Sphingosine 1-Phosphate Triggers Both Apoptotic and Survival Signals for Human Hepatic Myofibroblasts*

Hepatic myofibroblasts (hMFs) are central in the development of liver fibrosis during chronic liver diseases, and their removal by apoptosis contributes to the resolution of liver fibrosis. We previously identified Edg receptors for sphingosine 1-phosphate (S1P) in human hMFs. Here, we investigated the effects of S1P on hMF apoptosis. S1P reduced viability of serum-deprived hMFs by an apoptotic process that was unrelated to the conversion of S1P into sphingosine and ceramide. The apoptotic effects of S1P were receptor-independent because dihydro-S1P, an Edg agonist, had no effect. S1P also stimulated a receptor-dependent survival pathway, revealed by enhanced activation of caspase-3 by S1P in the presence of pertussis toxin. Cell survival relied on two pertussis toxin-sensitive events, activation of ERK and activation of phosphatidylinositol 3-kinase (PI3K)/Akt by S1P. Both pathways were also activated by dihydro-S1P. Blunting either ERK or PI3K enhanced caspase-3 stimulation by S1P, and simultaneous inhibition of both pathways resulted in additive effects on caspase-3 activation. In conclusion, S1P induces apoptosis of human hMFs via a receptor-independent mechanism and stimulates a survival pathway following activation of Edg receptors. The survival pathway arises from the sequential activation of Gi/Goproteins and independent stimulations of ERK and PI3K/Akt. Therefore, blocking Edg receptors may sensitize hepatic myofibroblasts to apoptosis by S1P.

Liver fibrosis is the common response to chronic liver injury and is characterized by increased deposition and altered composition of extracellular matrix. This fibrogenic process is consecutive to intense proliferation and accumulation of myofibroblasts that synthesize fibrosis components and proinflammatory cytokines (1). Apoptosis of hepatic myofibroblasts is emerging as a key event in the regression of liver fibrogenesis. Indeed, it has been shown in experimental models of liver fibrosis that withdrawal of the offending agent is associated with apoptosis of hepatic myofibroblasts, followed by activation of fibrolysis mechanisms and regression of fibrosis (2). Therefore, identification of factors that govern death and survival of liver fibrogenic cells is of crucial interest for the design of antifibrotic therapies. Several apoptotic factors for these cells have recently been identified, such as soluble Fas (CD95/ APO-1) ligand, nerve growth factor, benzodiazepines, gliotoxin, and 15-deoxy-⌬ 12,14 -prostaglandin J2 (3)(4)(5)(6)(7)(8).
Recent lines of evidence suggest that sphingolipids, in addition to being structural constituents of cell membranes, play a key role as signaling molecules. In this respect, metabolites of sphingolipids, including ceramide, sphingosine, and sphingosine 1-phosphate (S1P), 1 have emerged as a new class of lipid messengers that regulate cell proliferation, differentiation, and survival (9 -11). Ceramide and sphingosine are generated after receptor-coupled activation of sphingomyelinase and ceramidase, respectively, and have generally been linked to cell growth arrest and apoptosis (9 -11). S1P is a downstream metabolite of ceramide that is produced after phosphorylation of sphingosine by sphingosine kinase (10). In contrast to ceramide and sphingosine, S1P generally elicits mitogenic and antiapoptotic effects (12), although growth-inhibitory and apoptotic effects have occasionally been described (13)(14)(15)(16)(17)(18). S1P modulates cell function by two distinct mechanisms, either as an intracellular messenger or as a ligand of a family of G proteincoupled receptors, the Edg (endothelial differentiation gene) receptors, which includes Edg-1, -3, -5, -6, and -8 (10,19). S1P acts as a messenger for various stimuli, including growth factors, cytokines, or oxidized low density lipoproteins (11,12), and its intracellular effects are mimicked by exogenous addition of micromolar concentrations of the sphingolipid (11,12). The intracellular pathway is involved in modulation of cell proliferation, positive or negative regulation of apoptosis, expression of adhesion molecules, and cytoskeletal remodeling (11,12). In contrast, Edg receptor-dependent actions of S1P are elicited by submicromolar concentrations of S1P and include angiogenic effects, vascular maturation (Edg-1), positive or negative regulation of cell migration, proliferation, and survival (19).
