Tumor Necrosis Factor-α-stimulated Cell Proliferation Is Mediated through Sphingosine Kinase-dependent Akt Activation and Cyclin D Expression*

Tumor necrosis factor-α (TNF-α) has been shown to activate sphingosine kinase (SphK) in a variety of cell types. The extent to which SphK signaling mediates the pleiotropic effects of TNF-α is not entirely clear. The current study examined the role of SphK activity in TNF-α-stimulated cell proliferation in 1321N1 glioblastoma cells. We first demonstrated that pharmacological inhibitors of SphK markedly decrease TNF-α-stimulated DNA synthesis. Signaling mechanisms through which SphK mediated the effect of TNF-α on DNA synthesis were then examined. Inhibition of Rho proteins with C3 exoenzyme or of Rho kinase with Y27632 attenuated TNF-α-stimulated DNA synthesis. However, RhoA activation by TNF-α was not blocked by SphK inhibition. ERK activation was also required for TNF-α-stimulated DNA synthesis but likewise TNF-α-induced ERK activation was not blocked by inhibition of SphK. Thus, neither RhoA nor ERK activation are the SphK-dependent transducers of TNF-α-induced proliferation. In contrast, TNF-α-stimulated Akt phosphorylation, which was also required for DNA synthesis, was attenuated by SphK inhibition or SphK1 knockdown by small interfering RNA. Furthermore, cyclin D expression was increased by TNF-α in a SphK- and Akt-dependent manner. Additional studies demonstrated that TNF-α effects on DNA synthesis, ERK, and Akt phosphorylation are not mediated through cell surface Gi -coupled S1P receptors, because none of these responses were inhibited by pertussis toxin. We conclude that SphK-dependent Akt activation plays a significant role in TNF-α-induced cyclin D expression and cell proliferation.

Tumor necrosis factor-␣ (TNF-␣) 4 is a pleiotropic cytokine that can affect diverse cellular responses. Although TNF-␣ is cytotoxic in many systems and can promote apoptosis when protein synthesis is inhibited, TNF-␣ can also act as an antiapoptotic signal and promote cell survival and proliferation. This is particularly evident in cancer cells, which tend to proliferate in response to TNF-␣ and are often resistant to TNF-␣induced apoptosis (1)(2)(3)(4). We studied a human glioblastoma cell line (1321N1) in which TNF-␣ acts in this manner and examined the basis for its proliferative effects.
TNF-␣ (5-7) and a range of agonists, including tyrosine kinase growth factors (8,9), G protein-coupled receptor ligands (10,11), and antigen receptor agonists (12,13), have been shown to activate SphK. SphK has emerged as an important mediator of cell proliferation, particularly in the context of cancer. There is considerable evidence suggesting that SphK can act as an oncogene, contributing to cell transformation (14). Overexpression of SphK1 increases proliferation of C6 glioma (6) and MCF-7 breast cancer cells (15) and increases cell cycle progression and promotes tumors in mice (15). In addition, siRNA-mediated blockade of SphK1 attenuates cell proliferation and arrests the cell cycle in breast cancer cells (16) and in U-1242 MG and U-87 MG glioblastoma cell lines (17). Although SphK activation is clearly linked to abnormal cell growth, potential involvement of SphK in TNF-␣-induced cell proliferation has not been clarified nor has the mechanism by which SphK enhances proliferation been fully elucidated.
Sphingosine 1 phosphate (S1P) is formed when SphK is activated. S1P, similar to TNF-␣, can mediate diverse cellular responses, including cell proliferation and protection from apoptosis as well as cell migration and cytoskeletal reorganization (18). S1P can be released from cells to act on cell surface S1P receptors, of which five distinct subtypes coupled to multiple heterotrimeric G proteins have been described (19). S1P has also been suggested to serve as an intracellular signal working through receptor-independent pathways (9,20,21).
