TNF-α-Induced Sphingosine 1-Phosphate Inhibits Apoptosis Through a Phosphatidylinositol 3-Kinase/Akt Pathway in Human Hepatocytes¹

Tumor necrosis factor-α is a multifunctional cytokine that is involved in inflammation, immunity, anti-viral responses, and a variety of diseases. In the liver, TNF-α modulates hepatocyte responses, depending on the physiological circumstances. TNF-α is involved in viral hepatitis, alcoholic liver disease, and fulminant hepatitis, and its toxicity is particularly important in the pathophysiology of hepatocytes (1). TNF-α is known to activate a variety of components implicated in cellular signal transduction. Binding of TNF-α to TNFR-1 (p55) results in trimerization of its C-terminal cytoplasmic “death domain” and recruitment of some intracellular proteins involved in apoptotic signal transduction (2, 3, 4). However, hepatocytes are normally resistant to the cytotoxicity of TNF-α. Administration of TNF-α alone does not induce hepatocyte apoptosis in mice (5), suggesting that TNF-α also induces the molecules that protect cells from apoptosis by TNF-α.

The self-protective signaling pathway of TNF-α previously has been reported. TNF-α strongly activates NF-κB through a second class of adaptor protein TNFR-associated factors, and this transcriptional factor regulates the expression of antiapoptotic gene products (6, 7, 8), such as antiapoptotic members of the Bcl-2 family (9) and the inhibitor of apoptosis proteins (IAP)³ c-IAP1 and c-IAP2 (10). Therefore, apoptosis of rat and mouse hepatocytes can be induced by TNF-α when NF-κB activation or RNA transcription is repressed (6, 11, 12, 13).

In addition to activating the NF-κB pathway, TNF-α recently was shown to activate other antiapoptotic signaling pathway(s). In human endothelial cells (14), TNF-α activates sphingosine kinase (SphK), which converts sphingosine to sphingosine 1-phosphate (S1P). This lipid-derived mediator also is shown to prevent the cytotoxic action of TNF-α. In hepatoma HTC4 cells (15), exogenous S1P exerts its antiapoptotic activity, which is mediated by G protein-coupled receptors encoded by the endothelial differentiation gene (Edg) 3 and 5. Furthermore, in several types of cells such as HeLa and human endothelial cells, TNF-α appears to activate serine/threonine kinase Akt via phosphatidylinositol 3-kinase (PI3K; Refs. 16, 17, 18, 19), which protects cells from apoptosis caused by TNF-α (18, 19, 20, 21).

In the present study, we have examined the interrelationships among the antiapoptotic signaling components NF-κB, SphK, and PI3K/Akt in human hepatocyte Huh-7 and Hc cells stimulated with TNF-α. TNF-α was observed to activate SphK to generate S1P, which then acts as a protective factor against apoptosis. This antiapoptotic effect of S1P was dependent on the PI3K/Akt pathway but independent of the NF-κB pathway. These results indicate that TNF-α simultaneously but independently activates at least these two survival signaling pathways. Human hepatocytes can be sensitized to TNF-α-mediated apoptosis by blocking these pathways.

Huh-7 cells, a human hepatoma cell line, were obtained from the Japanese Cancer Research Resources Bank (Tokyo, Japan). Hc cells (normal human hepatocytes) were purchased from the Applied Cell Biology Research Institute (Kirkland, WA). Cell culture media for Huh-7 (RPMI 1640) and Hc cells (CS-C complete) were obtained from Life Technologies (Rockville, MD) and from Cell Systems (Kirkland, WA), respectively. Recombinant human TNF-α and TNF-related apoptosis-inducing ligand (TRAIL) were obtained from Genzyme (Cambridge, MA). Fas ligand (FasL) was obtained from Upstate Biotechnology (Lake Placid, NY). N,N-dimethylsphingosine (DMS), d,h-threo-dihydrosphingosine (DHS), sphingosine, and S1P were obtained from Matreya (Pleasant Gap, PA). Pertussis toxin (PTX), PD 98059, and GF 109203X were obtained from Calbiochem-Novabiochem (La Jolla, CA). LY 294002 was obtained from Alexis (San Diego, CA). Hoechst 33258 (bisbenzimide) staining dye was obtained from Wako (Osaka, Japan). A broad spectrum caspase inhibitor, z-VAD-FMK, was obtained from the Peptide Institute (Osaka, Japan). [α-³²P]dCTP was obtained from ICN Biomedicals (Costa Mesa, CA). [γ-³²P]ATP was obtained from NEN Life Science Products (Boston, MA). [¹⁴C]Serine was obtained from American Radiolabeled Chemicals (St. Louis, MO). Abs against NF-κB components and poly(ADP-ribose) polymerase (PARP) were obtained from Santa Cruz Biotechnology (Santa Cruz, CA) and those against phosphorylated Akt (Ser⁴⁷³) and Akt were obtained from Cell Signaling Technology (Beverly, MA). Anti-phosphotyrosine Ab (PY20) was obtained from Transduction Laboratories (Lexington, KY). Anti-rabbit IgG HRP-coupled secondary Ab and the ECL Western blot detection system were obtained from Amersham-Pharmacia Biotech (Buckinghamshire, U.K.). High-performance TLC (HPTLC) plates were obtained from Merck (Darmstadt, Germany). The adenovirus 5 (Ad5) variant expressing a green fluorescent protein (GFP) was provided by Hisanori Kojima (Gifu International Institute of Biotechnology, Gifu, Japan). All other reagents used were of the highest analytical grade available.

