Resistance to the Cytotoxic Effects of Tumor Necrosis Factor α Can Be Overcome by Inhibition of a FADD/Caspase-dependent Signaling Pathway*

Tumor necrosis factor (TNF) α initiates the activation of a pro-apoptotic pathway involving the recruitment of the death domain containing protein FADD and the subsequent activation of specific proteases (caspases). Many cells are resistant, however, to the cytotoxic effects of TNFα due to the concurrent activation of pro-survival pathways involving the transcription factor NFκB and TRAF2. Here we show that the TNFα-activated FADD/caspase pathway can also exert an unexpected pro-survival effect. Inhibition of this pathway in NIH3T3 fibroblasts or U937 leukemic cells by peptide caspase inhibitors or expression of dominant-negative FADD leads to rapid death following treatment with TNFα, whereas control cells are TNFα-resistant. FADD/caspase-inhibited cells die by a non-apoptotic mechanism caused by increased production of reactive oxygen species which precedes loss of the mitochondrial membrane potential. Cytotoxicity can be prevented by preincubation with antioxidants including reduced glutathione or by expression of a dominant-negative Rac GTP-binding protein. These results indicate that caspase activation in response to TNFα has anti-necrotic as well as pro-apoptotic effects and extend our understanding of the biological role of these proteases.

Tumor necrosis factor (TNF) ␣ initiates the activation of a pro-apoptotic pathway involving the recruitment of the death domain containing protein FADD and the subsequent activation of specific proteases (caspases). Many cells are resistant, however, to the cytotoxic effects of TNF␣ due to the concurrent activation of prosurvival pathways involving the transcription factor NFB and TRAF2. Here we show that the TNF␣-activated FADD/caspase pathway can also exert an unexpected pro-survival effect. Inhibition of this pathway in NIH3T3 fibroblasts or U937 leukemic cells by peptide caspase inhibitors or expression of dominant-negative FADD leads to rapid death following treatment with TNF␣, whereas control cells are TNF␣-resistant. FADD/ caspase-inhibited cells die by a non-apoptotic mechanism caused by increased production of reactive oxygen species which precedes loss of the mitochondrial membrane potential. Cytotoxicity can be prevented by preincubation with antioxidants including reduced glutathione or by expression of a dominant-negative Rac GTP-binding protein. These results indicate that caspase activation in response to TNF␣ has anti-necrotic as well as pro-apoptotic effects and extend our understanding of the biological role of these proteases.
Tumor necrosis factor ␣ (TNF␣) 1 is a cytokine that exerts a variety of biological effects and is involved in processes such as the host response to infection, in regulation of the immune system and in inflammation (reviewed in Ref. 1). TNF␣ has cytotoxic effects on some transformed cells but the majority of human tumor cells and normal tissues are resistant to TNF␣induced death (2,3). It has been known for some years that blocking protein synthesis sensitizes most cells to TNF␣-induced apoptosis indicating that protein synthesis-dependent protective mechanisms normally antagonize protein-synthesis independent cytotoxicity (1,4). The molecular basis of the TNF␣-induced pro-apoptotic pathway has been characterized recently and involves a group of adaptor proteins containing a conserved motif known as a death domain (DD) and a family of proteases known to play a central role in apoptosis (caspases) which themselves require proteolytic activation (5).
The adaptor protein TRADD (TNF-R1-associated death domain protein) (6) binds to the intracytoplasmic portion of the TNFR1/p55 and associates via its DD with the DD of the adaptor FADD (Fas-associated death domain protein) (7)(8)(9). FADD, which is also involved in transducing apoptotic signals from Fas/CD95, binds via its death effector domain to caspase-8 (10) resulting in its proteolytic activation most likely by autoprocessing. Activated caspase-8 is able to proteolytically activate downstream caspases such as caspase-3/CPP32 which can result in apoptotic cell death (11). The nature of the antiapoptotic pathway induced by TNF␣ has also been unravelled. TNF␣ activates the NFB transcription factor by a pathway involving the death domain kinase RIP (as shown by defective activation of this pathway in rip null mice) (12) leading to the activation of a kinase cascade resulting in the phosphorylation and degradation of the inhibitory IB protein (13,14). This pathway exerts a significant anti-apoptotic effect by regulating the transcription of target genes. Some of these have been identified and include IEX-1L, TRAF1/2, and c-IAP1/2 (15,16). The NFB pathway is considered to be the main mechanism opposing TNF␣ induced cytotoxicity (5,17,18) but there is also a significant role for an independent survival pathway downstream of TRAF2 (19 -21).
