Regulation of Autophagy and Its Associated Cell Death by “Sphingolipid Rheostat”

Background: The sphingolipids ceramide and sphingosine 1-phosphate (S1P) control various cellular functions, including proliferation, cell death, and autophagy. Results: Binding of S1P to its receptor S1P3 counteracts ceramide-mediated autophagy by activating the mammalian target of rapamycin (mTOR) pathway. Conclusion: Sphingolipid rheostat between ceramide and S1P plays an important role in regulating mTOR-controlled autophagy. Significance: We provide new insights into novel regulatory mechanisms in autophagy induction. The role of “sphingolipid rheostat” by ceramide and sphingosine 1-phosphate (S1P) in the regulation of autophagy remains unclear. In human leukemia HL-60 cells, amino acid deprivation (AA(−)) caused autophagy with an increase in acid sphingomyleinase (SMase) activity and ceramide, which serves as an autophagy inducing lipid. Knockdown of acid SMase significantly suppressed the autophagy induction. S1P treatment counteracted autophagy induction by AA(−) or C2-ceramide. AA(−) treatment promoted mammalian target of rapamycin (mTOR) dephosphorylation/inactivation, inducing autophagy. S1P treatment suppressed mTOR inactivation and autophagy induction by AA(−). S1P exerts biological actions via cell surface receptors, and S1P3 among five S1P receptors was predominantly expressed in HL-60 cells. We evaluated the involvement of S1P3 in suppressing autophagy induction. S1P treatment of CHO cells had no effects on mTOR inactivation and autophagy induction by AA(−) or C2-ceramide. Whereas S1P treatment of S1P3 overexpressing CHO cells resulted in activation of the mTOR pathway, preventing cells from undergoing autophagy induced by AA(−) or C2-ceramide. These results indicate that S1P-S1P3 plays a role in counteracting ceramide signals that mediate mTOR-controlled autophagy. In addition, we evaluated the involvement of ceramide-activated protein phosphatases (CAPPs) in ceramide-dependent inactivation of the mTOR pathway. Inhibition of CAPP by okadaic acid in AA(−)- or C2-ceramide-treated cells suppressed dephosphorylation/inactivation of mTOR, autophagy induction, and autophagy-associated cell death, indicating a novel role of ceramide-CAPPs in autophagy induction. Moreover, S1P3 engagement by S1P counteracted cell death. Taken together, these results indicated that sphingolipid rheostat in ceramide-CAPPs and S1P-S1P3 signaling modulates autophagy and its associated cell death through regulation of the mTOR pathway.

The role of "sphingolipid rheostat" by ceramide and sphingosine 1-phosphate (S1P) in the regulation of autophagy remains unclear. In human leukemia HL-60 cells, amino acid deprivation (AA(؊)) caused autophagy with an increase in acid sphingomyleinase (SMase) activity and ceramide, which serves as an autophagy inducing lipid. Knockdown of acid SMase significantly suppressed the autophagy induction. S1P treatment counteracted autophagy induction by AA(؊) or C 2 -ceramide. AA(؊) treatment promoted mammalian target of rapamycin (mTOR) dephosphorylation/inactivation, inducing autophagy. S1P treatment suppressed mTOR inactivation and autophagy induction by AA(؊). S1P exerts biological actions via cell surface receptors, and S1P 3 among five S1P receptors was predominantly expressed in HL-60 cells. We evaluated the involvement of S1P 3 in suppressing autophagy induction. S1P treatment of CHO cells had no effects on mTOR inactivation and autophagy induction by AA(؊) or C 2 -ceramide. Whereas S1P treatment of S1P 3 overexpressing CHO cells resulted in activation of the mTOR pathway, preventing cells from undergoing autophagy induced by AA(؊) or C 2 -ceramide. These results indicate that S1P-S1P 3 plays a role in counteracting ceramide signals that mediate mTOR-controlled autophagy. In addition, we evaluated the involvement of ceramide-activated protein phosphatases (CAPPs) in ceramide-dependent inactivation of the mTOR pathway. Inhibition of CAPP by okadaic acid in AA(؊)-or C 2 -ceramide-treated cells suppressed dephosphorylation/inactivation of mTOR, autophagy induction, and autophagy-associated cell death, indicating a novel role of ceramide-CAPPs in autophagy induction. Moreover, S1P 3 engagement by S1P counteracted cell death. Taken together, these results indicated that sphingolipid rheostat in ceramide-CAPPs and S1P-S1P 3 signaling modulates autophagy and its associated cell death through regulation of the mTOR pathway.
