Anoikis mechanisms.

Abstract

Anoikis is defined as apoptosis that is induced by inadequate or inappropriate cell-matrix interactions. It is involved in a wide diversity of tissue-homeostatic, developmental and oncogenic processes. The central problem of anoikis is to understand how integrin-mediated cell adhesion signals control the apoptotic machinery. In particular, the initiation of the caspase cascade in anoikis remains to be explained.

Full text

Page 1

555
Anoikis is defined as apoptosis that is induced by inadequate
or inappropriate cell–matrix interactions. It is involved in a wide
diversity of tissue-homeostatic, developmental and oncogenic
processes. The central problem of anoikis is to understand
how integrin-mediated cell adhesion signals control the
apoptotic machinery. In particular, the initiation of the caspase
cascade in anoikis remains to be explained.
Addresses
The Burnham Institute, 10901 North Torrey Pines Road, La Jolla,
California 92037, USA
*Correspondance: e-mail: sfrisch@burnham-inst.org
Current Opinion in Cell Biology 2001, 13:555–562
0955-0674/01/$ — see front matter
© 2001 Elsevier Science Ltd. All rights reserved.
Abbreviations
DLC1
EGF
EGFR
ERK
FADD
FAK
IGFRs
IAP
ILK
JNKs
PAK
PDGF
PKA
PI3K
SGKs
dynein light chain 1
epidermal growth factor
EGF receptor
extracellular signal-regulated kinase
fas-associated death domain
focal adhesion kinase
insulin-like growth factor receptors
inhibitor of apoptosis
integrin-linked kinase
Jun N-terminal kinases
p21-activated kinase
platelet-derived growth factor
cAMP-dependent protein kinase A
phosphoinositide-3 kinase
serum- and glucorticoid-inducible kinases
Introduction
Normal cell and tissue homeostasis reflects a dynamic bal-
ance of cell proliferation, differentiation and apoptosis.
Anoikis — the subset of apoptosis triggered by inadequate or
inappropriate cell–matrix contacts — maintains the correct
cell number of high-turnover epithelial tissues. The clearest
evidence for this is that the breakdown of anoikis contributes
to neoplasia. Accordingly, this process is discussed first.
Recent evidence for a role of anoikis resistance
in malignancy
Anoikis was first documented in both epithelial cells —
the precursors of most human cancers — and endothelial
cells. In these early reports [1,2], the expression of certain
oncogenes was shown to render normal epithelial cells
resistant to anoikis. More recent reports confirm that the
breakdown of anoikis contributes prominently to the
malignancy of mammary and colon cancers [3–5]; a similar
role has recently been reported for lung carcinomas (see
Now in press).
The breakdown of anoikis is expected to confer a selective
advantage upon pre-cancerous epithelial cells, affording
them an increased survival time in the absence of matrix
attachment, facilitating eventual reattachment and
colonization of secondary sites. Interestingly, a second
selective advantage has been revealed by two recent
studies [6,7]. In A431 epidermal carcinoma cells [6] and
MDA-MB468 mammary carcinoma cells [7], stimulation
with epidermal growth factor (EGF) induced either cell
cycle progression or cell rounding, which triggered anoikis.
This observation suggests that excessively stimulated
growth factor signaling pathways, which commonly occur
in tumor cells, may cause cytoskeletal perturbations capa-
ble of initiating anoikis. This phenomenon would provide
a strong and immediate selective pressure for cells to
become resistant to anoikis, perhaps by oncogene activa-
tion. Moreover, it may explain why several individual
oncogenes simultaneously activate survival and cytoskele-
ton-altering pathways (e.g. ras.) These specific examples
also underline the importance of coupling integrin- and
growth-factor-signaling pathways for producing the appro-
priate cellular response.
Surprisingly, resistance to anoikis also appears to be
involved in the evolution of certain non-epithelial cancers
such as melanoma [8,9] and neuroblastoma [10], where
de-differentiation confers resistance. These results sug-
gest that the circumvention of anoikis may be involved
not only in carcinomas but also in other commonly occur-
ring human tumors, although the mechanisms leading to
aberrant survival may vary greatly among these cell types.
Role of protein kinase signaling pathways in
anoikis
Numerous kinase/phosphatase signaling molecules have been
implicated in anoikis as central regulators. Because ras activa-
tion prevents anoikis [1] and integrins can stimulate various
aspects of the ras pathway, anoikis research has focused on the
two major ras effectors, the kinases PI3K (phosphoinositide-3
kinase) and raf. Rather than cataloguing the signaling mole-
cules that we may infer to control anoikis, the next section
focuses on some recent findings regarding selected signaling
pathways; it is not intended to be comprehensive.
Phosphoinositide-3 kinase-related signaling
Akt is involved in diverse survival signaling scenarios, includ-
ing anoikis [11]. The three primary integrin signaling
molecules that have been linked to cell survival are FAK
(focal adhesion kinase), Shc and ILK (integrin-linked kinase)
[12-14]. Each of these proteins may impinge upon the
PI3K/Akt pathway, although recent evidence shows that they
can also signal by distinct mechanisms. The evidence follows.
ILK interacts with the cytoplasmic tails of β1 and β3 integrin
subunits and is activated transiently by cell–matrix adhesion or
growth factor stimulation (the latter presumably through a
Nck- and ILK-interacting adaptor called Pinch.) When
Anoikis mechanisms
Steven M Frisch* and Robert A Screaton

