Mapping Pathways Downstream of Sphingosine 1-Phosphate Subtype 1 by Differential Chemical Perturbation and Proteomics*

Sphingosine 1-phosphate subtype 1 (S1P1) receptor agonists alter lymphocyte trafficking and endothelial barrier integrity in vivo. Among these is the potent, non-selective agonist, FTY720-P, whose mechanism of action has been suggested to correlate with S1P1 down-regulation. Discovery of the in vivo active S1P1-selective agonist, SEW2871, has broadened our understanding of minimal requirements for S1P1 function while highlighting differences regarding agonist effect on S1P1 fate, because SEW2871 does not degrade S1P1. To further understand the mechanism of agonist-induced S1P1 down-regulation, we compared signaling and fate of human S1P1-green fluorescent protein (GFP) in stable 293 cells, using AFD-R, a chiral analog of FTY720-P, SEW2871, and S1P. Although all agonists acutely internalized S1P1 to late endosomal vesicles and activated GTPγS35 binding and pERK to similar maxima, only AFD-R led to significant S1P1 down-regulation, as shown by GFP immunoprecipitation studies. Down-regulation was time- and concentration-dependent, was partially blocked by proteasomal inhibition and reversed by chloroquine and an antagonist to S1P1. All agonists induced a receptor-associated increase in ubiquitination, with AFD-R inducing 3-fold more accumulation than S1P and being 3–4 logs more potent than SEW2871. The formation of AFD-R-receptor ubiquitin complex was inhibited by antagonist and chloroquine and was enhanced by proteasomal inhibition. Identification of proteins by PAGE liquid chromatography-tandem mass spectrometry in cells treated with AFD-R confirmed the co-migration of ubiquitin peptides with those of S1P1 and GFP, relative to vehicle alone. These data suggest that the hierarchy of ubiquitin recruitment to S1P1 (AFD-R > S1P > SEW2871) correlates with the efficiency of lysosomal receptor degradation and reflects intrinsic differences between agonists.

Trafficking of agonist-stimulated G protein-coupled receptors (GPCRs) 2 classically proceeds through time-dependent steps starting by acute (within minutes) internalization of receptors from plasma membrane into cytoplasmic vesicular compartments and followed by receptor sorting into either recycling or degradative pathways, which usually take place within hours of agonist treatment. The long term effects of agonists on GPCR fate are dependent on both the nature of the agonist as well as the cellular environment and are believed to be important determinants of agonist effectiveness. Several mechanisms implicated in early GPCR trafficking have been described and involve phosphorylation of agonist-stimulated receptor by GPCR kinases and subsequent binding of arrestin proteins to phosphorylated receptor sites and to adapter proteins such as clathrin and AP-2, which form coated-pit receptor complexes (reviewed in Refs. 1 and 2). Based on the stability of the arrestin-receptor complexes, GPCRs have been separated into class A receptors, which favor the recycling pathway, and class B receptors, which recycle slowly and are instead destined for degradation (3).
The ubiquitin-proteasome pathway has been shown to be involved in the trafficking and degradation of some GPCRs (reviewed in Refs. 4 and 5). Ubiquitin, a 76-amino acid protein, is known to conjugate via a conserved three-step enzymatic reaction to lysine residues of proteins that are destined for proteasomal degradation (6). As such, ligand-dependent GPCR ubiquitination has been shown to impact the down-regulation of CXCR 4 , protease-activated receptor-2, V 2 vasopressin receptor, and ␤ 2 -adrenergic receptors, in some cases, by functioning as a sorting signal for lysosomal receptor targeting (7)(8)(9)(10).
Sphingosine 1-phosphate (S1P) is a secreted lipid that binds with nanomolar affinity to a family of five GPCRs, referred to as S1P [1][2][3][4][5] . Receptor coupling for S1P subtypes includes inhibition of adenylyl cyclase, activation of the small G proteins Rac and Rho, and activation p42/p44 mitogen-activated protein and AKT kinases and calcium release (reviewed in Ref. 11). The physiological functions of S1P extend to multiple systems, including cardiovascular, lymphoid, and auditory, and are being unveiled through both genetic and pharmacological * This work was supported by National Institutes of Health Grants RO1 AI055509 and NIMH 074404 (to H. R.) and Grant HL70694 (to T. H.) and a fellowship from the Genomics Institute of the Novartis Research Foundation (to P. J. G.-C.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. approaches (11,12). A well known outcome of S1P 1 receptor agonist administration in vivo is the inhibition of lymphocyte egress from lymph node and thymus. Originally described for FTY720 (13), this pro-drug, once converted into its active phosphate-ester form (FTY720-P) by cellular sphingosine kinase 2, acts as a potent and non-selective agonist (activates S1P receptors 1, 3, 4, and 5), which induces and maintains blockade of lymphocyte egress. Studies in the S1P 1 genetic knock-out mouse (14) and the discovery of S1P 1 -selective agonists (15,16), later singled out the S1P 1 receptor as the primary mediator of S1P signals that alter lymphocyte recirculation. In fact, we have provided evidence that administration of the selective S1P 1 agonist, SEW2871, discovered from high throughput screening, induces dose-dependent and reversible lymphopenia in mice, with onset kinetics and magnitude similar to the FTY720-P chiral analog, AFD-R (15). Additional studies, intended to compare S1P 1 agonist signaling in cell lines stably expressing S1P 1 , determined that SEW2871 recapitulates S1P effectors signaling and overlaps with S1P for key S1P 1 pocket interactions, although at lower potency (17). Interestingly, fate of the receptor was significantly different with different agonists, and while stimulation with the physiological ligand S1P or SEW2871 supported S1P 1 -GFP recycling, FTY-720-P-treated cells did not lead to recycling, suggesting the existence of liganddependent differences in receptor fate within the same cellular environment. In the present study, we have investigated whether differences in agonist-induced receptor ubiquitination could account for this discrepancy in fate.
