Death cycle and Swiss army knives

Article metrics

Cytochrome c leads a double life. When a cell is called on to commit apoptotic suicide, cytochrome c relocalizes from the mitochondria to the cytosol. There, it helps to activate the foot-soldiers of apoptosis — the death proteases known as caspases1. How cytochrome c escapes from the mitochondria is still a matter of debate1, but it is clear that certain elements within the apoptotic regulatory hierarchy do not condone such behaviour. In particular, overexpression of the cell-death suppressors Bcl-2 and Bcl-xL prevents the release of cytochrome c, suggesting that these proteins act upstream of cytochrome c in the pathway to cell death2,3. However, on pages 449 and 496 of this issue, Zhivotovsky et al.4 and Rossé et al.5 show that Bcl-2 can also protect cells downstream of cytochrome c release, forcing a re-evaluation of this newly acquired dogma.

Caspases do the brunt of the work in apoptosis. Activated through proteolytic processing, they cleave a limited number of apoptotic substrates and cause, directly or indirectly, most of the changes that are characteristic of apoptosis (for example, see ref. 6). So how are caspases activated? In the nematode worm Caenorhabditis elegans, the caspase homologue CED-3 is activated through a physical interaction between proCED-3 and CED-4. Interaction with CED-4 seems to be the only mechanism available for CED-3 activation, because there is no programmed cell death in ced-4 mutant animals. In cells that should survive, CED-9 (which belongs to the Bcl-2 family) binds to CED-4 and holds it in an inactive conformation, thereby preventing CED-4-mediated activation of proCED-3. The ability of CED-9, CED-4 and CED-3 to exist in a multiprotein complex has led to the ‘apoptosome’ model of cell-death regulation (reviewed in ref. 7).

A series of elegant papers by Wang and colleagues indicated that at least some caspases are activated through a similar mechanism in mammals. This group purified, from human cell extracts, three proteins that can activate procaspase-3 in the presence of dATP. The molecular nature of two of these apoptotic protease activating factors — or Apafs — made perfect sense: Apaf-3 turned out to be caspase-9, which has a similar structure to CED-3 (ref. 8); and Apaf-1 was the long-sought mammalian homologue of CED-4 (ref. 9). As expected from sequence similarities with the worm proteins, Apaf-1 binds to Apaf-3/caspase-9 and promotes its proteolytic activation.

But there are two interesting twists to this story. First, unlike CED-4, Apaf-1 needs a cofactor in order to bind to and activate caspase-9. To general stupefaction, this cofactor — Apaf-2 — turned out to be the humble cytochrome c, which is normally present on the outer surface of the inner mitochondrial membrane. Here, it usually shuttles electrons between systems III and IV of the mitochondrial electron-transport chain.

The second twist concerns the mechanism of cell-death suppression by the Bcl-2 oncoprotein (and its homologues). Bcl-2 interacts with at least half a dozen proteins (not counting other Bcl-2-family members10), but it is still not clear how it prevents death. Although it would be attractive to use the C. elegans data to postulate that Bcl-2 will bind to Apaf-1 and/or other adaptors of its ilk, evidence for such a model is woefully lacking. Rather, last year two groups11,12 showed that high levels of Bcl-2 can prevent the release of cytochrome c by mitochondria, providing an attractive biological — if not molecular — function for the protein (Fig. 1a). Could Bcl-2, prodded perhaps by the recruitment in mammals of a new apoptotic cofactor, have adopted a new function? Has the C. elegans model reached the limits of its usefulness?

Figure 1: Models for the function of Bcl-2 in apoptosis.

Based on the genetic similarity with CED-9 from C. elegans, both models incorporate the idea that Bcl-2 can prevent apoptosis by interacting with caspase activators. a, The ‘Swiss army knife’ model10. Bcl-2 prevents the release of cytochrome c (perhaps by antagonizing Bax), as well as caspase activation in the presence of cytosolic cytochrome c. b, The ‘death cycle’ model. Bcl-2 blocks caspase activation, preventing amplification of pro-apoptotic signals. Signals that generate high levels of active caspases might be able to bypass the amplification step and activate the downstream caspases through a direct caspase cascade.

Maybe not. The studies of Rossé et al.5 and Zhivotovsky et al.4 now suggest a more complicated picture. Using two different approaches, these groups find that high levels of Bcl-2 can delay death, even when cytochrome c is already in the cytosol. Rossé et al. used transient transfection of Bax, a pro-apoptotic member of the Bcl-2 family, to bring about cytochrome- c release and apoptosis. Co-transfection of Bcl-2 prevented Bax-induced apoptosis, yet cytochrome c nonetheless transferred to the cytosol. So, at least in this case, Bax seems to act upstream of cytochrome c, which in turn acts upstream of Bcl-2. Zhivotovsky et al. used a more direct, but also more invasive, approach — they microinjected high concentrations of cytochrome c into the cytosol. They found that high levels of Bcl-2 block most of the deaths, indicating that Bcl-2 can act downstream of cytochrome c.