The studies reported here have explored the role of SphK activation in the proliferative effects of TNF-␣, examined intracellular signaling mechanisms underlying the requirement for SphK, and asked whether SphK-mediated TNF-␣ signaling involves cell surface S1P receptors.

EXPERIMENTAL PROCEDURES
Materials-The plasmid for the glutathione S-transferase fusion proteins of the Rho binding domain of the Rho effector rhotekin was provided by Dr. M. Schwartz (22). Human TNF-␣ was purchased from R & D Systems and used at 100 ng/ml. S1P was purchased from Avanti Polar Lipids and used at 5 M. Pertussis toxin (PTX) was from Calbiochem and used at 100 ng/ml. C3 exoenzyme was from Cytoskeleton, Inc., and was used at 40 g/ml. N,N-Dimethylsphingosine (DMS) was from Biomol and used at 1 M. Sphingosine kinase inhibitor (SKI) and PD98059 were purchased from Calbiochem and used at 1 and 5 M, respectively. Akt inhibitors III and V were also purchased from Calbiochem and used at 10 and 1 M, respectively. Y27632 was from Calbiochem and used at 10 M. Anti-RhoA and anti-cyclin D antibodies were from Santa Cruz Biotechnology. Antibodies to phosphorylated p42/44 ERK, total p42/44 ERK, phosphorylated Akt (Ser-473), and total Akt were all purchased from Cell Signaling Technology. [␥-32 P]ATP (3000 Ci/mmol) was purchased from PerkinElmer Life Sciences. Silica gel 60A plates were from Whatman (Florham Park, NJ).
Cell Culture-1321N1 human astrocytoma cells were originally isolated from primary cultures of a human cerebral glioblastoma multiforme. The cells were cultured in Dulbecco's modified Eagle's medium supplemented with 5% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin in a 37°C 10% CO 2 -humidified environment, as described in several previous publications (23,24).
DNA Synthesis-Cells were plated in 24-well dishes and serum-starved for 48 h prior to experiment. Agonists were added for 24 h in the presence or absence of inhibitors. Cells were then labeled with 1 Ci of [ 3 H]thymidine/well for 6 h, washed with phosphate-buffered saline, and fixed with cold methanol. Proteins were precipitated with a series of trichloroacetic acid washes. Following the addition of 1 N NaOH for 30 min and then 1 N HCl, lysates were collected and [ 3 H]thymidine incorporation was assessed by scintillation quantification.
RhoA Activation-1321N1 cells were grown to confluence on 10-cm plates and serum-starved for 24 h prior to the experiment. Agonists were added in the presence or absence of inhibitors for the indicated times. Cells were washed with cold TBS, and lysis buffer (50 mM Tris-HCl, pH 7.4, 10% glycerol, 0.1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 100 mM NaCl, 5 mM MgCl 2 , 10 mg/ml aprotinin, and 10 mg/ml leupeptin) was added. The cells were scraped, collected, and homogenized. An aliquot of the lysate was reserved for total RhoA determination, and the remaining lysate was incubated with Rho binding domain to immunoprecipitate activated, GTPbound Rho for 45 min at 4°C. Beads and immunoprecipitated proteins were then subjected to a series of washes with lysis buffer and centrifugations. Laemmli buffer was added to the samples prior to separation on SDS-PAGE, as described above. Following immunoblotting with a RhoA antibody, activated GTP-bound RhoA was detected and normalized to total RhoA.
Whole Cell Lysate Preparation and Immunoblotting-Cells were grown to confluence on 10-cm plates and serum-starved for 24 h prior to the start of the experiment. Following treatment with agonists, the cells were washed once with cold phosphate-buffered saline, and then lysis buffer was added. Cells were scraped, collected, and incubated at 4°C with rotation for 30 min. Following centrifugation, supernatants were collected, 4ϫ Laemmli buffer was added, and samples were boiled for 5 min. Whole cell lysates were subjected to SDS-PAGE analysis.
