Apoptotic Effect of Sphingosine 1-Phosphate and Increased Sphingosine 1-Phosphate Hydrolysis on Mesangial Cells Cultured at Low Cell Density*

The lipid mediator sphingosine 1-phosphate (S1P) may alter the proliferation of mesangial cells during pathophysiological processes. Here, S1P stimulated proliferation of rat mesangial cells and phosphorylation of MAPKs at subconfluent cell density. Both effects were inhibited by pertussis toxin treatment. Mesangial cells expressed several S1P receptors of theendothelial differentiation gene family: EDG-1, -3, -5, and -8. Conversely, S1P induced apoptosis at low cell density (2 × 104cells/cm2), which was demonstrated by flow cytometry and Hoechst staining. Apoptosis was observed also in quiescent or growing cells and was not reverted by lysophosphatidic acid or platelet-derived growth factor. S1P enhanced phosphorylation of SAPKs. Incubation with [33P]S1P, [3H]S1P, and [3H]sphingosine demonstrated increased S1P hydrolysis, resulting in enhanced intracellular sphingosine levels and decreased S1P levels. A rise in total ceramide levels was also observed; however, ceramide did not originate from [3H]sphingosine, and S1P-induced apoptosis was not inhibited by fumonisin B, precluding involvement of de novo ceramide synthesis in apoptosis. Therefore, we suggest that sphingosine accumulation and decreased S1P are primarily responsible for S1P-induced apoptosis. In conclusion, incubation of low-density mesangial cells with S1P results in apoptosis, presumably due to increased S1P hydrolysis.

Glomerular diseases are often characterized by the increasing proliferation rate of mesangial cells, also responsible for altered composition of the extracellular matrix, with both mechanisms leading to progressive fibrosis (1)(2)(3)(4). Because of this, the identification of agents regulating the growth of mesangial cells has been the topic of numerous studies. The prominent members of the family of lysophospholipids (LPLs), 1 the structurally related molecules sphingosine 1-phosphate (S1P) and lysophosphatidic acid (LPA; 1-acyl-2-hydroxy-sn-glycerol 3-phosphate), have attracted attention in the last decade as potent growth factors (see reviews in Refs. [5][6][7][8][9]. Activated platelets or blood cells secrete S1P and LPA, and so mesangial cell may be a target for LPLs during inflammatory and/or degenerative processes. Several studies have documented the proliferative effect of LPLs on these cells (10 -16). LPLs act through homologous G protein-coupled receptors of the endothelial differentiation gene (EDG) family, several members of which are known today. S1P preferentially activates EDG-1, -3, -5, -6, and -8, whereas LPA activates EDG-2, -4, and -7 (7). LPLs promote cell growth by activating several EDG-evoked signaling pathways. G i/o induces mitosis via the mitogen-activated protein kinase (MAPK) pathway; G q mediates mobilization of intracellular calcium; and G 12/13 activates Rho-related pathways. In addition, S1P may act as an intracellular second messenger (17). S1P activates multiple responses in various cells. For example, S1P induces migration of endothelial cells, and it is regarded, together with the expression of its receptor EDG-1, as a factor promoting angiogenesis (18). S1P is also considered as a factor of cell survival. This is the case for both exogenous S1P (9, 19 -21) as well as S1P generated by activation of sphingosine kinase by growth factors or by overexpression of the enzyme (22,23). Contrary to this, cell proliferation is negatively influenced by other sphingolipids such as ceramides and sphingosine (24 -27). Thus, the balance between S1P and ceramide/ sphingosine cellular levels is probably crucial to control the proliferation of mesangial cells (28,29).
We previously observed that the effect of S1P on mesangial cell proliferation is variable and depends upon the density of the cell culture (15). In the present study, we documented that an increased hydrolysis of S1P to sphingosine occurred in mesangial cells cultured at low cell density, leading to increased sphingosine and decreased S1P intracellular levels. This phenomenon, in association with the increased production of ceramide, seems a key factor in inducing the apoptosis of mesangial cells. To our knowledge, this is the first time that increased S1P hydrolysis as a mechanism leading to apoptosis has been observed in cells derived from a primary culture. It might also be relevant under pathophysiological conditions, where excessive mesangial cell proliferation is present.
