Matrix Biology 21 (2002) 525–532

E-cadherin is a ligand for integrin a2b1

John D. Whittarda, Susan E. Craiga, A. Paul Moulda, Alexander Kochb,1, Olivier Pertzb,2,
¨
Jurgen Engelb, Martin J. Humphriesa,*
a
Wellcome Trust Centre for Cell-Matrix Research, School of Biological Sciences, University of Manchester, 2.205 Stopford Building,
Oxford Road, Manchester, M13 9PT, UK
b
Department of Biophysical Chemistry, Biozentrum, University of Basel, Klingelbergstrasse 70, CH-4056 Basel, Switzerland

Received 1 May 2002; received in revised form 10 June 2002; accepted 11 June 2002

Abstract

E-cadherin is a 120-kDa transmembrane glycoprotein expressed mainly on the surface of epithelial cells. The best characterised
function of E-cadherin is homotypic, calcium-dependent cell–cell adhesion; however, the observation that E-cadherin is also
capable of interacting with the aEb7 integrin to mediate leukocyte cell–cell adhesion wNature 372 (1994) 190x suggests that it
also participates in heterotypic interactions. To investigate the possibility that E-cadherin may interact with integrins expressed on
non-leukocytic cells, cell adhesion and solid-phase receptor–ligand binding experiments were performed using a pentameric E-
cadherin construct designed to detect low affinity, high avidity interactions. HT1080 human fibrosarcoma cells specifically adhered
to pentameric E-cadherin, and this adhesion was inhibited by anti-functional monoclonal antibodies directed against the integrin
a2 and b1 subunits, but not by a series of antibodies recognising other subunits. This suggested that the E-cadherin receptor was
a2b1, a previously characterised collagenylaminin receptor. Pentameric E-cadherin, but not monomeric E-cadherin, specifically
bound, in a divalent cation-dependent manner, to both purified a2b1 and to a recombinant form of the A-domain of the a2
subunit, which has been shown to be a major ligand-binding site within this and other integrins. These findings demonstrate that
E-cadherin can interact with a2b1 and suggest that heterotypic interactions between E-cadherin and integrins may be more
common than originally thought.
䊚 2002 Elsevier Science B.V. and International Society of Matrix Biology. All rights reserved.

Keywords: Adhesion; E-cadherin; Integrin a2b1

1. Introduction highly conserved cytoplasmic domain associates with

E-cadherin is an epithelial cell adhesion molecule that
mediates the establishment and maintenance of cell–cell
contacts (Takeichi, 1995). E-cadherin contains an extra-
cellular domain of five, tandemly repeated immunoglob-
ulin-like modules, a single membrane-spanning region
and a cytoplasmic tail (Geiger and Ayalon, 1992). The
*Corresponding author. Tel.: q44-161-275-5071; fax: q44-161-
275-1505.
E-mail address: martin.humphries@man.ac.uk (M.J. Humphries).
1
Current address: Department of Biochemistry and Molecular
Biology, Mt. Sinai School of Medicine, Box 1130, 1425 Madison
Ave, Room 16-02, New York, NY 10029, USA.
2
Current address: Department of Cell Biology, The Scripps
Research Institute, 10550 North Torrey Pines Road, La Jolla, CA
92037, USA.
a-catenin, b-catenin, plakoglobin (g-catenin), and p120
to form complexes that link cadherins with the actin
cytoskeleton (Kemler, 1993; Aberle et al., 1996; Aplin
et al., 1998). The three-dimensional structures of the
amino-terminal regions of N- and E-cadherin have been
determined (Overduin et al., 1995; Shapiro et al., 1995;
Nagar et al., 1996), and crystal packing data suggest
that the molecules may form a dimeric structure at the
same plasma membrane through the lateral interaction
of two cadherin subunits. Ca2q-dependent hydrogen
bonding (Nagar et al., 1996), tryptophan intercalation
(Shapiro et al., 1995) and use of a conserved HAV
motif (Blaschuk et al., 1990; Nose et al., 1990) have
all been proposed to contribute to the dimerisation
process. It has also been suggested that two N-cadherin