The aim of this study was to evaluate the role of inhibiting caspase activity in modifying TNF␣-induced cellular responses, in particular cell death by apoptosis. Unexpectedly, we found that inhibition of caspase activity in a tumor-derived human pro-monocytic cell line sensitized to the cytotoxic effects of TNF␣ in the absence of protein synthesis inhibition. Similar effects were seen in murine NIH3T3 fibroblasts with cells dying by a non-apoptotic mechanism dependent on the excessive formation of reactive oxygen species (ROS). Therefore, TNF␣induced caspase activity has an anti-necrotic as well as a proapoptotic effect and inhibition of this pathway in certain cells sensitizes them to TNF␣-induced cell death.

EXPERIMENTAL PROCEDURES
Cell Lines, Reagents, and Expression Vectors-The U937 cell line was maintained in RPMI plus 10% fetal calf serum and the NIH3T3 cell line in Dulbecco's modified Eagle's medium plus 10% fetal calf serum. zVAD.fmk was from Bachem and all other peptide caspase inhibitors were from Enzyme Systems Products. Human recombinant TNF␣ was from Peprotech; reduced glutathione ethyl ester and N-acetylcysteine were from Sigma. The DN-FADD construct was a gift from V. Dixit and has been previously described (8). The SEK-AL vector was a gift from J. Woodgett and the mutant H-Ras and Rac vectors were as described previously (22).
Cell Transfection and Survival Assay-NIH3T3 cells were plated at 10 5 cells/well of a 6-well plate and transfected the following day with 1.5 g of the test plasmid DNA plus 0.5 g of a GFP plasmid (pEGFP-N1, CLONTECH, to identify transfected cells) using the Superfect reagent (Qiagen) according to the manufacturers instructions. Typical transfection rates at 24 h ranged from 20 to 40%. After 24 h cells were treated with diluent control, TNF␣ (50 ng/ml) alone, zVAD.fmk (50 M) alone, or zVAD.fmk for 30 min followed by TNF␣. After 8 or 24 h, cells were detached, washed, and resuspended in phosphate-buffered saline containing 1 g/ml propidium iodide. Samples were analyzed by flow cytometry (Epics, Coulter) for GFP positivity and the ability to exclude propidium iodide. The numbers of propidium iodide excluding, GFP positive cells in the TNF␣ or zVADϩ TNF␣ samples, were compared with those in diluent control treated cells and the results expressed as % survival. Control cells were transfected with plasmid vector plus GFP.
Measurement of Caspase-8 Activity-U937 cells were incubated with TNF␣ (50 ng/ml) after incubation for 45 min with cycloheximide (5 g/ml), zVAD.fmk (50 M), or diluent control in RPMI/10% fetal calf serum at a cell density of 10 6 per ml. At indicated time points 5 ϫ 10 6 cells were washed in phosphate-buffered saline, lysed, and processed for caspase-8 activity using the ApoAlert caspase-8 colorimetric assay kit (CLONTECH) according to the manufacturers instructions. Each assay point was run in duplicate using lysates from 2 ϫ 10 6 cell equivalents.
Measurement of Reactive Oxygen Species-NIH3T3 cells were plated at 10 5 cells/well of a 6-well plate and treated with diluent control, TNF␣ (50 ng/ml) alone, zVAD.fmk (50 M) alone, or zVAD.fmk for 30 min followed by TNF␣. At the stated time points, dihydrorhodamine 123 (DHR123, Molecular Probes) was added to a final concentration of 2 M for 30 min. Cells were then washed and harvested and analyzed by flow cytometry. ROS production was compared with that in control cells by setting an appropriate gate (Fig. 4) and considering cells with a greater fluorescence than this to be positive. In experiments investigating induction of the mitochondrial permeability transition and production of ROS simultaneously, cells were incubated for 30 min with 50 nM 3,3Јdiethyloxacarbocyanine (DiOC 6 (3)) to assess mitochondrial membrane potential and 2 M hydroethidine to measure ROS production.
Measures of Apoptosis and Cell Survival-DNA fragmentation was quantified using the Cell Death Detection enzyme-linked immunosorbent assay (Roche Molecular Biochemicals) as described previously (23). For propidium iodide staining to detect hypodiploid DNA cells were fixed overnight in 70% ethanol at Ϫ20°C, washed, and incubated with RNase A for 10 min and then propidium iodide for 5 min before analysis by flow cytometry. For clonogenic survival assays, NIH3T3 were plated at 200 or 500 cells per 5-cm tissue culture plate in standard medium and fixed/stained with methanol/crystal violet 10 -14 days later.