Page 2

overexpressed in cell lines, ILK activates Akt activity either
direct-ly or indirectly. Inactivation of the phospholipid phos-
phatase PTEN, which occurs frequently in prostate cancer cell
lines, constitutively activates both ILK and Akt (both kinases
possess lipid-binding pleckstrin homology (PH) domains,
which probably explains this activation).Interestingly, transfec-
tion of a dominant-negative ILK into PTEN-deficient prostate
carcinoma cells or treatment of these cells with a specific
chemical inhibitor of ILK decreases serine-473 phosphoryla-
tion of Akt and, in turn, its kinase activity [15]. This implicates
ILK as an in vivo activator of Akt, which was recently
confirmed biochemically [16]. Overexpression of ILK can also
suppress aniokis in certain epithelial cell lines [17].
FAK may also activate Akt through direct PI3K activation as
well as indirectly through a p130cas-crkII-DOCK180-rac
pathway [18]. To date, only FAK has been demonstrated to
regulate the expression of caspase inhibitors of the IAP
(inhibitor of apoptosis) family [19], by a proposed mechanism
involving PI3K/Akt activation of the NF-κB pathway. Future
experiments will determine whether or not IAPs are induced
in response to the other integrin signaling molecules.
Although Shc is thought to activate the MAP kinase path-
way through grb2, recent results indicate that Shc is also a
potent PI3K/Akt activator [20]. In hematopoietic cells,
activated Shc recruits gab2 through grb2. Gab2 in turns
556Cell-to cell contact and extracellular matrix
Figure 1
Cytoskeletal alterations potentially involved
in anoikis. Diagrams illustrating the
alterations in (a) attached cells and
(b) suspended cells are shown. The various
signaling molecules that are controlled by
integrins and growth factor receptors have
been omitted, so as to focus on cytoskeletal
functions. (a) In the attached cells, the
integrins are complexed with extracellular
matrix and growth factor receptors, and the
three intact cytoskeletal systems are linked
through plectin. They sequester the BH3
domain proteins Bim and Bmf, the JNK
pathway components JNK1, SEK1 and
MLK2, and p150-Spir is able to stabilize
actin polymers through Arp2/3; TNFR2 (not
shown) may be kept inactive through
interaction with intermediate filaments. Most
of the caspase-8 is bound to mitochondria,
as discussed in the text. (b) Upon
suspension, the initiating signal for anoikis is
not yet known. However, cytoskeleton of all
types are presumably perturbed, which may
cause the release of Bmf from actin, thus
neutralizing bcl-2, and allowing cytochrome c
release from mitochondria. Accompanying
this, caspase-8 is activated, which promotes
further cytochrome c release through Bid,
plectin cleavage and activation of effector
caspases. These in turn cleave filamin and
cytokeratin intermediate filaments. The fate
of actin polymers after detachment is not
well understood but activation of the JNK
pathway may inhibit Arp2/3 function through
p150-Spir, leading to disassembly.
IFs
Integrin
GFR
Actin
Microtubule
MLK2
SEK1
JNK
M
Filamin
bcl-2
c8
c8
c8
Plectin
Bmf
bcl-2
Arp2/3
Cytochrome c
Apoptosis
Suspended
p150
Current Opinion in Cell Biology
Attached
IFs
Actin
MLK2
JNK
Filamin
SEK1
Plectin
Bmf
bcl-2
M
c8
p150
Microtubule
IntegrinGFR
(a)
(b)

