Regulation of Autophagy by Sphingosine Kinase 1 and Its Role in Cell Survival during Nutrient Starvation*

The sphingolipid ceramide induces macroautophagy (here called autophagy) and cell death with autophagic features in cancer cells. Here we show that overexpression of sphingosine kinase 1 (SK1), an enzyme responsible for the production of sphingosine 1-phosphate (S1P), in MCF-7 cells stimulates autophagy by increasing the formation of LC3-positive autophagosomes and the rate of proteolysis sensitive to the autophagy inhibitor 3-methyladenine. Autophagy was blocked in the presence of dimethylsphingosine, an inhibitor of SK activity, and in cells expressing a catalytically inactive form of SK1. In SK1wt-overexpressing cells, however, autophagy was not sensitive to fumonisin B1, an inhibitor of ceramide synthase. In contrast to ceramide-induced autophagy, SK1(S1P)-induced autophagy is characterized by (i) the inhibition of mammalian target of rapamycin signaling independently of the Akt/protein kinase B signaling arm and (ii) the lack of robust accumulation of the autophagy protein Beclin 1. In addition, nutrient starvation induced both the stimulation of autophagy and SK activity. Knocking down the expression of the autophagy protein Atg7 or that of SK1 by siRNA abolished starvation-induced autophagy and increased cell death with apoptotic hallmarks. In conclusion, these results show that SK1(S1P)-induced autophagy protects cells from death with apoptotic features during nutrient starvation.

Macroautophagy (hereafter referred to as autophagy) is a lysosomal catabolic pathway for macromolecules and organelles (1)(2)(3). The discovery of ATG 2 in yeast has revealed the evolutionarily conserved mechanism for the formation of autophagosomes that sequester cytoplasmic material before they fuse with the endo/lysosomal compartment (4,5). Stimulation of autophagy during the limitation of nutrient supply is a cell-defense mechanism by which amino acids and other intracellular nutrients are recycled to maintain survival (6,7) and to repress the induction of apoptosis (8). The role of starvation-induced autophagy is vital for newborn mice after the maternal nutrient supply is interrupted (9).
Several years ago, cell death was divided into three types (10 -13). Type I is apoptosis, type II is autophagic cell death, and type III is cytoplasmic cell death. Type II is characterized by the accumulation of autophagic vacuoles. Recent studies have shown that autophagic machinery plays a role in cell death occurring in the absence of pro-apoptotic members of the Bcl-2 family (14) and when caspases are inhibited (15). However, in many situations apoptosis and autophagy may both contribute to cell death making difficult to clearly distinguish type I from type II cell death (16 -19).
One characteristic of autophagy is its dependence on several tumor suppressor genes acting both in the formation of the autophagosome and in signaling pathways controlling autophagy (20,21). Autophagy is thought to be a tumor suppressor mechanism, because cancer cells often display reduced autophagic capacities (20,21). Beclin 1 is a tumor suppressor gene product that plays an important role, together with the class III phosphatidylinositol 3-kinase (PI3K), during the formation of the autophagosome (22,23). Beclin 1 haploinsufficiency induces a high level of tumor development in various mouse tissues (24,25). The tumor suppressor PTEN (phosphatase and tensin homolog deleted on chromosome 10), which deactivates the class I PI3K signaling pathway, has a stimulating effect on autophagy (26). Its deletion favors the inhibition of autophagy by the class I PI3K signaling pathway by activating mTOR (target of rapamycin), a conserved nutritional sensor that represses autophagy (6,27,28).
However, the role of autophagy during tumor progression is probably more complex, because autophagic vacuoles have been observed in several types of human tumors (29). Whether these autophagic vacuoles are associated with cell survival or cell death remains to be clearly demonstrated.
