S1P is a circulating lipid that binds to five G protein-coupled receptors (GPCRs) termed S1P_1-5. S1P₁ selectively regulates physiological functions in the immune and cardiovascular systems, including immune cell trafficking (Rosen and Goetzl, 2005; Rosen et al., 2007) and maintaining endothelial integrity (Lee et al., 1999; Sanchez et al., 2003; Dudek et al., 2004; Sanna et al., 2006; Foss et al., 2007). Pharmacological studies with the sphingosine analog immunosuppressant prodrug FTY720 indicated that administration of FTY720 or S1P decreased lymphocyte counts in the mouse draining lymph node, resulting in lymphopenia (Mandala et al., 2002). Moreover, FTY720 was shown to be rapidly phosphorylated, with the phosphorylated species (FTY720-P) acting as a potent agonist on S1P_{1, 3-5} subtypes. Discovery of SEW2871 (Sanna et al., 2004), a selective agonist for S1P₁ that replicated the lymphopenic actions of FTY720, plus development of selective S1P₁ antagonists with in vivo activity, later demonstrated that S1P₁ was the primary mediator of lymphocyte sequestration in secondary lymphoid organs and lymphopenia.

Mechanistic insights into S1P₁-mediated lymphopenia came from combining genetics, pharmacological tools, and two-photon imaging (Wei et al., 2005; Sanna et al., 2006). The last of these has allowed studying real-time lymphocyte dynamics in the intact lymph node, where SEW2871 infusion reduces lymphocyte egress in the medulla, whereas a selective S1P₁ antagonist (W146) fully reverses agonist actions.

In the thymus, S1P₁ agonism enhances late thymocyte maturation yet inhibits subsequent thymocyte egress (Rosen et al., 2003; Alfonso et al., 2006; Weinreich and Hogquist, 2008). Consequently, S1P agonist administration reduces naive blood CD4⁺ and CD8⁺ thymocytes and B cells by nearly 90 and 70%, respectively.

Regulation of lymphocyte trafficking with S1P₁ agonists has opened the possibility for clinical modulation of lymphocyte dynamics, exemplified by the positive results reported in multiple sclerosis (MS) clinical trials with FTY720 (Hiestand et al., 2008). Although the mechanism of FTY720 in ameliorating MS symptoms is not clear, it may encompass multiple targets, both systemic (including S1P₁-mediated lymphopenia) and local, by actions on neurons (Kataoka et al., 2005; Balatoni et al., 2007; Miron et al., 2008). Vascular integrity of the blood-brain barrier (BBB) is likely to participate in the mechanism of FTY720's efficacy in MS (Foster et al., 2008). Thus, understanding the contribution of systemic and local effects of S1P₁ modulation would be relevant for developing efficacious therapies for autoimmune disease.

Previous studies aimed at dissecting the S1P-S1P₁ binding pocket have provided strong evidence for a model of S1P ligand interaction with three charged residues on S1P₁ (Parrill et al., 2000; Wang et al., 2001). Two residues, Arg¹²⁰ and Glu¹²¹, which are present on TM3 of all S1P receptors, interact with the phosphate headgroup and the ammonium moiety of S1P, respectively. The importance of Arg¹²⁰ and Glu¹²¹ interactions with S1P has been demonstrated by mutagenesis. Substitution of Arg¹²⁰ or Glu¹²¹ with alanine results in total loss of [³³P]S1P binding to S1P₁ and, consistent with the lack of binding, neither of the TM3 mutants internalizes after an S1P challenge or stimulates S1P-mediated guanosine 5′-O-(γ-[³⁵S]thio)triphosphate binding. Thus, the strong zwitterionic nature of S1P requires hydrophilic headgroup interactions on S1P₁ TM3 Arg and Glu residues to achieve high-affinity binding and receptor functions.

We have shown that besides S1P, other S1P₁ agonists, including AFD-R (an FTY720-P analog) and SEW2871, also require Arg¹²⁰ and Glu¹²¹ for full activation of intracellular pathways (Jo et al., 2005). It is noteworthy that although SEW2871 lacks structurally charged headgroups, its binding model to S1P₁ and activation of P42/p44 MAPK and AKT pathways is dependent on Arg¹²⁰ and Glu¹²¹, likely by replacing the salt-bridge polar interactions by ion-dipole interactions. SEW2871 has recently been shown to make additional hydrophobic interactions deep in S1P₁ TM5, suggesting that it overlaps both the hydrophilic pocket and a hydrophobic pocket on S1P₁ (Fujiwara et al., 2007).

Because the Arg and Glu residues are invariant across S1P receptors, developing efficacious ligands that bind solely in deep hydrophobic pockets could provide advantages for selectivity, mechanism-based toxicity, and the ability of these molecules to potentially traverse impermeant barriers, such as the BBB.

Here we report development and characterization of a potent and moderately water-soluble small molecule S1P₁-agonist (CYM-5442) that does not require interactions with either Arg120 or Glu121 of S1P₁ headgroup yet is fully active in vivo for inducing lymphopenia. It is noteworthy that CYM-5442 partitions significantly into brain tissue. Development of in vivo-active S1P₁ agonists that do not require headgroup interactions, such as CYM-5442, reveal a discrete novel hydrophobic pocket on S1P₁ to be further explored in therapeutics.