0945-053X/02/$ - see front matter 䊚 2002 Elsevier Science B.V. and International Society of Matrix Biology. All rights reserved.
PII: S 0 9 4 5 - 0 5 3 X Ž 0 2 . 0 0 0 3 7 - 9
526 J.D. Whittard et al. / Matrix Biology 21 (2002) 525–532
dimer interfaces could combine to form zipper-like
superstructures at the intercellular space (Shapiro et al.,
1995; Boggon et al., 2002), and that E-cadherin mole-
cules form active pairs at the cell surface that are
capable of combining with E-cadherin pairs on the
opposing cell surface (Tomschy et al., 1996; Pertz et al.
1999).
For some time, cadherins were only thought to be
involved in homophilic interactions; however, the inter-
action between intraepithelial lymphocytes and mucosal
epithelial cells was found to be mediated by E-cadherin
and the integrin aEb7 in humans (Cepek et al., 1994;
Higgins et al., 1998) and in mice (Karecla et al., 1995).
The interaction of the integrin aEb7 with E-cadherin
may function as a mechanism for the retention of
intraepithelial lymphocytes within the mucosal epitheli-
um (Cepek et al., 1994; Farstad et al., 1996). The
integrins are a large family of a,b heterodimeric, trans-
membrane cell-adhesion receptors that recognise both
cell-surface and extracellular matrix molecules. At pres-
ent, 24 different integrin receptors have been identified
in higher vertebrates, each drawn from a collection of
18 a subunits and eight b subunits (Hyne, 1992;
Humphries, 2000). The cellular distribution of integrins
varies considerably, with the b1 and b3 integrins gen-
erally having a widespread distribution, and the b2
integrins being restricted to leukocytes (Springer, 1990).
The aE subunit is a member of a subset of integrins,
including those containing the a1, a2, a10, a11, aM,
aL, aX and aD subunits, which all contain an inserted
domain (A-domain) of approximately 200 amino acids
in their N-terminal region (Shaw et al., 1994). This
insert shares sequence homology to the ‘A’ domains of
von Willebrand factor, matrilins, type VI collagen, and
complement factors B and C2 (Colombatti et al., 1993).
To date, the aE A-domain has not been expressed,
but the A-domains of a1, a2, aL, aM and aX have
been shown to represent a major ligand-binding site
within the a subunit, as recombinant versions bind
ligands with virtually the same cation dependence and
specificity as their parent molecules (Randi and Hogg,
1994; Ueda et al., 1994; Zhou et al., 1994; Tuckwell et
al., 1995). An acidic amino acid, E31, which is located
in the B–C loop of the N-terminal domain of E-cadherin,
is required for aEb7 binding (Karecla et al., 1996), as
are residues in the adjacent F–G loop (Taraszka et al.,
2000). The role of E31 is reminiscent of the well-
characterised contribution of acidic residues in binding
of VCAM-1 and MAdCAM-1 to a4 integrins (Newham
et al., 1997), and collagen to a2b1 (Emsley et al.,
2000). Consistent with this, mutation of the metal ion-
dependent adhesion site (MIDAS) in the aE A-domain
(D190A) abolished E-cadherin binding, and a docking
model of the A-domain with E-cadherin domain 1
indicated that the side-chain of E31 could potentially
co-ordinate the aE MIDAS metal ion (Higgins et al.,
2000). Recently, disruption of either Ca2q-dependent
interactions in the domain 1–2 junction of E-cadherin
or W2-dependent cross-intercalation of E-cadherin mon-
omers, both of which have been suggested to mediate
dimerisation (Shapiro et al., 1995; Nagar et al., 1996;
Pertz et al., 1999), were shown to inhibit aEb7 binding
(Corps et al., 2001). This finding suggests that E-
cadherin dimerisation might be needed to generate
favourable binding kinetics.
The observation that E-cadherin is capable of inter-
acting with the aEb7 integrin to mediate cell–cell
adhesion raised the possibility that E-cadherin may also
interact with other integrins. In this regard, experimental
studies using dominant negative E-cadherin mutants
have revealed cross-talk between the cadherin and inte-
grin families (Hodivala and Watt, 1994). In the present
study, we have examined the possibility that E-cadherin
may bind to other A-domain-containing integrins. Our
results demonstrate that a2b1 is able to bind E-cadherin,
and that the binding is mediated by the a2A-domain.
2. Experimental procedures
2.1. Cell culture
A375-SM cells, a human metastatic melanoma cell
line, were a gift from I.J. Fidler (MD Anderson Hospital
and University of Texas, Houston, TX). The human
fibrosarcoma cell line HT1080 was obtained from the
European Collection of Animal Cell Cultures (Porton
Down, Salisbury, UK). A375-SM cells were cultured in
Eagle’s minimal essential medium (MEM) with Earle’s
salts, supplemented with 10% (vyv) foetal calf serum
(FCS), 1.5= MEM vitamins, 1= non-essential amino
acids, 2 mM L-glutamine, 1 mM sodium pyruvate and
50 mgyml gentamycin (all from GIBCO-BRL, Paisley,
UK). HT1080 cells were cultured in Eagle’s MEM
supplemented with 10% (vyv) FCS, 1 mM sodium
pyruvate and 2 mM L-glutamine.
2.2. Reagents
Mouse monoclonal antibody HP2y1, directed against
the human a4 integrin subunit, was purchased from
Serotec (Oxford, UK). Rat monoclonal antibodies
mAb13, directed against the human b1 integrin subunit,
and mAb16, directed against the human a5 integrin
subunit, were gifts from S.K. Akiyama (National Insti-
tute of Environmental Health Sciences, Research Trian-
gle Park, NC) and K.M. Yamada (National Institute for
Dental and Craniofacial Research, Bethesda, MD).
Mouse monoclonal antibody GoH3, directed against the
human a6 subunit, was a gift from A. Sonnenberg
(Netherlands Cancer Institute, Amsterdam, Nether-
