Sphingosine 1-phosphate and platelet-derived growth factor (PDGF) act via PDGF beta receptor-sphingosine 1-phosphate receptor complexes in airway smooth muscle cells.

Platelet-derived growth factor (PDGF) and sphingosine 1-phosphate (S1P) act via PDGF beta receptor-S1P(1) receptor complexes in airway smooth muscle cells to promote mitogenic signaling. Several lines of evidence support this conclusion. First, both receptors were co-immunoprecipitated from cell lysates with specific anti-S1P(1) antibodies, indicating that they form a complex. Second, treatment of airway smooth muscle cells with PDGF stimulated the phosphorylation of p42/p44 MAPK, and this phosphorylated p42/p44 MAPK associates with the PDGF beta receptor-S1P(1) receptor complex. Third, treatment of cells with antisense S1P(1) receptor plasmid construct reduced the PDGF- and S1P-dependent activation of p42/p44 MAPK. Fourth, S1P and/or PDGF induced the formation of endocytic vesicles containing both PDGF beta receptors and S1P(1) receptors, which was required for activation of the p42/p44 MAPK pathway. PDGF does not induce the release of S1P, suggesting the absence of a sequential mechanism. However, sphingosine kinase 1 is constitutively exported from cells and supports activation of p42/p44 MAPK by exogenous sphingosine. Thus, the presentation of sphingosine from other cell types and its conversion to S1P by the kinase exported from airway smooth muscle cells might enable S1P to act with PDGF on the PDGF beta receptor-S1P(1) receptor complex to induce biological responses in vivo. These data provide further evidence for a novel mechanism for G-protein-coupled receptor and receptor tyrosine kinase signal integration that is distinct from the transactivation of receptor tyrosine kinases by G-protein-coupled receptor agonists and/or sequential release and action of S1P in response to PDGF.

The S1P 1 receptor is implicated in regulating smooth muscle cell migration, proliferation, and vascular maturation. Insight into the function of the S1P 1 receptor was obtained by studies showing that disruption of the s1p 1 gene by homologous recombination in mice results in extensive intra-embryonic hemorrhaging and intrauterine (16). This is caused by incomplete vascular maturation due to the failure of mural cells, vascular smooth muscle cells, and pericytes to migrate to arteries and capillaries and to reinforce them properly. Interestingly, the disruption of the PDGF␤ receptor gene in mice results in a similar phenotype (17,18). Hobson et al. (16) showed that PDGF-stimulated cell motility is S1P 1 receptor-dependent in human embryonic kidney 293 cells (HEK 293 cells) and mouse embryonic fibroblasts. Therefore, a sequential model has been proposed in which PDGF stimulates S1P synthesis/release, which in turn binds to and activates the S1P 1 receptor to induce activation of Rac, PKB, and cell motility (16). In contrast, we have reported that the PDGF␤ receptor forms a complex with the S1P 1 receptor in HEK 293 cells transfected with recombinant forms of these receptors (19,20). We showed that the association of these recombinant receptor types enables more efficient tyrosine phosphorylation of G i ␣ by the PDGF␤ receptor kinase, possibly via proximity-induced effects. We showed that the tyrosine phosphorylation of G i ␣ was required to stimulate the p42/p44 MAPK pathway by PDGF or S1P, consistent with an integrative signal model (19). Freedman et al. (21) recently confirmed the involvement of G i in PDGF␤ receptor signaling. We also reported that ␤-arrestin I/GRK2 associate with the PDGF␤ receptor-S1P 1 receptor complex in HEK 293 cells (19) and that this might represent an important step regulating p42/p44 MAPK signaling. Freedman et al. (21) recently confirmed the association of GRK2 with the PDGF␤ receptor.
The question as to which model, sequential versus integrative signaling, is the major mechanism of action of PDGF and S1P is very important and still remains to be fully investigated.