Materials and Methods

Reagents

S1P was purchased from BIOMOL Research Laboratories (Plymouth Meeting, PA). The selective S1P₁ agonist SEW2871 was purchased from Maybridge (Cornwall, UK). The selective S1P₁ antagonist W146 was from Avanti Polar Lipids (Alabaster, AL). The selective S1P₂ antagonist JTE-013 was from Cayman Medical Company. [³²P]Orthophosphate was purchased from PerkinElmer Life and Analytical Sciences. The MEK1 inhibitor U0126 was from Calbiochem (San Diego, CA).

Cell Lines

The generation of stable CHO-K1-S1P receptor cell clones and the conditions for CRE and nuclear factor of activated T cells reporter assays have been documented (Schurer et al., 2008). The Tango S1P₄ and Tango S1P₅ stable cell lines were obtained from Invitrogen and assayed according to Invitrogen's protocols with 1 μM S1P as positive control (Schurer et al., 2008). HEK293 cells stably expressing human S1P₁ tagged with C-terminal GFP (S1P₁-GFP) were a gift from Timothy Hla (University of Connecticut Health Center, Farmington, CT). S1P₁-GFP cells were grown as reported (Gonzalez-Cabrera et al., 2007). Parental CHO-K1 cells used in transient transfection experiments were purchased from the American Type Culture Collection (Manassas, VA).

Chemical Synthesis of CYM-5442

The schematic steps (i-iv) of CYM-5442 synthesis are indicated in Fig. 2 and are as follows.

i: 1-Hydroxy-2,3-dihydro-1H-indene-4-carbonitrile. To a stirred suspension of 1-oxo-2,3-dihydro-1H-indene-4-carbonitrile (1.0 equiv, 0.4 M) and silica gel (catalytic) in ethanol at 0°C was added NaBH₄ (0.33 equiv). The reaction was allowed to warm up to room temperature and stirred for 2 h. The solvent was removed under reduced pressure, and the product was purified by C.C. in hexane/EtOAc (5:5) to offer 1-hydroxy-2,3-dihydro-1H-indene-4-carbonitrile as white solid in 80% yield. ¹H NMR (300 MHz, CDCl₃): δ 7.62 (d, J = 7.5 Hz, 1H), 7.54 (d, J = 7.8 Hz, 1H), 7.33 (t, J = 7.8 Hz, 1H), 5.28 (t, J = 6.3 Hz, 1H), 3.28-3.18 (m, 1H), 3.02-2.92 (m, 1H), 2.63-2.52 (m, 1H), 2.06-1.99 (m, 1H).

ii: 1,N-Dihydroxy-indan-4-carboxamidine. To a stirred suspension of hydroxylamine hydrochloride (1.1 equiv) and Na₂CO₃ (1.1 equiv) in ethanol was added, in one portion, the benzonitrile prepared in the previous step (1 equiv). The mixture was refluxed for 6 h followed by addition of another portion of hydroxylamine hydrochloride (1.1 equiv) and Na₂CO₃ (1.1 equiv). The reaction was refluxed for an additional 6 h. The suspension was cooled to room temperature and filtered. The solid was washed with ethanol, and the filtrate was concentrated under reduced pressure. The amidoxime-crude was recrystallized from EtOAc/hexanes and used without further purification.

iii: 4-[5-(3,4-Diethoxy-phenyl)-[1,2,4]oxadiazol-3-yl]-indan-1-ol. In a microwave vial, a stirring solution of 3,4-diethoxybenzoic acid (1 equiv, 0.2M) in dimethylformamide was treated with hydroxybenzotriazole (1.3 equiv) and 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (1.3 equiv) at room temperature. The reaction was stirred for 20 min followed by addition, in a single portion, of amidoxime (prepared in previous step) (1.1 equiv). The reaction was stirred for additional 30 min at room temperature and then heated to 130°C for 30 min. The reaction was diluted using a saturated solution of NaCl and extracted with EtOAc (3×). The organic phase was dried over Na₂SO₄ anhydrous and concentrated under reduced pressure. The product was purified by C.C. using CH₂Cl₂:MeOH (9:1) to offer diaryloxadiazole as white solid in good yield. ¹H NMR (400 MHz, CDCl₃): δ 8.10 (d, J = 7.6, 1H), 7.78 (dd, J1 = 1.6 Hz, J2 = 8 Hz, 1H), 7.67 (d, J = 1.6 Hz, 1H), 7.56 (d, J = 7.6 Hz, 1H), 7.39 (t, J = 7.6 Hz, 1H), 6.97 (d, J = 8.0 Hz, 1H), 5.29 (t, J = 6.4 Hz, 1H), 4.19 (q, J = 7.2 Hz, 2H), 4.18 (q, J = 7.2 Hz, 2H), 3.51-4.43 (m, 1H), 3.22-3.14 (m, 1H), 2.59-2.51 (m, 1H), 2.04-1.97 (m, 1H), 1.5 (t, J = 7.2 Hz, 3H), 1.49 (t, J = 7.2, 3H); ¹³C NMR (100 MHz, CDCl₃): δ 175.2, 168.9, 152.8, 148.9, 146.6, 143.3, 128.9, 127.4, 127.0, 123.8, 122.2, 116.7, 112.7, 112.4, 76.2, 64.9, 64.8, 35.7, 31.5, 14.9, 14.8. Mass spectrometry (EI) m/z 367 (M⁺), high-resolution mass spectrometry (EI) for C₂₁H₂₂N₂O₄ (M⁺): calculated, 367.1652; found, 367.1653.

