Comparison of Intrinsic Activities of the Putative Sphingosine 1-Phosphate Receptor Subtypes to Regulate Several Signaling Pathways in Their cDNA-transfected Chinese Hamster Ovary Cells*

We examined the actions of sphingosine 1-phosphate (S1P) on signaling pathways in Chinese hamster ovary cells transfected with putative S1P receptor subtypes, i.e.Edg-1, AGR16/H218 (Edg-5), and Edg-3. Among these receptor-transfected cells, there was no significant difference in the expressing numbers of the S1P receptors and their affinities to S1P, which were estimated by [3H]S1P binding to the cells. In vector-transfected cells, S1P slightly increased cytosolic Ca2+ concentration ([Ca2+] i ) in association with inositol phosphate production, reflecting phospholipase C activation; the S1P-induced actions were markedly enhanced in the Edg-3-transfected cells and moderately so in the AGR16-transfected cells. In comparison with vector-transfected cells, the S1P-induced [Ca2+] i increase was also slightly enhanced in the Edg-1-transfected cells. In all cases, the inositol phosphate and Ca2+ responses to S1P were partially inhibited by pertussis toxin (PTX). S1P also significantly increased cAMP content in a PTX-insensitive manner in all the transfected cells; the rank order of their intrinsic activity of S1P receptor subtypes was AGR16 > Edg-3 > Edg-1. In the presence of forskolin, however, S1P significantly inhibited cAMP accumulation at a lower concentration (1–100 nm) of S1P in a manner sensitive to PTX in the Edg-1-transfected cells but not in either the Edg-3 or AGR16-transfected cells. As for cell migration activity evaluated by cell number across the filter of blind Boyden chamber, Edg-1 and Edg-3 were equally potent, but AGR16 was ineffective. Thus, S1P receptors may couple to both PTX-sensitive and -insensitive G-proteins, resulting in the selective regulation of the phospholipase C-Ca2+ system, adenylyl cyclase-cAMP system, and cell migration activity, according to the receptor subtype.

We examined the actions of sphingosine 1-phosphate (S1P) on signaling pathways in Chinese hamster ovary cells transfected with putative S1P receptor subtypes, i.e. Edg-1, AGR16/H218 (Edg-5), and Edg-3. Among these receptor-transfected cells, there was no significant difference in the expressing numbers of the S1P receptors and their affinities to S1P, which were estimated by [ 3 2؉ ] i increase was also slightly enhanced in the Edg-1-transfected cells. In all cases, the inositol phosphate and Ca 2؉ responses to S1P were partially inhibited by pertussis toxin (PTX). S1P also significantly increased cAMP content in a PTX-insensitive manner in all the transfected cells; the rank order of their intrinsic activity of S1P receptor subtypes was AGR16 > Edg-3 > Edg-1. In the presence of forskolin, however, S1P significantly inhibited cAMP accumulation at a lower concentration (1-100 nM) of S1P in a manner sensitive to PTX in the Edg-1-transfected cells but not in either the Edg-3 or AGR16-transfected cells. As for cell migration activity evaluated by cell number across the filter of blind Boyden chamber, Edg-1 and Edg-3 were equally potent, but AGR16 was ineffective. Thus, S1P receptors may couple to both PTX-sensitive and -insensitive G-proteins, resulting in the selective regulation of the phospholipase C-Ca 2؉ system, adenylyl cyclase-cAMP system, and cell migration activity, according to the receptor subtype.

H]S1P binding to the cells. In vector-transfected cells, S1P slightly increased cytosolic Ca 2؉ concentration ([Ca 2؉ ] i ) in association with inositol phosphate production, reflecting phospholipase C activation; the S1P-induced actions were markedly enhanced in the Edg-3transfected cells and moderately so in the AGR16transfected cells. In comparison with vector-transfected cells, the S1P-induced [Ca
Sphingosine 1-phosphate (S1P), 1 one of the sphingolipid me-tabolites, has recently been suggested to affect a variety of cellular processes (1,2). These cellular responses elicited by S1P have first been ascribed to the intracellular action of the lipid, because S1P accumulated in the cells in response to some kinds of cytokines, and moreover, S1P induced Ca 2ϩ mobilization in a cell-free system (3)(4)(5). On the other hand, these S1Pinduced responses are also accompanied by the stimulation of several early signaling events that are usually regulated by cell-surface receptors. These signaling events include activation of PLC (6 -9), an increase in [Ca 2ϩ ] i (10 -12), regulation of adenylyl cyclase (6,9,10,13), and Rho activation (14,15). The presence of the latter mechanism has been supported by the recent identification of several cDNAs encoding G-protein-coupled receptors for S1P, i.e. Edg-1, AGR16/H218, and Edg-3 (16 -23).