Here, we provide biochemical and proteomic evidence that S1P 1 agonists induce recruitment of ubiquitin to S1P 1 , although by different magnitudes depending on the choice of agonist and despite similar ligands' efficacy at acute receptor activation. We found that there is enhanced efficacy of receptor ubiquitination by AFD-R relative to SEW2871 or S1P, and this is strongly associated with receptor sorting to lysosomes and receptor downregulation. Our model proposes that the extent of ubiquitin recruitment by distinct ligands, and thus receptor degradation, represents a ligand-regulated step in determining the fate of S1P 1 .

EXPERIMENTAL PROCEDURES
Materials-GTP␥S 35 was obtained from PerkinElmer Life Sciences. S1P was obtained from Biomol. The selective S1P 1 agonist, SEW2871, was from Maybridge. The S1P receptor agonist, AFD-R (the phosphate-ester of the prodrug amino alcohol AAL(R)) and AAL(S) were gifts from Dr. Volker Brinkmann (Novartis Pharma). Anti-GFP antibodies and the mannose 6-phosphate (M6P) receptor antibody were from Abcam, antiubiquitin P4D1 antibody from Santa Cruz Biotechnology (Santa Cruz, CA), and ERK antibodies from Cell Signaling. The proteasomal inhibitors (MG132 and lactacystin) were obtained from Calbiochem. Chloroquine and cycloheximide (CHX) were from Sigma-Aldrich.
Detection of S1P 1 -GFP and Evaluation of Ligand-dependent Down-regulation-Confluent cells expressing S1P 1 -GFP or vector-GFP control grown in 6-well plates were washed twice in ice-cold PBS, and lysates were obtained by incubation in radioimmune precipitation assay buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 1% Triton x-100) containing 1 mM NaVO 4 , 1 mM NaF, 0.5 M ␤-glycerol phosphate and protease inhibitor mixture (Roche Applied Science). Cellular lysates were cleared by centrifugation (16,500 ϫ g, 15 min), and the protein concentration of lysates supernatants was determined by the BCA (Pierce) method. Equal amounts of protein lysates were incubated overnight at 4°C with a monoclonal GFP antibody (1 g of antibody per 400 g of protein), followed by incubation (2 h, 4°C) with protein-A-Sepharose beads. The beads were recovered by centrifugation (10,000 ϫ g, 1 min) and washing: three times in radioimmune precipitation assay buffer: PBS (1:1) without protease inhibitors and twice in PBS. The beads were suspended in Laemmli buffer containing 2-mercaptoethanol and boiled for 10 min, and proteins in the beads were separated by SDS-PAGE using either 4 -12% gradient NuPage gels (Invitrogen) or 10% linear gels. The gels were transferred to polyvinylidene difluoride membranes, which were blocked in 5% milk and subsequently probed with a polyclonal GFP antibody (1:10,000; 1 h at room temperature) for detection of S1P 1 -GFP. Horseradish peroxidase-labeled goat anti-rabbit (1:5,000) antibodies were visualized by ECL chemiluminescence (Amersham Biosciences).
Agonist-induced down-regulation was measured in cells treated with 15 g/ml CHX to prevent new receptor synthesis. Cells were exposed to either 10 M SEW2871, 500 nM AFD-R, 500 nM S1P, or vehicle (0.1% fatty acid-free BSA) for the indicated times, and lysed as described above. S1P 1 -GFP downregulation was analyzed by comparing the density of immunoprecipitated S1P 1 -GFP bands in agonist versus vehicle-treated cells, quantified by scanning densitometry using Kodak-1D Application software.
The effect of proteasomal inhibition, chloroquine, or the S1P 1 competitive antagonist W146 (19) on AFD-R-stimulated S1P 1 -GFP down-regulation was studied by comparing intensity of S1P 1 -GFP bands in cells incubated with the respective agents (5 M MG132, 10 M lactacystin, 80 M chloroquine, and 10 M W146) for 30 min prior to and during incubation with AFD-R or vehicle for an additional 4 h.
Cells were transfected in 6-well plates with either 3, 2, or 1 g of receptor plasmid for 48 h using Lipofectamine 2000. Following transfection, cells were incubated with vehicle or 500 nM AFD-R for 1 h, and cellular lysates were obtained as described above. Receptor expression and AFD-R-stimulated receptor ubiquitination were determined by immunoblotting with an anti-HA (Bio-Rad) antibody at 1:1000 dilution.
Acute Receptor Activation Experiments-Activation of ERK phosphorylation (pERK) by S1P 1 agonists was determined by incubating cells with either 10 M SEW2871, 500 nM AFD-R, or 500 nM S1P for the indicated times. The conditions for determining activation of pERK were done as described previously (17). The potency (EC 50 ) and maximal responses of S1P, AFD-R, and SEW2871 in activating GTP binding were determined in S1P 1 -GFP membranes using GTP␥S 35 . Membrane preparation and conditions for binding were done as reported before (15), using 40-g membranes per well. Data analysis was performed using GraphPad Prism (San Diego, CA).