iv: 2-{4-[5-(3,4-Diethoxy-phenyl)-[1,2,4]oxadiazol-3-yl]-indan-1-ylamino}-ethanol. A solution of alcohol (1 equiv), at 0°C, was treated with SOCl₂ (1.1 equiv) and pyridine (1.1 equiv) in CH₂Cl₂. The reaction was stirred at room temperature for 2 h. The reaction was diluted with CH₂Cl₂ and washed with NaHCO₃ (2×). The organic phase was dried over sodium sulfate and concentrated under reduced pressure. The crude was dissolved in dimethylformamide and treated with the ethanolamine (2 equiv) and diisopropylethylamine (2.0 equiv). The reaction was stirred at 50°C for 48 h. The reaction was diluted with H₂O and the product extracted with EtOAc (3×). The product was purified by C.C. using CH₂Cl₂/MeOH (9:1) to offer aminoethanol CYM-5442 as white solid in 60% yield. ¹H NMR (500 MHz, CDCl₃): δ 8.12 (d, J = 7.5 Hz, 1H), 7.79 (dd, J1 = 2.0 Hz J2 = 8.5 Hz, 1H), 7.68 (d, J = 2.0 Hz, 1H), 7.63 (d, J = 7.5 Hz, 1H), 7.40-7.37 (m, 1H), 4.49-4.47 (m, 1H), 4.23-4.16 (m, 4H), 3.78-3.70 (m, 1H), 3.53-3.46 (m, 1H), 3.29-3.22 (m, 1H), 2.96-2.94 (m, 4H), 2.56-2.50 (m, 1H), 2.09-2.03 (m, 1H), 1.52-1.49 (m, 6H); ¹³C NMR (125 MHz, CDCl₃): δ 175.03, 168.66, 152.71, 148.87, 143.76, 128.71, 127.11, 123.89, 122.04, 116.67, 112.63, 112.49, 104.66, 64.84, 64.61, 62.70, 60.34, 47.98, 31.90, 29.69, 14.72, 14.64. Mass spectrometry (EI) m/z 410 (M⁺), high-resolution mass spectrometry (EI) for C₂₃H₂₇N₃O₄ (M⁺): calculated, 410.2074; found, 410.2077.

Evaluation of CYM-5442 Agonist Properties

Details of the screening CRE reporter assays can be found in Schurer et al., (2008). For the CRE reporter assay studies, forskolin was included at a 2 μM final concentration before addition of the agonists. Agonist-mediated inhibition of forskolin induced cAMP accumulation was read on an Envision fluorescent plate reader (384-well format) after 2-h agonist incubation and with SEW2871 as the positive control. Agonist-induced S1P₁-GFP internalization and ubiquitination assays have been documented previously (Gonzalez-Cabrera et al., 2007). Description of agonist-mediated S1P₁-GFP phosphorylation experiments can be found in Schurer et al., (2008).

ELISA Determination of p42/p44 MAPK Activity

CYM-5442 and S1P-mediated p42/p44 MAPK activation was measured using an ELISA kit (Cell Signaling Technologies, Danvers, MA) in CHO-K1 cells transiently transfected with either the WT S1P₁ receptor cDNA or the single-point S1P₁ mutant (R¹²⁰A or E¹²¹A) cDNAs (gifts from Gabor J. Tigyi, University of Tennessee Health Science Center, Memphis, TN). For transfections, cells were plated on 10-cm dishes at 80% confluence and transfected with 12 μg of each plasmid using Fugene HD (Roche Applied Science, Indianapolis, IN) in Opti-MEM (Invitrogen). After 16 h, each pool-transfected dish was split into a six-well plate and allowed to incubate for additional 24 h. The cells were next incubated for 4 h in serum-free Dulbecco's modified Eagle's medium before addition of the agonist. In the antagonist (W146)/MEK1 inhibitor (U0126) experiments, W146 or U0126 was incubated for 30 min (U0126) or 1 h (W146), respectively, at 10 μM before agonist treatment. Cells were then stimulated for 5 min (determined empirically to be maximal for both agonists) with increasing concentrations of CYM-5442 or S1P and activation of p42/p44 MAPK phosphorylation was assayed according to manufacturer instructions. For each condition (WT, R¹²⁰A, or E¹²¹A), both CYM-5442 and S1P agonists were tested in parallel, in the presence and absence of W146. In addition, for each of the conditions tested, the concentration response curves for agonist-mediated activation of p42/p44 MAPK phosphorylation were plotted as a percentage of the agonist eliciting the maximal response, and the potency (EC₅₀), maximal response (E_max), and goodness of fit (R²) values were determined using Prism (Graphpad Software, San Diego, CA).

Pharmacokinetic Studies

Pharmacokinetics of CYM-5442 was assessed in Sprague-Dawley rats. The compound was formulated at 1 mg/ml [10:10:80 dimethyl sulfoxide/Tween 80/water (v/v/v)] and dosed at 1 mg/kg intravenous (i.v.) into the jugular or 2 mg/kg by oral gavage (p.o.). Blood was obtained at t = 5, 15, or 30 min or 1, 2, 4, 6, and 8 h into EDTA-containing tubes, and plasma was generated by standard centrifugation methods. Separate studies to evaluate brain exposure were done in C57BL6 mice in which CYM-5442 was dosed at 10 mg/kg by intraperitoneal (i.p.) route. Blood and brains were obtained at t = 2 h to assess the amount of compound in the target organ. All procedures and handling were according to standard operating procedures approved by the Institutional Animal Care and Use Committee (IACUC).

