Role of sphingosine-1-phosphate phosphatase 1 in epidermal growth factor-induced chemotaxis.

Sphingosine-1-phosphate (S1P) is the ligand for a family of specific G protein-coupled receptors that regulate a wide variety of cellular functions, including cytoskeletal rearrangements and cell motility. Because of the pivotal role of S1P, its levels are low and tightly regulated in a spatial-temporal manner through its synthesis catalyzed by sphingosine kinases and degradation by an S1P lyase and specific S1P phosphatases (SPP). Surprisingly, down-regulation of SPP-1 enhanced migration toward epidermal growth factor (EGF); conversely, overexpression of SPP-1, which is localized in the endoplasmic reticulum, attenuated migration toward EGF. To determine whether the inhibitory effect on EGF-induced migration was because of decreased S1P or increased ceramide as a consequence of acylation of increased sphingosine by ceramide synthase, we used fumonisin B1, a specific inhibitor of ceramide synthase. Although fumonisin B1 blocked ceramide production and increased sphingosine, it did not reverse the negative effect of SPP-1 expression on EGF- or S1P-induced chemotaxis. EGF activated the epidermal growth factor receptor to the same extent in SPP-1-expressing cells, yet ERK1/2 activation was impaired. In agreement, PD98059, an inhibitor of the ERK-activating enzyme MEK, decreased EGF-stimulated migration. We next examined the possibility that intracellularly generated S1P might be involved in activating a G protein-coupled S1P receptor important for EGF-directed migration. Treatment with pertussis toxin to inactivate Galpha(i) suppressed EGF-induced migration. Moreover, expression of regulator of G protein signaling 3, which inhibits S1P receptor signaling and completely prevented ERK1/2 activation mediated by S1P receptors, not only reduced migration toward S1P but also markedly reduced migration toward EGF. Collectively, these results suggest that metabolism of S1P by SPP-1 is important for EGF-directed cell migration.

The bioactive sphingolipid, sphingosine-1-phosphate (S1P), 1 is a ligand for a family of five specific G protein-coupled recep-tors (S1P 1 , S1P 2 , S1P 3 , S1P 4 , and S1P 5 ) (1) that are coupled to various G proteins. These receptors regulate diverse cellular processes and have been implicated in cytoskeletal rearrangement and regulation of cell movement (2,3). The founding member of this receptor family, S1P 1 , couples to a G i pathway (4), and its gene disruption in mice demonstrated an essential function for vascular maturation (5). More recently, several studies have established that S1P 1 is also essential for lymphocyte recirculation and that it regulates egress from peripheral lymphoid organs (6,7). In addition to its extracellular functions, S1P acts intracellularly to regulate cell survival (8), yet its direct targets have not yet been elucidated.
As with other potent lipid mediators, levels of S1P are low and tightly regulated in a spatial-temporal manner. Sphingosine kinase (SphK) catalyzes the synthesis of S1P (2), which can be irreversibly degraded by S1P lyase (9 -11) and reversibly converted back to sphingosine by S1P phosphatases (SPPs) (12)(13)(14)(15). Recently, two isoforms, designated SPP-1 (12,13,16) and SPP-2 (15), have been identified. Both belong to the family of type 2 lipid phosphate phosphohydrolases (LPPs) (17,18) which are magnesium-independent, membrane-associated, and N-ethylmaleimide-insensitive. Although specific biological roles for the LPPs have not yet been established, the Drosophila gene wunen, which encodes a homologue of LPP1 and LPP2, negatively regulates primordial germ cell migration (19); LPP3 mediates gastrulation and axis formation, probably by influencing the canonical Wnt signaling pathway (20). Except for the conserved residues within three domains present in the active sites of LPPs (17), the two S1P phosphatases have little overall homology to other known LPPs and, in contrast to the broad specificity of the other LPPs (18), are specific sphingoid base phosphate phosphatases.
