Characterization of a novel sphingosine 1-phosphate receptor, Edg-8.

Three G protein-coupled receptors (Edg-1, Edg-3, and Edg-5) for the lysolipid phosphoric acid mediator sphingosine 1-phosphate have been described by molecular cloning. Using a similar sequence that we found in the expressed sequence tag data base, we cloned and characterized of a fourth, high affinity, rat brain sphingosine 1-phosphate receptor, Edg-8. When HEK293T cells were co-transfected with Edg-8 and G protein DNAs, prepared membranes showed sphingosine 1- phosphate-dependent increases in [(35)S]guanosine 5'-(3-O-thio)triphosphate binding with an EC(50) of 90 nm. In a rat hepatoma Rh7777 cell line that exhibits modest endogenous responses to sphingosine 1-phosphate, this lipid mediator inhibited forskolin-driven rises in cAMP by greater than 90% when the cells were transfected with Edg-8 DNA (IC(50) 0.7 nm). This response is blocked fully by prior treatment of cultures with pertussis toxin, thus implicating signaling through G(i/o)alpha proteins. Furthermore, Xenopus oocytes exhibit a calcium response to sphingosine 1-phosphate after injection of Edg-8 mRNA, but only when oocytes are co-injected with chimeric G(q/i)alpha protein mRNA. Membranes from HEK293T and Rh7777 cell cultures expressing Edg-8 exhibited high affinity (K(D) = 2 nm) binding for radiolabeled sphingosine 1-phosphate. Rat Edg-8 RNA is expressed in spleen and throughout adult rat brain where in situ hybridization revealed it to be associated with white matter. Together our data demonstrate that Edg-8 is a high affinity sphingosine 1-phosphate receptor that couples to G(i/o)alpha proteins and is expressed predominantly by oligodendrocytes and/or fibrous astrocytes in the rat brain.

Sphingosine 1-phosphate (S1P) 1 is a potent, extracellular lysolipid phosphoric acid mediator that is released, for example, during platelet activation (1). S1P elicits a wide variety of responses by cells; prominent among these are cell proliferation (2)(3)(4) and anti-apoptosis (5,6) as well as a wide variety of other effects. S1P and the structurally related lysolipid mediator, lysophosphatidic acid (LPA), are recognized now to signal cells through a set of G protein coupled receptors known colloquially as the "Edg" receptors. Discovered initially as "orphan" receptors (7,8), three members of the group, Edg-1, Edg-5, and Edg-3, have been shown to be S1P receptors. For example, Edg-1 mediates S1P activation of mitogen-activated protein kinase and inhibition of adenylyl cyclase in a pertussis toxindependent manner (9,10). S1P activation of Edg-3 results calcium mobilization in a pertussis toxin-independent manner (11), while others found that this receptor coupled also, for example, to inhibition of adenylyl cyclase via G i/o ␣ protein (12). Likewise, several groups have shown that a third S1P receptor, Edg-5, couples also to G q/11 ␣ proteins (13,14). All three S1P receptors signal in Xenopus oocytes, although Edg-1 signaling was dependent on co-injection with a chimeric G q ␣/G i ␣ protein (15). These data suggest that the three known S1P receptors interact with different signal transduction pathways. Indeed recent biochemical evidence supports this notion, Edg-1 interacts with G i ␣, whereas Edg-3 and Edg-5 interact with G q ␣ and G 13 ␣ as well as G i ␣ in a ligand-dependent manner (14). Three of the remaining four Edg family proteins (Edg-2, -4, and -7), which form a cluster as judged by amino acid sequence similarity, are LPA receptors, while a fourth member (Edg-6) remains an orphan receptor. In this paper, we report the existence of another S1P receptor, Edg-8, which we found first as a deposition in the expressed sequence tag data base and later cloned from rat brain. Edg-8 binds S1P with high affinity, couples to G␣ i/o proteins, and is localized prominently to white matter in the adult rat brain.

EXPERIMENTAL PROCEDURES
Cloning of Rat Edg-8 cDNA-We found a partial rat nucleotide sequence (GenBank TM accession number AI317881) similar to Edg-1, Edg-5, and Edg-3 during a routine search of updates to the GenBank TM data base of expressed sequence tags (ESTs) using the FAST_PAN program (16). We used this DNA sequence to design oligonucleotide primers that were in turn used to amplify a fragment of rat brain cDNA. * This work was supported in part by Research Grants R01 GM52722 (to K. R. L.) and R01 DK4569 (to T. H.) from the National Institutes of Health. 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  After verifying that the partial cDNA had the same nucleotide sequence as the EST DNA, the former was used to screen a rat brain cDNA library as we have reported previously (17). Three independent cDNA clones were identified and their nucleotide sequences were determined.