To assess in vivo pharmacokinetic parameters an LC-tandem mass spectrometry bioanalytical method was developed in which 25 μl of plasma was treated with 125 μl of acetonitrile containing an internal standard in a Multiscreen Solvinter 0.45-μm low binding polytetrafluoroethylene hydrophilic filter plate (Millipore, Billerica, MA) and allowed to shake at room temperature for 5 min. The plate was then centrifuged for 5 min at 4000 rpm in a tabletop centrifuge, and the filtrate was collected in a polypropylene capture plate. The filtrate (10 μl) was injected using an HPLC (1200; Agilent Technologies, Santa Clara, CA) equipped with a Betasil C18 HPLC column 5 μ (50 × 2.1 mm; Thermo Fisher Scientific, Waltham, MA). Mobile phase A was water with 0.1% formic acid. Mobile phase B was acetonitrile with 0.1% formic acid. Flow rate was 375 μl/min using a gradient of 90% A/10% B from 0 to 0.5 min, ramped to 5% A/95% B at 2 min, held at 5% A/95% B until 3.0 min, ramped to 90% A/10% B at 4 min, and held at 90% A/10% B until 7 min. An LC-tandem mass spectrometry system (API 4000; Applied Biosystems/MDS Sciex (Foster City, CA), equipped with a turbo ion spray source, was used for all analytical measurements. Instrument settings had curtain gas set to 10, gas 1 and gas 2 set to 45, ionization energy set to 5500 V, drying gas temperature 550°C, resolution was set to unit, and dwell time was set to 100 ms. A positive ion MRM method was developed. CYM-5442 was quantitated from 2 to 1000 ng/ml. Peak areas of the m/z 410.1→193.1 product ion of CYM-5442 (dicluster potential = 45, collision energy = 30) were measured against the peak areas of the internal standard (sunitinib) m/z 399→283 product ion (DP = 86, CE = 41). Data were fit using WinNonLin (Pharsight Corporation, Mountain View, CA).

Similar conditions were used to determine brain levels of CYM-5442 except that the brain samples were frozen upon collection. When analyzed, the frozen samples were weighed and acetonitrile was added [10× (w/v)]. The samples were sonicated to extract the compound from the brain matrix and then filtered as described above. Brain samples were analyzed against a standard curve generated in blank brain matrix.

Measurement of CYM-5442-Mediated Lymphopenia

Male C57BL6 mice weighing 30 g were used for all experiments. All animal studies were approved by the IACUC. Animals were injected intraperitoneally with a volume of 300 μl with the indicated dose of CYM-5442 or vehicle. For W146 experiments, mice were administered with 20 mg/kg W146 or vehicle for 30 min before CYM-5442 administration. Vehicle consisted of sterile water (CYM-5442) or 10% dimethyl sulfoxide and 25% Tween 20 in sterile water (W146). After incubation at the indicated times, animals were euthanized, and blood was collected into tubes containing EDTA. White blood cell count and lymphocyte counts were obtained using an automated veterinary Hemoanalyzer (Hospitex Diagnostics, Webster, TX) by reading 30-μl blood samples. The remaining blood was used for determination of drug serum levels (see below), or to determine, by FACS, the number of B cells (using a fluorescein isothiocyanate-conjugated B220 antibody marker; 1:100 dilution; BD Diagnostics, Franklin Lakes, NJ) and T cells (using Pacific Blue CD4⁺ and Per-CP-Cy5.5 CD8⁺ antibodies; both at 1:100; BD Diagnostics). FACS analysis was done in FlowJo (Tree Star, Ashland, OR).

CYM-5442 and W146 Serum Concentration. Blood samples used in lymphopenia studies were further processed for measuring serum CYM-5442 and W146 concentrations. Samples (100 μl of plasma) were extracted with 400 μl of ice-cold methanol (stored at -20°C), vortexed for 1 min, allowed to sit at 4°C for 30 min, and centrifuged at 16,400 rpm for 5 min. The supernatant was evaporated to a volume near dryness, and methanol was added to bring the final volume to 50 μl. For calibration standards, clean plasma was spiked with increasing concentrations of CYM-5442 or W146 and processed the same way as the samples. Samples were analyzed on an Agilent LC-tandem mass spectrometry system using an 1100 LC stack-mated with a 6410 triple quadrupole mass spectrometer. For CYM-5442 quantification, the transition of m/z 410 -> 349.2 was monitored, and for W146, the transition of m/z 343→138 was monitored. The column used was an Agilent Zorbax SB-C18 (2.1 × 75 mm). Flow rate was 250 μl/min. Mobile phase A = water/0.1% formic acid; mobile phase B = acetonitrile/0.1% formic acid. Gradient: 15% BatT = 0 ramped to 98% B at T = 8; 5 μl injected.

