THE JOURNAL OF BIOLOGICAL CHEMISTRY
© 1997 by The American Society for Biochemistry and Molecular Biology, Inc.
Vol. 272, No. 38, Issue of September 19, pp. 23912–23920, 1997
Printed in U.S.A.
Identification of a Region in the Integrin b3 Subunit That Confers
Ligand Binding Specificity*
(Received for publication, May 21, 1997)
Emme C. K. Lin‡, Boris I. Ratnikov, Pamela M. Tsai, Christopher P. Carron§, Debra M. Myers§,
Carlos F. Barbas III¶, and Jeffrey W. Smithi
From the Program on Cell Adhesion and the Extracellular Matrix, The Burnham Institute, La Jolla Cancer Research
Center, and the ¶Department of Molecular Biology, The Scripps Research Institute, La Jolla, California 92037, and
§G. D. Searle Research and Development, Monsanto Co., St. Louis, Missouri 63198
Many integrin adhesion receptors bind ligands containing the Arg-Gly-Asp (RGD) peptide motif. Most integrins exhibit considerable specificity for particular
ligands and can distinguish among the many conformations of RGD. In this study we identify the domain of the
integrin b subunit involved in determining ligand binding specificity. Chimeras of b3 and b5, the most homologous integrin b subunits, were expressed with av on
the surface of human 293 cells. The ligand binding phenotype of each chimera was assessed using the ligands
Fab-9 and fibrinogen, both of which have a binding preference for avb3. The results of the study show that when
exons C and D of the b3 subunit (residues 95–233) are
substituted into b5, the chimera gained the ability to
bind Fab-9 with an affinity close to that of wild-type
avb3. This chimera was able to mediate cell adhesion to
fibrinogen. Furthermore, the swap of only a 39-residue
segment of this larger domain, b3 residues 164 –202, into
the backbone of b5 enabled the chimeric integrin to bind
soluble Fab-9. This small domain is highly divergent
among the integrin b subunits, suggesting that it may
play a role in determining ligand selection by all
integrins.
Integrins are transmembrane ab heterodimers that are responsible for most of the physical contacts between cells and
the extracellular matrix. Integrins are involved in a number of
tissue remodeling events including embryogenesis, angiogenesis, wound repair, and bone resorption (1–5). Integrins are also
causally linked to pathological conditions such as tumor progression, thrombosis, inflammation, and osteoporosis (6 – 8).
The integrin protein family contains 13 a and 9 b subunits.
These subunits can cross-pair to form heterodimers with different expression patterns and distinct ligand binding profiles
(9, 10). Each integrin binds to only a limited series of ligands,
ensuring that cell adhesion and migration are precisely regulated. Such precision is key to tissue remodeling.
Based on their ligand recognition specificity, integrins can be
divided into two broad sub-families. About one-half of the integrins bind to ligands that lack any consensus binding motif.
* This work was supported in part by National Institutes of Health
Grants CA56483 and AR42750 (to J. W. S.). The costs of publication of
this article were defrayed in part by the payment of page charges. This
article must therefore be hereby marked “advertisement” in accordance
with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ Supported in part by a postdoctoral fellowship from the Cancer
Research Institute and by a fellowship from Searle.
i Established Investigator of the American Heart Association and
Genentech. To whom correspondence should be addressed: The Burnham Institute, La Jolla Cancer Research Center, 10901 N. Torrey Pines
Road, La Jolla, CA 92037. Tel.: 619-646-3121; Fax: 619-646-3192.
The remaining integrins bind to the RGD1 consensus sequence.
Importantly though, even the integrins that bind to RGD display a great deal of specificity for matrix ligands (11, 12). Some
of this specificity is determined by contacts that are ancillary to
the RGD sequence (13, 14), but much of the specificity is determined by the shape of the RGD motif and the residues in its
immediate vicinity. Thus, to understand the structural basis of
matrix selection, it is important to understand how integrins
distinguish the many conformations of RGD.
The goal of the present study was to identify domain(s)
within the integrin b subunit that confer ligand binding specificity. As a model system, we compared the ligand binding
domains of the two most homologous integrins, avb3 and avb5.
The b3 and b5 subunits have 56% identity at the amino acid
level (15, 16). Although both avb3 and avb5 recognize the RGD
motif, the avb3 integrin binds to at least six more ligands than
avb5, indicating that avb3 has a more relaxed ligand binding
specificity (17). This difference in the way the two integrins
bind to ligand may ultimately be related to their distinct biological functions. For example, while both avb3 and avb5 can
mediate cell adhesion, only avb3 mediates cell spreading and
migration in the absence of an external stimulus (18, 19).
In the present investigation, we constructed a series of chimeras between b3 and b5 and assessed the ability of the
chimeric integrins to bind to two ligands that exhibit significant binding preference for avb3 over avb5. One ligand, Fab-9,
is an engineered Fab that contains an RGD sequence within
the heavy chain complementarity determining region 3. Fab-9
was selected from a phage display library for the ability to bind
the b3-integrins with high affinity (20). In fact, this ligand
exhibits greater than 1000-fold higher affinity for avb3 than for
avb5 (20). Another ligand, fibrinogen, is a protein involved in
blood clotting and is thought to be a physiologic ligand for avb3
on endothelial cells (21, 22). Prior study has shown that it fails
to bind to avb5 (17). Unlike Fab-9, fibrinogen contains multiple
integrin-binding sites: two RGD motifs and an additional adhesive domain in the carboxyl terminus of its g chain (23, 24).
Analysis of the b3/b5 chimeras shows that exons C and D of the
b3 subunit are key to determining the binding specificity for
Fab-9 and fibrinogen. However, within this larger domain, a
small region of only 39 amino acid residues has a significant
impact on ligand selection. An alignment of this small domain
in all integrin b subunits shows it to be highly divergent.
Structural algorithms also predict the region to be conformationally flexible. This is the first identification of a domain
within the integrin b subunit involved in ligand selection. In
1
The abbreviations used are: RGD, arginine-glycine-aspartate;
FACS, fluorescence-activated cell sorting; mAb, monoclonal antibody;
PAGE, polyacrylamide gel electrophoresis; PCR, polymerase chain reaction; PD, pixel density.