Ceramide initially emerged as a lipid mediator of apoptotic cell death, and many molecules including cathepsin D, ceramide-activated protein phosphatases (CAPPs), and PKC were shown to be involved in its signaling (9 -11). Ceramide was also reported to promote the formation of autophagic vacuoles by up-regulating beclin 1 (12) and to play a role in regulating autophagy and its associated cell death (13)(14)(15). However, the molecular mechanism by which ceramide modulates autophagy signaling has not been elucidated fully. S1P is formed by the sequential catalytic metabolism of ceramide. Ceramidase degrades ceramide to form sphingosine, which in turn is converted to S1P by sphingosine kinase (3,16). Extrinsic S1P is involved in highly diverse cellular functions including adherent junction assembly, cytoskeleton changes, cell migration and proliferation through specific G protein coupled-membrane receptors (S1P [1][2][3][4][5] ) (17)(18)(19)(20)(21)(22). Interestingly, S1P intrinsically targets specific intracellular proteins and exerts biological responses independently of S1P receptors (23). In terms of autophagy, overexpression of sphingosine kinase 1 was reported to induce autophagy independently of Akt/protein kinase B in MCF-7 cells (24). Furthermore, Lépine et al. (25) showed that deletion of sphingosine-1-phosphate phosphohydrase-1, which is a metabolic enzyme of S1P, induces autophagy without the involvement of the mammalian target of rapamycin (mTOR) and type III phosphoinositide 3 (PI3)-kinase-beclin-1 pathways. That study demonstrates that intrinsic, but not extrinsic, S1P serves as an inducing lipid. However, recent studies have shown that extrinsic S1P activates the mTOR pathway through S1P receptors (26 -28), and it was assumed that extrinsic S1P counteracts autophagy induction by activating its receptor-mTOR pathway. S1P and ceramide are biologically interconvertible lipids (8), and it has been proposed that their relative levels determine cell fate (i.e. life or death) (29,30). The relevance of this "sphingolipid rheostat" in regulating cell fate has been demonstrated in many different cell types (31). In the present study, we demonstrate that the sphingolipid rheostat also modulates autophagy.
Cell Culture and Autophagy Induction-Human leukemia HL-60 cells were cultured in RPMI 1640 medium containing 10% (v/v) fetal calf serum (FCS). Autophagy was induced by amino acid deprivation (AA(Ϫ)). Cells were washed three times with phosphate-buffered saline (PBS) and incubated with amino acid (L-glutamine, L-leucine, L-lysine, L-methionine)free RPMI1640 medium (Nissui, Tokyo, Japan) in the absence of FCS at 37°C in a humidified atmosphere containing 5% CO 2 . Chinese hamster ovary (CHO) and CHO/S1P 3 cells were cultured in ␣-minimal essential medium containing 10% (v/v) FCS (32). For autophagy induction, CHO cells were incubated in serum-free ␣-minimal essential medium for 3 h, and then the medium was replaced with AA(Ϫ) medium.
Electron Microscopy-After treatment, cells were harvested, washed twice with PBS, and fixed with 2.5% (w/v) glutaraldehyde in PBS. Fixed cells were treated with 1% (w/v) osmium tetraoxide in PBS and dehydrated in ethanol, followed by epoxy-1,2-propane treatment. Ultrathin sections obtained with a diamond knife were counterstained with uranyl acetate in ethanol solution, followed by lead citrate. Images were collected with a CCD camera.
Labeling of Autophagic Vacuoles with MDC-Cells were incubated with 50 M MDC (33) in PBS at 37°C for 10 min following autophagy induction. After incubation, the cells were washed three times with PBS, placed on glass slides by cytospinning and immediately analyzed by fluorescence microscopy using an inverted microscope equipped with a filter system. Images were obtained with a CCD camera and processed using the program Cool Snap (Roper Industries, Sarasota, FL). MDC incorporated into cells was expressed as the fluorescence intensity (arbitrary units) per one cell, and at least 200 cells were calculated by the program MacSCOPE (Mitani, Fukui, Japan).