The transfection experiments of these S1P receptor subtypes demonstrated that these putative S1P receptors can actually couple to multiple signaling pathways. For example, transfection of Edg-1 induced the inhibition of cAMP accumulation in HEK293 cells (22), Sf9 cells (18), and Chinese hamster ovary (CHO) cells (20), extracellular signal-regulated kinase activation in COS-1 cells (19), COS-7 cells (18), and CHO cells (20), and activation of Rho, resulting in a morphological change in HEK293 cells (19,22). The transfection of Edg-1 also caused activation of PLC and [Ca 2ϩ ] i increase in CHO cells and HEL cells (20), although the overexpression of Edg-1 in HEK293 cells and in Sf9 cells failed to affect the S1P-induced activation of PLC-Ca 2ϩ system (18,22). The expression of Edg-3 and AGR16 resulted in the activation of a serum response elementdriven transcriptional reporter gene in Jurkat cells and the stimulation of Ca 2ϩ flux in Xenopus oocytes in response to S1P (17). Coupling of Edg-3 (21) and AGR16 (23) to the Ca 2ϩ signaling has recently been confirmed in CHO cells and K562 cells.
Thus, the previous transfection experiments suggest the involvement of these putative S1P receptor subtypes in the regulation of multiple signaling pathways. However, in these experiments, the responses to S1P were observed in the different species of cells, and the receptor numbers expressed in the cells have not always been estimated. Thus, it is difficult to evaluate and compare the intrinsic activity of the respective receptor subtype to regulate a specific signaling pathway from the previous transfection experiments. In the present study, we prepared the CHO cells that permanently express the respective S1P receptor subtype to a comparable level. This made it possible to compare their intrinsic activity to regulate the respective signaling pathway. Our data suggested that S1P selectively regulates multiple signaling pathways according to the receptor subtype.

EXPERIMENTAL PROCEDURES
Materials-1-Oleoyl-sn-glycero-3-phosphate (lysophosphatidic acid) and D-erythro-sphingosine were purchased from Sigma; sphingosine 1-phosphate (S1P) was purchased from Cayman Chemical Co.; sphingosylphosphorylcholine was purchased from Biomol Research Laboratories, Inc.; Fura 2/AM was purchased from Dojindo (Tokyo); and myo-[2-3 H]inositol (23.0 Ci/mmol) and [ 3 H]sphingosine (20.0 Ci/mmol) were purchased from American Radiolabeled Chemicals, Inc. U73122 was generously provided by the The Upjohn Co. [ 3 H]S1P was enzymatically synthesized from [ 3 H]sphingosine by sphingosine kinase-catalyzed phosphorylation as described previously (12). [ 3 H]S1P was separated on silica gel 60 high performance thin-layer chromatography plates (Merck) in a solvent system of butanol-water-acetic acid (3:1:1). By this method, we could obtain [ 3 H]S1P with the same specific activity as the labeled sphingosine. The sources of all other reagents were the same as described previously (21, 24 -27) Cell Cultures-CHO cells that had been transfected with the pEFneo empty vector (28) or a pEFneo S1P receptor expression vector were cultured in DMEM containing 10% (v/v) fetal calf serum (Life Technologies, Inc.) in a humidified air/CO 2 (19:1) atmosphere. The cells were maintained on 10-cm dishes for inositol phosphate response, [Ca 2ϩ ] i response, and adenylyl cyclase activity and on 12-multiplates for measurements of cAMP content and S1P receptor binding unless otherwise specified. Twenty-four h before the experiments, the medium was changed to fresh DMEM (without serum) containing 0.1% (w/v) bovine serum albumin (fraction V). In the case of the inositol phosphate response, the medium was changed to the inositol-free DMEM containing 20 Ci of [ 3 H]inositol (in 6 ml) and 0.1% (w/v) bovine serum albumin (fraction V). PTX treatment of the cells was performed by adding the toxin (100 ng/ml) to the medium 24 h before the experiments. Unless otherwise specified, the cells were harvested from the dishes and used in suspension for inositol phosphate and Ca 2ϩ responses, and the cells attached to the plates without cell harvest were used for the cAMP response and S1P receptor binding.