Isolation of Rat Renal Glomeruli and Culture of Mesangial Cells-Renal glomeruli were isolated from kidneys from 2-month-old male Wistar rats weighing 150 -200 g. Most of the procedure was adapted from Striker and Striker (30) as previously described (31). Briefly, the kidneys were removed from their capsule, and the cortex fragments (3-4 mm 3 ) were submitted to four repeated enzymatic digestions with collagenase at 37°C in PBS solution. The glomeruli purified by successive mechanical sieving were retained on the 100-, 80-, and 60-m pore stainless steel sieves. Glomerular mesangial cells were obtained from glomerular preparations and cultured under standard conditions at 37°C in a humidified 5% (v/v) CO 2 incubator in RPMI 1640 medium supplemented with 10% (v/v) heat inactivated FCS plus 100 units/ml penicillin and 100 g/ml streptomycin. Passages of primary cultures were done after 3-4 weeks. To eliminate contamination by either epithelial or endothelial cells, the culture was performed in RPMI 1640 medium with D-valine until passage 5. Thereafter, experiments were performed between passages 5 and 20. Mesangial cells were characterized by their morphological appearance in phase contrast and their biological and biochemical properties as previously described (31): 1) positive staining with antibodies against myosin or vimentin and negative staining with antibodies against factor VIII or cytokeratins by standard immunofluorescence; 2) ability to contract in response to 10 Ϫ5 M angiotensin; and 3) ability to grow in selective D-valine-containing medium. In all experiments, cells were made quiescent by serum starvation in RPMI 1640 medium with 0.5% (v/v) FCS for 48 h. When testing the effect of cell density, equivalent numbers of cells were seeded on variable culture areas. Consequently, for analysis, we used equal cell material originating from the cell at different densities. Cell density varied between 1 ϫ 10 4 and 1.7 ϫ 10 5 cells/cm 2 as indicated below and in the figure legends.
[ 3 H]Thymidine Incorporation-Mesangial cells were seeded with RPMI 1640 medium containing 10% (v/v) FCS in tissue culture dishes at different densities as indicated. After 24 h, the medium was changed for 48 h to RPMI 1640 medium containing 0.5% (v/v) FCS to make the cells quiescent. The agents to be tested were then added for 24 h. At this time, [ 3 H]thymidine (0.5 Ci/ml) was added, and the incubation was continued for 4 h. Cells were washed with PBS, and the DNA was precipitated with 10% (w/v) trichloroacetic acid. The trichloroacetic acid-precipitable material was dissolved in 0.5 M NaOH, and the radioactivity was measured with a liquid scintillation counter. Data from triplicate values are expressed as a percentage of the average control value.
MTT Cell Proliferation Assay-Cells were cultured as described above for thymidine incorporation, except that the agents to be tested were added for 36 h. Cells were incubated with 5 g/ml MTT for 2 h at 37°C. The medium was aspirated, the formazan crystals formed were dissolved in isopropyl alcohol with 60 mM HCl, and the absorbance at 570 nm was determined with a spectrophotometer (Beckman Instruments). Data from triplicate values are expressed as a percentage of the average control value.
Lactate Dehydrogenase Assay-Lactate dehydrogenase release from damaged cells was used to assess cytotoxicity. Quiescent mesangial cells were cultured in 12-well dishes and treated with the stimulus as indicated. Lactate dehydrogenase activity was determined in both the supernatant and cells according to the manufacturer's instructions as previously described (11). Lactate dehydrogenase activity in the supernatant is expressed as percent of the total release (lactate dehydrogenase released from the supernatant and cells).
Flow Cytometry Analysis-Apoptotic cells were quantitatively evaluated by flow cytometry. Quiescent cells (5 ϫ 10 5 ) were cultured at high or low density and treated with the agonist or vehicle. At the indicated times, both adherent and non-adherent cells were harvested, washed twice with PBS, and fixed in 70% cold ethanol. Briefly, cells were labeled with propidium iodide as described previously (32) and analyzed in a flow cytometer (Coulter ELITE).
Hoechst Staining Assay-Intact cells were stained with Hoechst 33342 (5 g/ml) and propidium iodide (10 g/ml) for 10 min and analyzed in a fluorescence microscope (Zeiss Axioskop) with excitation at 360 nm. Under these conditions, the Hoechst dye stained all nuclei, whereas propidium iodide stained the nuclei of only necrotic cells with disrupted plasma membranes. With Hoechst staining, viable cells were observed with intact nuclei, and apoptotic cells with fragmented or condensed nuclei.
Measurement of S1P Lipid Phosphate Phosphohydrolase Activity in Mesangial Cells-Quiescent cells were seeded, cultured at low or high density, and treated with 10 M [ 33 P]S1P (10 Ci/mol) in 0.1% (w/v) BSA. At the indicated times, 0.5 ml of the incubation medium was transferred into 0.5 ml of 1 M HClO 4 to precipitate proteins and the labeled lipid. The 33 P i present in the medium was measured as described previously (34). The radioactivity was counted in a scintillation counter. The initial rate was deduced from experimental data obtained between 0 and 10 min in the linear part of the activity curve. Activity is expressed as nanomoles of inorganic phosphate released per 10 6 cells.