The human hepatoma cell line Huh-7 was cultured in RPMI 1640 medium supplemented with 1% (v/v) FBS, antibiotics (penicillin and streptomycin), and 2 mg/ml lactoalbumin. Hc cells were cultured in CS-C complete medium supplemented with the same antibiotics. A total of 5 × 10⁵ and 1 × 10⁶ cells were plated on 60- and 100-mm dishes, respectively. After 24 h of incubation with growth medium, the cells were washed twice with PBS, and the medium was changed to serum-free RPMI 1640. Then cells were incubated for another 24 h in the presence or absence of recombinant adenoviruses at a multiplicity of infection of 25. Before stimulation with 20 ng/ml TNF-α, 20–100 ng/ml TRAIL, or 20–100 ng/ml FasL, the cells were washed twice with PBS and, if necessary, incubated for 1 h in serum-free RPMI 1640 containing the indicated agent(s): 10–50 μM DMS or DHS, 25 μM LY 294002, 100 μM PD 98059, or 100 μM z-VAD-FMK. In some experiments, cells were pretreated with 100 ng/ml PTX for 24 h before TNF-α or S1P stimulation.

Nuclear extracts were prepared by the method of Schreiber et al. (22), with slight modifications (6). NF-κB-binding consensus single-strand oligonucleotide (5′-TAGTTGAGGGGACTTTCCCAGG-3′) was first annealed with the complement oligonucleotide (5′-TGCCTGGGAAAGTCCCCTCAACTA-3′). The annealed DNA fragment was labeled with [α-³²P]dCTP by Klenow DNA polymerase. Nuclear protein (20 μg) was incubated with 2.5 ng of ³²P-labeled double-strand oligonucleotide probe for 30 min at room temperature. For the Ab supershift assays, 2 μl of the Ab, anti-p50-, or anti-p65-specific polyclonal Ab against the individual NF-κB components, was added to the reaction mixture and incubated for another 30 min. The reaction mixture was electrophoresed on 4% polyacrylamide gels with 0.5 × Tris-borate-EDTA buffer at 4°C. Gels were dried and exposed to film for autoradiography.

For quantitation of apoptotic cells, cells were stained with Hoechst 33258, and nuclear morphological changes were examined. Briefly, harvested cells were fixed at 4°C for 30 min with 1% glutaraldehyde in PBS, stained with 1 mM Hoechst 33258 for 30 min, and examined under a fluorescent microscope (BX60; Olympus, Tokyo, Japan) with excitation at 360 nm. For the analysis of DNA fragmentation, cells were resuspended in ice-cold lysis buffer containing 50 mM Tris-HCl (pH 7.8), 10 mM EDTA, 0.5% (w/v) sodium N-lauroylsarcosinate, and 1% Triton X-100. The lysates were centrifuged at 20,000 × g for 20 min, and the resulting supernatants were treated with proteinase K (0.1 mg/ml) and DNase-free RNase (10 μg/ml) for 2 h at 37°C. DNA was precipitated with ethanol and electrophoresed on 2% agarose gels containing 0.023% (v/v) ethidium bromide. The DNA fragmentation pattern was detected by UV transillumination.

