Sphingosine 1-Phosphate, Present in Serum-derived Lipoproteins, Activates Matriptase*

We describe here a novel biological function of sphingosine 1-phosphate (S1P): the activation of a serine protease, matriptase. Matriptase is a type II integral membrane serine protease, expressed on the surface of a variety of epithelial cells; it may play an important role in tissue remodeling. We have previously reported that the activation of matriptase is regulated by serum. We have now identified the bioactive component from serum. First, the activity was observed to co-purify with lipoproteins by conventional liquid chromatography and immunoaffinity chromatography. The ability of lipoproteins to induce the activation of matriptase was further confirmed with commercial preparations of low density lipoprotein (LDL) and very low density lipoprotein (VLDL). Next, we observed that the bioactive component of LDL is associated with the phospholipid components of LDL. Fractionation of lipid components of LDL by thin layer chromatography (TLC) revealed that the bioactive component of LDL comigrates with S1P. Nanomolar concentrations of commercially obtained S1P were then observed to induce the rapid activation of matriptase on the surfaces of nontransformed human mammary epithelial cells. Other structurally related sphingolipids, including dihydro-S1P, ceramide 1-phosphates, and sphingosine phosphocholine as well as lysophosphatidic acid, can also induce the activation of matriptase, but at significantly higher concentrations than S1P. Furthermore, S1P-dependent matriptase activation is dependent on Ca2+ but not via Gi protein-coupled receptors. Our results demonstrate that bioactive phospholipids can function as nonprotein activators of a cell surface protease, suggesting a possible mechanistic link between S1P and normal and possibly pathologic tissue remodeling.

We describe here a novel biological function of sphingosine 1-phosphate (S1P): the activation of a serine protease, matriptase. Matriptase is a type II integral membrane serine protease, expressed on the surface of a variety of epithelial cells; it may play an important role in tissue remodeling. We have previously reported that the activation of matriptase is regulated by serum. We have now identified the bioactive component from serum. First, the activity was observed to co-purify with lipoproteins by conventional liquid chromatography and immunoaffinity chromatography. The ability of lipoproteins to induce the activation of matriptase was further confirmed with commercial preparations of low density lipoprotein (LDL) and very low density lipoprotein (VLDL). Next, we observed that the bioactive component of LDL is associated with the phospholipid components of LDL. Fractionation of lipid components of LDL by thin layer chromatography (TLC) revealed that the bioactive component of LDL comigrates with S1P. Nanomolar concentrations of commercially obtained S1P were then observed to induce the rapid activation of matriptase on the surfaces of nontransformed human mammary epithelial cells. Other structurally related sphingolipids, including dihydro-S1P, ceramide 1-phosphates, and sphingosine phosphocholine as well as lysophosphatidic acid, can also induce the activation of matriptase, but at significantly higher concentrations than S1P. Furthermore, S1P-dependent matriptase activation is dependent on Ca 2؉ but not via G i protein-coupled receptors. Our results demonstrate that bioactive phospholipids can function as nonprotein activators of a cell surface protease, suggesting a possible mechanistic link between S1P and normal and possibly pathologic tissue remodeling.
Physiologic and pathologic tissue remodeling involve an interplay of processes involving proteases and growth factors. Whereas growth factors induce cell proliferation and motility and function as chemoattractants, proteases directly promote cell migration by degrading the surrounding extracellular matrix barrier. However, the participation of proteases in physiologic and pathologic tissue remodeling is more complex, since proteolytic cleavage also serves to expose cryptic sites of extracellular matrix molecules, to activate latent growth factors, and to activate other protease cascades (1,2). This complexity is exemplified by the epithelium-derived serine protease, matriptase (3,4) (also known as membrane type serine protease-1 (5) or as epithin, its mouse homologue (6)). Matriptase has been demonstrated to activate the extracellular matrixdegrading protease, urokinase-type plasminogen activator, the potent, motility-inducing hepatocyte growth factor/scatter factor, and the calcium-regulating protease-activated receptor-2 (7,8). Therefore, matriptase may play an important role in tissue remodeling (7).