Ceramide Measurements-After extracting lipids according to the Bligh and Dyer method (34,35), ceramide levels were determined by the diacylglycerol kinase assay using [␥-32 P]ATP and Escherichia coli diacylglycerol kinase, which converts ceramide and diacylglycerol to ceramide 1-phosphate and phosphatidic acid, respectively (35). Radioactivity of ceramide corresponding to ceramide 1-phosphate was detected and quantified with the BAS-2000 (Fujifilm, Tokyo, Japan). Amounts of ceramide were normalized with phospholipid phosphate.
Transfection with Small Interfering RNA (siRNA)-Cells were transfected with 40 nM double-strand siRNAs for scrambled sequence or acid SMase using MultiFectam (Promega) according to the manufacturer's instructions. After 72 h, cells were washed and treated with AA(ϩ) or AA(Ϫ) to induce autophagy. Table 1 shows sequences of acid SMase siRNA.
Q-RT-PCR was performed using an Applied Biosystems 7300 Real-time PCR System, according to the standard TaqMan PCR kit protocol. TaqMan probes of S1P receptors were purchased from Applied Biosystems. The GAPDH was used as an internal reference control to normalize relative levels of mRNA expression.
Confocal Microscopy-For endogenous LC3 detection, cells were cultured at various conditions, washed twice with cold PBS, and attached onto glass slides by cytospinng. Cells were fixed with 1% (w/v) paraformaldehyde for 15 min at 4°C and permeabilized with 0.1% (v/v) Triton X-100 for 5 min at room temperature. Cells were then blocked with PBS containing 2% (w/v) BSA and 0.1% (v/v) Tween 20 for 20 min and treated with anti-LC3 antibody for 90 min at room temperature. Cells were washed five times and treated with anti-mouse IgG Alexa Fluor 488. Images were obtained by confocal microscopy using the Leica TSC SP2 (Leica microsystems, Wetzlar, Germany).
Detection Autophagic Flux Using RFP-GFP-LC3-The CHO/ S1P 3 cells were transiently transfected with tfLC3 vector, and treated with various conditions after 24 h transfection. Cells were washed twice with cold PBS and fixed with 2% (w/v) para- formaldehyde for 20 min at 4°C. Images were gained with fluorescent microscopy (Nikon Instruments Inc., NY). Determination of Serine/Threonine Protein Phosphatase Activity-Cells were harvested and treated with the hypotonic buffer containing 10 mM HEPES-KOH (pH 7.4), 10 mM KCl, and 2 mM MgCl 2 for 30 min on ice. The swollen cells were homogenized using a 27-gauge needle, and the cell lysates were centrifuged at 10,000 ϫ g for 10 min at 4°C. The supernatants (0.3 g of protein) were used for serine/threonine protein phosphatase assay. The proteins were incubated with or without 500 nM okadaic acid prior to performing the measurement of serine/threonine protein phosphatase activity using a ProFluorSer/Thr PPase assay kit (Promega). The protein phosphatase activity insensitive to okadaic acid was subtracted from a okadaic acid-sensitive one, giving specific PP1/PP2A activity.
DAPI Staining-Cells were harvested and then fixed in 1% (w/v) paraformaldehyde for 30 min at room temperature, followed by DAPI staining as described previously (37). At least 200 cells were counted under a fluorescent microscope, and cells with nuclear condensation and fragmentation were designated as dead cells.
Statistical Analysis-Comparisons between two groups were carried out using the unpaired Student's t test.