Metabolism of [ 3 H]S1P in Mesangial
Cells-Mesangial cells were seeded, cultured at low or high density, and treated with the agonists or vehicle in RPMI 1640 medium plus 0.01% (w/v) BSA as indicated. Lipids were then extracted from the extracellular medium by the Bligh and Dyer method (35). Lipids were extracted from cells as follows. Cells were washed twice with PBS and scraped in 0.75 ml of methanol. The plates were washed with an additional 0.75 ml of methanol, and 0.75 ml of chloroform was added to the pooled methanol extracts. The samples were stored overnight at Ϫ20°C, and the precipitated proteins were removed by centrifugation. S1P, sphingosine, and ceramides were analyzed by thin-layer chromatography with butanol-1/acetic acid/water (3:1:1, v/v) as solvent. Radioactivity was detected with an automatic TLC linear analyzer (Bertold, Wildbad, Germany). Spots corresponding to the different lipids were scraped, and the radioactivity was counted in a scintillation counter. A control incubation with radiolabeled substrates without cells demonstrated that ϳ5% of the compounds remained bound to the plate, with no clear time dependence. This value was considered as the zero value and was subtracted from all experimental data.
Measurement of Cellular Ceramides by the Diacylglycerol Kinase Assay-The cellular ceramide level was measured according to a method described previously (36) with slight modifications. Quiescent cells were seeded, cultured at low or high density, and treated with S1P as indicated. Cells were washed twice with PBS and scraped in 2 ml of PBS. Lipids were extracted according to the Bligh and Dyer method (35). Samples from the lower chloroform phase were dried and incubated with 0.08 N NaOH in chloroform/methanol (2:0.5, v/v) at 37°C for 30 min for alkaline hydrolysis of diacylglycerol. The reaction was stopped by lipid extraction, and ceramide was converted to ceramide 1-[ 33 P]phosphate by E. coli diacylglycerol kinase in the presence of [␥-33 P]ATP (25 Ci/mmol) as described above, except that the reaction was carried out for 30 min with 1 IU/ml diacylglycerol kinase. The labeled lipids were extracted by the Bligh and Dyer method (35) and separated by thin-layer chromatography in chloroform/methanol/acetone/acetic acid/water (10:2:4:2:1,v/v). The R F value for ceramide phosphate was 0.20 in this system. Radioactivity was measured with a PhosphorImager 445 SI (Molecular Dynamics, Inc.). Spots corresponding to ceramide phosphate were scraped, and the radioactivity was counted in a scintillation counter. Addition of C 16 -ceramide to the initial medium as a control testified for ϳ70% recovery during the overall procedure. This rate of recovery was taken into account to estimate the quantity of ceramide initially present in the assay medium.
Statistics-Results (expressed as means Ϯ S.E.) are representative of at least three experiments as indicated in the figure legends. Comparisons were done using Student's t test. Results were considered to be significant when p Ͻ 0.05.

Cell Density and Opposite Effects of S1P on Mesangial Cell
Proliferation-Although S1P acts as a mitogen toward diverse cell types, few reports have described an antiproliferative effect on cells derived from a primary culture (37,38). In a preliminary study, we observed that S1P exerts opposite effects on the proliferation rate of mesangial cells, depending upon the density of the cell culture (15). Therefore, our first aim was to assess precisely the effect of cell density on mesangial cell proliferation. We tested a range of cell densities varying between 1 ϫ 10 4 and 1.7 ϫ 10 5 cells/cm 2 (Fig. 1). The effect of S1P on [ 3 H]thymidine incorporation strongly depended upon the cell density, whereas under the same conditions, PDGF was constantly proliferative (data not shown). At subconfluent cell density, 10 M S1P increased DNA synthesis by up to 6-fold. The effect of 10 M S1P or 10 M LPA tested under the same conditions was similar (Fig. 2C). Above 1.5 ϫ 10 5 cells/cm 2 , the decrease in DNA synthesis was interpreted logically as a confluence-dependent inhibition of cell growth. When decreasing the cell density, the effect of S1P progressively reverted toward an unexpected antiproliferative one at the lowest densities. Because of these first results, we tested the effect of cell density more extensively. In the following experiments, low density corresponded to distinct cells with rare membrane contacts (1-2 ϫ 10 4 cells/mm 2 ), and high density corresponded to subconfluent cells (1-1.2 ϫ 10 5 cells/mm 2 ). S1P-induced Proliferation at High Cell Density-At high cell density (1 ϫ 10 5 cells/cm 2 ), the treatment of mesangial cells with S1P increased DNA synthesis as demonstrated by [ 3 H]thymidine incorporation ( Fig. 2A). Dose dependence was demonstrated by a significant difference between the effect of 1 and 10 M S1P (p Ͻ 0.05). The proliferative effect of S1P was confirmed by the MTT technique. Fig. 2B shows a 2.5-fold increase with 10 M S1P, with a 3-fold increase being induced by 10 M LPA as a positive control. Pretreatment of mesangial cells with pertussis toxin almost completely inhibited the S1Pinduced mitogenic response, as it did for 10 M LPA (Fig. 2C), suggesting the involvement of a G i -coupled pathway for the proliferative response to S1P. Phosphorylation of p42 and p44 MAPKs (ERK1/2) showed a two-wave phosphorylation profile induced by 10 M S1P that persisted for Ͼ2 h, as tested by Western blot analysis using anti-phospho-p42/44 antibody (Fig.  2D). This is in agreement with the results of the thymidine incorporation assay. In addition, activation of ERK1/2 was partly inhibited by pertussis toxin treatment (Fig. 2D).