S1P formation was measured as described previously (14) with slight modifications. For the radio-labeling of sphingolipids, the cells (1 × 10⁶ cells/100-mm dish) were incubated with regular growth medium containing [¹⁴C]serine (1 μCi/ml) for 48 h. Then, the medium was changed to serum-free RPMI 1640 medium containing [¹⁴C]serine, and the cells were incubated for another 24 h. The radiolabeled cells were stimulated by TNF-α with or without LY 294002 or DMS preincubation. Cellular lipids were extracted by the method of Bligh and Dyer (23) and separated on HPTLC plates in the solvent system of 1-butanol/acetic acid/water (60:20:20, v/v; Ref. 24). The radioactive S1P spot, identified by comigration with an authentic standard, was scraped off the plate, and the radioactivity was measured in a liquid scintillation counter (LS-6500; Beckman Coulter, Fullerton, CA).

SphK activity was measured as described previously (14). Briefly, cells were washed with ice-cold PBS and sonicated in lysis buffer (10 mM Tris-HCl, pH 7.4, 10% glycerol, 0.5 mM MgCl₂, 0.5 mM EDTA, 0.5 mM DTT, 1 mM PMSF, 1 mM sodium orthovanadate, 10 mM sodium fluoride, 0.1 mM sodium molybdate, and 0.5 mM 4-deoxypyridoxine). After ultracentrifugation at 100,000 × g for 30 min, SphK activity in the supernatant was measured by incubation with 20 μM sphingosine-BSA complex and [γ-³²P]ATP (1 μCi/assay) for 30 min at 37°C in reaction buffer (20 mM Tris-HCl, pH 7.4, 2.5 mM MgCl₂, 0.5 mM EDTA, 20% glycerol, 1.2 mM DTT, and 0.5 mM 4-deoxypyridoxine). The reaction was stopped by adding HCl to obtain a 0.1 M final concentration. Radiolabeled lipids were separated on HPTLC plates in the solvent system described above. After autoradiography, the spot corresponding to S1P was scraped off the plate and the radioactivity was measured. The SphK activity was normalized based on the total protein. Protein concentrations were assayed by using the Bradford protein assay reagent with BSA as a standard.

Cytosolic proteins were used for the Western blot analysis of Akt and phospho-Akt, and the total cellular protein extracts were used for PARP detection. For isolation of cytosolic proteins, cells were sonicated in lysis buffer (10 mM Tris-HCl, pH 7.4, 10% glycerol, 0.5 mM MgCl₂, 0.5 mM EDTA, 0.5 mM DTT, 1 mM PMSF, 1 mM sodium orthovanadate, 10 mM sodium fluoride, 0.1 mM sodium molybdate, and 0.5 mM 4-deoxypyridoxine) and ultracentrifuged at 100,000 × g for 30 min. The resulting supernatant was used for the cytosolic fraction. For the preparation of total cell proteins, cells were sonicated in RIPA buffer (50 mM Tris-HCl, pH 8.8, 150 mM NaCl, 10 mM EGTA, 1% Triton X-100, 0.1% SDS, 1% deoxycholic acid, 0.3 mM PMSF, and 30 μg/ml (l-3-trans-carboxyoxirane-2-carboryl)-l-leucyl-agmatine. The proteins were separated by SDS-PAGE and were electrophoretically transferred onto polyvinylidene difluoride membranes. The membranes were probed with the Abs against Akt, phospho-Akt, and PARP, and then incubated with the anti-rabbit IgG HRP-coupled secondary Ab. Detection was performed with an ECL system.

PI3K activity was measured as described previously (25). Briefly, Hc cells were washed twice with PBS and lysed in lysis buffer (1% Nonidet P-40, 20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM EDTA, 1 mM PMSF, and 1 mM sodium orthovanadate). After centrifugation at 3000 × g for 10 min, the supernatant was incubated with anti-phosphotyrosine Ab at 4°C for 2 h. The immunocomplex was precipitated with a mixture of protein G plus A-Sepharose. The immunocomplex was incubated with 200 μg/ml phosphatidylinositol and 10 μM ATP including [γ-³²P]ATP (1.0 μCi/assay) for 30 min at 25°C in 50 μl of reaction buffer (20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 0.5 mM EDTA, and 20 mM MgCl₂). Labeled lipids were extracted and separated on HPTLC plates in a solvent system of chloroform/methanol/25% ammonia/water (43:38:5:7, v/v; Ref. 25). After autoradiography, the radioactive phosphatidylinositol 3-phosphate spot was scraped off the plate and the radioactivity was measured.