Recently, it has been shown that SPP-1 and SPP-2 are localized to the endoplasmic reticulum (ER) where they degrade S1P to terminate its actions (13)(14)(15). These findings have far reaching implications, because the ER also contains the enzymes of ceramide biosynthesis, and suggest an important role for SPP in the regulation of de novo ceramide biosynthesis. Indeed, we have recently found that SPP-1 functions in an unprecedented manner to up-regulate biosynthesis of long chain C16-ceramide (14), which has been implicated in mitochondrial-dependent apoptosis (21).
Epidermal growth factor (EGF) plays an important role in proliferation and migration of diverse cell types and cancers. These processes are mediated through binding to the EGF receptor (EGFR), a transmembrane protein with tyrosine kinase activity, and the consequent regulation of various downstream targets (22,23). Of particular relevance to this study, many signaling events of G protein-coupled receptors (GPCRs) in diverse cell types transduced by potent lipids, such as LPA and S1P, are dependent on the function of the EGFR (24,25). EGFR transactivation is important for LPA-induced head and neck cancer cell proliferation and motility (26). Because a recent study suggested that S1P also activates ERK1/2 by transactivating EGFR (27), we examined the role of SPP-1 in EGFinduced cell locomotion. We found that overexpression of SPP-1 abolished the chemotactic effect of EGF, and, conversely, knockdown of SPP-1 enhanced migratory responses to EGF. Our results suggest that intracellular metabolism of S1P may play an important role in EGF-directed chemotaxis.

MATERIALS AND METHODS
Materials-[␥-32 P]ATP (3000 Ci/mmol) was purchased from Amersham Biosciences). S1P and fumonisin B1 (FB1) were from Biomol Research Laboratory (Plymouth Meeting, PA). Sphingosine was from Avanti Polar Lipids (Alabaster, PA). Serum and media were from Biofluids (Rockville, MD). G418 and EGF were obtained from Invitrogen. Collagen type I was purchased from BD Biosciences (Bedford, MA). Polycarbonate filters were from Neuro Probe (Gaithersburg, MD). PD98059 was purchased from Calbiochem (La Jolla, CA). Other chemicals were from Sigma.
To measure cellular S1P phosphohydrolase activity, HEK 293 cells (5 ϫ 10 5 ) were permeabilized with 50 M digitonin and incubated with [ 32 P]S1P or [ 32 P]phosphatidic acid (500,000 cpm, 1 M). Loss of membrane integrity was determined by the inability of cells to exclude trypan blue (Ͼ95% when permeabilized with digitonin compared with Ͻ5% in untreated cells). Lipids were extracted from 1 ml of medium with 2.7 ml of methanol/CHCl 3 /HCl (100/200/2, v/v). 1.2 ml of 2 M KCl and 1.2 ml of CHCl 3 were then added for phase separation. The organic layer was dried under nitrogen and resuspended in CHCl 3 /methanol (95/5, v/v). Lipids were separated by TLC and the remaining [ 32 P]S1P quantified as described above.
Chemotaxis Assay-Chemotaxis was measured in a modified Boyden chamber using polycarbonate filters (25 ϫ 80 mm, 12-M pore size) coated with collagen type I (50 g/ml in 5% acetic acid) as described previously (28). Briefly, chemoattractants were added to the lower chamber, and cells (5 ϫ 10 4 /well) were added to the upper chamber. After the indicated times, non-migratory cells on the upper membrane surface were mechanically removed, and the cells that traversed and spread on the lower surface of the filter were fixed and stained with Diff-Quik (VWR, Buffalo Grove, IL). Migratory cells were counted using a microscope with a ϫ10 objective. Each data point is the average number of cells in three random fields and is the mean Ϯ S.D. of three individual wells.