Isolation of cDNAs for S1P Receptors and Construction of Expression Plasmid--The cDNAs for putative S1P receptors were cloned by reverse transcription-polymerase chain reaction for Edg-1 (29) and AGR16 (30) (H218, an almost identical cDNA to AGR16, differs from AGR16 in only 2 nucleotides but has the same amino acid sequence (31)) from the total RNA of rat brain. Edg-3 (32) was cloned from the total RNA of HEK 293 cells as described previously (21). The 5Ј primers contain a restriction enzyme site (HindIII or EcoRI) and a Kozak sequence (CCACC) before the N-terminal region of receptor proteins. The 3Ј primers contained a restriction enzyme site (XbaI) and a stop codon in addition to the C-terminal region of the receptor proteins. The amplified fragments were digested with the restriction enzymes as described above and put in the pBluescript II plasmids (Stratagene), and the DNA sequence was checked.
To construct the S1P receptor expression plasmid, the amplified fragment was inserted into the EcoRI/XbaI site of the pEFneo expression plasmid for Edg-3 and into the HindIII/XbaI site of pEFneo expression plasmid for Edg-1 and AGR16. Of the three types of the putative S1P receptors, the wild type CHO cells expressed the mRNA of AGR16 (30) at 3.1 kilobases but not Edg-3 or Edg-1 (their expected size is 2.8 kilobases for Edg-3 (32) and 3.0 kilobases for Edg-1 (31)) (21). The CHO cells were transfected with pEFneo empty vector alone or the pEFneo vector containing Edg-1, AGR16, or Edg-3, and the neomycin (G418 sulfate at 1 mg/ml)-resistant cells were selected. We prepared three batches of CHO cells that were transfected with the empty vector or the respective receptor subtype cDNA. Since there was no appreciable difference in the amount of expression of S1P receptor transcript at around 1.8 kilobases between three batches of transfected cells regardless of the receptor subtype, all the data presented in the present study were from one batch of transfected cells. As shown in Fig. 1, there was no significant difference in the number of S1P binding to the cells among these receptor cDNA-transfected cells.
Measurement of S1P Receptor Binding-This was performed by the methods slightly modified from those previously described (19). The cells were washed twice with an ice-cold Tris-buffered medium consisting of 20 mM Tris-HCl (pH 7.5), 100 mM NaCl, 15 mM NaF, and 0.4% (w/v) bovine serum albumin (fraction V). The medium was replaced with a fresh medium containing [ 3 H]S1P at 3.125 to 100 nM. The plates were kept on ice for 30 min, and the cells were then washed twice with the same ice-cold medium to remove the unbounded ligand. The cells were solubilized with a cell-solubilizing solution composed of 0.1% SDS, 0.4% NaOH, and 2% Na 2 CO 3 , and the radioactivity was counted. The specific S1P binding to its receptor was estimated by subtracting the radioactivity in the presence of 10 M unlabeled S1P.