Peptide Caspase Inhibitors Can Sensitize U937 and NIH3T3
Cells to the Cytotoxic Effects of TNF␣-Like many cell lines, U937 human leukemic cells are normally resistant to the cytotoxic effects of TNF␣ but will undergo apoptosis if protein synthesis is inhibited by cycloheximide with over 95% cell death within 6 h (data not shown). While investigating the role of caspase inhibitors in modifying TNF␣-mediated apoptosis we found that preincubation of U937 cells with the broad spectrum peptide caspase inhibitor zVAD.fmk (24) sensitized cells to TNF␣ cytotoxicity in the absence of cycloheximide (Fig. 1). This effect was rapid with over 30% of cells unable to exclude propidium iodide or trypan blue after 6 h. Prolonged incubation showed that less than 5% of cells were alive at 48 h in optimal TNF␣ and zVAD.fmk concentrations. In order to quantify caspase activation in response to TNF␣ alone or in cells preincubated with either cycloheximide or zVAD.fmk, we measured cleavage of the preferred caspase-8 substrate IETD-pNA using cell extracts made at time points up to 6 h after TNF␣ addition. Fig. 1 shows that in cells preincubated with cycloheximide there is a marked increase in caspase-8 activity that peaks 2-3 h after TNF␣ addition. The addition of TNF␣ alone results in a small but consistent increase in caspase-8 activity which is between 8 and 16% of the response seen with cycloheximide plus TNF␣. Although the magnitude of the increase with TNF␣ alone is relatively modest, the increase over baseline activity at time points between 120 and 360 min is highly statistically significant (p Ͻ 0.01 by paired t test compared with time 0) and is inhibited by preincubation with the caspase inhibitor zVAD.fmk.
In addition to the findings in U937 cells we detected a similar response to zVAD.fmk plus TNF␣ in NIH3T3 fibroblasts (Fig.  1). In the absence of cycloheximide these cells were resistant to the cytotoxic effects of TNF␣ at concentrations up to 100 ng/ml. However, in the presence of 50 M zVAD.fmk, the addition of as little as 0.1 ng/ml TNF␣ resulted in the death of 68 Ϯ 5% of cells within 24 h.
Testing of a panel of peptide caspase inhibitors showed that the broad spectrum zVAD.fmk and boc.D.fmk inhibitors were most effective at promoting TNF␣ cytotoxicity (Fig. 2). IETD-.fmk, which has been shown to inhibit caspase-8 activity (25), was also effective. The inhibitor YVAD.fmk was ineffective in this assay but showed biological activity by inhibiting Jurkat cell apoptosis in response to an agonistic Fas antibody (viability after 24 h in control cells 95%, anti-Fas 19%, YVAD ϩ anti-Fas 90%). Morphological examination showed that whereas U937 treated with cycloheximide plus TNF␣ showed typical features of apoptosis, zVAD ϩ TNF␣-treated cells underwent non-apoptotic death with loss of membrane integrity, prominent vacuolation, and absence of nuclear/cytoplasmic condensation (Fig.  2). In contrast to the findings in U937 and NIH3T3 cells, we found that zVAD.fmk, boc.D.fmk, and IETD.fmk were effective inhibitors of TNF␣-induced apoptosis in the HeLa cell line as has been reported previously (Ref. 26 and data not shown).
Inhibition of FADD Function Results in Non-apoptotic Death in TNF␣-treated NIH3T3 Cells-The FADD protein links the TNFR1 to caspase-8 by binding via its carboxyl-terminal DD to homologous sequences in the adaptor protein TRADD. fects of inhibiting FADD-mediated caspase recruitment to the TNFR1 complex in NIH3T3 cells. Cells were transiently transfected with a DN-FADD cDNA plus a GFP vector to positively identify transfected cells. Cell expression of GFP and cell viability (detected by uptake of propidium iodide) were quantified by flow cytometry. Transfection of dominant-negative FADDsensitized NIH3T3 cells to TNF␣ with only 35 Ϯ 7% of transfected cells surviving at 8 h after addition of TNF␣ (Fig. 3). In contrast, using the same assay technique, the transient expression of DN-FADD protected HeLa cells from both Fas and TNF-induced apoptosis as has been described previously (data not shown) (26).
