Activation of tumor necrosis factor receptor 1 or Fas leads to the generation of reactive oxygen species, which are important to the cytotoxic effects of tumor necrosis factor α (TNF-α) or Fas ligand. However, how these radicals are generated following receptor ligation is not clear. Using primary hepatocytes, we found that TNF-α or anti–Fas antibody–induced burst of oxygen radicals was mainly derived from the mitochondria. We discovered that Bid—a pro-death Bcl-2 family protein activated by ligated death receptors—was the main intracellular molecule signaling the generation of the radicals by targeting to the mitochondria and that the majority of oxygen radical production was dependent on Bid. Reactive oxygen species contributed to cell death and caspase activation by promoting FLICE-inhibitory protein degradation and mitochondrial release of cytochrome c. For the latter part, the oxygen radicals did not affect Bak oligomerization but instead promoted mitochondrial cristae reorganization and membrane lipid peroxidation. Antioxidants could reverse these changes and therefore protect against TNF-α or anti–Fas-induced apoptosis. In conclusion, our studies established the signaling pathway from death receptor engagement to oxygen radical generation and determined the mechanism by which reactive oxygen species contributed to hepatocyte apoptosis following death receptor activation. (HEPATOLOGY 2004;40:403–413.)
The pathogenesis of many human diseases, including liver injuries, involves inappropriate activation of apoptosis.1–3 A common apoptotic signal is mediated by the death receptors, tumor necrosis factor receptor 1 (TNF-R1) and Fas. Tumor necrosis factor α (TNF-α) is a cytokine produced by a variety of cells. Its expression is up-regulated in a number of stressful and pathological conditions that can lead to liver injury when coupled with other detrimental factors.3, 4 On the other hand, hepatocyte apoptosis and liver injury could be also caused by Fas activation induced by Fas ligand–bearing T cells or agonistic antibodies.2, 3
Although multiple death mechanisms can be activated by the death receptors, generation of reactive oxygen species (ROS) is a critical one,5–13 and antioxidants have been shown to alleviate cell death induced by TNF-α9, 14 or Fas7, 15 activation. Despite the importance of ROS, the mechanisms of their generation remain controversial and unresolved. Although the evidence for mitochondrial origin of ROS generation is strong,5, 9, 12, 14 extramitochondrial ROS production involving various cytosolic oxidative pathways has also been indicated.6, 10, 13 However, in none of these cases is the intracellular signaling pathway from the receptor engagement to ROS generation defined. Moreover, how ROS affect the cell death program has yet to be clearly determined.
The Bcl-2 family proteins play important roles in apoptosis regulation by targeting the mitochondria.16 They can regulate the mitochondrial release of apoptogenic factors such as cytochrome c; in addition, Bcl-2 and Bcl-xL can function as antioxidants to exert their antiapoptotic activities.12, 17–18 On the other hand, the pro-death Bcl-2 molecules may work as pro-oxidants. Overexpression of Bax caused an enhanced production of superoxide radicals.19 Deletion of Bax prevented ROS burst and apoptosis in nerve growth factor–deprived sympathetic neurons.20 Bid, a BH3-only member of the Bcl-2 protein family, is a substrate of caspase 8 and is responsible for the mitochondrial activation following death receptor ligation.21 Bid connects the death receptor apoptosis pathway to the mitochondria apoptosis pathway, which is critical to the hepatotoxicity of TNF-α or Fas agonist.22–25 Thus it is possible that Bid may also mediate TNF-R1/Fas-induced ROS generation, which in turn contributes to Bid-induced apoptotic events.
The present study examined this hypothesis and determined the mechanism of ROS generation and ROS contribution to apoptosis in TNF-R1/Fas-stimulated hepatocytes. We show that Bid is responsible for the majority of ROS production following death receptor activation and that ROS are important to caspase activation and cell death by promoting FLICE-inhibitory protein (FLIP) degradation, mitochondrial cristae reorganization, membrane lipid peroxidation, and cytochrome c release.
Wild-type and bid-deficient mice were maintained in a C57BL/6 background. All animals received humane care according to the guidelines of the National Institutes of Health and the University of Pittsburgh.
