MOLECULAR AND CELLULAR BIOLOGY, Feb. 2011, p. 766–782
Copyright © 2011, American Society for Microbiology. All Rights Reserved.
Vol. 31, No. 4
Phosphorylation of Trask by Src Kinases Inhibits Integrin Clustering
and Functions in Exclusion with Focal Adhesion Signaling?
Danislav S. Spassov,1,4Ching Hang Wong,1,4Natalia Sergina,1,4Deepika Ahuja,1,4Michael Fried,2,4
Dean Sheppard,1,3and Mark M. Moasser1,4*
Departments of Medicine,1Cancer Research Institute,2Lung Biology Center,3and Helen Diller Family Comprehensive Cancer Center,4
University of California, San Francisco, San Francisco, California 94143
Received 21 July 2010/Returned for modification 2 September 2010/Accepted 6 December 2010
Trask is a recently described transmembrane substrate of Src kinases whose expression and phosphoryla-
tion has been correlated with the biology of some cancers. Little is known about the molecular functions of
Trask, although its phosphorylation has been associated with cell adhesion. We have studied the effects of
Trask phosphorylation on cell adhesion, integrin activation, clustering, and focal adhesion signaling. The small
hairpin RNA (shRNA) knockdown of Trask results in increased cell adhesiveness and a failure to properly
inactivate focal adhesion signaling, even in the unanchored state. On the contrary, the experimentally induced
phosphorylation of Trask results in the inhibition of cell adhesion and inhibition of focal adhesion signaling.
This is mediated through the inhibition of integrin clustering without affecting integrin affinity state or ligand
binding activity. Furthermore, Trask signaling and focal adhesion signaling inactivate each other and signal
in exclusion with each other, constituting a switch that underlies cell anchorage state. These data provide
considerable insight into how Trask functions to regulate cell adhesion and reveal a novel pathway through
which Src kinases can oppose integrin-mediated cell adhesion.
Src family kinases (SFKs) are a family of nonreceptor pro-
tein tyrosine kinases with a domain structure consisting of a
highly conserved kinase domain as well as an SH2 domain and
an SH3 domain, a C-terminal negative-regulatory tyrosine res-
idue, and an N-terminal myristoylation site. Three members of
the family, Src, Yes, and Fyn, are ubiquitously expressed, while
the expression of the other members is largely restricted to
specific hematopoietic cell lineages. SFKs participate in numer-
ous cellular pathways in association with growth factor receptors,
G protein-coupled receptors, steroid hormones, STAT transcrip-
tion factors, and integrin receptors (14, 17, 34). The role of
SFKs in regulating cell adhesion signaling at sites of adhesion
to the extracellular matrix (ECM) is particularly well estab-
lished. Upon integrin engagement with the ECM and clus-
tering of integrins at sites of cell adhesion to matrix, mac-
romolecular complexes are assembled in association with
the intracellular tails of activated integrins (30, 40, 48). Within
these focal adhesions, focal adhesion kinase (FAK) is activated
by autophosphorylation at tyrosine 397, creating a binding site
for the Src SH2 domain (30, 37). Upon binding to FAK, Src is
activated and phosphorylates a number of additional tyrosine
residues on FAK, creating additional binding sites for SFKs
and other proteins. Activated Src also phosphorylates a num-
ber of additional cytoskeletal proteins, including paxillin and
p130Casand proteins involved in regulating the RhoA, Rac1,
and Cdc42 GTPases (23). These events function to stabilize
focal adhesions, generating a force-induced mechanical link
with the actin cytoskeleton, and regulate the surrounding
SFKs are required for proper establishment of focal adhe-
sions, as fibroblasts deficient in Src kinases have significantly
reduced tyrosine phosphorylation at focal contacts and defec-
tive cell adhesion to matrix (7, 26, 47). Although this loss-of-
function model supports the current molecular models of focal
adhesion establishment, the conclusions are not reciprocated
by gain-of-function experiments. The constitutively activated
v-src oncogene product interacts with focal contacts, phosphor-
ylating target proteins within them (20, 33). However, the
activities of the v-src product are destructive to focal adhesions,
and in fact, v-src-transformed cells appear to have significantly
reduced focal adhesions (11). Therefore, the evidence suggests
that SFKs are capable of promoting both adhesive and an-
tiadhesive functions, and most current models reconcile this
by proposing that SFKs function in focal adhesion turnover
(15). The mechanisms that mediate the antiadhesive func-
tions of SFKs are less well understood. Some evidence sug-
gests that SFKs can mediate focal adhesion disassembly
through a RhoA- and mDia1-mediated pathway or through a
calpain-mediated pathway (18, 51). In this paper, we describe
a novel mechanism by which SFKs can negatively regulate
focal adhesion assembly.
We have been studying a novel substrate of SFKs named
Trask. Trask is a recently described 140-kDa transmembrane
protein with little homology to known families of proteins. It
has a large extracellular region containing CUB domains and a
smaller intracellular region containing five tyrosines (5). Trask
is widely expressed in epithelial cells and tissues as a variable
blend of 140-kDa and 85-kDa forms, the latter due to proteo-
lytic cleavage of its distal extracellular region by serine pro-
teases, including the membrane bound MT-SP1 (5, 42). Trask
is phosphorylated in vitro by SFKs, including Src and Yes, and
is also phosphorylated by SFKs in cells, and its phosphorylation
can be inhibited by all classes of SFK inhibitors (5). The phos-
* Corresponding author. Mailing address: Box 1387, University of
California, San Francisco, San Francisco, CA 94143-1387. Phone:
(415) 476-0158. Fax: (415) 353-9592. E-mail: email@example.com.
?Published ahead of print on 28 December 2010.
phorylation of Trask is exclusively dependent on SFKs, since it
fails to undergo phosphorylation in Src/Yes/Fyn knockout cells
(SYF cells) unless transfected with an SFK member (50).
Trask has also been independently identified as a cancer-asso-
ciated gene by other groups. In a microarray analysis of colon
cancers, it was identified as a transcript with increased expres-
sion in tumors compared with that in adjacent normal tissues
and was named CDCP1 (39). In another line of study, a sub-
tractive immunization screen designed to identify antibodies
against more metastatic variants of HEp-3 carcinoma cells
identified a surface protein that was named SIMA135, which is
identical to Trask/CDCP1 (22).
The suggestion that Trask/CDCP1 is important in cancer
progression has been further supported by correlative studies
of human tumors, although the data are mixed and the nature
of this association and the cellular role of Trask/CDCP1 in
cancer is a matter of ongoing interest and investigation. In an
extensive analysis of Trask expression and phosphorylation in
human tissues, we found that Trask is widely expressed in most
epithelial tissues; however, the SFK phosphorylation of Trask
is restricted to physiological circumstances of detachment,
such as in mitotically detached cells in the colonic crypts (42).
However, in a large survey of human tumor sections, we found
that Trask is phosphorylated in many epithelial cancers at all
stages, including preinvasive cancers such as tubular adeno-
mas, but not in their normal-tissue counterparts (50). In other
studies, the elevated expression of Trask/CDCP1 has been
linked with poorer prognosis in cancers of the lung, kidney,
and pancreas (3, 24, 31) but with better prognosis in endo-
metrioid cancer (28).
The phosphorylation of Trask is linked with cell adhesion
such that Trask is phosphorylated almost instantly when epi-
thelial cells detach from matrix and is dephosphorylated when
cells readhere to matrix (42). In the current study, we mecha-
nistically studied the link between Trask phosphorylation and
cell adhesion through loss-of-function and gain-of-function
studies, looking at cell adhesiveness, at the affinity, binding,
and clustering of integrins, and at focal adhesion assembly and
signaling. We report that the SFK phosphorylation of Trask
inhibits cell adhesion through the inhibition of integrin binding
activity. This is mediated through the inhibition of clustering,
but not regulation of affinity state, and consequent inhibition of
focal adhesion assembly and signaling. We also found that the
SFK phosphorylation of Trask functions in opposition to and
in exclusion with focal adhesion signaling. These two signaling
pathways oppose and inactivate each other, defining a switch
that regulates cell anchorage state.
MATERIALS AND METHODS
Cell culture and immunoblotting. Cell lines were obtained from the American
Type Culture Collection. To force cells into suspension, cells were washed in
phosphate-buffered saline (PBS), exposed to a 2 mM solution of EDTA in Hanks
buffer, and when fully detached, cultured in growth media in ULC plates (Corn-
ing) for 2 h. ULC plates are not permissive to cell adhesion.
Antiphosphotyrosine antibodies (PY99), anti-FAK, anti-p-Y397 FAK, anti-
paxillin, and anti-p130Casantibodies were from Santa Cruz Biotechnology, Inc.
(Santa Cruz, CA). Generation of polyclonal and monoclonal anti-Trask and
anti-p-Y743 Trask antibodies were previously described (50). PP1 and PP2 were
from EMD-Calbiochem (San Diego, CA). Anti-integrin ?1 antibodies were
mouse monoclonal anti-integrin ?1 (BD Biosciences) for immunoprecipitation
and immunoblotting studies and anti-integrin ?1 clone P5D2 for integrin-acti-
vating and flow cytometry studies (6, 13, 52).