In addition to its catalytic serine protease domain, matriptase contains a transmembrane domain and two putative regulatory domains: two tandem repeats of a CUB (C1r/s, Uegf, and bone morphogenic protein-1) domain and four tandem repeats of a low density lipoprotein (LDL) 1 receptor domain (4). Matriptase is synthesized as a single-chain zymogen, which is presented on the surfaces of cells, where it is activated (9). The cognate inhibitor of matriptase, hepatocyte growth factor activator inhibitor-1 (HAI-1), can then bind to the active form of matriptase, resulting in the inhibition of its the proteolytic activity (10). The activity of matriptase at the surface of cells is further controlled by shedding of the activated protease (9). However, the mechanism of regulation of the initial activation step of matriptase is not known. During its activation, the zymogen form of matriptase is cleaved at its canonical activation motif to generate a disulfide-linked, two-chain form of the protease. This activation process results in conformational changes, creating new immunologic epitopes (9). Using an antibody that specifically recognizes the activated, twochain form of matriptase, we have recently demonstrated that exposure of human mammary epithelial cells to serum induces this activation step (9). We have now identified the factor present in serum responsible for the activation of matriptase. Our results show that sphingolipid metabolite, S1P, is a potent, specific activator of matriptase in intact epithelial cells.
Purification of the Serum-derived Inducer of Matriptase Activation-100 ml of human serum was diluted with 300 ml of phosphate-buffered saline (PBS). Ammonium sulfate powder was added to the diluted serum with continuous mixing to 60% saturation. The protein precipitate was centrifuged at 5000 ϫ g for 20 min. The pellet was dissolved and dialyzed against 20 mM Tris-HCl, pH 8.0. The majority of human IgG was first removed by passing serum through a protein A column. The flow-through was then loaded onto a DEAE-Sepharose Fast Flow column (Amersham Biosciences, Inc.), equilibrated with 20 mM Tris-HCl, pH 8.0. After washing, the column was eluted with a linear gradient of 0 -0.5 M NaCl in DEAE equilibration buffer. The fractions containing bioactivity were pooled. Some glycoproteins and the majority of human albumin were removed by passing the sample through a concanavalin A-Sepharose column (Pharmacia Biotech) and Cibacron blue 3GA-agarose column (Sigma). The partially purified serum factor was further subjected to gel filtration through a Sephacryl S-300 HR column (Amersham Biosciences), equilibrated with PBS. The column was eluted with PBS.
To further purify the serum factor using immunoaffinity chromatography, bioactive fractions from an S-300 HR column were used as immunogen to immunize mice, and a panel of mAbs against the subunits of this 160-kDa complex was generated using hybridoma technology, as described previously (10). Monoclonal antibody L24 was coupled to Sepharose 4B (5 mg/ml gel), based on the manufacturer's instructions. The L24 immunocolumn, equilibrated in PBS, was used to further purify the 160-kDa complex from S-300 column fractions. The bound protein was eluted with 0.1 M glycine buffer, pH 2.4, and immediately neutralized by Tris buffer.
For LDL extraction, the LDL sample was first extracted with 1 volume of chloroform/methanol; the resulting aqueous phase and interphase were either left at pH 7.4 or acidified to pH 4.5 and extracted with 1 volume of 1-butanol. The fractions were vacuum-dried and resuspended in ethanol.
Identification of the Active Components of LDL-LDL (5 mg) was extracted with chloroform, methanol, 0.5 M NaOH (3:2:1, v/v/v), and the phases were separated, as previously described (11). The lower phase, containing phospholipids and neutral lipids, was re-extracted. The basic aqueous phase, containing S1P and other lysophospholipids, devoid of sphingosine and the majority of phospholipids, was transferred to a siliconized glass tube. The organic phases were reextracted with 1 ml of methanol, 1 M NaCl (1:1, v/v) plus 50 l of 3 N NaOH, and the aqueous fractions were combined and reextracted by the addition of 50 l of 2 N HCl and 0.5 ml of CHCl 3 /methanol/concentrated HCl (100:200:1, v/v/v). Phases were separated by the addition of 250 l of CHCl 3 and 250 l of 1 M NaCl. After centrifugation, an aliquot of the organic phase was separated by TLC, using a solvent system consisting of 1-butanol/acetic acid/water (3:1:1). One-centimeter bands were scraped, and lipids were eluted with chloroform/methanol/HCl (100:100:1) and dried. Lipids were resuspended in PBS, containing 0.4% bovine serum albumin (1 ml), for measurements of matriptase activation as described above. S1P was identified by its co-migration with [ 32 P]S1P, as an internal standard (11), which was visualized and quantified with a Molecular Dynamics Storm PhosphorImager (Sunnyvale, CA).