In this article, we have shown that native functionally active PDGF receptor-S1P 1 receptor complexes are expressed in airway smooth muscle cells. Moreover, PDGF stimulated the phosphorylation of p42/p44 MAPK, and this phosphorylated p42/p44 MAPK associates with the PDGF␤ receptor-S1P 1 receptor complex. These findings show that the integrative signaling model is a major mechanism of action of PDGF and S1P in cells. Consistent with an integrative model, we have also established that these receptors are co-internalized together as a functional signaling unit to regulate the p42/p44 MAPK pathway.
Cell Culture-The preparation of primary cultures of guinea pig airway smooth muscle (ASM) cells has been described previously (23). Their identity was confirmed to be smooth muscle by the presence of ␣-actin using smooth muscle-specific mouse anti-␣-actin monoclonal antibodies. Cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) fetal calf serum and 10% (v/v) horse serum. Cells were routinely used at passage 3-4. Cells were placed in Dulbecco's modified Eagle's medium supplemented with 0.1% (v/v) fetal calf serum and 0.1% (v/v) horse serum for 24 h before experimentation. HEK 293 cells were maintained in culture as previously reported by us (19).
Transfection-ASM cells were transiently transfected with antisense/sense S1P 1 receptor, sense PDGF␤ receptor, FLAG-tagged hSK1 WT , or FLAG-tagged hSK1 G82D plasmid constructs as required. Cells at 90% confluence were placed in minimum Eagle's medium containing 2% fetal calf serum and transfected with 2 g of plasmid construct after complex formation with LipofectAMINE TM 2000 according to the manufacturer's instructions. The cDNA-containing media was removed after incubation for 24 h at 37°C, and the cells were incubated for a further 24 h in serum-free medium before agonist additions.
p42/p44 MAPK, PKB, and p38 MAPK Assays-The phosphorylated forms of p42/p44 MAPK, PKB, and p38 MAPK were detected by Western blotting cell lysates with the respective anti-phospho-specific antibodies. Anti-p42 MAPK, PKB, and p38 MAPK antibodies were also used for Western blotting to establish equal loading of protein in each sample.
Blotting-Immunoblotting was performed as described by us previ-ously (19). Immunoreactive proteins were visualized using enhanced chemiluminescence detection. S1P Measurements-Cells grown to confluency on multiple 6-well plates (1.2 ϫ 10 6 cells/well) were serum-starved overnight and equilibrated in 1 ml of Dulbecco's modified Eagle's medium/fatty acid-free bovine serum albumin (2 mg/ml) at 37°C for at least 1 h before stimulation. [ 3 H]Sphingosine was resuspended with phosphate-buffered saline containing fatty acid-free bovine serum albumin (2 mg/ml) by sonication for 10 min. The reactions were started by the addition of [ 3 H]sphingosine (222,000 dpm/well, final concentration 5ϳ10 nM) and various agonists for the periods of time indicated.
Stimulation of cells was stopped by removal of the medium and the addition of 0.5 ml of ice-cold methanol. Cells were then scraped into glass vials, and an equal volume of chloroform was added. Samples were vortexed vigorously and maintained at room temperature for at least 20 min to allow extraction of cellular lipids. The samples were dried using a SpeedVac centrifuge. Samples were then re-dissolved in 80 l of CHCl 3 :CH 3 OH (19:1, v/v) and resolved by TLC using silica gel G60 LK6D TLC plates (Whatman) and solvent system CHCl 3 :CH 3 OH: CH 3 COOH:H 2 O (25:10:1:2, v/v). A [ 32 P]S1P standard was resolved in parallel. Sample profiles were obtained by excising 1-cm fractions of the silica and quantification of the radioactivity by liquid scintillation counting.
Immunofluorecence-Cells were grown on 12-mm glass coverslips to 60 -90% confluence and transfected as described. After treatment with S1P and/or PDGF for the times indicated, cells were fixed in 3.7% formaldehyde in PBS for 10 min then permeabilized in 0.1% Triton X-100 in phosphate-buffered saline for 1 min. Nonspecific binding was reduced by preincubating cells in blocking solution containing 5% fetal calf serum, 1% bovine serum albumin in phosphate-buffered saline for 1 h. Cells were incubated in primary antibodies (1:100 dilution in blocking solution) for 1-2 h and then incubated with the appropriate fluorescein isothiocyanate (rabbit)-or TRITC (mouse)-conjugated secondary antibodies (1:100) for 1 h. Cells were mounted on glass slides using Vectashield mounting medium with 4,6-diamidino-2-phenylindole and visualized using a Nikon E600 epifluorescence microscope.

