Receptors to die for

Death is as intrinsic to life as birth, yet most people choose to ignore its looming presence. It must be this aversion that has allowed cellular proliferation and differentiation to take centre stage in scientific research for decades, while little or no attention was given to death. The discovery of specific apoptosis regulators, apoptotic enzymes (caspases) and death receptors only two decades ago have clearly changed this imbalance and has now resulted in a place for necrophilic research at the forefronts of science. The current issue of Cell Death and Differentiation underscores this notion and exemplifies the complexity that has evolved since the identification of the prototypic death receptor CD95 (Fas/APO-1).

At the time of its cloning, CD95 was recognised to share a death domain with its older brother TNF-R1. Soon thereafter this domain was reported to mediate the assembly of the death-inducing signalling complex (DISC) after trimerisation of the receptor. This allows the death receptors to activate the initiator caspase-8 and release the active enzyme into the cytoplasm (reviewed in Peter and Krammer¹).

In essence this model still stands, but quite a few additions have been made through the years. First of all, clinical and biochemical evidence have provided clear evidence that besides caspase-8, caspase-10 is an essential part of the DISC. This is best exemplified by the dramatic effects displayed by patients carrying a mutation in this caspase. These patients develop an ALPS-type of disease (autoimmune lymhoproliferative syndrome) that is reminiscent of the classical ALPS patients that contain a mutation in CD95 or CD95L (reviewed in Rieux-Laucat et al.²). Secondly, more and more data argue for an important role for the membrane microdomain rafts in death receptor signalling as DISC formation is preferentially seen in these rafts (reviewed in Heuber³). Whether this is because of the formation of higher-order oligomers in the rafts is not clear yet, but it seems very likely that the original receptor trimerisation hypothesis needs to be revised.

Thirdly, FLIP_L, the molecule that saw the light as the inhibitor of the DISC, should now be seen as an essential player that keeps the DISC running. Indeed, recent evidence indicates that it teams up with caspase-8 to form an active complex similar to what has been suggested for the APAF-1/caspase-9 apoptosome. That is, one part of the complex has an active caspase-8 site, while the other site is the inactive FLIP_L part (reviewed in Peter and Krammer¹). Only at exceedingly high levels, FLIP_L prevents the DISC from giving a death signal. But even then it may not be silent, but rather shifts from death to (co-) stimulation. In agreement, T cells signal NFκB and ERK activation via CD95 and this requires the presence of FLIP_L.⁴

What then happens after caspase-8 activation? Initially, we could suffice by saying that other caspases were activated and apoptotic substrates were cleaved. However, as the list of substrates is growing to mind-blowing numbers (reviewed in Fischer and Schulze-Osthoff⁵), it becomes necessary to assign apoptotic meaning to the different substrates. Unfortunately, we have been rather unsuccessful at doing so and as far as experiments go, most substrates could simply be accidental victims of a raging enzyme. Only a few serve a definite role in the apoptotic process. Of these, ICAD is one of the clearest examples (reviewed in Nagata⁶). Cleavage of this protein allows ICAD's binding partner CAD, the apoptotic DNAse, to be released and start fragmenting DNA, one of the necessary hallmarks of apoptosis. Nevertheless, ICAD cleavage may be dispensable as mice deficient for the CAD DNAse activity show no phenotypic abnormalities. Recent studies have also assigned a role to the cleavage of the most frequently used apoptotic marker, PARP (reviewed in Fischer and Schulze-Osthoff⁵). In a cell that contains fragmented DNA, this enzyme would become highly activated and consume vast amounts of ATP. If therefore left uncleaved, it will likely drain the cell's energy stores and shift the balance from apoptosis to necrosis with its potentially harmful side effects to the surrounding tissue. As such PARP can be considered as an essential substrate. Most other substrates do however remain an enigma and many years will pass before we can discriminate between accidental, non'essential and essential substrates.

Also the mode of caspase activation downstream of caspase-8 has led to a bifurcation in our thinking. Upon activation, caspase-8 can activate caspase-3 and thereby induce all events necessary to obtain apoptosis (type 1 cells). However, direct cleavage of caspase-3 is very inefficient in several cell lines and cell types, such as hepatocytes. In these cells, which are now considered to be type 2, amplification occurs that involves the mitochondria, that is, the intrinsic death pathway. This amplification is initiated by a caspase-8-dependent cleavage of Bid, which then mobilises the apoptogenic substances from the mitochondria. Once released the best-studied substance, cytochrome c, allows caspase-9 and APAF-1 to interact via their CARD-domains (reviewed in Lahm et al.⁷) and to form the apoptosome. This then results in effective activation of downstream caspases and the amplification of the pathway. Bratton et al.⁸ in this issue now provide an interesting side step to this model. In their view the mitochondrial amplification may also serve to release Smac and/or Omi, which then prevent IAPs from inhibiting the downstream caspases. Cells with high XIAP would therefore need to mobilise Smac before caspase-8 can activate caspase-3 (type 2), while XIAP-low cells would do so directly (type 1). This clearly doesn't explain all facets of type1/2 cells, but it is an interesting addition to our thinking and may well be a second level at which type 1/2 is regulated.