lands). Mouse monoclonal antibodies 5E8 and HAS3,
both directed against the human a2 integrin subunit,
 J.D. Whittard et al. / Matrix Biology 21 (2002) 525–532
were gifts from R.B. Bankert (Roswell Park Memorial
Institute, Buffalo, NY; Chen et al., 1991) and F.M. Watt
(Keratinocyte Laboratory, Cancer Research UK, Lon-
don, UK), respectively.
The Hy120 variant of the HepIIyIIICS region of
human fibronectin was cloned and expressed as
described (Makarem et al., 1994). a2b1 integrins from
HT1080 human fibrosarcoma cells (Mould et al., 1995)
and the 80-kDa cell-binding fragment of fibronectin
(Garcia-Pardo et al., 1989) were purified as previously
described. Recombinant human integrin a2 A-domain
was a gift from D.S. Tuckwell (University of Manches-
ter, UK; Tuckwell et al., 1995). Recombinant a1 and
aM A-domains were gifts from D.A. Calderwood (Uni-
versity of Manchester, UK; Calderwood et al., 1997).
Rat cartilage oligomeric matrix protein (COMP), mono-
meric E-cadherin (ECAD; contains the five extracellular
domains of mouse E-cadherin), and pentameric E-cad-
herin (ECAD-COMP; contains the five extracellular
domains of mouse E-cadherin fused to the assembly
domain of rat cartilage oligomeric matrix protein) were
produced as recombinant proteins in human kidney 293
cells as described in Koch et al. (1997). Pepsinised rat-
tail type I collagen was purchased from Sigma (Poole,
UK).
2.3. Cell spreading
A modification of the assay described by Humphries
et al. (1986) was used. Cell spreading assays were
performed using flat-bottomed 96-well microtitre tissue
culture plates (Costar). Micotitre wells were coated with
100-ml aliquots of proteins diluted with Dulbecco’s
phosphate-buffered saline containing Ca2q and Mg2q
(PBSq; GIBCO-BRL) for a minimum of 1 h at room
temperature. Wells were then aspirated and washed with
PBSq. Sites on the plastic for non-specific cell adhesion
were blocked with 100 ml of 10 mgyml heat-denatured
crystalline bovine serum albumin (BSA; Calbiochem,
Nottingham, UK), for 30 min at room temperature. For
experiments examining the effects of antibodies on cell
spreading, 50-ml aliquots of antibody were added for 1
h prior to the addition of cells. Wells were subsequently
aspirated and washed three times with PBSq. Near-
confluent cultures of A375-SM or HT1080 cells were
detached with 1% (vyv) trypsin–EDTA and harvested.
Cells were resuspended to a density of 1=105 yml in
warm Dulbecco’s MEM containing 1 mM Mn2q with
25 mM N-w2-hydroxyethylx-piperazine-N9-w2-ethanesul-
fonic acidx (Sigma), and allowed to recover for 15 min
at 37 8C. Aliquots of cell suspension (100 ml) were
added to each well and incubated in a humidified
atmosphere containing 7% (vyv) CO2 for 20 min at 37
8C for the initial 15 min. For the subsequent 1 h 15
min, the plate was incubated in a humidified air atmos-
phere containing 5% (vyv) CO2 at 37 8C. Cells were
527
fixed by the addition of 10 ml of 50% (wyv) glutaral-
dehyde for 30 min at room temperature, and the per-
centage of cells spread in each well was determined
using phase contrast microscopy. A total of 400 cellsy
well from a number of randomly selected fields were
counted. The criteria for a spread cell included a phase
dark appearance and an area of visible cytoplasm encir-
cling the nucleus.
2.4. Solid-phase receptor–ligand binding
Proteins to be biotinylated were diluted to 0.5 mgy
ml in PBSq. A 100-ml aliquot of 0.5 M NaCl, 0.1 M
NaHCO3, pH 8.0, and 200 ml of an approximately 30-
fold molar excess of Immunopure sulfo-N-hydroxysuc-
cinimide ester–biotin (Pierce, Chester, UK) were added
and rotated for 30 min at room temperature. The mixture
was then dialysed against several changes of sodium
phosphate buffer, pH 7.2, to remove excess biotin. The
initial protocol for solid-phase receptor–ligand binding
assays followed the method of Charo et al. (1990) and
Mould et al. (1994), and was subsequently optimised
for each different experiment. Purified integrins were
diluted 1:100 with PBSq, and 100-ml aliquots added to
the wells of a 96-well enzyme-linked immunoabsorbent
assay plate (ELISA; Immulon 4 plates; Dynatech, Bil-
lingshurst, UK) overnight at 4 8C. Wells were then
blocked with 200 ml of 5% (wyv) Fraction V BSA
(Sigma), 150 mM NaCl, 0.05% (wyv) NaN3, 10 mM
Tris–HCl, pH 7.4, overnight at 4 8C. A 25-ml aliquot
of 5% (wyv) Fraction V BSA was added to each well
before the wells were washed three times with 200 ml
of 150 mM NaCl, 1 mM MnCl2, 25 mM Tris–HCl, pH
7.4, containing 1 mgyml BSA (buffer A). If anti-integrin
antibodies or other inhibitors were used, they were added
30 min before the addition of biotinylated ligands.
Aliquots of 100 ml of biotinylated ligands (ranging from
1 to 10 mgyml) diluted in buffer A were added, and the
plate was incubated at 30 8C for 3 h. Biotinylated
ligands were aspirated and the wells washed four times
with buffer A. Bound ligand was quantitated by the
addition of 100 ml of a 1:200 dilution of ExtrAvidin–
peroxidase conjugate (Sigma) in buffer A for 10–15
min. Wells were then washed four times with buffer A,
and colour was developed using 2,29-azino-bis(3-ethyl-
benzthiazoline-6-sulfonic acid) solution. The absorbance
of each well was measured at 405 nm using a multiscan
ELISA reader (Dynatech, Billingshurst, UK). The level
of non-specific binding was determined by measuring
the level of biotinylated ligand binding to wells coated
with BSA or GST alone. Values for receptor–ligand
binding were then determined by subtracting non-spe-
cific binding values.
528 J.D. Whittard et al. / Matrix Biology 21 (2002) 525–532