According to the C. elegans model of CED-9/CED-4/CED-3, this is exactly what we might have expected. But how do we explain the finding that the same protein can block death at two distinct steps in the apoptotic pathway? One possibility is that Bcl-2 is just very resourceful, and simultaneously performs many functions (the Swiss army knife model10; Fig. 1a). Or maybe the two activities are linked. For example, Bcl-2 has been reported to undergo a large-scale conformational change that can lead to its insertion into membranes and, possibly, the formation of Bcl-2 channels10. Such an insertion could allow the release of cytochrome c (by hampering the formation of Bcl-2/Bax heterodimers?), and prevent Bcl-2 from binding to Apaf-1. If not all of the Bcl-2 molecules insert, then the remaining ones could still bind Apaf-1 and prevent cell death. This might be particularly apparent if the apoptotic pathway is stimulated downstream of the events that usually lead Bcl-2 to insert into the membrane.

Alternatively, Bcl-2 might seem to act both up- and downstream of cytochrome- c release because it is involved in a circular pathway. Indeed, not only does cytochrome c stimulate caspase activation, but active caspases promote the release of cytochrome c from intact mitochondria. This suggests that mitochondria carry a caspase substrate that, when cleaved, promotes cytochrome- c release. Thus, mitochondria might act as apoptotic amplifiers, fostering a positive-feedback loop between cytochrome- c release and caspase activation (Fig. 1b). Any event that primes the loop — cytochrome- c release, caspase activation or otherwise — will initiate the vicious ‘circle of death’, eventually leading to large-scale caspase activation and apoptotic death.

For the death-cycle model to be viable, we need to add dampeners to the system. Otherwise, the slightest perturbation could be amplified, leading to unwarranted death. Good candidates are the inhibitors of apoptosis (IAP) family of proteins, which specifically inhibit particular caspases13. Bcl-2 could also be considered as a dampener, its role being to break the cycle — if the C. elegans model is right — by preventing caspase activation. If the cycle is broken early on, as might be the case with weak signals, there will be little release of cytochrome c, and Bcl-2 will seem to have acted upstream of cytochrome c.

In theory, Bcl-2 should also prevent strong pro-apoptotic signals, as long as they enter the cycle at the level of cytochrome- c release. But it might be a pyrrhic victory — such a cell might be technically alive, but if its mitochondria are mangled and its electron-transport chain disrupted, it is unlikely to thrive or divide. Neither Rossé et al.5 nor Zhivotovsky et al.4 report on the long-term prospects for the cells with cytosolic cytochrome c, leaving this question open.

At this point, both of the models can account for most of the published observations. Unfortunately, the proliferation of arrows (Fig. 1) makes falsification difficult, thereby limiting the predictive value of the models. In fact, probably the only safe prediction is that apoptosis will be a very complex process or, worse, a nonlinear one. Or both.


  1. 1

    Reed, J. C. Cell 91, 559–562 (1997).

  2. 2

    Golstein, P. Science 275, 1081–1082 (1997).

  3. 3

    Duckett, C. S.et al. Mol. Cell. Biol. 18, 608–615 (1998).

  4. 4

    Zhivotovsky, B., Orrenius, S., Brustugun, O. T. & Døskeland, S. O. Nature 391, 449–450 (1998).

  5. 5

    Rossé, T.et al. Nature 391, 496–499 (1998).

  6. 6

    Enari, M.et al. Nature 391, 43–50 (1998).

  7. 7

    Hengartner, M. O. Nature 388, 714–715 (1997).

  8. 8

    Li, P.et al. Cell 91, 479–489 (1997).

  9. 9

    Zou, H., Henzel, W. J., Liu, X., Lutschg, A. & Wang, X. Cell 90, 405–413 (1997).

  10. 10

    Reed, J. C. Nature 387, 773–776 (1997).

  11. 11

    Yang, J.et al. Science 275, 1129–1132 (1997).

  12. 12

    Kluck, R. M., Bossy-Wetzel, E., Green, D. R. & Newmeyer, D. D. Science 275, 1132–1136 (1997).

  13. 13

    Deveraux, Q. L., Takahashi, R., Salvesen, G. S. & Reed, J. C. Nature 388, 300–304 (1997).

Download references

Author information

Rights and permissions

Reprints and Permissions

About this article

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.