Ceramide Activates the Stress-activated Protein Kinases*

Tumor necrosis factor (cid:97) (TNF (cid:97) ) activates the stress-activated protein kinases (SAPKs, also known as Jun nuclear kinases or JNKs) resulting in the stimulation of AP-1-dependent gene transcription and induces the translocation of NF (cid:107) B to the nucleus resulting in the stimulation of NF (cid:107) B-dependent gene transcription. A poten-tial second messenger for these signaling pathways is ceramide, which is generated when TNF (cid:97) activates sphingomyelinases. We show that treatment of HL-60 human promyelocytic cells with exogenous sphingomyelinase leads to rapid stimulation of JNK/SAPK activity, an effect not mimicked by treatment with phospholipase A 2 , C, or D. Further, JNK/SAPK activity is stimulated 2.7- and 2.8-fold, respectively, in cells exposed to C 2 -ceramide (5 (cid:109) M ) or TNF (cid:97) (10 ng/ml). The prolonged stimulation of this kinase activity by C 2 -ceramide is similar to that previ- ously reported for TNF (cid:97) . In contrast, the related mitogen-activated protein kinases ERK1 and ERK2 are weakly stimulated following TNF (cid:97) treatment (1.5-fold) and are inhibited by C 2 -ceramide treatment. TNF (cid:97) also potently stimulates NF- (cid:107) ) and tissue injury (hypoxia, UV irradiation, and chemo-therapeutic agents) activate both a neutral sphingomyelinase and JNK/SAPK. This study provides a link between the sphin-gomyelinase/ceramide pathway and the JNK/SAPK pathway by demonstrating that ceramide functions upstream of JNK/ SAPK, resulting in activation of nuclear downstream effects.

Tumor necrosis factor ␣ (TNF␣) activates the stressactivated protein kinases (SAPKs, also known as Jun nuclear kinases or JNKs) resulting in the stimulation of AP-1-dependent gene transcription and induces the translocation of NFB to the nucleus resulting in the stimulation of NFB-dependent gene transcription. A potential second messenger for these signaling pathways is ceramide, which is generated when TNF␣ activates sphingomyelinases. We show that treatment of HL-60 human promyelocytic cells with exogenous sphingomyelinase leads to rapid stimulation of JNK/SAPK activity, an effect not mimicked by treatment with phospholipase A 2 , C, or D. Further, JNK/SAPK activity is stimulated 2.7-and 2.8-fold, respectively, in cells exposed to C 2 -ceramide (5 M) or TNF␣ (10 ng/ml). The prolonged stimulation of this kinase activity by C 2 -ceramide is similar to that previously reported for TNF␣. In contrast, the related mitogenactivated protein kinases ERK1 and ERK2 are weakly stimulated following TNF␣ treatment (1.5-fold) and are inhibited by C 2 -ceramide treatment. TNF␣ also potently stimulates NF-B DNA binding activity and transcriptional activity, but these effects are not mimicked by addition of C 2 -ceramide or sphingomyelinase to intact cells. Furthermore, TNF␣, sphingomyelinase, and C 2 -ceramide induce c-jun, a gene that is stimulated by the ATF-2 and c-Jun transcription factors. These data suggest that ceramide may act as a second messenger for a subset of TNF␣'s biochemical and biological effects. TNF␣ 1 is a multifunctional cytokine involved in inflamma-tion, infection, and cancer (1,2). TNF␣ has potent biological effects on cultured cells, including induction of apoptosis and differentiation (3,4). TNF␣ signals via unknown second messengers leading to activation of NFB (5,6) and JNKs/SAPKs (7-9) and a weaker stimulation of MAP kinase activity (7,8). Each of these pathways leads to activation of a specific set of transcription factors; JNK/SAPK phosphorylates and activates c-Jun (10,11) and ATF-2 (12), NF-B translocates to the nucleus where it functions as a transcription factor, and MAP kinase phosphorylates and activates Elk-1 (13,14) and other transcription factors.
TNF␣ stimulates sphingomyelinase activity, which results in the generation of ceramide (15,16). Treatment of cells with the cell-permeable C 2 -ceramide mimics TNF␣ in the induction of apoptosis (4), differentiation (15), and in the activation of a ceramide-activated protein phosphatase activity similar to PP2A (17,18). A ceramide-activated protein kinase that is proline-directed has been described, but its specific target(s) is unknown (19). Thus, many of the biological and biochemical effects of TNF␣ are mimicked by ceramide treatment, but the molecular links between TNF␣-induced ceramide production and transcriptional activation are not clear.
