Sage Journals: Discover world-class research

Abstract

Sphingosine-1-phosphate (S1P) is a bioactive metabolite with pleiotropic effects on multiple cellular processes in health and disease. Responses elicited by S1P are a result of binding to five specific G-protein–coupled receptors. We have developed multiple assays to systematically study the downstream signaling of these receptors, including early events such as direct receptor activation (GTPγS) as well as more distal events such as S1P₁ receptor degradation. Employing such assays, we have characterized and compared multiple S1P₁ agonists that are in clinical development including FTY720, BAF312, CS-0777, and other molecules from the S1P₁ patent literature. Our parallel assessment has allowed us to compare their potency against S1P₁, their selectivity against the four other S1P receptors, as well as species cross-reactivity. We note that all of the compounds studied signal in an identical manner through S1P₁, leading to receptor degradation.

Keywords

sphingosine-1-phosphate receptors functional antagonist GTPγS β-arrestin receptor down-regulation receptor degradation multiple sclerosis

Introduction

FTY720 (Fingolimod/Gilenya) is an approved drug for the treatment of multiple sclerosis, and this has established the S1P receptor (SIPR) family as an effective therapeutic intervention point for this disease. Although FTY720 is an agonist for S1P₁, S1P₃, S1P₄, and S1P₅, expression data and preclinical studies suggest that the efficacy is largely mediated by S1P₁.^1,2 Current research in the field has focused on developing molecules with an improved selectivity profile with the goal of reducing side effects observed with FTY720. These side effects, which may be mediated by the other S1P receptors, include bradycardia, macular edema, and hypertension, as outlined in the label. We have developed a set of complementary assays to study S1PR agonists, and here we compare their potency and selectivity in GTPγS, β-arrestin recruitment, calcium flux, cyclic AMP (cAMP) reduction, receptor down-regulation, and degradation assays. We demonstrate that all of the small molecules investigated signal through S1P₁ in a similar manner by causing receptor down-regulation and degradation. This behavior contrasts with the natural ligand (S1P), which induces receptor down-regulation from the cell surface but not degradation at physiological concentrations (at high concentrations of S1P, degradation of the receptor has been observed^3,4). We also identify some differences in the selectivity profiles across different S1P receptors, which may prove important as the origin of the side effects is further understood.

Materials and Methods

Cell Lines

Tango EDG1-bla U2OS cells (human S1P₁), Tango EDG3-bla U2OS cells (human S1P₃), Tango EDG6-bla U2OS cells (human S1P₄), and Tango EDG8-bla U2OS cells (human S1P₅) cell lines were purchased from Life Technologies (Carlsbad, CA) and cultured according to the manufacturer’s instructions. Tango rat S1P₁, (rS1P₁) and mouse S1P₁ (mS1P₁) were generated by Life Technologies and cultured according to the manufacturer’s instructions. Human S1P₁ (hS1P₁), human S1P₂ (hS1P₂), human S1P₃ (hS1P₃), human S1P₅ (hS1P₅), rat S1P₃ (rS1P₃), dog S1P₁ (dS1P₁), dog S1P₃ (dS1P₃), and dog S1P₅ (dS1P₅) receptors were subcloned into pcDNA3.2/V-5 Dest expression vector and transfected into the CHO-k1Gα₁₆ cell line using lipofectamine 2000 (Invitrogen 11668-019, Carlsbad, CA) according to the manufacturer’s instructions. They were either used directly or selected as pools of stably transfected cells (as indicated in the “Results” section). Cells were passaged twice a week in Hams F12 (Gibco 11765), 10% fetal bovine serum (Gibco 16140), penicillin/streptomycin, 0.2 mg/mL Zeocin (Invitrogen R25001), and 0.5 mg/mL Geneticin (Gibco 10131).

[³⁵S] GTPγS SPA Binding Assay

Human S1PR membranes for S1P₁, ₂, ₃, and ₅ (Millipore, Billerica, MA) were diluted to 2 µg/well for S1P₁ and S1P₃ and 3 µg/well for S1P₂ and S1P₅. A mixture (160 µL) of membrane suspension in assay buffer (25 mM Tris-HCl, pH 7.9, 100 mM NaCl, 3 mM MgCl₂, 0.2 mM EGTA, 0.25% bovine serum albumin [BSA], 10 µg/mL saponin, and 30 µM GDP) was added to a 96-well flexi-sample plate (Wallac) containing 10 µL of 20× compound, S1P (for 100% activation control), or DMSO only (no agonist control) for a final concentration of 1 µM for S1P and 1% DMSO. Cold GTPγS was included at 10 µM final concentration in the assay as a nonspecific binding control. Thirty microliters of [³⁵S] GTPγS (PerkinElmer, Waltham, MA), specific activity = 1250 Ci/mmol, was added for a final concentration of 0.10 nM. The assay plate was incubated at room temperature for 1 h followed by the addition of 50 µL of WGA PVT SPA beads (GE HealthCare RPNQ0001, Waukesha, WI) resuspended to 17.9 mg/mL in 25 mM Tris-HCl, pH 7.9, 100 mM NaCl, 3 mM MgCl₂, 0.2 mM EGTA. The plate was incubated in the dark at room temperature for an additional hour, sealed, and centrifuged for 10 min at 1500 rpm at room temperature. Light emission was then determined on the MicroBeta plate reader (PerkinElmer) reading at 3 min/well. Reagent sourcing was S1P, GDP, GTPγS, fatty acid–free BSA, DTT, sodium chloride, magnesium chloride, DMSO, and saponin (Sigma, St. Louis, MO), Tris-HCl and PBS (Gibco), and EGTA (Anaspec, Inc, Fremont, CA).