RESULTS
The Existence of Functional PDGF␤ Receptor-S1P 1 Receptor Complexes in ASM Cells-We set out to investigate whether the PDGF␤ receptor forms a complex with the S1P 1 receptor in cultured ASM cells and whether this represents a functional integrative signaling unit. ASM cells were chosen for several reasons. First, these cells express high levels of S1P 1 receptor (24). Second, pretreatment of cells with pertussis toxin reduced PDGF-AB-and S1P-stimulated activation of p42/p44 MAPK by 50 and 100%, respectively (24,25). Under the conditions used in these studies, we showed that pertussis toxin completely inactivates G i ␣. Therefore incomplete blockade of PDGF receptor signaling suggests that the receptor uses both G i -dependent and -independent pathways to regulate p42/p44 MAPK. Thus, a substantial fraction of the PDGF receptors might be associated with a GPCR in these cells, since pertussis toxin acts by uncoupling GPCRs from their respective G-protein to prevent signaling. Third, S1P and PDGF-AB use a common signaling pathway involving G i -regulated Grb-2-associated phosphoinositide 3-kinase, which lies up-stream of p42/p44 MAPK activation (24,25). Fourth, a combination of submaximal S1P and PDGF-AB induced a synergistic activation of p42/p44 MAPK (24). Fifth, S1P is co-mitogenic with PDGF-AB (24). These data are consistent with the possibility that the S1P 1 receptor might be associated with and exert cross-talk regulation with the PDGF receptor. Fig. 1a shows that the native PDGF␤ receptor is indeed associated with native S1P 1 receptors in ASM cells. Thus, the PDGF␤ receptor (molecular mass 180 kDa) and S1P 1 receptor (molecular mass 45 kDa) were co-immunoprecipitated from cell lysates with anti-S1P 1 receptor antibodies. Neither receptor was immunoprecipitated when the antibodies were omitted from the procedure. Stimulation of ASM cells with PDGF-AB did not promote further formation of the PDGF␤ receptor-SIP 1 receptor complex, suggesting that the complex is preformed in these cells. Transfection of ASM cells with recombinant PDGF␤ receptor did not increase the amount of PDGF␤ receptor present in the anti-S1P 1 receptor immunoprecipitates (Fig. 1a), suggesting that the endogenous S1P 1 receptor is limiting for complex formation with the PDGF␤ receptor (Fig. 1a). S1P 1 receptor was also co-immunoprecipitated with PDGF␤ receptor using anti-PDGF␤ receptor antibodies (data not shown). The interaction of the S1P 1 receptor with the PDGF receptor exhibits specificity, since S1P 1 receptors do not associate with the EGF receptor tyrosine kinase and there is no cross-talk regulation between EGF and S1P in terms of activating p42/p44 MAPK (Ref. 19 and data not shown). EGF signaling also differs from that of PDGF in that EGF-stimulated p42/p44 MAPK activation is insensitive to blockade by pertussis toxin and, therefore, appears to be independent of G i PCRs (19,26).
Several lines of evidence support a model in which association of the PDGF receptor with the S1P 1 receptor enables integrative signaling in response to both PDGF and S1P. First, the treatment of ASM cells with PDGF-AB stimulated the phosphorylation of p42/p44 MAPK, and this phosphorylated p42/p44 MAPK associates with the PDGF␤ receptor-S1P 1 receptor complex. Thus, substantially more phosphorylated p42/ p44 MAPK was co-immunoprecipitated with the PDGF␤ receptor-S1P 1 receptor complex using anti-S1P 1 receptor antibodies from cells treated with PDGF compared with controls ( Fig. 1b). p42 MAPK was also detected in anti-S1P 1 receptor immunoprecipitates from lysates of PDGF-treated cells using anti-p42 MAPK antibodies to probe the Western blots (data not shown). Similar results were obtained when S1P was used to stimulate cells (data not shown).