For wound healing experiments, near-confluent MCF10A cells were mechan-
ically scraped across the center of the well to generate a cell-free lane and
observed over the next 24 to 48 h for wound closure by cell migration. For
migration assays, the Transwell migration assay (Millipore) was used accord-
ing to manufacturer’s procedure. To the bottom wells, Dulbecco’s modified
Eagle’s medium (DMEM)–10% fetal bovine serum (FBS) containing 20 ng/ml
epidermal growth factor (EGF) was added. Cells were harvested, resuspended in
serum-free medium at densities of 5 ? 105cells/ml, and added to the top well.
When necessary, the cells were preincubated with anti-Trask or control antibod-
ies (80 ?g/ml) for 30 min at room temperature. After overnight incubation, the
cells were stained with crystal violet and enumerated.
Total cellular lysates were harvested in modified radioimmunoprecipitation
assay (RIPA) buffer (10 mM Na phosphate [pH 7.2], 150 mM NaCl, 0.1% SDS,
1% NP-40, 1% Na deoxycholate, protease inhibitors, 1 mM sodium orthovana-
date). For Western blotting, 50 ?g of each lysate was separated by SDS-PAGE,
transferred to membrane, and immunoblotted using appropriate primary and
secondary antibodies and enhanced chemiluminescence visualization. For immu-
noprecipitation studies, 300 ?g of lysate was incubated overnight with specific
antibodies, immune complexes were collected by protein G-Sepharose beads and
washed, and the denatured complexes were immunoblotted as described above.
Quantitative analysis of cell attachment. Wells of 96-well plates (not tissue
culture treated) were incubated overnight at 4°C with 100 ?l of matrix proteins
(5 ?g/ml fibronectin, 10 ?g/ml collagen type IV [Sigma], and 10 ?g/ml laminin).
MDA-468TR/WT Trask MDA-468TR/pcDNA4-TO cells were induced for 16 h
with or without doxycycline (Dox) and were plated at a density of 5 ? 105cells/ml
in 100 ?l per well. Cells were incubated at 37°C with 5% CO2for 1 h, followed
by three washes with PBS to remove unattached cells. Adhesion was assessed by
a colorimetric assay using crystal violet (Sigma), a cytochemical stain that binds
to chromatin. Cells were fixed in cold methanol for 15 min, and plates were then
air dried and stained with 0.1% crystal violet in PBS for 5 min at room temper-
ature. After removal of the crystal violet solution, the plates were washed ex-
tensively and dried, and the stain was released by using 2% SDS in PBS. Stain
intensity was quantified by spectrophotometry (at 570 nm) using a plate reader.
Generation of Trask-overexpressing cells. Generation of MDA-468 cells ex-
pressing doxycycline-inducible Trask was previously described (5). Several clones
were analyzed, all of which showed identical adhesion phenotypes when treated
with doxycycline. These cells were named MDA-468TR/Trask. Tyrosines 707,
734, and 743 in Trask were mutated to phenylalanine in the pcDNA4-Trask
vector by using the Stratagene Quik Exchange kit. The mutated insert was
sequenced for confirmation and stably transfected into MDA-468TR cells. Sev-
eral clones were expanded, and the doxycycline-inducible expression of mutant
Trask was confirmed by myc immunoblotting. All clones had similar phenotypes.
An individual clone (with tyrosine-to-phenylalanine [Y?F] mutations) was fur-
ther studied and named MDA-468TR/Y?F Trask.
Generation of Trask knockdown cells. Small hairpin RNA (shRNA) sequences
were cloned into pSico-RGFP and pSico-RNeo vectors. pSico-RGFP expresses
green fluorescent protein (GFP) as a selectable marker, and pSico-RNeo con-
tains a neomycin resistance cassette.
To create the shTrask-1 construct, the oligonucleotide 5?-TGAATGTTGCT
TTGGATCC-3? was annealed to 5?-TCGAGGATCCAAAAAAAATGTTGCT
and cloned into HpaI and XhoI sites of the pSicoR vector. To create the
shTrask-2 construct, the oligonucleotide 5?-TGATAGATGAGCGGTTTGCAA
GGCGCGCC-3? was annealed to 5?-TCGAGGCGCGCCAAAAAAAATAGA
TCATCTATCA-3? and cloned into HpaI and XhoI sites of the vector. For
generation of the nonsilencing construct, the oligonucleotides 5?-TGTCTCGCT
TTTTTGGCGCGCC-3? and 5?-TCGAGGCGCGCCAAAAAAATCTCGCTT
A-3? were similarly annealed and cloned. The shRNA constructs were
transfected into 293T cells along with the appropriate packaging vectors. The
resulting lentiviral particles were used to infect cells. Stably integrated GFP-
expressing MCF10A cell infectants were purified by flow sorting, and Trask
knockdown was confirmed in the GFP-expressing cell population by Western
blotting. MDA-468 cells were infected with pSico-RNeo shRNA constructs and
selected with G418 (300 ?g/ml).
Bead assays. To measure the integrin signaling in suspended cells in the
absence of cell spreading, polystyrene beads (Polysciences) were used. These
beads were precoated with 125 ?g/ml fibronectin (Sigma) or 50 ?g/ml anti-?1
integrin antibody (P5D2) (Santa Cruz Biotechnology) by rotation for 1 h at room
VOL. 31, 2011p-Trask INHIBITS INTEGRIN CLUSTERING 767
temperature in PBS. A total of 106beads were incubated with 106cells in a 2-ml
suspension in ULC plates for 3 h. At this time, aggregation between beads and
cells was observed under a microscope, and the cells remained with rounded
morphology. The cell-bead suspensions were pelleted and cells lysed in RIPA
To measure bead-cell aggregation, fluorescent beads and cells were used. The
beads were FluoSphere polystyrene microspheres, which had yellow-green fluo-
rescence (505/515), from Molecular Probes. Beads were precoated with 125
?g/ml fibronectin (Sigma) or anti-?1 integrin antibody (P5D2) at 50 ?g/ml
(Santa Cruz Biotechnology) by rotation for 1 h at room temperature in PBS.
Cells were fluorescently labeled with Vybrant DiD cell-labeling solution (Mo-
lecular Probes) according to the manufacturer’s instructions. For aggrega-
tion, 1 ? 106beads were incubated with 1 ? 106cells in a 2-ml suspension in
ULC plates for 3 h. The percent aggregation was determined by flow cytometry.
FACS analysis of cell surface ?1 integrin expression. Cell surface ?1 integrin
expression was assayed by flow cytometry. MDA-468TR/Trask cells were left
uninduced or induced with doxycycline overnight and subsequently stained with
anti-?1 integrin antibodies (conformation independent) and secondary fluores-
cent antibodies and quantified by fluorescence-activated cell sorter (FACS) anal-
Integrin conformation assays. Recombinant human fibronectin fragment 3,
containing type III domains 8 to 13 and the amino acid sequence between
Glu1266 and Pro1908 (R&D Systems), was fluorescently labeled with the Alexa
Fluor 488 microscale protein labeling kit (Invitrogen) according to the manu-
facturer’s instructions. For binding, cells were detached with EDTA, washed with
serum-free medium, and incubated with the fluorescent fibronectin fragment for
30 min at room temperature. Cells were washed with serum-free medium, fixed
with paraformaldehyde, and immediately analyzed by FACS analysis. Where
indicated, samples were incubated and washed in the presence of 2 mM MnCl2.
Integrin conformation was also assayed using the conformation state-specific
antibody 9EG7 (BD Biosciences) as follows. EDTA-detached cells were washed
with 1? Tris-buffered saline (TBS)–5% bovine serum albumin (BSA) and incu-
bated with 9EG7 in the presence of 1 mM calcium or 2 mM manganese in 1?
TBS–5% BSA for 1 h on ice. Cells were washed with 1? TBS–1% BSA and
consequently incubated with fluorescein isothiocyanate (FITC)-conjugated anti-
rat secondary antibodies for 30 min on ice. Cells were washed with 1? TBS,
resuspended, and immediately analyzed by FACS.
Microscopy. Tissue cultured cells were viewed and imaged by phase-contrast
microscopy using a Nikon TS-100F inverted microscope equipped with a Nikon
D100 digital camera attached to the photo port. Images were imported into
Photoshop software, converted to grayscale, and gamma adjusted for optimal
representation. Fluorophore-stained cells were imaged by fluorescence micros-
copy using an Axioplan 2 Zeiss microscope at the appropriate excitation wave-
lengths. Zeiss Fluar objective lenses were used with Immersol 518N as the
imaging medium. Pictures were taken at room temperature with an Axiocam
MRm camera and Axiovision 4.5 acquisition software, imported into Photoshop
software, and gamma adjusted for optimal representation.
TIRF microscopy. Cells were grown on number 1.5 coverslips, and induced
with or without Dox for 12 h and fixed with 4% paraformaldehyde. Cells were
stained with anti-integrin ?1, anti-FAK, or antipaxillin antibodies (Santa Cruz
Biotechnology) and secondary Alexa Fluor 546-conjugated antibodies (Invitro-
gen). Samples were mounted with water, illuminated with a Nikon laser total
internal reflection (TIRF) illuminator, and observed under a Nikon TE2000E
inverted microscope. Images were taken with the NIS-Elements Advanced Re-
search software and a 100? objective.