Separated proteins were transferred to Immobilon membranes, which were then incubated with blocking buffer for 2 h. Primary antibodies were added overnight at 4°C, and secondary IgG-horseradish peroxidase was added at 1:4000 for 1 h at room temperature. Proteins were visualized using ECL and quantitated using Geldoc software.
siRNA Nucleofection-Cells were grown to confluence, harvested, and counted. The cells were nucleofected with 21-nucleotide siRNAs using the Amaxa system according to the manufacturer's instructions using solution V and program T16. Silencer predesigned siRNAs targeting human SphK1 were from Ambion. The two sequences that were used were 5Ј-GCUUCCUUGAACCAUUAUG-3Ј and 5Ј-GGUGCACCC-AAACUACUUC-3Ј. The Silencer ␤-actin siRNA-positive control kit (Ambion) was also used, which contains a predesigned siRNA targeting ␤-actin-positive control and a negative control-scrambled sequence. Cells were tested for siRNA effects and used for experiments 48 h post-nucleofection.
Quantitative PCR-Total RNA was extracted and cDNA was generated using the Superscript III first strand synthesis system for reverse transcription-PCR (Invitrogen) according to the manufacturer's instructions. TaqMan Gene Expression Assays (Applied Biosystems) were used to amplify human SphK1 (Hs00184211_m1), human SphK2 (Hs00219999_m1), or human ␤-actin (4236315E-0504006) in conjunction with the platinum quantitative PCR SuperMix-UDG kit (Invitrogen) and analyzed by the MJ Research Opticon 2 and Opticon Monitor software.
Assay of Sphingosine Kinase Activity-SphK activity was assayed using a previously published method (25). Briefly, cells were washed in cold phosphate-buffered saline and then scraped in lysis buffer (20 mM Tris-HCl, pH 7.6, 250 mM NaCl, 3 mM EDTA, 3 mM EGTA, 20 mM ␤-glycerophosphate, 0.5% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, 1 mM disodium 4-nitrophenylphosphate, 100 M sodium orthovanadate, and 10 g/ml leupeptin), collected, and incubated at 4°C with rotation for 30 min. Following centrifugation, supernatants were collected and 180 l (30 g of protein) was mixed with 10 l of 1 mM sphingosine in 5% Triton X-100 and 10 l of [ 32 P]ATP (10 Ci, 20 mM) containing MgCl 2 (200 mM). The reaction was carried out at 30 min at 37°C and terminated by the addition of 0.8 ml of chloroform/methanol/ HCl (100:200:1, v/v). After vortexing, 250 l of chloroform and 250 l of 2 M KCl were added, and phases were separated by centrifugation. The labeled lipids in the organic phase were resolved by thin layer chromatography on Silica Gel 60A with chloroform/acetone/methanol/acetic acid/water (10:4:3:2:1, v/v) and visualized by autoradiography. The radioactive spots corresponding to S1P were quantified with a Molecular Dynamics Storm phosphorimaging device and compared with known radioactive ATP standards to calculate the SphK-specific activity, expressed as picomoles of sphingosine-1-phosphate formed per minute per milligram of protein.

RESULTS
To determine whether TNF-␣ induces cell proliferation through a pathway that involves SphK, two pharmacological inhibitors with different mechanisms of action were utilized.

Role of SphK/Akt Signaling in TNF-␣-stimulated Proliferation
DMS inhibits SphK by competing with its substrate sphingosine (26), whereas SKI is a selective, non-competitive inhibitor of the ATP binding site of SphK (27). Cells were treated with 1 M DMS or 1 M SKI or vehicle for 15 min and then stimulated with 100 ng/ml TNF-␣ for 24 h; DNA synthesis was assessed using [ 3 H]thymidine incorporation. Fig. 1, A and B, shows that TNF-␣ treatment increased [ 3 H]thymidine incorporation ϳ15-fold and that this response was significantly attenuated following treatment with 1 M DMS (55% inhibition) or 1 M SKI (70% inhibition). Treatment with SphK inhibitors alone did not significantly decrease DNA synthesis (Fig. 1, A and B) nor did either inhibitor increase apoptosis (data not shown). To rule out the possible effects of these inhibitors on other downstream kinases required for DNA synthesis, we examined the effects of DMS and SKI on the response to serum. Neither DMS nor SKI attenuated serum-stimulated [ 3 H]thymidine incorporation (Fig. 1C). Thus, SphK appears to be uniquely involved at an early point in the signaling cascade by which TNF-␣ stimulates DNA synthesis.