Results

CYM-5442 Is a Potent S1P₁-Selective Agonist. CYM-5442 (Fig. 1) is a chemically optimized version of an original hit (CYM-5181) from a screen aimed to discover novel S1P receptor agonists (Schurer at al., 2008). The chemical synthesis of CYM-5442 is depicted in Fig. 2. Even though CYM-5442 contains the privileged oxadiazole-based scaffold reported to fit S1P receptors (Schurer et al., 2008), CYM-5442 is selective for S1P₁, as measured by high-throughout agonist-antagonist formats across S1P receptors (Table 1). In high-throughput agonist format with SEW2871 as the positive control, CYM-5442 inhibited forskolin-stimulated CRE transcription in a concentration-dependent manner (Fig. 3), being a full agonist compared with SEW2871 but of higher potency (20- and 100-fold) relative to CYM-5181 and SEW2871, respectively.

Further characterization of the agonist properties of CYM-5442 was done using a receptor immunoprecipitation protocol previously validated for studying trafficking and fate of S1P₁-GFP during agonist stimulation (Gonzalez-Cabrera et al., 2007). Using this protocol and HEK293 cells stably expressing S1P₁ fused to GFP on the carboxyl terminus, we probed CYM-5442 for stimulating three agonist-S1P₁ activated steps: receptor phosphorylation, receptor internalization and receptor-ubiquitin recruitment. Figure 4A indicates that incubation of 500 nM CYM-5442 with [³²P]orthophosphate-labeled cells stimulated S1P₁ phosphorylation in a time-dependent manner, similar to that obtained with 500 nM S1P (at 30 min). CYM-5442 led to rapid S1P₁ phosphorylation that was sustained throughout the analysis. To confirm the involvement of S1P₁ for CYM-5442 phosphorylation, the S1P₁ selective antagonist W146 was used. Preincubation with 10 μM of W146 for 30 min before CYM-5442 treatment completely abolished CYM-5442-mediated S1P₁ phosphorylation, whereas antagonist alone had no effect on the response. In addition, silver staining of the same gel used to measure receptor phosphorylation indicated that the differences in S1P₁ phosphorylation across conditions were not due to differences in immunoprecipitated S1P₁ protein loading.

Membrane-associated GPCRs are usually internalized into cytosolic vesicles upon agonist stimulation. We and others have shown that S1P and S1P agonist analogs internalize S1P₁-GFP from the plasma membrane to cytoplasmic vesicles (Liu et al., 1999; Gonzalez-Cabrera et al., 2007; Oo et al., 2007). Figure 4B shows that incubation of S1P₁-GFP cells with 500 nM CYM-5442 stimulated the internalization of S1P₁-GFP receptor from a membrane-associated localization to an intracellular, multivesicular compartment. Similar internalization pattern was obtained by 0.5 μM S1P incubation (not shown). Internalization of S1P₁-GFP by CYM-5442 was completely blocked by preincubation with 10 μM W146, and, consistent with the phosphorylation data, W146 alone had no effect on internalization.

S1P₁ agonism with some ligands, termed supraphysiological because of their ability to alter receptor signaling reserve, results in degradation of S1P₁-GFP in lysosomes and proteasomes, and the magnitude of agonist-dependent receptor ubiquitination has been reported to influence receptor fate (Gonzalez-Cabrera et al., 2007; Oo et al., 2007). Thus, agonist recruitment of ubiquitin to S1P₁ can be used as a measure of ligand receptor activation. The ability of AFD-R, a highly efficient ubiquitinator (Gonzalez-Cabrera et al., 2007; Oo et al., 2007), and CYM-5442 to recruit ubiquitin onto immunoprecipitated S1P₁-GFP is shown in Fig. 4C. CYM-5442 stimulated receptor ubiquitination (UB-S1P₁-GFP) to AFD-R levels. Ubiquitination by the agonists resulted in high molecular mass smears (approximately from 97 to 180 kDa) of ubiquitin-receptor complexes running immediately above the immunoprecipitated receptor. Preincubation with the selective antagonist W146 was shown to nearly abolish CYM-5442-stimulated S1P₁ ubiquitination. Overall, the in vitro data indicate that CYM-5442 is a potent and selective S1P₁ agonist.

CYM-5442 Does Not Require Ionic S1P-receptor Headgroup Interactions for Activating S1P₁-Mediated p42/p44 MAPK Phosphorylation. S1P and synthetic S1P₁ agonists have been shown to activate the mitogenic p42/p44-MAPK pathway across many cell types (Sorensen et al., 2003; Jo et al., 2005; Ahmad et al., 2006). The ability of CYM-5442 and S1P to activate p42/p44-MAPK was studied by ELISA, using lysates of CHO-K1 cells transiently expressing either the human wild-type S1P₁ (WT) or the two separate single-point S1P₁ mutants (R¹²⁰A and E¹²¹A), known to abolish binding and activity of S1P on S1P₁. Time course studies demonstrated that maximal p42/p44-MAPK activity was achieved at 5 min in CHO-K1 cells expressing S1P₁ (Supplemental Fig. 1A). The time course also showed that treatment with 500 nM S1P for 5 min did not lead to measurable p42/p44 MAPK activity in mock-transfected CHO-K1 cells. S1P and CYM-5442 stimulated P42/P44 MAPK phosphorylation was completely blocked by preincubation with 10 μM U0126, a MEK1 inhibitor (Supplemental Fig. 1B).