Trask phosphorylation is linked with the state of anchorage
in vitro and in vivo. To interrogate changes in signaling asso-
ciated with the anchorage state, we compared adherent epi-
thelial cells with nonadherent (suspended) epithelial cells. The
unanchored state can be induced by brief exposure to EDTA
and subsequent culture in nonadherent plates. Trask was phos-
phorylated upon loss of anchorage in MCF10A immortalized
epithelial cells (Fig. 1A). Trask phosphorylation was similarly
induced if adhesion is severed by brief exposure to trypsin (Fig.
1B). Trypsin also cleaves p140Traskto its smaller 85-kDa form.
The cleavage is not linked with the state of anchorage or with
the phosphorylation of Trask, since EDTA disrupts cell adhe-
sion, inducing Trask phosphorylation without cleavage. The
phosphorylation of Trask is tightly linked with the state of
anchorage, occurring immediately upon loss of anchorage and
continuing for as long as the cells are maintained in suspension
and rapidly reversing upon respreading and adhesion (Fig.
1B). The association between Trask phosphorylation and the
loss of adhesion is not unique to these cells and was repro-
ducible in other epithelial cell lines, including immortalized
keratinocytes (Fig. 1C) and epithelial cancer cell lines (Fig.
1D). Primary keratinocytes also showed no phosphorylation
of Trask when cultured in the adherent state but immediately
phosphorylated Trask when deprived of anchorage (data not
shown). The SFK phosphorylation of Trask in these cells is not
an artifactual finding unique to circumstances of tissue culture
detachment, but rather it experimentally reproduces physio-
logical conditions of cell detachment in vivo. In human tissue
sections, the phosphorylation of Trask is rarely seen, since
almost the entire normal epithelium is anchored. However,
Trask phosphorylation can be seen in physiologically detached
mitotic cells (Fig. 1E). To look at a physiologic circumstance of
experimentally induced adhesion disruption in vivo, we studied
skin wounding in mice. Traumatic scalpel-inflicted wounding
and disruption of the dermal tissue architecture induced the
phosphorylation of Trask in mouse skin to levels comparable
to those in detached MCF10A cells (Fig. 1F). Trask phosphor-
ylation is also frequently seen in human cancers at all stages in
vivo (50). Therefore, the SFK phosphorylation of Trask is
inversely linked with the state of anchorage in all in vivo and in
vitro circumstances that we have observed. Anchorage depri-
vation in cultured cells provides a highly relevant experimental
model system for mechanistic exploration of this physiologic
finding, which we have pursued in the following studies.
Experimental Trask knockdown promotes adhesive signal-
ing. The loss of anchorage is associated with the phosphoryla-
tion of Trask as well as the concomitant dephosphorylation of
focal adhesion proteins, consistent with the dismantling of
focal adhesions (Fig. 2A). To study the function of Trask in cell
detachment, we generated MCF10A cells lacking Trask ex-
pression due to shRNA knockdown (MCF10A/shTrask) along
with control cells (MCF10A/shControl) (Fig. 2B). While con-
trol cells fully inactivated integrin signaling in the unanchored
state, MCF10A/shTrask cells showed persistent phosphoryla-
tion of FAK when detached, indicating a failure to properly
inactivate integrin signaling in the unanchored state (Fig. 2C).
The phenotypic consequence of this is evident during the pro-
cess of cell detachment. Treatment of control MCF10A cells
with trypsin resulted in the rapid loss of cell adhesions with
consequent cell retraction and rounding (Fig. 2D). However,
Trask knockdown cells (MCF10A/shTrask) were considerably
more persistent in adhesion to matrix than were controls and
were significantly retarded in their rates of cell detachment,
retraction, and rounding (Fig. 2D). Dissociation from the un-
derlying matrix was impaired in MCF10A/shTrask cells despite
the effective disruption of cell-cell associations (Fig. 2E). With
prolonged exposure to trypsin, MCF10A/shTrask cells eventu-
ally did go into suspension, but the process was much slower
than that for control cells. Similar results were seen with
MDA-468 breast cancer cells and when cell detachment was
induced by EDTA (discussed below; see Fig. 4D). Similar
768SPASSOV ET AL.MOL. CELL. BIOL.
results were also seen with transient transfection of Trask small
interfering RNA (siRNA) (data not shown).
Since proper focal adhesion turnover is important in cell
migration, we studied the role of Trask in cell migration in
wound healing assays. When a scraping wound was generated
in a monolayer of MCF10A cells, cells migrated and eventually
filled the wound over a period of 40 h. However, MCF10A/
shTrask cells failed to migrate into and close a wound (Fig. 3).
This effect is not due to clonal variation, as similar results were
seen with transiently transfected shRNA knockdown cells not
subjected to subcloning (not shown).
The reduced abilities of Trask knockdown cells to detach
from matrix and to effectively inactivate integrin signaling sug-
gest that the SFK phosphorylation of Trask may function to
oppose cell adhesion. To further study the hypothesis that the
SFK phosphorylation of Trask functions to oppose cell adhe-
sion, we moved to a cell model with which we could conduct
both loss-of-function and gain-of-function experiments. For
FIG. 1. Trask phosphorylation is tightly linked with the state of anchorage. IP, immunoprecipitation; IB, immunoblot. (A) MCF10A cells were
assayed while adherent (lane A) or detached by brief exposure to EDTA and cultured for 2 h in suspension in ULC plates (lane S). (B) MCF10A
cells were harvested in the adherent state (lane 1) or induced to detach by treatment with trypsin for 1 min (lane 2) or 5 min (lane 3), and suspended
(susp) cells were subsequently cultured in growth media for 2 h (lane 4) or 24 h (lane 5). Some cells were returned to tissue culture-treated plates
after 2 h of culture in suspension and proceeded to gradually adhere over the next 1 (lane 6) or 2 (lane 7) hours. Trask phosphorylation and
expression were assayed by IP/IB assays as shown. Trask is rapidly cleaved by trypsin and is seen predominantly in its 85-kDa cleaved form in cells
immediately after trypsin treatment. Continued Trask protein synthesis resulted in the reemergence of the 140-kDa uncleaved Trask form
minimally at 2 h and maximally at 24 h. (C) HaCaT immortalized keratinocytes were assayed while adherent (lanes A) or detached by brief
exposure to EDTA and cultured for 2 h in suspension in ULC plates (lanes S). (D) A panel of cancer cell lines was also assayed in the adherent
or suspended state. Lanes correspond to MDA-468 (1) and MDA-231 (2) breast cancer cells, Du145 (3) and PC3 (4) prostate cancer cells, and
DLD1 (5) and HCT116 (6) colon cancer cells. (E) Anti-phospho-Trask immunohistochemical staining of normal human colon. Trask phosphor-
ylation is seen during mitotic detachment. The cross sections show the lower third of the crypts where mitoses are frequently seen. The mitotic
nature and the colon epithelial nature of these cells were previously confirmed by numerous other immunostains (42). (F) A patch of mouse skin
was excised and the epidermis was wounded ex vivo by numerous rapid scalpel-induced cuts and the tissue snap-frozen within 2 min. Trask
phosphorylation was assayed in the tissue lysates as indicated. Shown here are results from unwounded (1) or wounded (2) skin, as well an IgG
IP control (3). Lysates from suspended MCF10A cells (4) were run for comparison. Arrows indicate the 85- and 140-kDa forms of Trask.
VOL. 31, 2011 p-Trask INHIBITS INTEGRIN CLUSTERING769
this, we chose MDA-468 breast cancer cells. In contrast to
untransformed MCF10A cells, MDA-468 cells have some basal
SFK-dependent phosphorylation of Trask, possibly due to the
activated state of SFKs in these cancer cells, and the overex-
pression of Trask in these cells leads to its constitutive phosphor-
ylation at much higher levels, similar to the detached state. For
loss-of-function experiments, MDA-468 cells were engineered
with near total knockdown of Trask (MDA-468/shTrask) along
FIG. 2. Trask knockdown in MCF10A cells promotes focal adhesion signaling and cell adhesion. (A) The phosphorylation of FAK or paxillin
was assayed in adherent (lanes A) or suspended (lane S) MCF10A cells by immunoprecipitation of FAK or paxillin followed by immunoblotting
with antiphosphotyrosine antibodies and compared with total expression by using anti-FAK or antipaxillin antibodies. (B) MCF10A cells were
engineered to stably express Trask shRNA or a control nonsilencing shRNA. Molecular masses (in kDa) are shown to the right. (C) Lysates from
control and Trask knockdown MCF10A cells were obtained while adherent (lanes A) or in suspension for 2 h (lanes S). Focal adhesion signaling
was assayed by the analysis of FAK phosphorylation by FAK IP followed by pTyr immunoblotting. (D) Trask shRNA- and nonsilencing
shRNA-expressing MCF10A cells and the parental MCF10A cells were simultaneously placed in identical warm solutions of trypsin and imaged
under phase-contrast microscopy at the indicated time points under ?200 magnification. (E) Magnification of ?400 is shown for MCF10A/shTrask
cells in trypsin for the indicated lengths of time. At higher magnification, it is apparent that the disruption of cell-cell contacts is preserved, but
it is cell retraction from the underlying matrix that is retarded in these Trask knockdown cells.