SPP-1 Regulates
Chemotaxis toward EGF-Similar to its effect on chemotaxis of many other cell types (29), low concentrations of S1P enhanced chemotaxis of HEK 293 cells (Fig. 1A). EGF also induced a dose-dependent increase in migration of HEK 293 cells, with an optimum effect at ϳ100 ng/ml (Fig. 1A). As expected, the effect of EGF was completely abolished by AG1478, a specific inhibitor of the intrinsic EGF receptor tyrosine kinase (Fig. 1B). Interestingly, pertussis toxin, which ADP ribosylates and inactivates G␣ i , decreased EGF-induced chemotaxis by almost 50% (Fig. 1B). Consistent with other studies (30 -32), migration toward S1P, but not fibronectin, was also inhibited by pertussis toxin (data not shown). These results suggest that EGF-directed migration of HEK 293 cells may involve, at least in part, activation of a G i -coupled receptor.
Recently, we reported that SPP-1 plays an important role in regulation of intracellular levels of S1P (14). Moreover, we found that although sphingosine kinase 1-generated S1P stimulates survival independently of S1PRs (8), it can act in an autocrine manner to stimulate the S1P 1 cell surface receptor and consequently regulate cytoskeleton rearrangement and focal adhesions (8,33,34) and chemotaxis (35). This "inside-out" signaling by S1P, which can regulate pathways downstream of S1P 1 receptors important for cell locomotion, prompted us to investigate the potential function of SPP-1 in chemotaxis. Surprisingly, we found that overexpression of SPP-1 drastically reduced migration of HEK 293 cells not only toward S1P but also toward EGF and serum ( Fig. 2A). In sharp contrast, IGF-1-stimulated migration was unaltered (Fig. 2B), suggesting that overexpression of SPP-1 does not disrupt all essential mechanisms of cellular locomotion.
To better understand the physiological functions of SPP-1, we also examined the effect of down-regulating its expression with siRNA on migration toward EGF. In agreement with a recent report (13), transfection with siRNA targeted to SPP-1, but not control siRNA, specifically reduced expression of SPP-1 as determined by semiquantitative RT-PCR (Fig. 3A). Importantly, knockdown of SPP-1 markedly enhanced EGF-directed motility but did not influence chemotaxis toward serum or fibronectin (Fig. 3B).
Role of Ceramide-Dephosphorylation of S1P to sphingosine by SPP-1 has been associated with a large increase of ceramide levels resulting mainly from acylation of increased sphingosine by ceramide synthase (14,16). Because ceramide inhibits signaling pathways downstream of the EGFR (36,37), it was of interest to determine whether the inhibitory action of SPP-1 on chemotaxis was mediated by changes in ceramide biosynthesis. To this end, we used FB1, a specific inhibitor of ceramide synthase. In agreement with our previous results (12), FB1 blocked ceramide production in SPP-1-overexpressing cells (Fig. 4A) and concomitantly increased sphingosine (Fig. 4B). It did not reverse the negative effect of SPP-1 expression on EGFand S1P-induced chemotaxis (Fig. 4, C and D), suggesting that, in fact, inhibition of chemotaxis by SPP-1 is not a result of ceramide or sphingosine accumulation but rather dephosphorylation of S1P. Moreover, treatment with FB1 did not influence migration of vector cells toward EGF yet somewhat reduced chemotaxis toward S1P (Fig. 4D). S1P Is Cleaved Intracellularly by SPP-1-Although SPP-1 is mainly localized in the ER (13,14), it is possible that similar to other LPPs (38) a small fraction could also be localized to the plasma membrane where it might act as an ecto-S1P phosphatase. To examine this possibility, we permeabilized the plasma membrane of cells expressing SPP-1 with digitonin, an approach developed by Brindley and co-workers (39) to determine ectophosphatase activity of the LPPs. If exogenous S1P were cleaved at the plasma membrane by ectophosphatase activity of SPP-1, then its rate of hydrolysis by cells overexpressing SPP-1 would be independent of permeabilization, as was shown to be the case for LPA hydrolysis by cells overexpressing LPP1 (39). In contrast to LPP1, we found that when HEK 293 cells expressing SPP-1 were incubated