23912
This paper is available on line at http://www.jbc.org
Ligand Binding Specificity of avb3
conjunction with our prior studies that map a region of the a
subunit that determines ligand binding specificity (25), these
results provide a structural basis for matrix selection by
integrins.
EXPERIMENTAL PROCEDURES
Antibodies and Synthetic Peptides—Monoclonal antibody L230 (antiav) was purified from the culture supernatant of the hybridoma obtained from ATCC. The anti-avb3 mAb LM609 was purchased from
Chemicon. The mAb P112-4C1 that is specific for the avb3 complex was
generated by conventional procedures at Searle/Monsanto. This antibody is specific for the complex between av and b3 and blocks the
integrin’s ligand binding function.2 The mAb RUU-PL7F12, which recognizes the b3 subunit, and mAb P1F6 (anti-avb5) were purchased
from Becton Dickinson. The mAb P3G2 (against avb5) was provided by
Dr. David Cheresh (Scripps Research Institute). The polyclonal antibody directed against the b5 cytoplasmic tail was a kind gift from Dr.
Lou Reichardt. Fab-9 is specific for the avb3 integrin ligand binding
site. The characterization of this antibody has been previously described (20, 26). Goat anti-human F(ab9)2 IgG was obtained from Pierce.
Nonspecific mouse IgG was purchased from Calbiochem. Synthetic
peptides with sequences GRGDSP and SPGDRG were synthesized by
Coast Scientific.
Construction of Chimeras between Integrin b3 and b5—Wild-type b3
and b5 cDNA were cloned using PCR and verified by sequencing.
Chimeras between b3 and b5 were designed based on the genomic
sequence of b3, of which the exon junctions (labeled A–N) have been
delineated (27). As b5 is highly homologous to b3 in amino acid sequence, we designed the chimeras based on their sequence homology
and placed the junctions of each chimera at or near the exon boundaries. Chimeras were constructed by PCR using b3/b5 hybrid primers
and primers that encompassed an endogenous restriction site. The
resultant hybrid cDNA, consisting of b3 and b5 sequence flanked by
restriction sites, was then cloned into the cDNA for b5 at endogenous
restriction sites to complete the full-length integrin subunit. b3/b5
hybrid primers (listed 59 to 39) used to construct the junctions (/) of the
chimeric integrin b subunits were as follows: b3-(A–B)/b5, GGTCTTGTCACC/TGGCCGGAGCCGGAGTGCAATCCT and CGGCTCCGGCCA/GGTGACAAGACCACCTTC; b3-(A–D)/b5, ACAGTCTGTGAT/GAGAAGATTGGCTGGCGAAAGGAT and GCCAATCTTCTC/ATCACAGACTGTAGCCTGCATGAT; b3-(C–D)/b5, CTTCGAATCATC/GGGCCGGAGGTTCACGGCAATCTC; b3-(A–J)/b5, ACCTTTAAGAAA/GATTGCGTCGAGTGCCCGCTGCTC and CTCGACGCAATC/TTTCTTAAAGGTGCAGGCATCTGG; b3-(A–L)/b5, AAGTCCATCCTG/SCCGTCCTCAGGGAGCCAGAGTGT and CCTGAGGACGGT/CAGGATGGACTTTCCACTAGAATC. To facilitate the cloning of the chimera containing
b3 residues 164 –202, a silent EcoRV restriction site was introduced
into b5 by mutagenesis. The point mutation at residue 587 (underlined)
was made using the primer GTTGATAAGGATATCTCTCCTTTCTCC
and the Chameleon Double-Stranded, Site-Directed Mutagenesis Kit
(Stratagene). Hybrid primers were then used to generate the b3/b5
junctions by PCR as follows: GATATCTCT/CCATACATGTATATCTCC
and GGTGACCCGCTTC/AATGAGGAAGTTCGGAAACAGAGG. The
chimera b3-(C–D)/b5 was made by subcloning the amino-terminal b5
end into the cDNA encoding b3-(A–D)/b5. The chimera b3-(A–I)/b5 was
made by replacing the amino-terminal fragment of b5 with b3 at the
endogenous NcoI restriction site that is located at the end of exon I in
both b3 and b5. The nucleotide sequence of each chimera was confirmed
by dideoxy sequencing.
Cell Lines and FACS Analysis—Human embryonic kidney 293 cells
obtained from ATCC were maintained in Dulbecco’s modified Eagle’s
medium (Irvine Scientific) supplemented with 10% fetal calf serum
(Irvine Scientific). The 293 cells contain av and express no detectable
avb3. The 293 cells express trace levels of avb5 (28). Chimeric cDNAs
consisting of human b3 and b5 cDNA were cloned into the expression
vector pcDNA3 (Invitrogen) and transfected into 293 cells using N[1-(2,3-dioleoyloxy)propyl]-N,N,N-trimethylammonium methylsulfate
transfection reagent (Boehringer Mannheim). Following selection in
500 mg/ml geneticin (Life Technologies, Inc.) for 3 weeks, the cells were
selected by FACS to obtain the top 5% of the expressing cells. These
were chosen by sorting with antibodies against avb3 (P112-4C1) or
anti-avb5 (P1F6). Cells expanded from the sorted population were
continuously monitored by FACS analysis for high expression of transfected integrin throughout the duration of the study.
2
E. C. K. Lin, unpublished data.
23913
Cell Surface Binding Studies—Ligand binding to human 293 cells
expressing avb3, avb5, or chimeras between b3 and b5 was measured
with 125I-vitronectin using a tracer format as described previously (28).
Briefly, vitronectin was labeled to high specific activity (150,000 cpm/
ng) with 125INa and allowed to bind to cells held in suspension. Incubations were carried out in Binding Buffer (Hanks’ balanced salt solution lacking MgCl2, CaCl2, and MnCl2 (Life Technologies, Inc.), 50 mM
HEPES, pH 7.4, 3 mg/ml bovine serum albumin) containing 0.5 mM
MgCl2 and 0.02 mM CaCl2 at 14 °C for 70 min. A range of unlabeled
vitronectin was included with the cells to compete for binding to 0.5 nM
125
I-vitronectin. Specific binding was determined by subtracting EDTAsensitive binding from total binding. Scatchard analysis was used to
calculate the affinity of chimeric integrin for vitronectin (29).