770 SPASSOV ET AL.MOL. CELL. BIOL.
ing MDA-468 cells into suspension led to hyperphosphorylation
of Trask (Fig. 4B) and the concomitant loss of focal adhesion
signaling as shown by the dephosphorylation of FAK and pax-
illin (Fig. 4C, lanes 2 and 4). However, Trask knockdown cells
failed to effectively inactivate focal adhesion signaling, as seen
by persistent FAK and paxillin phosphorylation in the sus-
pended state (Fig. 4C, lanes 6 and 8). This failure to inactivate
integrin signaling in detached Trask knockdown cells suggests
that the phosphorylated Trask (p-Trask) functions to oppose
integrin signaling. Consistent with this, MDA-468/shTrask cells
showed increased adhesiveness compared with control cells.
This could be seen both during cell detachment and during cell
reattachment. When induced to detach by treatment with
EDTA, MDA-468/shTrask cells had delayed detachment com-
pared with control cells (Fig. 4D), and when detached MDA-
468/shTrask cells were replated on fibronectin-coated (FC)
plates, they had more rapid attachment and spreading than
control cells did (Fig. 4E and F).
Experimental Trask phosphorylation inhibits integrin sig-
naling and cell adhesion. For gain-of-function experiments,
MDA-468 cells were engineered to express the tetracycline
(Tet) repressor and to overexpress myc-tagged Trask (MDA-
468TR/Trask) or vector control (MDA-468TR/vector) when
exposed to Dox (Fig. 5A). Dox induces the overexpression and
constitutive SFK phosphorylation of Trask at high levels (Fig.
5A), and this serves as a model of Dox-induced SFK phosphor-
ylation of Trask. The phosphorylation of Trask induced by its
overexpression is specifically due to SFKs, since it can be in-
hibited by Src-selective tyrosine kinase inhibitors, and we have
FIG. 3. Trask knockdown in MCF10A cells prevents cell migration. Confluent parental or shRNA-expressing MCF10A cells were linearly
scraped through the center and cultured for 2 more days. Images were taken at ?100 magnification. A magnification of ?400 is also shown for
the 20-h time point.
VOL. 31, 2011 p-Trask INHIBITS INTEGRIN CLUSTERING771
FIG. 4. Trask knockdown in MDA-468 cells increases focal adhesion signaling and cell adhesion. (A) MDA-468 cells were engineered to stably
express either of two Trask shRNA sequences (named MDA-468/shTrask-1 and MDA-468/shTrask-2) or a control nonsilencing shRNA (named
MDA-468/shControl). Arrows indicate the 85- and 140-kDa forms of Trask. (B) MDA-468 cells were assayed while adherent (lane A) or detached
by brief exposure to EDTA and cultured for 2 h in suspension in ULC plates (lane S). The phosphorylation of Trask was assayed as shown. (C) The
state of integrin outside-in signaling was investigated through the analysis of FAK and paxillin phosphorylation by IP/immunoblots in the indicated
control and Trask knockdown cells cultured in the adherent (lanes A) or suspended (lanes S) state. (D) Trask knockdown and control MDA-468
cells were simultaneously placed in a 2 mM solution of EDTA and allowed to detach slowly over the indicated periods of time. Images were taken
under phase-contrast microscopy at ?200 magnification at the indicated time points (T). The control cells at 30 and 40 min were mostly detached
and floating. Since this is not readily apparent in still pictures, we reimaged them at 45 min following a PBS wash, which removed all the detached
cells. (E and F) Suspended Trask knockdown and control MDA-468 cells were replated on fibronectin-coated plates, and the rates of reattachment
were compared by analysis of cell spreading and attachment at 2 h postreplating. Cell spreading (E) was quantified by microscopic analysis of 200
cells, and cell adhesion (F) was quantified by a colorimetric assay as described in Materials and Methods. Abs, absorbance.
FIG. 5. Trask phosphorylation counteracts cell adhesion and focal adhesion signaling. (A) MDA-468 cells were engineered to overexpress
full-length myc-tagged Trask (MDA-468TR/Trask) or vector control in a doxycycline (Dox)-inducible manner. The Dox-induced expression and
phosphorylation of myc-Trask are shown by IB and IP/IB methods. Arrows indicate the 85- and 140-kDa forms of Trask. (B) The indicated cells
were trypsinized and seeded onto fresh tissue culture plates in the presence or absence of doxycycline and imaged by phase-contrast microscopy
the following day. Results from quantitative analyses of cell spreading and attachment are also shown as averages of values from triplicate wells.
Error bars represent standard errors of the means (SEM). Scoring was by phase-contrast microscopic counting of 200 cells. (C) The experiment with
MDA-468TR/Trask was repeated with cells detached by EDTA rather than trypsin. EDTA-detached cells similarly failed to readhere if induced by
doxycycline. (D) MDA-468TR/Trask cells were left uninduced or induced with doxycycline overnight, and total cellular expression levels of laminin and
fibronectin were assayed by immunoblotting as indicated. (E) Ninety-six-well plates precoated with the indicated matrix proteins over a coating of BSA
were prepared fresh. The indicated cells were pretreated with doxycycline overnight and subsequently detached by trypsinization and replated on plates
coated with the indicated matrix proteins for 2 h. After unattached cells were washed, the relative fraction of attached cells was quantified by a
colorimetric assay of stained cells and is shown as the average of values from triplicates. (F) Focal adhesion signaling was assayed by the analysis of total
FAK and paxillin phosphorylation by IP/immunoblotting. FAK activity was also assayed by immunoblot analysis of its Y397 autophosphorylation site
using a p-Y397-specific antibody. Assays were done on adherent cells induced with doxycycline or control overnight. (G) Cell migration was quantitatively
studied in Transwell chambers and is shown as the average of values from triplicates with SEM.
VOL. 31, 2011 p-Trask INHIBITS INTEGRIN CLUSTERING773
previously shown that it fails to undergo phosphorylation in
Src/Yes/Fyn knockout (SYF) cells unless an SFK member is
cotransfected (50). The constitutive phosphorylation of over-
expressed Trask is likely due to the saturation of dephosphor-
ylation mechanisms (not shown). When control MDA-468 cells
were seeded onto tissue culture plates, they spread and ad-
hered within several hours of seeding (Fig. 5B), coincident with
the dephosphorylation of Trask seen with the onset of adhe-
sion in these and other cells. However, cells with Dox-induced
constitutively phosphorylated Trask failed to spread onto ma-
trix and remained in the suspended state indefinitely (Fig. 5B,
lower-right image). When Trask phosphorylation was induced
by doxycycline in already adherent cells, this also resulted in
the loss of adhesion to the underlying matrix, although with a
longer latency. A disruption of focal adhesions in already ad-
herent cells was evident before the eventual loss of adhesion
when observed under TIRF microscopy (discussed further be-
low; see Fig. 11). The antiadhesive phenotype of Trask-over-
expressing cells was not due to clonal selection, since 4 differ-
ent MDA-468TR/Trask clones showed identical biological
phenotypes (data not shown). Trask-overexpressing cells failed
to adhere, whether they were initially detached with trypsin
(Fig. 5B) or EDTA (Fig. 5C). The failure to adhere is not due
to reduced expression of cellular matrix proteins (Fig. 5D).
The failure to adhere is also not specific to tissue culture-
treated plates and was similarly seen with plates coated with
fibronectin, laminin, or collagen (Fig. 5E). This suggests that
the inhibition of cell adhesion by phosphorylated Trask is due
to the inhibition of integrin function. Consistent with this,
when Trask phosphorylation was experimentally induced in
adherent cells by doxycycline treatment, FAK and paxillin were
dephosphorylated (Fig. 5F). This led to a rapid reduction in
the number of focal adhesions (shown below; see Fig. 11) and
eventual loss of cell adhesion. The doxycycline induction of
Trask phosphorylation in MDA-468TR/Trask cells also inhib-
ited cell migration in Transwell chamber assays (Fig. 5G).
These findings were not unique to MDA-468 cells. Trask over-
expression and constitutive phosphorylation were also studied
in HEK293 cells and similarly inhibited cell adhesion and
dephosphorylated FAK and paxillin in these cells (data not
The loss-of-function experiments discussed above show that
Trask knockdown cells are unable to properly inactivate inte-
grin signaling, and the gain-of-function experiments show that
the SFK phosphorylation of Trask inactivates integrin signal-
ing. Taken together, these data show that phosphorylated
Trask negatively regulates integrin-mediated cell adhesion.
The fact that both loss-of-function and gain-of-function exper-
imental models of Trask impair cell migration (Fig. 3 and 5G)
is consistent with the fact that both adhesive and antiadhesive
mechanisms must function for proper focal adhesion turnover
in migrating cells.
A limitation of the MDA-468TR/Trask overexpression model
is that we cannot specifically attribute the results to the SFK
phosphorylation of Trask rather than the overexpression of
Trask. Therefore, we generated a full-length Trask construct
with tyrosine-to-phenylalanine (Y?F) mutations at Y707,
Y734, and Y743. This Y?F Trask mutant correctly localizeed
to the membrane but showed no detectable phosphorylation
when overexpressed (Fig. 6A). However, in contrast to the case
with wild-type Trask, overexpression of Y?F Trask did not
lead to dephosphorylation of the focal adhesion proteins FAK,
paxillin, and p130Cas(Fig. 6B), and it did not inhibit cell ad-
hesion (Fig. 6C) or cell migration (Fig. 6D). These data sup-
port the conclusion that the antiadhesive effects seen in Dox-
induced MDA-468TR/Trask cells are specifically due to the
phosphorylation of Trask and not the overexpression of Trask.