We previously identified S1P receptors (Edg-1, -3, and -5) in human hepatic myofibroblasts and noted that they relay the antiproliferative effect of S1P (13). In the present study, we show that S1P modulates apoptosis of hepatic myofibroblasts with a dual effect. The sphingolipid is proapoptotic via a receptor-independent pathway and antiapoptotic after activation of Edg receptors.
Cell Isolation and Characterization-Human hMFs were obtained by outgrowth of explants prepared from surgical specimens of normal liver, as described previously (20). This procedure was performed in accordance with ethical regulations imposed by French legislation. Cells were used between the third and seventh passage. The myofibroblastic nature of the cells was routinely evaluated by positivity for smooth muscle ␣-actin by immunohistochemistry, as described previously (20). The cultures were also found to express two markers of rat hepatic myofibroblasts, fibulin-2 and interleukin-6, but not the protease P100, a marker for rat hepatic stellate cells (21).
Cell Viability-Cells (7,000 cells/well in 96-well plates) were seeded in Dulbecco's modified Eagle's medium containing 5% human serum and 5% fetal calf serum (DMEM5/5), allowed to attach overnight, serum-starved for 48 h, and treated for the indicated periods of time with the indicated effectors. CellTiter 96 AQ ueous One Solution reagent was added to each well, and absorbance was recorded at 490 nm.
Apoptosis Assays-All the following techniques for measuring apoptosis were performed on nonconfluent cells allowed to attach overnight in DMEM5/5 and serum-starved for 48 h. Nuclear morphology was assayed by DAPI staining. Cells (10,000 cells/cm 2 ) in Lab-Tek chamber slides (Nalge Nunc International) were treated for the indicated periods of time with the effectors, fixed in 2% paraformaldehyde, stained with DAPI, and viewed under a fluorescence microscope (Zeiss). Caspase-3like activity was assayed on cell lysates obtained as follows. After treatment of cells (200,000 cells/60-mm dish) with the indicated effectors, floating cells were collected and centrifuged, and the pellet was lysed in 50 l of lysis buffer containing 50 mM HEPES, pH 7.4, 100 mM NaCl, 1% Nonidet P-40, 1 mM EDTA (pH 8.0), 1 mM dithiothreitol, 2 g/ml leupeptin, and 2 g/ml aprotinin. Adherent cells were washed three times with cold phosphate-buffered saline and lysed for 10 min on ice in 200 l of lysis buffer. Lysates from adherent and floating cells were pooled and centrifuged, and the supernatant was stored at Ϫ80°C until use. Caspase-3-like activity was measured in 200 l of assay buffer containing 100 mM HEPES, pH 7.4, 10% sucrose, 10 mM dithiothreitol, 500 M EDTA, 50 g of protein, and 20 M AC-DEVD-AFC as fluorogenic substrate. After 8 h at 37°C, the fluorescence of the reaction mixture was determined with a spectrofluorometer (FL600 Microplate Fluorescence Reader; BIO-TEK, Saint-Quentin, France), with excitation and emission wavelengths of 400 and 530 nm, respectively. DNA laddering was assayed by agarose gel electrophoresis of total DNA extracted from cells (500,000 cells/100-mm dish) treated with the indicated effectors. Total DNA was extracted using the Apoptotic DNA Ladder Kit according to the manufacturer's instructions and was further incubated with 20 g/ml RNase (DNase-free) for 10 min at 30°C. One g of DNA was electrophoresed on a 2% agarose gel stained with SYBR Green I and analyzed by PhosphorImager (Molecular Dynamics).
