and only speculations on the possible interaction surfaces exist (Eberstadt et al., 1998;
Weber and Vincenz, 2001a).
Functionally, all known death effector domains are involved in binding to other
death effectors. Despite their analogous fold and similar properties, no dimerization
between a death domain and a death effector domain has ever been described. The
reason for this specificity is still unclear and will probably remain so until the problem
of the interaction surfaces has been solved satisfactorily. Another mystery is the
stoichiometry of the DED interactions. Analogy to the case of the death domains
would suggest a trimerization of DEDs, an idea that is supported by the presence of
three death effector domains in the all-important interaction between FADD and
caspase-8.
3.3
The caspase recruitment domain (CARD)While the activation of caspase-8 was now understood as a consequence of the DED-
mediated recruitment to the FADD-containing receptor signaling complex, the
activation of the other, non-DED caspases remained unclear. Some light was shed
on this problem by the discovery of another adapter domain. The starting point was
the identification of a death-domain-containing adapter protein named RAIDD or
CRADD (Ahmad et al., 1997; Duan and Dixit, 1997). Similar to FADD,
ov erexp ressio n o f RAIDD i s able to induce apo pto sis, but, unli ke FADD, the RAIDD
protein does not bind directly to the death domain of a receptor but rather to the
RIP-containing signaling complex associated with the TNF receptor. A likely modus
operandi was immediately apparent from the properties of RAIDD’s N-terminal
region: it both has similarity to and interacts with the prodomain of caspase2 and
recruits this enzyme to the death-signaling complex (Ahmad et al., 1997; Duan and
Dixit, 1997). A further region of similarity was identified early on in the prodomain
of the C. elegans caspase ced-3, suggesting that this caspase can be activated in a similar
manner. The great importance of this homology domain became apparent later, again
by bioinformatical analysis, when regions with significant sequence similarity were
identified in many other caspases and several other proteins (Hofmann et al., 1997).
Since the general function of this new domain appeared to be the connection of
caspases to various upstream signaling complexes, the name CARD for ‘caspase
recruitment domain’ was proposed. Except for the downstream, or ‘executor’,
caspases, such as caspase-3, all mammalian caspases appear to have a cleavable
prodomain that comprises either two death effector domains or one CARD domain.
Following the paradigm of the structurally and functionally related death effector
domains, it was to be expected that the caspase-recruiting proteins should also contain
a CARD domain. This prediction has now been confirmed in several cases, APAF1
being the most important example (Zou et al., 1997). APAF1 was identified as a
component of the so-called apoptosome, a multiprotein complex working
downstream of the mitochondrium in apoptosis signal transduction (Zou et al., 1997;
1999). Nowadays, the function of this complex is fairly well understood: upon
FUNCTIONAL DOMAINS IN APOPTOSIS PROTEINS 83