cAMP Assays

S1P₁, ₂,₃ and ₅ CHO cells were plated at a density of 10,000 cells per well in Hank’s Balanced Salt Solution (HBSS), 0.1% BSA, in a white 384-well plate and incubated at 37 °C, 5% CO₂, for 2 h. Forskolin and agonists were diluted into HBSS, 0.1% BSA, 0.5 mM IBMX (3-isobutyl-1-methylxanthine), and added to the wells. The final concentration of forskolin was 10 µM. The cells were incubated at 37 °C, 5% CO₂, for another 2 h. The assay was stopped with the addition of cAMP detection reagents (PerkinElmer). The assay was read on a PE Envision (LANCE), and a ratio was calculated by dividing the 665 nm/620 nm channels. A total of 10 µM S1P was used as a positive control, and wells containing no agonist were used as negative controls. All compounds tested were subjected to a 10-point dose response in 0.5% DMSO. Compounds were also evaluated as antagonists by incubating the compound with the natural ligand, S1P, at its EC₈₀ for each receptor.

Calcium Flux Assays

S1P₁,₂,₃ and ₅ CHO cells were plated at a density of 50,000 cells per well into 96-well, black, clear-bottom tissue culture–treated plates (BD Biosciences, Franklin Lakes, NJ) and allowed to adhere overnight. The media were removed, and the cells were washed once with HBSS (Life Technologies). Calcium 4 dye (Molecular Devices, Sunnyvale, CA) was prepared according to the manufacturer’s instructions and added to each well. Agonists were diluted in HBSS and added to the cells on a FLIPR Tetra. Fluorescence intensity was measured for 180 s. Data are represented as relative fluorescent units by subtracting the minimum signal from the maximum signal. A total of 10 µM S1P was used as a positive control, and wells containing no agonist were used as negative controls. All compounds tested were subjected to a 10-point dose response in 0.5% DMSO.

β-Arrestin Recruitment Assays

β-arrestin recruitment assays for S1P₁, ₃, ₄ and ₅ (Tango) were performed according to the manufacturer’s instructions with the following modifications. Agonists were diluted in assay media (FreeStyle Expression Medium, Invitrogen) so that the final DMSO concentration in the assay was 1%. The agonists were routinely incubated on the cells for 18 h prior to addition of the GenBlazer substrate. Ratio values for each well were compared with 10 µM S1P (percentage of control [POC]). EC₅₀ values were calculated with XLfit using a four-parameter fit. Compounds were also evaluated as antagonists by incubating the compound with the natural ligand, S1P, at its EC₈₀ for each receptor.

S1P₁ Degradation

HS1P₁ CHO cells were seeded at 1 X 10⁶ cells per 35-mm well and incubated in serum-free F12 media overnight. Cells were preincubated with compounds for 30 min (final DMSO concentration of 1%), washed with PBS, and detached using 1 mL of Accutase (Millipore). They were then pelleted and resuspended in 100 µL of 1× Sample Prep Buffer (4XNuPAGE LDS, Invitrogen NP0007) containing reducing agent (10XNuPAGE Reducing Agent, Invitrogen NP0004) and Benzonase (Novagen 71206-3), vortexed, sonicated, and heated to 95 °C for 5 min, prior to Western analysis. Equivalent volumes were run on 4% to 12% Bis Tris NuPAGE gels (Invitrogen NP-323) in 1× MOPS Buffer (Invitrogen NP0001) at 180 V for ~1 h, transferred to nitrocellulose for 7 min at 20 V using the i-Blot Dry Transfer System (Invitrogen IB1001), and blocked for 1 h at room temperature in Li-Cor Blocking Buffer (Odyssey 927-40000). Blots were probed with rabbit Anti-EDG1 antibody (sc-25489, Santa Cruz Biotechnology, Santa Cruz, CA) at a 1:1000 dilution followed by goat Anti-Rabbit IR Dye 680 (Li-Cor 926-32221) diluted 1:10,000 for visualization of the S1P₁ receptor. Blots were stripped with Restore Western Blot Stripping Buffer (Thermo Scientific 21059, Waltham, MA) and then reprobed with monoclonal Anti-GAPDH, clone GAPDH-71.1 (Sigma G8795) diluted 1:20,000, followed by goat anti-Mouse IR Dye 800 CW (Li-Cor Odyssey 926-32210) for the visualization of GAPDH (loading control). All antibodies were diluted in Li-Cor Blocking Buffer containing 0.1% Tween-20 (Sigma P1379). The Li-Cor Odyssey was used to detect the S1P₁ receptor signal, which was quantified using densitometry and normalized to the GAPDH signal.