Trask phosphorylation is linked with cell adhesion and not
with cell spreading. The evidence that the SFK phosphoryla-
tion of Trask functions in opposition to another SFK target
function, outside-in integrin signaling, prompted us to study
these seemingly conflicting roles more directly. To do this, we
moved to a more direct experimental model of integrin signal-
ing. One of the limitations of studying cell adhesion on a flat
surface is that this experimental system involves the concomi-
tant activation of the interdependent but mechanistically dis-
tinct processes of cell adhesion (mediated through integrin
activation and focal adhesion signaling) and cell spreading
(mediated through remodeling of the actin cytoskeleton). In
order to more specifically study the functional link between the
SFK phosphorylation of Trask and integrin function without
the compounding effects of cell spreading, we set up a model of
integrin activation using FC beads in contact with cells under
conditions of anchorage deprivation. The beads are much
smaller than the cells and do not provide a surface for spread-
ing. However, the beads provide sufficient immobilization of
fibronectin to cluster and activate integrins, allowing fibronec-
tin-integrin engagement and signaling to be studied in the
absence of cell spreading. In this experimental model, integrin
outside-in signaling can be assayed biochemically by analysis of
p-FAK and p-paxillin, and integrin physical binding activity can
be studied by quantitative assays of cell-bead adhesion using
flow cytometry. Fibronectin immobilization on beads is nec-
essary, since soluble fibronectin in media is unable to acti-
vate outside-in integrin signaling (Fig. 7A). When unin-
duced MDA-468TR/Trask cells came into contact with FC
beads in the suspended state, Trask was dephosphorylated
(Fig. 7B, lane 2). This effect was reproduced by using beads
coated with antibodies that activate integrin ?1 (IAbC beads)
similar to fibronectin (Fig. 7B, lane 3). Therefore, the activa-
tion of integrins leads to the dephosphorylation of Trask, and
the dephosphorylation of Trask is not a consequence of cell
Trask signaling and integrin signaling act in opposition and
exclusion. FC beads or IAbC beads failed to dephosphorylate
Trask in cells with Dox-induced constitutive SFK phosphory-
lation of Trask (Fig. 7B, lanes 5 and 6); therefore, we looked to
determine whether integrin signaling and Trask signaling
occur simultaneously or whether they are mutually exclu-
sive. In MDA-468TR/vector cells or in uninduced MDA-
468TR/Trask cells, FC beads induced integrin signaling (as-
sayed by p-FAK/p-paxillin) (Fig. 7C, compare lane 1 with lane
3 and lane 5 with lane 7) concomitant with the dephosphory-
lation of Trask (7B, compare lane 1 with lane 2). If Trask could
not be dephosphorylated (due to doxycycline-induced expres-
sion/phosphorylation), then FC beads failed to induce integrin
signaling (7C, compare lane 6 with lane 8). This indicates that
the two signaling processes oppose each other. Similar results
were obtained when integrins were activated by IAbC beads
rather than FC beads (Fig. 7D).
774 SPASSOV ET AL.MOL. CELL. BIOL.
This inhibition of integrin signaling is specifically due to the
phosphorylation of Trask and not its overexpression, since the
doxycycline-induced overexpression of the phosphorylation-
defective Y?F Trask mutant failed to block integrin signaling
(Fig. 7C, compare lane 10 with lane 12). In fact, the Y?F Trask
mutant itself promoted an exaggerated integrin signaling re-
sponse. This is due to a partial dominant-negative effect of this
mutant, evidenced by the reduction of endogenous Trask phos-
phorylation upon induction of the Y?F Trask mutant (Fig. 7E,
lane 2). Trask knockdown cells similarly showed an enhance-
FIG. 6. The Trask inhibition of integrin function is due to its phosphorylation. (A) MDA-468 cells engineered to express a myc-tagged Y?F
mutant (phosphorylation-defective) Trask construct in a Dox-inducible manner were assayed as shown to confirm Trasks expression and lack of
phosphorylation. Arrows indicate the 85- and 140-kDa forms of Trask. Additionally, the same cells growing on coverslips were induced with Dox
or control, and the correct localization of the Y?F mutant Trask construct was verified by anti-myc immunostaining of paraformaldehyde-fixed
cells. (B) The state of integrin signaling was assayed following the Dox induction of Y?F Trask by analysis of FAK, paxillin, and p130Cas
phosphorylation. (C) Suspended MDA-468TR/Y?F Trask cells were replated on tissue culture plates and allowed to adhere overnight in the
presence or absence of Dox and imaged the following day by phase-contrast microscopy. Results of quantitative analyses of cell spreading and
attachment are also shown as averages of values from triplicate wells with SEM. (D) Cell migration was quantitatively studied in Transwell
chambers and reported as the average of values from triplicates with SEM.
VOL. 31, 2011 p-Trask INHIBITS INTEGRIN CLUSTERING775
ment of integrin signaling. In Trask knockdown cells, baseline
integrin signaling was elevated (Fig. 7F, compare lane 1 with
lane 3) and FC beads induced an exaggerated integrin signal-
ing response (7F, lane 4).
Therefore, we found that Trask signaling (i.e., phosphoryla-
tion) was inhibited by the experimental induction of integrin
signaling. However, we also found that integrin signaling was
blocked by the experimental induction of Trask signaling and
FIG. 7. SFK phosphorylation of Trask and integrin signaling oppose each other, and this is not linked with cell spreading. (A) MDA-468 cells
cultured in suspension for 2 h were exposed to a control (lane 1), soluble fibronectin in media at 50 ?g/ml (lane 2), beads previously coated in 50
?g/ml fibronectin (lane 3), or soluble anti-?1 integrin-activating antibodies (lane 4). p-paxillin was assayed by antipaxillin antibody immunopre-
cipitation followed by anti-pTyr immunoblotting. This experiment showed that soluble fibronectin or soluble integrin-activating antibodies (P5D2)
do not activate focal adhesion signaling. However, fibronectin immobilized on beads activates focal adhesion signaling. (B) MDA-468TR/Trask
cells were cultured in the absence or presence of doxycycline to express phosphorylated Trask and subsequently cultured in suspension in contact
with fibronectin-coated beads (FC-beads) or anti-?1 integrin antibody-coated beads (IAbC-beads) for 1 h. The phosphorylation of Trask was
assayed in cell lysates by IP/immunoblotting. (C) The induction of integrin signaling (FAK and paxillin phosphorylation) was assayed following cell
contact with FC beads for 1 h. The ability of phosphorylated Trask to inhibit the induction of integrin signaling was determined by the forced
induction of Trask phosphorylation in MDA-468TR/Trask cells (lanes 5 to 8). Control cells included vector transfectants (lanes 1 to 4) and cells
expressing the phosphorylation-defective Y?F Trask mutant (lanes 9 to 12). (D) The experiment described for panel B was repeated using IAbC
beads in place of FC beads. (E) The partially dominant-negative effect of the Y?F Trask mutant on phosphorylation of endogenous Trask was
studied by analysis of total cellular Trask phosphorylation. (F) FC-induced integrin signaling was studied in Trask knockdown cells in comparison
with nonsilencing shRNA control cells.
776 SPASSOV ET AL.MOL. CELL. BIOL.
was enhanced by eliminating Trask signaling. These results
were consistent and reproducible across all experiments done
either in monolayer adhesion models or in suspended cell-bead
models as described above. Taken together, these data reveal
that Trask signaling and integrin signaling functionally inacti-
vate each other and signal in a mutually exclusive fashion.
To determine whether phospho-Trask inhibits integrin sig-
naling through a physical interaction with integrin complexes,
we looked for the presence of Trask in integrin immune com-
plexes. Trask is indeed seen in ?1 integrin immune complexes
after Dox induction in MDA-468TR/Trask cells (Fig. 8A, lane
2). The interaction is specific for phosphorylated Trask and is
not seen in uninduced cells or in cells induced to express the
phosphorylation-defective Y?F Trask mutant (Fig. 8A, lane
5). Immunostaining experiments confirm the expression of
Trask at the cell membrane within the adhesion plane (Fig.
8B). This is consistent with its physical presence within integrin
immune complexes and its functional role in disrupting inte-
grin signaling in adherent cells. The Trask-integrin interaction
is not an artifact of the overexpression model and is also seen
with endogenous Trask when phosphorylated by SFKs in the
suspended state (Fig. 8C, lane 2). The interaction is specific for
phosphorylated Trask and is disrupted if Trask is dephosphor-
ylated by brief treatment with a Src inhibitor (Fig. 8D). Immu-
nostaining experiments confirm that p-Trask and ?1 integrin
both localize to the membrane in suspended MDA-468 cells
(Fig. 8E). This is consistent with the physical presence of Trask
within integrin immune complexes in suspended cells and its
functional role in inactivating integrin signaling during anchor-
Phosphorylation of Trask inhibits integrin binding activity.