AA(Ϫ) Induces Autophagy Concomitant with Ceramide
Accumulation in Human Leukemia HL-60 Cells-AA(Ϫ) promotes autophagy in many cell types such as liver, breast, and colon cancer cells (38 -40). Autophagy begins with the formation of double-membrane autophagosomes that engraft cyto-plasmic proteins and suborganelles. The autophagosomes are fused to lysosomes for bulk degradation. Herein, we showed that human leukemia HL-60 cells also underwent autophagy in response to AA(Ϫ) (Fig. 1. Electron microscopy revealed that AA(Ϫ) medium-cultured HL-60 cells were generated a myriad of vacuoles like autophagosomes, whereas no significant detection of autophagosomes was observed in regular AA(ϩ) medium-cultured cells (Fig. 1A). Autophagosome formation and progressive fusion to lysosomes are mediated by converting LC3 from a cytosolic form (LC3-I) to a phosphatidylethanolamine-conjugated membrane-bound form (LC3-II) (41). The cellular localization of the autophagy-related protein LC3 changes from a diffuse cytosolic pattern to a punctate pattern during autophagy (41). The formation of LC3 puncta (immunofluorescence) and LC3-II (immunoblotting) was promoted in response to AA(Ϫ) (Fig. 1B). Autophagy is also characterized by the incorporation of MDC as previously described (33), and MDC incorporation was significantly increased after AA(Ϫ) treatment (Fig. 1C). Similar to AA(Ϫ), autophagy was also induced in response to the cell permeable C 2 -ceramide (Figs. 1, A-C), which is a known autophagy-inducing lipid (12,15).
Ceramide was assumed to mediate autophagy induction following AA(Ϫ) treatment. To assess the effects of AA(Ϫ) on ceramide accumulation, ceramide levels were determined using the diacylglycerol kinase assay. Accompanied with an increase in MDC incorporation (Fig. 1D), ceramide was significantly accumulated in a time-dependent manner after exposure of cells to AA(Ϫ) (Fig. 1E). This increase in ceramide declined to baseline levels with AA(ϩ), when AA(Ϫ) conditions were restored by the addition of amino acids (data not shown). These results indicate that AA(Ϫ) facilitates the formation of ceramide, which stimulates autophagy induction in HL-60 cells.
Acid SMase Is Involved in Autophagy Induction-Ceramide is synthesized by three enzymatic pathways: (i) de novo pathway, (ii) the sphingomyelin cycle consisting of SMase and SMS, and (iii) the salvage pathway. In previous studies, ceramide synthases, which are involved in de novo and salvage ceramide synthesis, have been implicated in the formation of autophagyinducing ceramides (12,13,42). To elucidate the involvement of de novo and salvage pathways in autophagy, the effects of ceramide synthase inhibitor fumonisin B1 or ceramide de novo synthesis inhibitor myriocin on autophagy induction by AA(Ϫ) were tested in HL-60 cells. Neither treatment of HL-60 cells with fumonisin B1 nor myriocin affected the autophagy ( Fig.  2A). Next to investigate the involvement of the sphingomyelin cycle, we determined enzyme activities such as SMase or SMS after AA(Ϫ) treatment. As shown Fig. 2, B and C, AA(Ϫ) treatment significantly activated acid SMase but not neutral SMase after 30 min and 3 h. On the other hand, SMS activity did not change in the AA(Ϫ) condition (Fig. 2D). To investigate whether acid SMase-dependent formation of ceramide contributes to AA(Ϫ)-induced autophagy, the effects of acid SMase knock-down were tested (Fig. 2E). As compared with scrambled sequence-treated cells, acid SMase knock-down significantly suppressed autophagy induction (Fig. 2F). These data suggest that acid SMase plays a role in AA(Ϫ)-induced ceramide production and autophagy. S1P Treatment Counteracts Autophagy Induction by AA(Ϫ) or Ceramide-S1P is a potent proliferative lipid, and its effects are believed to result from its engagement with G-protein-coupled S1P receptors that generate mitogenic signaling. S1P signals protect cells from ceramide-dependent cell death signals (29), and that sphingolipid rheostat has been proposed to determine reciprocal cell fate. As ceramide was implicated in inducing autophagy and its associated cell death (13), we at first elucidated the effects of S1P on ceramide-dependent induction of autophagy. Treatment of HL-60 cells with S1P inhibited the formation of LC3 puncta and the generation of LC3-II induced by AA(Ϫ) (Fig. 3, A and B). Similarly, S1P treatment signifi- cantly attenuated AA(Ϫ)-induced incorporation of MDC in autophagosomes from 100 Ϯ 6.5 to 25 Ϯ 2.7 arbitrary units (Fig.  3C). On the other hand, treatment with S1P alone failed to form LC3 puncta in HL-60 cells (Fig. 3A). Next, we assessed whether S1P treatment affected AA(Ϫ)-induced formation of ceramide. As shown in Fig. 3D, S1P treatment had no effect on ceramide. These results suggest that S1P attenuates induction of autophagy by AA(Ϫ) without affecting ceramide formation in HL-60 cells.
