Toxins convert the hepatocellular response to tumor necrosis factor-α (TNF-α) stimulation from proliferation to cell death, suggesting that hepatotoxins somehow sensitize hepatocytes to TNF-α toxicity. Because nuclear factor-κB (NF-κB) activation confers resistance to TNF-α cytotoxicity in nonhepatic cells, the possibility that toxin-induced sensitization to TNF-α killing results from inhibition of NF-κB-dependent gene expression was examined in the RALA rat hepatocyte cell line sensitized to TNF-α cytotoxicity by actinomycin D (ActD). ActD did not affect TNF-α-induced hepatocyte NF-κB activation but decreased NF-κB-dependent gene expression. Expression of an IκB superrepressor rendered RALA hepatocytes sensitive to TNF-α-induced apoptosis in the absence of ActD. Apoptosis was blocked by caspase inhibitors, and TNF-α treatment led to activation of caspase-2, caspase-3, and caspase-8 only when NF-κB activation was blocked. Although apoptosis was blocked by the NF-κB-dependent factor nitric oxide (NO), inhibition of endogenous NO production did not sensitize cells to TNF-α-induced cytotoxicity. Thus NF-κB activation is the critical intracellular signal that determines whether TNF-α stimulates hepatocyte proliferation or apoptosis. Although exogenous NO blocks RALA hepatocyte TNF-α cytotoxicity, endogenous production of NO is not the mechanism by which NF-κB activation inhibits this death pathway.

  • caspases
  • nitric oxide
  • inducible nitric oxide synthase
  • liver
  • hydrogen peroxide

tumor necrosis factor-α (TNF-α) activates a wide array of cellular signaling pathways that result in divergent biological responses depending on the physiological setting (30). Although considerable interest has centered on responses in transformed cells because of the potential of TNF-α as an anticancer agent, TNF-α also regulates normal cell function. In vivo studies in the liver have demonstrated that increased TNF-α expression modulates contrasting hepatocyte responses depending on the stimulus. During liver regeneration after partial hepatectomy, increased TNF-α production initiates or promotes hepatocyte proliferation (1, 36). In contrast, in the setting of liver regrowth following toxin-induced hepatocyte injury, TNF-α production triggers cell death rather than proliferation. TNF-α expression is increased in animal models of toxic liver injury (4, 8, 10, 21) and in humans during alcohol-induced liver disease (26, 37). Although TNF-α production is a normal response to tissue injury, TNF-α in the setting of toxin-induced liver injury acts as a hepatocyte cytotoxin. Investigations employing neutralizing antibodies or soluble receptors to block biological activity of TNF-α have shown that TNF-α neutralization significantly reduces the liver damage (4, 11, 18) and mortality (11) induced by a variety of toxins. These findings indicate that the degree of cell death following toxin-induced liver injury depends largely on the cytotoxic effects of TNF-α rather than on the direct biochemical effects of the toxin or its metabolites. The mechanism by which hepatotoxins sensitize hepatocytes to further injury from TNF-α is unknown.

Studies in many cell systems, including hepatocytes, have shown that inhibition of RNA or protein synthesis sensitizes cells to TNF-α toxicity, presumably by interfering with the upregulation of a TNF-α-inducible protective factor (23). Liver toxins also interfere with hepatocyte macromolecular synthesis, suggesting that they may sensitize hepatocytes to TNF-α cytotoxicity by a similar mechanism. Investigations in nonhepatic cells suggested that failure to upregulate specific antioxidant defenses sensitized these cells to TNF-α toxicity (20, 34). However, prior studies have demonstrated that the failure to upregulate hepatocellular antioxidant defenses such as manganese superoxide dismutase (9) or glutathione (Xu, unpublished data) cannot account for the sensitization of hepatocytes to TNF-α-induced cell death. Recently it has been shown that blocking the normal activation of the transcription factor nuclear factor-κB (NF-κB) by TNF-α sensitizes fibroblasts, macrophages, and several transformed cells to TNF-α cytotoxicity (2, 25, 31, 32). It has been proposed that after TNF-α stimulation NF-κB activation is critical in determining whether a cell enters into a pathway leading to survival or to cell death (38). We hypothesized that toxins may therefore sensitize hepatocytes to TNF-α toxicity by blocking the upregulation of an NF-κB-dependent protective gene. These investigations assessed the role that NF-κB activation plays in the sensitization of hepatocytes to TNF-α cytotoxicity by actinomycin D (ActD).



All reagents were from Sigma (St. Louis, MO), unless otherwise indicated.

Cells and culture conditions.