Table 1 cadherin, has previously been observed (Cepek et al.,

Cell spreading on E-cadherin 1993).
Ligand Spreading (%) ("S.D.) To identify the integrin(s) responsible for mediating
HT1080 fibrosarcoma cell spreading on ECAD-COMP,
A375-SM HT1080 the effects of a panel of anti-integrin monoclonal anti-
COMP 0.0"0.0 0.0"0.0 bodies on cell adhesion were studied. As positive con-
ECAD 2.5"0.6 2.3"1.7 trols, the inhibitory effects of the anti-integrin
ECAD-COMP 8.3"3.8 57.5"5.0 monoclonal antibodies on HT1080 cell spreading on
Hy120 78.5"4.7 68.5"2.9
type I collagen (a2b1-mediated), the 80-kDa fragment
Spreading of A375 melanoma and HT1080 fibrosarcoma cells on of fibronectin (a5b1-mediated) and the Hy120 variant
COMP, monomeric E-cadherin (ECAD), pentameric E-cadherin of fibronectin (a4b1-mediated) were determined.
(ECAD-COMP) and Hy120 variant of fibronectin in the presence of
1mM Mn2q. Ligands were coated at concentrations that gave maximal HT1080 cell spreading on ECAD-COMP was almost
spreading (20 mgyml for ECAD-COMP and 10 mgyml for Hy120 completely abolished by either anti-b1 or anti-a2 mon-
variant of fibronectin). COMP and ECAD were coated at concentra- oclonal antibodies, but not affected by any other anti-
tions of 20 mgyml. integrin monoclonal antibodies (Fig. 2). Thus, HT1080
cell spreading on ECAD-COMP appeared to be entirely
3. Results and discussion attributable to the a2b1 integrin.

3.2. Solid-phase receptor–ligand binding assays to

3.1. Cell adhesive activity of ECAD-COMP
examine E-cadherin– a2b1 integrin binding

To mimic the clustering of E-cadherin at the cell To investigate the interaction between purified a2b1
surface, a pentameric E-cadherin construct (ECAD- integrin from HT1080 fibrosarcoma cells and ECAD-
COMP) containing the five extracellular domains of COMP in a cell-free system, solid-phase receptor–ligand
mouse E-cadherin fused to the assembly domain of rat binding assays were carried out. ECAD was unable to
cartilage oligomeric matrix protein (COMP) was gen- bind to a2b1, whereas binding of ECAD-COMP was
erated (Tomschy et al., 1996). In the resulting chimeric comparable to that of type I collagen (Fig. 3). Thus,
protein, five E-cadherin strands are linked at their C- results from cell-based and cell-free assays suggested
terminus by the five-stranded COMP domain. To assess that multimerisation of E-cadherin was necessary for
whether E-cadherin was capable of interacting with significant integrin-binding activity.
integrins other than aEb7, in the absence of homotypic
binding, initial investigations focused on its ability to
support A375 melanoma and HT1080 fibrosarcoma cell
adhesion. These cell lines were chosen as they express
a wide range of integrins. In addition to ECAD-COMP,
recombinant monomeric E-cadherin (ECAD) was also
tested in these assays. Cell spreading assays were carried
out in the presence or absence of 1 mM Mn2q, since
Mn2q-occupied integrins generally bind their ligands
with higher affinity, and therefore it was hoped to
maximise adhesive activity (Gailit and Ruoslahti, 1988;
Karecla et al., 1995; Mould et al., 1995).
A375 melanoma and HT1080 fibrosarcoma cells were
unable to spread on COMP or ECAD (Table 1). ECAD-
COMP, however, supported the spreading of both cell
types, but HT1080 fibrosarcoma cells spread to a much
greater extent. The requirement for pentamerisation
implied that the avidity of the interaction was an
important factor for determining whether the cells were
able to adhere to E-cadherin. Furthermore, since the
level of HT1080 cell spreading on ECAD-COMP was
approximately doubled when 1 mM Mn2q was present
(Fig. 1), this suggested that the interaction might be
Fig. 1. Spreading of HT1080 fibrosarcoma cells on pentameric E-
mediated by integrins. A similar stimulatory effect of cadherin (ECAD-COMP), coated at varying concentrations, in the
Mn2q on T-cell hybridoma adhesion to epithelial cells, presence of 1 mM Mn2q (j) and in the absence of Mn2q (d). Values
an interaction mediated by the aEb7 integrin and E- shown are mean"S.D.
 J.D. Whittard et al. / Matrix Biology 21 (2002) 525–532 529

phase receptor–ligand binding results were therefore in

broad agreement with results derived from cell-based
assays; both suggested that integrin binding to ECAD-
COMP was specific.