Numerous drugs have been reported to trigger cell death with autophagic features in cancer cells (20,21,30). However, the events that lead to the accumulation of autophagic vacuoles by cancer treatments have only begun to be identified (30). The antiestrogen drug tamoxifen triggers autophagic cell death in human breast cancer MCF-7 cells (31). We have recently shown that the sphingolipid ceramide is a downstream target of tamoxifen during autophagy in MCF-7 cells (32). Ceramide stimulates autophagy by relieving the class I PI3K/Akt/PKB pathway and provoking the accumulation of the autophagy gene product Atg6/ Beclin 1. In line with these findings, a cell-permeable analog of ceramide has recently been shown to induce autophagic cell death in glioma cells * This work was supported by institutional funding from the Institut National de la Santé et de la Recherche Mé dicale (INSERM) and from University Paris-Sud, by grants from the Association pour la Recherche sur le Cancer (n°3503 to PC) and a bilateral French-Italian exchange program (Galileo). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1  by increasing the expression of the mitochondrial bound BH3-only BNIP3 protein (33). Sphingolipids are ubiquitous constituents of eukaryotic membranes. Their metabolism is a highly dynamic process generating second messengers, including ceramide, sphingosine, and sphingosine 1-phosphate (S1P) (34 -38). Ceramide is deacylated by ceramidases to generate sphingosine, which is then phosphorylated by sphingosine kinases to produce S1P (36). It is now generally admitted that ceramide, sphingosine, and S1P have contrasting effects on the response to stress (34 -38). Ceramide and sphingosine are often associated with cell-growth arrest and the induction of cell death, whereas S1P favors cell proliferation and cell survival. Thus, the dynamic balance between ceramide, sphingosine, and S1P is important in determining whether cells survive or die (36,39). The existence of this sphingolipid rheostat stress regulator led us to consider the role sphingosine kinase 1 (SK1) that produces S1P, in the control of autophagy.
In the present work we show that SK1 modulates two major players of autophagy: the mTOR signaling pathway and the expression of Beclin 1. Stimulation of autophagy by SK1 is dependent upon the production of S1P. In contrast to ceramide-induced autophagy, SK1(S1P)-induced autophagy did not induce cell death but has a protective effect toward apoptosis during nutrient starvation. These findings, and others in the literature (32,33), show that both the tumor-suppressor ceramide and the tumor-promoter S1P (38) are able to trigger autophagy but with different outcomes on cancer cell survival and death.

EXPERIMENTAL PROCEDURES
Reagents C 2 -Cer and DMS were from Calbiochem (VWR International). Fumonisin B1 (FB1) was purchased from Sigma. 3MA was from Fluka. z-VAD-fmk was from Apotech. Cell culture medium and fetal bovine serum were from Invitrogen. Nitrocellulose membranes were from Schleicher & Schüll. Ceramide from porcine brain, used as an internal standard, was from Avanti Polar Lipids Inc. The radioisotopes L-[U-14 C]valine (5.47 GBq/mmol) and [␥ 32 -P]ATP (110 TBq/mmol) and [9,10-3 H]myristic acid (1.69 TBq/mmol), the ECL TM Western blotting detection kit, and the donkey anti-rabbit antibody were purchased from BD Biosciences. The mouse monoclonal anti-Beclin1 was from Amersham Bioscience; the mouse monoclonal anti-PARP and the rabbit polyclonal anti-phospho-PKB Ser 473 were from Cell Signaling. Goat anti-PKB was from Santa Cruz Biotechnology. The mouse monoclonal anti-actin was from Chemicon International Inc. The rabbit polyclonal anti-p70S6K, anti-phospho-p70S6K Thr 389 , anti-phospho-S6 Ser 235 , and anti-4E-BP1 were from Ozyme, and anti-FLAG M2 was from Sigma. The rabbit polyclonal anti-Atg7 was kindly provided by W. A. Dunn (University of Florida, Gainesville). Goat anti-mouse and swine anti-goat antibodies were obtained from Bio-Rad and Caltag, respectively.