Human umbilical vein endothelial cells (HUVECs) were obtained from Lonza (Basel, Switzerland), cultured in EBM-2 medium (Lonza #CC-3156), and seeded at 0.5 X 10⁶ cells per 35 mm well and incubated in serum-free EBM-2 media overnight. HUVECs were incubated with compound for 4 h as this was the time when maximum degradation was observed in a time course study (data not shown), rather than 30 min, the time of maximum degradation observed in S1P₁ CHOs (data not shown). HUVECs were processed in the same manner as the S1P₁ CHO cells after compound treatment.

Receptor Down-regulation Assay (Flow Cytometry)

HS1P₁ CHO cells were plated overnight (50,000–75,000 cells per 150 µL/well) in 96-well round-bottom plates in serum-free F12 media. The next day, cells were incubated with compounds for 30 min at 37 °C in a total volume of 200 µL (final DMSO concentration was 1%), detached with Accutase, washed, and stained with Cell Staining Buffer (Biolegend 420201, San Diego, CA). S1P₁ receptors were detected on the cell surface by staining with mouse anti-human-S1P₁ (Santa Cruz Biotechnology, sc-73923) at a 1:10 dilution for 30 min on ice, washed, and stained with Allophycocyanin goat anti-mouse IgG (Biolegend 405308) at a 1:50 dilution. Stained cells were analyzed on the FACS Canto II (BD Biosciences). Mean fluorescent intensity (MFI) was calculated using FlowJo 7.6.3. The MFI from untreated (DMSO) cells (maximum expression) was set to 100% and was used to calculate the POC for the experimental samples.

Compounds and Data Handling

Select chemicals were purchased from commercial sources, for example, (S)-FTY720-P (Toronto Research Chemicals, Toronto, Canada; referred to as FTY720-P in the text), KRP-203-P (Cayman Chemical, Ann Arbor, MI, provides a racemic mixture that was used here), and SEW2871 (Sigma-Aldrich). The remaining chemicals were synthesized according to published methods, for example, compound X (US2009/275554), compound Y (US 2010/0249071), CS-0777 (US2007/191468), PF-991 (WO2009/060278), and BAF312 (WO2004/103306 and WO2010/071794).^5,6

POC was calculated for all wells from the raw data of the test compounds, S1P as a positive control, and no agonist as a negative control. EC₅₀ values were calculated using XLfit (IDBS, Guildford, Surrey, UK) with a four-parameter fit.

Results

We developed and evaluated multiple assay formats for measuring S1P receptor response to natural and chemical ligands, both to allow us to compare our data to previously published data and to choose the most informative and robust assays for compound profiling. As shown in Figure 1 , our optimized GTPγS assays for S1P_{1, 3} (A) and S1P_{2, 5} (B) demonstrated a dose response to the natural ligand (S1P). Although the fold stimulation (signal to background ratio) was larger for S1P₁ and S1P₃ (at 10X and 14X, respectively) than for S1P₂ and S1P₅ (both at ~2X), the signals were reproducible regardless of membrane batch of the same receptor tested. The dose-response data for FTY720-P and BAF312 are shown for S1P₁ and S1P₃ in Figure 1C and D . Consistent with published reports, FTY720-P⁷ and BAF312⁵ are potent and efficacious S1P₁ agonists (EC₅₀s of 1.4 nM and 0.14 nM, respectively). In addition, we confirmed that FTY720-P is a partial agonist with shifted potency against S1P₃ (EC₅₀: 13 nM; 30% efficacy), whereas BAF312 does not act as an agonist of S1P₃ ( Fig. 1D ) up to 10 µM in experiments not shown. We performed dose responses for all compounds in each assay format based on availability of subtype and species for the S1PRs ( Tables 1 and 2 ). S1P receptor agonists tested were the natural ligand S1P, FTY720-P (the biologically active phosphate metabolite of FTY720), BAF312, CS-0777, PF-991, SEW 2871, compound Y (a representative compound from the Exelixis S1P patent literature), KRP203-P, and compound X (a representative compound from the ONO S1P patent literature). It was not possible to present data from the unphosphorylated (prodrug) and phosphate form (more active form) of all compounds because of patent restrictions. Compound phosphorylation will likely occur in cells, and this needs to be taken into consideration when comparing assay formats. All chemical structures are shown in Table 1 . Selectivity profiles are discussed later in the results. The second assay format evaluated was β-arrestin recruitment to the receptor (TANGO), which was developed for S1P₁, ₃, ₄ and _5. This format has been previously reported,⁸ and one important advantage is that the signal measured is due entirely to signaling through the transfected receptor and is not due to signals from other endogenously expressed S1P receptors. We optimized the assay to maintain the assay window while using 1% DMSO by increasing the incubation time with ligand from 3 to 18 h ( Fig. 1E ). The assay window increases from 14 to 30 (S1P) with a longer incubation time. This assay format has a large fold increase of between 5X and >20X for all four receptors tested (data not shown) and has no background in parental cells due to the specificity of the β-arrestin readout in transfected cells, making it a robust format for compound profiling. As shown in Figure 1F , these changes did not affect the EC₅₀s obtained in the assay for FTY720-P: EC₅₀ values of 0.08 nM and 0.09 nM were obtained at 3 and 18 h, respectively. In Figures 1G and H , the EC₅₀s for FTY720-P and BAF312 are compared for S1P₁ and S1P₃ and are consistent with the GTPγS and cAMP results in that FTY720-P is a potent (0.09 nM) S1P₁ agonist and a less potent (36 nM) partial (37% efficacy) S1P₃ agonist. In contrast, BAF312 is completely inactive against S1P₃, whereas it is potent against S1P₁ (2.9 nM). All compounds were profiled against all four receptors ( Tables 1 and 2 ) in both agonist and antagonist mode.