3.3. Solid-phase receptor–ligand binding assays to

examine ECAD-COMP– a2 A-domain binding

Both the aE and a2 subunits possess A-domains and,

Fig. 2. Effect of anti-integrin antibodies on HT1080 fibrosarcoma cell
spreading on ECAD-COMP (coated at 20 mgyml) in the presence of
1 mM Mn2q, and to the 80-kDa fragment of fibronectin (coated at
10 mgyml), Hy120 variant of fibronectin (coated at 10 mgyml) and
type I collagen (coated at 10 mgyml) in the absence of Mn2q. Proteins
were coated at concentrations that gave maximal cell spreading. Cells
were incubated with anti-b1 antibody mAb13, anti-a2 antibody 5E8,
anti-a4 antibody HP2y1, anti-a5 antibody mAb16 and anti-a6 anti-
body GoH3 at concentrations of 10 mgyml. Values shown are
mean"S.D.
to test whether this region contained the E-cadherin
binding site, the binding of biotinylated E-cadherin and
collagen type I binding to immobilised a1, a2 and aM
A-domain fusion proteins was determined in the pres-
ence of 1 mM Mn2q. The level of biotinylated collagen
type I binding was greater for the a1 than the a2 A-
domain fusion protein, in agreement with a previous
report (Calderwood et al., 1997; Fig. 4). Binding to aM
A-domain fusion protein was only just above back-
ground binding to BSA. The binding profile of biotiny-
lated ECAD-COMP to the a1, a2 and aM recombinant
A-domains was different to that of type I collagen.
Biotinylated ECAD-COMP binding was highest for the
a2 A-domain, followed by a1 A-domain, and the aM
A-domain was inactive.
Some of the binding observed between the a2 A-
domain and ECAD-COMP was due to the non-specific
interaction between the GST element of the fusion
protein and ECAD-COMP, as a small amount of ECAD-

To test the specificity of the interaction between

ECAD-COMP and purified a2b1 integrin, the effects
of EDTA, anti-b1 and anti-a2 integrin monoclonal
antibodies were studied. Integrin–ligand interactions are
known to be dependent on the presence of divalent
cations (Gailit and Ruoslahti, 1988), and thus integrin
function should be abolished by the presence of EDTA.
Biotinylated type I collagen binding to the a2b1 integrin
was almost completely inhibited by 10 mM EDTA and
by the anti-functional a2 monoclonal antibody 5E8, and
substantially inhibited by the anti-functional b1 mono-
clonal antibody mAb13, but the negative control anti-
functional a5 monoclonal antibody mAb16 had no
significant effect (Table 2). The binding of ECAD-
COMP to purified a2b1 was inhibited by 45% by the
addition of 10 mM EDTA (Table 2); however, it is
possible that the presence of EDTA may inhibit E-
cadherin function, as well as integrin function, since the
presence of calcium ions is important for the confor-
mation of the extracellular region of cadherins (Pokutta Fig. 3. Solid-phase assay showing the binding of biotinylated type I
et al., 1994), and hence for adhesiveness. An excess of collagen (j), monomeric E-cadherin (ECAD) (m) and ECAD-
unlabelled ECAD-COMP (40% inhibition), anti-func- COMP (d) at varying concentrations to a2b1 integrins purified from
HT1080 fibrosarcoma cells in the presence of 1 mM Mn2q. The level
tional anti-b1 (60% inhibition) and anti-a2 monoclonal of non-specific binding, determined from ligand binding to wells coat-
antibodies (50% inhibition) also inhibited ECAD- ed with BSA alone, was subtracted. Values shown are mean"S.D. of
COMP binding to a2b1 (Table 2). Data from solid- five replicate wells.
530 J.D. Whittard et al. / Matrix Biology 21 (2002) 525–532

Table 2
Specificity of E-cadherin–integrin binding

Inhibitor Control binding (%) ("S.D.)

