Integrin receptor specificity for human red cell ICAM-4 ligandCritical residues for aIIbb3 and aVb3 binding Patricia Hermand1, Pierre Gane1, Isabelle Callebaut2, Nelly Kieffer3, Jean-Pie
Trang 1Integrin receptor specificity for human red cell ICAM-4 ligand
Critical residues for aIIbb3 and aVb3 binding
Patricia Hermand1, Pierre Gane1, Isabelle Callebaut2, Nelly Kieffer3, Jean-Pierre Cartron1and Pascal Bailly1 1
INSERM U76, Institut National de la Transfusion Sanguine, Paris, France;2De´partement de Biologie Structurale, LMCP, CNRS UMR 7590, UP6/UP7, Paris, France;3Laboratoire de Biologie et de Physiologie inte´gre´e (CNRS/GDRE-ITI), Universite´ du Luxembourg, Luxembourg
The red cell intercellular adhesion molecule-4 (ICAM-4)
binds to different members of the integrin receptor families
To better define the ICAM-4 integrin receptor specificity, cell
transfectants individually expressing various integrins were
used to demonstrate that aLb2, aMb2, and aIIbb3(activated)
bind specifically and dose dependently to the recombinant
ICAM-4-Fc protein We also show that cell surface ICAM-4
interacts with the cell surface aVb3 integrin In addition,
using a a4b1cell transfectant and b2integrin-deficient LAD
cells, we show here that ICAM-4 failed to interact with a4b1
even after a4b1 activation by phorbol ester or with the
monoclonal antibody TS2/16 (+ Mn2+) ICAM-4 amino
acids that are critical for aIIbb3 and aVb3 interaction were
identified by domain deletion analysis, site-directed
muta-genesis and synthetic peptide inhibition Our results provide evidence that the b3integrin binding sites encompass the first and second Ig-like domains of ICAM-4 However, while the
aIIbb3contact site comprises the ABED face of domain D1 with an extension in the C¢-E loop of domain D2, the aVb3
contact site comprises residues on both faces of D1 and in the C¢-E loop of D2 These data, together with our previous results, demonstrate that different integrins bind to different but partly overlapping sites on ICAM-4, and that ICAM-4 may accommodate multiple integrin receptors present on leukocytes, platelets and endothelial cells
Keywords: adhesion; ICAM-4; integrins; site-directed muta-genesis; structure model
1
The LW (Landsteiner–Wiener) blood group glycoprotein
has been renamed ICAM-4 (CD242) based on sequence
similarities with the family of intercellular adhesion
mole-cules (ICAMs) [1], a group of molemole-cules which plays a
crucial role in cell–cell interactions in the immune system
[2–4] ICAM
2 genes are all clustered on chromosome 19
except ICAM-2 which is located on chromosome 17 All
ICAM proteins exhibit several Ig-like domains, two for
ICAM-4 and ICAM-2 They bind to ab-heterodimeric
integrins and have a differential pattern of expression and
cellular distribution [5,6] Data consistent with this first
showed that ICAM-4 binds to the integrins aLb2(LFA-1,
CD11a/CD18) and aMb2 (Mac1, CD11b/CD18) [1,7]
Based on a three-dimensional model of ICAM-4 deduced
from the crystal structure of ICAM-2 [8] and using
mutated chimeric ICAM-4-Fc proteins, we have previously
shown that aLb2binds to the first Ig-domain of ICAM-4,
whereas the aMb2 binding site encompassed both
Ig-domains [9] These interactions were also reported to
be highly sensitive to the presence of divalent cations for high affinity binding to b2 integrins In parallel, it was established that ICAM-4 on red blood cells (RBCs) and transfected fibroblasts interacted specifically with recom-binant and purified I-domains of aLb2and aMb2integrins [10] More recently, using ICAM-4 positive and negative RBCs, and normal or type-I Glanzmann’s thrombastenia platelets lacking b3 integrins, we identified the platelet fibrinogen receptor integrin aIIbb3 (GPIIb-IIIa, CD41/ CD61) in its high affinity state as the receptor for RBC ICAM-4, suggesting a potential physiological significance
of the ICAM-4 mediated RBC–platelet interaction in hemostasis and thrombosis [11] Other studies, based on adhesion of hemopoietic and nonhemopoietic cells, repor-ted that ICAM-4 might also interact with aVb1 (CD51/ CD29) and aVb5integrins
these integrins is distinct but adjacent to binding sites for
b2integrins [13]
However, controversy still exists regarding the ligand specificity of a4b1, as ICAM-4 was considered as a ligand for this integrin by some authors [12] but not by others [7,9]
As some of these results were obtained with erythroid and non-erythroid cell lines that display a large number of integrins, we have developed cell adhesion assays using stable recombinant cell lines expressing unique human integrins obtained by cotransfection with appropriate cDNA encoding the a- and b-subunits We also used a permanent cell line, established from a patient with leuko-cyte adhesion deficiency [14], lacking b2 integrins but strongly expressing the a4b1 integrin Moreover, using ICAM-4 mutants generated by site-directed mutagenesis of
Correspondence to J.-P Cartron, INSERM U76, INTS, 6 rue
Alex-andre Cabanel, 75015 Paris, France Fax: +33 1 43 06 50 19,
Tel.: +33 1 44 49 30 00, E-mail: cartron@idf.inserm.fr
Abbreviations: a IIb b 3 CHO*, activated a IIb b 3 CHO cells; a IIb b 3 *,
activated a IIb b 3 integrin; LAD, leukocyte adhesion deficiency; L-cells,
L929 mouse fibroblast cells; LW, Landsteiner–Wiener; mAb,
mono-clonal antibody; MAdCAM-1, mucosal addressin cell adhesion
molecule 1; PMA, 4b-phorbol 12-myristate 13-acetate; RBC, red
blood cell; VCAM-1, vascular cell adhesion molecule 1.