Page 3

recruits the regulatory subunit of PI3K, p85, leading to
PI3K activation. Whether an analagous pathway pertains to
integrin signaling — wherein Shc is recruited to integrins
via fyn and caveolin — will be interesting to test.
Growth factor inputs into integrin-mediated
phosphoinositide-3 kinase/Akt survival signaling
As insulin and IGF receptors are major activators of Akt
activity and their ligands are critical survival factors for
numerous cell types, several laboratories have addressed
the role of these factors in anoikis. Two studies demon-
strated that although IGFs protected serum-starved mouse
embryo fibroblasts [21] and LNCap prostate carcinoma
cells [22] from anoikis, they did not inhibit anoikis in
primary mammary epithelial cells — even though the
mammary cells normally require insulin for survival when
attached to the extracellular matrix [23]. When mammary
cells attached to collagen (a non-permissive matrix
required for their survival) were stimulated with insulin,
the insulin receptor was still capable of autophosphoryla-
tion. Further signaling, however, was blocked: the receptor
was not able to phosphorylate and recruit the adapter
IRS-1 or activate downstream PI3K/Akt survival signaling
[23]. Thus, in this cell system, integrin ligation is a prereq-
uisite for insulin signaling. Whether or not similar
requirements exist in animal models remains to be con-
firmed. Nevertheless, the interdependence of integrins
and certain growth factor receptors (insulin receptor,
PDGF (platelet-derived growth factor) receptor, and the
EGF receptor) warrants further examination of the possi-
bility, in some systems at least, that anoikis is partly due to
an inhibition of growth factor signaling.
Interestingly, EGFR and integrins form complexes in cer-
tain cell types in which extensive cross-activation of the
two receptor types is observed [24,25]; uncoupling cooper-
ative signaling by growth factor and matrix receptors could
dysregulate cell survival by dissociating growth factor sig-
naling from cytoskeletal alterations. It is in this context
that the prevalence of FAK overexpression in human
tumor cells [26] — which prevents anoikis [27] — may be
appreciated. In the case of the PDGF receptor, the inter-
action between the receptor and integrin is extracellular,
raising the possibility of a direct collaboration or competi-
tion between matrix molecules and the growth factor for
receptor binding [28].
In the LNCap system, IGF levels are reported to dissipate
slowly over the course of 16 hours of cell detachment,
which is immediately followed by anoikis [22]. The
authors attribute the observed anoikis in LNCap cells to
this decrease in IGF. Moreover, the absence of IGFs was
proposed to initiate a decline in Akt activity and the
following cascade of events: activation of Gsk-3,
phosphorylation-driven degradation of cyclin D, Rb hyper-
phosphorylation and consequent repression of the IGF
promoter. With the IGF promoter repressed, the cells are
committed stably to anoikis. Confirmation of such a model
for anoikis in LNCap cells will require a direct assessment
of the levels of IGF protein itself during cell detachment
(because only mRNA levels were examined). As loss of
membrane integrity frequently occurs late in or after the
completion of apoptosis in most systems, the release of
transfected luciferase protein from apoptotic LNCap cells
may not accurately reflect the kinetics of anoikis.
Furthermore, cellular transcription and translation are
largely inhibited in suspended cells. Given that the
kinetics of caspase activation (and the consequent
pro-apoptotic effects) following cell detachment are
generally more rapid than transcriptional or translational
alterations, the impact of transcriptional changes on
anoikis may be secondary. A precise determination of the
role of Akt activity in anoikis is complicated by the fact
that Akt itself is a target for activated caspases [29]. Thus
the loss of Akt activity may be a consequence and not a
cause of anoikis. Nevertheless, the role of soluble survival
factors in anoikis warrants further examination.
Akt phosphorylates regulators of apoptosis such as (human)
caspase-9 and Bad (reviewed in [30]), as well as transcrip-
tion factors whose target genes program the survival
phenotype (e.g. Forkhead factors). However, Akt-mediated
survival is likely to involve as yet unidentified novel Akt
substrates as well as kinases possessing Akt-like activity.
Notably, Akt is also required for the activation of p21-acti-
vated kinases (PAK) by ras [31•], thus PAK itself or some
PAK activator may be a critical new Akt substrate. Several
anti-apoptotic effects of PAK — which, like Akt, can phos-
phorylate BAD [32•] — have been reported (e.g. [33]; see
section below.) Forkhead transcriptional regulators are also
substrates for a set of Akt-related kinases known collective-
ly as serum- and glucorticoid-inducible kinases (SGKs);
both Akt and SGK1 can inhibit Forkhead function by pro-
moting cytoplasmic accumulation of the phosphorylated
Forkhead product [34]. Whether the other Akt substrates
are also SGK substrates, and the relevance of SGKs to
anoikis, remains to be determined.
Raf–ERK-related signaling
There is much evidence to implicate ERK (extracellular
signal-regulated kinase) activation in both cell cycle pro-
gression and survival (e.g. [35]). It is well established that
raf-1 function is critical for ERK activation; a possible alter-
native function and novel activation mechanism for raf-1
are considered here.
The Bcl-2-interacting protein BAG-1 activates raf-1 kinase
activity and targets it to mitochondria, where its survival effect
is apparently enhanced [36]. Interestingly, BAG-1 also interacts
with the p53-inducible inhibitor of the ras–raf–MAP-kinase
pathway, Siah-1, and prevents Siah-1 from inducing growth
arrest [37]. Whether this phenotypic effect of Siah-1 is depen-
dent on association with raf-1 through BAG-1 is not yet clear.
In any event, the existence of critical mitochondrial raf-1
substrates is strongly implied by these results.
Anoikis mechanisms Frisch and Screaton 557