The rat hepatocyte cell line RALA255-10G (6) was cultured in DMEM (GIBCO BRL, Grand Island, NY) supplemented with 4% fetal bovine serum (HyClone, Logan, UT), 2 mM glutamine, and antibiotics (GIBCO BRL). These cells are conditionally transformed with a temperature-sensitive T antigen. At the permissive temperature of 33°C they express T antigen, remain undifferentiated, and proliferate. Culture of the cells at the restrictive temperature of 37°C suppresses T antigen expression, markedly slows growth, and allows differentiated gene expression (6, 13). For these experiments, the cells were cultured at 33°C until confluent, trypsinized, and replated at 0.5 × 106 cells/dish on 35-mm plastic dishes (Falcon, Becton Dickinson, Lincoln Park, NJ). After 24 h, the medium was changed to DMEM supplemented with 2% fetal bovine serum, glutamine, antibiotics, and 1 μM dexamethasone, and the cells were placed at 37°C. After 3 days of culture at 37°C, the cells received fresh serum-free medium containing dexamethasone. Medium was supplemented with dexamethasone to optimize hepatocyte differentiation as previously described (6). Cells were pretreated 24 h later with ActD (15 ng/ml) for 30 min, and then mouse recombinant TNF-α (R&D Systems, Minneapolis, MN) was added to some dishes, at a concentration of 10 ng/ml unless otherwise indicated.

To inhibit caspase activity, cells were pretreated for 30 min before the addition of ActD with 100 μM Ac-Tyr-Val-Ala-Asp-chloromethylketone (YVAD-CMK) or Ac-Asp-Glu-Val-Asp aldehyde (DEVD-CHO) (BACHEM, Torrance, CA), and 25 μMN-[(indole-2-carbonyl)alaninyl]-3-amino-4-oxo-5-fluoropentanoic acid (IDN-1529) orN-[(1,3-dimethylindole-2-carbonyl)valinyl]-3-amino-4-oxo-5-fluoropentanoic acid (IDN-1965) (IDUN Pharmaceuticals, La Jolla, CA) dissolved in DMSO. IDN-1529 has broad anti-caspase activity, inhibiting caspase-1, caspase-3, caspase-6, and caspase-8, whereas IDN-1965 selectively inhibits caspase-6 and caspase-8 (J. Wu, personal communication). The nitric oxide (NO) donorS-nitroso-N-acetylpenicillamine (SNAP; Molecular Probes, Eugene, OR) was dissolved in DMSO and added at a concentration of 750 μM.N 6-(iminoethyl)-l-lysine hydrochloride (l-NIL) at 0.25–1.0 mM andN G-methyl-l-arginine acetate (l-NMMA) at 1.5–5.0 mM (Molecular Probes) were employed to inhibit inducible NO synthase (iNOS). The higher concentrations of either iNOS inhibitor blocked >95% of the NO produced by RALA hepatocytes following combined stimulation with TNF-α (10 ng/ml) and interleukin-1β (National Cancer Institute Biological Response Modifiers Program, Frederick, MD; 30 ng/ml).

[3H]thymidine incorporation.

Cells were incubated with 2.5 μCi of [3H]thymidine (DuPont-NEN, Boston, MA) for 2 h. The medium was removed, and the cells were washed in PBS and then homogenized in 1 ml of 0.33 N NaOH; 0.3 ml of 40% TCA-1.2 M HCl was added, and the solution was centrifuged. The pellet was redissolved in 0.33 N NaOH, and the incorporated counts/min (cpm) were determined by scintillation counting. The DNA content was determined by means of bisbenzimidazole spectrofluorometry (24), and the amount of [3H]thymidine incorporation was calculated as cpm/mg of DNA.

MTT assay.

The amount of cell death was determined by examining cell number with the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay (12). The cell culture medium was removed, and serum-free medium containing 1 mg/ml MTT was added to the cells. After a 1-h incubation, this medium was removed and the formazan product was solubilized inn-propanol. This solution was alkalinized with sodium hydroxide, and the absorbance at 560 nM was measured in a spectrophotometer. The percent cell survival was calculated by taking the optical density (OD) reading of cells given a particular treatment, dividing that number by the OD reading for the untreated, control cells, and then multiplying by 100. The accuracy of the MTT assay was validated by comparing it with the number of trypan blue-excluding cells recovered from trypsinized dishes.

Microscopic determination of apoptosis.

Phase-contrast and fluorescence microscopy were conducted as previously described (35). The relative number of apoptotic cells was determined by fluorescent costaining with acridine orange and ethidium bromide as previously employed (35). The percent of cells with apoptotic morphology (nuclear and cytoplasmic condensation, nuclear fragmentation, membrane blebbing, and apoptotic body formation) was determined by examining >400 cells/dish. Necrosis was excluded by the absence of ethidium bromide staining.