The affinity of the chimeric integrins for Fab-9 was measured using
cell surface binding studies essentially as described (28). Briefly, cells
(3 3 106/ml) in suspension were incubated with increasing concentrations of 125I-Fab-9 at 14 °C for 70 min. Nonspecific binding was measured by including 5 mM EDTA in the reaction mixture. Previous studies
in this laboratory have shown that competition with an excess of unlabeled Fab-9 yields nearly identical values for nonspecific binding as
does inclusion of EDTA. Following incubation, free 125I-Fab-9 was separated from cell-bound ligand by centrifugation through 20% sucrose,
50 mM Tris-buffered saline, pH 7.4, at 14,000 rpm for 3 min in disposable microcentrifuge tubes (Fisher). The bottoms of the tubes containing the cell pellets were cut off and counted in a gamma counter. All
data represent the average of triplicate measurements. All assays were
repeated at least twice yielding nearly identical results. The affinity of
the chimeric integrin for Fab-9 was calculated by Scatchard analysis
(30).
Immunoprecipitations—Immunoprecipitations used to assess the expression of chimeras on the surface of kidney 293 cells were performed
as described previously except that 75 mM octylglucoside was used as
the sole detergent (31). A modified protocol was used to measure the
binding of Fab-9 to the chimeras. Cells expressing chimeric integrin
were labeled with 125I using the lactoperoxidase method (32). Cells were
lysed in Lysis Buffer (75 mM octylglucoside (Calbiochem), 25 mM Tris,
pH 7.7, 150 mM NaCl, 1 mM phenylmethylsulfonyl fluoride, 0.5 mM
MgCl2, and 0.05 mM MnCl2). Cell lysates were precleared with Pansorbin (Calbiochem) and then lipid was removed using Seroclear (Calbiochem). Lysates were immunoprecipitated with P112-4C1 (anti-avb3
mAb) for 18 h at 4 °C. Chimeric integrins were eluted from the P1124C1-linked Affi-Gel (Bio-Rad) with Elution Buffer (50 mM octylglucoside, 10 mM sodium acetate, pH 3.0). The eluted integrins were
immediately neutralized by addition of 3 M Tris, pH 8.8. This material
was dialyzed extensively against 50 mM octylglucoside, 100 mM NaCl,
25 mM Tris, pH 7.4, 0.5 mM MgCl2, and 0.05 mM MnCl2. The partially
purified integrins were then immunoprecipitated with a concentration
range of Fab-9 in Lysis Buffer containing 1 mg/ml bovine serum albumin. The complex between integrin and Fab-9 was precipitated by
addition of protein A-agarose (Sigma) and a saturating amount of goat
anti-human F(ab9)2 IgG. The protein bound to protein A-agarose was
washed in Lysis Buffer and then with Tris-buffered saline containing
0.5 mM MgCl2 and 0.05 mM MnCl2. Samples were analyzed on SDSPAGE under non-reducing conditions. Chimeric integrins precipitated
by Fab-9 were visualized by exposing the dried gel to a PhosphorImager
(Bio-Rad). Each chimeric b band was quantified as pixel density (PD)
volume which was calculated as the mean PD multiplied by the area
(mm2) of the band. The apparent affinity of Fab-9 for the chimeric
integrin was calculated as the concentration of Fab-9 that gave 50% of
the maximal PD units at a saturating concentration of Fab-9. Quantitative immunoprecipitations were always done in parallel with cells
expressing avb3.
Cell Adhesion Assay—The cell adhesion assay was conducted
essentially as described (33). Cells were harvested with 200 mM
EDTA/phosphate-buffered saline and resuspended in Hanks’ balanced salt solution (Sigma) supplemented with 25 mM MnCl2. 2 3 105
cells were added to each well containing immobilized fibrinogen (20
nM) or vitronectin (10 nM). Cells were allowed to adhere for 45 min at
37 °C. Nonadherent cells were removed by gentle washing, and the
adherent cells were quantitated using a colorimetric assay for lysosomal acid phosphatase (34).
RESULTS
Construction and Expression of Chimeras between b3 and
b5—To identify domains of the integrin b subunit involved in
ligand binding specificity, we constructed chimeras between
the b3 and b5 subunits. As some models suggest that domains
23914
Ligand Binding Specificity of avb3
of the b subunit near the plasma membrane are linked to
amino-terminal domains and contribute to ligand binding (35,
36), we constructed several chimeras that contained relatively
broad regions of b3 (Fig. 1). Most chimeras were constructed
based on the known exon junctions for b3 (27). For example,
b3-(A–L)/b5 is a chimera containing b3 exons A through L (the
entire extracellular domain) and the b5 transmembrane and
cytoplasmic domains. The chimera b3-(C–D)/b5 encompasses
most of the regions of b3 thought to be part of the ligand
binding site.
Expression of Chimeric Integrins on Kidney 293 Cells—The
cDNAs encoding each chimera were used to transfect human
embryonic kidney 293 cells. The stable expression of each chimera was monitored by FACS. For this analysis, we used
monoclonal antibodies (mAb) P112-4C1 (anti-avb3) and P1F6
(anti-avb5). These antibodies are specific for the complex between av and the respective b subunit. Both antibodies also
block the ligand binding function of integrin; hence, their ability to bind to the chimeras can be taken as evidence that the
domain swaps do not grossly alter the fold of the hybrid inte-
FIG. 1. The chimeric b3/b5 subunits. Chimeras between the b3
and b5 integrin subunits were designed based on the exon junctions of
b3 (27). The name of each chimera is derived from the exons or amino
acid residues of b3 that were substituted into b5. Regions corresponding
to the RGD cross-linking sites (bars) (38, 40) and other ligand contact
sites (asterisks) (42, 45) are noted above the figure.
grins. Human 293 cells transfected with the control expression
vector produced little endogenous avb3 or avb5 (Fig. 2A). Cells
transfected with cDNAs encoding wild-type b3 (Fig. 2B) and
wild-type b5 (Fig. 2E) expressed substantial levels of each
integrin. The chimera containing b3 exons A through I was
detected with antibody P112-4C1 (Fig. 2C). The chimera containing b3 exons A and B was detected with mAb P1F6 (Fig.