Page 4

Although the precise mechanism of raf-1 kinase activation
remains unclear, it is known that the PAKs, whose role in
integrin signaling is well recognized, can activate raf-1 by
direct phosphorylation of serine 338 [38]. PAK activity is
controlled both by phosphoinositides and PI3K, as the for-
mer interact with the PH domains of the rac/cdc42 GEFs,
and the latter may directly activate PAK through the adap-
tor protein PIX [39]. Integrin ligation to fibronectin can
stimulate raf phosphorylation at serine 338 by this path-
way, which provides an interesting new mechanism for the
control of raf-1 activity by integrin stimulation of PI3K
activity. It may also explain the critical involvement of
PAKs in the stimulation of ERK activity by integrins.
Recently, an indirect mechanism for raf-mediated survival was
indicated: the induction of transforming growth factor-α
(TGF-α) expression [40]. In MCF10a epithelial cells, a signif-
icant increase in TGF-α was noted in response to expression of
a tamoxifen-inducible raf-1 protein. Exogenous addition of this
cytokine sufficed for anoikis rescue, and inhibition of the
TGF-α receptor with AG1478 abrogated rescue by raf or
TGF-α. Anoikis rescue by EGF has also been demonstrated in
keratinocytes [41]. As raf-1 can in turn be activated by EGFRs,
this result implicates a positive feedback loop that promotes
cell survival involving raf and autocrine growth stimulation.
Finally, the ERK pathway is also affected by the state of the
actin cytoskeleton. In 3T3 cells, both detachment and treat-
ment of attached cells with cytochalasin D is sufficient to
prevent ERK translocation into the nucleus — and the con-
sequent phosphorylation of its target, elk-1 — in response to
mitogenic stimulation [42•]. It is difficult to reconcile the
adhesion requirement for ERK translocation with the abili-
ty of EGF to rescue cells in suspension from anoikis,
although cell-type differences between these studies may
contribute. Nevetheless, this highlights the important
challenge of elucidating how integrins control ERK nuclear
import or export through cytoskeletal regulation.
Jun N-terminal kinases
Numerous reports have suggested both pro- and anti-apop-
totic roles for Jun N-terminal kinases (JNKs), creating some
degree of complexity in the field. Early reports indicated
that detachment of epithelial cells from matrix strongly and
rapidly induces the activation of JNKs, which may con-
tribute to anoikis [43]. Recent evidence supporting a
pro-apoptotic role for JNKs comes from the observation that
mouse embryo fibroblasts lacking both JNKs 1 and 2 (due to
a compound knockout) were resistant to UV-, anisomycin-
and DNA damage-induced apoptosis [44]. These effects
correlated with a failure to release cytochrome c from mito-
chondria or activate Bid. Interestingly, anoikis also involves
mitochondrial cytochrome c release [45]. Perhaps determin-
ing the effect, if any, of the JNK1 and 2 gene knockouts on
cellular response to matrix detachment will clarify this issue.
In contrast to previous reports, Almeida and co-workers [46]
observed that serum-starved fibroblasts activated JNKs in