The data discussed above show that p-Trask inhibits outside-in
integrin signaling. In these experiments, integrin function was
assayed through the phosphorylation of focal adhesion pro-
teins, which occurs downstream of activated integrin receptors.
The exact point of inhibition can actually be at several events
that occur prior to activation of focal adhesion signaling. p-
Trask could inhibit the adoption of high-affinity integrin con-
formations (typically referred to as inside-out integrin signal-
ing), it could inhibit the engagement of activated integrins with
ECM, or it could inhibit the clustering of activated integrins.
To more specifically determine the point at which p-Trask
interferes with integrin activation and signaling, we performed
several additional experiments. We first repeated the FC bead
assays with the goal of assaying cell-bead binding activities
rather than cellular signaling events. In this experiment,
cell-bead binding was quantitatively assayed by two-color
flow cytometry using fluorescent cells and fluorescent FC
beads (schematically described in Fig. 9A). The loss-of-func-
tion and gain-of-function cell models of Trask, along with their
FIG. 8. Trask is found in integrin complexes when phosphorylated. (A) Anti-?1 integrin (?1-integ) immunoprecipitates from the indicated cell
types were immunoblotted as shown. The induced Trask constructs can be identified by immunodetection of their myc tags. (B) Doxycycline-
induced MDA-468TR/Trask cells were fixed and immunostained using anti-myc antibodies and imaged by confocal microscopy. The image on the
right shows a z section in the midplane of the cells, while the image on the left shows a z section corresponding to the basal surface of the cells,
at the plane of adhesion to the coverslip. (C) The interaction between ?1 integrin and Trask was studied in nonengineered MDA-468 cells in the
adherent (lane A) or suspended (lane S) state. (D) The interaction between ?1 integrin and Trask was studied again in suspended MDA-468 cells
with or without a 1-h pretreatment with the Src inhibitor dasatinib at 1 ?M. (E) MDA-468 cells were fixed in paraformaldehyde after being cultured
for 2 h in suspension, permeabilized in methanol, and immunostained with anti-p-Y743 Trask or anti-?1 integrin antibodies. Secondary antibodies
were Alexa Fluor 546 (red) and Alexa Fluor 488 (green) conjugates. Imaging was performed under an LSM510 Carl Zeiss confocal microscope,
at ?630 magnification, and images were captured by using the LSM Image browser.
VOL. 31, 2011 p-Trask INHIBITS INTEGRIN CLUSTERING777
controls, were assayed in this experiment. The Dox-induced
overexpression and phosphorylation of Trask did not alter the
surface expression of ?1 integrin (Fig. 9B); therefore, the cell-
bead aggregation data from these experiments reflect the bind-
ing activity of integrins and not their surface expression. Rel-
ative to the basal state, shRNA knockdown of Trask increased
FC bead binding by ?2-fold (Fig. 9C, compare column 1 with
columns 2 and 3). Conversely, the Dox-induced SFK phosphor-
ylation of Trask significantly reduced FC bead binding (Fig. 9,
compare column 6 with column 7). The suppression of binding
activity is due to the phosphorylation of Trask and not its
overexpression, since it was not seen with overexpression of the
phosphorylation-defective Y?F Trask mutant (Fig. 9, columns
8 and 9). The apparent partial increase in binding activity with
the Y?F Trask mutant is consistent with a partial dominant-
negative effect of this mutant on Trask signaling (discussed
previously). This experiment shows that p-Trask interferes with
integrin binding activity.
Phosphorylation of Trask inhibits integrin clustering but
does not regulate the affinity state. p-Trask may interfere with
integrin binding activity either through the inhibition of inte-
grin conformational activation and ligand binding or the inhi-
bition of integrin clustering. To differentiate between these
possibilities, we looked more specifically at the affinity state
and ligand binding of integrins and the clustering of integrins.
We assayed the affinity state of integrins by two methods. With
the first method, we studied cell binding to a labeled monova-
lent soluble fibronectin fragment. This assay showed that the
expression or phosphorylation of Trask had no effect on the
fibronectin binding activity of integrins (Fig. 10A). Further-
more, phosphorylated Trask did not block the Mn-induced
conformational activation of integrins (Fig. 10A). With a sec-
ond method, we studied the activation state of ?1 integrin by
using an activation state-specific antibody. This assay similarly
showed that the expression or phosphorylation of Trask had no
effect on the conformation of ?1 integrin and that phosphory-
lated Trask did not block the Mn-induced conformational ac-
tivation of ?1 integrin (Fig. 10B). These assays show that
p-Trask does not affect the affinity state of integrins, or their
ligand binding activity, and taken together with all the previous
FIG. 9. SFK phosphorylation of Trask inhibits binding to FC beads. (A) A two-color flow cytometry assay was used to determine the binding
activity of integrins. In the cell-bead mixture, the green-fluorescent FC beads are easily identified by one channel, and the red-fluorescence-labeled
experimental cells are identified by the second channel. A bead bound to a cell is called by both channels and identifies the binding of integrins
binding to fibronectin. (B) MDA-468TR/Trask cells were induced with Dox overnight (line 3) to express phosphorylated Trask. The following day,
cells were stained with anti-?1 integrin antibodies (conformation independent) and secondary fluorescent antibodies and assayed by FACS
analysis. Controls include cells without Dox induction (line 2) and cells without primary antibody staining (line 1). (C) Aggregation to FC
beads was assayed in the indicated 9 experimental cell types. Cell-bead aggregation was quantified by FACS analysis and is shown in graphical
format as the average of values from triplicates with SEM. Representative histograms corresponding to columns 1, 2, 6, and 7 are shown
778 SPASSOV ET AL.MOL. CELL. BIOL.
FIG. 10. Trask phosphorylation does not affect the integrin affinity or conformational state. (A) The affinity state of integrins was interrogated
in the indicated cell types by binding activity with a monovalent Alexa Fluor 488-labeled fibronectin fragment. Binding activity was assayed by flow
cytometry. Increased binding of the labeled ligand is seen as a right shift. Mn treatment induced integrin activation into the high-affinity
conformation and functions as a positive control. (B) The conformational state of ?1 integrin was interrogated in the indicated cell types by using
a conformation state-specific anti-?1 integrin antibody. Manganese induces the active conformation of integrins and was used as a positive control
in contrast to the negative control, calcium. For Trask-overexpressing cells, the data are shown in two formats to highlight that there was no change
in integrin conformation with doxycycline induction of Trask (middle data set) and to highlight the Mn-induced activation of integrins (lower data
set). For each row, the color legend is identified on the right. Ab, antibody.
data, suggest that phosphorylated Trask inhibits cell adhesion
through the inhibition of integrin clustering. We looked more
directly at integrin clustering and focal adhesion complexes by
total internal reflection fluorescence (TIRF) microscopy. The
induction of Trask hyperphosphorylation in MDA-468TR/
Trask cells led to a marked reduction in integrin clustering and
focal adhesion complexes (Fig. 11). Continued Trask hyper-
phosphorylation eventually leads to the loss of all focal adhe-
sions and loss of cell adhesion, as described earlier. Therefore,
p-Trask inhibits cell adhesion by inhibiting the clustering of
activated integrins and negatively affecting focal adhesion as-
Much is known about the signaling mechanisms that regu-
late focal adhesions, although at this time we know much more
about the mechanisms that promote focal adhesion formation
upon integrin activation and clustering and much less about
the mechanisms that disrupt them. In particular, SFKs are
critically involved in regulating focal adhesion formation and
signaling downstream of integrin receptors. Activated Src lo-
calizes to focal adhesions, and this is an event that is required
for cell spreading on fibronectin (16, 25, 26, 43). Fibroblasts
deficient in Src kinases have focal contacts but reduced ty-
rosine phosphorylation at focal adhesions and defective cell
adhesion to matrix (7, 26, 27, 47). Upon integrin clustering,
FAK is recruited to the cytoplasmic tails (c-tails) of ?-inte-
grins, where it undergoes autophosphorylation at Y397 fol-
lowed by binding and conformational activation of Src and
reciprocal full activation of FAK by Src through phosphoryla-
tion of several of its tyrosine residues (30, 35). Src can also be
directly activated by its interaction with ? integrins (1, 40). The
FAK-Src complex phosphorylates a number of substrates, in-
cluding paxillin and p130Cas, further solidifying and expanding
the focal adhesion complex and ultimately linking it with the
actin cytoskeleton (4, 10, 21, 38, 44).
In this study, we describe an effector of SFKs that negatively
regulates focal adhesion formation, thereby mediating an an-
tiadhesive function. When phosphorylated by SFKs, Trask ap-
pears in complexes containing ?1 integrin, interfering with
integrin clustering and preventing the mechanical and signal-
ing events that link the intracellular cytoskeleton with the
ECM. Future studies will focus on determining the molecular
mechanisms through which p-Trask prevents integrin cluster-
ing. It is now well established that the activities of the integrin
heterodimer are regulated by intracellular proteins interacting
with the ? integrin cytoplasmic tail. In particular, proteins that
can induce the separation of the intracellular tails of the ? and
? integrins promote an extended integrin conformation in the
integrin extracellular domains (ECDs) with high affinity for
ligand binding (2, 32). The binding of several intracellular
proteins, including talin, kindlins, Dok1, tensin, and Numb, is
mediated through the integrin NPxY motif (9, 29, 49). While
the c-tail interactions described to date have provided great
insight into how intracellular proteins can regulate the affinity
state of the integrin heterodimer, much less is known about
inside-out signaling mechanisms that may regulate integrin
clustering. p-Trask does not regulate the affinity state of the
integrin heterodimer, as we have shown using assays that mea-
sure monovalent ligand binding and integrin conformation.