Primary Hepatocyte Cultures and Analysis for Apoptotic and Necrotic Death.
As described previously,25 murine hepatocytes were isolated by a retrograde, non-recirculating perfusion of livers with 0.05% Collagenase H (Roche Molecular Biochemicals, Indianapolis, IN). Cells were cultured in William's medium E with 10% fetal bovine serum but no other supplements for 2 hours for attachment. Cells were then cultured in the same medium without serum overnight before treatment. Analyses were conducted at various time points after the treatment. Apoptotic and necrotic death were determined as previously described26 with modifications. Briefly, cells were costained with Hoechst 33342 (5 μg/mL) and propidium iodide (1 μg/mL) for 10 minutes and analyzed via digital microscopy. Cells stained with Hoechst 33342 showing condensed and/or fragmented nuclei were considered apoptotic, whereas cells stained with propidium iodide but showing no signs of nuclear condensation and/or fragmentation were considered necrotic. Viable cells were propidium iodide–negative and did not show apoptotic nuclear changes by Hoechst staining.
Measurement of ROS Production.
Superoxide anions (O) were detected as previously defined.5 Briefly, treated cells were incubated with dihydroethidium (2.5 μM) for 30 minutes at 37°C. (Dihydroethidium is converted to a red fluorescent product, ethidium, in the presence of O2−.5) The cells were washed and resuspended in phosphate-buffered saline for flow cytometry. For hydrogen peroxide (H2O2) detection, cells were loaded with 2′,7′-dichlorodihydrofluoresein diacetate (DCFH-DA) (2.5 μM), which is cleaved by intracellular esterases to become dichlorodihydrofluoresein and further oxidized to dichlorofluorescein (DCF) by H2O2.9, 12 Cells were counterstained with Hoechst 33342 and images were taken under a fluorescence microscope (Nikon Eclipse TE 200, Melville, NY). For quantification, cells were lysed using 0.1% Triton X-100, and the fluorescence intensity of formed DCF was determined with a fluorometer (Tecan GENios, Phenix Research Products, Hayward, CA) at excitation and emission wavelengths of 488 nm and 535 nm, respectively.
Analysis of Cytochrome c Release, Bak Oligomerization, and Mitochondria Morphology.
Immunostaining and immunoblotting analyses of cytochrome c release in cultured cells were conducted as previously described.25, 28 Procedures of liver mitochondria isolation and in vitro induction of mitochondrial cytochrome c release were performed as previously described.27, 28 Bak oligomerization was determined via chemical cross-linking followed by immunoblotting with an anti-Bak antibody (Upstate Biotechnology, Lake Placid, NY).25, 28 Electron microscopy was performed as described29 except that the mitochondria treatment was conducted in 250 mM sucrose buffer with 4 mM MgCl2.27
Analysis of Caspase Activation and Activities.
As previously described,25 caspase activation was determined via immunoblotting with antibodies against caspase 8 (Dr. Wim Declercq, University of Ghent, Belgium), caspase 9 or caspase 3 (Cell Signaling, Beverly, MA). Caspase activities were measured using 30 μg of proteins and 20 μM of fluorescent substrates (Ac-DEVD-AFC, AC-IETD-AFC, and Ac-LEHD-AFC for caspase 3, 8, and 9, respectively). The fluorescence signals were detected by a fluorometer (Tecan GENios) at excitation and emission wavelengths of 400 nm and 510 nm, respectively.
Measurement of Lipid Peroxidation.
Malondiadehyde (MDA), a common end product of lipid peroxidation, was determined with an assay kit (Lipid Peroxidation Kit, Oxford Biomedical Research, Oxford, MI). Materials were prepared according to the manufacturer's instructions. The amount of MDA was quantified with the use of MDA standards provided in the kit.
ROS Play an Important Role in TNF-α or Anti–Fas-Induced Apoptosis.