with [ 32 P]S1P there was no detectable hydrolysis over a 20-min period as long as cell integrity was maintained (Fig. 5A). However, when the plasma membrane was permeabilized with digitonin, S1P was rapidly dephosphorylated by SPP-1 transfectants compared with vector-transfected cells and almost completely disappeared within 20 min (Fig. 5B). In agreement with the substrate specificity of FIG. 6. SPP-1 overexpression abrogates EGF-induced ERK 1/2 activation but not EGFR tyrosine phosphorylation. HEK 293 cells stably expressing vector or SPP-1 were serum-starved for 16 h and treated without or with EGF (100 ng/ ml) for 5 min. A, cells were lysed, and equal amounts of proteins were immunoprecipitated with anti-EGFR and analyzed by Western blotting with anti-phosphotyrosine antibody (upper panel). Blots were then stripped and reprobed with anti-EGFR (bottom panel) as a loading control. B, equal amounts of total lysates were also analyzed by immunoblotting with anti-phosphotyrosine; activation of ERK (C) and p38 (D) was determined with phospho-specific anti-ERK1/2 and p38 antibodies, respectively. Blots were stripped and reprobed with ERK2 or p38 antibodies to demonstrate equal loading. Numbers indicate relative intensities determined by densitometry of scanned blots.  (14), HEK 293 cells overexpressing SPP-1 and permeabilized with digitonin failed to cleave nonsphingoid base phospholipids, such as [ 32 P]phosphatidic acid (Fig. 5A). These results indicate that SPP-1 is unable to cleave S1P at the plasma membrane and is active only when S1P is produced inside the cells or transported there.

SPP-1 Expression Does Not Abrogate EGF-induced Tyrosine Phosphorylation of EGFR yet Inhibits ERK Activation-Be-
cause SPP-1 expression potently inhibited migration toward EGF, it was possible that this was caused by either reduction in EGFR expression, interference with activation of the EGFR, and/or inhibition of its downstream signaling. However, no significant changes in EGFR expression were detected in cells transfected with SPP-1 (Fig. 6A). Moreover, EGF stimulated tyrosine phosphorylation of EGFR to the same extent in vector and SPP-1 transfectants (Fig. 6A). We noticed, however, that expression of SPP-1 decreased levels of two tyrosine-phosphorylated proteins with molecular masses of 42 and 44 kDa, likely phospho-ERK1/2.
Previous studies suggest that S1P activates the EGFR through a protein kinase C-dependent pathway that links Ras signaling to the activation of ERK1/2 in Rat-2 cells (27). Although overexpression of SPP-1 did not significantly decrease EGF-induced tyrosine phosphorylation of EGFR (Fig. 6, A and  B), it markedly reduced ERK1 activation and had a lesser effect on ERK2 phosphorylation (Fig. 6C) without affecting p38 phosphorylation (Fig. 6D).
Involvement of S1P Receptors in EGF-mediated Chemotaxis-Collectively, our data suggest that the intracellular level of S1P, regulated by SPP-1, is important for EGF-directed chemotaxis. Although it is well established that S1P acts as a ligand for S1P receptors to regulate cytoskeletal changes and migratory re-FIG. 7. Involvement of ERK1/2 in EGF-and S1P-induced chemotaxis. HEK 293 cells stably expressing vector (V) or SPP-1 were serum-starved for 16 h and treated without or with S1P (100 nM) for 10 min (A) or pretreated without or with FB1 (25 M) for 16 h and then treated with EGF (100 ng/ml) for 10 min (B). Cells were lysed and ERK activation was determined with phospho-specific anti-ERK1/2 antibodies. Blots were stripped and reprobed with ERK2 antibody (A) or anti-ERK1/2 antibodies (B) to demonstrate equal loading. Numbers indicate relative intensities determined by densitometry of scanned blots. C, naïve HEK 293 cells were treated without or with the indicated concentrations of PD98059 and then stimulated with EGF (100 ng/ml) for 10 min, and ERK1/2 activation was determined as above. D, HEK 293 cells were pretreated without or with PD98059 (5 M) and allowed to migrate for 4 h toward EGF (100 ng/ml) or S1P (10 nM). Data are means Ϯ S.D. of three individual determinations. Similar results were obtained in at least three independent experiments. *, p Ͻ0.05 by t test.