Measurement of [ 3 H]Inositol Phosphate Production-The [ 3 H]inosi
tol-labeled cells were harvested from the 10-cm dishes with trypsin (0.05% in phosphate-buffered saline containing 0.53 mM EDTA) and washed by sedimentation (250 ϫ g x 5 min) and resuspension in the Hepes-buffered medium. The Hepes-buffered medium consisted of 20 mM Hepes (pH 7.5), 134 mM NaCl, 4.7 mM KCl, 1.2 mM KH 2 PO 4 , 1.2 mM MgSO 4 , 2 mM CaCl 2 , 2.5 mM NaHCO 3 , 5 mM glucose, and 0.1% (w/v) bovine serum albumin (fraction V). The washing procedure was repeated, and the cells were finally resuspended in the same medium. Unless otherwise specified, the cells (about 5 ϫ 10 6 cells) were preincubated for 5 min with 10 mM LiCl in polypropylene vials (20 ml) in a final volume of 2.0 ml. The test agents (ϫ10) were then added to the medium, and the cells were further incubated for the indicated time. The cell suspension (0.5 ml) in triplicate was transferred to tubes containing 1 ml of CHCl 3 /CH 3 OH/HCl (100:100:1). [ 3 H]-Labeled respective inositol phosphate including IP 1 , IP 2 , and IP 3 was separated as described previously (24). Where indicated, the results were normalized to 10 5 dpm of the total radioactivity incorporated into the cellular inositol lipids. The radioactivity of the trichloroacetic acid (5%)-insoluble fraction was measured as the total radioactivity.
Measurement of [Ca 2ϩ ] i -The cells were harvested from the dishes with trypsin as described above, and then [Ca 2ϩ ] i was measured as described previously (21).
Accumulation of cAMP in Intact Cells-The cells were washed once and preincubated for 10 min at 37°C in the Hepes-buffered medium. The cells were then incubated with test agents in the presence of 0.5 mM IBMX to estimate the stimulatory activity of adenylyl cyclase (Fig. 6) or in the presence of 0.5 mM IBMX and 10 M forskolin to estimate the inhibitory activity of adenylyl cyclase (Fig. 5). After a 10-min incubation, the reaction was terminated by adding 100 l of 1 N HCl. Cyclic AMP in the acid extracts was measured as described previously (9).
Adenylyl Cyclase Activity in Cell-free System-Crude membranes were prepared as described previously (26). Briefly, the CHO cells transfected with the respective S1P receptor were harvested from the dishes and washed once with phosphate-buffered saline. The cells were suspended in 50 mM Hepes (pH 7.4) containing 50 mM sucrose, 1 mM EGTA, and 2 g/ml aprotinin, then homogenized in a Physcotron homogenizer (NS-310E, Niti-on, Tokyo, Japan) for 30 s, and centrifuged at 500 ϫ g for 5 min. The supernatant was recentrifuged at 10,000 ϫ g for 10 min, and the resultant pellet was used as crude plasma membranes. These membrane preparations (around 15 g) were incubated 37°C for 10 min with 10 nM GTP␥S and/or 3 M S1P in the same Hepes buffer containing 0.5 mM IBMX, 0.2 mM ATP, 7 mM phosphocreatine, 30 units/ml creatine phosphokinase, 100 mM NaCl, 2.5 mM MgCl 2 , and 0.1% bovine serum albumin (Fraction V) in a final volume of 200 l. The reaction was terminated by adding 50 l of 1 N HCl and boiling for 1 min. Cyclic AMP in the acid extracts was measured as described above.
Cell Migration-The migration of cells were quantified using a blind Boyden chamber apparatus (Neuro Probe Inc., Gaithersburg, MD) as described previously (13). Briefly, the lower well of the chamber was filled with DMEM containing 0.1% bovine serum albumin (fraction V) with the indicated concentration of S1P. The wells were subsequently covered with a polyvinylpyrrolidone-free filter with 8-mm pores (Neuro Probe) that were precoated for 72 h at 4°C with 100 g/ml type I collagen (Vitrogen; Collagen Corp., Palo Alto, CA). The cells were trypsinized, washed once in DMEM containing 0.1% bovine serum albumin (fraction V), and resuspended. Cells (3 ϫ 10 4 cells in 50 l) were then loaded into the upper wells of the Boyden chamber. Migration was allowed to proceed for 4 h at 37°C under a humidified air/CO 2 (19:1) atmosphere. Cells remaining on the upper surface of the filters were mechanically removed, and then filters were fixed in methanol and stained with Diff-Quik Solution II. The number of cells that had migrated to the lower surface was determined by counting under microscopy at ϫ400 magnification.