Type I collagen ECAD-COMP
A
None 100"10.1 100"10.6
10 mM EDTA 16.83"13.0 55.49"10.4
Anti-b1 57.26"10.4 37.60"7.9
Anti-a5 87.30"10.7 85.47"6.5
Anti-a2 21.00"5.1 47.91"0.4
10= wECAD-COMPx – 58.28"7.6
10= wCOMPx – 106.14"15.3
B
None 100"5.1 100"10.6
10 mM EDTA 1.67"0.5 y21.8312.4
Anti-a5 99.42"1.7 104.16"10.3
Anti-a2 9.06"7.5 43.75"7.8
Effect of inhibitors on the binding of (A) biotinylated type I col-
lagen (used at 5 mgyml) and ECAD-COMP (used at 15 mgyml) to
a2b1 integrins purified from HT1080 fibrosarcoma cells in the pres- Fig. 4. Solid-phase assay showing the binding of biotinylated ECAD-
ence of 1 mM Mn2q and (B) biotinylated type I collagen (used at 5 COMP (at 5 mgyml) to a2, a1 and aM A-domain fusion proteins,
mgyml) and ECAD-COMP (used at 5 mgyml) to a2 A-domain fusion in the presence of 1 mM Mn2q. The level of non-specific binding,
protein (coated at 5 mgyml) in the presence of 1 mM Mn2q. The determined from ligand binding to wells coated with GST alone, was
background binding of type I collagen to GST was effectively zero subtracted. Values shown are mean"S.D. of five replicate wells.
in (A) and 35% of control in (B) due to high non-specific binding
of ECAD-COMP to GST. This explains the reduction in ECAD-
COMP binding by EDTA below zero. Wells were incubated with 10 nition of the a2 A-domain. However, interaction with
mM EDTA or anti-integrin antibodies mAb13 (anti-b1), mAb16 a2b1 integrin and the a2 A-domain only occurred when
(anti-a5) or 5E8 (anti-a2) at a concentration of 10 mgyml. The level
of non-specific binding, determined from ligand binding to wells coat-
E-cadherin was in a pentameric state of ECAD-COMP.
ed with BSA alone, was subtracted. Values shown are mean"S.D. of The high activity is best explained by interactions of
five replicate wells.