The full translational open reading frame (1203 nucleotides) from one of these cDNAs was subcloned into the plasmid expression vector pcDNA3 for further analysis.
Transient Expression in HEK293T Cells-Edg-8 DNA was mixed with an equal amount of DNA encoding a mutated (C351F) rat G i2 ␣ protein as well as DNAs encoding cow ␤ 1 and ␥ 2 G proteins and used to transfect monolayers of HEK293T cells using the calcium phosphate precipitate method (18). After about 60 h, cells were harvested, membranes were prepared, aliquoted, and stored at Ϫ70°C until use.
Stable Expression in Rh7777 Cells-Rat hepatoma Rh7777 cell monolayers were transfected with Edg-8-pcDNA3 DNA using the calcium phosphate precipitate method and clonal populations expressing the neomycin phosphotransferase gene were selected by addition of geneticin (G418) to the culture medium. The Rh7777 cells were grown in monolayers at 37°C in a 5% CO 2 , 95% air atmosphere in a growth medium consisting of: 90% minimal essential medium, 10% fetal bovine serum, 2 mM glutamine, and 1 mM sodium pyruvate.
Measurement of Calcium Transients and cAMP Accumulation-Assays of calcium mobilization and adenylyl cyclase activity were performed as described previously by us (19). Briefly, intracellular calcium fluxes were measured on Rh7777 cell populations (2-4 ϫ 10 6 cells) that had been loaded with the calcium sensitive fluorophore, indo-1, in the presence of 2 mM probenecid. Responses were measured in a temperature-controlled fluorimeter (Aminco SLM 8000C, SLM Instruments, Urbana, IL). Lipids were delivered as aqueous solutions containing 0.1% (w/v) fatty acid-free BSA; this vehicle did not elicit a response. Assays of adenylyl cyclase activity were conducted on populations of 5 ϫ 10 5 cells stimulated with 1 M forskolin in the presence of the phosphodiesterase inhibitor isomethylbutylxanthine. cAMP was measured by automated radioimmunoassay (gammaflow).
Oocyte Expression-Using the T7 RNA polymerase and Edg-8 DNA as template DNA, we transcribed Edg-8 mRNA in vitro in the presence of a capping analog. This mRNA was mixed with mRNA (each 20 nl at 1.0 g/ml) encoding one of several G proteins (including chimeric G q/i ␣ protein or chimeric G q/o ␣ protein (20)) and injected these into defolliculated stage V-VI Xenopus laevis oocytes prepared as described previously (15). After 48 h, oocytes were injected with aequorin (20 nl at 1.0 mg/ml) in calcium-free water and after 12 h placed individually in a 3.0-ml vial containing 450 l of OR2 buffer (82.5 mM NaCl, 2.0 mM KCl, 1.0 mM MgCl 2 , 5.0 mM HEPES, pH 7.5) supplemented with 0.01% BSA. Light emission was detected using a luminometer (Turner Design).
Membrane Preparation for Radioligand Binding-The Edg-8 plasmid was introduced into HEK293 cells by DNA-mediated transfection using the CaCl 2 procedure from Specialty Media (Lavellette, NJ). Cells were washed three times in cold phosphate-buffered saline 48 h after transfection and then lysed in hypotonic buffer consisting of 1 mM Tris/Cl, pH 7.2, and a protease inhibitor mixture for 10 min at 4°C. Cell debris was removed by centrifugation at 1000 ϫ g for 5 min at 4°C, and the supernatant fluid was recentrifuged at 40,000 ϫ g for 30 min. The pellet was suspended at a protein concentration of approximately 2 mg/ml in 40 mM Tris/Cl, pH 7.5, 200 mM NaCl, 30 mM NaF, 4 mM deoxypyridoxine, protease inhibitor mixture and stored frozen at Ϫ70°C until use.
Binding was performed for 60 min at room temperature with gentle mixing and terminated by collecting the membranes onto GF/B filter plates with a Packard Filtermate Universal Harvester. After drying the filter plates for 30 min, 50 l of Microscint 20 was added to each well, and filter-bound radionuclide was measured on a Packard Top Count. Nonspecific binding was defined as the amount of radioactivity remaining in the presence of 100-fold excess unlabeled S1P.