Rather, it inhibits integrin clustering and establishment of fo-
cal adhesions. Future studies will seek to more specifically
determine how Trask phosphorylation by SFKs inhibits inte-
grin clustering. One possibility is that phosphorylation of Trask
by SFKs promotes the SFK phosphorylation of the ? integrin
c-tail, promoting or disrupting integrin c-tail interactions that
are important in clustering. In preliminary attempts thus far,
we have not been able to identify p-Trask-induced tyrosine
phosphorylation of ?1 integrin, but the interactions of the ?
integrin c-tail are known to be regulated through tyrosine
phosphorylation. In particular, the integrin NPxY motif binds
some proteins specifically when it is phosphorylated and other
FIG. 11. Trask phosphorylation inhibits integrin clustering. MDA-
468TR/Trask cells grown on coverslips were induced to express phos-
phorylated Trask by doxycycline treatment for 12 h. Cells were then
fixed and immunostained using the indicated antibodies and visualized
under TIRF microscopy. Images are at ?1,000 magnification. Two
representative images are shown for each condition.
780 SPASSOV ET AL.MOL. CELL. BIOL.
proteins preferentially in its unphosphorylated state (29). The
tyrosine kinases that phosphorylate these sites have not been
well established. v-Src has been shown to phosphorylate the
NpxY motif within the ? integrin c-tail, disrupting certain
interactions and partly explaining the adhesion defects in v-src-
transformed cells (36). However, this appears to be one of
many promiscuous functions of v-src, and an analogous phys-
iological role for c-Src or other SFKs has not been described.
It remains possible that this phosphorylation activity of cellular
SFKs is induced specifically in the unanchored state, facilitated
by the appearance of p-Trask. The inhibitory effects of p-Trask
on integrin clustering and binding activity are almost surely
mediated through the phosphorylation of its intracellular ty-
rosines and not through any functions attributed to its ECD,
since Trask truncation mutants lacking the entire ECD inhibit
cell adhesion when overexpressed and phosphorylated identi-
cally to what is seen with full-length Trask (unpublished data).
Of particular interest is the fact that the inhibitory activities
of p-Trask and focal adhesion signaling are reciprocal in na-
ture, as each of these two signaling mechanisms turns off the
other, and they function in a mutually exclusive fashion. The
SFK phosphorylation of Trask, whether induced physiologi-
cally by loss of anchorage or experimentally by Trask overex-
pression, inactivates focal adhesion signaling. The reverse also
holds true, as the activation of integrin signaling, whether in-
duced physiologically by cell adhesion to matrix or experimen-
tally by fibronectin-induced activation of integrin receptors,
promotes the dephosphorylation of Trask. It is apparent from
all our experimental model systems and from all observed
physiological states that Trask phosphorylation and focal ad-
hesion signaling inactivate each other. Therefore, integrin sig-
naling and Trask signaling inhibit each other and constitute
mutually exclusive and opposing signaling programs, defining a
switch that determines the cell anchorage state. Although there
are five tyrosines within the intracellular domain of Trask, in
extensive tyrosine mutation studies, we have found that the
phosphorylation of Trask occurs as an all-or-none event con-
sistent with its function as a switch (unpublished data). The
mechanisms that mediate the dephosphorylation of Trask al-
most surely involve the regulation of specific protein tyrosine
phosphatases (PTPs). It is likely that adhesion promotes the
activation of a PTP or the interaction of a PTP with Trask,
thereby shifting the phosphorylation state equilibrium toward
the dephosphorylation of Trask.
Interest in Trask (also known as CDCP1) has grown because
of a growing body of evidence suggesting that it may have
cellular functions that are particularly important in tumor pro-
gression. A number of studies have documented increased
expression of Trask/CDCP1 in some cancers of the lung, kid-
ney, and pancreas, with a poorer prognosis conferred by its
elevated expression (3, 24, 31). Some studies have suggested
that Trask/CDCP1 promotes anoikis resistance in cancer cells
(45). This may be a cell-specific finding, since anoikis resis-
tance was not affected when Trask was knocked down in
MDA-468 breast cancer cells (data not shown). Other studies
suggest that Trask/CDCP1 may be more important for tumor
metastasis. Consistent with this, a number of experimental
studies using overexpression, knockdown, or monoclonal anti-
body targeting approaches show that the functions of Trask/
CDCP1 are important in tumor invasion and metastasis (12,
19, 41, 46).
The link between Trask/CDCP1 and cell adhesion has been
reported by others as well. Brown et al. reported that Trask/
CDCP1 undergoes tyrosine phosphorylation upon loss of ad-
hesion, but they reported that the phosphorylation event is
specifically linked with the proteolytic cleavage of the Trask/
CDCP1 ECD and does not occur with EDTA-induced detach-
ment (8). However, this study was done without the benefit of
anti-Trask/CDCP1 antibodies, and the data were only indi-
rectly inferred from the cross-reactive activities of an anti-p-
FAK antibody. Our current data obtained using specific anti-
Trask/CDCP1 antibody reagents as well as myc-tagged Trask/
CDCP1 constructs now clearly show that the phosphorylation
of Trask/CDCP1 is tightly linked with the state of anchorage
and is not a consequence of proteolytic cleavage (Fig. 1A, C,
D, and F, 4B, 5A, and 6A). In panels of the cell types studied,
we found that the relative expression levels of the 85- and
140-kDa forms of Trask vary considerably among different cell
types but that both forms undergo phosphorylation when an-
chorage is lost. Both examples were seen in this study. MDA-
468 cells express predominantly the 85-kDa form of Trask,
which undergoes phosphorylation upon detachment (Fig. 4B),
mouse keratinocytes express predominantly the 140-kDa form
of Trask, which undergoes phosphorylation upon scraping
(Fig. 1F), and MCF10A cells express a mixture of the 85- and
140-kDa forms of Trask, both of which undergo phosphoryla-
tion upon detachment (Fig. 1A). The evidence is now clear that
the phosphorylation of Trask/CDCP1 is not a consequence of
the proteolytic cleavage of its ECD.
It should be noted that overexpressed Trask does not follow
the same cleavage ratio as endogenous Trask. Although MDA-
468 cells express predominantly the 85-kDa form of Trask, its
overexpression by transfection produces both forms. This is
likely due to the fact that the overexpression exceeds the cel-
lular capacity for full cleavage. However, both forms are phos-
phorylated and both forms are found in integrin immune com-
plexes when expressed (Fig. 8A). Therefore, Trask cleavage
does not appear to play a role in the regulation of integrins.
The adhesion functions of Trask appear to be entirely medi-
ated through the tyrosine phosphorylation of its intracellular
This work was funded by the National Institutes of Health (grant
CA113952 to M.M.M.). D.S.S. is funded by a Susan G. Komen for the
Cure Postdoctoral Fellowship. C.H.W. was funded by a California
Breast Cancer Research Program Postdoctoral Fellowship.
We thank Michael McManus and the UCSF Sandler Lentiviral
RNAi core facility. Data for the TIRF microscopy analysis were ac-
quired at the Nikon Imaging Center at UCSF/QB3.
We have no conflicts of interest to declare.
1. Arias-Salgado, E. G., et al. 2003. Src kinase activation by direct interaction
with the integrin beta cytoplasmic domain. Proc. Natl. Acad. Sci. U. S. A.
2. Arnaout, M. A., B. Mahalingam, and J. P. Xiong. 2005. Integrin structure,
allostery, and bidirectional signaling. Annu. Rev. Cell Dev. Biol. 21:381–410.
3. Awakura, Y., et al. 2008. Microarray-based identification of CUB-domain
containing protein 1 as a potential prognostic marker in conventional renal
cell carcinoma. J. Cancer Res. Clin. Oncol. 134:1363–1369.
4. Bellis, S. L., J. T. Miller, and C. E. Turner. 1995. Characterization of
VOL. 31, 2011 p-Trask INHIBITS INTEGRIN CLUSTERING781
tyrosine phosphorylation of paxillin in vitro by focal adhesion kinase. J. Biol.
5. Bhatt, A. S., H. Erdjument-Bromage, P. Tempst, C. S. Craik, and M. M.
Moasser. 2005. Adhesion signaling by a novel mitotic substrate of src ki-
nases. Oncogene 24:5333–5343.
6. Blaschke, F., et al. 2002. Hypoxia activates beta(1)-integrin via ERK 1/2 and
p38 MAP kinase in human vascular smooth muscle cells. Biochem. Biophys.
Res. Commun. 296:890–896.
7. Bockholt, S. M., and K. Burridge. 1995. An examination of focal adhesion
formation and tyrosine phosphorylation in fibroblasts isolated from src-, fyn-,
and yes- mice. Cell Adhes. Commun. 3:91–100.
