1
Exploring the TRAILs less travelled:
TRAIL in Cancer Biology and Therapy
Silvia von Karstedt
1, 2
*, Antonella Montinaro
1,
* and Henning Walczak
1
1
Centre for Cell Death, Cancer and Inflammation, UCL Cancer Institute, University College London, 72
Huntley Street, London WC1E 6DD, UK
2
Current address: The Francis Crick Institute, 1 Midland Road, London NW1 1AT, UK
* These authors contributed equally
Correspondence to H. W. e-mail: h.walczak@ucl.ac.uk
2
The discovery that the tumour necrosis factor-related apoptosis-inducing ligand (TRAIL) can induce
apoptosis of cancer cells without causing toxicity in mice has led to the in-depth study of pro-apoptotic
TRAIL receptor (TRAIL-R) signalling and the development of biotherapeutic drug candidates that activate
TRAIL-Rs. The outcome of clinical trials with these TRAIL-R agonists has, however, been disappointing so
far. Recent evidence indicates that many cancers, in addition to being TRAIL-resistant, employ the
endogenous TRAIL–TRAIL-R system to their own advantage. However, novel insight at two fronts: how
resistance of cancer cells to TRAIL-based pro-apoptotic therapies might be overcome, and how the pro-
tumourigenic effects of endogenous TRAIL might be countered, gives reasonable hope that the TRAIL
system can be harnessed to treat cancer. In this review we assess the status quo of our understanding of
the biology of TRAIL–TRAIL-R system – as well as the gaps therein – and discuss the opportunities and
challenges in effectively targeting this pathway.
3
Although unknown at the time, tumour necrosis factor (TNF) first entered the world stage of cancer
therapy towards the end of the 19
th
century when William Coley found that sarcomas shrunk with
certain bacterial infections
1,2
. It was not until well into the 20
th
century that this effect was found to be
due to the induction of TNF,
3
which caused tumours to become necrotic, a feature that coined the name
of the protein. Initial enthusiasm following the discovery of TNF was, however, dampened by the
demonstration that systemic TNF treatment induced a lethal inflammatory shock syndrome
4
. In search
for another molecule with similar anti-tumour properties, the attention turned to CD95 (also known as
FAS and APO-1) , a receptor homologous to TNF receptor 1 (TNFR1) and TNFR2 that can potently trigger
apoptosis in many cancer cells
5,6
. However, systemic treatment with CD95 agonists led to fulminant liver
toxicity in mice within hours of treatment
7,8
, again excluding a TNF-like molecule for therapeutic use.
Third time lucky: another TNF superfamily (TNFSF) member termed TNF-related apoptosis-inducing
ligand (TRAIL, also known as TNFSF10 and APO2L) was discovered a few years later
9, 10
, and this factor
was capable of killing tumour cells, importantly however without causing the lethal adverse effects
encountered with TNF or CD95 agonists
11,12
.
Although these promising findings resulted in the development of TRAIL-receptor (TRAIL-R) agonists for
clinical use, this happened at a time when toxicity of pro-apoptotic TNF-like factors in general, but also
of TRAIL specifically
13,14
, was a concern as some recombinant forms of TRAIL had shown potential for
liver toxicity at high doses
13-15
. Moreover, when the decision to take particular molecular entities
forward for clinical development was made, the biology of TRAIL and its receptors in cancer, as well as
inflammation and immunity, was still underexplored and could therefore not adequately be taken into
consideration. Since then, this has substantially changed. It is therefore timely to take a step back and
revise our current understanding of the biology of the TRAIL–TRAIL-R system in order to come forward
with novel and effective therapeutic strategies harnessing this system for cancer therapy.