Assay for Matriptase Activation-Immortalized 184A1N4 human mammary epithelial cells were maintained for 48 -72 h in 0.5% fetal bovine serum and then incubated for 1 h with IMEM containing the various purified fractions or tested compounds. Cells were lysed in 1% Triton X-100-PBS buffer, and equal amounts of whole cell lysates were resolved by SDS-PAGE. Activated matriptase and total matriptase were detected by Western blotting, using monoclonal antibodies M69 (activated matriptase) and M32 (total matriptase) (9), followed by a goat anti-mouse horseradish peroxidase-linked secondary antibody (Bio-Rad).
Immunofluorescence-Cells were fixed in 2% paraformaldehyde in PBS for 10 min at room temperature. Activated matriptase was stained using monoclonal antibody M69, followed by a goat anti-mouse fluorescein isothiocyanate-labeled secondary antibody (Jackson Immunoresearch, West Grove, PA). For actin staining, cells were fixed in 2% paraformaldehyde, permeabilized with 0.05% Triton X-100 in PBS, and actin was visualized with Texas Red-conjugated phalloidin (Molecular Probes, Inc., Eugene, OR). Cells were mounted using Prolong Antifade (Molecular Probes) and observed on an Olympus IX70 confocal microscope, using the 60ϫ (NA1.4) Olympus objective, and images were generated with the Olympus Floview system.
Expression of Matriptase in MDA-MB-435 Breast Cancer Cells-The full-length matriptase coding region was subcloned into the pcDNA3.1 expression vector (Invitrogen, Carlsbad, CA) and used to transfect matriptase-negative MDA-MB-435 breast cancer cells. Stable matriptase clones were isolated, after selection in hygromycin-containing media, and matriptase expression was confirmed by Western blotting and immunofluorescence, using the matriptase-specific monoclonal antibody M32.

Identification of Matriptase-activating Activity in Lipoproteins-
We have previously shown that exposure of 184A1N4 and MCF 10A human mammary epithelial cells to serum induces the activation of matriptase (9). Briefly, after 2 days of serum starvation of cells in culture, matriptase was detected, mainly in its single chain, latent form. This single chain form was detected by the anti-total (latent plus activated) matriptase mAb M32 but not by the anti-activated matriptase mAb M69 (Fig. 1). Activation of matriptase was induced by the addition of fresh medium containing 1% human serum. Activated matriptase was detected, either in its 70-kDa noncomplexed form, which migrates slightly slower in SDS-PAGE, compared with latent matriptase (9), or in the 120-kDa complex with its full-length cognate inhibitor, termed HAI-1 (Fig. 1). FIG. 4. Purification of the matriptase-activating activity using immunoaffinity chromatography. A panel of monoclonal antibodies was generated using the 160-kDa protein complex as immunogens. One of these mAbs, L24, was used to purify the 160-kDa complex, and the subunit profile of the 160-kDa complex was analyzed by SDS-PAGE under nonreduced conditions (A). The matriptase-activating activity of this 160-kDa complex was examined using 184A1N4 cells. Equal amounts of cell lysates from 160-kDa complex-treated cells (B, lane 2) and control cells (B, lane 1) were examined by immunoblot using antitwo-chain matriptase mAb M69. The bioactivity is associated with a 160-kDa (determined using a sizing column) protein complex, containing two major subunits of 27-and 15-kDa (on SDS-PAGE), under nonreduced conditions. The ratio of noncomplexed, activated matriptase to complexed, activated matriptase varied from experiment to experiment.
To identify the factor(s) in serum responsible for the activation of matriptase, we fractionated human serum by DEAE chromatography and examined chromatography fractions for their ability to induce activation of matriptase in 184A1N4 cells (Fig. 2). Although most of the human serum albumin was separated from the serum factors by DEAE chromatography, albumin still represented the major protein in the pooled DEAE fractions (data not shown). Since Cibacron blue dyeagarose and concanavalin A-agarose did not absorb the serum factor(s), we used both columns to remove most of the human serum albumin and some concanavalin A-binding proteins from the serum factor(s). Further purification of the serum factor(s) was carried out using an S-300 gel filtration column. Examination of these fractions indicated that the bioactivity was associated with a protein complex of ϳ160 kDa (Fig. 3). This complex contained several associated proteins, including two major polypeptides of 27 and 15 kDa, as determined by SDS-PAGE, under nonreducing conditions. Alternative chromatographic techniques, such as zinc chelating and hydroxyapitide chromatographies, resulted in the purification of a similar bioactive complex (data not shown). Further purification by immunoaffinity chromatography, using monoclonal antibodies directed against the 160-kDa complex, confirmed the role of the complex in the activation of matriptase (Fig. 4). The 27-kDa component of the 160-kDa complex was subjected to matrixassisted laser desorption ionization mass spectrometry (Howard Hughes Medical Institute and W. M. Keck Foundation Biotechnology Resource Laboratory, Yale University). Seventeen unique tryptic peptides matched 60% of the matrix-assisted laser desorption ionization mass spectrometry pattern for human apolipoprotein A1, one of the protein components of lipoproteins. This result suggested that the factor we purified was a lipoprotein.