8. Brown, T. A., et al. 2004. Adhesion or plasmin regulates tyrosine phosphor-
ylation of a novel membrane glycoprotein p80/gp140/CUB domain-contain-
ing protein 1 in epithelia. J. Biol. Chem. 279:14772–14783.
9. Calderwood, D. A., et al. 2002. The phosphotyrosine binding-like domain of
talin activates integrins. J. Biol. Chem. 277:21749–21758.
10. Chodniewicz, D., and R. L. Klemke. 2004. Regulation of integrin-mediated
cellular responses through assembly of a CAS/Crk scaffold. Biochim. Bio-
phys. Acta 1692:63–76.
11. David-Pfeuty, T., and S. J. Singer. 1980. Altered distributions of the cytoskel-
etal proteins vinculin and alpha-actinin in cultured fibroblasts transformed
by Rous sarcoma virus. Proc. Natl. Acad. Sci. U. S. A. 77:6687–6691.
12. Deryugina, E. I., et al. 2009. Functional role of cell surface CUB domain-
containing protein 1 in tumor cell dissemination. Mol. Cancer Res. 7:1197–
13. Dittel, B. N., J. B. McCarthy, E. A. Wayner, and T. W. LeBien. 1993.
Regulation of human B-cell precursor adhesion to bone marrow stromal
cells by cytokines that exert opposing effects on the expression of vascular
cell adhesion molecule-1 (VCAM-1). Blood 81:2272–2282.
14. Engen, J. R., et al. 2008. Structure and dynamic regulation of Src-family
kinases. Cell. Mol. Life Sci. 65:3058–3073.
15. Fincham, V. J., and M. C. Frame. 1998. The catalytic activity of Src is
dispensable for translocation to focal adhesions but controls the turnover of
these structures during cell motility. EMBO J. 17:81–92.
16. Fincham, V. J., et al. 1996. Translocation of Src kinase to the cell periphery
is mediated by the actin cytoskeleton under the control of the Rho family of
small G proteins. J. Cell Biol. 135:1551–1564.
17. Frame, M. C. 2004. Newest findings on the oldest oncogene; how activated
src does it. J. Cell Sci. 117:989–998.
18. Frame, M. C., V. J. Fincham, N. O. Carragher, and J. A. Wyke. 2002. v-Src’s
hold over actin and cell adhesions. Nat. Rev. Mol. Cell Biol. 3:233–245.
19. Fukuchi, K., et al. 2010. Inhibition of tumor metastasis: functional immune
modulation of the CUB domain containing protein 1. Mol. Pharm. 7:245–
20. Glenney, J. R., Jr., and L. Zokas. 1989. Novel tyrosine kinase substrates from
Rous sarcoma virus-transformed cells are present in the membrane skeleton.
J. Cell Biol. 108:2401–2408.
21. Harte, M. T., J. D. Hildebrand, M. R. Burnham, A. H. Bouton, and J. T.
Parsons. 1996. p130Cas, a substrate associated with v-Src and v-Crk, local-
izes to focal adhesions and binds to focal adhesion kinase. J. Biol. Chem.
22. Hooper, J. D., et al. 2003. Subtractive immunization using highly metastatic
human tumor cells identifies SIMA135/CDCP1, a 135 kDa cell surface phos-
phorylated glycoprotein antigen. Oncogene 22:1783–1794.
23. Huveneers, S., and E. H. Danen. 2009. Adhesion signaling—crosstalk be-
tween integrins, Src and Rho. J. Cell Sci. 122:1059–1069.
24. Ikeda, J., et al. 2009. Expression of CUB domain containing protein
(CDCP1) is correlated with prognosis and survival of patients with adeno-
carcinoma of lung. Cancer Sci. 100:429–433.
25. Kaplan, K. B., et al. 1994. Association of the amino-terminal half of c-Src
with focal adhesions alters their properties and is regulated by phosphory-
lation of tyrosine 527. EMBO J. 13:4745–4756.
26. Kaplan, K. B., J. R. Swedlow, D. O. Morgan, and H. E. Varmus. 1995. c-Src
enhances the spreading of src-/- fibroblasts on fibronectin by a kinase-inde-
pendent mechanism. Genes Dev. 9:1505–1517.
27. Klinghoffer, R. A., C. Sachsenmaier, J. A. Cooper, and P. Soriano. 1999. Src
family kinases are required for integrin but not PDGFR signal transduction.
EMBO J. 18:2459–2471.
28. Mamat, S., et al. 2010. Prognostic significance of CUB domain containing
protein expression in endometrioid adenocarcinoma. Oncol. Rep. 23:1221–
29. McCleverty, C. J., D. C. Lin, and R. C. Liddington. 2007. Structure of the
PTB domain of tensin1 and a model for its recruitment to fibrillar adhesions.
Protein Sci. 16:1223–1229.
30. Mitra, S. K., and D. D. Schlaepfer. 2006. Integrin-regulated FAK-Src sig-
naling in normal and cancer cells. Curr. Opin. Cell Biol. 18:516–523.
31. Miyazawa, Y., et al. 2010. CUB domain-containing protein 1, a prognostic
factor for human pancreatic cancers, promotes cell migration and extracel-
lular matrix degradation. Cancer Res. 70:5136–5146.
32. Moser, M., K. R. Legate, R. Zent, and R. Fassler. 2009. The tail of integrins,
talin, and kindlins. Science 324:895–899.
33. Nigg, E. A., B. M. Sefton, T. Hunter, G. Walter, and S. J. Singer. 1982.
Immunofluorescent localization of the transforming protein of Rous sar-
coma virus with antibodies against a synthetic src peptide. Proc. Natl. Acad.
Sci. U. S. A. 79:5322–5326.
34. Parsons, S. J., and J. T. Parsons. 2004. Src family kinases, key regulators of
signal transduction. Oncogene 23:7906–7909.
35. Playford, M. P., and M. D. Schaller. 2004. The interplay between Src and
integrins in normal and tumor biology. Oncogene 23:7928–7946.
36. Sakai, T., R. Jove, R. Fassler, and D. F. Mosher. 2001. Role of the cytoplas-
mic tyrosines of beta 1A integrins in transformation by v-src. Proc. Natl.
Acad. Sci. U. S. A. 98:3808–3813.
37. Schaller, M. D., et al. 1994. Autophosphorylation of the focal adhesion
kinase, pp125FAK, directs SH2-dependent binding of pp60src. Mol. Cell.
38. Schaller, M. D., and J. T. Parsons. 1995. pp125FAK-dependent tyrosine
phosphorylation of paxillin creates a high-affinity binding site for Crk. Mol.
Cell. Biol. 15:2635–2645.
39. Scherl-Mostageer, M., et al. 2001. Identification of a novel gene, CDCP1,
overexpressed in human colorectal cancer. Oncogene 20:4402–4408.
40. Shattil, S. J. 2005. Integrins and Src: dynamic duo of adhesion signaling.
Trends Cell Biol. 15:399–403.
41. Siva, A. C., et al. 2008. Targeting CUB domain-containing protein 1 with a
monoclonal antibody inhibits metastasis in a prostate cancer model. Cancer
42. Spassov, D. S., F. L. Baehner, C. H. Wong, S. McDonough, and M. M.
Moasser. 2009. The transmembrane src substrate Trask is an epithelial
protein that signals during anchorage deprivation. Am. J. Pathol. 174:1756–
43. Timpson, P., G. E. Jones, M. C. Frame, and V. G. Brunton. 2001. Coordi-
nation of cell polarization and migration by the Rho family GTPases requires
Src tyrosine kinase activity. Curr. Biol. 11:1836–1846.
44. Turner, C. E. 2000. Paxillin and focal adhesion signalling. Nat. Cell Biol.
45. Uekita, T., et al. 2007. CUB domain-containing protein 1 is a novel regulator
of anoikis resistance in lung adenocarcinoma. Mol. Cell. Biol. 27:7649–7660.
46. Uekita, T., et al. 2008. CUB-domain-containing protein 1 regulates perito-
neal dissemination of gastric scirrhous carcinoma. Am. J. Pathol. 172:1729–
47. Volberg, T., L. Romer, E. Zamir, and B. Geiger. 2001. pp60(c-src) and
related tyrosine kinases: a role in the assembly and reorganization of matrix
adhesions. J. Cell Sci. 114:2279–2289.
48. Vuori, K. 1998. Integrin signaling: tyrosine phosphorylation events in focal
adhesions. J. Membr. Biol. 165:191–199.
49. Wegener, K. L., et al. 2007. Structural basis of integrin activation by talin.
50. Wong, C. H., et al. 2009. Phosphorylation of the SRC epithelial substrate
Trask is tightly regulated in normal epithelia but widespread in many human
epithelial cancers. Clin. Cancer Res. 15:2311–2322.
51. Yamana, N., et al. 2006. The Rho-mDia1 pathway regulates cell polarity and
focal adhesion turnover in migrating cells through mobilizing Apc and c-Src.
Mol. Cell. Biol. 26:6844–6858.
52. Yokosaki, Y., et al. 1994. The integrin alpha 9 beta 1 mediates cell attach-
ment to a non-RGD site in the third fibronectin type III repeat of tenascin.
J. Biol. Chem. 269:26691–26696.
782SPASSOV ET AL.MOL. CELL. BIOL.