Immunoblot Analysis of the Phosphorylation State of ERK and Akt-Electrophoresis was performed on 10% SDS-polyacrylamide gel (25 mA/gel). Proteins were transferred to Protran nitrocellulose transfer membranes (Schleicher & Schü ll) by electroblotting using Tris/glycine/ SDS buffer containing 20% methanol (120 mA for 3 h at 4°C). Blots were incubated for 1 h at room temperature in TBST (10 mM Tris-HCl, pH 8, 150 mM NaCl, and 0.05% Tween 20) supplemented with 5% nonfat dry milk and incubated overnight with primary antibodies against phospho-ERK, phospho-Akt, or ␤-actin (1:1,000 dilution) in TBST supplemented with 5% nonfat dry milk at 4°C. Membranes were washed three times with TBST and incubated with peroxidase-conjugated affinipure donkey anti-rabbit IgG (1:10,000 dilution) for 1 h at room temperature in TBST containing 5% nonfat dry milk. Membranes were washed three times with TBST, and the peroxidase activity was determined using the enhanced chemiluminescence Western blot detection system (ECL ϩ ; Amersham Biosciences). Equal loading of proteins in each lane was confirmed by checking ␤-actin expression .
Statistics-Results are expressed as the mean Ϯ S.E. of n experiments. Results were analyzed by repeated-measures ANOVA or twotailed Student's t test, as appropriate, and p Ͻ 0.05 was considered significant.

Sphingosine 1-Phosphate Induces Apoptosis of Human
hMFs-Sphingosine 1-phosphate strongly reduced viability of serum-deprived hMFs (Fig. 1). Indeed, after 16 h, S1P caused shrinkage, rounding, and detachment of hMFs, as assessed by phase-contrast analysis, whereas serum deprivation alone (control) did not affect cell morphology (Fig. 1A). S1P also reduced cell viability by 80% (Fig. 1B). Addition of 5% human serum totally protected human hMFs from the cytotoxic effect of S1P (data not shown), as described in other cell types (22) and as reported previously in human hMFs when using 15deoxy-⌬ 12,14 -prostaglandin J2 as an apoptotic factor (8). Further experiments indicated that apoptosis was responsible for hMF cell death. After DAPI staining, 80% of the cells exhibited condensed nuclei after 24 h of treatment with 15 M S1P, whereas serum-deprived cells displayed normal nuclei ( Fig.  2A). S1P-treated cells showed dramatic DNA laddering on agarose gel electrophoresis, whereas control serum-deprived cells exhibited intact DNA (Fig. 2B). To further characterize the apoptotic effect of S1P on hMFs, we assessed the effects of the general caspase inhibitor ZVAD-fmk. As shown in Fig. 2, AϪC, ZVAD-fmk reduced S1P-induced nuclei fragmentation, DNA laddering, and cell death. Accordingly, S1P caused a timedependent activation of caspase 3-like activity, as assessed by DEVDase activity. Maximal activation occurred after 8 -10 h of stimulation and was blunted by ZVAD-fmk (Fig. 2D).
Taken together, these data demonstrate that S1P induces caspase-dependent apoptosis of human hMFs.
The Apoptotic Effect of S1P Is Not Related to Its Conversion into Sphingosine or Ceramide-Although S1P is mostly known as an antiapoptotic factor, apoptotic effects have occasionally been described. For example, in mesangial cells, the cytotoxic properties of S1P have been attributed to its conversion into sphingosine (14). In addition, S1P may also be ultimately converted into ceramide, a potent apoptotic factor for several cells (23). Indeed, both ceramide and sphingosine induced apoptosis of human hMFs, as shown by increased nuclei fragmentation (Fig. 3A), DNA laddering (Fig. 3B), and decreased cell viability (Fig. 3C). Therefore, we investigated whether the cytotoxic effects of S1P in human hMFs result from its conversion into either sphingosine or ceramide. As shown in Fig. 3C, apoptotic effects of ceramide and sphingosine required high micromolar concentrations like S1P, arguing against an apoptotic effect of S1P related to metabolite conversion. Moreover, an inhibitor of ceramide synthase, fumonisin B1, did not affect S1P-induced cell death (Fig. 3, A and C), and ceramide triggered activation of caspase-3-like activity only after 18 h of incubation, in contrast to S1P, which maximally activated the enzyme within 8 h (Fig. 3D). Finally, sphingosine did not affect caspase-3-like activity (Fig. 3D), nor did Z-VADfmk alter sphingosine-induced apoptosis (data not shown).