sponses (reviewed in Ref. 29), a few studies suggest that the inhibitory effect of S1P on chemotactic motility is likely mediated through intracellular actions rather than through cell surface receptors (42,43). However, the observation that PTX reduced chemotaxis toward EGF (Fig. 1B) strongly argued against this possibility. To further separate responses dependent on intracellular S1P from those that are receptor-mediated, we specifically blocked signaling of the heterotrimeric G proteins that S1PRs couple to and signal through. Regulators of G protein signaling, such as RGS3, function as GTPase-activating proteins for G␣ i and G␣ q subunits of heterotrimeric G proteins, resulting in their inactivation (44). It was previously shown that transfection of HEK 293 cells with RGS3 completely prevented signaling mediated by activation of S1PRs with exogenous S1P (45). In agreement, we found that S1P-induced ERK1/2 activation was blocked in HEK 293 cells overexpressing RGS3 (Fig. 8A). RGS3 also reduced ERK activation induced by EGF. Importantly, expression of RGS3 not only reduced migration toward S1P, it also markedly reduced migration toward EGF (Fig. 8B). DISCUSSION Our results suggest that SPP-1-regulated intracellular levels of S1P, acting through G protein signaling, plays an important role in EGF-directed cell motility. Several lines of evidence support this notion. First, overexpression of SPP-1, which decreases intracellular levels of S1P, suppressed motility toward EGF. Conversely, down-regulation of SPP-1 enhanced chemotaxis toward S1P and EGF. Third, PTX treatment, which inhibits G i signaling, markedly reduced motility toward EGF, although the EGFR is a tyrosine kinase and does not utilize G proteins for signaling. Finally, blocking signaling of all the heterotrimeric G proteins that S1PRs couple to and signal through also markedly reduced migration toward EGF. This receptor communication between EGFR and S1PRs is distinct from the well established transactivation of the EGFR by other GPCRs (24,25,46).
Although SPP-1 overexpression is also associated with an increase in ceramide that is inhibited by the ceramide synthase inhibitor FB1 (14), FB1 did not attenuate the inhibitory effect of SPP-1 on EGF-induced chemotaxis. These results suggest that SPP-1 acts mainly by reducing S1P accumulation in response to EGF, rather than by formation of sphingosine or by increasing its conversion into ceramide or further ceramide metabolites. SPP-1 can therefore modulate inside-out signaling mediated by S1P through S1PRs important for EGF-induced chemotaxis. In agreement, Obeid and co-workers (13) have recently shown that not only were intracellular levels of S1P increased by siRNA knockdown of endogenous SPP-1 in HEK 293 cells, it also induced an increase in secretion of S1P into the extracellular medium.
How ER-resident SPP-1 is able to reduce S1P levels at the plasma membrane and inhibit S1PR activation is an intriguing conundrum. It has been suggested that close contacts exist between the ER and plasma membrane (47). If so, it is possible that SPP-1 might dephosphorylate S1P at specific sites of contact between these two organelles. Alternatively, similar to calnexin and calreticulin, other ER-resident proteins that have been observed at the cell surface (48), it is also possible that SPP-1 can function at the plasma membrane as an ectophosphatase similar to other LPPs (39). There are several other possibilities that can explain how overexpression of SPP-1 inhibits exogenous S1P-induced chemotaxis and ERK activation. First, it is possible that overexpressed SPP-1 inhibits S1Pinduced chemotaxis not by dephosphorylating S1P at the cell surface but by decreasing intracellular S1P formed in response to activation of sphingosine kinase by S1P itself. Meyer zu Heringdorf et al. (49) showed that exogenous S1P induced calcium mobilization in HEK 293 cells through sphingosine kinase activation and that this was PTX-sensitive, indicating that intracellularly generated S1P was acting in an autocrine manner to activate cell surface G i -coupled S1P receptors. Secondly, it is also possible that there is a small fraction of the total SPP-1 that is expressed on the plasma membrane that is capable of reducing local concentrations of S1P near receptors, yet assays are not sensitive enough to detect such small changes. Consistent with this notion, we have shown previously that SPP-1-transfected HEK cells have a 3-fold increase in plasma membrane S1P phosphohydrolase activity compared with 5-fold in the ER and that the total activity is about 20% of that of the ER (14). In agreement, most of the increase in ceramide after S1P treatment in these cells occurred in the internal membrane fraction, although there was also a small but significant increase of ceramide in the plasma membrane (14).