(Received 2 July 2004, accepted 29 July 2004)
Trang 2surface exposed residues as well as synthetic peptides of
ICAM-4, we have identified critical amino acids involved in
the binding sites of aIIbb3
synthetic peptide inhibition that ICAM-4–b3integrin
inter-actions might be modulated, at least in vitro
Experimental procedures
Reagents
Specific monoclonal antibodies (mAbs) used include
clones Lia1/2, 7E4 and SZ21 specific for the b1-chain
(CD29), b2-chain (CD18), and b3-chain (CD61),
respect-ively; clones 25.3.1, BEAR-1, AMF7, HP2/1, SAM1 and
SZ22 recognizing the aL-chain (CD11a), aM-chain
(CD11b), aV-chain (CD51), a4-chain (CD49d), a5-chain
(CD49e) and aIIb-chain (CD41) in the absence of the
b3-chain, respectively, which were purchased from
Coulter/Immunotech (Marseille, France) mAb AP-2
specific for an epitope of the complex aIIbb3 integrin
came from GTI (Brookfield, WI, USA) Stimulatory
anti-b1 (CD29) mAb TS2/16 came from Endogen (Woburn,
MA, USA) Murine mAbs BS46/56 to ICAM-4/LWab
were described previously [15] ImmunoPure mouse IgG
and human Fc fragment came from Pierce (Rockford, IL,
USA) and ImmunoReseach-Jackson (West Grove, PA,
USA), respectively The human plasma fibronectin, and
fibrinogen depleted of vWF and fibronectin were
purchased from Life Technologies (Cergy Pontoise,
France) and Enyme Research Laboratories Inc (South
Bend, IN, USA), respectively Calcein AM was purchased
from Molecular Probes (Eugene, OR, USA) Peptides
Gly65–Val74 (residues 65–74 of ICAM-4, hereafter called
G–V), and peptide Phe26–Ser40 (residues 26–40 of
ICAM-4, hereafter called F–S) and corresponding random
peptides were synthesized and purified by Neosystem
(Strasbourg, France)
Soluble recombinant Fc proteins
Chimeric ICAM-pIgI constructs derived from intact or
mutagenized ICAM-4 (LWaallele) carrying the two Ig-like
domains D1 and D2 (residues 1–208) and deletion mutants
D1-ICAM-4 (residues 1–101) or D2-ICAM-4 (residues 102–
208) were used to produce soluble Fc-fusion proteins in
COS-7 cells as described [9] Similarly, ICAM-1- and
ICAM-2-pIgI constructs (kindly provided by D Simmons
and E Ferguson
5 , University of Oxford, UK) were used to
produce ICAM-1- and ICAM-2-Fc proteins Human
vascular cell adhesion molecule 1 (VCAM-1)-Fc
purchased from R & D Systems Europe (Abingdon,
Oxfordshire, UK)
Transfectants, cell lines and adhesion assays
L cell transfectants expressing a4b1or aVb3 integrins were
generated by cotransfection into L929 mouse fibroblast cells
(hereafter called L cells) of the CD49d (a4-chain)-pFNEO
[16] plus the CD29 (b1-chain)-pECE [17], or the CD51
(aV-chain)-pcDNA I NEO [18] plus the CD61 (b3
-chain)-pcDNA3.1 expression constructs, respectively Stable
expression of ICAM-4 in L cells was performed by
transfecting the cells with pcDNAI-ICAM-4 (LWa) [19] Transfectants resistant to G418 (geneticin) were selected
by immunomagnetic separation for a4b1 or aVb3 integrin expression or ICAM-4 expression Stable clones were isolated and the antigen expression of CD49d, CD51 and ICAM-4 was analysed by flow cytometry, as reported [11]
L cell transfectants expressing aLb2or aMb2integrins and the unactivated or dithiothreitol-activated aIIbb3CHO cell transfectants (activated cells hereafter referred to as aIIbb3 CHO* and activated integrin as aIIbb3*) were also used
in these studies [9,11] All transfectants were grown in Iscove’s modified Dulbecco medium with Glutamax-1 (Life Technologies, Inc.) supplemented with G418 600 lgÆmL)1 and 1% (v/v) amphotericin-B/penicillin/streptamycin
10% (v/v)
7,8 fetal bovine serum
Briefly, for cell adhesion assays to immobilized proteins under static conditions, indicated amounts of purified wild type or mutant ICAM-Fc, VCAM-1-Fc, fibrinogen and fibronectin in 25 mM Tris, pH 8.0, 150 mM NaCl, 2 mM
MgCl2plus 2 mMCaCl2were adsorbed overnight to flat-bottom 96-well microtiter plates (Nunc A/S, Roskilde, Denmark) The wells were then blocked for 2 h at 22C with 1% (w/v) nonfat milk in the same buffer and washed once with binding buffer A (RPMI containing 10 mM
Hepes, 2 mM MgCl2 and 2 mM CaCl2) Transfectants (1· 105per well) in 100 lL of binding buffer were added
to washed protein coated wells for 45 min at 22C before washing to remove nonadherent cells and carry out microscopic observation, as described previously [9,11] For cell–cell adhesion assays under static conditions, parental L cells and transfected aVb3 L cells (5· 104 per well) resuspended in cell culture medium, were added to each well Twenty-four hours later, the wells were washed three times with binding buffer A before adding parental
L cells and ICAM-4 L cell transfectants (1· 105per well in