We have previously shown that the small G protein Rho cooperates with Ras to allow cell cycle progression and DNA synthesis in 1321N1 cells (24,28). The role of Rho in TNF-␣mediated DNA synthesis was investigated by inhibiting Rho function through ribosylation with C3 exoenzyme (29). C3 treatment significantly attenuated TNF-␣-stimulated [ 3 H]thymidine incorporation ( Fig. 2A), indicating Rho involvement. TNF-␣-induced DNA synthesis was also attenuated when the Rho effector Rho kinase was inhibited with Y27632 ( Fig. 2B). RhoA activation was examined by immunoprecipitating the activated GTP-bound RhoA using the Rho effector rhotekin. Surprisingly, TNF-␣ did not activate RhoA at short times (3.5 min) or at 3 h, but RhoA activation was significantly elevated at 6 h (Fig. 3, A and B). Thus, RhoA activation is not an early event in TNF-␣ signaling but may instead be a secondary response elicited through TNF-␣ signaling pathways. To determine whether SphK activation was required, TNF-␣-induced RhoA activation was assessed following treatment with the SphK inhibitors DMS or SKI. TNF-␣-stimulated RhoA activation was not attenuated following SphK inhibition with DMS ( Fig. 3C) or SKI (data not shown). Thus, RhoA activation by TNF-␣ does not require SphK or its product S1P.
We next determined whether ERK activation was involved in TNF-␣-mediated DNA synthesis and, if so, whether TNF-␣stimulated ERK signaling was SphK-dependent. Inhibition of MEK (mitogen-activated protein kinase/extracellular signalregulated kinase kinase, the kinase upstream of ERK) with 5 M PD98059 significantly attenuated TNF-␣-stimulated [ 3 H]thymidine incorporation (Fig. 4A). TNF-␣ increased ERK phosphorylation at 15 min and 6 h (Fig. 4B). This response was not diminished in cells pretreated with DMS or SKI (Fig. 4B). Thus, although ERK signaling is involved in the proliferative effects of
Because Akt signaling has been shown to be aberrantly regulated in human glioblastomas (30) and can play a role in regulating cell proliferation by modulating cell cycle progression (31), we examined the involvement of Akt in TNF-␣-induced cell proliferation. Two commercially available Akt inhibitors (Akt III and V) and the PI3K inhibitor LY29002 were used to determine whether Akt signaling was involved in the TNF-␣stimulated proliferative response. First, the PI3K inhibitor LY29002 and Akt inhibitors III and V were shown to block TNF-␣-stimulated Akt phosphorylation (Fig. 5A). Both Akt inhibitors also significantly attenuated TNF-␣-stimulated [ 3 H]thymidine incorporation, as did the PI3K inhibitor LY29002 (Fig. 5B), implicating PI3K and Akt activation in TNF-␣-stimulated DNA synthesis.
To determine whether Akt might be the downstream target of SphK activation by TNF-␣, we examined the effects of the SphK inhibitors on TNF-␣-stimulated Akt phosphorylation (Fig. 6). In contrast to what was observed for ERK and Rho activation, inhibition of SphK with DMS or SKI attenuated TNF-␣-induced Akt phosphorylation (Fig. 6A). Quantitation of results from pooled experiments demonstrated a 45-50% inhibition of TNF-␣-stimulated Akt phosphorylation by DMS and SKI treatment, respectively (Fig. 6B).