Using the 5-min incubation time, we sought to determine the requirement of S1P₁ Arg¹²⁰ and Glu¹²¹ residues for CYM-5442 and/or S1P activation of P42/P44 MAPK. For these experiments, concentration response curves with the agonists were performed on cells expressing each of the single-point mutants (R¹²⁰A and E¹²¹A), and these responses were compared with those in the WT receptor, in both the presence and the absence of the selective S1P₁ antagonist W146. Figure 5 shows the plots of the average concentration responses in the three transfectants, whereas the potency (EC₅₀), maximal response (E_max), and goodness of fit (R²) value estimates derived from curve fitting are included in Table 2. In WT cells, both S1P and CYM-5442 led to the concentration-dependent activation of P42/P44 MAPK phosphorylation, with EC₅₀ values of 2.3 and 46 nM, respectively. In WT transiently transfected CHO-K1 cells, S1P was a full agonist (E_max of 1.0 or 100%), whereas CYM-5442 was a partial agonist (E_max of 0.7 or 70% of S1P). Incubation of 10 μM W146 before addition of S1P resulted in a significant rightward shift (60-fold, EC₅₀ of 140 nM) in potency of S1P versus S1P alone (Fig. 5). On the other hand, preincubation with W146 in WT cells led to the complete inhibition of CYM-5442-mediated p42/p44 MAPK phosphorylation (Fig. 5). These results indicate that W146 is a competitive antagonist with S1P and a noncompetitive antagonist against CYM-5442.

As expected, substitution of S1P₁ Arg¹²⁰ for alanine (R¹²⁰A) resulted in a near loss of p42/p44 MAPK activity for S1P. Curve-fitting the S1P data in this mutant indicated relatively poor fits (R² < 0.4). It is noteworthy that the R¹²⁰A mutant was still able to maintain p42/p44 MAPK activity when incubated with CYM-5442; as a result, the data were plotted relative to CYM-5442 maximum. The EC₅₀ for CYM-5442 in the R¹²⁰A mutant was determined to be not significantly different from that of WT CYM-5442 cells (R¹²⁰A EC₅₀, 67 nM; WT EC₅₀, 46 nM). Unlike WT cells, preincubation of 10 μM W146 in mutant R¹²⁰A cells did not significantly alter the p42/p44 MAPK activity by CYM-5442, and led to a modest (3-fold) rightward shift in EC₅₀ (67 nM), suggesting that in the absence of headgroup localization, W146 is a weak competitive antagonist of the hydrophobic site.

Consistent with the notion that S1P makes a functional headgroup interaction with Glu¹²¹, S1P did not lead to significant p42/p44 MAPK activation in E¹²¹A S1P₁ mutant cells. On the other hand, activation of p42/p44 MAPK by CYM-5442 in E¹²¹A S1P₁ cells was concentration dependent, with a mean EC₅₀ value of 134 nM. W146 preincubation led to a 10-fold rightward shift in potency of CYM-5442 for activating p42/p44 MAPK phosphorylation in E¹²¹A transfected cells. These results indicate that for S1P₁-dependent p42/p44 MAPK activation, CYM-5442 does not require the Arg¹²⁰ and Glu¹²¹ S1P₁ residues that make up functional S1P-S1P₁ headgroup interactions.

CYM-5442 Pharmacokinetics. We next evaluated the pharmacokinetics of CYM-5442 in rats. Measures included are depicted in Table 3. Overall, CYM-5442 was modestly orally bioavailable (F = 26%). Routes of delivery influenced the half-lives (t_½) of 50 min (intravenous) and 3 h (oral), which supported its use in vivo. It is noteworthy that in mice, CYM-5442 administration was highly central nervous system-penetrant. A dose of 10 mg/kg i.p. for 2 h resulted in a 13.7 ± 2.9 μM concentrations in brain compared with 1.08 ± 0.3 μM in plasma. This brain-to-plasma ratio of approximately 13:1 suggests that CYM-5442 may be a useful tool for studying the roles of S1P₁ in the central nervous system.

CYM-5442 Induces and Maintains Lymphopenia in Mice through S1P₁ Activation. Administration of S1P₁ agonists such as SEW2871 or FTY720-P induces rapid and reversible lymphopenia in mice (Mandala et al., 2002; Sanna et al., 2004). To determine whether CYM-5442 could lead to the induction of lymphopenia, we treated mice with a dose of 10 mg/kg i.p. and compared whole-blood white blood cell count and the number of circulating B and T cells against vehicle-treated mice. A 5-h treatment protocol was chosen based on the pharmacokinetic data (Table 3). White blood cell counts in CYM-5442 animals were decreased by 64% versus vehicle at 5 h (Table 4). FACS analyses of whole blood from treated animals indicated that CYM-5442 decreased B220⁺ B cells by 63% compared with vehicle, whereas CD4⁺ and CD8⁺ T-cell counts were decreased by 83% and 84% of vehicle, respectively. These data indicate that a single dose of CYM-5442 induces acute lymphopenia in mice.