100 lL of buffer A) labeled with calcein AM, as described previously [20] After 45 min at 22C, nonadherent cells were removed as indicated above using NaCl/Pi solution containing MgCl2and CaCl2 Labeled cells that adhered to the well walls were recovered by vigourous shaking and counted by flow cytometry as described [11] For blocking experiments, transfectants or protein coated wells were pretreated for 30 min at room temperature with specific mAbs (25 lgÆmL)1) or peptides (31–250 lM final concen-tration)
The Epstein–Barr virus-transformed B lymphoblastoid cell line established from a patient with leukocyte adhesion deficiency (LAD) (gift of C Figdor, Netherland Cancer Institute, Amsterdam, the Netherlands) [14] was grown in RPMI with 10% (v/v) fetal bovine serum LAD cells (1· 105 per well) in 100 lL of binding buffer A supple-mented with 5% (v/v) fetal bovine serum were used in cell adhesion assays as indicated above For blocking experi-ments, cell lines were pretreated with specific mAbs (25 lgÆmL)1) For phorbol ester activation, cells were pretreated for 15 min at 37C with 80 lM 4b-phorbol 12-myristate 13-acetate (PMA; Sigma), as described [12] For mAb anti-b1(TS2/16) activation [21], cells were washed twice in Hank’s balanced salts (Sigma–Aldrich), then resuspended in Hank’s balanced salts containing 1.0 mM
Mn2+ and the TS2/16 mAb (final concentration 1.0 lgÆmL)1) for 30 min at room temperature
Trang 3Characterization of the integrin transfectants
and cell lines
We first established a stable transfectant expressing
human a4b1or aVb3that was obtained by cotransfection
of CD49d (a4-chain) plus CD29 (b1-chain) or CD51 (aV
-chain) plus CD61 (b3-chain), respectively, into L cells
Then, all transfectants, including those established earlier
and expressing aLb2, aMb2 and aIIbb3 [9], as well as the
LAD cell line, were characterized by flow cytometry
analysis with specific mAbs Each transfectant reacted
strongly with the respective mAbs against the transfected
human a-chains, and the estimated copy number for each cell type was determined (Table 1) Parental L cells and CHO cells were nonreactive with the mAbs used (not shown) As expected, LAD cells that lack the b2-chain (CD18) had no detectable surface expression of aL and
aMintegrins (Table 1) LAD cells were negative for aIIb -and aV-chains but had a high copy number of a4-chain, and the b1 integrin was also strongly expressed (not shown) These findings indicate that a4b1 integrin is the predominant ab heterodimer expressed on LAD cells
Table 1 Integrin profiles of cell lines Integrin subunit expression was
quantified by analysis using flow cytometry Cell lines were stained
with mAbs anti-CD49d (HP2/1), anti-CD49e (SAM1), anti-CD51
(AMF7), anti-CD41 (SZ22), anti-CD11a (25.3.1) and anti-CD11b
(BEAR-1) followed by phycoerythrin-goat anti-mouse IgG (Fab¢2)
fragments The estimated number of copies per cell, measured as
specific antibody binding capacity per cell, was deduced from a
calibration curve obtained with Qifikit calibration beads (Dako,
Denmark).
Cell lines
Integrin a-subunits expressed Copies per cell Transfectants
a 4 b 1 L cell a 4 (CD49d) 10 200
a L b 2 L cell a L (CD11a) 32 000
a M b 2 L cell a M (CD11b) 46 000
a V b 3 L cell a V (CD51) 62 300
a IIb b 3 CHO a IIb (CD41) 18 600
LAD cells
a 4 (CD49d) 40 000
a L (CD11a) < 300
a M (CD11b) < 300
a 5 (CD49e) 1100
a V (CD51) 1300
a IIb (CD41) < 100
Fig 1 Specific adhesion of b-integrin transfectants to ICAM-4-Fc.
(A) Dose-dependent cell adhesion a IIb b 3 -CHO* transfectant ( ), L cell
transfectants expressing a 4 b 1 (d), a M b 2 (,), a L b 2 (n) or a V b 3 (e).
Controls (Ctrl.) including parental L cell (h), parental CHO cell
dithiothreitol-treated
24 or not, and native a IIb b 3 -CHO transfectant (s)
to ICAM-4-Fc coated to plastic wells at varying concentrations The
results are expressed as mean percentage of bound cells ± SEM of
three experiments (B) Effect of mAbs on adhesion of L cell
trans-fectants expressing a L b 2 , a M b 2 , a V b 3 and a IIb b 3 CHO*-transfectant
(dithiothreitol-activated) to coated ICAM-4-Fc (500 ngÆwell)1) Cell
transfectants were pretreated or not with saturating concentrations of
indicated mAbs specific for the a- or b-chains of each integrin (see text
for specificities) and ICAM-4 (BS46/56) Controls included unrelated
mouse IgG antibody (ctrl.IgG) and wells without Fc-protein The
re-sults are expressed as the relative percentage of bound transfectants,
where 100% is calculated from the total number of transfectants
bound to ICAM-4-Fc without mAb The mean ± SEM from three
experiments is shown By Student’s t-test analysis, ***P < 0.001, and
*P < 0.05.