Data Presentation-All experiments were performed in duplicate or triplicate. The results of multiple observations were presented as the mean of two separate experiments or means ϮS.E. of at least three separate experiments unless otherwise stated. Fig. 1, S1P binding activity was measured in the respective S1P receptor subtype-transfected cells. In the control (empty vector-transfected) cells, only a small S1P binding activity was detected (Fig. 1A), and there was no significant correlation between B and B/F in the Scatchard plot (Fig. 1B). Since AGR16 mRNA expression is detected in the native CHO cells (27), the small binding activity might reflect binding to the endogenous S1P receptor. It should be noted, however, that in this experiment we used intact cells to lower the nonspecific binding due to the lipophilic nature of the lipid (19). Therefore, we could not discriminate the activity of the ligand uptake into the cells from the binding to the cell-surface receptors even though the experiments were performed on ice to lower the ligand uptake. On the other hand, in the cells transfected with receptor cDNAs, the apparent specific S1P binding activity was clearly increased (Fig. 1A), and significant correlation was observed between B and B/F in the Scatchard plot regardless of the receptor subtypes ( Fig. 1, C--E). There were no appreciable differences in the maximal binding activity (4.25ϳ5.69 pmol/ mg) and K d values (13.2ϳ26.2 nM) among these receptor-transfected cells. These results suggest that a comparable number of S1P receptors with similar affinity to the ligand is expressed in these receptor subtype-transfected cells.

Expression of a Comparable Amount of S1P Receptors in CHO Cells Transfected with the Respective S1P Receptor Subtype-In
Activation of PLC-Ca 2ϩ System by Edg-3 and AGR16 -Consistent with the previous study (21), the S1P-induced accumulation of inositol phosphate was remarkable in the Edg-3-transfected cells (Figs. 2, left panels and 3D). The AGR16transfected cells (Figs. 2, left panels and 3B), but not Edg-1transfected cells (Figs. 2, left panels and 3C), also displayed a higher ability than the vector-transfected cells to produce inositol phosphate in response to S1P. The difference in the inositol phosphate response is specific to S1P; the response to UTP, a P 2 purinergic agonist, was hardly affected by the transfection of any receptor subtype (Fig. 2, right panels). PTX treatment suppressed by 50ϳ80% the S1P action (Fig. 3). Thus, for the intrinsic activity to activate PLC, Edg-3 was the highest followed by AGR16, and a significant effect was not detected by Edg-1.
The activation of PLC is usually accompanied by an increase in [Ca 2ϩ ] i. As expected from the inositol phosphate response, S1P induced a small but significant [Ca 2ϩ ] i increase in the vector-transfected cells, possibly through endogenous S1P receptors (Fig. 4A). The S1P-induced increase in [Ca 2ϩ ] i was markedly enhanced in the Edg-3-transfected cells (Fig. 4D) and moderately so in the AGR16-transfected cells (Fig. 4B). In this case, the S1P-induced [Ca 2ϩ ] i increase was slightly higher in the Edg-1-transfected cells (Fig. 4C) than in the vector-transfected cells (Fig. 4A). The change in the Ca 2ϩ response was specific to S1P, as evidenced by the observation that the lysophosphatidic acid-induced action was hardly affected by any receptor transfection (Fig. 4E). PTX treatment partially suppressed the S1P-induced actions in all cases.
Regulation of Adenylyl Cyclase by S1P Receptor Subtypes-To estimate the activity of adenylyl cyclase in intact cells, the change in cAMP content was measured in the presence of IBMX, a potent inhibitor of phosphodiesterase. In the experiments shown in Fig. 5, forskolin, an activator of adenylyl cyclase, was also supplemented in the incubation medium to evaluate the inhibitory ability of S1P against adenylyl cyclase. In the vector-transfected cells, S1P had no significant effect at less than 100 nM S1P but significantly inhibited it at more than 1 M (Fig. 5A). This inhibitory action was completely reversed by the treatment of the cells with PTX (Fig. 5A), suggesting the G i /G o -protein-mediated action. When the cells were transfected with the respective receptor subtype, the pattern of the cAMP response to S1P was changed in a manner specific to each receptor. At the concentration lower than 100 nM S1P, where the lipid exerted no detectable effect on forskolin-induced cAMP accumulation in the control vector-transfected cells, S1P slightly enhanced it in the AGR16-transfected cells (Fig. 5B), conversely inhibited it in the Edg-1-transfected cells (Fig. 5C), and exerted no apparent effect in the Edg-3-transfected cells (Fig. 5D). The inhibitory effect of S1P at lower concentration (1-100 nM) in the Edg-1-transfected cells was reversed by PTX treatment (Fig. 5C). This suggests the coupling of Edg-1 to G i /G o -proteins, resulting in the inhibition of adenylyl cyclase. At concentrations of S1P higher than 1 M, the lipid-induced inhibitory action was apparently attenuated in both cells expressing AGR16 or Edg-3 (Fig. 5, B and D). Interestingly, in AGR16-and also Edg-3-transfected cells, which were treated with PTX, the cAMP level was significantly increased by increasing the concentration of S1P (Fig. 5, B and D).