Sphingosine 1-Phosphate Is the Serum Factor that Induces the Action of Matriptase-The ability of lipoproteins to induce the activation of matriptase was then confirmed by treating human mammary epithelial cells with commercial, purified lipoprotein preparations (Fig. 5). Both LDL and VLDL induced the activation of matriptase to the same extent as whole serum (Fig. 5A). Furthermore, charcoal-stripped serum, which is depleted of lipids (11), failed to induce the activation of matriptase (Fig. 5A). Inactivation of apolipoprotein function, by either boiling (Fig. 5B) or by protein acetylation (data not shown), did not decrease the ability of LDL to induce the activation of matriptase, suggesting that the bioactive factor is a low molecular weight component of the lipoproteins. Further extraction of LDL and serum with organic solvents (12) suggested that a phospholipid was the active component (Fig. 5C).
To examine which lipid in the LDL was responsible for matriptase activation, the lipid components of LDL were separated by extraction. The lipid fractions, containing fatty acids, neutral lipids, and phospholipids, had no matriptase-activating activity. However, the majority of the activity coincided with the lipids soluble in alkaline aqueous solution. This fraction is known to contain mainly lysophospholipids, such as S1P and LPA. To determine which of these lipids are active, this fraction was further separated by TLC, and the lipids were extracted from the silica gel (Fig. 6A). The migration of the matriptaseactivating activity coincided exactly with the migration of standard S1P (Fig. 6B). This result strongly suggested that S1P is the active component of LDL responsible for the activation of matriptase. Therefore, we next tested the ability of S1P to induce the activation of matriptase in our mammary epithelial cell culture system. S1P, at a concentration of 10 ng/ml (26 nM), could mimic serum-and LDL-induced activation of matriptase in 184A1N4 cells (Fig. 7A). Matriptase activation was detectable within 5 min of S1P addition, reaching a maximum within 10 min (Fig. 7B). Similar activity and potency of S1P was observed when S1P was dissolved in albumin solution or in ethanol.
In addition to S1P, we also determined whether other sphingolipids and lysophospholipids induce the activation of matriptase (Table I). A weaker response for the activation of matriptase was observed in varieties of phosphorylated sphingolipid metabolites (e.g. ceramide 1-phosphates, dihydro-S1P, and sphingosine phosphocholine). C2 ceramide-1-phosphate (237 nM) is more potent than C8 ceramide-1-phosphate (1 M). Overall, 1-alkyl-2-acetoyl-sn-glyco-3-phosphocholine (platelet activation factor) and 1-alkyl-2-hydroxy-sn-glyco-3-phosphocholine (lyso-platelet activation factor) have similar structures to SPC; however, the ether and amino differences found in these molecules are likely to result in their inactivity. In contrast, use of sphingolipid metabolites without phosphate groups (e.g. sphingosine, D-glucosyl-␤1-1Ј sphingosine, D-galatosyl-␤1-1Ј sphingosine (psychosine), and D-lactosyl-␤1-1Ј sphingosine, N,N-dimethylsphingosine, N,N,N-trimethylsphingosine, and ceramide), at concentrations up to 1 g/ml, failed to induce the activation of matriptase. These results suggest that phosphate modification of sphingolipids is essential for their ability to activate matriptase. Among other lysophopholipids tested, only LPA exhibited the bioactivity of matriptase activation but required a higher concentration. S1P also induced the activation of matriptase, in a similar manner, in an independently derived human mammary epithelial cell line, MCF-10A (data not shown). These results suggest that the ability of S1P to induce matriptase activation is highly specific to this phospholipid structure. S1P-dependent Matriptase Activation Occurs on the Surfaces of Cells-Immunofluorescence studies indicated that activated matriptase, following S1P treatment, was located on the surfaces of 184A1N4 epithelial cells (Fig. 8). Similar to what has been described in fibroblasts, S1P induced actin cytoskeletal reorganization (stress fiber formation) in 184A1N4 human mammary epithelial cells (Fig. 8). This actin cytoskeletal reorganization may result in more cell-cell contacts, where the activated matriptase was detected. S1P-dependent Matriptase Activation May Not Be via a Direct Interaction between S1P and Matriptase-We next wanted to test whether S1P could directly interact with matriptase, leading to its activation. We examined this possibility in stably transfected, matriptase-overexpressing