Imaging-Single S1P 1 -GFP cells grown in 0.2% gelatincoated coverslips were used to study ligand-induced localization with the late endosomal M6P-receptor marker. Incubation with agonists (500 nM AFD-R, 500 nM S1P, or 10 M SEW2871) was terminated by removal of medium and washing with PBS. Cells were fixed in 3.7% paraformaldehyde in PBS for 10 min, permeabilized in PBS/0.1% Triton X-100 (PBST) for 30 min, and blocked for 30 min in PBST containing 1% BSA and 5% normal goat serum. Primary antibody incubation (1:1000) was performed in blocking buffer overnight at 4°C. Secondary antibody (goat-anti-mouse Alexafluor-546) incubation was performed in blocking buffer for 30 min at room temperature. Coverslips were washed three times with PBS and mounted onto slides by using Gel Mount (Biomeda Corp.) mounting media. Cells were scanned with an Olympus BX61 scanning confocal fluorescence microscope. For detecting GFP, fluorescence was excited by using an argon laser at a wavelength of 488 nm, and the absorbed wavelength was detected at 510 -520 nm. For detecting Alexa Fluor 546, fluorescence was excited by using a helium-neon laser at a wavelength of 522 nm. In experiments using Lysotracker Red (Cambrex) dye (75 nM) was added to the medium 15 min before the end of agonist incubation. Cells were then washed and fixed with 3.7% paraformaldehyde in PBS for 10 min. For all experiments, images were processed with Adobe Photoshop 6.0.
Chromatography and Mass Spectrometry-AFD-R-stimulated ubiquitin recruitment to S1P 1 -GFP was determined using a scale-up GFP immunoprecipitation using identical conditions as above. Briefly, vehicle and AFD-R lysates isolated from five 150-mm plates per condition (vehicle or 500 nM AFD-R, 1-h incubation) were immunoprecipitated with anti-GFP (400 g of protein to 1 g of antibody), separated by SDS-PAGE (4 -12% NuPage gels), and the gel was subsequently stained by Colloidal Blue. In-gel stained proteins derived from three separate gel fragments (bottom, 73-85 kDa; medium, 100 -115 kDa; top, 150 -170 kDa, see Fig. 8) were cut out from either vehicle or AFD-R lanes and sent for LC-MS/MS analysis. The gel fragments were chosen by imaging against a matched AFD-R-induced "ladder-like" ubiquitin receptor complex (run in an adjacent well of the same gel and derived from the same lysates).
As controls, corresponding vehicle-only gel fragments were analyzed. The gel bands were excised and treated with 10 mM dithiothreitol to reduce disulfide linkages. Alkylation was performed with 55 mM iodoacetamide (Sigma-Aldrich) before digestion with trypsin (Promega) overnight at 37°C using an estimated (1:30) enzyme to substrate ratio in 50 mM ammonium bicarbonate. The LC separation was performed on a laser-pulled 100-m inner diameter C 18 column with a tip of Ͻ5 m that is also used as a nanoelectrospray emitter. Gradient elution was used with 0.1% formic acid/acetonitrile as the mobile phases, from 5% to 60% acetonitrile in 90 min, and then maintained for an additional 20 min with flow rates of ϳ200 nl/minute. The MS/MS analysis was performed on a linear ion mass spectrometer (LTQ, ThermoFisher Corp.). Datadependent scanning was used to maximize the number of peptides sequenced in the highly complex mixture. This mode of operation uses preset criteria to select unique peptides on-the-fly for undergoing MS/MS. Over 10,000 MS/MS spectra were obtained during the run. These were searched using Mascot (Matrix Science, Ltd.) and Sequest (University of Washington, WA) search engine software using the NCBInr (non-redundant data base). To improve searching efficiency, the taxonomic category was limited to mammalian proteins. Only peptides producing good quality fragmentation spectra and scoring higher than the threshold required for 95% confidence level for Mascot were used for protein identification.

RESULTS
We have shown previously (17) that stable HEK293 cells expressing C-terminal GFP-tagged human S1P 1 differ in trafficking pattern when stimulated with FTY720-P (which induces receptor degradation), compared with S1P or SEW2871 (which induce receptor recycling). To further investigate agonist-induced receptor fate differences, we used the same HEK293-S1P 1 -GFP cell line (18) to compare three agonists (AFD-R, SEW2871, and S1P) for stimulating 1) receptor down-regulation, 2) acute receptor signaling (GTP␥S 35 binding and pERK activation), and 3) short-and long-term receptor trafficking. Additionally, we augmented biochemical and pharmacological data with a proteomics identification approach in exploring possible mechanism(s) of agonist-induced S1P 1 fate.
Agonist-stimulated S1P 1 -GFP down-regulation was evaluated by immunoprecipitating the receptor with GFP antibodies (Fig. 1A). We could detect S1P 1 -GFP in transfected S1P 1 -GFP cells, but not in vector-GFP-expressing cells, as a band that migrated between 64 and 82 kDa, corresponding to the GFPtagged human S1P 1 (44-kDa S1P 1 plus 27-kDa GFP). Heavyand light-IgG chains were visible in both samples, and a GFP band (immediately above the light chain) was detected in vector-GFP-expressing cells.
Down-regulation was measured as the loss of receptor band to agonist stimulation in experiments in the presence of CHX. Treatment of S1P 1 -GFP cells with either AFD-R (500 nM), S1P (500 nM), SEW2871 (10 M), or vehicle (0.1% fatty acid-free BSA) for 4 h led to the finding that AFD-R down-regulated approximately half (44 Ϯ 2%) of total S1P 1 -GFP expression versus vehicle-treated cells (Fig. 1B, bottom). On the other hand, there were no significant differences in S1P 1 -GFP expression in cells incubated for the same time with either SEW2871 or S1P.