Electrophoretic mobility shift assays.

Nuclear proteins were isolated by the method of Schreiber et al. (28) with slight modification as previously described (35). Electrophoretic mobility shift assays were performed with the use of a commercially supplied oligonucleotide for the NF-κB consensus sequence (Santa Cruz Biotechnology, Santa Cruz, CA). The DNA binding reaction was performed at room temperature for 20 min in a 20-μl reaction mixture consisting of 5 μg of nuclear extract, 50 μg/ml polydeoxyinosinic-deoxycytidylic acid, 10 mM Tris (pH 7.5), 100 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol (DTT), 1 mg/ml BSA, 10% glycerol, and 25,000 cpm 32P-end-labeled oligonucleotide. After incubation, the samples were resolved on a 4% polyacrylamide gel, dried, and subjected to autoradiography. For supershift assays, 8 μg of anti-p50 NF-κB, anti-p65 NF-κB, or anti-Stat3 antibody (Santa Cruz Biotechnology) were added to the reaction mixture, and the incubation time was extended for an additional 20 min.

Transfections and luciferase assays.

RALA hepatocytes were transfected with NF-κB-Luc, a luciferase reporter gene driven by three NF-κB binding sites (16), using Lipofectamine (GIBCO-BRL). All treatments were initiated 48 h after the time of transfection. To assay luciferase activity, cells were washed in PBS, lysed with a buffer containing 1% Triton X-100 (Promega, Madison, WI), scraped from the dish, and centrifuged, and the cell extract was assayed for luciferase activity in a luminometer. All luciferase values were normalized for extract protein concentration.

Adenovirus construction and infection.

A recombinant replication-deficient adenovirus, Ad5IκB, was constructed as previously described (19). This adenovirus contains an IκB in which serines 32 and 36 are mutated to alanines, driven by the cytomegalovirus promoter-enhancer. The presence of the mutant IκB sequence packaged into the recombinant Ad5 virus (Ad5IκB) was confirmed by PCR and by Western blotting. Ad5IκB was grown in 293 cells and purified by banding twice on CsCl gradients. Titers of viral particles were determined by optical densitometry, and recombinant virus was then stored in 10% (vol/vol) glycerol at −20°C. A control virus, Ad5LacZ, which contains the Escherichia coli β-galactosidase gene, was also grown and purified as described above.

Twenty-seven hours before the addition of TNF-α, 35-mm dishes were infected with 5 × 109particles of Ad5LacZ or Ad5IκB (∼2.5 × 103 particles/cell or 10–25 plaque-forming units/cell). Three hours after the addition of virus, the medium was changed to serum-free medium containing dexamethasone identical to the treatment given noninfected cells.

Protein isolation and Western blot analysis.

Cells were scraped in the medium and centrifuged. The cell pellet was resuspended in lysis buffer containing 10 mM HEPES (pH 7.4), 42 mM MgCl2, 1% Triton, 1 mM phenylmethylsulfonyl fluoride, 1 mM EDTA, 1 mM DTT, and 2 μg/ml pepstatin A, leupeptin, and aprotinin. The solution was then mixed at 4°C for 30 min. After centrifugation, the supernatant was collected and the protein concentration was determined by the Bio-Rad protein assay (Bio-Rad, Hercules, CA).

Fifty micrograms of protein were heated in 1× SDS gel loading buffer [50 mM Tris (pH 6.8), 100 mM DTT, 2% SDS, 0.1% bromphenol blue, and 10% glycerol] at 100°C for 2 min. The samples were subjected to 10% SDS-PAGE and subsequently transferred to a nitrocellulose membrane (Schleicher & Schuell, Keene, NH) in a transfer buffer containing 39 mM glycine, 48 mM Tris (pH 8.3), 0.037% SDS, and 15% methanol, using a Bio-Rad Trans-blot SD semidry transfer cell to which 50 mA was applied for 18 h. All membranes were stained with Ponceau red to ensure equivalent amounts of protein loading and electrophoretic transfer among samples. Blocking of the membranes was performed using a solution (TBS-T) of 5% dry milk, 10 mM Tris (pH 8.0), 0.15 M NaCl, and 0.05% Tween 20 for 1 h. A rabbit anti-caspase-1 polyclonal IgG (Santa Cruz Biotechnology), rabbit anti-caspase-2 polyclonal IgG (Santa Cruz Biotechnology), rabbit anti-caspase-3 polyclonal IgG (33), or rabbit caspase-8 polyclonal IgG (29) was used as primary antibody at 1:1,000, 1:2,000, 1:4,000, and 1:2,000 dilutions, respectively, in 5% dry milk-TBS-T for 2 h. A goat anti-rabbit IgG conjugated with horseradish peroxidase (GIBCO BRL) was used as a secondary antibody at a 1:10,000 dilution in 5% dry milk-TBS-T blocking solution for 1 h. After use of a chemiluminescence detection system (Supersignal Ultra, Pierce, Rockford, IL), the membranes were exposed to Reflection film (DuPont-NEN).