TNF-α induced NF-κB activation within 5 min in both human hepatoma Huh-7 cells (Fig. 1,A) and normal human hepatocyte Hc cells (Fig. 1,B). The activated NF-κB complex in Huh-7 cells was mainly composed of p50-p65 heterodimers, whereas Hc cells contained p50-p50 homodimers as well as heterodimers, as determined by supershifts. NF-κB activation by TNF-α was almost abolished when Huh-7 and Hc cells were infected with Ad5IκB but not with control adenovirus Ad5GFP (Fig. 1). Because of missense mutations at phosphorylation sites, where serines 32 and 36 are replaced with alanines, the mutant IκB irreversibly binds to NF-κB and prevents its activation (6).

Normal mouse or rat hepatocytes are usually resistant to the cytotoxicity of TNF-α. However, infection with Ad5IκB sensitizes hepatocytes to TNF-α-mediated apoptosis (6, 13, 26). To assess whether the NF-κB inactivation sensitizes human hepatocytes to apoptosis by TNF-α, Huh-7 and Hc cells were infected with Ad5IκB. The adenovirus infection or TNF-α alone did not cause apoptotic changes (data not shown). TNF-α treatment induced only 10–20% apoptosis in Huh-7 and Hc cells infected with Ad5IκB at 24 h, as inferred by Hoechst 33258 staining (Table I). The sensitizing effect by Ad5IκB was much less in human hepatocytes than in the mouse and rat hepatocytes observed in previous reports (6, 26).

The above results indicate that survival signal(s) other than NF-κB may be activated by TNF-α treatment of human hepatocytes. Previous studies demonstrated that SphK (14), Akt (18, 19), and extracellular signal-regulated kinase (ERK) (27), which have been proposed to mediate antiapoptotic actions in several cell types, were activated by TNF-α. To gain further insight into the mechanisms of self-protection by TNF-α in human hepatocytes, the roles of SphK, PI3K, and ERK were examined in the presence of their specific or selective inhibitors. Pretreatment of Ad5IκB-infected Huh-7 cells with SphK inhibitors (DMS and DHS) greatly enhanced TNF-α-induced apoptosis (Fig. 2,A). The extent of apoptotic cell death was 71.4 and 50.1% at 24 h in the presence of 10 μM DMS or 10 μM DHS, respectively. In contrast, death rate was 12.0% in the absence of DMS and DHS. The PI3K inhibitor LY 294002 (25 μM) also brought about a fourfold increase in cell death. In contrast, PD 98059 (100 μM), a commonly used inhibitor of mitogen-activated protein kinase/ERK kinase (MEK) 1, had no effect on TNF-α-induced apoptosis in Ad5IκB-infected cells (Fig. 2,A), although activation of ERK1/2 was abrogated, as assessed by phosphorylation with anti-phospho-ERK1/2 Abs (data not shown). Apoptotic death of Hc cells infected with Ad5IκB also was augmented by 10 μM DMS and 25 μM LY 294002 (Fig. 2,B). However, NF-κB activation induced by TNF-α was not prevented by either DMS or LY 294002 (Fig. 2,C), indicating that NF-κB activation was independent of SphK and PI3K/Akt. These data suggest that SphK and PI3K play important roles in mediating survival signaling in an NF-κB-independent manner. Interestingly, DMS (10 μM) pretreatment sensitized both Huh-7 and Hc cells to TNF-α-induced apoptosis without Ad5IκB infection (Table II). In contrast, LY 294002 itself did not sensitize to TNF-α-mediated apoptosis without simultaneous Ad5IκB infection (Table II).

Previous studies have demonstrated that TNF-α causes caspase activation and PARP cleavage in rat hepatocyte RALA255-10G cells infected with Ad5IκB (6, 28). To assess whether caspase activation occurs, cleavage of PARP, a substrate for caspases, was examined by Western blot analysis. PARP cleavage occurred within 24 h after TNF-α treatment (Fig. 3,A). Moreover, a broad-spectrum caspase inhibitor, z-VAD-FMK, completely prevented cell death and PARP cleavage (Fig. 3,B). This finding indicates that apoptosis induced by TNF-α is dependent on caspase activation. Apoptotic cell death and its enhancement by DMS and LY 294002 also were confirmed by fragmentation of chromosomal DNA (Fig. 3 C).