Taken together, these results strongly suggest that the apoptotic effects of S1P are not due to its conversion into sphingosine and ceramide.
The Apoptotic Effects of S1P are Edg-independent-We have previously identified Edg-1, Edg-3, and Edg-5 in human hepatic myofibroblasts and shown their coupling to pertussis toxin-sensitive G proteins (13). To determine whether S1P triggers apoptosis via Edg, we used dihydro-S1P, a structural analog of S1P that binds and activates Edg receptors and has no intracellular effects. Dihydro-S1P did not affect either hMF cell viability (data not shown) or caspase-3 activity (Fig. 4). Furthermore, pertussis toxin (PTX) did not block caspase-3 activation by S1P. In contrast, surprisingly, PTX pretreatment strongly enhanced stimulation of caspase-3 by S1P and slightly increased basal activity (Fig. 4). PTX maximally increased the enzyme activation by 10 -11-fold, as compared with 2-2.5-fold in cells exposed to S1P alone (Fig. 4).
These results suggest that the apoptotic effect of S1P occurs independently of Edg receptor activation. They also reveal an antiapoptotic effect of S1P, via a PTX-sensitive pathway. In the next series of experiments, we further characterized the mechanisms of this antiapoptotic pathway. An Edg-dependent Pathway Limits the Apoptotic Effect of S1P via Activation of ERK and PI3K/Akt-The antiapoptotic functions of ERK and PI3K/Akt are well established (24,25). To gain further insight into the survival mechanisms by which S1P counteracts its own apoptotic effect, we investigated the respective role of ERK and PI3K/Akt using selective inhibitors.
The ERK inhibitor UO126 potentiated caspase-3 activation by S1P, without significantly affecting basal activity (Fig. 5A). Whereas S1P alone maximally increased caspase-3 activity by 2-2.5-fold, addition of the ERK inhibitor enhanced activation of caspase-3 by S1P to 5-6-fold (Fig. 5A). Similar results were obtained with PD98059, another ERK inhibitor (Fig. 5A, inset). These results suggested a role for ERK in the survival signal elicited by S1P. Accordingly, submicromolar concentrations of S1P stimulated phosphorylation of ERK, an effect mimicked by dihydro-S1P (Fig. 5B). Stimulation of ERK by S1P was blunted in pertussis toxin-treated hMFs (Fig. 5B) and in cells pretreated with UO126 or PD98059 (Fig. 5B). Collectively, these data suggest that the survival pathway triggered by S1P is Edg-dependent and involves activation of ERK via a pertussis toxin-sensitive G protein.
Blocking PI3K by its inhibitor LY294002 also enhanced activation of caspase-3 by S1P and had a minimal effect on basal activity (Fig. 6A). In the presence of LY294002, S1P maximally increased caspase-3 activation by 5-fold, as compared with 2-2.5-fold in its absence (Fig. 6A). Similar results were observed with wortmannin, another PI3K inhibitor (Fig. 6A, inset). These results suggested that a PI3K/Akt pathway is involved in the survival effect of S1P. Accordingly, both S1P and dihydro-S1P stimulated phosphorylation of Akt (Fig. 6B). Phosphorylation of Akt by S1P was blunted in pertussis toxin-treated cells and in cells pretreated with LY294002 or wortmannin (Fig. 6B). These data indicate that, in addition to ERK, the survival pathway of S1P also involves activation of PI3K/Akt, following activation of Edg coupled to G i /G o proteins. To investigate whether S1P regulates ERK and PI3K/Akt independently, activation of caspase-3 by S1P was measured in the presence of a combination of ERK and PI3K inhibitors. As shown in Fig. 7A, pretreatment of cells with both UO126 and LY294002 strongly potentiated activation of caspase-3 by S1P. Combination of the ERK and PI3K inhibitors exerted additive effects on caspase-3 activation, which was increased by 13-14-fold in S1P-treated cells, as compared with 5-6-fold in cells exposed to either UO126 or LY294002 (Fig. 7A). Similar results were observed when using a combination of UO126 and wortmannin or of PD98059 and LY294002 (Fig.  7B). Therefore, ERK and PI3K independently contribute to the antiapoptotic effect of S1P. Accordingly, blocking ERK with UO126 or PD98059 did not affect phosphorylation of Akt by S1P (Fig. 7C); conversely, blocking PI3K by LY294002 or wortmannin did not alter phosphorylation of ERK by S1P (Fig. 7D).