For poly(ADP-ribose) polymerase (PARP) immunoblots, the protein isolation procedure was altered slightly. Cultured cells were washed twice in cold PBS, scraped, and centrifuged. The cell pellet was resuspended in 100 μl of lysis buffer composed of 62.5 mM Tris (pH 6.8), 2% SDS, 10% glycerol, 2% β-mercaptoethanol, and the proteinase inhibitors described above. After a 20-min incubation on ice, an equal volume of 2× loading buffer was added and the samples were placed in a 100°C water bath for 5 min. Thirty microliters of each sample were subjected to 8% SDS-PAGE as described above. For detection of PARP protein, a mouse anti-PARP monoclonal antibody (Pharminogen, San Diego, CA) was used at a 1:1,000 dilution, followed by a goat anti-mouse IgG conjugated with horseradish peroxidase (Transduction Laboratories, Lexington, KY) at a 1:10,000 dilution.

Statistical analysis.

All numerical results are reported as means ± SE and represent data from a minimum of three independent experiments.


Coadministration of toxin converts the RALA hepatocyte TNF-α response from proliferation to cytotoxicity.

The effects of treatment with TNF-α alone and in combination with the toxin ActD were determined. Treatment with 10–50 ng/ml TNF-α for 24 h was nontoxic to RALA hepatocytes by microscopic examination and MTT assay. TNF-α alone did stimulate a proliferative cellular response, as determined by [3H]thymidine incorporation. A 24-h treatment with 10 ng/ml TNF-α increased [3H]thymidine incorporation to 237% of control levels (639 ± 61 cpm/μg DNA for control vs. 1516 ± 191 cpm/μg DNA for TNF-α treated cells). Cell number also increased slightly at 24 h despite the use of confluent cultures (111 ± 2.0% of control). Treatment with ActD alone caused 14 ± 3.1% cell death, whereas combined ActD-TNF-α administration increased cell death to 32 ± 2.5%. Cell death resulted from apoptosis, as determined by light and fluorescence microscopy (data not shown). Thus, consistent with in vivo studies, stimulation with TNF-α alone elicited a proliferative response, whereas TNF-α in combination with a toxin caused increased cell death.

ActD does not affect NF-κB activation but inhibits NF-κB-regulated gene expression.

The finding that resistance to TNF-α cytotoxicity in nonhepatic cells is mediated by NF-κB activation (2, 25, 31, 32) suggested that a potential mechanism by which ActD could sensitize hepatocytes to TNF-α-induced apoptosis is through the inhibition of expression of an NF-κB-dependent protective gene. To examine whether ActD directly interfered with NF-κB activation, the effects of ActD on TNF-α-induced increases in NF-κB DNA binding were determined. DNA gel shifts demonstrated that protein binding to an NF-κB consensus oligonucleotide was equivalent in cells treated with either TNF-α alone or ActD-TNF-α at a variety of time points (Fig.1). The NF-κB complex activated by TNF-α treatment of RALA hepatocytes was composed of p50-p65 dimers, as determined by supershifts (Fig.2).

Fig. 1.

Actinomycin D (ActD) treatment did not affect tumor necrosis factor-α (TNF-α)-induced nuclear factor-κB (NF-κB) activation. Nuclear extracts were isolated at times indicated from untreated RALA hepatocytes or cells that were treated with TNF-α alone or ActD + TNF-α. Extracts were then used for electrophoretic mobility shift assays with an NF-κB consensus oligonucleotide, as described inmethods. Solid arrow, NF-κB binding complex; open arrow, free probe.

Fig. 2.

Activated NF-κB was composed of p50-p65 dimers. Nuclear extract was isolated from RALA hepatocytes treated with TNF-α for 4 h. After incubation with radiolabeled probe, extract was further incubated in absence or presence of anti-p50 NF-κB, anti-p65 NF-κB, and anti-Stat3 antibodies (Ab) as indicated. Thin solid arrows, appropriate supershifted complexes; thick solid arrow, NF-κB-binding complex; open arrow, free probe.