Trang 4Adhesion of integrin transfectants to ICAM-4-Fc
Cell adhesion assays of parental and transfected L or CHO
cells to the chimeric ICAM-4-Fc protein coated on plastic
were performed in the presence of Mg2+ and Ca2+
(Fig 1A) Moreover, dithiothreitol activation of the aIIbb3
integrin (aIIbb3*) was performed as reported [11] All
transfectants dose-dependently adhered to coated
ICAM-4-Fc, except a4b1L cells, nonactivated aIIbb3CHO cells and
all parental cells whether treated or not with dithiothreitol
(Fig 1A) For all transfectants that bound to ICAM-4-Fc,
a plateau was reached at 500 ngÆwell)1 For all cell clones
tested, the extent of adhesion correlated with the level of
integrin expression and the relative binding: aVb3> a
IIb-b3* > aMb2> aLb2 In contrast, despite their high copy
number of a4-chains, a4b1-transfected L cells did not bind
to ICAM-4 (Fig 1A)
In order to control the specificity of the ICAM-4–integrin
interactions, the effect of different mAbs on the adhesion of
transfectants to coated ICAM-4-Fc was investigated
(Fig 1B) mAbs against either the a- or b-chains, or against
ICAM-4 blocked approximately by 50–70% the binding of
aLb2, and aMb2L cells Binding of aIIbb3CHO* cells was
blocked (70%) by mAb AP2 against a complex-specific
epitope of the aIIbb3integrin and by mAbs BS46/56 (50%)
against ICAM-4, but was not blocked by other mAbs, as
reported previously [11] In contrast, mAb SZ21, specific for
the b3-chain alone, did not block binding of aIIbb3CHO*
cells, but efficiently inhibited (60%) binding of aVb3L cells
to ICAM-4 Similarly, mAbs AMF7 and BS46/56 against
aV-chain and ICAM-4, respectively, inhibited binding of
aIIbb3CHO* cells These data strongly support a specific interaction between ICAM-4 and aVb3 As expected, control mouse IgG had no effect
As a specific interaction between red cell ICAM-4 and platelet aIIbb3 was demonstrated previously [11], we provide further evidence that cell surface ICAM-4 might interact with cell surface aVb3 Accordingly, a cell–cell adhesion assay using ICAM-4 and aVb3 L cell transfect-ants, was performed Figure 2 shows that ICAM-4
L cells did not bind to immobilized parental L cells, but did bind to immobilized aVb3 L cells This cell–cell adhesion was efficiently blocked ( 57%) by BS56 mAb specific for ICAM-4 while control mouse IgG had no effect As another control, parental L cells did not bind significantly to immobilized aVb3 L cells, as expected Altogether, our findings indicate that ICAM-4 interacts specifically with the four integrins aLb2, aMb2, aIIbb3* and aVb3 but not with the a4b1 integrin This raised the possibility that a4b1 might need some activation inducing
a favorable conformational change to interact with ICAM-4 To test this hypothesis, the effect of PMA stimulation [22] and of antibody activation on a4b1 -mediated cell adhesion [21,23] was investigated As shown in Fig 3A, after activation with either PMA or mAb TS2/16 (+ Mn2+), a4b1 L cells still failed to bind
to coated ICAM-4-Fc, while both native and stimulated
a4b1 L cells showed adhesion to coated VCAM-1-Fc (from 10 to 32%) and fibronectin (from 27 to 39%), two well known ligands of a4b1[16,24,25] PMA pretreatment had a 1.4-fold stimulatory effect on adhesion to fibro-nectin, but no significant enhancement on adhesion to VCAM-1-Fc By contrast, stimulation by mAb TS2/16 (+Mn2+) enhanced adhesion to VCAM-1-Fc and fibronectin by 3.2- and 1.2-fold, respectively We also found that even under experimental conditions (buffer composition and cations) similar to those described by Mankelow et al [13], a4b1 L cells did not react with ICAM-4-Fc (not shown) As a control for these experi-ments, no binding of native or treated parental L cells to coated ICAM-4-Fc, VCAM-1-Fc and fibronectin was observed
Adhesion of LAD cells to ICAM-4-Fc
To obtain further evidence that ICAM-4 may or may not interact with a4b1, we examined the binding of LAD cells, which strongly express a4b1, but not b2 and b3 integrins (Table 1), to ICAM-4, VCAM-1 and fibronectin Figure 3B shows that whatever the experimental conditions, including activation by PMA or mAb TS2/16, LAD cells did not bind
to ICAM-4-Fc, but efficiently bound to VCAM-1-Fc (39–79%) and fibronectin (11–25%) In the presence of mAb TS2/16, LAD cell adhesion to VCAM-1 was stimu-lated twofold, whereas PMA-stimustimu-lated cell adhesion to fibronectin was only 2.2-fold LAD cells carry a4-, b1-, a5 -and aV-chains (Table 1) and therefore probably express the three integrins a4b1, a5b1and aVb1 Figure 3B shows that binding of LAD cells to VCAM-1 was almost completely blocked by mAbs specific for a-chain (clone HP2/1) or
Fig 2 Adhesion of ICAM-4 L cell transfectant to immobilized a V b 3
L cell transfectant ICAM-4 L cells (black bars) or parental L cells
(white bar) (10 5 cellsÆwell)1) labeled with the fluorescent dye calcein (#)
were deposited onto confluent a V b 3 L cell or parental L cell
mono-layers ICAM-4 L cells, either native or pretreated with saturating
concentrations of mAbs to ICAM-4 (BS56) or with unrelated mouse
IgG antibody (ctrl.IgG), were used At the end of the incubation period
and after washings, adherent calcein-labeled cells were collected and
counted by flow cytometry analysis The results are expressed as mean
percentage of bound cells ± SEM of three experiments By Student’s
t-test analysis, *P < 0.05.
Trang 5b1-chain (clone Lia1/2), whereas binding to fibronectin was
only weakly affected As a control, mouse IgG had no
effect These data provide clear evidence that, native and
stimulated L cell transfectants or LAD cells expressing an
active form of a4b1 integrin are unable to interact with
ICAM-4, although they interact with VCAM-1 and
fibro-nectin Identical results were obtained when the same buffer
and cation composition described by Mankelow et al [13]
was used (not shown)
b3Integrin binding sites on ICAM-4
Previous studies from our laboratory have shown that the
b2integrins aLb2and aMb2, interact with ICAM-4 through
distinct contact sites [9] The fact that mAb SZ21 did not
block aIIbb3* binding to ICAM-4 but inhibited aVb3
binding suggested that these integrins may also establish
distinct contact sites with ICAM-4 To determine which
ICAM-4 Ig domains contain aIIbb3and aVb3binding sites,
ICAM-4 domain deletion mutants lacking either domain
D1 or D2 were used in cell adhesion assays The binding of
aIIbb3CHO* and aVb3L cells required the presence of both
extracellular Ig domains because the absence of either
domain D1 or D2 was associated with a 55% decrease in
adhesion (Fig 4) To better define the binding sites of both
b3 integrins, we tested 32 ICAM-4-Fc mutant proteins,
obtained by site-directed mutagenesis of the most prominent
surface exposed residues, which could participate in
molecular interactions
located on the ABED and CFG faces of domain D1 and on the C¢-E loop of domain D2, and have previously been used
to define critical amino acids involved in the interaction between ICAM-4 and b2integrins [9]
As shown in Fig 4A, aIIbb3CHO* cell adhesion to six ICAM-4-Fc mutants, namely the mutants Gln30fi Ala, Gly32fi Ala, Lys33 fi Ala, Gln36 fi Ala, Trp77 fi Ala and Glu151fi Ala, was reduced by more than 55% as compared to wild type ICAM-4 Five out of the six mutated residues were located in domain D1, in the A-B loop (Gln30fi Ala, Gly32 fi Ala), at the beginning of the B strand (Lys33fi Ala, Gln36 fi Ala) and in the middle of the E-F loop (Trp77fi Ala), while the sixth residue (Glu151fi Ala) was located in domain D2, at the top of the C¢-E loop Based on these results, we conclude that the six amino acids critical for ICAM-4 binding to aIIbb3* integrin are spatially close and located at the bottom of the ABED face of domain D1 with an extension in the domain D2 through residue Glu151 (Fig 5, left)
Ten mutations located within domain D1 in the B and E strands at positions Lys33, Trp66, Tyr69 and Asp73, in the
C, F and G strands at positions Arg52, Leu80 and Arg97, in the E-F loop at position Trp77 and in the C¢-E loop of domain D2 at positions Glu151 and Thr154 caused a reduced adhesion of aVb3L cells by at least 40% (Figs 4B and 5, right)
10 These data suggest that the aVb3binding site
on ICAM-4 comprises residues on both the ABED and CFG faces of domain D1 and in the C¢-E loop of domain D2 (Fig 5, right) In addition, the three-dimensional
Fig 3 Adhesion of a 4 b 1 L cell transfectant
and LAD cells to CAM-Fc and fibronectin.