SphK knockdown with siRNA was used as an additional approach to inhibit SphK and confirm that TNF-␣ effects on Akt phosphorylation were SphK-dependent. Control experiments were first performed to optimize conditions for siRNA knockdown of SphK1 gene expression without affecting SphK2 or actin gene expression. Two distinct SphK1 siRNA sequences   were tested, one of which (5Ј-GCUUCCUUGAACCAUU-AUG-3Ј) demonstrated more significant knockdown than the other (data not shown). Only this latter sequence was used in further studies. SphK1 siRNA specifically attenuated SphK1 gene expression compared with the mock-nucleofected control (Fig. 7A), with no significant effect on SphK2 (Fig. 7B) gene expression. Similar results, with up to 90% down-regulation of SphK1 mRNA, were obtained in additional experiments. To confirm that knockdown of SphK resulted in the attenuation of enzymatic activity, SphK activity was assessed following nucleofection of SphK1 siRNA. Knockdown of SphK1 resulted in a significant decrease of SphK activity (Fig. 7C). TNF-␣-stimulated Akt phosphorylation was then examined following 48 h of treatment with SphK1 siRNA or scrambled siRNA. In mocknucleofected cells, TNF-␣ induced an ϳ2-fold increase in Akt phosphorylation (Fig. 7D). In cells that were treated with SphK1 siRNA, TNF-␣-induced Akt phosphorylation (Fig. 7D) was significantly reduced compared with that seen in mock-nucleofected cells. The scrambled siRNA had no effect on TNF-␣stimulated Akt phosphorylation (Fig. 7D). These data confirm the findings with pharmacological inhibition of SphK in demonstrating a significant contribution of SphK1 to TNF-␣-induced Akt activation.
The effect of Akt on cell proliferation has been suggested to involve actions on cell cycle proteins such as p27, p21, and cyclin D (31). We therefore asked whether the SphK-Akt pathway stimulated by TNF-␣ could alter expression of these cell cycle proteins. TNF-␣ treatment decreased levels of the cell cycle inhibitor p27, but SphK inhibitors had no effect on this response (data not shown). However, TNF-␣ also enhanced the expression of cyclin D, a protein that mediates G 1 to S phase transition, promoting cell cycle progression (Fig. 8). Notably SphK inhibition with either DMS or SKI attenuated induction of cyclin D by TNF-␣ (Fig. 8). Treatment with the Akt inhibitors III and V caused comparable attenuation of TNF-␣-stimulated cyclin D expression (Fig. 8).
The data above demonstrate a role for SphK signaling and Akt activation in the proliferative response to TNF-␣. One possible mechanism by which this could occur is via the effects of SphK-generated S1P on G protein-coupled S1P receptors. Indeed, the addition of exogenous S1P (5 M) to 1321N1 cells enhanced [ 3 H]thymidine incorporation to a similar extent as TNF-␣ (Fig. 9, A and B). S1P also stimulated ERK and Akt  . SphK siRNA attenuates SphK activity and TNF-␣-induced Akt phosphorylation. SphK1 siRNA or scrambled (scr.) siRNA (all 50 nM) were nucleofected into 1321N1 cells using the Amaxa system, as described under "Experimental Procedures." For mock nucleofection (NF), cells were exposed to nucleofection solution V and nucleofected using the T16 program in the absence of siRNA. SphK1, SphK2, and actin (not shown) gene knockdown was assessed by quantitative PCR 48 h post-nucleofection, as described under "Experimental Procedures." Data are expressed as SphK1/actin (A) and SphK2/ actin (B). C, cells were nucleofected with siRNAs and 48 h later assessed for SphK activity, as described under "Experimental Procedures." *, p Ͻ 0.05 versus mock nucleofection. D, cells were nucleofected with siRNAs and 48 h later stimulated with vehicle or 100 ng/ml TNF-␣ for 15 min. Cells were lysed and subjected to SDS-PAGE followed by immunoblotting for P-Akt or total Akt. Following densitometric analysis, P-Akt was normalized to total Akt, and fold TNF-␣ over vehicle was calculated for each condition. *, p Ͻ 0.05 versus mock nucleofection.  