In previous studies, we have shown that c-Src is activated by PDGF, and this can be blocked by pretreating ASM cells with pertussis toxin (25). Moreover, in the same study we showed that pretreating ASM cells with PP1, an inhibitor of c-Src, reduced the activation of p42/p44 MAPK in lysates of ASM cells stimulated with PDGF (25). We now show that the c-Src inhibitor PP2 ablated the PDGF-dependent phosphorylation of the p42/p44 MAPK that associates with PDGF␤ receptor-S1P 1 receptor complex (Fig. 1b). Similar results were observed in cell lysates (Fig. 1b). PP2 did not reduce the amount of the PDGF␤ receptor associated with the S1P 1 receptor in anti-S1P 1 receptor immunoprecipitates nor did it affect PDGF␤ receptor levels in cell lysates (Fig. 1b).
Second, the transfection of a S1P 1 receptor antisense plasmid construct reduced the PDGF-AB-and S1P-dependent activation of p42/p44 MAPK (% inhibition of PDGF-AB-and S1P-stimulated p42 MAPK activation (normalized for p42 MAPK levels) were: PDGF, 48.3 Ϯ 10.6%; S1P, 65 Ϯ 14%, n ϭ 3, p Ͻ 0.05 antisense-treated agonist response versus agonist response) and had no significant effect on phorbol ester (PMA)stimulated p42/p44 MAPK activation (% inhibition of PMAstimulated p42 MAPK activation (normalized for p42 MAPK levels) was 12 Ϯ 20%, n ϭ 3), which does not involve GPCRs (Fig. 2a). The lack of significant effect of the antisense construct on PMA-stimulated p42/p44 MAPK activation indicates specificity against S1P 1 receptor-mediated responses. The an-FIG. 1. The association of S1P 1 receptors with PDGF␤ receptors and phosphorylated p42/p44 MAPK. ASM cells were pretreated with the c-Src inhibitor PP2 (10 M, 10 min) before stimulation with or without PDGF-AB (10 ng/ml, 10 min) as indicated. In certain cases, cells were transfected with PDGF␤ receptor plasmid construct. a, Western blots (WB, showing the co-immunoprecipitation of PDGF␤ receptor (molecular mass 180 kDa) and S1P 1 receptor (molecular mass 45 kDa) from ASM cell lysates using anti-S1P 1 receptor antibodies (Ab). Anti-S1P 1 receptor immunoprecipitates were resolved on SDS-PAGE and Western-blotted with anti-PDGF␤ receptor and anti-S1P 1 receptor antibodies. ASM cells in which recombinant PDGF␤ receptor (PDGF␤R) were transfected were also used to show that endogenous S1P 1 receptors are limiting for the binding of PDGF␤ receptor. b, Western blots showing the PDGF-dependent phosphorylation of p42/ p44 MAPK and association with the PDGF␤ receptor-S1P 1 receptor complex. Also shown is the effect of pretreating cells with the c-Src inhibitor, PP2, on receptor-associated phosphorylated p42/p44 MAPK and PDGF␤ receptor-S1P 1 receptor complexes. Anti-S1P 1 receptor immunoprecipitates were resolved on SDS-PAGE and Western-blotted with anti-PDGF␤ receptor and anti-phosphorylated p42/p44 MAPK antibodies. This figure also shows a Western blot of the corresponding effects of PP2 on the PDGF-dependent phosphorylation of p42/p44 MAPK and PDGF␤ receptor levels in total cell lysates. These are representative results of an experiment performed three times. IP, immunoprecipitation; T/F, transfection. tisense S1P 1 receptor plasmid construct reduced S1P 1 receptor (molecular mass 45 kDa) expression (Fig. 2b), thereby also validating the specificity of the anti-S1P 1 receptor antibody for S1P 1 receptor. The antisense construct had no effect on cell viability under the conditions used (data not shown). Finally, S1P does not cause the release of PDGF, based on previous results obtained by us, showing that S1P does not stimulate PDGF receptor tyrosine phosphorylation (24).