(A) Adhesion of a 4 b 1 L cells (gray bars) and
parental L cell (black bars) to coated
ICAM-4-Fc, VCAM-1-Fc and fibronectin (FN)
(500 ngÆwell)1) in binding buffer A containing
2 m M MgCl 2 plus 2 m M CaCl 2 Cells were
untreated (left), pretreated with 4b-phorbol
12-myristate 13-acetate (PMA) (middle) or
used under optimal binding conditions in the
presence of mAbTS2/16 and 1.0 m M Mn2+
(right) ***P < 0.001 vs parental L cell (B)
Adhesion of LAD cells to coated ICAM-4-Fc
(black bars), VCAM-1-Fc (hatched bars) and
fibronectin (FN, white bars) in binding
buf-fer A containing 2 m M MgCl 2 plus 2 m M
CaCl 2 Cells were pretreated or not with
sat-urating concentrations of indicated mAbs
specific for a 4 -chain (HP2/1), b 1 -chain (Lia1/
2) (left), pretreated with PMA (middle) or used
under optimal binding conditions as indicated
(right) Control including unrelated mouse
IgG antibody and Hank’s balanced salts
binding buffer without cations and
stimula-tory mAb The results are expressed as mean
percentage of bound cells ± SEM of three
experiments.
Trang 6ICAM-4 model reveals that six of 10 critical residues are
clustered between the CFG face of D1 and the C¢-E loop of
D2 which are spatially close (Fig 5, far-right at 90)
Regarding residues 97 and 151, we found that
substitu-tion of a charged residue (Glu or Arg, respectively) by the
opposite charge had little or no effect on cell adhesion,
whereas substitution by an apolar amino acid (Ala)
inhibited adhesion Moreover, replacement of the aromatic
residue (Trp) at position 77 by a small apolar (Ala), but not
by another aromatic residue (Phe), reduced adhesion of
aIIbb3CHO* cells, although this may not be true for avb3
L cell adhesion (Fig 4) The replacement of the apolar
residue (Leu) at position 80 by another apolar residue (Gly)
had no effect, whereas substitution by an aromatic residue
(Phe) severely reduced avb3 L cell adhesion These
obser-vations suggest the importance of the surface exposed
residues over their electric charge or hydrophobicity in
ICAM-4/b3 integrin interaction, as already found for b2
integrin interaction [9]
Inhibition of ICAM-4–b3integrin interaction
by ICAM-4 peptides
To confirm the results of site-directed mutagenesis, blocking
experiments with ICAM-4 synthetic peptides (31–250 l )
covering the binding regions involved in b3 integrin interaction were performed We used the Phe26–Ser40 peptide (F–S) covering residues Gln30, Gly32, Lys33 and Gln36 involved in aIIbb3* interaction, and the Gly65–Val74 peptide (G–V) covering residues Trp66, Tyr69 and Asp73, involved in aVb3interaction
Figure 6A shows that adhesion of aIIbb3CHO* cells to ICAM-4 was efficiently inhibited by 63% and 78% with the F–S and G–V peptides (125 lM), respectively, whereas corresponding random peptides used as controls had no effect The G–V peptide inhibited adhesion of aIIbb3CHO* cells to fibrinogen by 64%, whereas the F–S peptide did not Because the G–V peptide harbors a Gln-X-X-Asp-Val motif involved in the aIIbb3*–fibrinogen interaction [26] and because mutations at position 69 and 73 did not affect adhesion of aIIbb3 CHO* cells to ICAM-4 (Fig 4A), we conclude, contrary to our previous speculation [11], that the
aIIbb3* binding site resides along the A-B strand (peptide F–S) but not on the E strand (peptide G–V) in domain D1
of ICAM-4 In addition, F–S peptide inhibited the binding
of aIIbb3 CHO* cells to ICAM-4 in a concentration-dependant manner (Fig 6, bottom)
Adhesion of aVb3 L cells to ICAM-4 was efficiently inhibited by 55% and 90% with 125 and 250 lMof the G–V peptide, respectively, whereas the F–S peptide and corres-ponding random peptides had no inhibitory effect (Fig 6B) These data are consistent with those of site-directed mutagenesis (Fig 4B) and strongly suggest that the aVb3 binding site partly resides along the E strand
Discussion
Current experimental evidence suggests that red cell ICAM-4 is a ligand for b2integrins, which is of potential significance in a variety of physiological processes, inclu-ding erythropoiesis and red cell destruction [7,9,11,27] The physiological relevance of these molecular interactions is also supported by the fact that, under depressed venous shear rates, adhesion of RBCs to activated adherent neutrophils is mediated through aMb2–ICAM-4 interaction [28] More recently, we found that ICAM-4 interacts with the activated platelet fibrinogen receptor, integrin aIIbb3* [11], suggesting that this interaction might be responsible for platelet–red cell aggregate formation in patients with sickle cell anemia, further contributing to the vaso-occlu-sion events characteristic of this disease [29,30] Here, we have confirmed and further extended these findings by providing the first demonstration that human aVb3integrin expressed in L cells functions as a receptor for recombinant ICAM-4 protein and ICAM-4 expressed in L cells, thus suggesting a potential role in the physiology and pathology implicating endothelial cells and platelets Indeed, prelim-inary data indicate that ICAM-4 might be the sickle red cell protein that mediates binding to endothelial aVb3 integrin through a protein kinase A
pathway [31] and that mAbs specific to aVb3inhibit sickle red blood cell–endothelium interaction induced by platelet-activating factor [32] Our current results extend the data from Spring et al [12] who showed that ICAM-4 is a novel ligand for the aV integrin family, although their studies only identified aVb1 and aVb5, but not aVb3, as ICAM-4 receptors
Fig 4 Adhesion of a IIb b 3 CHO* and a V b 3 L-cell transfectants to
ICAM-4 mutants Equivalent amounts of wild type ICAM-4-Fc,
ICAM-4-Fc domain deletion mutants (D1 or D2) and 32 ICAM-4-Fc
mutants (indicated using one letter amino acid code), were coated on
plastic well (500 ngÆwell)1) and used in a cell adhesion assay with a IIb b 3
CHO* (dithiothreitol-activated) (A) and a V b 3 L cells (B) Negative
control (Ctrl.) was carried out in the absence of ICAM-4-Fc protein.