To ascertain whether the effects of CYM-5442 were dose-dependent, we employed a 5-h treatment and a dosing range of 0.3 to 10 mg/kg. The dose-response of CYM-5442 for inhibiting CD4⁺ and CD8⁺ cell populations is shown in Fig. 6A, and a representative example of the FACS scatter plots in individual mice is shown in Fig. 6B. The data indicates that CYM-5442 decreased CD4⁺ and CD8⁺ cell counts in a dose-dependent manner, attaining near-maximal effects on inhibition of these cell populations at serum CYM-5442 levels that ranged between 50 and 100 nM. Overall, the CYM-5442 mediated inhibition of B cell and T cell number had an estimated ED₅₀ of 0.5, 2.0, and 1.0 mg/kg for CD4⁺, CD8⁺, and B220⁺, respectively (Fig. 6C).

Next, we measured the time course for CYM-5442 mediated lymphopenia and whether CYM-5442 effects could be reversed upon agonist clearance from plasma. These studies were performed employing a 16-h window, and both B- and T-cell counts were determined in conjunction with measures of serum CYM-5442 drug levels from the same animals (Fig. 6D). A representative example of individual animal FACS scatter plots is shown in Fig. 6E. With a fixed dose of 10 mg/kg, shown in Fig. 5A to be maximal for CYM-5442 induction of B-cell and T-cell lymphopenia, the data indicates that CYM-5442 CD4⁺ and CD8⁺ counts dropped significantly during the first 3 h of administration, and seemed to be maintained at these low levels from 3 h to 5 h. After 5 h, the B- and T-cell counts began to reverse toward basal levels; consistent with the time in which CYM-5442 began to disappear from plasma. Consistent with the dose-response data on Fig. 5A, CYM-5442 induction and maintenance of lymphopenia in the time course study occurred at serum levels of CYM-5442 ranging between 50 and 100 nM. The direct relationship between B- and T-cell recovery and loss of plasma serum CYM-5442 content thus suggested that CYM-5442-mediated lymphopenia can be reversed upon CYM-5442 clearance from the organism.

The minimal signal for lymphocyte sequestration and induction of lymphopenia by S1P₁ agonists depends on activation of S1P₁. To ascertain whether the in vivo effects of CYM-5442 were due to S1P₁ receptor activation, we used the selective S1P₁ antagonist, W146, shown here to block all of the measures of CYM-5442 activation in vitro. For antagonist studies, we used a 5-h window and a dose of CYM-5442 of 2 mg/kg, which is close to the ED₅₀ for CYM-5442 induction of B- and T-cell lymphopenia (Fig. 6, A and C). A competing dose of W146 of 20 mg/kg was first administered to mice and allowed to equilibrate for 30 min before CYM-5442 administration. Table 5 shows that at near ED₅₀, CYM-5442 led to the reduction of CD4⁺, CD8⁺, and B220⁺ cell counts by approximately 50% of vehicle, whereas W146 was able to partially recover CYM-5442 effects on B-220⁺, CD4⁺, and CD8⁺ cell counts. It is noteworthy that neither administration of W146 alone nor of vehicle had measurable effects on cell counts (data not shown). This indicates that CYM-5442-mediated lymphopenia is dependent on S1P₁ receptor activation.

Discussion

Using a chemical approach, we have characterized a small molecule compound, CYM-5442, as a selective and efficacious S1P₁ agonist for inducing and maintaining lymphopenia in vivo. It is of interest that the pharmacological properties of CYM-5442 for activating S1P₁-dependent pathways are distinct from S1P, in that the ligand has no groups capable of mimicking the headgroups of S1P. The data strongly suggest that CYM-5442 interacts with S1P₁ in a binding pocket separate from key S1P₁ residues Arg¹²⁰ and Glu¹²¹ essential for high-affinity S1P binding and receptor activation (Parrill et al., 2000; Wang et al., 2001; Jo et al., 2005). Alternative residues of the putative hydrophobic pocket interacting with oxadiazole series of compounds, such as the original S1P₁ agonist screening hit CYM-5181, have been recently published by Schurer et al., (2008). Because CYM-5442 has an EC₅₀ within a factor of 3 of CYM-5181, the binding free energies are within experimental error of each other, and thus the docking of the ligand CYM-5442 into the receptor pocket is the same as the model shown for CYM-5181 (Fig. 6; Schurer et al., 2008) and distinct from the head-group constrained interactions of S1P (Supplemental Fig. 5; Schurer et al., 2008). These differences between S1P and CYM-5442 find expression quantitatively by the retention of CYM-5442 activation of the E¹²¹A and R¹²⁰A receptor mutants that fail to respond to S1P, as well as by the different mode of inhibition for CYM-5442 compared with S1P using the S1P₁ antagonist W146. Mutagenesis of either Arg¹²⁰ or Glu¹²¹ residues, present on the extracellular face of TM3, to alanine has been shown to abolish S1P-S1P₁ binding and subsequent receptor activation (Parrill et al., 2000; Wang et al., 2001; Jo et al., 2005), due to the inability of the charged phosphate and ammonium headgroups in S1P to make polar interactions with neutral alanine on the R¹²⁰A and E¹²¹A mutants. Whereas S1P stimulated MEK1-dependent p42/p44 MAPK phosphorylation in WT cells in a concentration-dependent manner, and with a potency consistent with literature values (Mandala et al., 2002; Sanna et al., 2004), neither of the mutant receptors was able to elicit significant signal when treated with S1P. On the other hand, the same single-point mutants shown to be unresponsive to S1P, were still able to fully activate MEK1-dependent p42/p44 MAPK activity when treated with CYM-5442, with potency values found to be not significantly different between WT and R¹²⁰A receptors, although the E¹²¹A receptors displayed slightly lower potency than WT. These data are consistent with different receptor binding requirements for the amphiphilic orthosteric agonist, S1P, and the hydrophobic selective agonist, CYM-5442.