Extracellular S1P is known to act through receptors of the EDG family. mRNA expression of EDG receptors by mesangial cells was examined using reverse transcription-PCR. As shown in Fig. 2E, most of the identified receptors for S1P were well expressed by mesangial cells, i.e. EDG-1, -3, -5, and -8, with EDG-1, -3, and -5 being expressed the most. As the sequence of the rat EDG-6 receptor was unknown, we amplified it using primers for human EDG-6. We could amplify a homologous form of human EDG-6 from genomic rat DNA. However, we were unable to amplify EDG-6 from either rat or human mesangial cell mRNAs. Altogether, these results suggest that mesangial cells do not express EDG-6. The ubiquitous LPA receptor EDG-2 was also well expressed by mesangial cells. S1P-induced Apoptosis at Low Cell Density-As mentioned above, S1P inhibited the proliferation of rat mesangial cells at the lowest cell densities (1-2 ϫ 10 4 cells/cm 2 ) (Fig. 1). This effect started at 1 M, was significant at 5 M S1P, and was at maximum at Ͼ10 M S1P (Fig. 3A). Moreover, S1P significantly decreased the [ 3 H]thymidine incorporation at 24 h when incubated simultaneously with mitogenic agents such as PDGF (40 ng/ml) and FCS (10%, v/v) (Fig. 3, B and C). The hypothesis of DNA synthesis being simply delayed was proved to be wrong in the case of longer S1P incubation (up to 30 h) (data not shown). To clarify the mechanism involved in the deleterious effect of S1P, we first excluded the toxic effect of S1P. Lactate dehydrogenase measured in the supernatant of cells stimulated with 10 M S1P precluded cell lysis even after 15 h, with the values being similar to those obtained in the control (7.64 Ϯ 1.07% versus 6.87 Ϯ 1.02% of lysed cells, respectively). Apoptosis was studied with Hoechst 33342 staining. Condensed chromatin fragments were observed in cells treated with S1P for 24 h (Fig. 4A). Under the same conditions, propidium iodide remained excluded from the cells, precluding significant necrosis. After 24 h, the percentage of apoptotic cells was 24.9% in cells treated with S1P versus 4.3% in control cells. Another piece of evidence that S1P promoted DNA fragmentation was obtained by flow cytometry analysis after propidium iodide staining (Fig. 4B, upper panel). After 24 h, hypodiploid DNA content increased by up to 18% in S1P-processed cells versus 7% in control cells. The discrepancy between Hoechst staining and cytometry analysis is probably due to the fact that part of the non-adherent apoptotic cells are lost during harvesting before cytometry analysis. In addition, condensed (but not fragmented) nuclei were probably not detected by cytometry. S1P was also tested in cells maintaining a growing mode in the presence of 5% FCS. Although a percentage of the cells in G 2 -S phase were detected under these conditions, the apoptotic effect of S1P remained significant, with the same being valid in the presence of PDGF (Fig. 4B, lower panel). S1P was also apoptotic in quiescent cells in the presence of 10 M LPA (Fig.  4C, upper panel), which acts as a growth factor for mesangial cells (11). LPA decreased the percentage of apoptosis in quiescent cells, but did not prevent apoptosis in the presence of S1P. Apoptosis persisted in growing cells in the presence of PDGF (Fig. 4C, lower panel). These data demonstrate that the apoptotic effect of S1P at low cell density at concentrations Ͼ1 M is not dependent on the privation of growth factors and is not reverted by them.
Phosphorylation and activation of SAPKs are important signaling pathways for programmed cell death. SAPK phosphorylation was detected in S1P-treated cells only when cultured at low density, with activation lasting for Ͼ2 h (Fig. 4C).