Taken together, these results suggest that the antiapoptotic effect of S1P arises from the sequential activation of Edg receptors and G i /G o proteins and independent stimulation of ERK and PI3K/Akt. DISCUSSION We show here that sphingosine 1-phosphate exerts dual opposite effects on apoptosis of hepatic myofibroblasts. The sphingolipid is proapoptotic via a receptor-independent mechanism and stimulates a survival pathway after activation of Edg receptors (Fig. 8).
Hepatic myofibroblasts play a central role in the development of liver fibrosis. During this process, they proliferate, accumulate in the diseased liver, and secrete fibrosis components (1). Therefore, blockade of hepatic myofibroblast proliferation and enhancement of their apoptosis are considered as potential key targets for the development of antifibrotic drugs. We previously reported the growth-inhibitory properties of endothelin-1, tumor necrosis factor ␣, C-type natriuretic peptide, and sphingosine 1-phosphate in human hMFs (13, 26 -29). We show here that in addition to its growth-inhibitory properties, S1P is also a potent apoptotic factor for hepatic myofibroblasts, based on several lines of evidence ( Figs. 1 and 2). First, S1P induces hMF rounding and detachment. Second, S1P-treated hMFs exhibit condensed nuclei upon DAPI staining. Finally, S1P stimulates caspase-3-like activity, followed by loss of hMF viability, and the caspase inhibitor ZVAD-fmk blunts S1Pinduced cell death. Overall, these features are the hallmarks of programmed cell death. Our results therefore identify S1P as a novel apoptotic stimulus for human hepatic myofibroblasts. Other factors that trigger apoptosis of liver fibrogenic cells include soluble Fas (CD95/APO-1) ligand, nerve growth factor, benzodiazepines, gliotoxin, and 15-deoxy-⌬ 12,14 -prostaglandin J2 (3)(4)(5)(6)(7)(8).
Our results suggest that the apoptotic effects of S1P may be due to an intracellular effect of the sphingolipid, rather than to activation of plasma membrane Edg receptors. Indeed, the apoptotic effect of S1P requires micromolar concentrations, whereas we previously showed that, in human hMFs, Edg receptors are activated by submicromolar concentrations of the sphingolipid (13). In addition, dihydro-S1P, the S1P analog that binds to Edg receptors but is devoid of intracellular effects (30), did not induce hMF apoptosis. Edg-independent apoptotic effects of S1P have recently been reported in mesangial cells and attributed to its conversion into sphingosine (14). However, in human hMFs, both sphingolipids required similar high concentrations to induce cell death, arguing against the possibility that micromolar concentrations of S1P are being totally converted to micromolar concentrations of sphingosine. Furthermore, our data indicate that S1P and sphingosine induce apoptosis via distinct mechanisms. Whereas S1P-induced apoptosis was caspase-dependent, the cytotoxic effect of sphingosine was caspase-independent, raising the possibility that sphingosine may trigger apoptosis via a caspase-independent release of apoptosis-inducing factor, as described for ceramide in lymphoid cells (31). Intracellular mechanisms not related to metabolic conversion of S1P have also been implicated in S1Pinduced cell death. Thus, apoptosis triggered by S1P relies on a calcium-dependent pathway in hippocampal neurons or ovarian cancer cells (15,17), leading to activation of the protein phosphatase calcineurin (17). However, the calcium chelator BAPTA and the calcineurin inhibitors FK 506 and cyclosporin A did not affect S1P-induced hMF death (data not shown). We also investigated additional pathways, such as cyclooxygenase-2, which is central in the growth-inhibitory effects of S1P in human hMFs (13), and oxidative stress, which mediates apoptosis of these cells in response to 15-deoxy-⌬ 12,14 -prostaglandin J2. However, the cyclooxygenase-2 inhibitors NS-398 and ibuprofen or the antioxidant N-acetyl-cysteine did not affect the apoptotic effect (data not shown). The downstream pathways associated with the apoptotic effect of S1P are currently under investigation.