2D). Results from FACS analysis of the remaining chimeras are
summarized in Table I. Additional antibodies LM609 (antiavb3) and P3G2 (anti-avb5) reported identical expression levels as P112-4C1 and P1F6, respectively (Table I). The anti-b3
mAb RUU-PL7F12 did not bind to chimeras containing small
regions of b3 within exons A–D. Three other chimeras that
were constructed, b3-(A–E)/b5, b3-(C–E)/b5, and b3-(A–H)/b5,
were not expressed as a heterodimer with av on the cell surface,
even though reverse transcriptase-PCR analysis showed that
mRNA for these b subunits was produced by transfected cells.
Immunoprecipitation of Chimeric Integrins—Chimeras were
analyzed by immunoprecipitation to confirm their association
with av and to quickly assess their ability to bind to Fab-9, a
ligand specific for the b3-integrins. Cells were radiolabeled
with 125I and then precipitated with four different antibodies
(Fig. 3). These include a control IgG (lane 1), an antibody
against the cytoplasmic tail of b5 (lane 2), mAb L230 against av
(lane 3), and Fab-9 (lane 4). Human 293 cells transfected with
the pcDNA3 vector express little av in complex with b subunit
(A, lane 3). Cells transfected with cDNA encoding b3 (B), b5
(E), or a chimeric b subunit (C and D) expressed substantial
levels of an integrin complex on the cell surface. It is important
to note that although all of the chimeras contained variable
regions of b3 ectodomain, they all contained the transmembrane and cytoplasmic domains of b5. Thus, the expression of
the b3/b5 chimeras can be verified by precipitating with an
antibody against the cytoplasmic tail of b5 (lane 2, C and D).
Under the conditions used to lyse the cells, the complex
between the a and b subunit is not disrupted, so this assay also
allowed us to quickly screen the chimeras for the ability to bind
to Fab-9. Lane 4 of each panel in Fig. 3 shows an immunoprecipitation using Fab-9 at a concentration of 190 nM. This level
of Fab-9 is known to be 19-fold above the KD of Fab-9 for
wild-type avb3 (see Table I). As expected, Fab-9 precipitated
avb3 (B) but not avb5 (E). One chimera, b3-(A–B)/b5 (C) failed
to bind Fab-9, indicating that exons A and B of b3 are not
sufficient to confer ligand binding specificity. In contrast, the
chimeras containing b3 exons A–L, A–J, A–I, A–D (D), C–D,
and b3 residues 164 –202 were all precipitated by Fab-9, indicating that each has b3-like binding specificity (not all data
shown). The affinity of these chimeras for Fab-9 was subsequently studied in detail (see below).
The Chimera Containing Exons A and B of b3 Can Bind
Vitronectin but Not Fab-9 —One chimera, b3-(A-B)/b5, was recognized by the function-blocking antibodies P1F6 (Fig. 2, Table
I) and P3G2 (not shown), both of which are specific for the
FIG. 2. Expression of chimeric integrins on human 293 cells. Chimeric integrin subunits were transfected into human embryonic kidney
293 cells. Cells selected for stable expression of the transfected integrin were analyzed by flow cytometry. The antibodies P1F6 against avb5
(striped peak) and P112-4C1 against avb3 (shaded peak) were used in this analysis. Mouse IgG (clear peak) was used as a negative control. The
binding of each antibody to cells was detected with fluorescein isothiocyanate-conjugated goat anti-mouse mAb. Representative FACS profiles are
shown for cells transfected with the following: A, control expression vector; B, b3; C, b3-(A–I)/b5; D, b3-(A–B)/b5; or E, b5. Results from FACS
analysis of all chimeras are summarized in Table I.
Ligand Binding Specificity of avb3
23915
TABLE I
Expression of transfected integrin b subunits as measured
by FACS analysis
FACS Reactivity with Monoclonal Antibodiesa
Transfected
integrin
P1F6b
LM609
P112–4C1 RUU-PL7F12
(anti-avb5) (anti-avb3) (anti-avb3)
(anti-b3)
mean fluorescence intensity
Vector only
WT b3
b3(A-B)/b5
b3(164–202)/b5
b3(C-D)/b5
b3(A-D)/b5
b3(A-I)/b5
b3(A-J)/b5
b3(A-L)/b5
b5
1/2
2
111
2
2
2
2
2
2
111
2
11
2
11
1
111
11
11
11
2
2
11
2
11
1
111
11
11
11
2
2
11
2
2
2
2
11
11
11
2
a
FACS reactivity is quantitated by subtracting the mean fluorescence intensity of the nonspecific mouse IgG control from the mean
fluorescence intensity of the anti-integrin mAb: 2, 1–5 (not different
than nonspecific mouse IgG); 1/2, 5–10; 1, 11–30; 11, 31–50; 111,
greater than 50.
b
Monoclonal antibody P3G2 (anti-avb5) gave results identical to
mAb P1F6 (not shown).
complex between av and b5. b3-(A–B)/b5 was also found to lack
the ability to bind Fab-9 (Fig. 3), indicating that it does not
have the ligand binding specificity of avb3. To confirm that
b3-(A–B)/b5 had ligand binding function, we measured its ability to bind soluble vitronectin. We have previously shown that
soluble vitronectin has high affinity for avb5 but much less
affinity for avb3 (28). The b3-(A–B)/b5 chimera bound to 125Ivitronectin with an affinity only 3.2-fold lower than that of
wild-type avb5 (29 versus 9 nM, respectively) (Fig. 4). The
binding was almost completely abolished with antibody L230,
which blocks ligand binding to all av-integrins (37), and by
RGD peptide (not shown). Because the b3-(A–B)/b5 chimera
was capable of binding to vitronectin, but incapable of binding
to Fab-9, we conclude that exons A and B of the integrin b
subunit do not substantially contribute to ligand binding specificity. This conclusion is also supported by the inability of cells
expressing this chimera to adhere to fibrinogen (see below).