Although ActD failed to affect NF-κB activation, ActD inhibits RNA synthesis and may therefore act directly to block increases in gene expression that normally occur from TNF-α stimulated NF-κB activation. To determine whether ActD cotreatment inhibited TNF-α-induced NF-κB dependent gene expression, RALA hepatocytes were transiently transfected with NF-κB-Luc, a luciferase reporter gene driven by an NF-κB binding site in the promoter. At the onset of cell death 8 h after TNF-α treatment, luciferase expression was increased to 252 ± 15% of control levels in cells treated with TNF-α alone but only to 142 ± 6% of control in cells treated with ActD-TNF-α. Thus, although ActD did not interfere with NF-κB activation, this toxin did inhibit NF-κB-regulated gene expression, potentially sensitizing RALA hepatocytes to TNF-α-induced cell death by preventing upregulation of NF-κB-controlled protective genes.

Inhibition of NF-κB activation sensitizes RALA hepatocytes to TNF-α cytotoxicity.

To assess whether the failure to increase NF-κB-regulated gene expression sensitized hepatocytes to TNF-α toxicity, the effect of blocking NF-κB activation during TNF-α stimulation was determined. NF-κB activation was blocked by infecting cells with an adenovirus expressing a mutant IκB (Ad5IκB). This mutant IκB lacks the normal IκB phosphorylation sites so that it irreversibly binds NF-κB, preventing its activation. Although TNF-α treatment led to high levels of NF-κB activation in both uninfected cells and cells infected with a control adenovirus containing the β-galactosidase gene (Ad5LacZ), NF-κB DNA binding activity was low in cells infected with Ad5IκB (Fig. 3). Inhibition of NF-κB activation in this fashion sensitized RALA hepatocytes to rapid cell death from TNF-α treatment in the absence of ActD. Only 6 h after TNF-α administration, there was 50 ± 1.9% cell death in Ad5IκB-infected cells compared with only 5 ± 0.9% death in Ad5LacZ-infected controls (Fig.4). After 24 h of treatment, cell survival was still markedly decreased in Ad5IκB-infected cells (Fig.4). Fluorescent staining to determine the mode of cell death demonstrated that Ad5IκB-infected cells underwent rapid apoptosis following TNF-α treatment (Fig.5). Treated cells were examined for the presence of PARP cleavage as an additional indication of apoptosis. PARP cleavage occurred after TNF-α treatment in Ad5IκB-infected cells but not in uninfected or Ad5LacZ-infected control cells (Fig.6). Inhibition of NFκB-dependent gene expression was therefore sufficient by itself to sensitize RALA hepatocytes to TNF-α cytotoxicity.

Fig. 3.

Infection with Ad5IκB blocked TNF-α-induced NF-κB activation. Nuclear extracts were isolated from uninfected RALA hepatocytes (no virus) or cells infected with Ad5LacZ or Ad5IκB following 4 h of treatment with TNF-α. Solid arrow, NF-κB binding complex; open arrow, free probe.

Fig. 4.

Inhibition of NF-κB activation sensitized RALA hepatocytes to TNF-α-induced cell death. Cells were uninfected (no virus) or infected with Ad5LacZ or Ad5IκB as described inmethods. Cells were then treated with TNF-α, and percent cell survival relative to untreated cells was determined at 6 h (solid bars) and 24 h (hatched bars) by MTT assay. Data are from 4 independent experiments.

Fig. 5.

TNF-α induced apoptosis in Ad5IκB-infected cells. RALA hepatocytes infected with Ad5LacZ (•) or Ad5IκB (○) were treated with TNF-α, and numbers of apoptotic cells were determined at subsequent times by fluorescent costaining with acridine orange and ethidium bromide. Results are from 3 independent experiments.

Fig. 6.

TNF-α induced poly(ADP-ribose) polymerase (PARP) cleavage in Ad5IκB-infected cells. RALA hepatocytes that were uninfected or infected with Ad5LacZ or Ad5IκB were untreated or treated with TNF-α for 6 h as indicated. Aliquots of cell lysates were subjected to SDS-PAGE and immunoblotting with an anti-PARP antibody as described in methods. Intact 116-kDa and cleaved 85-kDa proteins are indicated.

NF-κB activation is a common cellular response to an environmental stress, and the upregulation of NF-κB-dependent gene expression may therefore regulate a common, final pathway inhibiting cell death from apoptosis. The role of NF-κB activation in other forms of hepatocyte apoptosis was therefore examined. The reactive oxygen intermediate H2O2and the metal Cu both induced NF-κB activation and hepatocyte death from apoptosis (data not shown). However, cells infected with either Ad5LacZ or Ad5IκB underwent equivalent amounts of cell death after exposure to toxic levels of H2O2or Cu. Infection with Ad5IκB also failed to sensitize RALA hepatocytes to cell death from nontoxic concentrations of H2O2or Cu, indicating that NF-κB activation did not play a protective role in these forms of apoptosis (Czaja, unpublished data). NF-κB activation therefore triggers a protective response that is specific for TNF-α-induced apoptosis.