NIH3T3 cells treated with cycloheximide plus TNF␣ undergo apoptosis. After incubation with zVAD.fmk and TNF␣, cells undergo a non-apoptotic form of death (Figs. 2 and 3). There is no evidence of DNA fragmentation or appearance of a hypodiploid DNA containing fraction (Fig. 3) and no early exposure of membrane phosphatidylserine as measured by annexin V staining (data not shown). The lack of features of typical apoptosis could be explained by the presence of the broad spectrum caspase inhibitor preventing the full appearance of the apoptotic cell phenotype. However, cells transfected with DN-FADD and incubated with TNF␣, in the absence of any peptide caspase inhibitor, also do not undergo apoptosis. There is no evidence of significant DNA fragmentation, no hypodiploid cells, and no early exposure of annexin V-binding sites (Fig. 3).

TNF␣-induced Cytotoxicity in FADD/Caspase-inhibited Cells Is Due to the Excessive Generation of Reactive Oxygen
Species-In view of previous evidence implicating the production of ROS in some forms of TNF␣-induced cytotoxicity (27) we investigated whether ROS may be involved in the form of non-apoptotic death seen here. Measurement of intracellular ROS using the fluorescent probe dihydrorhodamine-123 shows that there is normally a modest transient production of ROS in response to TNF␣ in NIH3T3 cells (Fig. 4). This returns to a basal level by 4 h. In the presence of zVAD.fmk, the response to TNF␣ is accentuated and prolonged and is detected in viable cells prior to the onset of death. Cells transfected with DN-FADD show a similar response when incubated with TNF␣ alone (Fig. 4).
Production of ROS has also been shown to occur in some forms of apoptosis and may result from loss of mitochondrial membrane potential and uncoupling of the respiratory chain (28,29). We investigated the temporal relationship between loss of transmembrane potential as measured by DiOC 6 (3) fluorescence and ROS production in NIH3T3 that had either been triggered to undergo apoptosis by cycloheximide ϩ TNF␣ or treated with zVAD ϩ TNF␣. The results show that whereas in cycloheximide ϩ TNF␣-treated cells, DiOC 6 (3) low cells appear prior to the detectable production of ROS, the situation is reversed with zVAD ϩ TNF␣ (Fig. 4).
To assess whether the production of ROS plays a causal role in this form of cell death, NIH3T3 and U937 cells were preincubated with the antioxidants reduced glutathione and N-acetylcysteine prior to the addition of zVAD.fmk and TNF␣. Fig. 5 shows that both of these agents could inhibit the cytotoxicity of TNF␣ under these conditions. Similar results were also seen with butylated hydroxyanisole (data not shown). Long-term (2 week) clonogenic assays confirmed that there was a genuine reduction in cytotoxicity (Fig. 5). Similarly, reduced glutathione protects cells expressing DN-FADD from the cytotoxic effects of TNF␣ (Fig. 5).
These results indicate that TNF␣ activates a FADD/caspaseindependent pathway leading to the production of ROS but that the subsequent down-regulation of ROS production is dependent on FADD/caspase activity. TNF␣ has been shown to In the experiment shown, cell viability was 63% for Chx ϩ TNF␣, 58% for zVAD.fmk ϩ TNF␣, 71% for DN-FADD ϩ TNF␣ and greater than 93% for all other conditions. activate multiple signaling pathways. The GTP-binding protein Rac is a critical component of the phagocytic cell NADPH oxidase system and has been implicated in the production of ROS in response to cytokines, growth factors, and oncogenes in many cell types (30 -32). Transient transfection of a dominantnegative N17Rac in NIH3T3 cells attenuated the cytotoxic response of zVAD.fmk plus TNF␣ (Fig. 6). Co-transfection of N17Rac with DN-FADD also inhibited the subsequent cytotoxicity seen with TNF␣ treatment (Fig. 6). Expression of a constitutively active V12Rac did not promote TNF␣-induced cytotoxicity suggesting that although activated Rac is required for TNF␣-induced death it is not sufficient on its own to give this response. In some situations, Rac may lie downstream of the Ras GTPase (22). Expression of a dominant-negative Ras in NIH3T3 cells did not affect zVAD.fmk plus TNF␣-induced cytotoxicity. Expression of a dominant-negative SEK, which inhibits Jun kinase (JNK) activation in response to TNF␣ and which has been implicated in some forms of cell death (33, 34) (Fig. 6) and a pharmacological inhibitor of the p38 stress kinase (SB203580) did not decrease cell death (not shown). DISCUSSION We have shown that activation of a FADD/caspase pathway by TNF␣ leads to complex effects on components of the cell survival machinery. In the presence of inhibitors of protein synthesis such as cycloheximide, stimulation with TNF␣ results in apoptotic cell death. However, most cells are resistant to the apoptotic effects of TNF␣ due to the dominant anti- apoptotic effect of NFkB which is activated in parallel. In U937 cells, incubation with TNF␣ results in a low level increase in caspase-8 activity (in the absence of cycloheximide) under conditions in which there is no detectable increase in apoptosis above background levels. This suggests that a limited increase in caspase-8 activity need not necessarily result in a full-blown activation of an execution caspase cascade and that the NFBdependent pathways involved in preventing TNF␣-induced apoptosis can exert at least some of their effects downstream of caspase-8. Activation of the FADD/caspase pathway by TNF␣ not only mediates a pro-apoptotic pathway but also, by downregulating the excessive production of ROS, exerts an antinecrotic effect. The relative importance of this appears to vary between different cell types but it could serve in some situations to promote a controlled form of cell death.