Cell Culture and Transfection
Human breast cancer MCF-7 cells were maintained at 37°C in 10% CO 2 in DMEM, supplemented with 10% fetal bovine serum and 100 ng/ml each of penicillin and streptomycin as previously reported (32). Beclin 1 MCF-7/tet-off (MCF-7.beclin 1) cells kindly provided by B. Levine (University of Texas Southwest Medical Center, Dallas) were used during starvation-induced autophagy. Briefly, cells were maintained in DMEM with 200 g/ml hygromycin B and 2 g/ml tetracycline (41). Expression of Beclin 1 was analyzed by Western blotting 5 days after tetracycline withdrawal. Starvation was induced by incubating cells in nutrient-free medium (Hanks' balanced salt solution plus 0.1% bovine serum albumin). Cells were transfected by using FuGENE 6 transfection reagent (Roche Applied Science), as recommended by the manufacturer. Briefly, cells (0.5 ϫ 10 6 cells/well) were plated in 6-well plates. Two days later, 1 g of plasmid and 3 l of FuGENE 6 were suspended in 100 l of DMEM and added to the culture medium. When cells were co-transfected with SK1, PLD1 (or empty vector), and GFP-LC3 vectors, the transfection efficiency was 40% as determined by counting GFP-positive cells. Cells were analyzed 24 or 48 h after cotransfection as detailed below. siRNA SK1 and Atg7 knockdowns were accomplished by transfecting MCF-7/GFP-LC3 cells with siRNAs. The RNAi target sequences were: sense, 5Ј-GGGCAAGGCCUUGCAGCUC-3Ј for SK1 (42) and sense, 5Ј-CCAACACACUCGAGUCUUU-3Ј for Atg7. As a control, siRNA targeting the unrelated protein phosphomannomutase-2 (sense, 5Ј-CUGGGAAAUGAUGUGGUUG) was used. siRNA were purchased from Eurogenetec (Seraing, Belgium). MCF-7 cells were seeded at 8 ϫ 10 4 /cm 2 . After 24 h, cells were transfected for 4 h with siRNA using Oligofectamine (Invitrogen) as recommended by the manufacturer. Cells were cultured for 24 h before analysis.

Autophagic Parameters
GFP-LC3 Staining-GFP-LC3 staining was carried out essentially as described previously (43). At the indicated time after transfection GFP-LC3 staining was visualized using an Axioplan Zeiss microscope, and the number of GFP-LC3-positive cells with GFP-LC3 dots was determined.
Proteolysis-Proteolysis was determined as described previously (44). Briefly, cells were incubated for 24 h at 37°C in normal culture medium containing 7400 Bq/ml of [ 14 C]valine. Cells were rinsed three times with PBS, pH 7.4, and then the cells were incubated for 1 h in complete medium supplemented with 10 mM valine. After incubating for 1 h, during which time short-lived proteins had been degraded, the medium was replaced by fresh chase medium for another 4 h. When required, C 2 -Cer (75 M), FB1 (100 M), DMS (1.5 M), and 3MA (10 mM) were added to the chase medium. For the PLD1-or SK1-overexpressing cells, radiolabeling with [ 14 C]valine was started 24 h after transfection. Cells and radiolabeled proteins from the 4-h chase medium were precipitated in trichloroacetic acid at a final concentration of 10% (v/v) at 4°C. Radioactivity was determined by liquid scintillation counting. Protein degradation was calculated by dividing the acid-soluble radioactivity recovered from both cells and medium by the radioactivity contained in the precipitated proteins from both cells and medium.

Cell Death Assay
Cell viability was determined by the trypan blue exclusion test as previously described (32). When required, 40 M z-VAD-fmk was added prior to incubation in nutrient-free medium (Hanks' balanced salt solution plus 0.1% of bovine serum albumin).

Quantification of Ceramide
Ceramide levels were measured using the DGK assay as described previously (45). Briefly, cells were collected, and lipids were extracted according to a previous study (46). The organic phase was divided into 1/2 and 1/6 aliquots, dried, and used for ceramide and total phospholipid measurements (47), respectively. Briefly, 30 nmol of extracted lipids was incubated at room temperature for 45 min in the presence of ␤-octylglucoside/dioleoylphosphatidylglycerol micelles, 2 mM dithiothreitol, 5 g of proteins of DGK containing membranes (Calbiochem), and 1 mM ATP mixed with [␥-32 P]ATP (4.8 ϫ 10 5 Bq/ml) in a final volume of 0.1 ml. At the end of the reaction, lipids were extracted, and [ 32 P]ceramide 1-phosphate was determined by TLC separation in chloroform/acetone/methanol/acetic acid/water (10:4:3:3:1, v/v). The radioactivity associated with ceramide 1-phosphate spots was determined after scraping and counting in a scintillation counter. Ceramide levels were expressed in terms of the total phospholipid content.
For quantification of S1P, MCF-7 cells were labeled for 7 h in serumfree medium with 0.3 Ci/ml D-erythro-[3-3 H]sphingosine (PerkinElmer Life Sciences). After washing with PBS, cells were scraped and lysed in water. Radiolabeled S1P was extracted by phase partition as previously described (49,50) and counted by liquid scintillation.