response to attachment to, rather than detachment from,
the matrix (fibronectin). In these experiments, as reported
previously, JNK activation was stimulated by cotransfection
of FAK, accompanied by enhanced survival. Thus, a sur-
vival pathway from integrins — through FAK — to JNKs
was proposed. There is evidence to suggest that JNKs are
in fact involved in normal cell cycle progression in fibrob-
lasts: adherent 3T3 cells show a spike of JNK activity in the
G1phase of the cell cycle, and transfection of a dominant-
negative version of MKK4 — a positive upstream regulator
of JNK activity — causes 3T3 cells to growth-arrest in G1
[47]. In this study, JNK activity was stimulated both by
attachment to the matrix (in HUVEC, 3T3 and 293 cells)
and by FAK (in 293 cells), further implicating JNKs in
adhesion-dependent cell cycle progression and survival.
JNK was assayed immediately after detachment (time zero)
versus various time points after reattachment. However,
because JNK activity is stimulated rapidly by detachment
[43] and cells reattach and spread slowly, it is possible that
the stimulation of JNK observed in these [47] and other
studies [48] was due to detachment — indicating a
proapoptotic effect of JNKs — rather than reattachment.
Clearly, JNK regulation is significantly different in fibrob-
lasts versus epithelial cells: in fibroblasts but not in
epithelial cells, ras stimulates JNK activity, and cell cycle
progression by dominant-negative MKK4 is inhibited. In
fact, anoikis is likely to proceed by different mechanisms
in these two cell types altogether, given that fibroblasts
must be deprived of growth factors in order to respond to
matrix detachment. This might reflect the different
circumstances under which fibroblasts and epithelial cells
have evolved to undergo anoikis in vivo. Differences
amongst epithelial cell lines may exist as well.
Role of the cytoskeleton
The readily apparent differences between the cytoskeletal
structures of attached versus suspended cells suggest that
survival signaling in anoikis is likely to be extensively regu-
lated by cytoskeleton. Such regulation may be affected by
the multiple cytoskeletal changes apparent in transformed
cells [49]. Indeed, substantial evidence now exists that both
signaling molecules and apoptosis regulators are associated
with the cytoskeleton, and as such may together regulate
anoikis by serving as sensors of cytoskeletal integrity.
For example, several interesting connections exist between
the cytoskeleton and the JNK pathway. First, MKK4/SEK1
interacts directly with the integrin-associated cytoskeletal
protein ABP280/filamin [50]. Filamin-deficient melanoma
cells are refractory to JNK stimulation, and reconstitution of
filamin rescued JNK activity, an observation that implicates
the cytoskeleton in the control of JNK activation. Moreover,
filamin is a caspase substrate, and cleavage of filamin may
play a role in the regulation of JNKs [51]. Second, JNKs may
control cytoskeletal arrangement by direct association with a
novel WASP family member (a family of effectors for rho-
like GTPases that regulate actin cytoskeleton), p150Spir
558Cell-to cell contact and extracellular matrix