RNA Analysis, Northern Blotting-RNA extraction, Northern blotting, and hybridization of radiolabeled Edg-8 DNA were as described previously by us (17).
RNA Analysis, in Situ Hybridization-Brains were removed from anesthetized adult, male Harlan Sprague-Dawley rats and frozen rapidly. Coronal sections (10 m) were post-fixed in paraformaldehyde then dehydrated through graded ethanol. Sections were stored in 95% ethanol until required. Oligonucleotides (antisense) to bases 891-937 and 1000 -1044 (see Fig. 1A) were synthesized and 3Ј-end-labeled with ␣-35 S-labeled dATP using a 30:1 molar ratio of radionuclide:oligonucleotide. Labeling was catalyzed using terminal deoxynucleotidyltransferase for 15 min at 37°C, and radiolabeled oligonucleotide was separated from unincorporated nucleotide on Sephadex TM spin columns. Sections were removed from alcohol, air-dried, and 2 ϫ 10 5 cpm of each 35 S-labeled probe in 100 l of hybridization buffer was applied to each slide. Nonspecific hybridization was defined by inclusion of a 100-fold excess of unlabeled antisense oligonucleotide. Sections were covered with parafilm coverslips and incubated overnight (16 h) at 37°C. Following hybridization, the sections were washed for 1 h at 57°C in 1 ϫ saline sodium citrate (SSC) then rinsed briefly in 0.1 ϫ SSC, dehydrated through a series of alcohols, air-dried, and exposed to x-ray film (Amersham Hyperfilm ␤max). Pseudo-color images were generated from the film using an M5 Image Analysis System (Imaging Research Inc., Ottawa, Canada).
Sources of Materials-Rh7777 (CRL 1601) cells and the EST cDNA were from the American Type Culture Collection (Manassas, VA), HEK293T cells were a gift from Dr. Judy White (University of Virginia), sphingolipids were purchased from Biomol (Plymouth Meeting, PA), [ 35 S]GTP␥S and [␥-33 P]ATP were from NEN Life Science Products, [␣-35 S]dATP was from Amersham Pharmacia Biotech, Geneticin, cell culture media, and sera were from Life Technologies, Inc., oligonucleotides were from Operon Technologies (Alameda, CA), expression plasmids were from Invitrogen (La Jolla, CA), the HPLC column was from Metachem Technologies Inc. (Torrance, CA), and other chemicals were from Sigma. Fig. 1A shows the DNA sequence and deduced amino acid sequence of rat Edg-8. This sequence is from a single cDNA, but is the same sequence that we found in three independent cDNAs and the EST cDNA in regions of overlap. Therefore the sequence we report most likely represents the coding region of the rat brain Edg-8 mRNA population. The longest cDNA that we isolated contained about 80 nucleotides upstream of the putative initiation codon, but no in-frame termination codon was found in this region. Although we could not be certain that we have isolated the full translational region from our data alone, a recent report (22) of the cloning of the same sequence as an orphan receptor (named nrg-1, see "Discussion") described a longer cDNA that does contain an in-frame termination codon at position Ϫ102. Using the BLAST (23) and FASTA (24) search algorithms, we found only the single EST (Gen-Bank TM accession number AI317881) as a record of a rat Edg-8 DNA sequence in the EST division of the GenBank TM data base. Recently, however, using FAST_PAN (16) with a G protein-coupled receptor (GPCR) query set that included three Edg family members, we found another deposition closely related to rat Edg-8. This was a human sequence in the HTGS (high through-put genome sequence) division (GenBank TM accession number AC011461) that contained an intron-less translational open reading frame encoding what is probably the human ortholog of Edg-8. The human amino acid sequence conceptualized from this deposition is 87% identical to rat Edg-8 and thus might be the human ortholog of rat Edg-8.