NF-κB inactivation during TNF-α stimulation results in caspase activation.

TNF-α-induced cytotoxicity occurs through the mechanism of caspase activation (38). To determine whether NF-κB inactivation sensitized cells to TNF-α-induced cell death in a caspase-dependent manner, the ability of caspase inhibitors to block this cell death was examined. TNF-α-induced cell death in Ad5IκB-infected cells was inhibited by the conventional caspase inhibitors YVAD-CMK and DEVD-CHO, which target but are not specific for caspase-1 and caspase-3, respectively (Table1). Cell death was even more effectively inhibited by the experimental caspase inhibitors IDN-1529, which has additional activity against caspase-6 and caspase-8, and IDN-1965, a specific inhibitor of caspase-6 and caspase-8. (Table 1).

View this table:
Table 1.

Effects of caspase inhibitors on TNF-α-induced cell death at 6 h in Ad5IκB-infected RALA hepatocytes

The ability of caspase inhibitors to block TNF-α cytotoxicity in RALA hepatocytes sensitized by NF-κB inactivation suggested that caspase activation occurred in these cells. The effects of blocking NF-κB activation on caspase-1, caspase-2, caspase-3, and caspase-8 activation were determined in Ad5LacZ- and Ad5IκB-infected cells. Caspase-1 was not detected in RALA hepatocytes by Western blot analysis, although the anti-caspase-1 antibody employed did detect this protein in cardiac myocytes (data not shown). Caspase-2, caspase-3, and caspase-8 were present in RALA hepatocytes, and Ad5LacZ-infected cells showed no evidence of caspase-2, caspase-3, or caspase-8 activation 4 and 6 h after TNF-α administration (Fig.7). Although untreated Ad5IκB-infected cells had levels of unprocessed caspase-2, caspase-3, and caspase-8 similar to those of Ad5LacZ-infected cells, TNF-α treatment of Ad5IκB-infected cells led to activation of all three caspases, as reflected in decreased levels of the proenzymes (Fig. 7). The presence of the small, processed caspase subunits was difficult to detect in these cells, perhaps due to secondary proteolysis. A time course of caspase activation revealed caspase-2, caspase-3, and caspase-8 processing in Ad5IκB-infected cells within 2–4 h after TNF-α treatment (Fig. 8). The inhibition of NF-κB activation following TNF-α stimulation led therefore to caspase activation, causing apoptotic cell death.

Fig. 7.

TNF-α treatment induced caspase-2, caspase-3, and caspase-8 activation in Ad5IκB-infected cells. RALA hepatocytes were infected with Ad5LacZ or Ad5IκB and left untreated or treated with TNF-α for 4 or 6 h. Aliquots of total cell lysates were subjected to SDS-PAGE, and immunoblotting was performed using anti-caspase-2, anti-caspase-3, and anti-caspase-8 antibodies as described inmethods.

Fig. 8.

Time course of caspase-2, caspase-3, and caspase-8 activation. RALA hepatocytes were infected with Ad5LacZ or Ad5IκB and treated with TNF-α for indicated times. Immunoblotting was performed with anti-caspase-2, anti-caspase-3, and anti-caspase-8 antibodies as previously described.

NO prevents TNF-α-induced cell death in the absence of NF-κB activation.

As a transcriptional regulator, NF-κB presumably mediates hepatocellular resistance to TNF-α toxicity by increasing expression of protective cellular genes. The gene product iNOS is TNF-α inducible and is regulated by NF-κB, and NO, the product of iNOS, has been implicated as a protective factor against TNF-α toxicity in several cell types, including hepatocytes (14, 22). Thus iNOS could be the NF-κB-inducible protective gene mediating resistance to TNF-α toxicity. The ability of NO to replace the protective function of NF-κB activation was therefore determined. Ad5IκB-infected cells were pretreated for 18 or 2 h before TNF-α administration with the NO donor SNAP. Treatment with 750 μM SNAP led to a 56.7 ± 10.8% inhibition of cell death after an 18-h pretreatment and to an 85.0 ± 3.0% inhibition with a 2-h pretreatment. The protective effects of SNAP were dose dependent (data not shown).

SNAP inhibited cell death at 6 h even when administered 2 h after TNF-α treatment or when coadministered with the protein synthesis inhibitor cycloheximide (data not shown). Administration of 5 mM cGMP, a common intracellular mediator of the effects of NO, had no protective effect against TNF-α-induced cell death. These data suggested that protection by NO was independent of protein synthesis and was mediated at a posttranslational level. An examination of the effects of NO on caspase activation in Ad5IκB-infected cells demonstrated that SNAP treatment was as effective as caspase inhibitors in blocking caspase-2, caspase-3, and caspase-8 activation in TNF-α-treated cells (Fig. 9).