The TNF␣-activated pathway leading to the production of ROS is independent of FADD/caspase and involves the Rac GTPase. The FADD/caspase pathway is involved in damping the production of ROS but the precise molecular mechanism for this has not been identified. The effect of ROS on cell prolifer-ation and survival are known to be complex. The positive effects of certain mitogens and oncogenes on growth have been shown to involve the generation of ROS (32) and TNF␣-induced activation of the anti-apoptotic NFB has been shown in some instances to be inhibited by antioxidants (35). In other settings TNF␣-induced cytotoxicity can be inhibited by increasing antioxidant levels, for example, by overexpression of manganese superoxide dismutase (36,37). Other workers have shown that in some cell types, TNF␣ induces apoptosis by a mechanism involving the production of ROS that can be inhibited by peptide caspase inhibitors (38,39). While this work was in progress, Vercammen and colleagues (40) reported that murine L929 fibrosarcoma cells, which undergo rapid necrosis due to excessive formation of ROS in response to TNF␣, are greatly sensitized by preincubation with peptide caspase inhibitors or by expression of CrmA, a serpin-like caspase inhibitor of viral origin. These findings in conjunction to those reported here indicate a significant role for a caspase mediated activity in the regulation of metabolism of ROS.
Caspases themselves can be modified by oxidation leading to a marked inhibition of their proteolytic activity (41). Thus, there is likely to be a complex inter-relationship between the production of ROS, caspase activity, and eventual outcome in terms of cell death or survival and the molecular mechanisms underlying this will require further characterization. FADD has been considered to be a prototypical example of a molecule with function restricted to a pro-apoptotic pathway. Recent studies have shown that although FADD is required for apoptosis induction via Fas/CD95 it also plays unexpected roles in embryonic development (42) and in promoting T-cell survival and proliferation (43). Whether the role for FADD in regulating the production of ROS described here is involved in mediating any of these effects remains to be determined.
The ability of caspase inhibitors to sensitize tumor cells to TNF␣ may also have relevance for the treatment of malignant disease. It has been postulated that combining TNF␣ with molecular inhibitor(s) of the NFB pathway may be used as an approach to overcome resistance in tumor cells (44). However, this approach is likely to be limited by toxicity to normal cells. It has been shown previously that pretreatment of mice with the peptide caspase inhibitor zVAD.fmk can ameliorate the tissue damage normally seen with TNF␣ administration (45). Further studies to evaluate the effects of the combination of caspase inhibitors and TNF␣ on normal tissues and transformed/tumor cells are in progress. Although targeting of the FADD/caspase pathway as described here would promote a non-apoptotic form of tumor cell death with potential toxicity, recent data indicate that this may promote the host anti-tumor immune response (46). Further studies will be needed to clarify the potential application of this approach to tumor therapy and to enable a re-evaluation of the role of TNF␣ in the treatment of malignant disease.
In conclusion, we have demonstrated that a FADD/caspase pathway can regulate the production of ROS in response to stimulation with TNF␣ and that this pathway serves to inhibit the process of cellular necrosis. This suggests a dual role for caspases: first as mediators of apoptosis, a highly controlled form of cell death, and second as inhibitors of necrosis, a form of cell death that is likely to be injurious to the organism.