these structural rearrangements, the CD95 ICD can then interact, via weak molecular interactions, with another CD95 ICD brought into close proximity by the trimerized CD95L. Interactions between different CD95 ICDs stabilize the open conformation and facilitate the recruitment of adaptor molecules.

FADD is recruited by a homotypic interaction between its DD and the DD of the CD95 ICD. The following recruitment of the initiator caspase-8 and -10 again requires homotypic interactions, this time between the respective DEDs of FADD and the caspases. We describe the process leading to the activation of caspase-8 because it is virtually identical to that of caspase-10. Homotypic interaction of the DEDs of FADD and caspase-8 results in dimerization of caspase- 8. Recruitment and dimerization induce a conformational change that allows caspase-8 to become enzymatically active. It is important to note that it is not the cleavage of caspase-8 that activates it, but rather the conformational changes induced by its recruitment to the DISC and the juxtaposition with a second caspase- 8 monomer at this protein complex. Active caspase-8 then cleaves itself, but most importantly, it proteolytically activates the downstream effector caspase-3. These effector caspases then perform the proteolysis of vital cellular proteins, including structural components such as lamins and gelsolin, but also other proteins such as poly(ADP)-ribose polymerase and the inhibitor of caspase-activated DNAse. The latter event liberates the caspase-activated DNAse from cytosolic retention and thereby allows for one of the hallmarks of apoptosis, the cleavage of nuclear DNA, to take place. The proteolysis of effector caspase substrates is responsible for the characteristic biochemical and morphological hallmarks of apoptosis. The antiapoptotic factor cFLIP can prevent CD95-induced apoptosis at the level of caspase- 8. This protein is structurally similar to caspase-8 and -10 as it contains two tandem N-terminal DEDs. However, unlike these cysteine proteases, it lacks a cysteine in what otherwise would be its active center. Hence cFLIP lacks enzymatic activity as a protease. Three different splice variants of cFLIP have been described – cFLIPL, cFLIPS, and cFLIPR – and they may exert their inhibitory effects on CD95-induced apoptosis differentially.

The proapoptotic BH3-only family member Bid is a critical substrate of caspase-8 and -10. Caspase- 8/10 cleaved, truncated Bid (tBid) translocates from the cytosol to the outer mitochondrial membrane where it can induce mitochondrial outer-membrane permeabilization (MOMP) if the molecular composition with respect to other members of the Bcl-2 protein family allows it to do so. These processes are discussed

in detail in other chapters of this book. In the context described here, it suffices to point out that BID and its cleavage by caspase-8 or -10 are what link the death receptor apoptosis pathway with the mitochondrial pathway of apoptosis induction. However, one crucial consequence of activation of the mitochondrial apoptosis pathway can be decisive for the outcome of CD95 stimulation, at least in cells referred to as type II cells. MOMP induces the release of proteins from the mitochondrial intermembrane space. Most importantly, these proteins are cytochrome c as the first caspaseactivating factor, but also the second mitochondrial activator of caspases (SMAC), also known as DIABLO. Whereas cytochrome c release triggers the formation of the apoptosome resulting in activation of caspase- 9, release of SMAC induces the neutralization of the X- linked inhibitor of apoptosis protein (XIAP). Once XIAP is inhibited by SMAC, caspase-3, -7, and -9, which are all inhibited by XIAP, are released from inhibition, and cell death can finally ensue (Figure 3-2). Thus, in cells that express high levels of XIAP, the direct activation of caspase-3 by caspase-8 is blocked so that these cells require the pro-mitochondrial changes brought about by the cleavage of BID and its proapoptotic activity on mitochondria to succumb when CD95 is activated. It therefore seems that the expression of XIAP as compared with lack thereof explains the dichotomy of cells with respect to their categorization as type I and type II cells for CD95-mediated apoptosis. Differences in the extent of CD95 DISC formation, first thought to be the sole cause for the type I/type II distinction, may contribute to this.

2.2. The TRAIL (Apo2L) system

Although CD95L itself most likely will not become a major drug in cancer therapy, its discovery paved the road to the identification of a new member of the TNF cytokine family. In 1995, two groups, one at Immunex in Seattle, Washington, and one at Genentech in San Francisco, California, independently found that there was an expressed sequence tag (EST) in the public database that was even annotated as being homologous to CD95L. The TNF-related apoptosis-inducing ligand (TRAIL), or Apo2L, as it was named by these two groups, respectively, seemed to specifically kill cancer cells. A number of cancer cell lines were susceptible to TRAILinduced apoptosis, whereas the normal cells tested were not. So could it be that TRAIL would finally fulfill the hopes placed initially on TNF and then on CD95L? The answer came in 1999 by studies from both the Immunex and the Genentech groups: systemic treatment of tumor-bearing mice with recombinant TRAIL, which,

XIAP

Apoptosis

importantly, was also capable of binding to and killing mouse cells, killed tumor cells in vivo without harming normal tissue and thereby ablated tumor growth (Walczak et al., 1999). By demonstrating that a TNF-like cytokine can be used systemically in vivo to specifically kill tumor cells, these results represented the culmination of decades of research into the agonistic action of TNF family members.

Given these encouraging results, Immunex and Genentech decided to join forces so that together they would be able to fully explore the clinical potential of this promising new avenue in the treatment of cancer. In the meantime, Apo2L/TRAIL is in various clinical trials, and it is clear from these trials that there is clinical efficacy. However, it is also apparent from these trials that we are only beginning to understand the clinical potential of this drug and, in fact, a whole new class of cancer drugs, which we will refer to as TRAIL receptor agonists. This is partly due to the receptor promiscuity of TRAIL, which we discuss next.