[H1] The TNF superfamily and TRAIL–TRAIL-Rs
[H3] The TNF superfamily
TNF is the canonical member of the TNFSF of which TRAIL and the CD95 ligand (CD95L, also known as
FASLG and APO-1L) are closely related members. Apart from lymphotoxin-α (also known as TNFSF1) and
vascular endothelial growth inhibitor (also known as TNFSF15), which are encoded as soluble proteins,
all other members of this family are encoded, and if not further cleaved, expressed as type II
transmembrane proteins
16
. Some members, including TNF, CD95L and TRAIL, can subsequently be
released from the cell surface through the action of proteases, and therefore can occur as both
membrane-bound and soluble proteins. The proteases ADAM10 (a disintegrin and metalloproteinase
domain-containing protein 10,) and ADAM17 (also known as TACE) have been identified to cleave CD95L
and TNF to generate their respective soluble forms in a process termed shedding
17,18
. The generation of
soluble TRAIL through shedding also involves cysteine protease activity
19
, but the identity of the
responsible protease(s) remains unknown. Soluble TRAIL is present in the plasma of a healthy adult at
approximately 100 pg/ml
20
, a concentration at which TRAIL fails to induce apoptosis in most cell lines in
vitro
21
.
4
For CD95L, only the membrane-bound protein can induce apoptosis whilst the soluble form has cancer-
promoting effects
22
. For TRAIL, this is less clear. It has, however, been shown that liposome-bound
TRAIL, which mimics membrane association, is more active in killing cancer cells than its soluble
counterpart
23,24
. In this context it is conceivable that recombinant forms of TRAIL that comprise the
extracellular domain fused to motifs that enable stabilisation and multimerisation might mimic the
membrane-bound conformation. This is likely the reason why different recombinant forms which
contain such motifs
11,13,25
are, by several orders of magnitude, more potent inducers of apoptosis than
recombinant TRAIL preparations that lack such additional motifs
12
.
TNFSF members bind to a corresponding family of receptors, referred to as the TNFR superfamily
(TNFRSF), which comprises more members than the TNFSF. Hence, some ligands have several receptors.
Eight TNFRSF members, including TNFR1 (also known as TNFRSF1A), CD95, TRAIL-R1 (also known as DR4
and TNFRSF10A) and TRAIL-R2 (also known as DR5 and TNFRSF10B)
26
contain an intracellular domain
required for cell death induction, consequently referred to as the death domain (DD).
[H3] The TRAIL–TRAIL-R system
Amongst the TNFSF, human TRAIL is unique in that it binds four membrane receptors and one soluble
receptor (Figure 1a). The human TRAIL-Rs can be subdivided into two classes: the full-length intracellular
DD-containing receptors TRAIL-R1
27
and TRAIL-R2,
28-34
which are capable of inducing apoptosis and are
most widely expressed, and the alternative receptors TRAIL-R3 (also known as DCR1 and TNFRSF10C)
33-36
TRAIL-R4 (also known as DCR2 and TNFRSF10D)
37,38
and osteoprotegerin (OPG, also known as
TNFRSF11B), which also functions as a soluble receptor for receptor activator of nuclear factor-κB ligand
(RANKL, also known as TNFSF11)
39
. TRAIL-R3 is glycosylphosphatidylinositol (GPI)-anchored to the
plasma membrane, hence lacks an intracellular domain, and TRAIL-R4 contains a cytoplasmic domain
capable of inducing nuclear factor-κB (NF-κB) activation but not apoptosis as it only encodes a truncated
DD. At 37°C, TRAIL binds TRAIL-R2 with higher affinity than the other membrane-expressed TRAIL-Rs
40
. It
is therefore likely that under physiological conditions binding to TRAIL-R2 would be favoured, especially
when endogenous TRAIL is limited.
All of the alternative TRAIL-Rs were proposed to act as TRAIL “decoys”, i.e. their binding to TRAIL would
lower the concentration of TRAIL available for binding to the pro-apoptotic receptors TRAIL-R1 and
TRAIL-R2 and, thereby, negatively regulate apoptosis induction by TRAIL. Whereas in-vitro
overexpression results and additional correlative data in favour of this concept were presented
41
, it
remains to be seen whether this function is indeed exerted by any of these receptors in cancer cells
under endogenous expression levels
42-45
.