MDA-MB-435 human breast cancer cells, which do not naturally express matriptase (13). Despite a high level of expression of latent matriptase protein, the activated form of matriptase was detected neither in cell lysates (Fig. 9) nor in the conditioned media of matriptase-overexpressing MDA-MB-435 cells (data not shown). Consistent with this observation, we found no increase in binding of S1P to matriptase-overexpressing MDA-MB-435 cells, compared with matriptase-negative parental MDA-MB-435 cells (data not shown). The lack of binding of S1P to matriptase was also observed in transiently matriptase-expressing MCF-7 breast cancer cells, where endogenous matriptase is also presented and activated (data not shown). These observations suggest that activation of matriptase by S1P appears not to be simply the interaction between matriptase and S1P.
Ca 2ϩ Is Required for the S1P-dependent Matriptase Activation-Because S1P has been shown to transiently increase intracellular Ca 2ϩ (14) and because the noncatalytic CUB and LDL receptor domains of matriptase may need Ca 2ϩ for their proper conformations and functions, we further examined the role of Ca 2ϩ in the S1P-dependent matriptase activation. Depletion of free intracellular Ca 2ϩ by pretreating cells with the intracellular calcium chelator 1,2-bis(o-aminophenoxy)ethane-N,N,NЈ,NЈ-tetraacetic acid tetraactetoxymethyl ester blocked S1P-induced matriptase activation (Fig. 10A). Similarly, extracellular calcium depletion with EGTA prevented matriptase activation, an inhibition reversible by the addition of calcium in the medium (Fig. 10B). However, treatment of 184A1N4 cells with the calcium ionophore ionomycin failed to mimic S1P-dependent matriptase activation (Fig. 10C), indicating that, although Ca 2ϩ is required for matriptase activation, S1P most likely does not induce the activation of matriptase through a mere increase of its free, intracellular concentration. FIG. 7. S1P induces the activation of matriptase. A, serum-starved 184A1N4 cells were treated for 1 h with the indicated concentrations of S1P. Equal amounts of whole cell lysates were analyzed by Western blot with mAb M69 (activated matriptase), mAb M32 (total matriptase), and M19 (HAI-1). B, serum-starved 184A1N4 cells were treated for the indicated amount of time with 10 ng/ml S1P. Equal amounts of whole cell lysates were analyzed by Western blotting with mAb M69 (activated matriptase). TC-MTP, twochain matriptase; SC-MTP, single-chain matriptase.

TABLE I Phospholipids inducing the activation of matriptase
Serum-starved 184A1N4 cells were stimulated for 1 h with the indicated concentrations of lipids. Activation of matriptase was evaluated by Western blotting of total cell lysates with M69 Ab. Ϫ, no response; ϩ, activation; ND, not determined. 1 ng/ml 10 ng/ml 100 ng/ml 500 ng/ml 1 g/ml 2 g/ml Considering that the activation of matriptase can be induced by nanomolar concentrations of S1P, the G protein-coupled receptors of endothelial differentiation gene (EDG) family may be involved in this process. We therefore tested whether S1Pdependent matriptase activation can be blocked by two anti-G protein drugs. First, we employed pertussis toxin, which ADP-ribosylates the G proteins, G i /G o , and uncouples these G proteins from EDG receptors. Second, we used suramin, which can also uncouple G proteins from receptors (15). Preincubation of cells with toxin or with suramin failed to block the activation of matriptase, induced by S1P (Fig. 11). In addition, both pertussis toxin and suramin failed to suppress the LPA-induced matriptase activation (Fig. 11). These results suggest that S1Pdependent matriptase may not be via a G i protein-coupled, receptor-dependent mechanism in immortalized mammary epithelial cells. DISCUSSION Matriptase, a New Cellular Effector of Sphingosine 1-Phosphate-S1P is recognized as a bioactive phospholipid mediator for a wide range of biological responses (16 -18). A variety of intracellular molecules have been identified as the downstream effectors of S1P signaling, via EDG receptors, such as adenylate cyclase, phospholipase C, c-Src, and Rho. While the intracellular targets of S1P have not yet been identified, S1P also acts intracellularly to regulate cellular proliferation and growth, and these activities are induced by S1P at micromolar levels (19). Higher (micromolar) concentrations of S1P also induce calcium mobilization, activation of phospholipase D, and tyrosine phosphorylation of p125 focal adhesion kinase (19), most likely in an EDG receptor-independent manner (20).