Because protein ubiquitination has been implicated in regulating GPCR trafficking and fate, we employed an antibody that recognizes ubiquitinated substrates (P4D1) and probed the same GFP immunoprecipitates used for determining agonist-induced receptor down-regulation. Fig. 1B shows that agonist incubation was associated with increases in the ubiquitination of high (ϳ115-180 kDa) molecular mass protein(s), which we have referred to as the "ubiquitinated receptor complex." Ubiquitinated receptor complex was dependent upon agonist stimulation in receptor-transfected cells only, because it was not found to be associated with either vehicle-treated S1P 1 -GFP cells or vector-GFP stimulated (using 500 nM S1P) cells. Interestingly, at the 4-h agonist incubation studied, the ubiquitinated receptor complex was found to differ in magnitude depending on agonist utilized, with AFD-R stimulating significantly (1.8-and 2.5-fold) higher ubiquitination relative to S1P and SEW2871, respectively.
A time course (0 -6 h) of agonist-induced receptor downregulation and associated ubiquitinated receptor complex is shown in Fig. 2. Again, depending on choice of agonist, two main differences in the measures were observed. First, only cells receiving AFD-R (500 nM) displayed significant receptor downregulation relative to untreated cells. Down-regulation by AFD-R was apparent at 4 h (with a loss of approximately half of the total S1P 1 expression), and continued at 6 h. In contrast, SEW2871 (10 M) or S1P (500 nM) incubations did not significantly alter S1P 1 -GFP expression up to 6 h, relative to unstimulated cells, respectively. Second, and consistent with results from Fig. 1, AFD-R stimulated the highest magnitude of ubiq-FIGURE 1. AFD-R promotes S1P 1 -GFP down-regulation and enhanced protein ubiquitination. Equal amounts of 293-vector-GFP or 293-S1P 1 -GFP cell lysates were immunoprecipitated (IP) and immunoblotted (IB) with GFP antibodies to detect S1P 1 -GFP expression. A, S1P 1 -GFP expression was detected as a band running between 64 and 82 kDa (lane 2), whereas vector-GFP cells without insert did not express receptor (lane 1). B, vector GFP or S1P 1 -GFP cells were incubated with S1P 1 agonists: AFD-R (500 nM), S1P (500 nM), SEW2871 (10 M), or vehicle (0.1% fatty acid-free BSA) for 4 h in the presence of CHX. Lower panel, GFP immunoprecipitation-immunoblotting experiments from equivalent protein amounts revealed AFD-R-mediated S1P 1 -GFP down-regulation versus vehicle (Veh) treatment (n ϭ 3). Upper panel, membranes were subsequently probed with P4D1 antibody for detection of agonist-induced ubiquitination. Note that P4D1 only recognized protein ubiquitination in ligand-stimulated, S1P 1 -GFP-expressing cells. The positions of molecular mass markers are indicated on the right (in kilodaltons). C, densitometric analysis of S1P 1 -GFP expression (left graph) and agonist-induced ubiquitinated-receptor complex (right graph) in cells treated with S1P 1 agonists for 4 h. *, p Ͻ 0.05 relative to vehicle alone (left graph); *, p Ͻ 0.05 versus AFD-R treatment (right graph). Bars represent the mean Ϯ S.E. of three independent experiments. FIGURE 2. Time course of agonist-stimulated ubiquitin receptor complex and receptor down-regulation. 293-S1P 1 -GFP cells were incubated for the indicated times with either 500 nM AFD-R, 500 nM S1P, or 10 M SEW2871 in the presence of CHX. A, equal amounts of protein lysates were immunoprecipitated (IP) with a GFP antibody, and the membranes were immunoblotted (IB) with either anti-GFP (lower panels) to detect receptor-GFP expression or P4D1 (upper panels) for determination of agonist-induced ubiquitinated receptor complex. B, densitometry analysis of receptor-GFP and ubiquitinated receptor complex following either AFD-R (f), SEW2871 (OE), or S1P () stimulation. Time courses were plotted as the fraction of each agonist maximum, relative to unstimulated (time 0) cells. The graphs are from a representative experiment that was repeated twice with identical conditions. uitinated receptor complex throughout the entire time course, relative to SEW2871 or S1P treatments, respectively. For all agonists, the time courses of ubiquitin receptor complex were biphasic, and of similar onsets (30 min), yet AFD-R induced sustained higher ubiquitination levels throughout most of the study, in comparison to SEW2871 or S1P.
The time course indicated that the agonist's maximal ubiquitination of cellular substrate(s) occurred at ϳ1-h incubation; thus we compared the concentration dependence of agonistinduced ubiquitinated signaling in cells treated for 1 h with each agonist. Fig. 3 shows that AFD-R was 1,000-and 10,000fold more potent in stimulating ubiquitin receptor complex formation than S1P and SEW2871, respectively, whereas no effect on amount of receptor (aside from the 50 nM S1P lane, which reflects a minor loading defect) was apparent at the 1-h incubation time. The potencies for AFD-R, S1P, and SEW2871 in stimulating ubiquitination at 1 h were 0.5 nM, 0.3 M, and 2.5 M, respectively. In addition, Fig. 3B indicates that AFD-R was a full agonist in activating receptor-complex ubiquitination relative to S1P and SEW2871 responses.