JANUARY 12, 2007 • VOLUME 282 • NUMBER 2 phosphorylation to a similar extent as TNF-␣ (Fig. 9C). Inhibition of G i signaling by pertussis toxin (PTX) pretreatment completely blocked all of the responses to S1P (Fig. 9), indicating that extracellular S1P stimulates DNA synthesis, ERK, and Akt through receptors coupled to G i signaling. If TNF-␣ generates S1P, which elicits these responses via S1P receptors, then TNF-␣ actions should also be sensitive to PTX. As demonstrated in Fig. 9B, however, TNF-␣-stimulated DNA synthesis was insensitive to PTX treatment. TNF-␣-stimulated ERK and Akt phosphorylation were likewise unaffected by PTX treatment (Fig. 9C). These data argue against the possibility that TNF-␣ treatment generates S1P, which in turn regulates ERK, Akt, and DNA synthesis through extracellular G i -coupled S1P receptors. Further evidence that responses to TNF-␣ do not involve signaling through S1P receptors is that exogenous S1P elicits robust inositol phosphate formation and pronounced Ca 2ϩ mobilization in 1321N1 cells, whereas TNF-␣ does not (data not shown).

DISCUSSION
The data presented here demonstrate that TNF-␣-mediated DNA synthesis in 1321N1 astrocytoma cells involves several signaling pathways, including ERK, Rho, and SphK/PI3K/Akt. Involvement of ERK signaling in ligand-induced DNA synthesis is now well established (32), and considerable data from our group (23,24) and others (33,34) implicates Rho signaling in ligand-induced cell proliferation. In contrast, the involvement of SphK1 as a mediator of TNF-␣-induced cell proliferation, and in particular its ability to affect this mitogenic response to TNF-␣ through Akt and cyclin D, is novel and unexpected.
Akt is most often considered for its effects on cell survival, but a role for Akt in cell proliferation is well documented in cancer cells (31,35). Mechanistically, this has been linked to the ability of Akt to inhibit the expression of or mislocalize the cyclin-dependent kinase inhibitors p27 Kip1 and p21 WAF1/CIP1 , resulting in cell cycle progression (36,37). Akt has also been shown to contribute to cell cycle progression by phosphorylating glycogen synthase kinase-3␤, blocking its ability to target cyclin D for degradation (38,39). Subsequent cyclin D accumulation drives the G 1 to S phase transition (38). Data from the current study shows that cyclin D is a downstream target of Akt activated through SphK and TNF-␣. These findings extend previous work showing that overexpression of SphK1 increases cyclin D expression (15) by demonstrating that this can occur in response to SphK activation by an agonist.
Dysregulation of PI3K-Akt signaling is evident in several types of human cancers including glioblastomas, which have been shown to have mutations in PTEN, the phosphatase that converts phosphatidylinositol 1,4,5-trisphosphate back to phosphatidylinositol 1,4,5-diphosphate (40). As indicated earlier, SphK is also dysregulated in cancer cells. SphK1 is highly expressed in rat colon adenocarcinoma cells compared with normal mucosal cells (41). In addition, high expression of SphK1 was correlated with survival in patients with glioblastomas (17). Although it is not known whether 1321N1 cells have genetic alterations that would confer constitutive activation of Akt or enhance SphK signaling, the marked ability of TNF-␣ to signal through these pathways suggests that their up-regulation could play a significant role in directing TNF-␣ signaling toward mitogenesis.