In addition, we determined the effects of S1P on C 2 -ceramide-induced autophagy. Immunocytochemistry of LC3 showed that S1P treatment inhibited the formation of LC3 puncta induced by C 2 -ceramide (Fig. 4A). Similar to LC3 results, C 2 -ceramide-induced incorporation of MDC was also significantly decreased by simultaneous treatment with 10 M S1P from 45 Ϯ 3.9 to 21 Ϯ 2.7 arbitrary units (Fig. 4B). Even 20 M S1P treatment also had inhibitory effects on autophagy induction. These data suggest that S1P offsets ceramide-mediated induction of autophagy. S1P Treatment Activates the mTOR Pathway to Counteract Ceramide-mediated Induction of Autophagy-mTOR is conserved as a key molecule for cell responses to extracellular amino acids and growth factors in eukaryotic evolution (43,44). In an amino acid-rich environment, mTOR is phosphorylated and activated, thereby suppressing autophagy induction. Amino acid withdrawal or ceramide treatment have been reported to inactivate mTOR signaling to promote autophagy, thereby raising possibilities that S1P is inhibiting autophagy by activating mTOR signaling. At first, we examined the effects of S1P on mTOR activation. AA(Ϫ) treatment decreased the levels of phosphorylated/activated mTOR at Ser-2448, and simultaneous treatment with S1P restored mTOR phosphorylation (Fig. 5A). In addition, we determined the kinase activity of mTOR in S1P-treated cells. An in vitro kinase assay using 4E-BP1 as a substrate revealed that S1P treatment significantly increased mTOR activity in a time-and dose-dependent manner (Fig. 5B). Consistently, S1P treatment potently promoted the phosphorylation of mTOR downstream targets such as p70 S6K and 4E-BP1 (␤/␥ forms) (Fig. 5E). These data suggest that S1P signaling targets the mTOR pathway. To further confirm this, we tested the effect of rapamycin on S1P-induced mTOR activation and autophagy inhibition. Rapamycin directly inhibits mTOR and shuts down its signaling pathway, inducing autophagy (45). As shown in Fig. 5, C and D, S1P treatment failed to counteract LC3-II formation and autophagy induction by rapamycin. This rapamycin treatment also blocked S1P-mediated phosphorylation of p70 S6K and 4E-BP1 (Fig. 5E). These results strongly suggest that S1P potentiates mTOR signaling and consequently counteracts autophagy induction by amino acid deprivation. S1P-S1P 3 Receptor Signaling Offsets Autophagy-S1P binds to its cell surface receptors or intracellular target proteins, thereby generating biological responses. Therefore, we investigated how S1P affects mTOR-controlled autophagy. HL-60 cells have been reported to predominantly express S1P 3 (46). To confirm the gene expression profile of S1P receptors, Q-RT-PCR was performed. S1P 3 was predominantly expressed in HL-60 cells (Fig. 6A), confirming previous reports (47). To address whether S1P 3 plays a key role in autophagy inhibition by S1P, we examined the effect of the S1P 3 antagonist CAY10444 on S1P-mediated counteraction of autophagy. Antagonism of S1P 3 receptor by CAY10444 offset an inhibitory effect of S1P on LC3-II formation by AA(Ϫ) (Fig. 6B), suggesting a novel role of S1P 3 in suppressing autophagy in HL-60 cells. To further corroborate the importance of S1P 3 , we employed S1P 3 -overexpressing CHO cells (CHO/S1P 3 ) (32). To examine S1P 3 involvement in offsetting autophagy, we employed the S1P 3 overexpression system with CHO cells. AA(Ϫ) treatment decreased mTOR phosphorylation in CHO and S1P 3 -overexpressing CHO cells (CHO/S1P 3 ). Subsequent S1P treatment increased phosphorylation of mTOR in a dose-dependent manner, which was specific to CHO/S1P 3 cells. Its EC 50 was ϳ2.93 M. Akt and ERK are known as downstream factors of the S1P-S1P receptor (28), those kinases were also activated by S1P treatment in CHO/S1P 3 but not CHO cells (Fig. 7A). S1P treatment also increased phosphorylation/activation of mTOR downstream proteins such as p70 S6K or 4E-BP1 in CHO/S1P 3 cells (Fig. 