Phospholipase D Activity-PLD activity was determined as described previously (51) with some modifications. Briefly, 2 ϫ 10 6 cells were plated on 25-cm 2 flasks for 24 h; when required, cells were transfected with PLD1 vectors (wild-type and K898R mutant). After 24 h, cells were preincubated in 0.5% fetal bovine serum containing DMEM for 1 h. After removing the preincubation medium, cells were labeled with [ 3 H]myristic acid (Amersham Biosciences) for 1 h at 1.11 ϫ 10 5 Bq/ml in 0.5% fetal bovine serum containing DMEM. Cells were rinsed with PBS and incubated for 30 min in complete medium containing 0.5% butanol. Cells were rinsed with PBS and scraped in methanol/water (98:2, v/v). Cells were rinsed with methanol, and the lipids were extracted by adding equal amounts of chloroform and water. After centrifuging, the lower phase was collected and dried under vacuum before being redissolved in chloroform. Lipids were separated by TLC on silica-coated plastic sheets (Merck) in a solvent system of chloroform:

Statistical Analysis
Statistical analysis of the differences between the groups was performed using Student's t test. p Ͻ 0.05 was considered statistically significant.

Sphingosine Kinase 1 Overexpression Stimulates Autophagy-MCF-7
cells were transfected either with the cDNA encoding the wild-type SK1 (SK1 wt ) or that encoding a mutant inactive form (SK1 G82D ). The activity of SK was significantly increased in SK1 wt -overexpressing cells when compared with that observed in control cells transfected with an empty vector or in SK1 G82D -expressing cells (Fig. 1A). Accordingly, SK1 wtoverexpressing cells displayed a higher level of S1P compared with control cells (Fig. 1A). In the presence of the SK inhibitor DMS, both S1P production and SK activity were dramatically reduced in these cells. In contrast, no change in the level of ceramide was detected in SK1 wtoverexpressing cells at any time points after transfection (Fig. 1, A  and B).
Autophagy was first analyzed after transfection of GFP-LC3 ( Fig. 2A). LC3 is a reliable marker of autophagosomes in mammalian cells, and its localization changes from a diffuse cytosolic pattern to a punctuate pattern representing its recruitment to the autophagosomal membrane during the induction of autophagy (43). An accumulation of autophagic vacuoles was observed in the SK1 wt -overexpressing cells but not in the SK1 G82D -expressing cells ( Fig. 2A). Accordingly, proteolysis sensitive to 3MA, another hallmark of autophagy (52), was increased in SK wt -overexpressing cells (Fig. 2B). In addition, the autophagic phenotype was abolished when SK1 wt -overexpressing cells were treated with 1.5 M DMS. However, DMS did not block C 2 -Cer-induced autophagy in MCF-7 (Fig. 2), suggesting that this drug does not inhibit autophagy in an unspecific manner. FB1, an inhibitor of CoA-dependent dihydroceramide/ceramide synthase, which inhibits C2-Cer-induced autophagy, by impairing its conversion to long-chain ceramides (see ref. 32), did not block autophagy in SK1 wt -overexpressing cells (Fig. 3). Overall, these findings led us to conclude that the production of S1P by SK1 has a stimulatory effect on autophagy, which is not merely a consequence of an increase in ceramide level.
Role of mTOR in Controlling SK1(S1P)-induced Autophagy-On the basis of genetics studies and the use of the immunosuppressor rapamy-cin, it has been demonstrated that the inhibition of TOR plays a central role in stimulating autophagy throughout the eukaryotic kingdom (6,27,28). To confirm that a reduction of mTOR signaling activity was involved in the control of SK1(S1P)-induced autophagy, we investigated the phosphorylation of two known downstream effectors of mTOR, namely p70S6K and the eukaryotic translation inhibitor, 4E-BP1 (Fig.  4A). The phosphorylation of p70S6K Thr 389 , the major rapamycin-sensitive site (53), was greatly reduced in SK1 wt -overexpressing cells. In line with the mTOR-dependent phosphorylation of 4E-BP1 (54), the hyperphosphorylated state of 4E-BP1 was dephosphorylated to a fast-migrating band in SK1 wt -overexpressing cells (Fig. 4A). In contrast, mTOR signaling was neither affected in DMS-treated SK1 wt -overexpressing cells nor in SK1 G82D -expressing cells (Fig. 4A). However, the phosphorylation of Akt/PKB that regulates autophagy upstream of mTOR (26) was not significantly reduced in SK1 wt -overexpressing cells treated or not by DMS (Fig. 4B). In contrast, C 2 -Cer induced both an inhibition of Akt/PKB and mTOR signaling (see Ref. 32 and Fig. 4). These data suggest that SK1 may stimulate autophagy independently of the inhibition of the class I PI3K/Akt/PKB signaling pathway.