We next examined the ability of the respective receptor subtype to stimulate adenylyl cyclase. For this purpose, the experiments were done without forskolin addition (Fig. 6). Under these conditions, S1P hardly changed the cAMP level in vectortransfected cells, but its level significantly increased in response to S1P in all the cells transfected with the receptor subtype. The cAMP level increased around 2 times in the Edg-1-transfected cells (Fig. 6C) and roughly 10 times both in the AGR16-and Edg-3-transfected cells in response to 10 M S1P (Fig. 6, B and D). However, at lower concentrations of S1P of less than 100 nM, the S1P-induced cAMP accumulation was higher in AGR16-transfected cells than Edg-3-transfected cells (Fig. 6, B and D). PTX did not exert an appreciable effect on the stimulatory action of S1P on cAMP accumulation in all the cases. This striking stimulatory action on the adenylyl cyclase in the cells expressing AGR16 or Edg-3 might partly account for the disappearance of the inhibitory action of S1P (Fig. 5).
It has been reported that adenylyl cyclase activation is in some cases regulated secondarily to the change in Ca 2ϩ signaling (33). As shown in Fig. 7, the PLC inhibitor alone significantly increased the cAMP content by an unidentified mechanism, but the net S1P-induced cAMP accumulation was not appreciably affected by the enzyme inhibitor. Under these conditions, the S1P-induced [Ca 2ϩ ] i increase was inhibited by more than 80% (21). Thus, it is unlikely that the S1P-induced activation of the PLC-Ca 2ϩ system may be responsible for the lipid-induced cAMP accumulation.
To further provide more direct evidence that the S1P receptor itself is coupled to the adenylyl cyclase system, we measured the enzyme activity in a cell-free system in Fig. 8. Compared with the results in intact cells, S1P effect was small, but the lipid significantly increased adenylyl cyclase activity in the presence of GTP␥S in membranes from receptor-transfected cells; the rank order of the magnitude of the S1P-induced activity was roughly AGR16 Ͼ Edg-3 Ͼ Edg-1, which was comparable with that of the S1P-induced cAMP accumulation in intact cells. Thus, the S1P receptor subtype may couple to G-proteins, probably G s , resulting in activation of adenylyl cyclase.
Stimulation of Cell Migration by Edg-1 and Edg-3-S1P has been reported to induce the change in cell morphology and cell motility (2, 13-15, 19, 22, 34, 35). Here, we measured the activity of cell migration by a blind Boyden chamber method (Fig. 9A). For the vector-and AGR16-transfected CHO cells, S1P was ineffective in stimulating cell migration as determined by cell number across the filter of a Boyden chamber. However, S1P at 1-10 nM stimulated the migration activity of the Edg-1and Edg-3-transfected cells. At concentrations higher than 100 nM, however, S1P lost its activity. The bell shape migration pattern has also been observed for other chemoattractants (36). The treatment of the cells with PTX markedly suppressed not only the S1P-induced action but also the basal activity without S1P (Fig. 9, B and C). DISCUSSION In the present study, we compared the intrinsic activity of the putative S1P receptors, i.e. Edg-1, AGR16, and Edg-3 to induce activation of PLC-Ca 2ϩ system, inhibition of adenylyl cyclase, stimulation of adenylyl cyclase, and stimulation of cell migration activity in CHO cells that permanently express the respective S1P receptor subtype to a comparable level. Results are summarized in Table I. One of the universal actions of S1P is the activation of PLC and the subsequent increase in [Ca 2ϩ ] i (6 -12), although the involvement of PLC in the [Ca 2ϩ ] i increase has not always been demonstrated (6, 10 -12). Actually, transfection of Edg-1 (20), AGR16 (23), or Edg-3 (21) has been reported to enhance the S1P-induced activation of PLC and/or [Ca 2ϩ ] i increase. Thus, all the putative S1P receptors have the potential to activate the PLC-Ca 2ϩ system. However, the present study revealed