Co-internalization of PDGF␤ Receptor-S1P 1 Receptor Complexes in Endocytic
Vesicles-One prediction of the proposed PDGF␤ receptor-S1P 1 receptor complex association model is that stimulation of cells with PDGF is expected to cause the co-internalization of both PDGF␤ and S1P 1 receptors into the same endocytic vesicles. ASM cells were transfected with Myctagged S1P 1 receptor to capture PDGF␤ receptor and to allow detection of S1P 1 receptors with anti-Myc tag antibodies. Un- der these conditions, stimulation of cells with PDGF promoted the formation of endocytic vesicles containing both PDGF␤ receptor and Myc-tagged S1P 1 receptor (yellow vesicles, Fig.  3A, panels 6a and 6b). Inhibitors of clathrin-mediated endocytosis (e.g. concanavalin A and monodansylcadervine) blocked the internalization of the PDGF␤ receptor-S1P 1 receptor complex in response to PDGF (data not shown). We excluded the possibility that PDGF elicits the release of S1P (see later), which might have acted on S1P 1 receptors, causing their recruitment to the same endocytic vesicles as those containing PDGF␤ receptors. The treatment of cells with S1P also induced co-internalization of S1P 1 and PDGF␤ receptors, although the sensitivity of detection was considerably reduced (data not shown). This is explained by the fact that overexpression of recombinant S1P 1 receptors probably results in the expression of a significant amount of non-associated S1P 1 receptor, which might effectively buffer the PDGF␤ receptor-S1P 1 receptor complex pool from S1P.
We have previously shown that recombinant forms of the PDGF␤ receptor and S1P 1 receptor form functional associated signaling complexes in HEK 293 cells (19). Because the levels of the PDGF␤ receptor-S1P 1 receptor complex are very high in these cells due to overexpression of the two proteins, we predicted that this system might show increased sensitivity in terms of co-internalization of both receptors in response to S1P. Thus, overexpression of both receptors to increase their association might surmount the buffering effect of non-associated S1P 1 receptors on S1P observed in ASM cells. Consistent with this, we found that the stimulation of transfected HEK 293 cells with S1P induced an increase in the number of endocytic vesicles containing both recombinant PDGF␤ receptors and Myc-tagged S1P 1 receptors. PDGF also induced co-internalization of both receptors to the same endocytic vesicles (Fig. 3B,  yellow vesicles, panels h and l).
PDGF also promoted the formation of endocytic vesicles containing PDGF␤ receptors that were not associated with Myctagged S1P 1 receptors (green vesicles, Fig. 3B, panel l). These endocytic vesicles might contain recombinant PDGF␤ receptor that is either free or associated with endogenous S1P 1 receptors, for which transcript is expressed in HEK 293 cells (data not shown).
A combination of S1P and PDGF produced an increase in the number of endocytic vesicles containing both recombinant PDGF␤ receptors and Myc-tagged S1P 1 receptors compared with that induced with each agent alone (Fig. 3B, yellow vesicles, panel p). This is entirely consistent with a model in which combined stimulation of cells with PDGF and S1P will increase the number of PDGF␤ receptor-S1P 1 receptor complexes occupied with either or both ligands, leading to an increased number of ligand-bound PDGF␤ receptor-S1P 1 receptor complexes being internalized within common endocytic vesicles. These findings represent an entirely novel mechanism for co-internalization of receptor tyrosine kinase/GPCRs. PDGF-stimulated S1P Formation and p42/p44 MAPK Activation-The existence of a PDGF␤ receptor-S1P 1 receptor complex in ASM cells and the functional evidence provided here support a model in which S1P and PDGF use an integrative signaling mechanism in these cells. Nevertheless, it was necessary to formally evaluate whether part of the response to PDGF is mediated via the release of S1P, which might act on S1P 1 receptors in a sequential manner. We therefore compared the effect of (i) different PDGF isoforms and (ii) sphingosine kinase inhibitors and recombinant wild type and a dominant negative mutant hSK1 on PDGF-stimulated S1P formation, p42/p44 MAPK, PKB, and p38 MAPK activation.