After TRAIL was identified, the race for the cloning of its receptor began. At the time, in 1996 and 1997, many new human genes were either found in the public database of human EST sequences provided by the Human Genome Project or by Human Genome Sciences, a company that had its own private “little” human genome project. However, because it was clear that the apoptosis-inducing TRAIL receptor was going to be very valuable, as antibodies against it would potentially become new cancer drugs, other approaches also were pursued, including expression cloning and purification of the TRAIL receptor. In the end, purification and ESTbased techniques were successful. The EST approach was, however, first to discover an apoptosis-inducing receptor for TRAIL (now referred to as TRAIL-R1 or death receptor 4 [DR4]). Yet the purification approach followed only weeks later with the discovery of a different apoptosis-inducing receptor for TRAIL, the receptor now referred to as TRAIL-R2 or DR5. Shortly after that, TRAILR2 was also discovered by a number of other groups as

its sequence then appeared in the public and private EST databases only a few weeks after it was characterized and identified by purification.

However, the search for TRAIL receptors was not over yet. In the subsequent months, work by a number of groups led to the identification of two other cell-bound receptors for TRAIL, TRAIL-R3 (DcR1), and TRAIL-R4 (DcR2). These two receptors do not induce apoptosis, and it was first thought by some authors that they may exert a decoy function for TRAIL (hence the name “decoy receptor” [DcR]), which would be particularly expressed by normal cells and responsible for protecting them from TRAIL-induced apoptosis and for TRAIL’s tumorselective killing activity. However, an expression pattern of TRAIL-R3 and/or TRAIL-R4 as being present on normal but not on cancer cells was never found, putting the decoy concept for these receptors into question. Finally, it was found that TRAIL binds to a fifth receptor, osteoprotegerin (OPG). OPG is a soluble TNFRSF member that is mainly described as a regulator of the development and activation of osteoclasts in bone remodeling. It binds TRAIL only with low affinity, and its high-affinity ligand is the TNFSF member RANKL, which, apart from binding to OPG, also binds to the cell surface receptor activator of nuclear factor kappa B (NF- κB) (RANK), inducing this receptor’s osteoclast differentiation activity. It is, however, rather unlikely that the reported interaction of TRAIL with OPG is relevant in vivo because mice over-expressing TRAIL do not exhibit any bone-related phenotype, which would have been expected if TRAIL were capable of interacting with the bone-protective OPG in vivo.

In summary, TRAIL interacts with five receptors: the four membrane-bound TRAIL receptors, TRAIL-R1 to TRAIL-R4, and the soluble receptor OPG. TRAIL is therefore the most promiscuous of all TNFSF members. The biological basis for this promiscuity is still unclear. Whereas TRAIL-R1 (DR4) and TRAIL-R2 (DR5, Apo-2, KILLER, TRICK2) are DD-containing receptors capable of triggering apoptosis, TRAIL-R3 and TRAIL-R4 cannot do so because of the lack of an intracellular DD.

TRAIL-R1 and TRAIL-R2 share 58% sequence homology, and thus far it has not been possible to identify distinct functions of one receptor versus the other. They both trigger apoptosis via the same pathway, and this pathway is even identical to the one described above for CD95. TRAIL-R3 lacks an intracellular domain and is inserted into the plasma membrane via a GPI anchor. TRAIL-R4 has a cytosolic domain, but there is only a truncated DD of 15 instead of 80 amino acids, which is not capable of inducing cell death. However, TRAIL-R4 can activate NF-κB. As mentioned above,

TRAIL-R3 and TRAIL-R4 are often referred to as decoy receptors because they were shown in some of the cloning papers to sequester TRAIL on over-expression, thereby inhibiting TRAIL-induced apoptosis. To exert this death-inhibitory effect, TRAIL-R3 and TRAIL-R4 would have to present with a higher affinity for TRAIL or be expressed at substantially higher levels than TRAILR1 and/or TRAIL-R2. However, this is not the case. Others have proposed a model in which TRAIL-R3 and TRAIL-R4 interact via a pre-ligand assembly domain to inhibit ligand binding. A third notion suggests that the NF-κB–inducing activity of TRAIL-R4 may antagonize the death signal. To summarize this, we are still pretty much completely in the dark regarding the actual function of these two receptors in the biology of TRAIL, and not much progress has been made in the understanding of their function since they were cloned more than a decade ago. It will be important to study their function under non–over-expression conditions to uncover their physiologic role in TRAIL biology.

The biggest conundrum is still the difference between the TRAIL and the CD95 system with respect to the differential outcome of the in vivo application of agonists to CD95 as compared with agonists of TRAIL-R1 and/or TRAIL-R2. Despite the fact that no significant differences in the signaling pathways triggered by these two systems have been discovered to date, the outcome of their stimulation by systemic application of CD95 versus TRAILR1/2 agonists could not be more disparate. It remains one of the mysteries in apoptosis research today what the biochemical basis for this difference is, and it will be highly rewarding to identify its cause because it is likely to open the door to a more targeted application of TRAIL receptor agonists in specific cancer patients or patient groups.

With respect to this novel class of cancer drugs, apart from Apo2L/TRAIL itself, a total of five TRAIL-R2– and one TRAIL-R1–specific monoclonal antibodies are being developed. Many of them are already in various phase II clinical trials for the treatment of different cancer entities. This topic has recently been reviewed elsewhere, and we will therefore not go into further detail here (Johnstone et al., 2008; Papenfuss et al., 2008; Ashkenazi, 2008). Nevertheless, we would like to highlight one important and somewhat troubling aspect of these current trials. The one big absentee from the current clinical trials with TRAIL receptor agonists is a set of biomarkers guiding the selection of specific combinations of drugs that would be most likely to be effective in individual cancer patients who present with a particular genetic make-up of their cancer. To provide such (sets of) biomarkers for future trials and ultimately