COMP binding was observed to wells coated with GST

alone (Fig. 5). The specificity of the interaction between
ECAD-COMP or type I collagen and a2 A-domain
fusion protein was tested using EDTA and anti-a2
integrin monoclonal antibodies. The binding of type I
collagen and ECAD-COMP to a2 A-domain fusion
protein was completely inhibited by the presence of
EDTA, suggesting that the interactions of type I collagen
and E-cadherin were both cation-dependent (Table 2).
In addition, type I collagen binding to a2 A-domain
fusion protein was almost completely inhibited by the
anti-functional monoclonal antibody 5E8 (90% inhibi-
tion), while in comparison, ECAD-COMP binding was
inhibited by 55%. These results indicate that the E-
cadherin–a2 A-domain interaction is specific, but the
relative potency of 5E8 might be interpreted to suggest
that ECAD-COMP might bind to a slightly different
region of the a2 A-domain compared to collagen.
In conclusion, the major finding of this manuscript is
that E-cadherin, which usually functions in homotypic
interactions, can participate in a heterotypic interaction Fig. 5. Solid-phase assay showing the binding of 5 mgyml biotinylated
with the integrin a2b1, which is a widely expressed ECAD-COMP (d) and GST (j) to a2 A-domain fusion protein
coated at varying concentrations in the presence of 1 mM Mn2q. The
member of the integrin family. Interaction with a2b1 is level of non-specific binding, determined from ligand binding to wells
specific, as judged by the lack of inhibition by antibodies coated with BSA alone, was subtracted. Values shown are mean"S.D.
directed against a4, a5 and a6, and the specific recog- of five replicate wells.
 J.D. Whittard et al. / Matrix Biology 21 (2002) 525–532
several arms of the multivalent ECAD-COMP with
multiple integrins or domains at the surface of the ligand
coat. The binding site of a2b1 at the E-CAD12 dimer
has not been localised yet, but it may be speculated that
the glutamate at position 31 in the BC loop of the first
N-terminal domain of E-cadherin, which is required for
interaction with aEb7 (Karecla et al., 1996), also
contributes to a2b1 binding. The functional significance
of the interaction between E-cadherin and the a2b1
integrin is currently unclear, but epidermal differentia-
tion and morphogenesis may well be biological process-
es in which such interactions could take place. a2b1 is
apically expressed in basal keratinocytes and in supra-
basal layers. In these areas there is no obvious collage-
nous extracellular matrix between cells, and therefore
the function of the integrins is unclear. It is conceivable,
however, that interactions between a2b1 integrin and
E-cadherin on opposing keratinocytes may occur, and
might be required for regulation of epidermal stratifica-
tion or maintenance of epidermal structure. In this
regard, it is intriguing that re-expression of a2b1 in
poorly differentiated breast carcinoma cells resulted in
a dramatic reversion of the malignant phenotype to a
differentiated epithelial phenotype (Sun et al., 1998).
This possibility will be the subject of future
investigations.
Acknowledgments
We would like to express our thanks to S.K. Akiyama,
K.M. Yamada, A. Sonnenberg and R.B. Bankert for
providing antibodies, and D.A. Calderwood and D.S.
Tuckwell for providing A-domain fusion proteins. We
are grateful to P. Newham for advice and helpful
discussions. This work was supported by grants from
the Wellcome Trust (to M.J.H.). J.D.W. was supported
by a studentship from BBSRC.
References
Aberle, H., Schwartz, H., Kemler, R., 1996. Cadherin–catenin com-
plex: protein interactions and their implications for cadherin func-
tion. J. Cell. Biochem. 61, 514–523.
Aplin, A.E., Howe, A., Alahari, S.K., Juliano, R.L., 1998. Signal
transduction and signal modulation by cell adhesion receptors: the
role of integrins, cadherins, immunoglobulin-cell adhesion mole-
cules, and selectins. Pharmacol. Rev. 50, 197–263.
Blaschuk, O.W., Sullivan, R., David, S., Pouliot, Y., 1990. Identifi-
cation of a cadherin cell adhesion recognition sequence. Dev. Biol.
139, 227–229.
Boggon, T.J., Murray, J., Chappuis-Flament, S., Wong, E., Gumbiner,
B.M., Shapiro, L., 2002. Structure of the C-cadherin ectodomain
and implications for the mechanism of cell adhesion. Science
(published online 18 April 2002; 10.1126yscience.1071559)
¨
Calderwood, D.A., Tuckwell, D.S., Eble, J.A., Kuhn, K., Humphries,
M.J., 1997. The integrin a1 A-domain is a ligand binding site for
collagens and laminin. J. Biol. Chem. 272, 12311–12317.
Cepek, K.L., Parker, C.M., Madara, J.L., Brenner, M.B., 1993.
Integrin aEb7 mediates adhesion of T lymphocytes to epithelial
cells. J. Immunol. 150, 3459–3470.
531
Cepek, K.L., Shaw, S.K., Parker, C.M., et al., 1994. Adhesion between
epithelial cells and T lymphocytes mediated by E-cadherin and the
aEb7 integrin. Nature 372, 190–193.
Charo, I.F., Nannizzi, L., Smith, J.W., Cheresh, D.A., 1990. The
vitronectin receptor aVb3 binds fibronectin and acts in concert
with a5b1 in promoting cellular attachment and spreading on
fibronectin. J. Cell Biol. 111, 2795–2800.
Chen, F.A., Repasky, E.A., Bankert, R.B., 1991. Human lung tumor-
associated antigen identified as an extracellular matrix adhesion
molecule. J. Exp. Med. 173, 1111–1119.
Colombatti, A., Bonaldo, P., Doliana, R., 1993. Type A modules:
interacting domains found in several non-fibrillar collagens and in
other extracellular matrix proteins. Matrix 13, 297–306.
Corps, E., Carter, C., Karecla, P., Ahrens, T., Evans, P., Kilshaw, P.,
2001. Recognition of E-cadherin by integrin aEb7. Requirement
for cadherin dimerisation and implications for cadherin and integrin
function. J. Biol. Chem. 276, 30862–30870.
Emsley, J., Knight, C.G., Farndale, R.W., Barnes, M.J., Liddington,
R.C., 2000. Structural basis of collagen recognition by integrin
a2b1. 1. Cell 101, 47–56.
Farstad, I.N., Halstensen, T.S., Lien, B., Kilshaw, P.J., Lazarovitz,