Tissue Culture-HL-60 cells were maintained at a density of Ͻ1 ϫ 10 6 cells/ml for less than 1 month before discarding. Cells were cultured in RPMI 1640 (Life Technologies, Inc.) with 10% heat-inactivated fetal bovine serum (Gemini Bioproducts, Calabasas, CA) in a humidified atmosphere of 5% CO 2 , 95% air. For treatment with agonists, cells were resuspended in serum-free medium as described (18) at a density of 250,000 cells/ml and incubated for at least 2 h before treatment. L929 fibrosarcoma cells (21) were cultured in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 5% fetal bovine serum (Gemini).
Extract Preparation and in Vitro Kinase Assays-Whole cell extracts were prepared as described previously (22). For assessment of JNK/ SAPK activity, 50 g of extract was used in a solid state in vitro kinase assay as described (23) with GST-c-Jun (10) as the substrate. For assessment of MAP kinase activity, whole cell extracts were immunoprecipitated with antisera to ERK2, which cross-reacts with ERK1 (SC-94, Santa Cruz Biotechnology, Santa Cruz, CA). Immune complexes were collected with protein A/G-agarose (Oncogene Science, Uniondale, NY), washed extensively, and used in an in vitro kinase assay with 50 M ATP (5 Ci of [␥-32 P]ATP), 10 mM MgCl 2 , 1 mM dithiothreitol, 0.1 mM Na 3 VO 4 , 25 mM ␤-glycerophosphate, 10 mM p-nitrophenyl phosphate for 20 min at 30°C. 2 g of GST-ElkC (a glutathione S-transferase fusion protein containing the C terminus of Elk-1, which contains multiple MAP kinase phosphorylation sites) was used as substrate (14,24). Proteins were fractionated on 12.5% SDSpolyacrylamide gel electrophoresis and subjected to autoradiography.
Gel Mobility Shift Assays-NF-B gel mobility shift assays were performed as described previously (3). Antibodies used in binding assays (1 l/reaction) were obtained from Santa Cruz Biotechnology.

RESULTS AND DISCUSSION
We investigated the role of ceramide in TNF␣-stimulated signal transduction by comparing the effects of TNF␣, sphingomyelinase, and C 2 -ceramide on JNK/SAPK activity, NFB activity, and MAP kinase activity. Exogenous bacterial sphingomyelinase catalyzes the hydrolysis of cell membrane sphingomyelin and the formation of ceramide when added to intact HL-60 human myelocytic leukemia cells (31,32). Treatment of cells with exogenous sphingomyelinase potently activates JNK/ SAPK activity, while equivalent amounts of the lipases phospholipase C, phospholipase A 2 , or phospholipase D do not (Fig.  1A). Exogenously added C 2 -ceramide or TNF␣ stimulates JNK/ SAPK activity in HL-60 cells to comparable levels (Fig. 1A). The activation of JNK/SAPK activity by C 2 -ceramide is first detected in cell extracts after 10 min (Fig. 1B). In contrast to agents that transiently stimulate JNK/SAPK activity, such as epidermal growth factor and phorbol esters (8), ceramide elicits a prolonged activation (more than 2-fold over background levels) at least 2 h after stimulation. The kinetics of JNK/SAPK activation by ceramide parallel those of TNF␣ stimulation (data not shown and Ref. 8).
Increases in JNK/SAPK activity are observed at C 2 -ceramide concentrations of 1 M and above (Fig. 1C). 5 M C 2 -ceramide stimulates JNK/SAPK activity 2.7-fold over untreated cell levels (average of seven experiments), while TNF␣ (10 ng/ml) stimulates JNK activity 2.8-fold (average of four experiments). Uptake of C 2 -ceramide applied at 3 M (16, 20) results in an intracellular C 2 -ceramide concentration that is equimolar to the concentration of ceramides generated endogenously following cell treatment with TNF␣ (15).
Because C 2 -ceramide is an amphiphilic lipid analog that may have nonspecific activities, it was important to establish the specificity of its activation of JNK/SAPK, a family of kinases activated by a wide variety of stress stimuli (7). DL-erythrodihydro-C 2 -ceramide is a close structural analog of C 2 -ceramide lacking the trans-unsaturated bond in the sphingosine moiety. Previous studies (20) demonstrated that the uptake and metabolism of radiolabeled C 2 -ceramide and dihydro-C 2 -ceramide are similar. DL-erythro-dihydro-C 2 -ceramide fails to activate JNK/SAPK activity (Fig. 1C, lane 8). Therefore, the effects of C 2 -ceramide on JNK/SAPK are specific and suggest a specific interaction of ceramide with a component of the JNK/ SAPK pathway.