When TNFR1 was first crystallised complexed with its ligand, it formed receptor trimers with a ligand
trimer located in its core
46
. However, when not bound by its ligand, TNFR1 formed dimers
47
. Similar to
TNFR1, TRAIL-Rs also exist as preassembled multimers. In the TRAIL–TRAIL-R system, however, receptor
dimers are ligand-induced and present in high molecular weight fractions together with ligand-induced
trimers
48
. Adding another level of complexity, TRAIL-R1 and TRAIL-R2 can homo- and heterotrimerise to
form higher-order complexes. It has been suggested that such complexes can either involve trimer
5
multimerisation or crosslinking of neighbouring trimers via dimerisation between receptor interfaces
which are located opposite of the ligand-binding interfaces resulting in a hexameric honeycomb-like
structure
48
. The latter model received support through two recent studies showing that non-stabilised,
untagged TRAIL synergised with TRAIL-R2-specific antibodies to kill cells and that this was achieved
through a ternary complex crystal structure resembling the above mentioned honeycomb
21,49
.
The most obvious molecular difference between the two DD-containing TRAIL-Rs is that there is only
one splice variant for TRAIL-R1 whereas there are two for TRAIL-R2
50
. The long isoform of TRAIL-R2
contains an additional 29 extracellular amino acids, which are located immediately adjacent to the
membrane. As this polypeptide, rich in threonine, alanine, proline and glutamine (TAPE), also referred to
as the TAPE domain
34
, is thought to form a rigid stalk as described for a highly homologous polypeptide
in TNFR2
51
, it is likely that its presence results in protrusion of the long isoform from the glycocalyx [G].
It is therefore tempting to speculate whether only TRAIL-Rs whose extracellular domains protrude at
similar stalk-dependent heights may effectively heterotrimerise. If that were the case, TRAIL-R1, TRAIL-
R4 and the short isoform of TRAIL-R2, in addition to forming homotrimers, would be capable of forming
heterotrimers amongst each other whereas the long isoform of TRAIL-R2 would only form homotrimers.
According to this model, with five consecutive repeats of the TAPE domain
34
TRAIL-R3 would hover high
above the other TRAIL-Rs and therefore also only form homotrimers.
Despite surface expression of TRAIL-R2 in cell lines derived from pancreatic cancer, chronic lymphocytic
leukaemia or mantle cell lymphoma, these cells only employ TRAIL-R1 for apoptosis induction by TRAIL
52-
54
. In addition, apoptosis induction via TRAIL-R2 requires crosslinking of untagged soluble TRAIL
55
implying that TRAIL-R2 might have a higher apoptotic threshold than TRAIL-R1 . In several leukaemia and
lymphoma cell lines, however, antibody-mediated TRAIL-R2 triggering appears sufficient to induce
apoptosis without additional crosslinking
56
. Together, these data highlight that human TRAIL-R1 and
TRAIL-R2 fulfil partly overlapping but also distinct functions, of which many remain to be discerned.
In contrast to humans, mice only express a single TRAIL-R (mTRAIL-R, also known as MK) with an
intracellular DD which shares almost the same level of identity with human TRAIL-R1 (43% sequence
homology) and human TRAIL-R2 (49% homology); like its human counterparts, mTRAIL-R is capable of
inducing apoptosis
57
. Two further mouse TRAIL-Rs (mDcTRAIL-R1, also known as TNFRSF23 and
mDcTRAIL-R2, also known as TNFRSF22) were later described, but these lack an intracellular DD
58
(Figure
1b). They differ substantially in their amino acid sequence from human TRAIL-R3 and TRAIL-R4 and do
not induce apoptosis or NF-κB activation upon overexpression
58
. Notably, human TRAIL binds only
weakly to mTRAIL-R whereas mouse TRAIL has high affinity for the human TRAIL-Rs
59
. These findings
need to be considered when designing tolerability studies in mice.
Although studies in mTRAIL-R-deficient mice have shed light on the relevance of many TRAIL-R-induced
pathways in vivo, it remains mysterious why humans have evolved to express two DD-containing
receptors for TRAIL. One option to study this question further would be to develop a “humanised”
mouse expressing human DD-containing TRAIL-Rs.