7, A and B). Consistent with HL60 cells, S1P treatment is suggested to stimulate the mTOR pathway through the S1P 3 receptor. In CHO cells, C 2 -ceramide treatment also pro-    moted the dephosphorylation of mTOR and inactivated the mTOR pathway in the absence or presence of S1P, whereas CHO/S1P 3 cells stimulated with S1P showed increased phosphorylation in mTOR even in the presence of ceramide (Fig.  7C). These results indicate that extracellular S1P activates S1P 3 receptors and in turn counteracts the effects of ceramide. Consistent with mTOR phosphorylation, the increase in LC3-II formation by AA(Ϫ) or C 2 -ceramide was suppressed by treatment with S1P in CHO/S1P 3 cells, but not CHO cells (Fig. 7, B and C).
In addition, we examined the effects of S1P-S1P 3 receptors on autophagosome formation. tfLC3 (36) has green and red fluorescence in steady-state conditions but loses the green fluorescence in acidic compartments such as autophagosome lysosomes. As shown in Fig. 7, D and E, AA(Ϫ) and C 2 -ceramide treatment induced the disappearance of green fluorescence and puncta of red fluorescence in both CHO and CHO/S1P 3 cells. Selective to CHO/S1P 3 cells, S1P treatment inhibited AA(Ϫ)and C 2 -ceramide-induced formation of LC3 pancta. Similar to LC3 puncta formation, inhibition of autophagy by S1P treatment was confirmed to occur in only CHO/S1P 3 cells with the MDC incorporation assay (Fig. 7F). These data suggest that the S1P-S1P 3 signal activates the mTOR pathway, counteracting autophagy induction by the AA(Ϫ)-ceramide signal.
Involvement of CAPPs in mTOR Dephosphorylation/ Inactivation-The above data show that ceramide induces autophagy by promoting dephosphorylation/inactivation of mTOR signaling. Ceramide has been shown to activate serine/ threonine protein phosphatase containing PP1 and PP2A, which are called CAPPs (47,48). Thus we determined the activity of PP1/PP2A in AA(Ϫ) or C 2 -ceramide-treated cells. Both treatments with AA(Ϫ) and C 2 -ceramide increased PP1/PP2A activity as compared with AA(ϩ) treatment after 30 min (Fig.  8A). To clarify the involvement of PP1/PP2A in ceramide-mediated dephosphorylation of mTOR, we employed okadaic acid (49), which is a specific inhibitor for PP2A and PP1. As shown in Fig. 8B, C 2 -ceramide treatment clearly facilitated dephosphorylation of mTOR in CHO/S1P 3 cells, and okadaic acid treatment blocked its dephosphorylation in a dose-dependent manner. Furthermore, okadaic acid treatment also inhibited LC3 puncta formation (Fig. 8D). Similarly, okadaic acid treatment also suppressed both mTOR dephosphorylation and LC3 puncta formation induced by AA(Ϫ) (Fig. 8, C and E). These results indicate that okadaic acid-sensitive CAPPs are involved in dephosphorylation of mTOR induced by AA(Ϫ) and C 2 -ceramide followed by autophagy induction. Furthermore, okadaic acid treatment also suppressed cell death induced by both AA(Ϫ) and C 2 -ceramide (Fig. 8, F and G), implicating the involvement of a ceramide-dependent cellular signal in cell death concomitant with autophagy. S1P Protects Autophagy-associated Cell Death as a Survival Lipid-In mammalian cells, autophagy exerts cytoprotective effects against nutrient deprivation, whereas prolonged autophagy destroys the cytosol and organelles, inducing cell death, namely autophagy-associated cell death. To investigate whether autophagy has a role as cell death or cell survival in HL-60, we employed 3-MA, which is known as an autophagy inhibitor targeting class III PI 3-kinase. As shown Fig. 9A, 3-MA treatment was confirmed to inhibit LC3-II formation induced by AA(Ϫ). Next, we monitored cell death by DAPI staining in those conditions. Approximately 50% of cells displayed significant chromatin condensation after 2 h of AA(Ϫ) treatment (Fig.  9B). 