Figure 1.

Optimized GTPγS and β-arrestin (Tango) recruitment assays for S1P receptors. CPM, counts per million. GTPγS assays (A–D). (A) Concentration-response curves of the natural ligand, S1P, with S1P₁ (closed circles) and S1P₃ (open triangles) gave EC₅₀ values of 1.9 nM (fold stimulation = 10) and 2 nM (fold stimulation = 14), respectively. (B) Titration of S1P on S1P₂ (closed circles) and S1P₅ (open triangles) gave EC₅₀ values of 22 nM and 2 nM, respectively. The fold stimulation for both S1P₂ and S1P₅ was ~2. (C) Dose-response curves for FTY720-P with S1P₁ (closed circles) and S1P₃ (open triangles) resulted in EC₅₀ values of 1.4 nM and 13 nM, respectively. (D) Dose-response curves for BAF312 with S1P₁ (closed circles; EC₅₀ = 0.14 nM) and S1P₃ (open triangles; EC₅₀ >10,000 nM). n = 1 experiment, with duplicates represented here. β-arrestin (Tango) recruitment assays (E–H). Tango EDG1-bla U2OS cells (human S1P₁) were incubated with the natural ligand, S1P (E) or FTY720-P (F) for 3 h (closed circles) and 18 h (open triangles). The data are represented as a ratio of 460 nm/530 nm. Dose-response curves of FTY720-P (G) and BAF312 (H) were performed at 18 h with Tango EDG1-bla U2OS cells (human S1P₁, closed circles) and Tango EDG3-bla U2OS cells (human S1P₃, open triangles). EC₅₀ values for FTY720-P (G) are 0.09 nM and 36 nM for S1P₁ and S1P₃, respectively. EC₅₀ values for BAF312 (H) are 2.9 nM and >1000 nM for S1P₁ (closed circles) and S1P₃ (open triangles), respectively. n = 1 experiment, with duplicates represented here. Additional experiments are represented in the averaged numbers presented in the tables.

Table 1.

Comparison of EC₅₀s for S1P₁ across multiple assay formats.

Structure
Compound (n = X, Averages)	S1P	(S)-FTY720-P	BAF312	CS-0777	PF-991	SEW 2871	Y	KRP203-P	X
β-arrestin EC₅₀ nM (n = 3)	13	0.4	2.5	0.9	1.9	160	0.1	0.2	0.3
GTPγS EC₅₀ nM (n = 3)	1.2	2	0.2	255	0.2	100	0.32	0.2	0.3
cAMP EC₅₀ nM (n = 3)	480	0.04	0.01	1	0.06	ND	0.45	0.06	0.01
Receptor down-regulation EC₅₀ nM (n = 1)	25	0.7	0.5	5	11	76	7	2	0.2
Receptor degradation EC₅₀ nM (n = 1)	302	0.34	0.59	Not calculated	Not calculated	195	51	Not calculated	ND

Table 2.

Selectivity profile of S1P receptor agonists on S1P₂, ₃, ₄ and ₅ shown in multiple assay formats (EC₅₀, nM).

Compound	S1P	(S)-FTY720-P	BAF312	CS-0777	PF-991	SEW2871	Y	KRP203-P	X
S1PX/assay
S1P₂/GTPγS	22	>10,000	>10,000	>10,000	>10,000	ND	>10000	>10,000	>10,000
S1P₃/β-arrestin	460	54	>10,000	>10,000	>10,000	>10,000	110	>10,000 (1 nM antagonist)	>10,000
S1P₃/GTPγS	2	175	>10,000	>10,000	>10,000	>10,000	ND	>10,000	>10,000
S1P₃/cAMP	77	12	>10,000	>10,000	>10,000	ND	360	>10,000	>10,000
S1P₄ /β arrestin	1900	26	245	1.5	34	ND	18	>10,000	800
S1P₅/GTPγS	3	2.5	0.5	>10,000	12	>10,000	8	ND	0.6
S1P₅/β-arrestin	35	1.8	0.8	15	7	>10,000	18	0.2	0.1
S1P₅/cAMP	180	0.3	0.2	23	6	ND	ND	1.6	0.1