Alternatively, we reasoned that forced stimulation of mTOR should affect SK1(S1P)-induced autophagy. Recently, phosphatidic acid produced by phospholipase D has been shown to activate mTOR in a rapamycin-dependent manner in nutrient-rich medium (55)(56)(57). First MCF-7 cells were transfected with cDNAs encoding PLD1 wt and a mutant inactive form PLD1 K898R . An increase in PLD activity was only observed in the PLD1 wt -overexpressing cells (Fig. 5A). According to previous results (55), mTOR signaling was increased in MCF-7 cells overexpressing PLD1 wt as revealed by the increase of p70S6K Thr 389 phosphorylation and that of the p70S6K substrate ribosomal protein S6 (Fig. 5B). This stimulatory effect was abolished in the presence of 100 nM rapamycin (Fig. 5B). Moreover, the expression of PLD1 wt reversed the inhibition of mTOR signaling pathway in SK1 wt -overexpressing cells (Fig. 5B). Interestingly, expression of the PLD1 wt was able to abrogate both the accumulation of GFP-LC3 dots and the stimulation of [ 14 C]valine-labeled protein degradation observed in SK1 wt -overexpressing cells (Fig. 5C, lower panel). In contrast, expression of PLD1 K898R did not inhibit autophagy in SK1 wt -overexpressing cells, as exemplified by its lack of inhibition on proteolysis (Fig. 5C, lower panel).
Effect of SK1 on the Accumulation of Beclin 1-Ceramide-induced autophagy is characterized by the robust accumulation of the autophagy protein, Beclin 1 (32). A slight but reproducible accumulation of Beclin 1 was observed in SK1 wt -overexpressing MCF-7 but not in cells expressing the mutant form of SK1 or in cells transfected with an empty vector (Fig. 6). However, SK1-induced accumulation of Beclin 1 was less pronounced than cells having an elevated level of ceramide after C 2 -Cer supplementation ( Fig. 6 and   in Beclin 1 accumulation and inhibition of mTOR signaling pathway could be responsible for the stimulatory effect S1P on autophagy. SK1(S1P)-induced Autophagy Protects Cells During Starvation-Next, we investigated the physiological consequence of SK1(S1P)-induced autophagy. According to previous results, enforced expression of SK1 wt has no deleterious effect on cell viability in MCF-7 cells (58). In addition and as previously reported (58), the chronic supplementation of MCF-7 cells with 100 nM S1P during 10 days neither affected cell growth nor viability, yet autophagic vacuoles were detected (data not shown).
To investigate whether SK1(S1P)-induced autophagy can contribute to cell survival during starvation, we used stably transfected MCF-7.be-clin1 sensitive to tetracycline, which express an increased expression of Beclin 1 after removing tetracycline from the medium (Fig. 7A and Ref. 41). According to Liang et al. (41), autophagy was stimulated in MCF-7.beclin1 cells in the absence of nutrients (Fig. 8A). Inhibition of autophagy by 10 mM 3MA treatment or silencing the expression of the autophagy protein Atg7, involved in the formation of autophagosomes (59,60), reduced cell viability after 48 h in nutrient-free medium (Fig. 8,  A and B). When MCF-7.beclin1 cells were incubated for 48 h in nutrient-free medium, a 2-fold increase in SK activity was observed (Fig. 7B) without any significant change in ceramide level (Fig. 7C). Inhibition of SK activity by DMS or SK1-specific siRNA severely impaired starvationinduced autophagy and cell viability of MCF-7.beclin1 cells (Fig. 8, A  and B) in the absence of any effect on ceramide level (Fig. 7C). Whatever the approach used to inhibit starvation-induced autophagy, cell death was characterized by the appearance of the cleaved form of the caspase substrate PARP (Fig. 8). Both cell death and PARP cleavage were abolished in the presence of 40 M of the pan caspase inhibitor z-VAD-fmk (Fig. 8B). These results suggest that apoptosis is involved in the loss of viability of MCF-7.beclin1 cells during starvation when autophagy is inhibited by manipulating the regulation by SK1 or the downstream formation of autophagosomes. In contrast, FB1 did not inhibit autophagy ( Fig. 9) or cell viability (data not shown). These findings suggest that ceramide produced by the activity of CoA-dependent dihydroceramide/ceramide synthase is not required to control autophagy under these conditions.