A.I., Brandtzaeg, P., 1996. Distribution of b7 integrins in human
intestinal mucosa and organized gut-associated lymphoid tissue.
Immunology 89, 227–237.
Gailit, J., Ruoslahti, E., 1988. Regulation of the fibronectin receptor
affinity by divalent cations. J. Biol. Chem. 263, 12927–12932.
Garcia-Pardo, A., Ferreira, O.C., Valinsky, J., Bianco, C., 1989.
Fibronectin receptors of mononuclear phagocytes: binding charac-
teristics and biochemical isolation. Exp. Cell Res. 181, 420–431.
Geiger, B., Ayalon, O., 1992. Cadherins. Annu. Rev. Cell Biol. 8,
307–332.
Higgins, J.M.G., Mandelbrot, D.A., Shaw, S.K., et al., 1998. Direct
and regulated interaction of integrin aEb7 with E-cadherin. J. Cell
Biol. 140, 197–210.
Higgins, J.M.G., Cernadas, M., Tan, K., et al., 2000. The role of
alpha and beta chains in ligand recognition by b7 integrins. J.
Biol. Chem. 275, 25652–25664.
Hodivala, K.J., Watt, F.M., 1994. Evidence that cadherins play a role
in the downregulation of integrin expression that occurs during
keratinocyte terminal differentiation. J. Cell Biol. 124, 589–600.
Humphries, M.J., Akiyama, S.K., Komoriya, A., Olden, K., Yamada,
K.M., 1986. Identification of an alternatively spliced site in human
plasma fibronectin that mediates cell type-specific adhesion. J. Cell
Biol. 103, 2637–2647.
Humphries, M.J., 2000. Integrin structure. Biochem. Soc. Trans. 28,
311–340.
Hyne, S.R.O., 1992. Integrins; versatility, modulation, and signaling
in the cell adhesion. Cell 69, 11–25.
Karecla, P.I., Bowden, S.J., Green, S.J., Kilshaw, P.J., 1995. Recog-
nition of E-cadherin on epithelial cells by the mucosal T cell
integrin aM290B7 (aEb7). Eur. J. Immunol. 25, 852–856.
Karecla, P.I., Green, S.J., Bowden, S.J., Coadwell, J., Kilshaw, P.J.,
1996. Identification of a binding site for integrin aEb7 in the N-
terminal domain of E-cadherin. J. Biol. Chem. 271, 30909–30915.
Kemler, R., 1993. From cadherins to catenins: cytoplasmic protein
interactions and regulation of cell adhesion. Trends Genet. 9,
317–321.
Koch, A.W., Pokutta, S., Lustig, A., Engel, J., 1997. Calcium binding
and homoassociation of E-cadherin domains. Biochemistry 36,
7697–7705.
Makarem, R., Newham, P., Askari, J.A., et al., 1994. Competitive
binding of vascular cell adhesion molecule-1 and the HepIIyIIICS
domain of fibronectin to the integrin a4b1. J. Biol. Chem. 269,
4005–4011.
Mould, A.P., Askari, J.A., Craig, S.E., Garratt, A.N., Clements, J.,
Humphries, M.J., 1994. Integrin a4b1-mediated melanoma cell
532 J.D. Whittard et al. / Matrix Biology 21 (2002) 525–532
adhesion and migration on vascular cell adhesion molecule-1
(VCAM-1) and the alternatively spliced IIICS region of fibronec-
tin. J. Biol. Chem. 269, 27224–27230.
Mould, A.P., Akiyama, S.K., Humphries, M.J., 1995. Regulation of
integrin a5b1-fibronectin interactions by divalent cations. Evidence
for distinct classes of binding sites for Mn2q, Mg2q, and Ca2q. J.
Biol. Chem. 270, 26270–26277.
Nagar, B., Overduin, M., Ikura, M., Rini, J.M., 1996. Structural basis
of calcium-induced E-cadherin rigidification and dimerization.
Nature 380, 360–364.
Newham, P., Craig, S.E., Seddon, G.N., et al., 1997. 4 integrin
binding interfaces on VCAM-1 and MAdCAM-1. Integrin binding
footprints identify accessory binding sites that play a role in
integrin specificity. J. Biol. Chem. 272, 19429–19440.
Nose, A., Tsuji, K., Takeichi, M., 1990. Localization of specificity
determining sites in cadherin cell adhesion molecules. Cell 61,
147–155.
Overduin, M., Harvey, T.S., Bagby, S., et al., 1995. Solution structure
of the epithelial cadherin domain responsible for selective cell
adhesion. Science 267, 386–389.
Pertz, O., Bozic, D., Koch, A.W., Fauser, C., Brancaccio, A., Engel,
J., 1999. A new crystal structure, Ca2q dependence and mutational
analysis reveal molecular details of E-cadherin homoassociation.
EMBO J. 18, 1738–1747.
Pokutta, S., Herrenknecht, K., Kemler, R., Engel, J., 1994. Confor-
mational changes of the recombinant extracellular domain of E-
cadherin upon calcium binding. Eur. J. Biochem. 223, 1019–1026.
Randi, A.M., Hogg, N., 1994. I Domain of b2 integrin lymphocyte
function-associated antigen-1 contains a binding site for ligand
intercellular adhesion molecule-1. J. Biol. Chem. 269,
12395–12398.
Shapiro, L., Fannon, A.M., Kwong, P.D., et al., 1995. Structural basis
of cell–cell adhesion by cadherins. Nature 374, 327–337.
Shaw, S.K., Cepek, K.L., Murphy, E.A., Russell, G.J., Brenner, M.B.,
Parker, C.M., 1994. Molecular cloning of the human mucosal
lymphocyte integrin aE subunit. Unusual structure and restricted
RNA distribution. J. Biol. Chem. 269, 6016–6025.
Springer, T.A., 1990. Adhesion receptors of the immune system.
Nature 346, 425–434.
Sun, H., Santoro, S.A., Zutter, M.M., 1998. Downstream events in
mammary gland morphogenesis mediated by reexpression of the
a2b1 integrin: the role of the a6 and b4 integrin subunits. Cancer
Res. 58, 2224–2233.
Takeichi, M., 1995. Morphogenetic roles of classic cadherins. Curr.
Opin. Cell Biol. 7, 619–627.
Taraszka, K.S., Higgins, J.M.G., Tan, K., Mandelbrot, D.A., Wang,
J.H., Brenner, M.B., 2000. Molecular basis for leukocyte integrin
aEb7 adhesion to epithelial (E)-cadherin. J. Exp. Med. 191,
1555–1567.
Tomschy, A., Fauser, C., Landwehr, R., Engel, J., 1996. Homophilic
adhesion of E-cadherin occurs by a co-operative two-step interac-
tion of N-terminal domains. EMBO J. 15, 3507–3514.
Tuckwell, D.S., Calderwood, D.A., Green, L.J., Humphries, M.J.,
1995. Integrin a2 I-domain is a binding site for collagens. J. Cell
Sci. 108, 1629–1637.
Ueda, T., Rieu, P., Brayer, J., Arnaout, M.A., 1994. Identification of
the complement iC3b binding site in the b2 integrin CR3 (CD11by
CD18). Proc. Natl. Acad. Sci. USA 91, 10680–10684.
Zhou, L., Lee, D.H.S., Plescia, J., Lau, C.Y., Altieri, D.C., 1994.
Differential ligand binding specificities of recombinant CD11by
CD18 integrin I-domain. J. Biol. Chem. 269, 17075–17079.