We also evaluated cAMP and Ca²⁺ flux assays for S1P₁, ₂, ₃ and ₅ in CHO cells stably expressing the receptors. Figure 2 shows that although the fold increase is small for cAMP on S1P₁ and S1P₃, the natural ligand, S1P, gives a clear dose response with reproducible data and EC₅₀s of 580 nM and 98 nM, respectively. Similar to what we observed in the GTPγS format, both FTY720-P and BAF312 are potent S1P₁ agonists (EC₅₀s of 0.4 nM and 0.01 nM, respectively), whereas only FTY720-P is potent against S1P₃ (EC₅₀ of 11 nM; BAF312 shows no significant inhibition at less than 10 µM). In our efforts to develop the Ca²⁺ flux assay, we noticed that the parental CHO cells responded in a concentration-dependent manner to the S1P ligand, suggesting that we were measuring the response of hamster S1P receptors or other intracellular signaling of the ligand (Supplemental Fig. 1A). Although the absolute signal is higher in the S1P receptor–expressing cells, it does complicate compound profiling in terms of normalizing data to S1P and reporting accurate EC₅₀s. In the case of FTY720-P, (Supplemental Fig. 1B), a dose response can be observed in the S1P₁-transfected cells and is left shifted compared with the background response on parental cells (1 nM vs. 570 nM). However, the EC₅₀s obtained in this format are also right shifted compared with other formats ( Table 1 ). All the compounds tested herein were inactive versus S1P₂ in all assay formats (data not shown, see Table 2 for inactivity of all compounds in the GTPγS assay). The use of this assay format, in which endogenous S1P receptor expression interferes with the calculation of absolute EC₅₀s, needs to be carefully considered because of the likely expression of endogenous S1P receptors in CHO cells. For this reason, potency data generated for compounds in this assay format are reported only for reference in Supplemental Table 1.

Figure 2.

Cyclic AMP (cAMP) assays for human S1P₁ and S1P₃. (A) hS1P₁ CHO or hS1P₃ CHO were incubated with the natural ligand (S1P) in the presence of forskolin. cAMP levels are plotted as the ratio of 665 nm/620 nm for each concentration of S1P for S1P₁ (closed circles and left axis) and S1P₃ (open triangles and right axis). (B) Dose-response curves of FTY720-P (S1P₁, closed circles; S1P₃, open circles) and BAF312 (S1P₁, closed triangles; S1P₃, open triangles) in hS1P₁ CHO or hS1P₃ CHO in the cAMP assay. Ratio values for each well were compared with 10 µM S1P (percentage of control). IC₅₀ values for FTY720-P are 0.04 nM and 11 nM for S1P₁ and S1P₃, respectively. IC₅₀ values for BAF312 are 0.01 nM and >10,000 nM for S1P₁ and S1P₃, respectively. n = 1 experiment, with duplicates represented here. Additional experiments are represented in the averaged numbers presented in the tables.

We then developed a flow cytometry assay using the transfected hS1P₁ CHO cells to monitor receptor internalization from the cell surface as a surrogate for receptor down-regulation. As shown in Figure 3A , there is a fold increase in the assay of 8X for BAF312, in which the highest compound dose is compared with isotype control and background. The untransfected CHO cells have a staining pattern similar to the isotype control, demonstrating specificity of the antibody for human S1P₁ receptor (Supplemental Fig. 2). Figure 3B shows the dose-response data for BAF312 compared with no agonist (EC₅₀ = 0.5 nM). In parallel, we assessed the degree of S1P₁ receptor remaining in the cell pellets by Western analysis ( Fig. 3C , D ) and showed that the EC₅₀s for receptor down-regulation from the cell surface and degradation of receptor in the cell are of a similar magnitude. The EC₅₀ for BAF312 is 0.6 nM for S1P₁ degradation in S1P₁ CHO cells.

Figure 3.