First, we found that there was a correlation between the ability of PDGF-BB to stimulate S1P formation and to activate p42/p44 MAPK or to promote [ 3 H]thymidine incorporation in ASM cells. PDGF-BB (10 ng/ml) caused a rapid and transient increase in [ 3 H]S1P production in ASM cells (300% increase in S1P above basal at peak 3 min of cell stimulation). However, neither PDGF-AA nor PDGF-AB stimulated S1P formation (Fig. 4a) 10.52 Ϯ 0.6, n ϭ 3-4, p Ͻ 0.05 for PDGF-BB versus PDGF-AB. Therefore, the rank order activation of p42/p44 MAPK (at 10 ng/ml PDGF) was PDGF-BB Ͼ PDGF-AB Ͼ PDGF-AA (Fig. 5a, left bottom panel). This was correlated with PDGF␣/␤ receptor auto-tyrosine phosphorylation (Fig. 5a, left mid-panel) and PDGF-stimulated [ 3 H]thymidine incorporation (Fig. 5a, right panel). Fig. 4a shows the dose response of S1P production (peak at 3 min of cell stimulation) induced by PDGF-BB and the lack of effect of PDGF-AA and PDGF-AB. Half-maximal and maximal stimulation of [ 3 H]S1P production was observed at 1.5 Ϯ 0.6 and 6 ng/ml PDGF-BB, respectively (Fig. 4a). We therefore further investigated whether the increased efficacy of PDGF-BB on p42/p44 MAPK activation could be accounted for by potential release of S1P by PDGF-BB.
Sphingosine Kinase Release from Airway Smooth Muscle Cells-Despite the lack of evidence to support a model in which PDGF promotes the release of S1P from ASM cells, hSK1 is exported from these cells. Indeed, Ancellin et al. (27) also show that SK is exported from human umbilical vein endothelial cells and HEK 293 cells and may play a role in producing S1P at the cell surface. Moreover, these authors showed that extracellular-formed S1P induced a profound angiogenic effect via S1P receptors. We have therefore evaluated whether FLAG-tagged hSK1 WT is exported from ASM cells. Fig. 6a shows that FLAG-tagged hSK1 WT was exported into the medium. The transfection of FLAG-tagged hSK1 WT into ASM cells did not compromise cell integrity, as assessed by trypan blue exclusion (data not shown), and did not cause leakage of other proteins of similar molecular mass to hSK1, such as p42/p44 MAPK (Fig. 6a).
The exported SK is capable of converting sphingosine to S1P. This was supported by data showing that a low concentration of exogenous sphingosine stimulated p42/p44 MAPK activation in ASM cells (Fig. 6b). The activation of p42/p44 MAPK by sphingosine was significantly increased in ASM cells transfected with S1P 1 receptor (data not shown). These findings suggest that export of endogenous SK may convert sphingosine to S1P, which can subsequently stimulate p42/p44 MAPK. Indeed, the sphingosine-dependent activation of p42/p44 MAPK was abrogated by overexpression of FLAG-tagged hSK1 G82D (Fig. 6b). However, a role for exported SK in mediating part of the action of PDGF in ASM cells was excluded on the basis that PDGF does not promote export of hSK1 (data not shown) and SK inhibitors/hSK1 G82D did not abrogate PDGF receptor signaling (Fig. 5). DISCUSSION We have presented evidence that directly supports a model for signal integration by PDGF␤ receptor-S1P 1 receptor complexes in ASM cells. We have also demonstrated that PDGF does not promote the release of S1P to stimulate p42/p44 MAPK, PKB, or p38 MAPK in a sequential manner in these cells. The sequential model may be operative in other cell types where PKB is very sensitive to low levels of S1P release. hSK1 WT is constitutively exported from ASM cells and is not regulated by PDGF. Interestingly, the enzyme converts exogenous sphingosine to S1P, which in turn stimulates the p42/p44 MAPK pathway.