To determine the contribution of ROS to the cytotoxicity of TNF-α or anti-Fas in hepatocytes, we first examined whether or not two antioxidants—MnTBAP, a potent superoxide dismutase mimetic,30 and Trolox, a water-soluble vitamin E analogue31—could provide any protection for the treated hepatocytes. Death receptor activation–induced hepatocyte death was noticeable in 6 hours, and most of the cells died from apoptosis (Fig. 1A) but not necrosis (Fig. 1B). The antioxidants could significantly suppress the apoptosis without leading to extra necrosis in the effective time period (24–36 hours). In contrast, a pan-caspase inhibitor, z-VAD-fmk, suppressed apoptotic death for a longer duration; however, cells died from necrosis at the later stage (>36 hours). The protective effects of MnTBAP and Trolox were dose-dependent (Fig. 1C), and MnTBAP was more potent than Trolox, consistent with its stronger ability to scavenge ROS and to inhibit caspase activation (see below). In addition, the antioxidants inhibited TNF-α and anti–Fas-induced death to a similar extent (Fig. 1D). These results indicate the significance of ROS in promoting hepatocyte apoptosis following death receptor activation.
TNF-α and Anti–Fas-Induced ROS Production in Hepatocytes Is Dependent on Bid.
We then examined ROS production in hepatocytes more directly using dihydroethidium, a superoxide anion (O2−) probe, and flow cytometry. We found that both TNF-α and anti-Fas significantly induced ROS production in wild-type hepatocytes in 3 hours, which peaked by 6 hours (Fig. 2A). Superoxide anions are converted to H2O2, which can be detected with the use of DCFH-DA. The reactive product of the dye, DCF, was detected in TNF-α–treated wild-type cells (Fig. 2B). MnTBAP and Trolox could effectively scavenge ROS in treated hepatocytes (Fig. 2C).
ROS production at the peak time (6 hours) was almost completely inhibited in bid-deficient hepatocytes following death receptor activation (Fig. 2B and 2D), indicating that the majority of ROS were produced in a Bid-dependent manner. A prolonged treatment of bid-deficient cells with TNF-α for 12 hours led to a marginal increase in ROS generation (Fig. 2E), suggesting the presence of a minor Bid-independent mechanism. These data were consistent with the findings that anti-Fas or TNF-α–induced hepatocyte apoptosis was significantly contributed by the Bid-dependent mechanism (see Fig. 1D)25 and that antioxidants were able to considerably improve hepatocyte viability (see Fig. 1). z-VAD-fmk also dramatically suppressed TNF-α–induced ROS production as previously reported in HeLa cells,9 consistent with its ability to block caspase 8–induced Bid activation25 and inhibit cell death (see Fig. 1).
Because there is much less ROS production in bid-deficient hepatocytes, and Bid works on the mitochondria, it is likely that TNF-α and anti–Fas-induced ROS are mainly derived from the mitochondria. Indeed, the dramatically increased H2O2-sensitive DCF fluorescence in treated hepatocytes was distributed in a pattern consistent with that of mitochondria (Fig. 2F), and the DCF signal was overlapped with that of MitoTracker (Molecular Probes, Eugene, OR) (Fig. 2G). In contrast, there were no significant increases in DCF fluorescence in the mitochondria of TNF-α–treated bid-deficient cells (see Fig. 2F), which is consistent with the lack of mitochondrial activation in these cells following death receptor activation.22–25
Antioxidants Inhibit TNF-α or Anti–Fas-Induced Caspase Activation.
We next determined how ROS could promote cell death by examining whether or not the antioxidants could affect TNF-α or anti–Fas-induced caspase activation. The effector caspase 3 activity rose rapidly in treated wild-type cells and could be detected in 3 hours after treatment (Fig. 3A). It reached peak around 6 hours after anti-Fas treatment and around 12 hours after TNF-α treatment, consistent with previous findings.25 ROS generation had similar kinetics (see Fig. 2A), thus they could contribute to caspase 3 activation. Indeed, both MnTBAP and Trolox could significantly suppress caspase 3 activity in a dose-dependent manner (Fig. 3A and 3B), in parallel to their ability to suppress cell death (see Fig. 1A).
Caspase activation was largely dependent on Bid, and the already low caspase activities in bid-deficient cells could be further reduced by MnTBAP, suggesting that both Bid-dependent ROS and Bid-independent ROS could contribute to caspase activation (Fig. 3C–E).