Measuring the Affinity between Fab-9 and the Chimeras of b3
and b5—All of the chimeras except b3-(A–B)/b5 indicated some
ability to bind to Fab-9 using the immunoprecipitation assay at
a saturating Fab-9 concentration (Fig. 3). As a next step, we
used a cell surface binding assay to measure the binding affinity of these chimeras for Fab-9. The binding of 125I-Fab-9 to
cells expressing the chimeras was measured at 14 °C to prevent
the internalization of ligand. The binding isotherms between
125
I-Fab-9 and cells expressing avb3 or b3-(A-I)/b5 are shown
in Fig. 5 as an example. The chimera containing exons A–I of
b3 has an affinity for Fab-9 that is equivalent to that of wildtype avb3. Chimeras containing larger domains of b3 (exons
A–J and A–L) also bound to Fab-9 with an affinity identical to
wild type avb3 (Table II). Based on these observations, it can be
concluded that a swap of exons A–I in the b3 subunit is sufficient to change the integrin’s ligand recognition specificity. As
previously reported, Fab-9 failed to specifically bind to cells
expressing wild-type avb5 (28) and also failed to bind the
b3-(A–B)/b5 chimera in this quantitative assay (data not
shown). In conjunction with the data shown in Figs. 3 and 4,
which indicate that exons A and B are not involved in ligand
selection, these findings suggest that the domain encompassing
b3 exons C–I is important for determining complete ligand
binding specificity.
Residues 164 –202 of b3 Are Sufficient to Confer Partial Ligand Binding Specificity—Because the cell surface binding
assay resulted in high nonspecific binding at 125I-Fab-9 con-
FIG. 3. Immunoprecipitation of chimeric integrins. Human 293
cells transfected with the pcDNA3 vector (A), b3 (B), b3-(A–B)/b5 (C),
b3-(A–D)/b5 (D), or b5 (E) were analyzed by immunoprecipitation. Cells
were labeled with 125I and lysed with detergent. The lysates were
incubated with goat anti-human F(ab9)2 antibody (lane 1), polyclonal
antibody against the b5 cytoplasmic tail (lane 2), mAb L230 which binds
to av (lane 3), or Fab-9 and goat anti-human F(ab9)2 (lane 4). Immune
complexes were precipitated with protein A-agarose, washed, and analyzed by SDS-PAGE and autoradiography. Each panel is a representative of at least two such immunoprecipitations.
centrations above 30 nM, the affinity of Fab-9 for chimeras
exhibiting a low affinity for Fab-9 could not be determined
reliably with this assay. Consequently, a quantitative immunoprecipitation assay was used to measure the affinity of the
chimeras b3-(A–D)/b5, b3-(C–D)/b5, and b3-(164 –202)/b5 for
Fab-9. Cells expressing each chimera were surface labeled with
125
INa and then lysed in octylglucoside. Each chimera was
immunoprecipitated from the lysate with antibody P112-4C1
(against avb3), an antibody known to recognize each chimera
in FACS (Table I). Chimeric integrins were then eluted from
this antibody and re-precipitated with a concentration range of
Fab-9. Immunoprecipitated proteins were resolved by SDSPAGE, and the amount of each chimera bound to Fab-9 was
quantified using a PhosphorImager (Fig. 6A). To obtain a binding isotherm, the pixel density volume of the band corresponding to the b subunit was plotted against the concentration of
Fab-9 used in the immunoprecipitation. Plots of pixel density
volumes obtained from the a subunit gave nearly identical
results. Examples of binding isotherms from two separate experiments are shown in Fig. 6B. In this assay, the affinity
between avb3 and Fab-9 is 0.5 nM. This value is 10-fold higher
than that measured in the cell surface binding assay (Fig. 5)
but nearly equivalent to the value we have previously reported
for avb3 purified from placenta (20). Both the purified integrin
(38) and the integrin examined in the quantitative immunoprecipitation assay are subject to elution from antibody with
low pH. Therefore, we suspect that treatment with low pH
enhances the affinity of the integrin for ligand. Nevertheless, in
this study the affinity of the chimeric integrins was always
compared with avb3 measured in the same assay (Table II), so
the relative affinities are directly comparable.
23916
Ligand Binding Specificity of avb3
FIG. 4. The b3-(A–B)/b5 integrin binds to vitronectin. Human 293 cells expressing avb5 (A) or the b3-(A–B)/b5 chimera (B) were analyzed
for their ability to bind to soluble 125I-vitronectin. Cells were incubated with 0.5 nM 125I-vitronectin in suspension at 14 °C with increasing amounts
of unlabeled vitronectin (top panels). The mAb L230 was used in these binding studies to confirm that binding was mediated by av-integrin.
Antibody P3G2 is a function-blocking mAb against the avb5 complex and was applied as an antagonist to verify that binding was mediated by avb5
or b3(A-B)/b5. The binding isotherms were converted to Scatchard plots (bottom panels) to obtain binding affinity. Each binding assay was
performed at least twice yielding identical results.
The affinity of Fab-9 for b3-(A–D)/b5 and b3-(C–D)/b5 was
approximately 10-fold lower than for avb3 (Table II). Thus,
these domains contain nearly all of the binding contacts needed
to obtain specificity for Fab-9. The chimera containing b3 residues 164 –202, a region chosen for a swap because it is contained within exons C–D and was predicted to be a highly
flexible loop (28), also bound to Fab-9. The affinity of b3-(164 –
202)/b5 for Fab-9 was 35-fold lower than that exhibited by avb3
but still significantly higher than that of avb5, which exhibits
at least 1000-fold lower affinity for Fab-9. The results of all
binding studies with Fab-9 are summarized in Table II.