The results are expressed as in Fig 1B The mean ± SEM from three
determinations is shown.
Trang 7Our findings and those reported above clearly indicate
that ICAM-4 exhibits the unusual property of
accommo-dating a broad repertoire of integrin receptors involved in
different cellular functions [7,9,11–13] However,
contro-versial results have been published with respect to a4b1
interaction with ICAM-4 [9,12] Therefore, in order to
further delineate the integrin receptor specificity of
ICAM-4, we have developed cell adhesion assays using various cell
lines that express recombinant a4b1such as L cells, or LAD
cells that express a4b1, but lack b2integrins as well as aIIb
-and aV-chains Our results show that although both cell
lines strongly express a functional a4b1integrin, as assessed
by VCAM-1 and fibronectin binding after activation, they
do not bind to immobilized ICAM-4-Fc Thus, we conclude
that ICAM-4 is not a ligand for a4b1 integrin On a
structural basis, although a minimal adhesion motif LDV is
present on ICAM-4, it may not be sufficient for integrin
binding [33,34] Moreover, the critical motif IDSP required
for a4b1integrin binding to its known ligands [VCAM-1,
mucosal addressin cell adhesion molecule 1 (MAdCAM-1)
is missing on ICAM-4 [35]
Next, to complement previous studies defining the critical
residues of ICAM-4 involved in b2integrin interaction we
performed an analysis of ICAM-4–b3integrin interactions
at a molecular level with essentially the same panel of
ICAM-4 mutants [9] The mutated residues were chosen according to their significant solvent accessibility observed
on the modeled structure of ICAM-4 Mutations were thus expected to affect the function and interaction potential, rather than the structure itself However, an effect on the structure could not be excluded especially for the Tyr69 mutant, which exposes only its phenol hydroxyl to solvent These studies revealed that distinct but partly overlapping sites spread over the domains D1 and D2 of ICAM-4 mediate interactions with aIIbb3* and aVb3, as both integrin binding sites share some residues (Lys33, Trp77 and Glu151) The aIIbb3* binding site comprises at least six residues clustered at the bottom of the ABED face (Gln30, Gly32, Lys33, Gln36) and at the top of the C¢-E loop of domain D1 (Trp77) and D2 (Glu151), respectively (Fig 5) Sequence alignments revealed that these residues are not conserved within human ICAM proteins [5], explaining why
aIIbb3* integrin has not been found to interact with other members of the ICAM family
The aVb3 binding site comprises at least 10 residues, including eight distributed on the ABED face (Lys33, Trp66, Tyr69, Asp73) and the CFG face (Arg52, Trp77, Leu80 and Arg97) of domain D1 and two (Glu151 and Thr154) in the C¢-E loop of domain D2 (Fig 5) These data provide evidence for a more extensive binding area between
Fig 5 Critical residues of ICAM-4 involved in interaction with b 3 integrins Three-dimensional representation of an ICAM-4 model constructed on the basis of ICAM-2 structure A ribbon representation of the secondary structures of the ICAM-4 model is shown inside its transparent solvent-accessible surface Each domain (D1 and D2) shows two faces formed by the ABED strands of one antiparallel b-sheet and the CFG strands of the second antiparallel b-sheet Each strand is labeled by an italic capital letter Amino acids are designated by a one letter code and residue numbers The solvent-accessible surfaces of critical amino acid residues involved in the ICAM-4–a IIb b 3 (left side) and –a V b 3 interactions (right side, two views
at 90 to each other) are colored (red, acidic residues Q30, Q36, D73, E151; blue, basic residues K33, R52, R97; pink/purple, aromatic residues W66, Y69, W77; green, hydrophobic residue L80; yellow/orange, other residues G32, T154 Residues involved in both b 3 integrin binding sites are underlined This figure has been prepared using the SWISSPDB VIEWER software [45].