Caspases are present in inactivate zymogen forms in healthy cells. Activated caspases can undergo proteolysis, generating subunits. Immunoblot analysis of caspase 3, 8, and 9 indicated that the cleavage of the zymogen form of these caspases following death receptor ligation was significantly inhibited by the antioxidants (Fig. 4A). These data suggest that antioxidants could inhibit ROS-promoted caspase activation. Consistent with this notion, antioxidants did not seem to have any effects on the activities of already activated caspases, as indicated by the ability of recombinant caspase 8 to cleave caspase 3 or Bid in vitro in the presence of MnTBAP or Trolox (Fig. 4B). Nonetheless, they were able to significantly suppress such cleavage in vivo by blocking caspase 8 activation (Fig. 4A and 4C).
We then examined how ROS could affect caspase 8 activation. Caspase 8 is mainly activated by the ligated death receptors,32 which can be impeded by a negative regulator—FLIP.33 We found that TNF-α treatment led to the degradation of FLIP, which seemed to be mediated by ROS, because both MnTBAP and Trolox could reverse this process (Fig. 4D). ROS-mediated FLIP degradation could thus significantly contribute to caspase 8 activation. Interestingly, FLIP degradation was not observed in the absence of Bid (see Fig. 4D), consistent with the observation that most ROS were generated in a Bid-dependent way in TNF-α–treated hepatocytes (see Fig. 2D). Therefore, it seemed that ROS-promoted caspase 8 activation was mainly resulted from a positive feedback mechanism dependent on Bid activation—although it is not likely that this would be the only mechanism, because there was also Bid-independent ROS generation. Consistently, caspase 8 activity was much lower in bid-deficient cells than in the wild-type cells following death receptor activation (see Fig. 3D).22–25 Such a feedback amplification of caspase 8 activation led to more Bid cleavage (see Fig. 4C) and therefore a stronger apoptotic response, including more ROS generation.
Antioxidants Inhibit ROS-Promoted Cytochrome c Release.
ROS also promoted the activation of caspase 9, another initiator caspase whose activation depends on Apaf-1 and the mitochondrial-released cytochrome c.16 Previous work had shown that cytochrome c release following death receptor ligation was mainly dependent on Bid.22–25 We then investigated whether or not this process could be affected by ROS. Immunostaining assay indicated that the percentage of hepatocytes with a cytoplasmic distribution of—or a complete loss of—cytochrome c staining increased dramatically following TNF-α or anti-Fas treatment, which could be inhibited by the antioxidants (Fig. 5A). MnTBAP provided more sustained and more potent inhibitory effects on cytochrome c release than Trolox (Fig. 5B and data not shown). The results of immunostaining (see Fig. 5A) were further confirmed by immunoblot analysis (Fig. 5C), which also showed the association of cytochrome c release with the cleavage of 14-3-3ϵ, a caspase 3 substrate, both being affected by the antioxidants in parallel.
ROS Can Affect Mitochondrial Functions Directly.
The effect of ROS on cytochrome c release and the downstream caspase activation could be in part due to their effects on the activation of caspase 8 and Bid (see Fig. 4). However, ROS could also have a direct effect on mitochondria. To test this possibility, we examined whether the antioxidants could directly affect the ability of Bid to induce mitochondrial apoptotic events in vitro. As has been shown previously,27 recombinant tBid induced cytochrome c release from isolated murine liver mitochondria in a time-dependent manner (Fig. 6A). Both MnTBAP and Trolox could significantly suppress this process in a dose-dependent manner (see Fig. 6A), suggesting that they could possess a direct effect on Bid-induced mitochondrial alterations.