How is it possible that S1P produced in an intracellular compartment such as the ER can play a role in EGF signaling? Although it was originally considered that activated FIG. 8. Overexpression of RGS3 in HEK 293 cells blocks migration toward S1P and EGF. HEK 293 cells stably expressing vector or RGS3 were serum-starved for 16 h. A, cells were treated without or with S1P (100 nM) or EGF (100 ng/ml) for 5 min and lysed. ERK activation was determined with phospho-specific anti-ERK1/2 antibodies. Blots were stripped and reprobed with ERK2 antibody to demonstrate equal loading. B, duplicate cultures were allowed to migrate toward S1P (10 nM) or EGF (100 ng/ml) for 4 h, and chemotaxis was measured. Data are means Ϯ S.D. of three individual determinations. Similar results were obtained in at least three independent experiments. *, p Ͻ0.05 by t test.
EGFR is rapidly internalized in early endosomes as a mechanism of receptor inactivation, accumulating evidence indicates that the EGFR remains active within early endosomes (50). Moreover, endosomal EGFR signaling was sufficient to activate the major signaling pathways, including Ras, ERK, and Akt, but not PLC␥, leading to cell proliferation and survival (51). A recent functional imaging study demonstrated that internalized EGFR travels to the ER, where the resident protein tyrosine phosphatase-1␤ catalyzes its dephosphorylation, leading to inactivation of EGF signaling prior to its degradation in lysosomes (52). ER-resident SPP-1 might act in a similar manner as an off switch to terminate EGF signaling by degrading S1P. Spatial and temporal partitioning of S1P formation and degradation within cells could regulate the duration of signals and provide an added layer of control to on-off signaling for receptors that utilize intracellularly generated S1P as a signaling molecule.
In many cell types, activation of ERK1/2 is required for motility induced by EGF (26,41) and S1P (53). Although overexpression of SPP-1 did not affect EGFR activation, it reduced EGF-induced ERK1/2 phosphorylation, suggesting that S1P production may be required for EGFR downstream signals. In agreement, we found that overexpression of RGS3, which inhibits signaling through all of the S1P receptors expressed in HEK 293 cells (45), reduced not only ERK1/2 phosphorylation induced by EGF but also its chemotactic effects. In this regard, EGF-activated ERK1/2 in ovarian theca cells was PTX-sensitive (54). We also found that EGF-induced cell motility is inhibited by pretreatment with PTX, suggesting that intracellular S1P, whose level is regulated by SPP-1, can activate a heterotrimeric G␣ i protein-coupled S1PR, likely by an autocrine pathway. This is consistent with the observation that decreased expression of SPP-1 increases secretion of S1P (13). Although S1P is secreted by several types of cells in response to various agonists (55), little is known about the mechanisms of its release. The ATP-binding cassette transporter ABCB1 (previously called MDR-1 and P-glycoprotein), which catalyzes movement of platelet-activating factor in addition to amphiphilic drugs from the cytosol to the extracellular environment (56), has recently been implicated in transport of S1P out of cells (57). Transport of S1P out of cells is an important issue to resolve because it impinges on S1P actions at the cell surface and possibly inside cells.