Measuring the Ability of b3/b5 Chimeras to Mediate Cell
Adhesion to Fibrinogen—Cell adhesion assays were conducted
to determine which chimeras could mediate adhesion to fibrinogen, a natural adhesive ligand that will bind avb3 but not
avb5 (17). In general, the ability of the chimeras to mediate
adhesion to fibrinogen was similar to their ability to bind to
Fab-9. As shown in Fig. 7, cells expressing avb3, b3-(A–I)/b5,
and b3-(C–D)/b5 could all adhere to immobilized fibrinogen. As
predicted from the binding studies with Fab-9, cells expressing
avb5 or b3-(A–B)/b5 did not adhere to fibrinogen. Interestingly, cells expressing b3-(164 –202)/b5 also failed to adhere to
fibrinogen, even though this chimera could bind to Fab-9. All
cell lines were able to adhere to immobilized vitronectin (Fig.
7), indicating that the chimeric integrins were functional in
adhesive capability. Adhesion to both fibrinogen and vitronec-
tin was inhibited by inclusion of RGD peptide in the assay
(data not shown).
DISCUSSION
The goal of this study was to identify the domains within the
integrin b subunit that have an impact on ligand binding
specificity. This was achieved by examining the ligand binding
phenotype of chimeras between the integrin b3 and b5 subunits. The major findings of this study are as follows: 1) the
domain of the integrin b3 subunit involved in determining
complete ligand binding specificity encompasses residues 95–
537 (exons C-I); 2) chimeras containing smaller domains of b3
exhibited reduced affinity for b3-specific ligands (b3 5 b3-(A–
I)/b5 . b3-(C–D)/b5 . b3-(164 –202)/b5). This stepwise loss in
binding affinity suggests that optimal ligand binding specificity
is achieved through the contact of ligand with several discontinuous domains of the b subunit; 3) the replacement of a small
region of the b5 subunit with the homologous region of b3
(residues 164 –202) is sufficient to alter the ligand binding
phenotype of the integrin. This small domain is highly divergent among the integrin b subunits, suggesting that it is likely
to be a key structural element responsible for ligand selection
by all integrins.
The results of this study should be interpreted in the context
of the current understanding of how integrins bind their ligands. Several lines of evidence show that the integrin ligand
Ligand Binding Specificity of avb3
23917
FIG. 5. A comparison of the binding of Fab-9 to avb3 and the b3-(A–I)/b5 chimera. The binding of 125I-Fab-9 to 293 cells transfected with
b3 (A) or the b3-(A–I)/b5 chimera (B) was measured as described under “Experimental Procedures.” Representative binding isotherms are shown
(top graphs). Nonspecific binding (‚) was determined by inclusion of EDTA in the binding reaction. Specific (●) binding was calculated by
subtracting nonspecific binding from total (M) binding. These data were transformed to Scatchard plots (bottom graphs). The b3-(A–J)/b5 and
b3-(A–L)/b5 chimeras had similar binding affinities for Fab-9 (see Table II). Each binding assay was conducted three times, with similar results.
TABLE II
Binding affinities of chimeric integrins for Fab-9
Transfected
integrin
Affinity for Fab-9
Whole cell binding
Immunoprecipitation
nM 6 S.D. (n)
vector only
WT b3
b3(A-B)/b5
b3(164–202)/b5
b3(C-D)/b5
b3(A-D)/b5
b3(A-I)/b5
b3(A-J)/b5
b3(A-L)/b5
b5
a
b
c
NDa
10 6 2 (3)
NMb
NM
NM
NM
4 6 1 (3)
9 6 3 (3)
8 6 2 (3)
ND
ND
0.5 6 0.2 (4)
ND
17.7 6 3.6 (5)
6.2 6 0.8 (2)
4.3 6 0.7 (2)
NPc
NP
NP
ND
ND, not detectable.
NM, not measurable.
NP, not performed
binding pocket is comprised of amino acid residues on both the
a and b subunits (39, 40). Cross-linking studies localize the
RGD ligand binding site on the b3 subunit to residues 61–203
in the context of avb3 (38) and residues 109 –171 in the context
of aIIbb3 (41). Several other studies show that peptides derived
from these regions of b3 exhibit some ability to bind ligand
(41– 45). Collectively, these studies map the general location of
the ligand binding site, but they fail to address the basis for
ligand binding specificity. Here, we demonstrate that a small
domain consisting of b3 residues 164 –202 is involved in ligand
selection. We have previously analyzed the structure of this
region with the PHD algorithm, and it is predicted to be a
flexible loop (28). An alignment of this region among known
human integrin b subunits is shown in Fig. 8. The considerable
homology among all integrin b subunits of the areas flanking
b3 residues 164 –202 has implications for the structure of the
ligand binding pocket. First, the architecture of the site is
probably defined by the stretches of amino acid residues that
are highly homologous among all b subunits. Second, the binding pocket appears to contain a metal binding site comprised of
conserved liganding residues from noncontiguous domains (28,
46 – 48). Finally, the ligand binding site of each integrin b
subunit contains a domain corresponding to b3 residues 164 –
202 which is highly divergent and which normally encompasses
a disulfide-bonded loop of variable length. These data suggest
that this domain may have a substantial impact on the ligand
selection of all integrins. The association of such a divergent,
flexible structure with integrin ligand binding specificity is
reminiscent of other proteins in which highly variable loops
determine ligand binding preference. For example, the complementarity determining regions of many IgG proteins, including
antibodies and the T cell receptor, are the major determinants
of immunologic recognition (49, 50).