The specificity of AFD-R-induced down-regulation/ubiquitinated receptor complex was studied using AAL-R and AAL-S. The former compound is efficiently converted into AFD-R by the action of sphingosine kinase-2 (20), whereas the latter, the S-enantiomer of AAL, is not a sphingosine kinase substrate and does not get converted into AFD-R. Fig. 4A shows that S1P 1 -GFP down-regulation and ubiquitin receptor complex were only associated with AFD-R and AAL-R treatments (4 h), whereas AAL-S incubation had no effect on either measure. Additional proof of AFD-R effects on receptor down-reg-ulation and associated ubiquitin receptor complex came from experiments in which a selective S1P 1 competitive antagonist (W146) possessing potent in vivo blocking activity (19) was used. This antagonist blocks SEW2871-mediated receptor internalization in HEK293-S1P 1 -GFP cells (19) and abolishes intracellular calcium release elicited by AFD-R in CHO-S1P 1 cells. 3 In experiments using W146 (Fig. 4A), AFD-R alone (50 nM, 4 h) led to significant S1P 1 -GFP down-regulation and ubiquitination. Preincubation with 10 M W146 led to the complete  . Specificity of AFD-R-induced ubiquitinated receptor complex and S1P 1 -GFP down-regulation in 293-S1P 1 -GFP cells. 293-S1P 1 -GFP cells were incubated with either 0.1% fatty acid-free BSA (Veh) or drugs (500 nM AFD-R, the active S1P agonist pro-drug AAL-R or its inactive stereoisomer AAL-S) for 4 h, and equal protein lysates immunoprecipitated (IP) and immunoblotted (IB) with GFP antibodies (lower panels) and P4D1 (upper panels) for detection of agonist-induced receptor down-regulation and ubiquitinated receptor complex, respectively. A, note that only AFD-R and AAL-R stimulated receptor down-regulation and induce ubiquitinated receptor complex, relative to vehicle, whereas the inactive compound (AAL-S) did not. W146, a selective S1P 1 antagonist, reversed the AFD-R-mediated actions on downregulation and ubiquitin receptor complex. Results of a representative experiment (4 -12% gel) are shown (n ϭ 3). B, AFD-R-induced S1P 1 -GFP downregulation and ubiquitinated receptor complex are dependent on concentration of AFD-R (n ϭ 3). C, the potency of AFD-R-stimulated S1P 1 down-regulation was determined by plotting the intensity of S1P 1 -GFP versus agonist concentration (run in a 10% gel) as a percentage of S1P 1 -GFP density in unstimulated cells. Ϫ, refers to vehicle.
blockade of AFD-R-stimulated S1P 1 -GFP down-regulation and abolished detection of the ubiquitinated receptor complex.
AFD-R mediated S1P 1 -GFP down-regulation was also found to be concentration-dependent (Fig. 4B). The potency for inducing receptor down-regulation (AFD-R incubation was 4 h) was determined to be 7.1 nM. For these experiments, AFD-R-stimulated ubiquitin receptor complex was found to be ϳ10fold more potent than receptor down-regulation, consistent with results obtained at 1-h AFD-R incubation (Fig. 3).
The discrepancy in agonist action in stimulating ubiquitin receptor complex and receptor down-regulation could be explained in part by differences in agonist intrinsic activities. To explore this possibility, we compared agonists for activating proximal receptor pathways such as pERK activation and GTP␥S 35 binding. Fig. 5A shows the kinetics of SEW2871-, AFD-R-, and S1P-activated pERK. Agonists stimulated ERK phosphorylation in a time-dependent manner, and, despite subtle differences in onset and time of maximal activation, neither agonist activated pERK beyond 30 min. Agonist-stimulated GTP␥S 35 binding was found to be concentration-dependent (Fig. 5B), and although there were differences in potency among agonists (SEW2871, 0.42 M; AFD-R, 42 nM; and S1P, 0.9 nM), all agonists were found to be similarly efficacious (SEW2871, 90%; AFD-R, 100; and S1P, 90%; percent own maximal responses).
We next studied agonist-stimulated trafficking of S1P 1 -GFP by confocal microscopy, using two cytoplasmic vesicle markers: the mannose 6-phosphate (M6P) receptor, which associates with late endosomes (21), and Lysotracker Red, a pH-sensitive dye, which labels lysosomes (22). Two time points were chosen, a 1-h protocol, which corresponded to maximal agonist-induced ubiquitin receptor complex, and 4 h, where agonist differences in down-regulation were first noticed. Fig. 6A shows that at 1 h, all agonists internalized S1P 1 -GFP from membrane to cytoplasmic vesicles, which were found to be colocalized in part with M6P-positive vesicles. On the contrary, no colocalization was observed between S1P 1 -GFP vesicles and Lysotracker Red-stained vesicles at the 1-h treatment, irrespective of agonist utilized (Fig. 6B). The effect of 4 h agonist incubation on receptor-lysosome colocalization is shown in Fig. 6C. Here, S1P 1 -GFP vesicles internalized by AFD-R were found to completely colocalize with vesicles stained by Lysotracker Red, whereas SEW2871 incubation did not stimulate appreciable GFP-Lysotracker Red vesicle colocalization. A mixed population of Lysotracker Red-positive and -negative S1P 1 -GFP vesicles was observed in S1P-treated cells at the 4-h incubation.