Previous studies have shown that TNF-␣ activates SphK and that this results in the formation of S1P. Time course analysis has revealed that, in endothelial cells (5), C6 glioma cells (6), and human hepatocytes (7), SphK is significantly activated and S1P is increased following 10 -15 min of TNF-␣ exposure. Several lines of evidence suggest that S1P generated from SphK activation can act on cell surface S1P receptors (7,20,42). The study most relevant to our work is that of Osawa et al. (7), who demonstrate that S1P-mediated activation of G i -coupled cell surface receptors was required for TNF-␣ to activate Akt in hepatocytes. However, in contrast to these observations, we have demonstrated that PTX treatment abrogates S1P-stimulated DNA synthesis, Akt phosphorylation, and ERK phosphorylation but does not block responses stimulated by TNF-␣. Thus, in 1321N1 glioblastoma cells, the ability of TNF-␣ to activate Akt signaling via SphK appears to be independent of the G i -coupled cell surface receptors utilized by exogenous S1P.
Data from our study are consistent with other reports that SphK1 is critical for cell proliferation, but exerts its effects independently of G protein-coupled S1P receptors. Notably, FIGURE 9. Role of G i signaling on S1P-and TNF-␣-induced DNA synthesis, ERK, and Akt phosphorylation. 1321N1 cells were treated with 5 M S1P Ϯ 100 ng/ml PTX (A) or 100 ng/ml TNF-␣ Ϯ 100 ng/ml PTX (B) for 24 h. DNA synthesis was assessed by [ 3 H]thymidine incorporation. *, p Ͻ 0.001 versus control (Ctrl); ϩ, p Ͻ 0.001 versus agonist. C, 1321N1 cells were treated with 100 ng/ml PTX overnight followed by stimulation with 100 ng/ml TNF-␣ or 5 M S1P. Cells were lysed and subjected to SDS-PAGE followed by immunoblotting for phospho-p42/44 ERK or phospho-Akt (Ser-473).
Olivera et al. (9,20) demonstrate that overexpression of SphK1 promotes cell survival and proliferation even when various G␣ and G␤␥ subunits are blocked, as well as following treatment with PTX, and in MEFs from S1P 2/3 knock-out mice treated with PTX. Intracellular targets of S1P have been suggested. For example, there is evidence that S1P can mobilize intracellular Ca 2ϩ or affect store-operated calcium channels (11,43); however our data do not directly implicate S1P in the SphK-dependent effects of TNF-␣.
It is possible that the dependence of TNF-␣ action on SphK reflect other metabolic consequences of SphK activation. Sphingosine kinase has been described as a rheostat, adjusting the balance of proliferative and apoptotic signals through conversion of sphingosine and ceramide to S1P (18). Both ceramide and sphingosine have been shown to mediate apoptosis (44,45). Ceramide has also been reported to cause growth arrest and apoptosis by inhibiting Akt (46,47). Thus, in theory, the activation of SphK could decrease levels of its precursors sphingosine and ceramide, thereby enhancing Akt activation and cell proliferation. Conversely, pharmacologic inhibition of SphK should increase sphingosine and/or ceramide. However, TNF-␣ did not induce apoptosis in 1321N1 cells even when SphK was inhibited (data not shown). Thus, the SphK-dependent effects of TNF-␣ are likely a result of SphK-mediated Akt phosphorylation and cyclin D expression rather than secondary effects related to increased cellular ceramide or sphingosine.
The decision of a cell to die or to survive and replicate depends upon the extent to which apoptotic versus proliferative signals are activated. TNF-␣ is widely recognized to induce cell death in some settings and proliferation in others. The concept that TNF-␣ serves pro-survival and pro-proliferative functions in cancer cells is supported by this study. We suggest that it is the stimulation of SphK and subsequent Akt activation that tips this balance in favor of survival and proliferation in 1321N1 glioblastoma cells. Some reports suggest that TNF-␣ can act as a mitogen for astrocytes (48,49); however, we did not observe increased DNA synthesis in primary mouse astrocytes treated with this agonist (data not shown). Thus the proliferative effects of TNF-␣, mediated through SphK1-dependent Akt signaling and cyclin D, may be exaggerated in cancer cells.