The type1/2 distinction appears justified by the apoptotic phenotype of different cell types from either Bid-deficient or Bax/Bak double-deficient mice. Nevertheless, it should be noted that the type 1/2 is also a fiercely contested model. The usage of CD95 ligand instead of agonistic antibodies would annihilate the difference and turn all cells into type1.⁹ Moreover, recent data suggest that caspase-9/APAF-1 are dispensable for several apoptotic stimuli, while these stimuli are inhibited by Bcl-2 (reviewed in Baliga and Kumar¹⁰). One possible explanation of these results is that Bcl-2 prevents the activation of an apical caspase that can induce apoptosis in a mitochondrial-dependent and -independent fashion. Intriguingly, such a model was recently also postulated for caspase-2, which was positioned between stress signals and the mitochondria (reviewed in Troy and Shelanski¹¹). Caspase-2 may also play a role in CD95 and TNF-induced cytochrome c release and thus could close the circle. Whether this is true remains to be determined, but it seems to be supported by the fact that caspase-2, as well as several other possible upstream caspases, has a CARD-domain (reviewed in Lahm et al.⁷) and could thus form a premitochondrial apoptosome.

Even though this revised positioning of Bcl-2 is popular and would finally order mammalian cell death according to C. elegans, some caution is clearly warranted. For instance, Bcl-2 not only prevents the release of cytochrome c, but also prevents Smac from leaving the mitochondria. As discussed above, it could simply be this molecule that (inefficiently) triggers caspase activation in the APAF-1 or caspase-9 knockouts and allows apoptosis to occur. Similarly, AIF and endonuclease G could play a role in this setting. These observations therefore indicate that we are still scratching the surface and are far from agreeing on the role of the mitochondria in CD95-induced apoptosis.

How about the other death receptors? As mentioned, CD95 is considered to be prototypic for the family. Nevertheless, reviews in this issue again stress the point that generalisation of this family is no longer warranted. TNF-R1, for instance, has a signalling cascade that is so immensely diverse from CD95 that it is hard to compare them (reviewed in Wajant¹²). Similarly, DR6, TRAMP, EDA-R and NGF-R can be considered as different entities that rather signal towards gene transcription than death. As such, the only really common denominator in the family is the death domain itself and even this domain is apparently used differently by the separate members. In agreement, phylogenetic analysis of the death domains indicates that they can be separated into different clades that appear to serve different functions in either development or immunity, and may thus have emerged at separate points in evolution (reviewed in Bridgham¹³).

From a signalling point of view, CD95 is prototypic for the TRAIL receptors 1 and 2, as the TRAIL DISC appears identical to the CD95 DISC (reviewed in LeBlanc and Ashkenazi¹⁴). In spite of this homology, one can't go around the differences. First of all, TRAIL signalling may be heavily complicated by the so-called decoy receptors. Whether these really act as decoys can be debated, but they will have a reason-d'être and may influence signalling in the long run. Similarly, some cells are sensitive to TRAIL while prove resistant to CD95L-induced death, even though the receptors are expressed. In addition, the type 1/2 division does not seem to hold for TRAIL. These and other observations clearly indicate a qualitative and/or quantitative difference between CD95L and TRAIL that may explain the sensitivity of a large panel of tumors to TRAIL.

In spite of our failure to grasp death in all its aspects, researchers have been trying to target the death receptors for clinical use. The CD95 system, for instance, has been shown to play a role in diseases as diverse as stroke, toxic epidermal necrolysis (TEN) and multiple sclerosis (reviewed in French LE and Tschopp¹⁵). Mouse models for these diseases show beneficial effects of blocking CD95-induced apoptosis, but clinical applicability is still limited because of the lack of suitable blockers. Nevertheless, the presence of anti-CD95L antibodies in intravenous immunoglobulin preparations, routinely used to treat TEN, suggests that there is a window of opportunity that should be scrutinised.

On the other end of the spectrum, that is, in cancer, it is the induction and not the prevention of apoptosis that is highly desirable. Unfortunately, CD95-based therapies are hampered by severe (liver) toxicity. TRAIL on the other hand seems to stick to its acquired reputation as a cancer drug. Several mouse models have shown a clear efficacy in preventing tumor outgrowth of different TRAIL preparations (reviewed in Bridgham¹³, LeBlanc and Ashkenazi¹⁴). This may actually be (one of) the physiological roles of TRAIL as mice deficient for TRAIL develop tumors more rapidly, in part because of a failure of NK cells to kill these tumors. Despite the promising results, there are still some formidable hurdles to take. For instance, none of the mouse studies published actually rid the mice of the transplanted tumors. This is especially true when one looks at established tumors that are treated with TRAIL, that is, how one will supposedly apply the drug in the clinic. Another possible problem is that at this point relatively high doses are needed to obtain the effects and little attention has gone towards escape variants that are rendered TRAIL resistant. It is obvious that neither of these points will and should stand in the way of clinical development, but efforts to combine TRAIL with other therapies are likely to prove more viable in the long run. Similarly, the reported hepatocyte toxicity will not prevent further development as this depends on the type of preparation used. This does once more raise the point though that we should be aware of the different outcomes one can find when using different ligand preparations and/or agonistic antibodies. Apparently, there is more to death receptor stimulation than meets the eye. In this light, it is also interesting to note that soluble CD95L is reported to be stimulatory or inhibitory depending on how it is produced.

Taken together this issue shows that we have long left the era of simple signalling pathways. Instead we are filling in the details, cell type specificities and are attempting to understand the careful interplays that decide whether a cell lives or dies. Unfortunately, this always seems to result in fierce discussions on different viewpoints, which are often based on differences in cellular systems or protein preparations (recombinant TRAIL, CD95ligand versus antibodies). It is therefore time to step away from cell lines and our favourite death ligand preparations and move on to physiological relevant settings. Such studies are especially needed in the light of our eagerness to enter the clinic with death ligands and/or their antagonists. It is safe to say that such death receptor therapies will be feasible in the near future, even though we scientists may still be in the dark or disagree on how they work.