Apoptotic Effect of Sphingosine at Low Cell Density-As the proliferative effect of S1P reverted to apoptosis in cells cultured at low density, we supposed that the sphingolipid rheostat would be modified toward its apoptotic components under these conditions. We hypothesized that the conversion of S1P to sphingosine could be responsible for the apoptotic effect of S1P at low cell density (S1P possibly being dephosphorylated to sphingosine). Therefore, we tested the effect of sphingosine, which also appeared to be density-dependent, resulting in stimulated thymidine incorporation at high cell density and inhibition at low cell density (Fig. 5A). The effect of sphingosine on low-density cells was significant in the concentration range of 1-10 M (Fig. 5B). 1 M sphingosine inhibited the proliferation induced by PDGF (40 ng/ml), and significant apoptosis oc-curred with 10 M sphingosine, as demonstrated by flow cytometry analysis. (Note that the vehicle used in the sphingosine assay (Me 2 SO, 1:1000, v/v) was responsible for increased apoptosis in control cells.) S1P Phosphohydrolase Activity Involved in Apoptosis of Mesangial Cells-To test the hypothesis of increased conversion of S1P to sphingosine at low cell density, S1P hydrolysis was measured. [ 33 P]S1P was synthesized as described under "Experimental Procedures." The amount of 33 P i released in the medium over 1 h was significantly higher at low compared with high cell density (Fig. 6A), suggesting increased S1P hydrolysis. The initial kinetic rate was increased by 2-fold, being 21 Ϯ 2 nmol of P i /min/10 6 cells at high cell density versus 42 Ϯ 6 nmol at low cell density (means Ϯ S.E.). When co-incubated with S1P, LPA induced a significant dose-dependent decrease in S1P hydrolysis exclusively in cells cultured at low density (Fig. 6B).
Metabolism of S1P in Mesangial Cells at Low and High Density-To clarify the metabolic fate of S1P as a function of cell density, mesangial cells were incubated with [ 3 H]S1P. Cellular [ 3 H]S1P and [ 3 H]sphingosine levels increased progressively. At 120 min, the [ 3 H]S1P level was higher in highdensity cells. Conversely, the [ 3 H]sphingosine level increased more in low-than in high-density cells (Fig. 7, A and B). Over 120 min, the [ 3 H]S1P level decreased by ϳ25%/10 6 cells in the extracellular medium of low-density cells versus 18%/10 6  in cells cultured at low density (Fig. 7D). At 120 min, the [ 3 H]S1P level was significantly higher in high-than in lowdensity cells, whereas the [ 3 H]sphingosine level increased more in the low-than in the high-density cells (Fig. 7, E and F) (Fig. 7C). However, as ceramides are usually involved in the apoptosis process, a hypothesis could be made that ceramide can be independently generated. To clarify this point, the total ceramides were quantified by the diacylglycerol kinase technique during S1P-induced apoptosis. Upon S1P stimulation, a significantly higher increase in the ceramide level was observed in low-compared with high-density cells (Fig. 8). We then tested the hypothesis that an increased de novo synthesis of ceramide is involved in S1P-induced apoptosis. Low-density cells were subjected for 24 h to 10 M S1P in the presence of 15 M fumonisin B, used as a ceramide synthase inhibitor (20). Apoptosis induced by etoposide was used as a control for the effect of fumonisin B. logical situations (5)(6)(7)(8)(9). Regarding sphingolipids, the balance between S1P and sphingosine is considered as a rheostat for cells deciding between death and survival (17,20). We have previously observed that the effect of S1P on the proliferation of mesangial cells is variable (15). In that study, it was suggested that the role of the density of the cell culture is an important factor, something that we have tested further here.
We have confirmed in this study that the effect of S1P is highly dependent upon the density of the cell culture, as demonstrated by [ 3 H]thymidine incorporation assays. At subconfluent cell density, S1P stimulated the proliferation of rat mesangial cells in a dose-dependent fashion. S1P-induced proliferation involved the long-term phosphorylation of the MAPKs ERK1 and ERK2. This result is in agreement with other studies in mesangial cells where sustained activation of ERK1/2 has been related to mitogenesis induced by growth factors such as PDGF, LPA, and angiotensin II (11,39,40). The known S1P receptors of the EDG family are coupled with heterotrimeric G proteins. In this case, the steps between the receptor and Ras-ERK1/2 activation may involve various processes such as G i /␤␥-linked pathways and transactivation (41)(42)(43)(44). In our experiments, [ 3 H]thymidine incorporation and ERK1/2 phosphorylation were pertussis toxin-dependent, suggesting that G i -dependent pathways are involved in S1P-induced mesangial cell proliferation. S1P is a ligand of several receptors of the EDG family. We found that several of these receptors were well expressed by rat mesangial cells (i.e. EDG-1, -3, -5, and -8), some of them being potentially coupled to G i (7)(8)(9). As recently reviewed, the biological properties of S1P may be due to both intracellular and extracellular signaling (17). However, ERK1/2 activation by S1P is frequently pertussis toxin-dependent (8), as in our model; and in many cases, S1P-induced proliferation can be related to EDG-dependent signaling.