A number of mechanisms have been proposed for the mitochondrial release of cytochrome c induced by tBid, including Bak oligomerization.34 However, we did not find that the antioxidants were able to inhibit Bak oligomerization induced by tBid on isolated liver mitochondria (Fig. 6B). Another important mechanism in cytochrome c release is the alteration of the mitochondrial cristae structure, which can also be induced by Bid.29 We thus performed electron microscopic studies to examine the morphology of Bid-treated mitochondria in the absence or presence of antioxidants. The results indicated that while most untreated mitochondria had a condensed morphology, with the narrow electron-lucent cristae setting in a condensed electron-dense matrix (Fig. 6C), tBid-treated mitochondria displayed a series of morphological changes. Many mitochondria appeared to have undergone the so-called “cristae reorganization” as defined initially,29 manifesting widened and fused intracristae spaces and the “sausage-shaped” or “dot-shaped” matrix (see Fig. 6C). MnTBAP or Trolox could dramatically reduce this mitochondrial alteration, although not completely (Fig. 6C and 6D). Thus mitochondrial cristae reorganization could be exacerbated by ROS, which contributed to an enhanced release of cytochrome c.29
TNF-α Treatment Leads to Membrane Lipid Peroxidation.
One of the mechanisms by which ROS directly affect mitochondrial structures such as the cristae could be through its effect on membrane lipid peroxidation, which could lead to membrane structure alternation. To examine this possibility, we extracted total cellular lipids or mitochondrial lipids from TNF-α–treated hepatocytes for the presence of MDA, a common end product of lipid peroxidation. We found that peroxidation of both total cellular (Fig. 7A) and mitochondrial lipid (Fig. 7B) increased significantly in these cells following TNF-α treatment. In vitro incubation of isolated liver mitochondria with recombinant tBid also resulted in an approximately 30% increase of MDA formation (Fig. 7C). While Trolox consistently inhibited MDA formation to a significant degree, MnTBAP seemed to be less effective. This assay may be more sensitive for the assessment of the inhibitory effect of Trolox, because it is more potent against lipid peroxidation.31 It seemed that tBid-induced MDA formation in vitro was not as dramatic as that in TNF-α–treated hepatocytes, perhaps reflecting the lack of feedback amplification of ROS production and/or the lack of Bid-independent ROS production in the in vitro situation.
Previous studies of ROS generation in nonhepatocytes following death receptor activation indicated that mitochondria could be the major source of ROS. The evidence includes that ROS is produced in correlation with mitochondrial potential changes5, 11, 12 and that the inhibition of mitochondrial electron transport,14 the deletion of mitochondrial DNA,9 or the overexpression of Bcl-2 or Bcl-xL5, 12 could significantly suppress the production. How receptor engagement leads to mitochondrial generation of ROS is not known.
Using the physiologically relevant primary hepatocytes, we demonstrated that Bid, a pro-death Bcl-2 family protein, is the missing link between the receptor and the mitochondria (Fig. 8). Deletion of Bid eliminates the majority of ROS generation in hepatocytes treated with TNF-α or anti-Fas. Bid is proteolytically activated by caspase 8 following death receptor ligation, and tBid is translocated to the mitochondria to exert its activities.21–24 Inhibition of the initiator caspase 8 with z-VAD-fmk could thus also inhibit ROS generation (see Fig. 2).9 These studies provide further evidence that mitochondria is the major source of ROS and, more importantly, establish the signaling pathway from the receptor engagement to caspase 8 and Bid activation and then to the mitochondrial ROS generation. Although this seems to be the major pathway for ROS generation, a small amount of ROS could be produced in a Bid-independent way, as revealed in the bid-deficient cells. It is not clear how these ROS are generated; they could be derived from mitochondrial or extramitochondrial processes. It is possible that these mechanisms may involve the phospholipase A2-arachidonic acid-5-lipoxygenase pathway, the nicotinamide adenine dinucleotide phosphate oxidase pathway, or the ceramides as proposed previously.6, 8, 10, 13
How does Bid induce mitochondrial ROS generation? There are several possible mechanisms (see Fig. 8). Bid-induced cytochrome c release could lead to disturbance of respiration and the accumulation of single electrons at or upstream of complex III, which react with oxygen to form superoxide anions. Another mechanism involves Bid-dependent activation of effector caspases,22–24 which can further disrupt mitochondrial respiration at complex I or II in a feedback mode, promoting ROS generation.35 Finally, Bid could induce mitochondrial permeability transition,25, 36 which is capable of amplifying ROS production but could also be involved in the initial ROS generation.16, 37 In this scenario, ROS generation would be upstream of cytochrome c release.