that, in CHO cells expressing an almost equal number of S1P receptors, the intrinsic activity to activate the PLC-Ca 2ϩ system varied in a manner specific to the subtype, i.e. their order being Edg-3 Ͼ AGR16 Ͼ Edg-1. We detected a small but significant enhancement of the S1P-induced [Ca 2ϩ ] i increase in the Edg-1-expressing cells (Fig. 4) without a significant effect for the inositol phosphate response (Figs. 2 and 3). The failure of the Edg-1 effect on the inositol phosphate response might simply reflect a lower sensitivity for the detection of the PLC assay compared with the [Ca 2ϩ ] i measurement. Alternatively, it might indicate the existence of the inositol phosphate-indepen-

Stimulation of cell migration
Edg-1 ϭ Edg-3 AGR16 dent mechanism for [Ca 2ϩ ] i increase. In other cell types, however, the overexpression of Edg-1 failed to stimulate Ca 2ϩ signaling (18,22). Thus, Edg-1 has a potential activity to couple to the Ca 2ϩ signaling, but its intrinsic activity is very small compared with Edg-3 or AGR16, and hence this receptor subtype might not be important for the regulation of the Ca 2ϩ signaling in the native cells.
We have recently found that only Edg-3 mRNA expression was detected among three types of S1P receptors in undifferentiated HL-60 cells, and its expression was down-regulated in association with the attenuation of the S1P-induced Ca 2ϩ response during differentiation of the cells (37). This suggests that Edg-3 may couple to the PLC-Ca 2ϩ system in HL-60 cells.
In the native CHO cells, mRNA expression of AGR16, but neither Edg-1 or Edg-3, was detected (27). Similarly, in FRTL-5 cells, we detected only AGR16 mRNA expression (data not shown). In these cells, S1P seemed to increase [Ca 2ϩ ] i depending on PLC activation (9,21). Thus, both Edg-3 and AGR16 may be involved in the S1P-induced activation of the PLC-Ca 2ϩ system in the native cells as well. S1P-induced PLC activation and [Ca 2ϩ ] i increase have, in most cases, been reported to be attenuated by PTX treatment in native cells (6,7,10,11). Similarly, the S1P-induced actions in the present study were also partially suppressed by PTX treatment (Figs. 3 and 4). Although the complete ADP-ribosylation of G i /G o -proteins was not proven by the PTX treatment employed in the present study, the inhibitory action of S1P on the forskolin-induced cAMP accumulation was completely reversed (Fig. 5). This suggests that the function of G i /G o -proteins may be lost under these conditions. On the other hand, the stimulatory action of S1P on cAMP accumulation was unaltered by the toxin treatment (Fig. 6), excluding the possible nonspecific toxic action of PTX. Thus, Edg-3 and AGR16 may couple to both PTX-sensitive G i /G o -proteins and probably, the toxin-insensitive G q /G 11proteins, resulting in the activation of PLC and the subsequent Ca 2ϩ mobilization from the intracellular pool.
Consistent with the previous results (18,20,22), Edg-1 may couple to the inhibitory adenylyl cyclase system through PTXsensitive G i /G o -proteins (Fig. 5). On the other hand, we could not detect the inhibitory action of S1P in either the Edg-3-or the AGR16-expressing cells, in which S1P elicited a rather stimulatory action on cAMP accumulation (Fig. 6). However, the cAMP level was significantly higher in PTX-treated cells at lower concentrations of less than 100 nM S1P (Fig. 5, B and D), suggesting that the inhibitory S1P action was masked by the stimulatory S1P action in the cells not treated with PTX. In relation to this, it should be noted that in the native CHO cells, only AGR16 mRNA expression was detected among three types of the putative S1P receptor (27). Therefore, the inhibition of cAMP accumulation by higher concentrations (more than 1 M) of S1P, which was observed in the vector-transfected cells (Fig. 5), might be mediated by the endogenous AGR16. Further experiments are necessary to define conclusively the ability of Edg-3 and AGR16 to couple to the inhibitory adenylyl cyclase system.