The unexpected result of this study was that S1P became antiproliferative when the density of the cell culture was decreased. Under these conditions, S1P even inhibited the proliferation induced by PDGF, LPA, or serum. As analysis of the release of lactate dehydrogenase precluded any cell toxicity, S1P was supposed to induce apoptosis in that case. DNA fragmentation was evidenced in S1P-treated low-density cells by Hoechst staining and flow cytometry. Moreover, SAPK phosphorylation suggested that S1P elicited signaling pathways leading to a cell death program in low-density cells.
Our results have suggested a relationship between the apoptosis of mesangial cells cultured at low cell density and the formation of sphingosine. In a number of cell types, sphingosine induces either apoptosis or proliferation, specifically depending on the activity of sphingosine kinase (17). For instance, sphingosine and its methylated derivative N,Ndimethylsphingosine induce apoptosis of cancer cells and transformed mesangial cells (28). Conversely, activation of sphingosine kinase, allowing conversion of sphingosine to S1P, prevents apoptosis (22,23). In our study, we observed that sphingosine, like S1P, induced apoptosis of low-density mesangial cells in a dose-dependent manner. Therefore, we tested whether or not S1P hydrolysis was increased in cells cultured at low density. Enhanced S1P hydrolysis was demonstrated by incubation with [   concordant results. It is interesting to note that [ 3 H]sphingosine was more actively taken up by low-density cells (Fig. 7,  D and E). Moreover, a low conversion rate of [ 3 H]sphingosine to [ 3 H]S1P was observed, which was in favor of decreased sphingosine kinase activity in low-density cells. The down-regulation of sphingosine kinase at low cell density is in agreement with the increase in thymidine incorporation observed with sphingosine only in high-density cells.
In our opinion, the increased S1P hydrolysis as documented by [ 33 P]S1P incubation in Fig. 6 could account for the accumulation of sphingosine observed in low-density cells. We suggest that S1P hydrolysis and sphingosine uptake are crucial steps leading to the accumulation of [ 3 H]sphingosine in low-density cells. Moreover, this could be enhanced by the decrease in sphingosine kinase activity. Hence, sphingosine accumulation seems pertinent to explain the S1P-induced apoptosis of cells at low density. In this regard, it is worth noting that the dose responses of sphingosine and S1P were at a similar level because either 5 M S1P or 1 M sphingosine was sufficient to induce apoptosis.
A similar concentration has been used in another study of sphingosine-induced apoptosis (28). In that study, it was suggested that apoptosis of serum-starved cells may result from the mixed effect of treatment and starvation. Specially, the intracellular level of sphingosine may be modified by the downregulation of sphingosine kinase activity due to the starvation of growth factors like PDGF (22). However, it was demonstrated in the study of Sweeney et al. (28) that sphingosineinduced apoptosis may occur in various cell lines even in the presence of growth factors. This was the case in our study, where S1P-induced apoptosis persisted in growing cells and was reverted neither by 10 M LPA in quiescent cells nor by PDGF in growing cells (Fig. 4). Starvation of growth factors, which can be responsible for a modification of the "apoptosis threshold" (for instance, by modifying sphingosine kinase activity), was not necessary for apoptosis at low density. Therefore, one must suppose an additional effect of low density, such as increased S1P hydrolysis, to explain the S1P-induced apoptosis.
The role of intracellular sphingolipids in cell death is dependent upon two major interrelated components: ceramide and sphingosine. Both compounds may play an active part in apoptosis. Although the role of ceramide has been largely investigated in the past, more recent studies have pointed to the accumulation of sphingosine as being directly responsible for cell death (26,(45)(46)(47). Sphingosine may be directly involved in mitochondrion-dependent apoptosis, as recently suggested (26). In this regard, the relevance of sphingosine as an inductor of apoptosis is also related to a modification of the balance between sphingosine and S1P as discussed above. Alternatively, increased sphingosine levels may also enhance the synthesis of ceramide by ceramide synthase. Ceramide possesses apoptotic properties that are also intimately related to mitochondrial function, as recently reviewed (27).
In this study, we observed generation of ceramide, but no increased conversion of [ 3 H]sphingosine to [ 3 H]ceramide, during S1P-induced apoptosis. This suggests that de novo synthesis of ceramide from sphingosine, under the control of sphingosine N-acetyltransferase (ceramide synthase), is not significantly involved in this process. On the other hand, apoptosis was not inhibited in this study by incubation with fumonisin B, an inhibitor of ceramide synthase. On the contrary, fumonisin B inhibited etoposide-induced apoptosis, but slightly increased S1P-induced apoptosis. Logically, the blockade of ceramide synthase by fumonisin B would result in decreased ceramide levels and increased intracellular sphingosine and sphinganine levels. As discussed above, increased levels of sphingosine and sphinganine may be responsible for cell death. Hence, apoptosis seems more related to the accumulation of sphingosine than of ceramide in our model. Similarly, in the study of Sweeney et al. (28), N,N-dimethylsphingosine and sphingosine induced apoptosis in the absence of ceramide generation. Moreover, in a recent study, the generation of ceramide protected mesangial cells against sphingosine-induced cell death (29).