DISCUSSION
On the basis of overexpression and siRNA approaches, the present work shows the role of SK1 in the regulation of autophagy. Moreover, the production of S1P by SK1 is necessary to stimulate autophagy, because the pharmacological inhibition of SK activity or transfection of a catalytically inactive form of SK1 blunts autophagic capacities. However, a role for a metabolite derived from S1P cannot be totally excluded in the regulation of autophagy because of the rapid turnover of S1P (36,38). We had previously shown that an elevation of the level of ceramide triggers autophagy in MCF-7 cells and in human colon cancer cells (32). However, on the basis of measurement of ceramide level and inhibition of ceramide synthase by FB1, we conclude that SK1-induced autophagy is not merely due to an increase of ceramide level. Moreover, and in contrast to ceramide-induced autophagy, the activation of Akt/PKB was not impaired during S1P-induced autophagy. The absence of any effect of SK1 overexpression on Akt/PKB is consistent with previous data and with its role in cell survival and tumor growth (36). However, inhibition of mTOR signaling was observed during SK1(S1P)-induced autophagy. Interestingly, enforced expression of PLD1 that produces the mTOR activator phosphatidic acid (57) counteracts the SK1(S1P)-induced autophagy and relieves mTOR inhibition. These findings show that SK1 controls autophagy by inhibiting mTOR independently of the class I PI3K arm. Although, the inhibition of mTOR signaling by S1P seems odd at a first glance, in fact it is not incompatible with its stimulating effect on cell proliferation (36,61). mTOR is present in two complexes that have different functions. The raptor-mTOR complex is sensitive to rapamycin and controls cell growth, whereas the rictor-mTOR complex controls cell proliferation and activates Akt/PKB via a feedback loop (62). This complex has a low sensitivity to rapamycin inhibition. Further and SK1 G82D -expressing cells were analyzed by Western blotting for p70S6K phosphorylation at Thr 389 using a phosphospecific antibody. The expression of total p70S6K was detected by using an anti-pan p70S6K antibody. Densitometry was performed on the original fluorographs, and the ratio of phospho-Thr 389 /total p70S6K in controlcellswassetto1.Thephosphorylationof4E-BP1wasevaluatedbyphosphorylationdependent band shift using anti-4E-BP1. SK1 was detected in cell lysates of SK1 wt -overexpressing cells and SK1 G82D -expressing cells by anti-FLAG. Actin was used as loading control. When required, 1.5 M DMS was added. B, Akt/PKB activation. Cell lysates obtained as described above were analyzed by Western blotting for Akt/PKB phosphorylation at Ser 473 using a phosphospecific antibody. The expression of total Akt/PKB was detected by using an anti-pan Akt/PKB antibody. Densitometry was performed on the original fluorographs, and the ratio of phospho-Ser 473 /total Akt/PKB in control cells was set to 1. Values are the mean Ϯ S.D. of three independent experiments. *, p Ͻ 0.05 versus control cells. MARCH 31, 2006 • VOLUME 281 • NUMBER 13 S1P, which acts intracellularly, also binds to SP1 receptors at the cell surface (36). The role of intracellular and exogenous S1P on autophagy remains to be investigated. Preliminary data show that dihydrosphingosine, which activates cell-surface S1P receptors but does not control cell survival (63), did not trigger autophagy, suggesting that autophagy and cell survival are controlled by the second-messenger activity of S1P. However, the possibility that S1P may also have a role in controlling autophagy via its receptors cannot be ruled out, because extracellular S1P activates signaling pathways (36,61) such as ERK1/2, p38 mitogenactivated protein kinase, and Akt/PKB, which are known to modulate autophagy (26,64,65).