Given the close similarity of rat Edg-8 to known S1P receptors, we tested S1P as a potential ligand in a variety of assays. To accomplish this, we required systems wherein the ubiquitous endogenous S1P responses are small or negligible. One of these is a [ 35 S]GTP␥S binding assay using HEK293T cell membranes from cells transfected with Edg-8 DNA. Although this cell line responds to S1P 2 and expresses S1P receptors (26) 1. A, the nucleotide and conceptualized amino acid sequence of rat brain Edg-8. B, the multiple alignment was constructed using the PILEUP algorithm of the GCG program. Amino acid residues that are entirely conserved within this sequence set are starred. Our rat Edg-8 DNA sequence has been deposited with the GenBank TM (accession number AF233649).  Fig. 2 shows a dose-response curve of S1P-[ 35 S]GTP␥S binding from HEK293T membranes from cultures co-transfected with DNAs encoding rat Edg-8, rat G i2 ␣C351F protein, and cow ␤ 1 ␥ 2 G proteins. The EC 50 for S1P or dihydro-S1P in this assay was 90 nM; and while LPA and other glycerolbased lyso phospholipids at concentrations up to 10 M did not stimulate binding (not shown), sphingosylphosphorylcholine (SPC) did stimulate [ 35 S]GTP␥S binding, albeit with reduced potency and efficacy. Partial agonism by SPC was reported also with Edg-1, Edg-3, and Edg-5 receptors expressed in Xenopus oocytes (15).
To examine the signaling properties of recombinant Edg-8, we introduced this DNA into Rh7777 rat hepatoma cells by transfection and selected for Geneticin-resistant clonal populations. Rh7777 cells were used because they exhibit only modest endogenous responses to S1P (27). After transfection with Edg-8 DNA, however, S1P or dihydro-S1P treatment resulted in about 80% inhibition of forskolin-driven rises in adenylyl cyclase (Fig. 3), and the sensitivity to S1P increased about 2 log orders as compared with the endogenous response (IC 50 0.7 versus 100 nM). This adenylyl cyclase inhibition was blocked by prior treatment with pertussis toxin, suggesting the involvement of G i/o ␣ proteins. Like Edg-1, but in contrast to Edg-3, Edg-8 failed to confer on Rh7777 cells a calcium mobilization in response to S1P or dihydro-S1P (data not shown). As predicted from the results of the [ 35 S]GTP␥S binding assay with HEK293T cell membranes, SPC was distinctly less potent and less efficacious than S1P at inhibiting adenylyl cyclase activation (IC 50 120 nM). S1P, dihydro-S1P, and SPC all stimulated [ 35 S]GTP␥S binding in Rh7777 cell membranes (not shown) but, unlike HEK293T cells, co-transfection with G protein DNAs was not necessary to observe this activity. Other phospholipids, including lysophosphatidylcholine, -serine, -inositol, -ethanolamine, -glycerol, and lysophosphatidic acid did not inhibit adenylyl cyclase activation in Edg-8-expressing Rh7777 cells at concentrations up to 10 M (data not shown).
X. laevis oocytes generally do not respond to S1P. Although normally a tool for sensitive detection of calcium transients elicited by agonist occupation of G q ␣ protein-coupled receptors, the addition of various exogenous chimeric G␣ proteins permits non-G q ␣ protein-coupled receptors to trigger calcium mobilization (20). As expected, no calcium response was observed in response to S1P when the Edg-8 mRNA was expressed alone in the oocyte (not shown). However, when we injected Edg-8 and chimeric G q/i ␣ protein mRNAs into frog oocytes and after several days assayed an aequorin-mediated response to S1P, we observed robust calcium responses (EC 50 5.4 nM). The doseresponse curve generated with this system is presented as Fig.  4. In this assay, we found also that chimeric G q/o ␣ protein was functional, while the G q/s ␣ protein was not (data not shown). Furthermore, sphingosine, C 2 -ceramide, ceramide 1-phos- phate, and sphingomyelin were unable to stimulate Edg-8/G q/ i␣-expressing oocytes (at concentrations up to 1 M), while SPC was a low potency partial agonist (data not shown). In this assay, we found that a chimeric G q/o ␣ protein was functional also (data not shown). As with our other expression systems, dihydro-S1P was indistinguishable from S1P (data not shown). The requirement for co-expression of the chimeric G protein supports the conclusion that Edg-8 prefers coupling to a G protein of the G i/o ␣ type.
To measure the affinity of S1P for Edg-8, we prepared 33 P-labeled S1P and used this compound in a receptor binding assay. As shown in Fig. 5, radiolabeled S1P was displaced by S1P, dihydro-S1P, and SPC from membranes prepared from Edg-8 DNA-transfected HEK293T or Rh7777 cells. The binding constant (K D ) for S1P calculated from the displacement curve was 2 nM, and membranes from both cell types showed high affinity binding. As predicted by the aforementioned studies of receptor function (30), dihydro-S1P was equipotent to S1P, while SPC was relatively ineffective in displacing the radioligand.