Another pathway to ceramide production must be hypothesized to explain the progressive accumulation of ceramide in our study. Activation of acid sphingomyelinase is a frequent event during apoptosis following the recruitment of the adaptor proteins TRADD and FADD and activation of the early caspases (27,48). In this regard, SAPK phosphorylation as observed here is frequently associated with activation of acid sphingomyelinase (49). Therefore, the generation of ceramide by sphingomyelinase could explain the ceramide increase observed here, which could, in turn, be involved in later phases of the apoptosis process.
Thus, these data do not allow unambiguous interpretation of the respective roles of sphingosine and ceramide during S1Pinduced apoptosis of low-density mesangial cells. Nevertheless, FIG. 6. Effect of cell density on S1P hydrolysis. A, time course of S1P hydrolysis at low (2 ϫ 10 4 cells/cm 2 ) or high (1 ϫ 10 5 cells/cm 2 ) cell density. Activity was measured as the amount of 33 P i radioactivity released in the medium from mesangial cells treated with 10 M [ 33 P]S1P (10 nmol, 220,000 cpm). B, effect of LPA on S1P hydrolysis. S1P hydrolysis was measured as described for A at low (2 ϫ 10 4 cells/cm 2 ) or high (1 ϫ 10 5 cells/cm 2 ) cell density in the presence of variable concentrations of LPA or vehicle (0.01% (w/v) BSA in PBS) in the assay. *, p Ͻ 0.05. Results are means Ϯ S.E. from triplicate determinations, representative of three (A) or two (B) independent experiments. they do not suggest a major role for ceramide generation in the induction of apoptosis. Ceramides are not generated from the accumulated sphingosine. As sphingosine may alter mitochondrial function, sphingosine alone can start apoptosis. However, later production of ceramide by sphingomyelinase may be involved as an additional factor.
Alternatively, other mechanisms not related to the altered metabolism of sphingolipids have been proposed in a few cases where S1P induces cell growth arrest. Apoptosis or proliferation arrest has been observed in cells derived from primary cultures such as hippocampal neurons and hepatic myofibroblasts (37,38). It has also been described in several cell lines such as hepatoma cells, ovarian cancer cells, and HEK293 cells transfected with EDG receptors (50 -52). In hepatic myofibroblasts, the phenotype of which is similar to that of mesangial cells, S1P-induced growth arrest is related to cyclooxygenase-2 induction and prostaglandin E 2 production (38). However, in our work with mesangial cells, we ruled out such a mechanism, as incubation with ibuprofen (a cyclooxygenase-2 inhibitor) was of no benefit during S1P-induced apoptosis (data not shown). In other cases, S1Pinduced apoptosis may be related to receptor-dependent events. Overexpression of EDG-3 or EDG-5 receptors is responsible for apoptosis induced by 0.1 M S1P (52). S1P-induced apoptosis in hippocampal neurons is associated with activation of calcium-de-pendent phosphatases (37), a mechanism possibly related to activation of EDG-3/EDG-5 G q -coupled receptors. EDG receptors may also be involved in the expression of Bax protein observed during S1P-induced apoptosis of hepatoma cells (50). The hypothesis that EDG receptors are involved in the process has not been precisely tested here. However, we performed semiquantitative evaluations of the expression of EDG receptors and did not detect a clear difference between mesangial cells cultured at low and high density (data not shown). In addition, the similar effect observed with S1P and sphingosine did not favor a mechanism dependent on S1P receptors in our model.
Finally, S1P-induced apoptosis of mesangial cells may be closely compared with the modulated growth and adhesion of ovarian cancer cells (51). In that study, S1P induced cell death of cells in suspension, but stimulated growth of attached cells. This suggests that the presence of homologous cell-cell contacts may drastically modify the responsiveness to S1P. This allows a comparison with our model; however, the effect of sphingosine was not tested in the study of Hong et al. (51). Similarly, it has been demonstrated that up-regulation of the cadherincatenin complex decreases sphingosine-induced cell death (46), and a relationship between sphingosine and ␤-catenin-regulated pathways is supported by another recent study (47).