Trafficking of GPCRs within cells requires fusion of internalized receptor vesicles with those of acidic compartments (endosomal and lysosomal vesicles). We used the weak base, chloroquine, to determine the requirement of membrane fusion in AFD-R-stimulated ubiquitin receptor complex formation, and induced receptor down-regulation. In addition, because ubiquitination of proteins is known to lead to their degradation by the proteasome, we assessed the effect of two proteasomal inhibitors (MG132 and lactacystin) on stability of AFD-R-induced S1P 1 -GFP ubiquitin receptor complex and receptor down-regulation. Fig. 7 shows that 4 h after AFD-R stimulation, chloroquine inhibited the formation of receptor-ubiquitin complex and abolished AFD-R-mediated receptor down-regulation. Inhibition of AFD-R ubiquitin receptor complex by chloroquine was seen as early as 1 h (not shown). Incubation with either of the proteasomal inhibitors was shown to increase the molecular weight of the ubiquitinated receptor complex (relative to its migration versus AFD-R alone), while resulting in partial, but not significant inhibition of AFD-R-induced receptor down-regulation (by 35 Ϯ 24% with MG132) and (by 26 Ϯ 25% with lactacystin).
There is precedent for ligand-dependent GPCR down-regulation via recruitment of ubiquitin chains to lysine receptor residues; thus we looked for evidence of a ligand-induced receptor ubiquitination complex. Because the P4D1 antibody was not useful in our hands for immunoprecipitation, we explored this question using an LC-MS/MS proteomics scale-up method, and chose a 1-h AFD-R/vehicle incubation protocol (determined in previous experiments to be optimal for agonist P4D1 signaling). The results from LC-MS/MS analysis are summarized in Table 1. In all three fragments analyzed, there was detection of several peptides from human S1P 1 receptor and several peptides from GFP, in both the treated and untreated groups. The peptides detected from human S1P 1 spanned proximal, transmembrane, and distal regions of the FIGURE 5. Kinetics of activation of proximal receptor pathways by S1P 1 agonists in S1P 1 -GFP cells. A, cells were incubated with agonists for the indicated times, and Western blots were performed with phospho-specific ERK (pERK) antibodies using 20 g of total protein per lane. Membranes were reprobed with total ERK (ERK) to confirm equal loading. Two independent experiments are represented. B, membranes expressing S1P 1 -GFP were tested for agonism in a GTP␥S 35 binding assay. SEW2871 () was compared with AFD-R (f) and S1P (OE), and the results are expressed as counts per minute (cpm). Each point represents the average of three independent experiments performed in triplicate. FIGURE 6. S1P 1 -GFP trafficking following 1-or 4-h incubation with agonists. Cells expressing S1P 1 -GFP were incubated with either 500 nM AFD-R, 500 nM S1P, or 10 M SEW2871 for 1 h or 4 h. A, receptor-GFP (green) distribution was visualized by GFP fluorescence and distribution within late endosomes was detected using an antibody to M6P-Receptor (red). Note that, in unstimulated cells, S1P 1 -GFP fluorescence is primarily localized to the plasma membrane, whereas agonist incubation results in internalization of receptor into cytoplasmic vesicles, some of which are found to colocalize with M6P-receptor vesicles (merged images, yellow). B, receptor-lysosome interaction was determined using the pH-sensitive marker Lysotrackerா Red (LT-Red) as shown under "Experimental Procedures." Note the lack of colocalization between agonist-internalized GFP vesicles and LT-Red vesicles at 1-h agonist incubation. C, at 4-h AFD-R incubation, there was complete colocalization of S1P 1 -GFP vesicles with LT-Red vesicles, whereas internalized S1P 1 -GFP by SEW2871 and S1P did not show complete LT-Red colocalization. Shown are representative images of one of three independent experiments. receptor, and, as expected of tryptic fragmentation, all contained C-terminal KR residues.
The main difference between vehicle and AFD-R treatments was the identification of tryptic peptides corresponding to ubiquitin, found only in AFD-R treated cells. In fact, all three gel fragments isolated from the AFD-R-treated sample contained ubiquitin peptides (nearly 50% of ubiquitin content) relative to none detected in vehicle. The proteomics data corroborated the immunoblotting data and strongly suggests that ubiquitin is recruited to S1P 1 -GFP upon agonist treatment.
The scale-up proteomics method provided additional information about S1P 1 -GFP expression. We noticed that in both vehicle-and AFD-R-treated groups, S1P 1 -GFP was detected as a single running band (monomer S1P 1 -GFP running at ϳ71 kDa), and as a higher molecular mass multimer (likely a dimer, ϳ150 kDa). Besides proteomics, the putative receptor dimer was also resolved in the scaled-up preparation by immunoprecipitation-immunoblotting experiments (Fig. 8).
Additional supporting evidence of ligand-induced S1P 1 ubiquitination is shown in experiments in which HA-tagged S1P 1 (HA fused to the N terminus) was transiently transfected into 293-cells. In supplemental Fig. S1, HA-tagged S1P 1 was found to undergo AFD-R-stimulated HA-S1P 1 ubiquitination, relative to vehicle-treated transfectants, as shown by the presence of a high molecular weight smear when immunoblotted with the HA antibody.

DISCUSSION
There is increasing evidence linking agonist-promoted ubiquitination of GPCRs with receptor sorting and degradation. Examples of ubiquitinated GPCRs include the and ␦ opioid receptors (23), the CXCR 4 chemokine receptor (7), the ␤ 2 -adrenergic receptor (9), the V 2 vasopressin receptor (10), the sst 3 somatostatin receptor (24), and the protease-activated receptor-2 (8). Although these studies have demonstrated the involvement of the ubiquitin-proteasome pathway in regulat-   Table 1. A representative silver-stained gel of GFP immunoprecipitates and matched Western P4D1 immunoblot (IB) showing expression of S1P 1 -GFP in relation with AFD-R-induced ubiquitinated bands are shown. Heavy and light-IgG chains are shown for comparison.