To further explore the role of SphK, SphK activity and the amount of S1P produced were measured in Hc cells stimulated with TNF-α. The TNF-α treatment caused a rapid increase in SphK activity, which reached a maximum within 10 min (Fig. 4,A1). Consistent with SphK activation, the S1P level elevated rapidly, peaking at 10 min after TNF-α treatment (Fig. 4,B1). DMS inhibited TNF-α-induced SphK activation (Fig. 4,A2), which in turn led to reduced S1P generation (Fig. 4 B2) in a concentration-dependent manner in Hc cells.

The enhancement of TNF-α-induced apoptosis by LY 294002 suggests the involvement of PI3K in human hepatocyte resistance to TNF-α-mediated apoptosis. In fact, TNF-α effected a 3.5-fold increase in the activation of PI3K in Hc cells 10 min after the administration of TNF-α (Fig. 5,A). At 25 μM, LY 294002 completely abolished the TNF-α-induced activation of PI3K. Akt, a pivotal factor in cell survival (18, 19, 20, 21) and downstream effector of PI3K (29), has been reported to be activated by TNF-α (16, 17, 18, 19). In Hc cells, the phosphorylated (activated) form of Akt appeared 15 min after administration of TNF-α (Fig. 5,B), and 25 μM LY 294002 prevented TNF-α-induced activation of Akt (Fig. 5 B).

As described above, TNF-α protected hepatocytes from apoptosis by activating the SphK and PI3K/Akt pathways. To examine the relationship between SphK and PI3K/Akt in the TNF-α-mediated survival pathway, Ad5IκB-infected Hc cells were preincubated with both DMS and LY 294002, and then the number of apoptotic cells was measured 12 h after TNF-α treatment. The combined treatment did not show an additive effect on apoptosis (data not shown). DMS inhibited TNF-α-induced activation of PI3K (Fig. 6,A) and Akt (Fig. 6,B) in a concentration-dependent manner. A previous report demonstrated the inhibition of protein kinase C (PKC) by the addition of DMS (30). However, in the present study, PKC inhibitor GF 109203X had no effect on TNF-α-induced Akt activation (Fig. 6 C), although it abolished Akt activation by treatment with 100 ng/ml PMA (data not shown). Therefore, PKC did not mediate TNF-α-induced Akt activation. These findings indicate that TNF-α activates the PI3K/Akt pathway through SphK activation.

Accumulating evidence indicates that S1P formed by SphK acts not only as an autocrine and/or paracrine ligand via the Edg receptor(s), but also as an intracellular second messenger (31). Accordingly, the effect of exogenous S1P on the PI3K/Akt pathway was examined. At 1 μM, S1P caused PI3K activation, but it was inhibited by LY 294002 (25 μM) (Fig. 7,A). In contrast, 30 μM DMS could not inhibit the PI3K activation caused by S1P. At 1 μM, S1P was noted to activate Akt as early as 2 min after administration (Fig. 7,B). The involvement of Edg receptor(s) in the activation of Akt was examined in Hc cells pretreated with 100 ng/ml PTX for 24 h. Akt activation by exogenous S1P was completely blocked by PTX pretreatment, indicating that extracellular S1P acted through the Edg receptor(s) coupled with G_i/o. However, under the same conditions, the activation of Akt by TNF-α was only partially inhibited (Fig. 7,C). Moreover, exogenous S1P (1 μM) partially protected both Ad5IκB-infected and -uninfected Hc cells from apoptosis induced by TNF-α plus DMS (Table III). These data indicate that the SphK/S1P pathway play a critical role in protecting human hepatocytes against apoptosis by TNF-α.

To further investigate whether the survival signal via SphK/S1P pathway modulates hepatocyte apoptosis mediated by other death receptors, the effects of TRAIL and FasL were examined in Hc cells. Activation of NF-κB and Akt were discernible by stimulation with FasL at 100 ng/ml (Fig. 8, A and B). In contrast, TRAIL (20–200 ng/ml) failed to activate these survival factors (Fig. 8, A and B). FasL (100 ng/ml) alone could induce apoptosis of Hc cells, as we previously reported in in vivo mouse hepatic failure model (5). Pretreatment of Ad5IκB and/or DMS potentiated FasL-induced apoptosis (Table IV), and the effects of these factors were additive. Exogenous S1P (1 μM) significantly protected Hc cells from apoptosis induced by FasL in the presence of DMS (Table IV). In contrast, 20 ng/ml TRAIL failed to induce apoptosis of Hc cells, even when cells were treated with both Ad5IκB and DMS (data not shown).

