In this study, we investigated the role of caspases in TNF-mediated necrosis. First, we used the cowpox CrmA gene product as an inhibitor of a number of caspases. Surprisingly, expression of CrmA in L929 cells rendered them far more sensitive to TNF as compared to control cells not expressing CrmA. Furthermore, blocking of caspases by peptide inhibitors sensitized the cells to TNF-induced cytotoxicity. zDEVD-fmk and Ac-YVAD-cmk had moderate sensitizing activity, whereas zVAD-fmk and zD-fmk strongly potentiated TNF-mediated necrosis. In the latter case, the concentration of TNF required for half-maximal cytotoxicity decreased 1,000-fold. zDEVD-fmk and Ac-YVAD-cmk have a different specificity pattern, and when they were combined, they could not synergize with each other, suggesting the possibility that they inhibit consecutively acting caspases. TNF sensitization induced by zVAD-fmk was accompanied by an enhanced production of oxygen radicals, as measured by DHR123 oxidation. Scavenging of oxygen radicals by BHA completely abrogated the sensitizing effect of zVAD-fmk on TNF-induced necrosis. This indicates that enhanced oxygen radical production is the main cause of zVAD-fmk–mediated sensitization.
KYM and HeLa H21 cells respond to TNF treatment in an apoptotic way. When these cells were treated with TNF in the presence of zVAD-fmk, cell death was inhibited, revealing a fundamental difference between necrosis and apoptosis. In contrast to apoptosis, TNF-induced necrosis of L929 cells is not dependent on caspase activation; rather, the results shown here indicate a protective role for a low level of constitutively active caspase(s) in this mode of cell death. Alternatively, TNF may induce activation of a caspase that counteracts or deviates the pathway leading to mitochondrial oxygen radical production; this caspase activity would be at a level below the detection limit obtainable with fluorogenic substrates. TNF-induced cell death is primarily mediated by the p55 TNF receptor (21
), which contains a death domain (DD) in its intracellular part. Upon ligand-induced clustering of receptor DDs, other DD-containing components of the signaling pathway are recruited, leading to cell death (3
). In the case of TNF-mediated necrosis in L929 cells, the DD of the p55 TNF receptor has been shown to be necessary and sufficient for fully active TNF signaling to necrosis (26
). TRADD, which also has a DD, binds to the DD of clustered p55 TNF receptor, and is in turn necessary for recruiting the DD-containing FADD/MORT1 (25
). The latter was first identified as a factor recruited by Fas, another DD-containing receptor, upon activation (14
). In the case of the p55 TNF receptor, recruitment of FADD in the receptor complex has not yet been demonstrated at physiological receptor numbers. FADD/MORT1 possesses another domain that connects Fas and the TNF receptor complex to caspase-8 (12
) or the homologues caspase-10 and caspase-10b (15
). Caspase-8 contains a COOH-terminal caspase-3–like domain that is proteolytically released into the cytosol after stimulation of Fas or the p55 TNF receptor (14
). It is generally assumed that caspase-8 is the apex of a pathway leading to apoptosis in which the downstream executors are other caspases. The proteolytic activity of caspase-8 and caspase-10b is inhibited by zVAD-fmk, zDEVD-fmk, and CrmA, but not by Ac-YVAD-cmk (14
). In this study, we show that neither zVAD-fmk, zDEVD-fmk, nor CrmA block TNF signaling to necrosis, but, on the contrary, considerably enhance cytotoxicity. Obviously, TNF-mediated necrosis in L929 cells is not dependent on caspase-8/caspase-10, but in fact is attenuated by one or more caspases.
Our results suggest a new role for caspases as negative regulators of TNF-induced oxygen radical production and consequent necrosis. As shown previously, TNF-induced radical formation in L929 cells depends on an intact electron transport system in the mitochondria, and probably involves O2
−·, H 2
, and/or lipid hydroperoxides (5
). Although evidence for the existence of mitochondrial caspases has recently been reported (31
), a role for caspases in the electron transport system has not yet been demonstrated. However, since CrmA is probably located in the cytosol, it is unlikely that mitochondrial caspases are involved. Rather, it seems that one or more caspases interfere with the signal from the triggered receptor to the mitochondria. Alternatively, the production of oxygen radicals may be counteracted by caspases at the level of the mitochondria themselves (Fig. ).
Figure 6 Possible mechanisms of action in caspase inhibitor-mediated sensitization of TNF-induced necrosis in L929 cells. A putative caspase (CASP-X), inhibited by CrmA or zVAD-fmk, acts as a negative regulator of premitochondrial signaling (1) or (more ...)
A third hypothetical model is the following. Degradation of mitochondrial proteins has been documented both in physiological and pathological conditions (33
). This is especially the case when membrane proteins are damaged by oxygen radicals. In mitochondria of rat liver cells, increasing the radical production results in enhanced protease activity (34
). In addition, oxidative damage to intracellular proteins increases their susceptibility to proteolysis (35
). Although it is known that in some of these turnover processes, mitochondrial and/or cytosolic ATP-dependent protease complexes play an important role, there is also evidence for involvement of ATP-independent proteases in mitochondrial catabolism. Possibly, caspases could be key elements in such an intracellular mitochondrial quality control system. As cells increase their production of oxygen radicals in the mitochondria after p55 TNF receptor stimulation, oxidative damage of lipids and proteins accumulates; this results in occasional failure of the electron transport system, which leads to an amplified radical production. It is conceivable that such defective mitochondria are recognized and eliminated by a specific cellular mechanism, and this is where caspases could play a role. Elimination of such deficient but oxygen radical–producing mitochondria should then be beneficial for the cell to survive the deadly TNF signal. By inhibiting cytosolic caspase activity, this “rescue mechanism” would be impaired, and hence the cells would accumulate excessive reactive oxygen-producing mitochondria and would be far more sensitive to TNF-induced necrosis. Whatever the exact mechanism is, a low activity of caspases is implied, stressing the importance of a stringent control mechanism of caspase activity in healthy cells. Fig. illustrates alternative mechanisms for possible interference by caspases in TNF-induced mitochondrial production of reactive oxygen intermediates.
The results reported here prompt us to add a cautionary note. Indeed, caspases have already been shown to be essential mediators in illness-related cell death, such as neuronal damage following hypoxic-ischemic insult (36
) or fulminant liver destruction after anti-Fas injection (37
), and evidence exists for the implication of caspases in amyotrophic lateral sclerosis (38
) and Alzheimer's disease (39
). In the first two indications, inhibition of caspases by tripeptide derivatives protects treated mice against injury and death. However, considering the 1,000-fold sensitization of TNF-induced necrotic cell death by inhibitors of caspases, one should be cautious in cases where reactive oxygen-mediated necrosis may be involved, such as neutrophil-induced endothelial cell necrosis in the systemic inflammatory response syndrome (40
); liver necrosis after reperfusion, alcoholic liver disease, or hemochromatosis (iron overload) and Wilson's disease (copper overload) (41
); and myocardial ischemia and reperfusion injury (42
). It is not excluded that in these indications, administration of caspase inhibitors may rather have an adverse effect. Therefore, the mechanism leading to cell death should be taken into account when the use of caspase inhibitors would be considered as disease treatment.