3-MA treatment slightly increased cell death in control cells, whereas its treatment significantly suppressed AA(Ϫ)induced cell death. Those data suggested that autophagy triggers cell death induction. Moreover, we investigated whether S1P protects autophagy-associated cell death. S1P co-treatment decreased the number of DAPI-positive cells in AA(Ϫ)treated HL-60 cells (Fig. 10A). Treatment with C 2 -ceramide also increased cell death, whereas S1P treatment inhibited C 2 -ceramide-induced cell death in a dose-dependent manner. Similar to HL-60 cells, S1P treatment significantly blocked autophagy-associated cell death induced by both AA(Ϫ) and C 2 -ceramide in CHO/S1P 3 cells (Fig. 10, C and D, gray column). However, S1P could not prevent autophagy-associated cell death in CHO cells (Fig. 10, C and D, white column). Those results also indicate specific involvement of the S1P-S1P 3 pathway in the offset of autophagy-associated cell death activity.

DISCUSSION
In the present study, we uncovered the inhibitory effects of S1P on ceramide-mediated induction of autophagy. AA(Ϫ) stimulates acid SMase to generate ceramide that promotes the formation of autophagosomes, followed by autophagy-associated cell death. The regulation of mTOR by ceramide plays a key role in these biological responses, and S1P-S1P 3 signals counteract autophagy induction by activating the mTOR pathway. These findings demonstrate for the first time that cellular signaling arising from distinct actions of S1P and ceramide commonly target the mTOR pathway, leading to reciprocal regulation of autophagy and its associated cell death. Thus, the rheostat between sphingolipids is proposed to play an important role in autophagy regulation.
First, we showed that AA(Ϫ) induced an increase in ceramide. In apoptotic cell death, ceramide was mainly generated through activation of the de novo pathway or inhibition of the "sphingomyelin cycle" (50 -53). Some reports showed that autophagy accompanied ceramide formation through ceramide synthase, which mediates both de novo and salvage pathway (12,13,42). However, inhibition of ceramide synthases by fumonisin B had no effects on AA(Ϫ)-mediated autophagy in our system (Fig. 2). On the other hand, we found an increase of acid SMase activity with AA(Ϫ) treatment, and acid SMase knock-down suppressed autophagy induction (Fig. 2). These results suggested the implication of acid SMase in AA(Ϫ) stimulation of ceramide and autophagy induction. However, the mechanism by which AA(Ϫ) treatment activates acid SMase activation remains unclear.
Ceramide has been reported to inactivate the mTOR pathway or dissociate the beclin1-Bcl-2 complex, leading to autophagy (12,15,24). We have shown that ceramide elevation promotes mTOR dephosphorylation at Ser-2448 (Fig. 9). Okadaic acid-sensitive protein phosphatases such as CAPPs have been reported to interfere with the signal of mTOR whose upstream kinase is Akt. Ceramide is also known to promote the dephosphorylation and inactivation of Akt (54,55). Although we are yet to identify CAPP substrates, the ceramide-CAPP pathway is suggested to have an inhibitory effect on the mTOR pathway, which in turn induces autophagy. Our studies demonstrate, for the first time, an involvement of CAPPs in autophagy induction, and we propose that inactivation of the mTOR pathway by ceramide-CAPPs plays a key role in the induction of autophagy. S1P has been emerging as a cell proliferative lipid messenger. Intrinsic S1P-induced autophagy protects cells from cell death with apoptotic hallmarks during nutrient starvation (24). In contrast, we showed that extrinsic S1P inhibited ceramide-dependent induction of autophagy by engaging the S1P 3 receptors (Fig. 7). These reciprocal results raise the possibility that intrinsic regulation of autophagy by S1P is distinct from its extrinsic function. Interestingly, S1P/S1P 3 -mediated inhibition of autophagy led to the protection of cells from autophagy-associated cell death (Fig. 10). S1P is suggested to have reciprocal effects on autophagy induction, and even distinctly compartmentalized S1P exerted cell survival effects.