Fig. 9.

S-nitroso-N-acetylpenicillamine (SNAP) treatment inhibited caspase activation. RALA hepatocytes infected with Ad5IκB were untreated or treated with TNF-α for 4 h. Some cells were pretreated with caspase inhibitors IDN-1529 and IDN-1965 or with SNAP. Cell lysates were subjected to immunoblotting with anti-caspase-2, anti-caspase-3, and anti-caspase-8 antibodies.

If iNOS is the gene mediating resistance to TNF-α toxicity, then inhibiting endogenous NO production in RALA hepatocytes should sensitize these cells to toxicity from TNF-α alone. Cells were pretreated with two iNOS inhibitors at a variety of concentrations before TNF-α or ActD-TNF-α treatment. The administration of the iNOS inhibitors l-NIL andl-NMMA had no effect on TNF-α or ActD-TNF-α toxicity (data not shown). Thus exogenous NO is sufficient to prevent TNF-α-induced cytotoxicity, but endogenous NO is not the critical protective factor mediating RALA hepatocyte TNF-α resistance.


Inhibition of NF-κB activation sensitizes a variety of cell types to cytotoxicity from TNF-α (2, 25, 31, 32). Increased expression of an NF-κB-regulated gene(s) may serve as the central control point in determining whether the cellular response to TNF-α is survival or apoptotic cell death (38). This finding also suggests the possibility that the ability of toxins to sensitize normal cells to TNF-α cytotoxicity may be mediated through an inhibition of either NF-κB activation or upregulation of specific NF-κB-dependent cellular genes.

Our investigations examined the relationship of NF-κB activation to the toxin-induced sensitization of a hepatocyte cell line to TNF-α cytotoxicity because hepatocytes are known to be sensitized to TNF-α killing by toxins in vivo (4, 11, 18). Consistent with the hepatic effects of TNF-α in vivo (1, 4, 11, 18, 35), the present in vitro studies demonstrated that TNF-α alone stimulated RALA hepatocyte proliferation in the absence of cell death, whereas the coadministration of ActD and TNF-α led to cytotoxicity. This conversion from a proliferative to a cytotoxic response was not the result of toxin-mediated inhibition of NF-κB activation, since NF-κB DNA binding was unaffected by the addition of ActD. ActD did partially inhibit the upregulation of NF-κB-dependent gene expression, potentially sensitizing RALA hepatocytes to TNF-α toxicity by preventing expression of a protective gene.

NF-κB activation was essential for resistance to TNF-α toxicity, since blocking activation with an IκB superrepressor rapidly sensitized RALA cells to TNF-α-induced cell death in the absence of ActD. Previous in vivo investigations demonstrated that blocking the NF-κB activation that normally occurs in rats following partial hepatectomy converts a hepatocellular proliferative response into one of cell death, but the factor mediating this effect was not identified (19). The present in vitro studies demonstrate that it is TNF-α that acts as a hepatocyte mitogen in the presence of NF-κB activation but as a cytotoxin in the absence of NF-κB activation. Although our data cannot completely exclude concurrent proliferative and toxic effects, they demonstrate for the first time that selective inhibition of NF-κB activation is sufficient to convert a cellular TNF-α response from proliferation to apoptosis. Despite the multitude of complex cellular signaling events triggered by TNF-α stimulation, a single transcription factor ultimately determines whether the cell has the diametrically opposed responses of growth or cell death.

Previous studies in nonhepatic cells relating NF-κB activation to cellular resistance to TNF-α cytotoxicity did not address the mechanism by which NF-κB inactivation led to cell death from TNF-α. The present study demonstrates that the mechanism involves activation of the caspase family of cysteine proteases that are the effectors of apoptotic cell death (27). TNF-α-induced cell death mediated by NF-κB inactivation was blocked by caspase inhibitors. In addition, caspase-2, caspase-3, and caspase-8 were activated following TNF-α treatment in Ad5IκB-infected cells that underwent cell death but not in resistant cells infected with Ad5LacZ. Thus TNF-α stimulation in the absence of NF-κB activation led to caspase activation and apoptosis. In contrast to a recent report in a mouse hepatocyte cell line (3), infection with Ad5IκB in the absence of TNF-α administration did not result in caspase activation or cell death, suggesting that constitutive NF-κB activation does not play a role in preventing spontaneous hepatocellular apoptosis. The present data suggest that TNF-α-induced signaling must trigger caspase activation by a mechanism that is blocked when NF-κB activation upregulates a protective cellular factor. This protective NF-κB-regulated gene product may act directly on caspase-8, preventing its activation and subsequent processing of downstream caspase-2 and caspase-3. Alternatively, TNF-α can directly activate caspase-2 through the adaptor protein RAIDD/CRADD, and NF-κB expression may block this initial activation of caspase-2 and the subsequent processing of caspase-3 and caspase-8. Finally, an NF-κB-regulated gene product may directly affect a cellular process upstream of caspases in the cell death pathway, indirectly preventing caspase activation. For example, recent investigations suggest that NF-κB may regulate the inhibitor of apoptosis protein (IAP) family (7).