We investigated the expression pattern of the rat Edg-8 gene in several tissues. Northern blot analyses showed a prominent band migrating at 2.2 kilobases in extracts of various rat tissues, including throughout the brain and spleen (Fig. 6). To explore rat brain Edg-8 gene expression at higher resolution, we performed in situ hybridization with adult rat brain sections. This analysis (see Fig. 7) revealed a striking pattern, i.e. Edg-8 mRNA was expressed predominantly within white matter tracts of the rat brain. Corpus callosum, optic nerve, olfactory tract, anterior commisure, internal and external capsules, fimbra of the hippocampus, mammillary tract, stria medullaris, and white matter of the cerebellum and brain stem all showed specific signal. Within the striatum itself white matter fascicles could be identified. DISCUSSION Sphingosine 1-phosphate and the structurally similar lysolipid phosphoric acid mediator, lysophosphatidic acid, have been long suspected to signal cells at least in part through G protein-coupled receptors. Given the plethora of responses that these mediators elicit from cells and tissues, it is not surprising that there exists multiple receptor subtypes for both lipids. In the present study we have used two mammalian cell lines and frog oocytes to demonstrate unequivocally that Edg-8 is a fourth, high affinity sphingosine 1-phosphate receptor. Edg-8 is similar to Edg-1 in that both S1P receptors are predominantly G i ␣-linked and are unable to couple to the G q ␣ pathway. Rat tissue expression data, however, indicate that Edg-8 is not expressed nearly as widely as Edg-1. Like the other S1P receptors, Edg-8 does not discriminate between S1P and its reduced FIG. 7. In situ hybridization of Edg-8 oligonucleotides to thin sections of rat brain. See "Experimental Procedures" for details. form, dihydro-S1P (sphinganine 1-phosphate). The addition of a choline head group (SPC) results in a weak, partial agonist at this and other S1P receptors, and lysolipids that are not sphingosine-based are not active in our assays at concentrations up to 10 M.
The existence of Edg-8 was not predicted from existing knowledge of S1P biology, the responses of cells and tissues to this lipid mediator could apparently be explained by the three known S1P receptors. However, there exists an interesting correspondence between S1P and LPA signaling systems in many cell types. Interestingly, the LPA receptor, Edg-2, is a G i ␣ protein-linked receptor that is expressed predominantly by oligodendrocytes in rat (28) and mouse (29) brains. Perhaps the most intriguing aspect of our study is that Edg-8 is similar to Edg-2 in that expression in the CNS of both receptors is restricted to white matter. Although the present study does not allow discrimination between Edg-8 expression in oligodendrocytes versus fibrous astrocytes (or microglia or low level expression in neurons), the pattern of Edg-8 gene expression is perhaps yet another example of a parallel between the S1P and LPA signaling systems. We found marked expression of Edg-8 only in a single peripheral tissue (spleen). We do not understand the nature of the small hydrizing species that migrates at about 800 base pairs (Fig. 6) in extracts of spleen and intestine. Perhaps this signal arises from partly degraded mRNA or transcription of an overlapping gene.
Finally, while this manuscript was in preparation, Glickman et al. (22) published a report on the same rat cDNA sequence, which they named "nrg-1." Although these authors speculated that nrg-1 encodes a S1P receptor based on sequence similarity to the known Edg receptors, their paper contained no data concerning ligand identification. The two Edg-8 amino acid sequences are nearly identical (they differ by three amino acids, positions 131, 161, and 163; where our sequence has leucine, serine, and leucine; while the nrg-1 sequence reports isoleucine, leucine, and threonine, respectively). Our cDNA was from rat brain, while the nrg-1 cDNA was from the rat pheochromocytoma cell line, PC-12; perhaps the two sequences represent different alleles. We recapitulated their finding that nrg-1/Edg-8 RNA is found throughout the adult rat brain (as judged by Northern blotting, see Fig.  6). The nrg-1 gene was mapped to rat chromosome 8 where it was found tightly linked to rat Edg-5 (also known as H218). This linkage might be preserved in the human genome as both the Edg-8 and Edg-5 translational open reading frames are found in bacterial artificial chromosomes from human chromosome 19.