TNF-α is a potent mediator of hepatotoxicity in vivo and in cultured cells (5, 6, 11, 12). However, TNF-α alone cannot induce apoptosis in normal hepatocytes (5, 6). A sensitization step is required for the induction of apoptosis, because TNF-α also activates antiapoptotic signal pathway(s). Blockage of TNF-α-induced activation of the transcriptional factor NF-κB, which induces the expression of protective genes, is sufficient to induce apoptosis in cultured rat hepatocyte RALA255-10G cells (6) and in rat hepatocytes after partial hepatectomy (13). TNF-α treatment killed 88% of the primary cultured mouse hepatocytes (26) and 50% of the RALA255-10G rat hepatocytes (6) infected with Ad5IκB. However, in human hepatocyte cell lines (Huh-7 and Hc cells), inhibition of NF-κB by Ad5IκB was insufficient to induce massive cell death by TNF-α treatment (Table I). These results led us to consider the involvement of additional survival signaling factor(s) other than NF-κB. Therefore, we have attempted to reveal the mechanism(s) of resistance to TNF-α-mediated apoptosis in human hepatocytes. The results obtained in the present study indicate that TNF-α induces not only NF-κB activation but also S1P generation via SphK, which activates survival signals such as the PI3K/Akt pathway and protects human hepatocytes from TNF-α-induced apoptosis independently of NF-κB.

In several types of cells, such as HEK293 and COS7 cells, TNF-α induces ceramide formation by the activation of sphingomyelinase (32, 33). Ceramide is further hydrolyzed by ceramidase to sphingosine, which subsequently is converted to S1P by SphK. Ceramide is thought to be a second messenger involved in the apoptotic process (34, 35). Thus, the balance between intracellular concentrations of ceramide and S1P may be a critical factor in the determination of cell fate (36, 37). Therefore, perhaps the SphK inhibitor DMS enhances the ceramide level by sphingosine accumulation and leads to TNF-α-mediated apoptosis. However, ceramide accumulation was undetectable at least within the first 30 min after TNF-α treatment, and DMS did not alter the ceramide level in TNF-α-treated Hc cells (data not shown). Ceramide appears to induce hepatocyte apoptosis via a caspase-independent pathway (38). In contrast, the apoptosis enhanced by DMS was dependent on caspase activation (Fig. 3). Therefore, the potentiation of apoptosis by DMS is not attributable to ceramide accumulation, and the activation of SphK is one of the critical steps for the antiapoptotic action. However, the regulatory mechanisms of SphK activation and S1P formation remain unclear.

TNF-α activated PI3K and Akt in Hc cells. Activation of Akt via PI3K has been considered to protect cells from apoptosis induced by TNF-α (18, 19, 20, 21). In Ad5IκB-infected hepatocytes, the PI3K inhibitor LY 294002 potentiated the cytotoxicity of TNF-α. PI3K is activated by TNF-α though interaction with an adapter protein, namely, Grb2 (39) or Ras (40). In our system, Akt activation by TNF-α was blocked by DMS and exogenous S1P-activated Akt. Activation of SphK and S1P formation peaked at 10 min, thus preceding Akt activation (detectable at 15 min) after TNF-α treatment. S1P activates c-Src tyrosine kinases and promotes Grb2-PI3K complex formation (41). These findings suggest that S1P formation induced activation of Akt in TNF-α-treated human hepatocytes. S1P is reported to act as a ligand for the Edg receptor(s) and also as an intracellular second messenger (31, 42). For example, microinjected S1P induced DNA synthesis in Swiss 3T3 cells (43) and overproduced intracellular S1P in NIH3T3 fibroblasts and HEK293 cells by overexpression of SphK, which promoted cell growth and survival (42). In our system, exogenous S1P underwent Akt activation, but it was completely inhibited by PTX pretreatment. In contrast, PTX caused partial inhibition of the Akt activation because of treatment with TNF-α. Administration of exogenous S1P resulted in a partial rescue from death induced by TNF-α plus DMS. Thus, S1P produced by TNF-α treatment activates Akt intracellularly and also functions as an extracellular ligand for Edg receptor(s) in human hepatocytes. DMS also enhanced apoptotic death of Hc cells induced by another death receptor agonist, FasL, which caused Akt activation. However, exogenous S1P prevented apoptosis induced by FasL plus DMS. These results indicate that the survival signaling via the SphK/S1P pathway also may operate in Fas-mediated apoptosis of human hepatocytes.