Trang 8ICAM-4 and aVb3 than between ICAM-4 and other
integrins Accordingly, we conclude that the aVb3 site
comprises residues on both faces of domain D1 for
interaction with ICAM-4, as previously reported for
ICAM-3 binding to aLb2, and it was postulated that this
might reflect a distinct stoichiometry
However, the critical residues involved in aVb3interaction
appear to cluster on the edges of both b sheets, defining a
potential contact area (Fig 5)
Protein sequence examination suggests that aVb3does not
interact with other human ICAM molecules as the critical
residues identified here are not conserved among other
family members Different sites on immunoglobulin
super-family
15 proteins are known to mediate binding to integrins
(for review see [6]), defining for each protein unique
structural features which can be associated with its ligand
specifity: a flat surface on the D1 domain of ICAM-1 and
ICAM-2, with a glutamic acid playing a key role (see
below), a protruding CD loop on the D1 domain of
VCAM-1 and MAdCAM-1 including a critical aspartic
acid Interestingly, an unusually long D strand in domain
D2 of MAdCAM-1 appears to contribute to
integrin-binding [37] These two separate integrin-recognition motifs
encompassing both IgSF domains of MAdCAM-1 might
thus be compared to those defined here on ICAM-4 for
aIIbb3* and aVb binding, probably forming a cleft for ligand binding Binding sites involving two consecutive IgSF domains were already observed in other IgSF proteins, not involved in integrin binding, such as the cytokine receptor family (hormone-binding site I) [38] Importantly, except for positions 32 (aIIbb3), 66 and 80 (aVb3), most mutations affecting ICAM-4 interaction with b3 integrins involved amino acid residues different from those implica-ted in the binding of the monoclonal antibody BS46/56 that includes positions 19, 32, 56, 66, 70, 75, 80 and 91 in domain D1 [9] Therefore, it cannot be excluded that mutant L80G did not react correctly with aVb3due to protein misfolding,
as previously noted for interaction with b2 integrins [9] However, as discussed below, residues 32 and 66 belong to peptide regions involving amino acids 26–40 (peptide F–S) and 65–74 (peptide G–V), respectively, which are clearly involved in ICAM-4–aIIbb3and ICAM-4–aVb3interactions Inhibition studies with the synthetic peptides F–S (Phe26–Ser40) and G–V (Gly65–Val74) located on A-B loop/B strand and E strand, respectively, provided addi-tional support to the b3binding sites on ICAM-4 identified
by site-directed mutagenesis The results point to the critical role of F–S peptide in the ICAM-4–aIIbb3* interaction and
of G–V peptide in the ICAM-4–aVb3interaction However, the G–V peptide also inhibits ICAM-4–a b* interaction
Fig 6 Effects of ICAM-4 peptides on adhe-sion of b 3 transfectants to ICAM-4 and fibrin-ogen Adhesion of a IIb b 3 CHO* cells (dithiothreitol-activated) (A) and a V b 3 L cell transfectant (B) to coated ICAM-4-Fc and fibrinogen (500 ngÆwell)1) and effects of ICAM-4 peptides on cells binding Cell transfectants were pretreated or not with the peptides Phe26–Ser40 (F–S, black bars) and Gly65–Val74 (G–V, hatched bars) derived from ICAM-4 at different final concentrations (31–250 l M ) After 30 min incubation, the cells were tested for binding to coated ICAM-4-Fc or fibrinogen For each peptide tested, the corresponding random (rd, white bars) peptide was used as control The results are expressed as in Fig 1B 100% relative per-centages are in gray bars for each histogram The mean ± SEM from three experiments is shown.Student’st-test analysis,***P < 0.001.
Trang 9It is worth noting that these two peptides are adjacent on
the three-dimensional level (Fig 5) As the G–V peptide
harbors a Gln-X-X-Asp-Val motif involved in aIIbb3*–
fibrinogen interaction, it might occupy the fibrinogen
binding site of aIIbb3*, thus preventing the docking of
ICAM-4 This is consistent with the observation that the
peptide G–V, but not F–S, also prevents aIIbb3*–fibrinogen
interaction (Fig 6A) Therefore, it is assumed that these
peptides might be of interest to modulate ICAM-4–b3
integrin interactions occurring in some physiological as well
as pathological situations
Comparison of the present findings with our previous
results regarding ICAM-4–b2 integrin interaction [9] and
with very recent data on ICAM-4 interaction with aVb1and
aVb5[13] are summarized in
conclusions: (a) the binding sites of all integrins encompass
the two Ig-like domains of ICAM-4, except that for aLb2
which is restricted to domain D1, (b) the b2and b3(aIIband
aV) binding sites, which were defined with the same panel of ICAM-4 mutants, appear clearly distinct although they all share Trp77, thus suggesting that there is no consensus binding motif and that ICAM-4 binding might be modu-lated by I/A domain interactions (see below), (c) the E-F loop of domain D1 and the C¢-E loop of domain D2 which are spatially close (Fig 5) bear residues Trp77 and Glu151, respectively, that are crucial for binding to aMb2and b3(aIIb and aV) integrins and (d) the aVb3binding site defined here
is distinct from the reported binding site for aVb1and aVb5, although all share at least two residues (Trp66, Arg97), although this should be considered with caution as analyses were carried out with different sets of ICAM-4 mutants [13] With this restriction in mind, it appears that amino acids
Fig 7 Compilation of amino-acid residues which are involved in ICAM-4/b integrin interactions (A) Schematic two-dimensional representation of the D1 domain of ICAM-4 showing the the two antiparallel b-sheets with ABED and CFG strands, and the loops between strands (B) Compilation and location of relevant amino acids on ICAM-4 which caused reduced interaction with a L b 2 , a M b 2 , a IIb b 3 *, a V b 3 , a V b 1 and a V b 5 integrins Right column illustrates predicted localization of residues on ICAM-4 Strands are illustrated by vertical black bars and loops by open rectangles.