The findings presented here suggest that the ligand binding
site of the b3-(164 –202)/b5 chimera is quite similar to that of
23918
Ligand Binding Specificity of avb3
FIG. 6. Measuring the relative binding affinity of Fab-9 for chimeric integrins by quantitative immunoprecipitation. Human 293
cells expressing chimeric b subunits were labeled with 125I, lysed, and immunoprecipitated with mAb P112-4C1 (anti-avb3) linked to Affi-Gel
beads. The precipitated integrin was eluted with low pH, and the eluate was rapidly neutralized. This eluate was then immunoprecipitated with
a concentration range of Fab-9. The amount of each chimera bound to Fab-9 was quantified by analyzing the precipitate on SDS-PAGE and then
exposing the gel to a PhosphorImager plate (A). The pixel density (PD) volume of the b subunit (lower band) was calculated by multiplying the
mean PD by the area of the band (mm2). Representative images of avb3 (top), b3-(C–D)/b5 (middle), and b3-(164 –202)/b5 (bottom) are shown. The
binding isotherms (B) derived from scanned gels for two separate experiments are shown for avb3 (graphs 1 and 2), b3-(C–D)/b5 (graphs 3 and 4),
and b3-(164 –202)/b5 (graphs 5 and 6). The immunoprecipitation of b3-(A–D)/b5 with Fab-9 yielded results similar to b3-(C–D)/b5. Each assay was
repeated two to five times with comparable results.
b3 and unlike that of b5. How closely does the structure of
b3-(164 –202)/b5 mimic the structure of b3? This question can
be addressed by converting the binding constants between
Fab-9 and the integrins to theoretical free energy (DG0). Using
the affinity measurements shown in Table II and the equation
(DG0 5 21364 log(1/KD)) (51), the difference in DG0 for Fab-9
binding to avb3 versus the b3-(164 –202)/b5 integrin is only
about 22 kcal/mol. The reported energies of a single hydrogen
bond, hydrophobic interaction, or salt bridge are all between
21 and 24 kcal/mol (52). Therefore, the difference in the way
that avb3 and the b3-(164 –202)/b5 integrin bind to Fab-9 can
be theoretically equated to the loss of only one or two bonds.
Interestingly, the b3-(164 –202)/b5 integrin bound to soluble
Fab-9 but did not mediate cell adhesion to fibrinogen. Since the
affinity of this chimera for Fab-9 is 35-fold lower than avb3, the
simplest interpretation of these data is that the chimera simply
lacks sufficient affinity for fibrinogen to mediate cell adhesion.
However, these findings could also be interpreted differently
because there are important distinctions in the way that Fab-9
and fibrinogen bind to avb3. Fab-9 contains an RGD in complementarity determining region 3 that has been optimized to
bind the b3-integrins. This 9-residue loop is presumably the
only contact within the antibody for integrin (53). Fab-9 also
binds to integrins in a reversible manner (26). Although fibrin-
FIG. 7. Adhesion of cells expressing chimeric integrins to fibrinogen. The ability of 293 cells transfected with the indicated b
subunit to adhere to fibrinogen (dark) and vitronectin (striped) was
measured. Cells (2 3 105) were allowed to adhere to immobilized protein for 45 min at 37 °C. Adherent cells were quantified with a colorimetric assay for alkaline phosphatase (34). This experiment is representative of two similar repetitions.
Ligand Binding Specificity of avb3
23919
ing specificity (25), the present study indicates that the minimal integrin domains necessary for achieving ligand binding
specificity are aIIb residues 1–334 and b3 residues 164 –202.
The feasibility of expressing truncated integrins was recently
shown by McKay and colleagues (55), who were able to express
an integrin heterodimer consisting of aIIb residues 1–233 and
b3 residues 111–318. Such truncated integrins may be ideal for
structural analysis as they are likely to reveal the conformational differences that account for ligand selection.
Acknowledgment—We thank Dr. Erkki Ruoslahti for critical review
of the manuscript.
REFERENCES
FIG. 8. Comparison of the amino acid sequence of b3 residues
164 –202 with other integrin b subunits. The amino acid sequences
of human integrin b subunits b1–b8 corresponding to b3 residues
114 –252 were aligned using ClustalW (56). This domain contains 50
amino acids before and after the region of b3 164 –202. Amino acid
residues that are identical among all 8 subunits are shaded. Positions
containing conservative substitutions are boxed. The dashed lines above
the sequences show the position of b3 164 –202. Asterisks mark the
residues that have been identified as important for ligand and metal ion
binding function of several integrins (28, 46 – 48). Cysteines 177 and 184
in the b3 subunit are known to form a disulfide bond (36).
ogen also binds to avb3 through an RGD motif, it binds irreversibly (54). This irreversible binding is probably due to the
formation of additional contacts between integrin and ligand.
Therefore, it is possible that while the region encompassing b3
residues 164 –202 is capable of conferring binding to specific
presentations of the RGD motif (such as that reported by Fab9), it is not sufficient to make additional contacts essential for
stable binding of fibrinogen. Such a difference may account for
the inability of this chimera to promote cell adhesion. Importantly, the chimera that contained exons C and D, a somewhat
larger region of b3, did support adhesion to fibrinogen. Therefore, all of the structural information required for fibrinogen
binding is contained within these two exons.
Our study also provides information on the binding sites of
four monoclonal antibodies that block integrin function. Antibodies P112-4C1 and LM609 are specific for the avb3 complex
and block cell adhesion mediated by this integrin. Antibodies
P1F6 and P3G2 bind the complex between avb5 and inhibit
avb5-mediated cell adhesion. The smallest chimera examined,
b3-(164 –202)/b5, can bind to the avb3-specific antibodies and
has lost the ability to bind the avb5-specific antibodies. This
finding suggests that the domain corresponding to b3 residues
164 –202 may contain the epitope for many integrin complexspecific, function-blocking antibodies. Alternatively, this domain may contact the av subunit and regulate the conformation of the epitope by modifying av.
One current goal is to obtain the three-dimensional structure
of the integrin ligand binding site. Unfortunately, this is not
likely to be technically possible using full-length integrin subunits; it will have to be achieved with truncated forms of each
subunit. Ideally a truncated integrin would contain all of the
structural information required to confer ligand selection. In
conjunction with our prior report showing that the first two
cation binding sites of the aIIb subunit are necessary (in addition to the amino-terminal domain) for conferring ligand bind-
1. Albelda, S. M., and Buck, C. A. (1990) FASEB J. 4, 2868 –2880
2. Duband, J. L., Rocher, S., Chen, W. T., Yamada, K. M., and Thiery, J. P. (1986)
J. Cell Biol. 102, 160 –178
3. Brooks, P. C. (1996) Cancer Metastasis Rev. 15, 187–194
4. Grinnell, F. (1992) J. Cell Sci. 101, 1–5
5. Horton, M. A., and Davies, J. (1989) J. Bone Miner. Res. 4, 803– 808
6. Ruoslahti, E., and Giancotti, F. G. (1989) Cancer Cells 1, 119 –126
7. Phillips, D. R., Charo, I. F., and Scarborough, R. M. (1991) Cell 65, 359 –362
8. Engleman, V. W., Nickols, G. A., Ross, P. F., Horton, M. A., Griggs, D. W.,
Settle, S. L., Ruminski, P. G., and Teitelbaum, S. L. (1997) J. Clin. Invest.