Several lines of evidence were obtained to support the exist- ence of a functionally active native PDGF␤ receptor-S1P 1 receptor complex that can support integrative signaling by S1P/ PDGF in ASM cells. First, the PDGF␤ receptor and S1P 1 were co-immunoprecipitated from cell lysates with specific anti-S1P 1 /PDGF␤ receptor antibodies. Second, the treatment of ASM cells with PDGF stimulated the phosphorylation of p42/ p44 MAPK, and this phosphorylated p42/p44 MAPK associates with the PDGF␤ receptor-S1P 1 receptor complex. To our knowledge, this is the first report showing that phosphorylated p42/ p44 MAPK is present in a complex with receptor in response to growth factor. Current models have suggested that after ligand binding and the initiation of signaling the receptor is subsequently disconnected from events such as association of activated Raf-MEK (Raf-MAPK kinase 1) with p42/p44 MAPK. The close association of phosphorylated p42/p44 MAPK with the PDGF␤ receptor-S1P 1 receptor complex and attendant protein components is entirely compatible with an integrative signaling response to PDGF and S1P. Third, the treatment of cells with antisense S1P 1 receptor plasmid construct (which lowered S1P 1 receptor expression) reduced the PDGF-and S1P-dependent activation of p42/p44 MAPK. These findings support the requirement for cross-talk regulation between both receptors. They also suggest that the S1P 1 receptor might be partially constitutively active and capable of releasing active G-protein subunits for tyrosine phosphorylation by the PDGF␤ receptor kinase, as observed in HEK 293 cells (19). Fourth, S1P and PDGF promoted the co-internalization of PDGF␤ receptor and S1P 1 receptor in the same endocytic vesicles. The ability of S1P to induce internalization of the PDGF␤ receptor is important as it establishes further evidence for close association between the S1P 1 receptor and the PDGF␤ receptor. This model is, therefore, distinct from the sequential model, in which PDGF causes the release of S1P that subsequently acts at S1P 1 receptors in a manner that is completely independent of the PDGF receptor. Indeed, the results presented here clearly show an interaction between S1P and PDGF receptors that is independent of PDGF-stimulated S1P release from cells.
Part of the PDGF-stimulated response in ASM cells is insensitive to pertussis toxin and to S1P 1 receptor antisense plasmid construct. Thus, a fraction of the response does not involve G i or S1P 1 receptors. This is an interesting finding and is supported by other evidence that suggests that the GPCR input into signaling by receptor tyrosine kinases is dependent upon both receptor density and ligand concentration. Thus, high receptor tyrosine kinase density or ligand concentration has been shown to surmount the requirement for GPCR input (28,29). These findings suggest that receptor tyrosine kinases use at least two distinct signaling pathways to stimulate p42/p44 MAPK, one of which requires G-protein, whereas the other is probably initiated by auto-tyrosine phosphorylation of the growth factor receptor (28,29). Indeed, our previous findings showed that the G-protein-dependent activation of p42/p44 MAPK in response to PDGF occurs via a mechanism that does not require PDGF-dependent tyrosine phosphorylation of the PDGF␤ receptor (19). Moreover, we showed that overexpression of G i suppresses PDGF␤ receptor tyrosine phosphorylation while increasing PDGF-stimulated activation of p42/p44 MAPK (19). This is supported by studies from Freedman et al. (21), who have recently shown that GRK2-catalyzed phosphorylation of the PDGF␤ receptor suppresses PDGF-stimulated auto-tyrosine phosphorylation of the PDGF␤ receptor. Thus, G-protein input may switch growth factor receptor signaling such that stimulation of p42/p44 MAPK is less dependent upon recruitment of proteins to tyrosine phosphates on the growth factor receptor and is more dependent upon a G i ␣/GRK2/␤arrestin I-mediated pathway. Taken together with the findings presented in Alderton et al. (19) and in this current paper we propose that PDGF␤ receptor endocytic signaling mediated by G-protein might be initiated by GRK2 and ␤-arrestin I that are recruited indirectly to the PDGF␤ receptor via its association with the S1P 1 receptor. In this regard, Watterson et al. (30) have recently shown constitutive association of GRK2 with the S1P 1 receptors.