Hydrolysis of LPLs is dependent on lipid phosphate phospho- hydrolases (53). Three isoforms of the type 2 lipid phosphate phosphohydrolase (lipid phosphate phosphohydrolase-1, -2, and -3) have been cloned from mammalian cells, with these enzymes having broad substrate specificity with similar efficiencies against LPA, S1P, phosphatidic acid, ceramide 1-phosphate, and diacylglycerol pyrophosphate. However, genes encoding for phosphatases with a remarkable specificity for phosphorylated sphingoid bases have been identified in yeast; and recently, murine S1P phosphohydrolase-1, a homolog of the yeast phosphohydrolases, was cloned (54,55). Murine S1P phosphohydrolase-1 differs from the other mammalian lipid phosphohydrolases by sequence particularities and by its high specificity for S1P. In NIH 3T3 cells, overexpression of murine S1P phosphohydrolase-1 leads to a 4-fold increase in its activity, which is responsible for S1P-induced apoptosis. A 2-fold increase in S1P hydrolysis was observed at low cell density, which could explain the apoptotic effect of S1P. The nature of the enzyme involved has not been documented in this study. However, the lipid phosphohydrolase activity that increased at low density was probably not specific for S1P because LPA partly inhibited S1P hydrolysis in the in vitro assay (Fig. 6B). Even so, this would not preclude the involvement of S1P phosphohydrolase-1 in that case. We detected both lipid phosphate phosphohydrolase-1 and the rat homologous sequence of S1P phosphohydrolase-1 using semiquantitative reverse transcription-PCR. We observed no significant difference between cells cultured at low and high cell density (data not shown), but increased lipid phosphate phosphohydrolase activity may be due to post-translation regulation (53). Note that even if LPA phosphohydrolase activity is increased, it is probably not relevant in explaining the apoptosis occurring at low cell density.
LPA did not increase S1P-induced apoptosis (Fig. 4). Moreover, it is worth noting that S1P-induced apoptosis occurred in a serum-free medium in the absence of any other factor such as LPA contained in FCS.
Finally, the physiological relevance of these results may be considered. First, the notion of cell density is not easily transferable in vivo. However, a number of cell functions seem to be regulated by the density of cell cultures. For instance, the upor down-regulation of the tyrosine phosphorylation status of focal adhesion kinase or paxillin has been observed as a function of cell density (56). It is worth noting that the cadherincatenin complex, which is highly regulated by the level of homologous cell-cell contacts, modulates the apoptosis induced by sphingosine (46). This suggests that the effect of cell density observed here may be valid for other cells. The response of human osteoblasts to sphingolipids is also density-dependent, as determined by preliminary data. 2 Another point is related to the relevance of the concentrations able to induce apoptosis. The actual concentration of S1P in renal glomeruli is unknown. However, the S1P concentration can rise to micromolar levels in plasma (57). Platelet activation might account for significantly increased local S1P levels in several renal diseases. In addition, the hypothesis of paracrine production of S1P in vascular systems is strongly supported by the recent observation of the extracellular export of sphingosine kinase (58). Altogether, these data suggest the possibility of increased S1P concentrations in the vicinity of mesangial cells. Resident mesangial cells are normally maintained in a quiescent phenotype FIG. 8. Production of endogenous ceramide in mesangial cells depending on cell density. Shown is the time course of the production of endogenous ceramide in mesangial cells after treatment with 10 M S1P or vehicle (0.01% (w/v) BSA in PBS). Ceramide production was assessed in mesangial cells cultured at low (2 ϫ 10 4 cells/cm 2 ) or high (1 ϫ 10 5 cells/cm 2 ) cell density in the presence of 5% (v/v) FCS. Quantification of ceramide was performed with the diacylglycerol kinase method as described under "Experimental Procedures." C 8 -ceramide was used as a standard for quantification, and C 8 -ceramide phosphate was used as a standard for migration. Results are means Ϯ S.E. from triplicate determinations, representative of two independent experiments. and probably develop poor cell-cell contacts. In pathology, enhanced production of S1P may occur. Then S1P hydrolysis might act as one of the glomerular "self-defense mechanisms" (59), allowing apoptosis to be a component of the glomerular remodeling after injury.
In conclusion, it was demonstrated that S1P stimulated the proliferation of subconfluent mesangial cells, whereas at low density, increased S1P hydrolysis was responsible for phosphorylation of SAPKs and apoptosis. Apoptosis seemed to be mainly determined by the accumulation of sphingosine, together with decreased conversion to S1P. Ceramide that did not originate from the accumulated sphingosine was also generated. Thus, it was suggested that the sphingolipid rheostat was readjusted at low cell density, with up-regulation of S1P hydrolysis and down-regulation of sphingosine kinase. As S1P may be produced actively in glomerular diseases, it may be of interest in the future to assess the level of S1P hydrolysis in circumstances involving excessive proliferation of mesangial cells.