In addition to the inhibitory action on cAMP accumulation, S1P has in some cases been reported to stimulate cAMP accumulation reflecting activation of adenylyl cyclase, especially in smooth muscle cells (13). The order of the intrinsic activity of the stimulatory action on cAMP accumulation in the present receptor subtype-transfected cell system was AGR16 Ͼ Edg-3 Ͼ Edg-1. It is unlikely that the activation of adenylyl cyclase was the secondary action of the lipid-induced activation of PLC-Ca 2ϩ system (Fig. 7). Actually, at least a part of the cAMP response may reflect the lipid receptor/G s -protein-mediated activationofadenylylcyclaseasevidencedbytheguaninenucleotidedependent activation of the enzyme by S1P in a cell-free system (Fig. 8). However, we cannot completely rule out the possibility that S1P activated the enzyme partly through the production of autocrine stimulators such as prostaglandin in intact cells.
Finally, we evaluated the ability of the respective S1P receptor subtype to migrate the cells. Transfection of Edg-1 or Edg-3 into the CHO cells stimulated cell migration in response to S1P, whereas neither the vector nor AGR16 transfection appreciably affected the migration activity. Although the signaling mechanisms involved in the cell migration have not been well characterized, disassembly and assembly of actin filament may be involved in this process. The receptor-mediated rearrangement of actin filament has been shown to involve the G 12 /G 13 family of G-proteins and the Rho family of small Gproteins (14,15,38,39). In the present study, the S1P-induced action was markedly suppressed by PTX treatment. In this case, the basal activity was also suppressed by the toxin treatment (Fig. 9), suggesting that G i /G o -proteins might be absolutely necessary for the induction of cell migration. However, this result never rules out the possible involvement of G 12 /G 13proteins in the cell migration. Thus, both G i /G o -proteins and G 12 /G 13 -proteins might cooperatively regulate the cell migration activity. In the previous study, S1P has been shown to stimulate migration of T-lymphoma (15) similar to the S1P receptor-transfected CHO cells but conversely to inhibit migration of smooth muscle cells, neutrophils, and melanoma cells (2,13). This cell type-dependent discrepancy in S1P effects on cell migration is intriguing. The difference in the expression of the S1P receptor subtype might be responsible for the opposite direction of the response to S1P. This deserves further investigation in a future study.
At the present stage of investigation, AGR16 and Edg-1 have been identified from rat and AGR16, Edg-1, and Edg-3 from human. In the present study, we used rat AGR16, rat Edg-1, and human Edg-3. Therefore, it is not completely excluded that the difference in the intrinsic activity of the receptor subtype to regulate the several signaling pathways might in part reflect the differences in the animal species. However, the amino acid sequence is very similar between rat and human; the homology is 92 and 90% overall in the coding region of Edg-1 and AGR16, respectively (16,29,30). In a future study, the intrinsic activity should be compared within the receptors from the same animal species. In this way, we could know the portion or domain within the receptor that is important for the regulation of the respective signaling pathway by transfection experiments using the chimera and mutated receptors.
Last but not least, in the present study each S1P receptor subtype was expressed in the CHO cells that were expressing many types of cell surface receptors including one of the S1Preceptor, AGR16. The interaction or synergism between more than two receptors is one mode of signaling mechanism for some extracellular ligands including ATP and adenosine (40,41). Thus, there might be the interaction between the transfected receptors and the endogenous S1P receptors, which might result in the modification of the intrinsic activity of the transfected receptor, although such a cross-talk mechanism has not yet been reported in the case of the S1P receptors. Therefore, to evaluate a more accurate intrinsic activity of the respective S1P receptor subtype for the regulation of a specific signaling pathway, similar experiments may be necessary in the cells that are not expressing endogenous S1P receptors in a future study.
In conclusion, the putative S1P receptors that have recently been identified may couple to the PLC-Ca 2ϩ system, adenylyl cyclase-cAMP system, and actin rearrangement-cell motility system through PTX-sensitive and -insensitive G-proteins in a manner selective to their subtype.