TABLE 1 AFD-R induced recruitment of ubiquitin to S1P 1 -GFP
Lysates from 293-S1P 1 -GFP cells that were incubated with either vehicle or 500 nM AFD-R for 1 h were immunoprecipitated with a GFP antibody and separated by SDS-PAGE. Coomassie-stained gel bands (ϳ1 mm thick) were isolated from either vehicle or AFD-R lanes and analyzed by LC-MS/MS. The isolated bands were chosen based on matching against steps of an ubiquitin ladder, which was obtained from P4D1 immunoblotting of AFD-R GFP immunoprecipitates, and run in an adjacent lane of the gel (see Fig. 8). Indicated for each protein are the accession number (NCBI no.) and the tryptic peptides identified (ϩ) in each treatment group.

Protein NCBI no.
Vehicle AFD-R Tryptic peptides identified a Absent from analysis.
Mapping Pathways Downstream of S1P 1 MARCH 9, 2007 • VOLUME 282 • NUMBER 10 ing trafficking and fate of receptors, little is known regarding agonist efficacy of receptor ubiquitination and signaling and fate.
In the present study we used biochemical studies and proteomics to differentiate between two types of agonists with respect to S1P 1 fate. Using stable HEK293-S1P 1 -GFP cells, we propose that there are supraphysiological agonists, such as AFD-R, which promote S1P 1 down-regulation, and physiological-like agonists, such as S1P and SEW2871, which do not promote significant S1P 1 down-regulation when incubated continuously (up to 6 h; Figs. 1 and 2) in the presence of CHX.
Combining proteomics and immunoprecipitation studies, we determined that following internalization, S1P 1 agonists commonly stimulate ubiquitination of the receptor, which persists throughout trafficking to late endosomes. The main finding was the strong association between the supraphysiological agonist-induced receptor down-regulation and lysosomal targeting and its intrinsic ability to stimulate ubiquitin-receptor immunocomplexes at relatively low concentrations. In fact, the threshold concentration to achieve optimal S1P 1 ubiquitination (determined in the time course to be 1 h for all agonists, Fig.  2), was found to be significantly left-shifted (by 3 and 4 logs) in AFD-R-treated cells versus S1P and SEW2871, respectively (Fig. 3); thus, making AFD-R a high potency full agonist in recruiting ubiquitin S1P 1 -GFP relative to S1P or SEW2871. We found that receptor down-regulation occurred at 10-fold higher AFD-R concentration than optimal receptor ubiquitination, suggesting that a certain level of ubiquitin-receptor recruitment may be necessary for receptor degradation in lysosomes (i.e. beyond that stimulated by SEW2871 or S1P herein). Furthermore, it is interesting that receptor ubiquitination by AFD-R was readily detectable at a concentration as low as 0.5 nM, similar to the reported affinity of its chiral analog, FTY720-P, in activating recombinant systems (13), and within the reported EC 50 value for AFD-R in vivo actions on lymphocyte sequestration and CD69 thymocyte maturation, reported to be 0.7 nM (25).
Specificity of AFD-R functions on ubiquitin recruitment and down-regulation came from studies with the inactive (S)-isomer of the agonist AAL (26), which had no effect on receptor ubiquitination or down-regulation (Fig. 4A). Further, the S1P 1selective antagonist W146 (19) was able to reverse the actions of AFD-R on both measures. In addition, we were able to reproduce AFD-R-dependent receptor ubiquitination in transiently transfected HEK293-cells with HA-tagged receptors (see supplemental Fig. S1), ruling out an artifact induced by the S1P 1 -GFP fusion construct.
These differences in agonist-mediated S1P 1 fate and/or extent of receptor ubiquitination were found to be independent of agonist intrinsic activity (Fig. 5), because all agonists reached comparable maximal activation profiles in GTP binding and similar kinetics in activating ERK phosphorylation. The potency values calculated in the present study for activating agonist GTP binding and pERK kinetics are within reported values for drug actions in other recombinant systems and primary human umbilical vein endothelial cells (13,15,17). This suggested that agonist-stimulated receptor fate decisions seem likely to take place distal to receptor activation. In fact, incuba-tion with either agonist for 1 h (Fig. 6, A and B), which results in seemingly comparable rates of S1P 1 -GFP internalization from membrane to cytoplasmic vesicles, demonstrated similar compartmentalization at late endosomes. Late endosomes are specialized organelles reported to be essential for receptor sorting to either the recycling or the degradative pathways (27). One of the proposed mechanisms that dictate sorting relies on the relatively low pH of late endosomes, which has been shown to promote dissociation of ligand-receptor complexes; thus it is possible that AFD-R-bound receptor may be less susceptible than SEW2871 or S1P to low endosomal pH, and that ligandreceptor off-rates may constitute important determinants of receptor fate. Additionally, GTP␥S 35 binding experiments using membranes from cells pretreated with AFD-R for 4 h showed a complete loss of further agonist GTP␥S 35 binding, as opposed to vehicle-treated cells, which retain agonist binding (data not shown), suggesting that the degree of down-regulation resulting from prolonged AFD-R treatment is enough to impair receptor function. Lack of agonist effect in long term supraphysiological agonist-pretreated cells by means of downregulation may be separated from agonist-mediated desensitization, which seems to take place at relatively more acute agonist incubation times. For instance, a 10-min preincubation of human umbilical vein endothelial cells with either S1P or FTY720-P was reported to cross-desensitize receptor-mediated intracellular calcium release (13), likely through S1P 1 . Because S1P does not support significant down-regulation, yet is able to desensitize, the data suggest that receptor ubiquitination and desensitization may represent separate mechanisms, which depending on agonist, could act independently in terminating receptor function.