We have now identified matriptase in this study as a novel downstream effector on the surface of cells for this bioactive phospholipid. Our results suggest a new mechanism whereby S1P may modulate physiological and pathological states. Specifically, we have shown here that S1P activates matriptase on the surfaces of epithelial cells. This S1P-dependent activation of matriptase may serve subsequently to recruit and activate the two stromal derived matriptase substrates, the major extracellular matrix-degrading protease urokinase-type plasminogen activator, and the prominent cellular growth/motility factor, hepatocyte growth factor (7). Activated urokinase-type plasminogen activator and hepatocyte growth factor interact to regulate a wide range of cellular functions, including tissue morphogenesis, cellular motility, and matrix remodeling, and contribute to pathologic diseases such as cancer. S1P May Function as a Nonprotease Initiator for a Possible Serine Protease Cascade-Matriptase, a type II transmembrane serine protease, is biosynthesized as a single chain, latent zymogen and presented on the surfaces of epithelial and cancer cells (9). In nontransformed human mammary epithelial cells, the activation of matriptase depends on the exposure of cells to S1P. Like most serine proteases, activation of matriptase requires proteolytic cleavage within the canonical activation motif. For most serine proteases, this activation process is carried out by an upstream protease activator. Proteolytic activation of serine proteases by upstream activators is best illustrated by the complement system. In this system, an alternative mechanism of autocatalytic activation is adapted for the activation of the first protease in the cascade, C1r, which, in turn, triggers a proteolytic activation cascade, leading to the formation of the complement complex (21). The autoactivation of the C1r protease can be induced by the binding of C1 complex to an antibody-antigen complex (22), which serves a nonprotease function to induce the activation of the first protease in the complement system. For most serine protease cascades, however, little is known about the nature of the signal initiating the cascade. In the current study, we demonstrate that bioactive phospholipids, such as S1P, function as nonprotease initiators of a possible serine protease cascade (matriptase, urokinase-type plasminogen activator, and then plasminogen). Whether matriptase is the first protease in such a cascade to be activated by S1P and how matriptase becomes   1 and 3) and matriptase-transfected MDA-MB-435 clone C4 (lanes 2 and 4) were separated by SDS-PAGE and probed by Western blot using matriptase-specific monoclonal antibodies that recognize total (M32) (latent plus activated) or only activated (M69) matriptase. B, the latent form of the protease cannot be induced by the addition of S1P to matriptase-transfected MDA-MB-435 cells. S1P (10 ng/ml) was added to culture media, and cells were incubated for the indicated times prior to collection of whole cell lysates for Western blot analysis using M32 and M69 antibodies. activated remain to be further investigated. S1P-dependent Matriptase Activation May Not Be via the Direct Interaction of S1P with Matriptase or the Intracellular Binding Site of S1P-S1P has been shown to function both as an intracellular secondary messenger and as an extracellular signaling molecule (18,20,23). Therefore, there are at least three possible mechanisms by which S1P may activate matriptase. These include a direct interaction of S1P with matriptase, an indirect mechanism via intracellular matriptase-activating molecules, or indirectly by extracellular S1P receptors. Activation of a protease by the direct interaction with a lipid mediator was previously demonstrated for the activation of the lysosomal protease cathepsin D by ceramide (26). The fact that S1P failed to induce activation of matriptase in matriptase-overexpressing MDA-MB-435 human breast cancer cells suggests that direct exposure of matriptase to S1P is not sufficient for the activation of the protease. We have further observed that expression of matriptase, in naturally matriptase-negative, MDA-MB-435 cells, does not result in an increase in the binding of S1P to the cells compared with the matriptase-negative, parental MDA-MB-435 cells (data not shown). These results suggest that S1P may not directly bind to and activate matriptase. Furthermore, considering that the activation of matriptase by S1P can be induced by nanomolar concentrations of S1P, amplification of S1P signal by other molecules, leading to matriptase activation, may be needed in response to such a low concentration of S1P. Indeed, as in the case of ceramide-mediated activation of cathepsin D, much higher concentrations (up to micromolar levels) of ceramide are required to activate the protease in intact cells (26). The S1P analogue, dihydro-S1P, has been shown to bind to EDG receptors and to activate a receptor-mediated, extracellular S1P-dependent signaling pathway, but it does not elicit the intracellular action of S1P (23,27,28). Similar to S1P, exogenous dihydro-S1P induced the activation of matriptase in 184A1N4 cells (Table I). Consistent with what others have reported (28), dihydro-S1P was much less potent than S1P. Together, the low concentration (26 nM level) of exogenous S1P required to induce the activation of matriptase and the ability of dihydro-S1P to mimic the effect of S1P suggest that the S1P-mediated activation of matriptase does not occur via intracellular S1P action.