Receptor down-regulation and degradation assays for S1P₁ on S1P₁ CHO cells and human umbilical vein endothelial cells (HUVECs). (A) Measurement of S1P₁ on the cell surface in the presence of 100 nM BAF312. Media (dashed black), Isotype (tinted gray), and BAF312 (thick black line). Mean fluorescence intensity is shown. (B) Dose response of BAF312 (closed triangles) in inducing receptor down-regulation as compared with no agonist (percentage of control [POC]). (C) Western analysis of S1P₁ in response to BAF312 compared with GAPDH control. S1P₁ CHO cells were treated with BAF312 in a 10-point dose response starting at 100 nM. Samples were immunoblotted with anti-S1P₁, stripped, and reprobed with GAPDH. M = media only; D = DMSO. (D) Densitometric representation of the Western analysis shown in (C) and represented as a POC, DMSO only, and normalized to GAPDH. hS1P₁ CHO cells (E) or HUVECs (F) were treated with a 10-point dose response of FTY720-P starting at 100 nM or 1000 nM, respectively. Lysates were run on a Western blot and probed with anti-S1P₁ (top panel), stripped, and reprobed with GAPDH (lower panel). M = media only; D = DMSO. Densitometric representation of the Western analysis shown in (E) and (F) are shown in (G) and (H) and represented as POC, DMSO only. (I) Western blot of hS1P₁ CHO cells treated with compounds at their respective EC₅₀ and EC₉₀ concentrations for 30 min. Lanes 1 and 2, S1P at 25 nM and 1000 nM; lanes 3 and 4, KRP203-P at 2 nM and 100 nM; lanes 5 and 6, PF-991 at 10 nM and 1000 nM; lanes 7 and 8, BAF312 at 1 nM and 100 nM; lanes 9 and 10, CS-0777 at 10 nM and 1000 nM, respectively. (J) Western blot of HUVECs treated with compounds at their respective EC₅₀ and EC₉₀ concentrations for 4 h. Lanes 1 and 2, FTY720-P at 100 nM and 1 nM; lanes 3 and 4, compound Y at 100 nM and 0.3 nM; lanes 5 and 6, BAF312 at 100 nM and 0.3 nM, respectively. (K) Densitometric quantification of the ratio of S1P₁ to GAPDH per well of hS1P₁ CHO Western blot data represented in (J) plotted as POC, DMSO only. (L) Densitometric quantification of the ratio of S1P₁ to GAPDH per well of HUVEC Western blot data (M) is plotted as percentage of DMSO-only treatment.

We also evaluated receptor down-regulation and degradation on HUVECs, which endogenously express S1P₁⁹ and where the S1P₁ receptor degradation upon treatment with functional S1P₁ antagonists has previously been described.¹⁰ We confirmed S1P₁ and S1P₃ transcripts were present by quantitative PCR (qPCR; data not shown) and evaluated the cells for protein expression. We did not detect cell surface protein by flow cytometry but could detect the protein by Western blot. We therefore performed a receptor degradation assessment in parallel for hS1P₁ CHO cells and HUVECs and found similar EC₅₀s, as shown in Figure 3E , F (Western blots) Figure 3G , H (densitometry) for FTY720-P; EC₅₀s ranged from 0.2 to 0.6 nM in hS1P₁ CHO cells (n = 3) and from 0.3 to 1.3 nM in HUVECs (n = 3), illustrating the relevance of the hS1P₁ CHO cells for evaluating receptor degradation.

Given this correlation, we evaluated a number of other compounds in the hS1P₁ CHO cells by receptor degradation (complete dose response; Table 1 ) and by Western blot at more limited concentrations (EC₅₀ and EC₉₀ of each compound) to determine if receptor degradation is tracking with down-regulation as for FTY720-P ( Fig. 3I and J , Western blots; Fig. 3K and L , densitometry) or whether it is occurring at significantly higher concentrations, as was observed for the natural ligand S1P ( Table 1 ). We observed that the EC₅₀s for receptor down-regulation and degradation tracked for all S1P₁ agonists profiled.

Receptor selectivity for S1P_1–5 was tested using two overlapping sets of assays: GTPγS for S1P₁, ₂, ₃ and ₅ and β-arrestin for S1P_{1, 3}, ₄ and ₅ (as appropriate reagents were not available for GTPγS for S1P₄ or β-arrestin for S1P₂). As shown in Table 2 , although some variation exists between reported EC₅₀s in different assay formats, the overall selectivity profile is consistent. All compounds are inactive against S1P₂, GTPγS assay, also inactive, in other assay formats (data not shown in Table 2 ). FTY720-P and compound Y are potent against S1P₃, ₄ and ₅. This is consistent with published data for FTY720.¹¹ BAF312 is selective against S1P₁ and ₅ as reported by Novartis,⁵ has limited activity on S1P₄, and is completely inactive on S1P₂ and S1P₃. CS-0777, PF-991, and compound X have similar profiles: potent against S1P₅ and S1P₄ and inactive against S1P₂ and S1P₃. SEW2871 is described as an S1P₁ receptor agonist,¹² although is appears somewhat less potent in our hands. KRP203-P is potent against S1P₅ and shows no agonism against S1P₂, ₃ or ₄ ( Table 2 ). However, we observed that it is a potent antagonist against S1P₃ in the β-arrestin format (IC₅₀ of 1 nM). All compounds herein were profiled in both agonist and antagonist mode, and KRP203-P was the only compound that displayed antagonism versus S1P₃.

Several compounds were evaluated for their potency against rat, mouse, and dog S1P₁, ₃ and ₅ receptors ( Table 3 ). As reported, FTY720-P cross-reacts with all three species.¹³ We found that for all compounds tested, their species cross-reactivity was consistent with the human selectivity profile.

Table 3.

Species cross-reactivity of S1P receptor agonists.