In the present investigations, NO inhibited cell death induced by TNF-α in the absence of NF-κB activation. NO is known to have both pro- and antiapoptotic properties (reviewed in Ref. 15). NO has been reported previously to inhibit ActD-TNF-α-induced cell death (14,22). Kim et al. (22) demonstrated in rat primary hepatocytes that ActD-TNF-α-induced cell death was blocked by NO administration. The protective effect of NO in their study was dependent on a long pretreatment time (12–18 h) and protein synthesis and was linked to the induction of heat shock protein 70 expression (22). In our investigations employing an identical SNAP concentration, the protective effects of NO were independent of protein synthesis, present even when NO was given 2 h after TNF-α administration, and not reproduced by cGMP. These data suggest that the mechanism of NO protection in our studies was posttranslational. NO has been reported to directly inhibit caspase-3 activation byS-nitrosylation (14). NO prevented caspase-2 and caspase-3 activation and decreased caspase-8 activation in RALA hepatocytes, although our studies do not prove a direct effect of NO on these enzymes.

Inhibition of NO production did not sensitize RALA hepatocytes to cell death from TNF-α or ActD-TNF-α, demonstrating that endogenous NO does not modulate hepatocellular sensitivity to TNF-α toxicity. These data exclude iNOS as the NF-κB-dependent gene that mediates hepatocellular resistance to TNF-α cytotoxicity. These findings contrast with those of Kim et al. (22), who reported increased cell death in ActD-TNF-α-treated primary hepatocytes with iNOS inhibition. However, their studies actually demonstrated that additional stimulation of hepatocyte iNOS by the cytokines interferon-γ and interleukin-1β was protective against ActD-TNF-α-induced death. Their studies failed to show that iNOS inhibition caused cell death in hepatocytes treated with TNF-α alone or worsened cell death in cells treated with ActD-TNF-α. Hepatocyte iNOS expression and NO production are stimulated to much higher levels by a combination of cytokines than by TNF-α alone (17). Thus, although TNF-α alone may not stimulate sufficient NO to protect against cell death, exposure of hepatocytes to multiple cytokines in vivo may upregulate iNOS to sufficient levels to be cytoprotective. Alternatively, NO that is released in large quantities during injury by liver cells other than hepatocytes, such as Kupffer cells, may still modulate hepatocyte TNF-α toxicity. Consistent with this possibility are reports that inhibition of NO produced predominantly by Kupffer cells in vivo during toxin-induced liver injury increases hepatocyte cell death (5).

The identity of the NF-κB-regulated gene product(s) that blocks TNF-α-induced hepatocyte apoptosis remains to be determined. This protective gene must express a protein that is specific to the TNF-α death pathway. H2O2or Cu each induced NF-κB activation and apoptotic cell death, and H2O2-induced apoptosis was blocked by caspase inhibitors (Czaja, unpublished data). However, inhibition of NF-κB activation did not increase cell death induced by H2O2or Cu or sensitize cells to death from lower concentrations of these agents. Thus, although NF-κB activation is a common response to environmental stresses that result in hepatocyte apoptosis, NF-κB is not universally protective against all forms of caspase-dependent apoptosis.


We thank Amelia Bobe and Anna Caponigro for secretarial assistance, Dr. Joseph L. Goldstein for providing the anti-caspase-3 antibody, and Dr. Janice Chou for providing the RALA255-10G cells.


  • Address for reprint requests: M. J. Czaja, Marion Bessin Liver Research Center, Albert Einstein College of Medicine, 1300 Morris Park Ave., Bronx, NY 10461.

  • This work was supported in part by National Institutes of Health Grants DK-34987, GM-41804 (to D. A. Brenner), and DK-44234 (to M. J. Czaja), a grant-in-aid from the American Heart Association, New York City Affiliate (to R. N. Kitsis), a Howard Hughes predoctoral fellowship award (to S. Bialik), and an Australian National Health and Medical Research Council research scholarship (to B. E. Jones).

  • R. N. Kitsis is the Charles and Tamara Krasne Faculty Scholar in Cardiovascular Research of the Albert Einstein College of Medicine.

  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.


View Abstract