Activation of matriptase by S1P at nanomolar concentrations suggests that an amplification of S1P exogenous signal via receptor-like molecules may be needed. There are five EDG receptors (EDG1, -3, -5, -6, and -8) identified as high affinity receptors of S1P (29 -34). In addition to S1P, EDG1 acts as a low affinity receptor for LPA (K d 2.3 M) (35). EDG1 couples to a number of signaling pathways, mainly via pertussis toxinsensitive G i /G o proteins (30, 36, 37), but also through G z , which  11. S1P-dependent matriptase activation is pertussis toxin-and suramin-insensitive. Serum-starved 184A1N4 cells were pretreated with pertussis toxin (100 ng/ml) for 16 h or suramin (500 M) for 5 min and stimulated for 60 min with S1P or LPA. Equal amounts of total cell lysates were analyzed by Western blotting with mAb M69, directed against the activated matriptase. is insensitive to bacterial toxins (38). Therefore, although we have observed that pertussis toxin fails to suppress S1P-mediated activation of matriptase, we cannot exclude the role of EDG 1 in the activation of matriptase. Suramin acts as a specific functional antagonist for EDG 3 (39), and suramin also failed to block the activation of matriptase induced by S1P. This observation excludes EDG3 as a functional receptor for matriptase activation. Because EDG1, -3, and -5 were detected by RT-PCR in the matriptase-overexpressing MDA-MB-435 cells and because S1P failed to induce the activation of matriptase in this cell line, these results suggest that these three EDG receptors may not be involved in the activation of matriptase. The hypothesis is supported by the fact that pertussis toxin fails to suppress the S1P-dependent matriptase activation. Alternatively, this S1P-mediated activation of matriptase may require another, yet unidentified, cell surface receptor or anther unknown target to initiate the proteolytic activation cascade. Thus, the failure of S1P to induce activation of matriptase in matriptase-overexpressing MDA-MB-435 cells may be due to the lack of some of the cascade components in MDA-MB-435 cells.
In addition to S1P, SPC at 215 nM (100 ng/ml) and ceramide 1-phosphates at about 200 nM exhibit matriptase-activating activity; these concentrations are higher than those for S1P but low enough to suggest that their activity may be via non-EDG membrane receptors. Ovarian cancer G protein-coupled receptor 1 (OGR1) and G protein-coupled receptor GPR4 have been identified as high affinity (K d ϭ 33.3 nM for OGR1 and 36 nM for GPR4) receptors for SPC (25,40). 100 nM SPC is required to induce the activation of MAK kinase in HEK 293 cells transfected with OGR1 and GPR4 (25,40). A similar concentration of SPC is able to induce the activation of matriptase in 184A1N4 cells. The binding of SPC to OGR1 was not displaced by S1P (24). On the other hand, the binding of S1P to EDG3, but not to EDG1 and EDG5, can be displaced by SPC. These observations suggest that SPC may act through OGR1 or EDG3. As described above, EDG3 is unlikely to be the receptor involved in the activation of matriptase. Furthermore, the specific receptor of ceramide 1-phosphate has not yet been discovered. Therefore, matriptase activation can probably be induced by different sphingolipids, via different G protein-coupled receptors. Alternatively, these sphingolipids may act on an as yet unidentified receptor to induce the activation of matriptase.