Confocal microscopy studies demonstrated differences in agonist-mediated receptor compartmentalization at a late (4 h) incubation time (Fig. 6C), with AFD-R-internalized receptor being completely lysosomal. We used chloroquine to study the relationship between ubiquitination and lysosomal targeting. Chloroquine enters the lysosome and neutralizes the H ϩ -ion gradient, which results in the inhibition of lysosomal proteases that function optimally at acidic pH. The finding that chloroquine abolished AFD-R-mediated S1P 1 down-regulation and S1P 1 ubiquitination, without affecting internalization (not shown), indicate that 1) internalization is a prerequisite for ligand-induced ubiquitination, 2) membrane fusion is a requirement for lysosomal degradation, and 3) a strong relationship exists between magnitude of receptor ubiquitination and targeting for lysosomal degradation by supraphysiological agonism. Consistent with the notion that receptor fate decisions by agonists take place distal of receptor activation, such as endosomes, we found that chloroquine did not affect the kinetics or magnitude of a dynamic proximal receptor pathway, such as agonist-stimulated ERK phosphorylation (not shown).
In most cases, increasing the number of ubiquitin moieties that can be covalently tagged onto a protein has been shown to modulate the trafficking signal generated. For example, monoubiquitination usually signals internalization, whereas polyubiquitination tags them for destruction. In the case of S1P 1 , the presence of a smear of high molecular mass (from ϳ115 to 180 kDa) on Western blots suggests receptor polyu-biquitination, and our studies indicate that the amount of polyubiquitinated S1P 1 can be modulated depending on the choice of agonist.
The proteomic analysis of the ubiquitin ladder stimulated by AFD-R revealed the identity of ubiquitinated S1P 1 -GFP, as compared with vehicle control. Interestingly, we were able to resolve monomer and multimer versions of S1P 1 in the scale-up proteomic preparation. We attribute detection of the higher molecular weight S1P 1 -GFP version to the scale-up method, rather than a gel-running artifact, because the ligand-induced "ubiquitin ladder" pattern ran unchanged from previous results in which only monomers were detected, and the multimer S1P 1 -GFP version was also found to be ubiquitinated by AFD-R (see Fig. 8). Detection of multimeric GPCR versions by similar immunoprecipitation methods is not uncommon, as in the case of the ␤ 2 -adrenergic (9), V 2 vasopressin (28), and recently shown for S1P 1 (29).
There seems to be GPCR species differences in the utilization of the ubiquitin-proteasome pathway for receptor trafficking. For instance, ␤ 1 adrenergic receptors are entirely resistant to ubiquitination (30), whereas the ␤ 2 receptor is dependent on the ubiquitin-proteasome for internalization and trafficking (9). In some cases, such as the V 2 vasopressin receptor, blocking the proteasome is a requirement for detection of ligand-induced receptor ubiquitination (10). The use of proteasomal inhibitors in this study suggested that preserving ubiquitinated receptor (as seen by the shift in electrophoretic mobility of the ubiquitinated smear in Western blots), has a modest effect on AFD-R-induced receptor down-regulation. Because MG132 increased dose-dependently the accumulation of polyubiquitinated receptor in the absence of ligand (not shown), it is likely that the proteasome may also be involved in basal receptor turnover, as is the case for opioid and ␦ subtypes (23).
Our model suggests the involvement of a ligand-dependent ubiquitin-stimulated lysosomal targeting pathway that degrades S1P 1 . Accumulation of polyubiquitinated receptor by proteasome inhibition alone also suggests a ligand-independent pathway that may be responsible for maintaining balance between synthesis and degradation at equilibrium.
Taken together, these results strongly suggest that ligandinduced S1P 1 ubiquitination serves as a sorting signal for lysosomal receptor degradation and that ubiquitin recruitment to S1P 1 can be down-modulated using in vivo active physiologicallike agonists, such as SEW2871. We have shown recently that a selective competitive S1P 1 antagonist (19) reversed SEW2871mediated lymphopenia in vivo without having a measurable effect on lymphocyte recirculation when tested alone. In addition, two photon studies indicated that S1P 1 antagonism was able to restore SEW2871-arrested lymphocyte movement in lymph node medulla. These results, which have now been reproduced by an independent group using a chemically distinct S1P 1 -competitive antagonist (31), collectively disfavor functional antagonism as the mechanism of S1P 1 -mediated lymphopenia. Compelling evidence exists for FTY720-P actions on S1P 1 -rich endothelium (as opposed to lymphocytes that express few cell surface S1P 1 receptors), and a study by Singer et al. (32) demonstrated that mice treated with FTY720-P up-regulate endothelial junctional proteins (CD31, ␤-catenin, and ZO-1) and S1P 1 receptor expression in lymph node, presumably leading to an increased endothelial barrier integrity. Consistent with an endothelial target for S1P 1 agonists, S1P 1 antagonist studies have demonstrated a prominent role for S1P-S1P 1 tone in the maintenance of lung vascular endothelial integrity (19,31), suggesting that minimalist S1Plike agonists that do not significantly down-regulate S1P 1 may be better suited in long term studies to preserve lung endothelium receptor reserve. Finally, both SEW2871 (15,17) and S1P (13) define the minimal signaling requirements for inducing and maintaining reversible S1P 1 -induced lymphopenia, yet neither lead to significant S1P 1 down-regulation. These data suggest that AFD-R and SEW2871/S1P alter receptor fate and ubiquitination to different extents and that S1P 1 degradation is not an essential shared downstream outcome of agonist action; therefore, it is not essential for induction of lymphopenia and therapeutic efficacy.