Control of Macroautophagy by Sphingosine Kinase 1
Starvation stimulates both autophagy and endogenous SK activity. Inhibition of SK activity by DMS or silencing the expression of SK1 by siRNA strongly impairs starvation-induced autophagy in MCF-7.beclin 1 cells. These findings show that SK1 has to be added to the list of nutrient-sensitive proteins that regulate autophagy (66). Interestingly, the stimulation of SK activity during nutrient deprivation has also been reported in Saccharomyces cerevisiae (67) suggesting a general role for SK during starvation in eukaryotic cells. In line with the role of autophagy in maintaining cell survival and repressing apoptosis during starvation (7)(8)(9)68), cell death with biochemical hallmarks of apoptosis is detectable when SK activity is inhibited or after invalidation of the expression of SK1. These findings suggest that autophagy is a novel facet of the role of SK1(S1P) in cell survival (36).
The cell death mediator ceramide is also able to stimulate autophagy in human cancer cells. However, this autophagy can contribute to the execution of a cell-death program as recently shown in glioma cells (33). The role of the mitochondrial-associated cell-death protein BNIP3 downstream of ceramide during autophagic cell death has recently been proposed in glioma cells. The involvement of other sensors of the integrity of organelles such as death-associated protein kinase (69) is worth considering, because death-associated protein kinase, which is regulated by ceramide (70 -73), also triggers autophagic cell death in cancer cells (74).
"Experimental Procedures." B, analysis of mTOR signaling. Cell lysates of control cells, SK1 wt -overexpressing cells co-transfected or not with the cDNA encoding PLD1 wt were analyzed by Western blotting for p70S6K phosphorylation at Thr 389 using a phosphospecific antibody. The expression of total p70S6K was detected by using an anti-pan p70S6K antibody. Densitometry was performed on the original fluorographs and the ratio phospho Thr 389 /total p70S6K in control cells was set to 1. Ribosomal S6 protein phosphorylation at Ser 235 was detected using a phosphospecific antibody. Actin was used as loading control. When required 100 nM rapamycin was added for 2 h. C, autophagic capacities: Upper panels, GFP-LC3 staining and quantitative analysis in control MCF-7 cells, SK1 wt -overexpressing cells, PLD1 wt -overexpressing cells, SK1 wt -overexpressing cells co-transfected with the cDNA encoding PLD1 wt . The bar represents 10 m. Lower panel, degradation of [ 14 C]valine-labeled long-lived proteins was determined in control cells and in SK1 wt -overexpressing cells, PLD1 wtoverexpressing cells, PLD1 K898R -expressing cells, SK1 wt -overexpressing cells co-transfected with cDNA encoding PLD1 wt or PLD1 K898R as described under "Experimental Procedures." Values reported are the mean Ϯ S.D. of four independent experiments. *, p Ͻ 0.05 versus control cells.  The role of ceramide in autophagy and cell death points to cross-talk occurring between the control of autophagy and apoptosis (14,15,75). Recently, it has been reported that the interaction between Bcl-2 and Beclin 1 is required to keep autophagy in a physiological range (76). Unbalancing this interaction in favor of Beclin 1 triggers unchecked autophagy and autophagy gene-dependent cell death. It is of note that ceramide-induced autophagy is characterized by an accumulation of Beclin 1, whereas only a slight change in the accumulation of Beclin 1 is observed during SK1(S1P)-induced autophagy. Whether the changes in the ratio between ceramide and S1P can alter the Beclin 1/Bcl-2 rheostat is an attractive possibility that deserves further investigations.
In conclusion, the control of autophagy by ceramide and S1P suggests a novel cell function for the sphingolipid rheostat (36,39). In cancer cells, the survival role of SK1(SP1)-induced autophagy and the role of ceramide in cell death with autophagic features are compatible with their functions as tumor-promoting and tumor-suppressing molecules, respectively (38). The balance between ceramide (sphingosine) and S1P is under the control of sphingosine kinase and S1P-phosphatase. Overexpression of S1P phosphohydrolase induces apoptosis by increasing the intracellular level of ceramide (77). From this, it can be hypothesized that S1P phosphohydrolase could also trigger other forms of cell death, including autophagic cell death. Alternatively, S1P lyase  that hydrolyzes S1P into hexadecenal and ethanolamine phosphate has been shown to enhance stress-induced ceramide and apoptosis (78). Deletion of the gene encoding S1P lyase has been shown to protect yeast against nutrient-deprivation induced death (79). The study of enzymes that regulate the balance between S1P and ceramide is a promising route toward a better understanding of the regulation of autophagy by sphingolipids and its consequence for cell survival and cell death.