known members of the TNFR family act as receptors exclusively for members of the
TNF family, with p75NGFR being the only exception (see below). Like the TNF
family, the TNFR family is extensively discussed in two recent reviews (Locksley et
al., 2001; Bodmer et al., 2002); see also Table 2 for a list of known ligand/receptor
combinations. While VEGI is the only known orphan ligand, there exist at least four
orphan receptors (TRAMP/DR3, DR6, Troy/Taj, and RELT). Most members of
the TNFR family are integral membrane proteins of type I, bearing an N-terminal
signal sequence, followed by a more or less conserved ectodomain, a transmembrane
region, and a highly variable C-terminal signaling domain exposed to the cytoplasm.
Several TNFR proteins deviate from this architecture. Osteoprotegerin (OPG) and
Dcr3 are not anchored to the membrane at all, and TRAIL-R3/Dcr1 uses a GPI
anchor. These three proteins lack a signaling domain and thus are considered decoy
receptors. XEDAR, BCMA, TACI, and BAFFR are type III membrane proteins with
a topology equivalent to type I, but lacking a recognizable N-terminal signal sequence.
A defining feature of all members of the TNFR family is the presence of a particular
type of cysteine-rich domain (CRD), which is involved in ligand binding. The CRDs
are pseudo-repeats of 30–40 residues, which typically contain six Cys-residues
forming three disulfide bonds. The known TNFR members contain between one
and five copies of the CRD, which constitute the major part of the receptor’s
ectodomain. The CRDs can be classified into several subfamilies by virtue of their
cysteine spacing (Naismith and Sprang, 1998). Some atypical CRDs also exist, but
can be very hard to detect by sequence analysis methods due to their shortness and
high sequence divergence. Several three-dimensional structures of members of the
TNFR family have been solved, two them showing the receptor ectodomain in
complex with its ligands: TNFR/TNF (Banner et al., 1993) and DR5/TRAIL
(Hymowitz et al., 1999).
Members of the TNF ligand family form constitutive trimers, and there is little
doubt that an active signaling complex involves at least a ligand trimer bound by a
receptor trimer. It is less clear whether the receptors also have a tendency to trimerize
in the absence of the ligand (Bodmer et al., 2002). If this is the case, the ligand binding
would have to change the conformation of the intracellular region of the receptors
in order to elicit a signal. If, by contrast, the receptor trimerization is induced by
ligand binding, the induced proximity of the signaling domains would be the likely
trigger. It is this intracellular part of the receptor that decides whether it signals
apoptosis or another cellular event, such as NF-κB activation. In most receptors, the
cytoplasmic C-terminal region is very dissimilar and no sequence motifs can be
detected. However, there is one subset of receptors that harbors a well-defined
homology domain in this region, the ‘death domain’ (DD). This subset is responsible
for apoptosis induction and its members are frequently referred to as ‘death receptors’
(DRs). The death domain will be discussed in detail in the next paragraph. Currently,
at least five functional death receptors are known: Fas, TNF-R1, TRAMP/DR3,
TRAIL-R1/DR4, and TRAIL-R2/DR5. Two more death receptor candidates are
DR6 and EDAR, whose role in apoptosis has not yet been established. A further
protein, NGFR follows the typical death receptor architecture but is unusual in two
FUNCTIONAL DOMAINS IN APOPTOSIS PROTEINS 77