TNF␣ is a weak inducer of MAP kinase (ERK1 and ERK2) activity relative to its induction of JNK/SAPK or relative to the induction of MAP kinases by growth factors such as epidermal growth factor (7,8). In HL-60 cells, high concentrations of TNF␣ weakly activate MAP kinase activity using recombinant GST-Elk activation domain as a substrate in an immune complex kinase assay (1.5-fold over unstimulated levels) (Fig. 3,  lane 5). As reported elsewhere (35), high concentrations of exogenous sphingomyelinase activate MAP kinase activity (Fig. 3, lane 7). However, treatment of cells with C 2 -ceramide almost completely attenuates basal MAP kinase activity in a concentration-dependent manner (Fig. 3, lanes 2, 3, and 8), and this effect was not observed with dihydro-C 2 -ceramide-treated cells (Fig. 3, lanes 4 and 9). Similar results are obtained using myelin basic protein as a substrate for MAP kinase in an immune complex kinase assay or using an in-gel MAP kinase assay (36) with MBP incorporated into the gel (data not shown).
The ability of ceramide to activate JNK/SAPK predicts that ceramide would activate downstream targets of this kinase cascade. The c-jun gene contains two non-consensus AP-1 binding sites in its promoter, which are recognized with high affinity by protein complexes containing c-Jun and ATF-2 (37,38). The transcriptional activation domains of both c-Jun (11) and ATF-2 (12) are phosphorylated and activated by JNK/SAPK. TNF␣, C 2 -ceramide, and sphingomyelinase stimulate c-jun mRNA levels in HL-60 cells (Fig. 4A). Thus, all three agents that stimulate JNK/SAPK activity also lead to enhanced expression of the c-jun gene. The heightened ability of sphingomyelinase to activate c-jun transcription may be due to its activation of Elk-1 (Fig. 3), which can stimulate c-fos transcription and hence increase total AP-1 activity.
To further assess the functional sequela of stimulation by TNF␣ or bacterial sphingomyelinase, we measured transcriptional activities with reporter gene assays in L929 fibrosarcoma cells, which are efficiently transfected and are responsive to TNF␣ (21). TNF␣ and sphingomyelinase are equally effective in stimulating the reporter gene driven by the c-jun promoter (Jun-Luc, Fig. 4B, top panel), consistent with their stimulation of JNK/SAPK activity (Fig. 1) and c-jun mRNA level (Fig. 4A). The transcriptional activity of a fusion protein consisting of the Gal4 DNA binding domain and the Elk-1 transactivation domain (Gal-ElkC (14); a MAP kinase substrate), measured by a reporter plasmid containing a luciferase gene with Gal4 binding sites (5XGal-luciferase), is activated by sphingomyelinase and to a lesser degree by TNF␣ (Fig. 4B, middle panel), in accord with their stimulation of MAP kinase activity in vitro (Fig. 3). TNF␣ but not sphingomyelinase markedly stimulated an NF-B-responsive reporter gene ((KB)3luciferase, Fig. 4B, bottom panel), consistent with the effect of these agents on NF-B DNA binding activity (Fig. 2).
TNF␣ signals through two cell surface receptors, TNF-R1 (p55) and TNF-R2 (p75), which contain no apparent catalytic activity and whose intracellular domains are not homologous to characterized signaling proteins. Two proteins, TRAF1 and TRAF2, bind to the cytoplasmic domain of TNF-R2 (39). An unrelated protein, TRADD, binds to the TNF-R1-associated death domain and might be involved in TNF-induced apoptosis and NF-B activation (40). Most of TNF␣'s biological effects are mediated by TNF-R1 (41)(42)(43), but the mechanisms by which second messengers are recruited are unknown. Several of TNF␣'s second messengers are activated via TNF-R1, including the activation of a sphingomyelinase, protein kinase C, and phospholipase A 2 (41). By using differences in the species specificity of TNF␣ (44, 45), we have found that TNF␣ also activates JNK/SAPK through TNF-R1 (data not shown).