Mechanisms for the antiapoptotic effects of Akt activation have been reported previously. In some types of cells, NF-κB is a potential target for the PI3K/Akt pathway (16, 17, 21). However, in human endothelial cells, antiapoptotic action of Akt was independent of NF-κB activation (19). Similarly, in TNF-α-stimulated HeLa cells, wortmannin, a PI3K inhibitor, did not inhibit NF-κB activation (18). In our system, NF-κB activation was not prevented by LY 294002. Moreover, S1P did not induce NF-κB activation in Huh-7 cells (data not shown). These results suggest that survival mechanism(s) other than NF-κB exists downstream of PI3K/Akt. Pastorino et al. (18) reported that TNF-α induced phosphorylation of BAD through PI3K/Akt pathway and that phosphorylated BAD lost its ability to bind to Bcl-X_L, which is known to act on mitochondria to block the apoptotic signaling cascade (44). TNF-α was reported to induce apoptosis in hepatocytes via mitochondrial permeability transition (26). In contrast, overexpression of Bcl-X_L prevented the liver injury caused by TNF-α plus d-galactosamine (45). These findings lead us to speculate that Akt protects hepatocytes from TNF-α cytotoxicity through inhibition of apoptotic mitochondrial events.

The DMS-induced enhancement of apoptosis was not affected by the addition of LY 294002, and DMS inhibited TNF-α-induced activation of PI3K and Akt. Therefore, the effect of DMS is meditated by the inhibition of PI3K/Akt pathway through blockage of S1P formation. DMS pretreatment was sufficient to sensitize cells to TNF-α cytotoxicity without prior infection with Ad5IκB, whereas LY 294002 alone was unable to enhance apoptosis by TNF-α (Table II). LY 294002 exerted an enhancing effect on apoptosis induced by TNF-α in the presence of Ad5IκB. These data suggest that blockage of the PI3K pathway is insufficient to sensitize human hepatocytes to TNF-α-induced apoptosis and that the sensitizing effect of DMS alone on TNF-α-induced apoptosis may be attributable to mechanism(s) other than PI3K/Akt inhibition. In addition to Akt, S1P stimulates many signaling pathways, such as ERK, cAMP-dependent kinase, and focal adhesion kinase (31). Indeed, ERK was activated by S1P stimulation in Hc cells (data not shown). However, PD 98059, a selective inhibitor of MEK acting upstream of ERK, had no enhancing effect on apoptosis induced by TNF-α in Ad5IκB-infected hepatocytes. Therefore, the MEK/ERK cascade does not participate in the S1P-mediated survival signaling pathway in human hepatocytes. cAMP appears to inhibit apoptosis of primary rat hepatocytes by cAMP-dependent kinase activation (46, 47), and activation of focal adhesion kinase is an antiapoptotic effect in HL-60 cells (48). Therefore, it is reasonable to speculate that these survival signals also are involved in sensitizing effect by DMS. Their involvements in S1P-mediated antiapoptotic action in human hepatocytes should be examined to draw a complete picture of TNF-α signaling. An alternate interpretation, as Pitson et al. (49) reported, is that DMS inhibited the basal (housekeeping) SphK activity in unstimulated cells. DMS may have other, nonspecific effects in addition to SphK inhibition (29, 50). These effects of DMS also may be involved in sensitizing action of DMS. Although pathway(s) other than the PI3K/Akt cannot be excluded at the present stage, we would like to propose a new hypothesis that SphK-mediated formation of S1P plays an important role in the antiapoptotic signaling transduction mediated by TNF-α.

In summary, we have shown here that TNF-α activates the PI3K/Akt pathway via SphK activation and S1P formation in human hepatocytes, and that this pathway regulates apoptosis mediated by TNFR and Fas. This protective effect appears to be independent of NF-κB. The hypothetical signaling pathways were schematically summarized in Fig. 9. Therefore, regulation of S1P levels may present a new therapeutic approach for liver diseases.

We thank Dr. Hisanori Kojima for providing adenovirus and Dr. Yasuhiro Yamada for help with gel shift assay.