Overproduced ROS are clearly detrimental to cells.38 In the case of TNF-α or anti–Fas-induced hepatocyte apoptosis and liver injury, both in vitro (this study) and in vivo15 use of antioxidants inhibited cell death, reduced liver injury and improved animal survival. Conceivably, mitochondria themselves could be the major target. Mitochondrial membranes are susceptible to ROS-mediated lipid peroxidation (see Fig. 7). Cardiolipin, a special type of phospholipid present only in the mitochondria, contains a large number of unsaturated fatty acid chains and is associated with many respiratory proteins, making it prone to oxidation.39 Peroxidation of cardiolipin has been suggested to facilitate mitochondrial release of cytochrome c.40–42 Our data that ROS can promote membrane lipid peroxidation, mitochondrial cristae reorganization, and cytochrome c release thus suggest that there might be a mechanistic relationship among these events (see Fig. 8). In addition, although ROS does not seem to affect the oligomerization of Bak induced by tBid, ROS could promote Bax translocation to the mitochondria (data not shown), which is known to participate in death receptor activation–induced liver injury.23
A major extramitochondrial target of TNF-α and anti–Fas-induced ROS seems to be FLIP, which could be rapidly degraded in response to ROS in a Bid-dependent way. This would greatly increase caspase 8 activation and therefore function as a mechanism of feedback amplification (see Fig. 8). Bid-dependent enhancement of caspase 8 activation in this setting has been consistently observed (see Fig. 2).22–25 In addition to FLIP degradation, Bid-mediated effector caspase activation could also promote caspase 8 activation in a feedback loop.44, 45 It should be noted that our data do not exclude the possibility that some ROS, particularly those generated in Bid-independent way, can promote caspase 8 activation independently of Bid.
Superoxide anions (O2−) are likely the first oxygen radicals generated in the mitochondria following death receptor ligation,5, 14 which can affect mitochondria directly or other cellular targets indirectly after conversion to H2O2 and then hydroxyl radical (OH−) and singlet oxygen.38 MnTBAP belongs to a new generation of metalloporphyrin-based superoxide dismutase mimetic.30 It can effectively scavenge many types of ROS, including O2−, H2O2, ONOO−, NO−, and lipid peroxyl radicals.30 On the other hand, Trolox is a water-soluble vitamin E homologue that also has broad antioxidant effects, particularly in inhibiting lipid peroxidation.31 It is effective but not as potent as MnTBAP in most of the assays in this study, except for lipid peroxidation, perhaps reflecting that O is the most important ROS here, which is most effectively scavenged by MnTBAP. Previous studies had shown that depletion of glutathione sensitized hepatocytes to TNF-α killing26, 46 and thus demonstrated the contribution of peroxides in this setting. However, supplemental glutathione46 or overexpression of catalase47 did not inhibit TNF-α killing, indicating the involvement of other ROS species such as O2−, as supported by this study.
Although the current study indicates the importance of ROS in death receptor–mediated hepatocyte apoptosis, these findings need to be further verified in in vivo systems, because the culture system may be prone to oxidative stress.48 From this aspect, it is interesting to note that mice deficient in glutathione peroxidase 1 did not show an obvious increase of hepatocyte apoptosis following death receptor activation,49 suggesting that ROS, particularly peroxides, may not be a critical contributing factor in vivo. On the other hand, in vivo administration of MnTBAP has been shown to improve the survival of anti–Fas-treated mice and reduce liver injury,15 again suggesting the importance of O2−. Thus different approaches are required to determine ROS involvement in hepatocyte death.
In conclusion, we have demonstrated that Bid is the key molecule mediating ROS generation following death receptor activation and that the majority of ROS are likely derived from mitochondria following Bid activation. We have also found that the major effects of ROS are to reduce FLIP levels and promote cytochrome c release and caspase activation by enhancing mitochondrial cristae reorganization. These targets could therefore be therapeutically explored to alleviate TNF-α or Fas-induced hepatocyte toxicity, commonly seen in a variety of liver diseases.
The authors would like to thank Drs. W. Declercq and P. Vandenabeele (University of Ghent, Belgium) for the anti–caspase 8 antibody.