Endocytosis of the PDGF␤ receptor signal complexes is required for activation of p42/p44 MAPK (31), and our results suggest that this is driven by a G i PCR-dependent mechanism. Thus, increased internalization of PDGF␤ receptor-S1P 1 receptor complexes in response to combined stimulation with S1P and PDGF will improve the efficiency with which PDGF (and vice versa, S1P) stimulates the p42/p44 MAPK pathway.
We have also reported that PDGF stimulates a pertussis toxin-sensitive tyrosine phosphorylation of the Grb-2 associated binding protein, Gab1, in ASM cells (31). This appears to be dependent upon G i ␣ and c-Src (25,31). In previous studies, we have shown that c-Src is activated by PDGF, and this can be blocked by pretreating ASM cells with pertussis toxin (25). Furthermore, there is an indication that GRK2 is regulated by c-Src (21), thereby providing a functional link between the PDGF␤ receptor and GRK2. The tyrosine phosphorylation of Gab1 promotes the binding of phosphoinositide 3-kinase 1a (PI3K1a), which in turn induces association of dynamin II to the PI3K1a⅐Gab1⅐Grb-2 complex. This process appears to be essential for clathrin-mediated endocytosis of PDGF receptor signal complexes that include Raf-MAPK kinase 1 for re-localization with and activation of cytoplasmic p42/p44 MAPK (31). In this regard, dynamin II might function to promote "pinching off " of the endocytic vesicles containing active PDGF␤ receptor-S1P 1 receptor signal complexes for subsequent activation of p42/p44 MAPK in the cytoplasm.
Others report co-internalization of GPCR-receptor tyrosine kinase complexes in response to GPCR agonist. For instance, Maudsley et al. (32) show that isoprenaline induces transactivation of the EGF receptor, resulting in EGF receptor tyrosine phosphorylation and complex formation with ␤-adrenergic receptors. Under these conditions, isoprenaline induces the cointernalization of ␤-adrenergic receptors and EGF receptors to the same endocytic vesicles. However, Maudsley et al. (32) report that EGF alone does not induce internalization of the ␤-adrenergic receptor. This contrasts with our findings, where we show that PDGF induces co-internalization of both PDGF␤ receptor and the GPCR, S1P 1 . In addition, the internalization of the PDGF␤ receptor in response to S1P is significant because this does not involve PDGF release. We conclude that integrative signaling and co-internalization of both PDGF␤ receptor and S1P 1 receptor in response to PDGF and/or S1P is a novel important mechanism. It is clearly a different mechanism compared with receptor tyrosine kinase transactivation by GPCR agonists.
The formation of a complex between the PDGF␤ receptor and S1P 1 receptor may be a prototypical example for other growth factor receptors. This is supported by several studies showing that a number of growth factors use classic GPCR-mediated signaling pathways to stimulate p42/p44 MAPK in mammalian cells (33,34).
Finally, we show that hSK1 is constitutively exported and can support activation of p42/p44 MAPK by exogenous sphingosine. The export of hSK1 is not regulated by PDGF and, therefore, does not contribute to signaling by this growth factor. However, it is possible that the presentation of sphingosine on other cell types and its conversion to S1P by exported kinase from ASM cells might enable S1P to act with PDGF on the PDGF␤ receptor-S1P 1 receptor complex to induce biological responses in vivo. This would also represent a novel form of intercellular communication.
There is a possibility that a subfraction of ASM cells undergoes apoptosis in culture. This raises the intriguing question, Are cell death mechanisms associated with selective export of SK1 from cells, thereby reducing its ability to promote cell survival via an intracellular action? This is possible if there is hindered supply of sphingosine presented from other cells. However, contact with other cells that present sphingosine might effectively convert the export of SK into a cell survival action, since S1P (acting at cell surface S1P receptors) increases PDGF-stimulated p42/p44 MAPK activation and DNA synthesis in ASM cells (24).
In summary, our findings provide important information on integration of G-protein-mediated signals by receptor tyrosine kinases in mammalian cells by a mechanism that differs from sequential regulation and/or transactivation of receptor tyrosine kinases by GPCR agonists.