_sys and V_ss were determined by intraveneous administration.

^cValues were generated from a full 24 h time course of drug effect on lymphocyte lowering (LL).

Peripheral lymphocyte counts were measured over a 24 h time course following oral administration of APD334, in order to establish a PK/PD relationship. Lymphocyte lowering (LL) IC₅₀ values were calculated, which illustrated that relatively low blood concentrations (0.051 to 0.101 μM) are required to produce substantial lymphopenia in all species (Table 3). Notably, the calculated LL IC₅₀ values reflect total plasma concentration wherein APD334 is highly protein bound (97.8% human, 98.0% rat).

Constant infusion CNS distribution studies in male Sprague–Dawley (SD) rats were performed and APD334 was observed to have improved central exposure relative to its predecessor 3. After a 48 h constant infusion (14.5 μg/h/kg), APD334 reached a brain concentration of 0.21 μM, which is approximately twice that of the plasma concentration (0.11 μM) and translates to a brain-to-plasma (B/P) ratio of 1.9. Drug plasma concentration at the 48 h time point is 2-fold higher than the estimated rat lymphocyte lowering IC₅₀ value (0.051 μM).

APD334 was evaluated in a murine experimental autoimmune encephalomyelitis (EAE) model.²⁴ This model mimics MS by inducing a demyelinating autoimmune response by utilizing myelin/oligodendrocyte glycoprotein (MOG_35–55) as autoantigen. Prophylactically, APD334 prevented the onset and severity of disease relative to vehicle up to day 25, at which time dosing was discontinued. All treatment arms went on to develop severe disease as shown by their clinical scores (Figure 3). Therapeutic administration of APD334 was also examined (Figure 4). Treatment began at day 18, by which time all animals had developed severe disease, and APD334 was administered out to day 37. APD334 reversed disease relative to vehicle and was similar to the efficacy observed with fingolimod (1). Furthermore, histological examination of spinal cord and brain collected on day 37 revealed significantly fewer inflammatory foci compared to vehicle control, and a marked decrease in number of infiltrating lymphocytes was also observed (Figure 5).

Figure 3.

Prophylactic mouse experimental autoimmune encephalomyelitis (EAE): female C57Bl/6 mice, vehicle = 0.5% methylcellulose (MC); q.d. dosing began on day 3 and continued to day 25.

Figure 4.

Therapeutic mouse experimental autoimmune encephalomyelitis (EAE): female C57Bl/6 mice, vehicle = 0.5% methylcellulose (MC); q.d. beginning on day 18 and the arrow indicates the start of drug treatment.

Figure 5.

Immunofluorescent staining of cerebellum of therapeutically treated mice at 37 days postimmunization. Pink color indicates CD45⁺ T-cells, green CD3⁺ T-cells, and blue 4′,6-diamidino-2-phenylindole fluorescent stain (DAPI): (a) 0.5% methylcellulose; (b) 1.0 mg/kg 1; (c) 3.0 mg/kg APD334.

APD334 was similarly efficacious in a collagen induced arthritis (CIA) model. Prophylactic oral administration in female Lewis rats resulted in a significant reduction in ankle diameters (mean ± SEM) on the day 17 end point following a daily oral dose of 1.0 and 3.0 mg/kg (0.2803 ± 0.00570 and 0.2690 ± 0.0031 in., versus 0.3352 ± 0.0045 for vehicle) and weight regain (178.6 ± 4.1 and 182.3 ± 1.9 g, versus 160.3 ± 2.3 for vehicle) and was similar to that observed in rats treated with 1.0 mg/kg of fingolimod (0.2694 ± 0.0039 in., 181.9 ± 2.2 g) or 0.075 mg/kg of methotrexate (0.2795 ± 0.0066 in., 181.7 ± 2.9 g). Improvement in histological parameters in the knees and ankles of CIA rats was also observed, suggesting that inhibiting lymphocyte entry into arthritic joints with APD334 treatment suppresses CIA in rodents.

CYP inhibition IC₅₀ values were determined, and APD334 has a very low inhibitory potential (i.e., 3A4, 2D6, 2C9, 1A2, 2C19; IC₅₀ >50 μM), thus ruling out the likelihood of a drug–drug interaction. Selectivity was further examined in a broad panel of receptors, ion channels, and enzymes. APD334 exhibited weak or no activity, in more than 100 different binding and functional assays. The most potent off target interaction observed was for the human adenosine A3 receptor, which afforded an IC₅₀ value of 3.3 μM in the A₃ binding assay. Activation of inward-rectifying potassium channel (IKAch) expressed on human atrial myocytes was subsequently examined in vitro. APD334 and S1P were tested and were found to have EC₅₀ values of 29.9 and 2.1 nM, respectively.

Genotoxic potential was evaluated in an Ames bacterial reverse mutation assay, and APD334 was nonmutagenic both with and without rat liver S9 activation. [³H]-Astemizole binding and cardiac repolarization (patch clamp) assays ruled out any hERG channel interaction (IC₅₀ > 30 μM, respectively), and APD334 was subsequently evaluated in telemeterized dog. No changes in pulse pressure, heart rate, body temperature, or ECG parameter were observed at dose levels up to 40 mg/kg dose.

CNS and respiratory safety studies were performed in SD rats. No neurobehavioral effects or changes in respiratory function were observed up to the top dose of 350 mg/kg in both studies. Finally, a 14 day repeat dose rat toxicity assay was conducted using 30, 100, 300, and 1000 mg/kg does groups. APD334 was well tolerated up to 300 mg, with mortality and CNS clinical signs observed at the 1000 mg/kg dose. In a separate study, APD334 was found to be dose proportional up to 300 mg/kg in SD rat with high levels of exposure at the top doses (C_max = 135.7 μM at 300 mg/kg, and C_max = 182.7 μM at 1000 mg/kg) suggesting a comfortable safety window versus the pharmacologically effective plasma concentration of 0.051 μM in rat. Lastly, no safety related findings with respect to liver function were observed in rat.

In summary, we identified APD334, a structurally novel, selective, functional antagonist of S1P₁, which met our preclinical safety and efficacy criteria. Importantly, APD334 achieves good central exposure following oral dosing and possesses a favorable pharmacokinetic profile in multiple preclinical species. APD334 is currently in development for a number of indications related to autoimmune disease. The outcome of these studies and plans for future development will be reported in due course.

Glossary

Abbreviations

PPTS

pyridinium p-toluenesulfonate

DIEA

diisopropylethylamine

DMF

dimethylformamide

DIAD

diisopropyl azodicarboxylate

THF

tetrahydrofuran

CNS

central nervous system

hERG

human ether-à-go-go-related gene

ECG

electrocardiogram

CHO

chinese hamster ovary

CYP

cytochrome P450

CSF

cerebrospinal fluid

Supporting Information Available

Synthesis and analytical data for compounds 7–21 and APD334. This material is available free of charge via the Internet at http://pubs.acs.org.

The authors declare no competing financial interest.