Recent studies have delineated the steps immediately upstream of JNK/SAPK activation (46 -48), but connections between these cytoplasmic and nuclear kinases and upstream messengers leading to their activation are unresolved. This study demonstrates that ceramide is a link between binding of TNF␣ to TNFR1 and activation of a cytoplasmic kinase cascade that results in stimulation of JNK/SAPK activity and c-jun expression. TNF␣ and C 2 -ceramide activate JNK/SAPK to the same degree (Fig. 1) and stimulate activity with similar kinetics in HL-60 cells. We have previously demonstrated that JNK1 (also known as SAPK␥) is a major component of TNF␣-activated JNK/SAPK (8). By activating JNK/SAPK, TNF␣ and ceramide activate a subset of AP-1 transcription factors, such as c-Jun and ATF-2, which in turn will preferentially induce genes with specific non-consensus AP-1 binding sites, such as the c-jun gene itself (Fig. 4). Therefore, this study provides strong evidence that ceramide functions as the second messenger in TNF␣ signaling resulting in the activation of FIG. 3. MAP kinase activity is stimulated by sphingomyelinase but not ceramide. An immune complex MAP kinase assay was performed with HL-60 cell extracts prepared following cell treatment (Tx) with the indicated agonists for 10 min (lanes 1-5) or 2 min (lanes 6 -9). C5 and C10, 5 and 10 M D-e-C 2 -ceramide; C5(di), 5 M DL-erythrodihydro-C 2 -ceramide; SM, 100 milliunits/ml Staphylococcus aureus sphingomyelinase; TNF, 50 ng/ml recombinant human TNF␣; con, control. GST-ElkC was used as the substrate. Equal substrate loading was confirmed in all cases by Coomassie Blue staining of the gels prior to autoradiography.

FIG. 4. Stimulation of c-jun transcriptional activity.
A, c-jun mRNA levels are increased following TNF␣, sphingomyelinase, and C 2 -ceramide treatment. HL-60 cells were incubated with TNF␣ (TNF, 20 ng/ml), sphingomyelinase (SM, 100 milliunits/ml), C 2 -ceramide (CER, 10 M), or untreated (NONE) for 1 h, and total RNA was prepared followed by electrophoresis and Northern blotting. The blot was hybridized first with a radioactively labeled human c-jun probe (upper panel) and then with a labeled human glyceraldehyde-3-phosphate dehydrogenase (GAPDH) probe. Blots were quantitated by phosphor image analysis, and -fold induction of c-jun mRNA normalized to GAPDH mRNA (FOLD) is indicated. B, activity of luciferase reporter genes in the L929 fibrosarcoma cell line. Jun-luciferase (Luc), or 5XGalluciferase plus Gal-ElkC, or (KB)3-luciferase (three copies of the consensus NF-B response element linked to luciferase) was transfected into L929 cells. Cells were then treated with TNF␣ (20 ng/ml) or sphingomyelinase (100 milliunits/ml) for 5 h, and extracts were prepared for determination of luciferase activity. -Fold activation represents reporter activity relative to cells treated with vehicle only. Results from one representative experiment performed in duplicate are shown (ϮS.E.). Results were consistent over at least three experiments. JNK/SAPK and c-jun.
TNF␣ is not a potent activator of MAP kinases (7,8), and we demonstrate here that C 2 -ceramide actually decreases MAP kinase activity in HL-60 cells. Inhibition of MAP kinase activity may result from the activity of a ceramide-activated protein phosphatase (17,18) acting on a component of the MAP kinase cascade. Ceramide-activated protein phosphatase is a phosphatase of the PP2A class, and PP2A activity on both MAP kinases and MEKs has been described (reviewed in Ref. 49). The weak activation of MAP kinase activity by TNF␣ may be the result of ceramide-independent activation of a component (MEK or MAP kinase itself) downstream of Raf-1 and/or by activation of protein kinase C (50).
TNF␣ is a potent inducer of NFB (5, 51) (Figs. 2 and 4), perhaps through the activation of protein kinase C (52), and ceramide potentiates the stimulation of NFB by TNF␣ (3). Ceramide stimulates NF-B binding activity in permeabilized cells, which has been attributed to the activation of an acidic sphingomyelinase (32). However, extracts derived from intact HL-60 cells treated with exogenous sphingomyelinase or C 2ceramide have elevated JNK/SAPK activity but not NF-B DNA binding activity. Perhaps these different results reflect the existence of separate ceramide pools, such that C 2 -ceramide and the ceramide generated by exogenous bacterial sphingomyelinase are in a different cellular compartment than the ceramide generated by acidic sphingomyelinase.
The evidence is pointing to the existence of a novel signal transduction pathway that may be specifically involved in the stress response. A number of agents of systemic stress (e.g. TNF␣) and tissue injury (hypoxia, UV irradiation, and chemotherapeutic agents) activate both a neutral sphingomyelinase and JNK/SAPK. This study provides a link between the sphingomyelinase/ceramide pathway and the JNK/SAPK pathway by demonstrating that ceramide functions upstream of JNK/ SAPK, resulting in activation of nuclear downstream effects.