A major point of the present study is that, in addition to triggering apoptosis of human hMFs, S1P also stimulates a survival pathway that limits its apoptotic effect. The survival pathway involves activation of Edg receptors coupled to G i /G o proteins, as shown by the enhancement of caspase-3 activation in the presence of pertussis toxin. These results are in line with our previous findings showing that Edg receptors are expressed in human hMFs and coupled to PTX-sensitive G proteins (13). Investigation of the mechanisms by which S1P elicits antiapoptotic effects revealed the independent involvement of ERK and PI3K/Akt, two major protective mechanisms in cell survival and cell growth in response to growth factors or serum (24,25). Phosphorylation of ERK and that of Akt were activated by S1P and dihydro-S1P and blunted in pertussis toxintreated cells, reinforcing the implication of Edg receptors coupled to G i /G o proteins in the survival pathway elicited by S1P. Antiapoptotic effects of S1P were also described in endothelial cells via activation of Edg-1 (32,33) and in Edg-3-or Edg-5overexpressing cells (34). They involved activation of PTXsensitive pathway(s), namely ERK (30,(32)(33)(34). In contrast, stimulation of PI3K/Akt has merely been ascribed to its chemotactic effect (35)(36)(37). The fact that phosphorylation of ERK in response to S1P is not abolished by PI3K inhibitors and that, conversely, phosphorylation of Akt by S1P is not affected by ERK inhibitors suggests that both pathways are independently activated by S1P to mediate hMF survival. Other survival factors for hepatic myofibroblasts have been described, including insulin-like growth factor 1, integrin-mediated cell adhesion, and tissue inhibitor of metalloprotease-1 (38 -40), but the mechanism by which they protect hMFs from death has not been defined and may also involve ERK and PI3K/Akt.
Previous studies have shown that S1P may regulate both proliferation/survival and apoptosis in the same cell, depending on experimental conditions. Hence, S1P stimulates proliferation of mesangial cells or triggers apoptosis, depending on cell density (14). In ovarian cancer cells, S1P triggers death of cells in suspension but stimulates growth of attached cells (15). Positive and negative regulation of apoptosis in the same cell in response to a single stimulus has also been described. Thus, activation of ␤ adrenergic receptors in cardiomyocytes by norepinephrine stimulates a G i -dependent survival pathway that antagonizes its cAMP-dependent apoptotic effect (41). Tumor necrosis factor ␣, Fas ligand and thapsigargin also simultaneously generate survival and apoptotic signals (42)(43)(44). Interestingly, as described herein, blocking PI3K/Akt sensitizes human hepatocytes to apoptosis by tumor necrosis factor ␣ (45). In hMFs, we show that apoptosis and cell survival occur simultaneously in response to S1P, via distinct receptor-dependent and -independent mechanisms. Factors governing the balance between survival and apoptotic signaling pathways may therefore be critical to determine the final response to S1P.
Several lines of evidence indicate that apoptosis of hMFs may be beneficial for the resolution of liver fibrosis. First, accumulation of hMFs is one of the hallmarks of the fibrogenic process, and enhancement of their apoptosis may represent a promising therapeutic approach. Along these lines, recent studies indicate that resolution of hepatic fibrosis may occur upon apoptotic elimination of hMFs, and tissue inhibitor of metalloprotease-1 was recently identified as a critical factor in this process, but other factors such as S1P could also have been involved. In this respect, better knowledge of the regulation of S1P production by the liver is warranted. Evidence for secretion of S1P in cells and tissues is limited, probably due to its lipophilic nature and/or a high S1P degradative activity, although S1P has been detected in serum at a concentration of 0.5-1 M (46,47). Nevertheless, identification of Edg receptor antagonists may be useful to sensitize hepatic myofibroblasts to the apoptotic effect of S1P during chronic liver diseases.