Page 5

[52], with which activated JNK collaborated in transfection
experiments to cause cytoskeletal rearrangements. Third,
MLK2, a MAP kinase kinase kinase that activates JNK,
colocalized with JNK along microtubules and associated
with the kinesin superfamily motor KIF3 [53].
Another interesting cytoskeletal connection involves two
potently proapoptotic Bcl-2-family proteins of the ‘BH3-
only’ class, Bim and Bmf [54,55•,56]. In untreated cells, Bim
protein is apparently sequestered by microtubule-associat-
ed dynein light chain 1 (DLC1). When released from this
sequestration upon treatment with microtubule-disrupting
agent taxol, Bim interacts directly with Bcl-2 and stimulates
release of cytochrome c from mitochondria. Analogously,
Bmf — which was recently isolated by virtue of its interac-
tion with the Bcl-2 homolog Mcl-1 — is sequestered by the
actin/myosin-associated dynein light chain-2 (DLC-2) in
MCF7 cells. Cell suspension, cytochalasin treatment or
UV-irradiation all induce the release of Bmf from DLC-2,
which allows Bmf to complex with — and presumably neu-
tralize — Bcl-2. Interestingly, release of both Bim and Bmf
occurred independently of caspase activity. Thus, these
proteins may directly regulate the initial stages of apoptosis,
and perhaps anoikis, by serving as sensors for microtubule
and actin cytoskeleton integrity, respectively. Future func-
tional and genetic studies will help to establish the precise
role of these proteins in anoikis.
Another conceptually attractive connection between the
actin cytoskeleton and apoptosis is provided by the actin
severing/capping protein gelsolin [57]. Recent in vitro
experiments with isolated mitochondria have indicated
that full-length gelsolin prevents the Bax-stimulated
release of cytochrome c, whereas caspase-cleaved gelsolin
promotes apoptosis by an as yet unknown mechanism.
As will be discussed in the next section, there is emerging
evidence that some components of the anoikis pathway may
overlap with those of death receptor pathways.In this context,
it is interesting to note that mouse embryonic fibroblasts iso-
lated from keratin-8–/–mice are about 100-fold more sensitive
to killing by tumor necrosis factor (TNF) than normal cells
[58•]. An association between TNF receptor 2 and cytoker-
atins has been implicated in this effect, and these results
suggest that keratin intermediate filaments can serve to atten-
uate certain death receptor responses. Preliminary results
(S Frisch, R Oshima, unpublished data) also indicate that
these keratin 8-knockout cells are also sensitized to anoikis.
A final connection is provided by the cytoskeletal protein
plectin, which forms links with all three major cytoskeletal
systems (microtubules, intermediate filaments and actin).
Recently, the majority of procaspase-8 (the inactive
holoenzyme) was found associated with mitochondria in
unstimulated MCF-7 cells [59]. Remarkably, stimulation
of apoptosis with FAS ligand caused immediate processing
of mitochondrial procaspase-8, after which the active sub-
units of the enzyme translocated to and cleaved plectin.
These events occurred well before the cleavage of well
established initiator caspase targets could be detected,
including PARP poly (ADP-ribose) polymerase, lamin or
effector caspases. Moreover, embryonic fibroblasts from
plectin-deficient mice showed attenuated cytoskeletal
reorganization following FAS ligand treatment, indicating
that plectin degradation is a watershed event. It will be
interesting to test the plectin–/– cells for anoikis sensitivity
to elucidate a possible role for plectin.
Role of death receptors in anoikis
One of the major outstanding questions in anoikis research
is how the caspase cascade is initially activated by simple
detachment of cells from the matrix. One hypothesis is
that death receptors somehow become activated, either
through their propensity for self-association — which may
suffice for signaling — or by an interaction with endoge-
nous death ligands.
Recent results indicate that the death receptor adaptor
molecule FADD may be involved in anoikis, as a domi-
nant-negative truncated FADD containing only the death
domain inhibited anoikis [60,61]. This result does not
prove, however, that death receptors themselves are
involved in anoikis — only that death domains are.
Nevertheless, anoikis was also accompanied by an early
activation of caspase-8, as would be expected for death
receptor activation.
However, extensive attempts to document the assembly of
death-inducing signaling complexes (DISCs) in detached
epithelial cells have so far proven disappointing (S Frisch,
R Screaton, unpublished data). This suggests either that a
specific FADD-requiring death receptor that has not yet
been assayed is involved or that death receptors are not
involved at all. This remains to be resolved.
Recently — in support of the death receptor hypothesis —
it has been reported that anoikis in HUVECs requires inter-
actions between the endogenous death receptor FAS and its
cognate ligand, FAS ligand [62]. In this study, anoikis was
inhibited by FAS- or FAS-ligand-blocking antibodies, and
FAS was upregulated and associated with FADD after
detachment. However, our results (S Frisch, R Screaton,
unpublished data) indicate that HUVECs are devoid of
detectable FAS ligand protein. Moreover, anoikis of
HUVECs could not be blocked by anti-FAS antibodies, and
FAS is neither upregulated nor associated with FADD after
detachment; the reason for this discrepancy is not yet clear.
In summary, a role for death receptors in anoikis has yet to
be established. Interestingly, in T cells FAS is tightly asso-
ciated with the actin cytoskeletal protein ezrin, which is
required for FAS ligand responses of the receptor [63]. If
an analogous interaction occurs in other cell types, this may
provide a novel mechanism for the cytoskeleton to control
death receptor function positively (in detached cells) or
negatively (in attached cells).
Anoikis mechanisms Frisch and Screaton 559