Consistent with a previous report (46), we confirmed that HL-60 cells predominantly expressed S1P 3 (Fig. 6). Banno et al. (56) demonstrated that the S1P-S1P 3 receptor axis activated the PI 3-kinase-Akt pathway and promoted proliferation. Proliferative stimulation with growth factors activated Akt, which in turn phosphorylated mTOR (57). Similarly S1P-S1P 3 signaling had stimulatory effects on the phosphorylation of Akt at Ser-473 and mTOR at 2448, and its downstream signals such as ERK1/2, p70 S6 kinase, and 4E-BP1 and inhibited ceramide-dependent autophagy (Fig. 7). mTOR forms two functional complexes, the rapamycin-sensitive mTOR complex 1 (mTORC1) and rapamycin-insensitive mTORC2 (58,59). It was recently reported that the PI 3-kinase-PDK1-mTORC1 and mTORC2 pathways independently regulate the phosphorylation of Akt at Ser-473 and Thr-308 (60). The activation of Akt at Ser-473 was negatively involved in FOXO3a translocation to the nucleus, leading to apoptotic cell death resistance (54,60). These data may suggest that extrinsic S1P activates Akt downstream of mTORC2, but the precise involvement of mTORC2 as compared with mTORC1 in S1P-inhibited autophagy remains to be elusive. S1P receptors have nanomolar binding affinities for S1P (19,21,61,62). Our results showed that treatments of S1P 3 overexpressing CHO cells with S1P (0.2-10 M) activated mTOR, Akt, and ERK (Fig. 7), and we have determined the EC 50 from Western blot results (Fig. 7A). The EC 50 values of S1P for the phosphorylation of mTOR, Akt, and ERK are 2.93, 2.59, and 2.4 M, respectively. For the considered K d values (0.23 nM) between S1P and S1P 3 receptor (61), those EC 50 values appear to be high. Indeed, some reports also showed the high values for S1P EC 50 . Okajima et al. (63) showed that S1P treatment increases intracellular Ca 2ϩ in HL-60 cells, and its S1P EC 50 was 1-2 M. Moreover, S1P EC 50 value of A549 cells for prostaglandin E 2 production was also shown to be ϳ0.85 M (64). What decreases the susceptibility to S1P remains unknown.
Recently, Liu et al. reported that S1P 1 signaling activates the Akt-mTOR pathway to impede the development and function of regulatory T cells (65), but HL-60 cells had no detectable expression of S1P 1 receptors (Fig. 6). In S1P 2 or S1P 4 expressing CHO cells, S1P treatment failed to counteract AA(Ϫ)-or ceramide-induced autophagy, although S1P was capable of activating ERK2 through both S1P 2 and S1P 4 , whose mRNAs were faintly detectable in HL-60 cells (Fig. 6 and supplemental Fig.  S1). On the other hand, S1P 5 receptor mRNA was detectable in Q-RT-PCR (Fig. 6). Chang et al. showed that S1P 5 is involved in autophagy induction at the long time serum starvation condition in PC3 cells (66). However, S1P treatment had no effects on autophagy induced by AA(Ϫ) in S1P 5 overexpressing CHO cells (supplemental Fig. S2). Therefore, these results strongly indicate that inhibitory effects of S1P on ceramide-induced autophagy and its associated cell death in HL-60 cells were attributed to S1P 3 signaling. Furthermore, using CHO/S1P 3 cells, we also demonstrated that S1P-S1P 3 signaling facilitates the activation of the mTOR pathway, which lead to autophagy suppression. In conclusion, ceramide-CAPPs and the S1P-S1P 3 pathway reciprocally regulate autophagy by targeting the mTOR pathway. We proposed that the "sphingolipid rheostat" by ceramide and S1P serves as a novel biological system for