Assay EC₅₀ (nM), n = 1	S1P	(S)-FTY720-P	BAF312	CS-0777	PF-991	Y	KRP203-P	X
Stable transfected CHO (Ca²⁺)
hS1P₃	1	3	>10,000	>10,000	>10,000	54	>10,000	>10,000
rS1P₃	1	2	>10,000	>10,000	>10,000	6	>10,000	>10,000
β-arrestin in U2OS
hS1P₁	13	0.1	ND	ND	ND	ND	ND	ND
rS1P₁	82	0.02	ND	ND	ND	ND	ND	ND
mS1P₁	156	0.02	ND	ND	ND	ND	ND	ND
Transiently transfected CHO (Ca²⁺)
hS1P₁	1.3	30	22	ND	ND	ND	ND	ND
dS1P₁	3.3	18	9	758	52	42	84	1
dS1P₃	0.4	32	>10,000	>10,000	460	40	>10,000	1700
dS1P₅	8.4	159	890	>10,000	10	214	21	20

Discussion

We developed and compared different assays to monitor S1P receptor engagement, including GTPγS, cAMP reduction, calcium flux, β-arrestin recruitment, receptor down-regulation from the cell surface, and receptor degradation. We have used these assays to compare the mechanism of action and selectivity profile of S1P receptor family agonists including the clinically validated FTY720-P as well as other candidates and tool compounds. Furthermore, we demonstrate that although these compounds have different selectivity profiles, they all signal in a similar way through S1P₁, leading to receptor down-regulation from the cell surface, and in contrast to the natural ligand, S1P, they all lead to receptor degradation.

Independent of assay format, the EC₅₀s are remarkably conserved, despite different transfected cell lines used (three different backgrounds in these cases), and so we believe GTPγS and β-arrestin represent a useful pair of assays for assessing compound potency and relating early events in S1P₁ receptor activation to functional antagonism. Given the likely endogenous expression of S1P receptors in CHO cells, we do not believe that measurement of Ca2⁺ in those cells gives accurate estimations of compound potencies.

One challenge in working with this class of compounds is that they are often phosphorylated in cells/in vivo (by sphingosine or sugar kinases), leading to higher activity than the unphosphorylated compound, and hence they are described as prodrugs. Patent coverage by other companies prevents us from using both the prodrug and the more active (phosphate-P) form of all the compounds in parallel in this study. Because we are most interested in the cellular readout, we focused our attention on compound activity in longer-term cellular assays (e.g., 18 h incubation in β-arrestin, where phosphorylation of these compounds can likely occur). In these assays, we are likely measuring a combination of cell permeability and activity of phosphorylated/dephosphorylated compounds. Nonetheless, as we ultimately want to know how the compounds function in cells, we believe that the β-arrestin assay is the most relevant format.

Examples of this may be CS-0777 ( Table 1 ), which can be phosphorylated and does not appear as potent in GTPγS as in β-arrestin. This is consistent with SEW-2871 as a nonphosphorylatable compound whose potencies are relatively insensitive to assay format. We can also speculate that this is due to differences in assay conditions. A third possibility, which may also explain the somewhat divergent data in cAMP format, is that different compounds behave as biased agonists, activating one signaling pathway more than another. This has been alluded to for FTY720 on S1P₃,¹⁴ and differences in EC₅₀ values between different formats are indeed more pronounced for S1P₃ ( Table 2 ) than for S1P₅. For this reason, it is important to link the upstream signaling readout with a more biologically relevant downstream readout—receptor degradation, in this case.

To that end, we linked the initial events in cell signaling with the internalization of the S1P₁ receptor from the cell surface in a feasible assay format and demonstrate that this can be associated with receptor down-regulation depending on the compound profile. The down-regulation observed in the hS1P₁ CHO cells parallels that observed in HUVECs endogenously expressing S1P₁, building confidence that CHOs overexpressing S1P₁ can be used as biologically relevant reagents to measure this process. Calcium flux measurement, however, does not appear to represent a useful system because of background effects observed in CHO cells, and although the rank order of compound potencies was maintained in this format, they were significantly less potent, which may lead to loss of sensitivity/failure to identify hits in this format.

On a technical note, we observed that the S1P₂ and ₅ cellular assays had smaller-fold stimulations than their counterpart assays on S1P₁ and S1P₃ in all formats. This window is likely mediated by the degree of receptor expression in the different membrane preparations and cells. Although we cannot demonstrate this directly because of lack of specific antibodies for all S1P receptors, it is supported by qPCR on some of the cell lines (data not shown).

Our comparisons of different assay formats in parallel for S1P₁ allow us to compare different compounds directly, which have been profiled using different assays in the literature. This is important in understanding the relative selectivity profiles of tools and clinical candidates and the relative value of different assay systems.

We have also compared multiple compounds (or representative compounds of series) that are in clinical development in all our assay formats (in both agonist and antagonist mode, where possible) including across different species. This allows us and others to review and compare published in vitro data and reports and critically review in vivo pharmacology data, particularly when it comes to evaluating off-target effects based on differential selectivity profiles (e.g., the antagonism observed for KRP-203-P on S1P₃ may explain the relatively weaker impact on bradycardia in rats than that observed with FTY720¹⁵).