Trang 10critical for binding to b2and aV(b1and b5) integrins are
mainly located in the A and G strands of domain D1 and
share a common residue (Arg97)
I/A domains of integrins, including a metal-coordination
site, play a key role in ligand recognition Aspartate or
glutamate residues of ligands form part of the coordination
sphere and consequently are central to the integrin
recog-nition site [39] I/A domains are present on top of all the b
subunits, whereas they are present only in some a-subunit
(in aLand aM, but not in aIIband aV), inserted in the loop of
a b propeller structure In the aVb3experimental structure,
the propeller, lacking an I/A domain, is in complex with the
b I/A domain to form the ligand binding head of the
integrin [40] Binding of a cyclic peptide presenting an RGD
sequence involves major interface between the aV- and
b3-subunits and extensive contacts with both, including
formation of salt bridges between the arginine side chain
and aspartic residues on the integrin [41] A similar situation
might occur for the interaction of ICAM-4 with aVb3(with
a central role played by Arg97) and also with aIIbb3, both of
which lack I/A domains in the a-chains and for which we
have shown here that a relatively large surface of ICAM-4
might be involved
In contrast to integrins lacking an a I/A domain, only the
I/A domain inserted into the a-subunit appears to be
sufficient for integrin binding [39] Recombinant a I/A
domains can indeed bind ligand with the same affinity and
specificity as intact integrins Accordingly, ICAM-4 binding
to aLb2and aMb2integrins solely involves the a I/A domain
[10], similarly to what occurs for the ICAM-1–aLb2 or
–aMb2 interaction [42] However, the interface between
ICAM-4 and a I/A domain probably differs drastically
from that observed for ICAM-1 [42], due to the presence of
an arginine (Arg52) in ICAM-4, instead of the critical
ICAM-1 glutamic acid (Glu34) which coordinated the I/A
domain Mg2+ Regardless of the connection of I/A
domains to
17 a (aLb2/aMb2) or b (aIIbb3/aVb3) chains, the
exact nature of the different interfaces of ICAM-4 with
these domains remains to be determined Although it is
possible that an acidic residue of ICAM-4 in another
position might participate in the metal-binding site of I/A
domains, recent studies [13] combined with those reported
here do not support such an hypothesis, indicating that the
two acidic residues of domain D1 (Glu25 and Asp73) and
most of those from domain D2 (except Asp157 and Asp187
which were not tested) do not alter the interaction with
integrins However, ICAM-4–integrin interaction is clearly
divalent cation-dependant, as other ICAM–integrin
inter-actions [7,9], and the presence of both Ca2+and Mg2+is
required for maximal binding of ICAM-4 to the I domain
of b2 integrins [10] This is in line with our preliminary
observation showing that binding of anti-LW to the LW
antigen of red cells, as detected by agglutination tests, is
inhibited by EDTA [43] As EGTA
agglutination [43], current data suggest that chelation of
Mg2+is critical for both antigen and integrin interaction
and that calcium plays a critical role only for integrin
interaction Yet, the binding sites of these cations remain
unknown
The mutagenesis data should also be considered in light
of the dimeric state of ICAM-4 In RBCs, ICAM-4 is a
noncovalently linked accessory chain of the Rhesus
19membrane complex [44], but whether it is present as a
monomer or as a dimer is currently unknown According to the interface observed for the ICAM-1 dimer [42], which is suggested to mediate physiologic dimerization on the cell surface, ICAM-4 dimerization might also largely involve the ABED sheet of D1 domain However, in the cell adhesion assay used in our studies, the recombinant ICAM-4-Fc fusion protein is present as an obligate dimer and therefore impairment of interaction with different integrins for mutants in this region, as seen in cell adhesion assays, might reflect either an effect on the oligomerization or a direct effect on ligand binding Different stoichiometry, however, cannot be excluded, as already suggested by Bell and colleagues [36]
20
In conclusion, because of its broad repertoire of integrin receptors, ICAM-4 is likely to play a role in the normal physiology through interaction with macrophages, platelets and leukocytes that express a variety of integrins, and through abnormal interactions with blood cells and/or vascular endothelium in pathologic situations, as discussed above However, more studies are required to better document such interactions and their potential modulation
by synthetic peptides derived from ICAM-4 as additional specific tools for future therapeutic strategies in situations were these interactions may occur
Acknowledgements
This investigation was supported in part by the Institut National de la Sante´ et de la Recherche Me´dicale (INSERM) and the Institut National
de la Transfusion Sanguine (INTS).
We would like to thank Mrs M Huet (INTS, Paris) for technical assistance We also thank Dr P.J Newman (Blood Center of south-eastern Wisconsin, Milwaukee, WI), Dr D Cheresh (Scripps Reseach Institute, La Jolla, CA), Dr M Hemler (Dana Faber, Cancer Institute, Boston, MA) and Dr E Ruoslahti (Burnham Institute, La Jolla, CA) for the supply of the b 3 -chain-pcDNA3.1, a V -chain-pcDNAI-NEO,
a 4 -chain-pF-NEO and b 1 -chain-pECE constructs, respectively.
References
1 Bailly, P., Hermand, P., Callebaut, I., Sonneborn, H.H., Kha-mlichi, S., Mornon, J.P & Cartron, J.P (1994) The LW blood group glycoprotein is homologous to intercellular adhesion molecules Proc Natl Acad Sci USA 91, 5306–5310.
2 Springer, T.A (1990) Adhesion receptors of the immune system Nature 346, 425–434.
3 Luscinskas, F.W & Lawler, J (1994) Integrins as dynamic reg-ulators of vascular function FASEB J 8, 929–938.
4 Hayflick, J.S., Kilgannon, P & Gallatin, W.M (1998) The inter-cellular adhesion molecule (ICAM) family of proteins New members and novel functions Immunol Res 17, 313–327.
5 Gahmberg, C.G., Tolvanen, M & Kotovuori, P (1997) Leukocyte adhesion – structure and function of human leukocyte beta2-integrins and their cellular ligands Eur J Biochem 245, 215–232.
6 Wang, J & Springer, T.A (1998) Structural specializations of immunoglobulin superfamily members for adhesion to integrins and viruses Immunol Rev 163, 197–215.
7 Bailly, P., Tontti, E., Hermand, P., Cartron, J.P & Gahmberg, C.G (1995) The red cell LW blood group protein is an inter-cellular adhesion molecule which binds to CD11/CD18 leukocyte integrins Eur J Immunol 25, 3316–3320.