W146 has proved to be a reliable chiral-selective S1P₁ antagonist in vitro and in vivo (Sanna et al., 2006; Gonzalez-Cabrera et al., 2007; Yoon et al., 2008). As mentioned above, W146 antagonism of p42/p44 MAPK activity in these studies revealed two modes of action, competitive for S1P and noncompetitive for CYM-5442. W146 antagonism of S1P₁ signaling is therefore independent of whether S1P₁ activation is occurring through the orthosteric site or an adjacent hydrophobic site. Evidence for the spatial proximity of the orthosteric pocket to the hydrophobic one is provided by the competitive antagonist shift of W146 in the R¹²⁰A or E¹²¹A mutated receptors in response to activation by CYM-5442. Because W146 is an amino-phosphonate, capable of making Arg and Glu headgroup interactions, these data suggest that W146 is no longer constrained in the single-point mutants thus allowing more flexible binding within the hydrophobic pocket. Consistent with these data, noncompetitive inhibition of CYM-5442 function was also observed with W146 in U2OS cells coupled to Tango S1P₁-bla, whereas W146 in the same assay was a competitive inhibitor against S1P (not shown). Other S1P₁ selective agonists with similar structures to CYM-5442 also display noncompetitive W146 inhibition (not shown), suggesting that diaryl-oxadiazole agonists such as CYM-5442 interact with a common hydrophobic pocket on S1P₁ (Schurer et al., 2008).

The in vivo data with CYM-5442 demonstrates its usefulness for inducing and maintaining S1P₁-dependent lymphopenia, with a minimal serum concentration for achieving maximal lymphopenia of approximately 50 nM. Because systemic S1P₁ antagonism with W146 reversed CYM-5442-induced lymphopenia, and W146 alone had no effect on B- and T-cell counts, the data indicate that S1P₁ agonist signals are minimal and obligatory for lymphopenia induction. These data provide additional evidence against the functional antagonism hypothesis for S1P₁-mediated immunomodulation (Matloubian et al., 2004; Schwab and Cyster, 2007).

All effects of CYM-5442, both in vitro and in vivo, are reversed or abolished by the selective S1P₁ receptor antagonist W146. These data restrict the pharmacological effects of CYM-5442 to this single receptor subtype. Although CYM-5442 is an enantiomeric mixture of 2 isomers, this mixture does not seem to contribute to the pharmacology or selectivity based upon our selective antagonist studies. Work is still ongoing to resolve the enantiomers. Differential crystallization using chiral salt forms such as l-tartaric acid have thus far been unsuccessful. Separation using chiral columns by HPLC is currently ongoing with no positive results yet and an enantiospecific synthesis may be formally required.

CYM-5442 possesses additional properties that make it a useful S1P₁ chemical agonist. CYM-5442 was found to be 10,000-fold selective for S1P₁ over S1P₅. Besides selectivity, CYM-5442 is approximately 10- to 50-fold more potent than SEW2871 in vitro, depending on which agonist format was used, and induces lymphopenia at approximately 5- to 10-fold lower doses than SEW2871. Favorable pharmacokinetics and moderate water solubility would make CYM-5442 a useful tool for studying S1P₁ function in tissues in which drug formulation or penetrance may present challenges (i.e., lung and central nervous system). The levels of CYM-5442 in brain tissue, with a brain-plasma ratio of 13:1 after a bolus dose in mice, are notable. The beneficial effects of FTY720 in ameliorating MS symptoms have been associated, in part, with the inhibition of BBB permeability occurring during disease progression. Central nervous system-penetrant and efficacious nonprodrug S1P₁ agonists such as CYM-5442, that are also antagonist reversible in the periphery, may be good proof-of-concept chemical tools for investigating the contributions of S1P₁ to MS therapy. These allow the contributions of BBB integrity, lymphocyte sequestration, and glial and neuronal S1P₁ function to be assessed. Indeed, selective S1P₁ agonists and antagonists have shown to inversely modulate the integrity of endothelial barriers in vivo (Sanna et al., 2006; Foss et al., 2007). We have data indicating that endothelial integrity is tonically regulated by the S1P-S1P₁ axis (Sanna et al., 2006). This servo mechanism results in agonist enhancement of barrier function by inhibiting leakage induced by proangiogenic stimuli, including vascular endothelial growth factor and thrombin. Antagonism alone leads to barrier-opening effects. These actions have been well documented in lung and skin vasculature, and having a matched agonist-antagonist pair such as CYM-5442 and W146 would be optimal for further studying the hypothetical endothelial S1P-S1P₁ rheostat.

Highly selective and potent chemical tools in general have characteristically slow off-rates and have been of great interest in recent crystallographic studies (Cherezov et al., 2007; Rosenbaum et al., 2007) to define both orthosteric and spatially distinct hydrophobic binding pockets within this family of class A GPCRs and the minimalist changes in receptor structure necessary for full pharmacological efficacy. Such chemical tools, when deeply characterized and broadly available to the field (Rosen et al., 2008), can affect the progress of mechanistic understanding in physiology, pathology, structural biology, and, ultimately, therapeutics.