Based on our analysis, it may be difficult to differentiate any of these compounds from FTY720 from a safety point of view, if it is mediated by activity on S1P₁, as they all induced S1P₁ receptor degradation, which has been suggested to be causative in inducing vascular leak^16,17 and may be associated with the reported lung phenotypes. However, subtleties in agonism/antagonism on the other receptors (e.g., S1P₃ antagonism that may protect against cardiac effects and differences in agonism on S1P₄ that may lead to some positive pharmacological impact on lymphocytes) may lead to differences in the efficacy and safety profile of these compounds under clinical evaluation.

In conclusion, we recommend a detailed analysis of the downstream signaling events when developing agonists for G-protein–coupled receptors and a thorough evaluation of tool compounds in all available formats to characterize lead series. We conclude that for S1P receptors, a combination of GTPγS and β-arrestin assays can provide a robust data set to drive structure-activity relationships in medicinal chemistry programs.

Footnotes

Acknowledgements

We thank Nuruddeen Lewis for reading the manuscript.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Supplementary material for this article is available on the Journal of Biomolecular Screening Web site at .

References

Brinkmann

Sphingosine 1-Phosphate Receptors in Health and Disease: Mechanistic Insights from Gene Deletion Studies and Reverse Pharmacology. Pharmacol. Ther. 2007, 115, 84–105.

Matloubian

C. G.

Cinamon

. Lymphocyte Egress from Thymus and Peripheral Lymphoid Organs Is Dependent on S1P Receptor 1. Nature 2004, 427, 355–360.

M. L.

Thangada

M. T.

. Immunosuppressive and Anti-angiogenic Sphingosine 1-Phosphate Receptor-1 Agonists Induce Ubiquitinylation and Proteasomal Degradation of the Receptor. J. Biol. Chem. 2007, 282, 9082–9089.

Rosen

Gonzalez-Cabrera

Marsolais

. Modulating Tone: The Overture of S1P Receptor Immunotherapeutics. Immunol. Rev. 2008, 223, 221–235.

Legangneux

Gardin

Johns

Dose Titration of BAF312 Attenuates the Initial Heart Rate Reducing Effect in Healthy Subjects. Br. J. Clin. Pharmacol. 2012, 75, 831–841.

Luo

Smith

. Quantitative Determination of BAF312, a S1P-R Modulator, in Human Urine by LC-MS/MS: Prevention and Recovery of Lost Analyte due to Container Surface Adsorption. J. Chromatograph. B. 2012, 878, 583–589.

Brinkmann

Davis

M. D.

Heise

C. E.

. The Immune Modulator FTY720 Targets Sphingosine 1-Phosphate Receptors. J. Biol. Chem. 2002, 277, 21453–21457.

Wetter

J. A.

Revankar

Hanson

B. J.

Utilization of the Tango Beta-Arrestin Recruitment Technology for Cell-based EDG Receptor Assay Development and Interrogation. J. Biomol. Screen. 2009, 9, 1134–1141.

Lin

C. I.

Chen

C. N.

Lin

P. W.

. Sphingosine 1-Phosphate Regulates Inflammation-Related Genes in Human Endothelial Cells through S1P1 and S1P3. Biochem. .Biophys Res. Commun. 2007, 355, 895–901.

10.

LaMontagne

Littlewood-Evans

Schnell

. Antagonism of Sphingosine-1-Phosphate Receptors by FTY720 Inhibits Angiogenesis and Tumor Vascularization. Cancer Res. 2006, 66, 221–231.

11.

Mullershausen

Zecri

Cetin

. Persistent Signaling Induced by FTY720-Phosphate Is Mediated by Internalized S1P1 Receptors. Nat. Chem. Biol 2009, 6, 428–434.

12.

Sanna

M. G.

Gonzalez-Cabrera

P. J.

. S1P1-Selective In Vivo-Active Agonists from High-Throughput Screening: Off-the-Shelf Chemical Probes of Receptor Interactions, Signaling, and Fate. Chem. Biol. 2005, 6, 703–715.

13.

Troncoso

Kahan

B. D.

Preclinical Evaluation of a New Immunosuppressive Agent, FTY720. Clin. Biochem. 1998, 5, 369–373.

14.

Jongsma

van Unen

van Loenen

P. B.

. Different Response Patterns of Several Ligands at the Sphingosine-1-Phosphate Receptor Subtype 3 (S1P(3). Br. J. Pharmacol. 2009, 8, 1305–1311.

15.

Shimizu

Takahashi

Kaneko

. KRP-203, a Novel Synthetic Immunosuppressant, Prolongs Graft Survival and Attenuates Chronic Rejection in Rat Skin and Heart Allografts. Circulation 2005, 111, 222–229.

16.

Shea

B. S.

Brooks

S. F.

Fontaine

B. A.

. Prolonged Exposure to Sphingosine 1-Phosphate Receptor-1 Agonists Exacerbates Vascular Leak, Fibrosis, and Mortality after Lung Injury. Am. J. Respir. Cell Mol. Biol. 2010, 43, 662–673.

17.