99, 2284 –2292
9. Ruoslahti, E. (1991) J. Clin. Invest. 87, 1–5
10. Hynes, R. O. (1992) Cell 69, 11–25
11. Ruoslahti, E. (1996) Annu. Rev. Cell Dev. Biol. 12, 697–715
12. Smith, J. W. (1994) in Integrins, (Cheresh, D. A., and Mecham, R. P., eds)
pp. 1–32, Academic Press, New York
13. Bowditch, R. D., Hariharan, M., Tominna, E. F., Smith, J. W., Yamada, K. M.,
Getzoff, E. D., and Ginsberg, M. H. (1994) J. Biol. Chem. 269, 10856 –10863
14. Obara, M., Kang, M. S., and Yamada, K. M. (1988) Cell 53, 649 – 657
15. Ramaswamy, H., and Hemler, M. E. (1990) EMBO J. 9, 1561–1568
16. McLean, J. W., Vestal, D. J., Cheresh, D. A., and Bodary, S. C. (1990) J. Biol.
Chem. 265, 17126 –17131
17. Smith, J. W., Vestal, D. J., Irwin, S. V., Burke, T. A., and Cheresh, D. A. (1990)
J. Biol. Chem. 265, 11008 –11013
18. Klemke, R. L., Yebra, M., Bayna, E. M., and Cheresh, D. A. (1994) J. Cell Biol.
127, 859 – 866
19. Filardo, E. J., Deming, S. L., and Cheresh, D. A. (1996) J. Cell Sci. 109,
1615–1622
20. Barbas, C. F., Languino, L., and Smith, J. W. (1993) Proc. Natl. Acad. Sci.
U. S. A. 90, 10003–10007
21. Liaw, L., Lindner, V., Schwartz, S. M., Chambers, A. F., and Giachelli, C. M.
(1995) Circ. Res. 77, 665– 672
22. Kramer, R. H., Cheng, Y.-F., and Clyman, R. (1990) J. Cell Biol. 111,
1233–1243
23. Cheresh, D. A., Berliner, S. A., Vicente, V., and Ruggeri, Z. M. (1989) Cell 58,
945–953
24. Smith, J. W., Ruggeri, Z. M., Kunicki, T. J., and Cheresh, D. A. (1990) J. Biol.
Chem. 265, 12267–12271
25. Loftus, J. C., Halloran, C. E., Ginsberg, M. H., Feigen, L. P., Zablocki, J. A.,
and Smith, J. W. (1996) J. Biol. Chem. 271, 2033–2039
26. Smith, J. W., Hu, D., Satterthwait, A., Pinz-Sweeny, S., and Barbas, C. F., III
(1994) J. Biol. Chem. 269, 32788 –32795
27. Lanza, F., Kieffer, N., Phillips, D. R., and Fitzgerald, L. A. (1990) J. Biol.
Chem. 265, 18098 –18103
28. Lin, E. C. K., Ratnikov, B. I., Tsai, P. M., Gonzalez, E. R., Pelletier, A. J., and
Smith, J. W. (1997) J. Biol. Chem. 272, 14236 –142243
29. Rechler, M. M., Podskalny, J. M., and Nissley, S. P. (1977) J. Biol. Chem. 252,
3898 –3910
30. Scatchard, G. (1949) Ann. N. Y. Acad. Sci. 51, 660 – 672
31. Cheresh, D. A., and Spiro, R. C. (1987) J. Biol. Chem. 262, 17703–17711
32. Hudson, L. and Hay, F. C. (1980) Practical Immunology, pp. 243–244, Blackwell Scientific Publications, London
33. Smith, J. W., Piotrowicz, R. S., and Mathis, D. M. (1994) J. Biol. Chem. 269,
960 –967
34. Pratner, C. A., Plotkin, J., Jaye, D., and Frazier, W. A. (1991) J. Cell Biol. 112,
1031–1040
35. Calvete, J. J. (1995) Proc. Soc. Exp. Biol. Med. 208, 346 –360
36. Calvete, J. J., Henschen, A., and Gonzalez-Rodriguez, J. (1991) Biochem. J.
274, 63–71
37. Weinacker, A., Chen, A., Agrez, M., Cone, R. I., Nishimura, S., Wayner, E.,
Pytela, R., and Sheppard, D. (1994) J. Biol. Chem. 269, 6940 – 6948
38. Smith, J. W., and Cheresh, D. A. (1988) J. Biol. Chem. 263, 18726 –18731
39. Smith, J. W., and Cheresh, D. A. (1990) J. Biol. Chem. 265, 2168 –2172
40. D’Souza, S. E., Ginsberg, M. H., Lam, S. C.-T., and Plow, E. F. (1988) J. Biol.
Chem. 263, 3943–3951
41. D’Souza, S. E., Ginsberg, M. H., Burke, T. A., Lam, S. C., and Plow, E. F. (1988)
Science 242, 91–93
42. D’Souza, S. E., Haas, T. A., Piotrowicz, R. S., Byers-Ward, V., McGrath, D. E.,
Soule, H. R., Ciernieswki, C., Plow, E. F., and Smith, J. W. (1994) Cell 79,
659 – 667
43. Pasqualini, R., Koivunen, E., and Ruoslahti, E. (1995) J. Cell Biol. 130,
1189 –1196
44. Alemany, M., Concord, E., Garin, J., Vincon, M., Giles, A., Marguerie, G., and
Gulino, D. (1996) Blood 87, 592– 601
45. Charo, I. F., Nannizzi, L., Phillips, D. R., Hsu, M. A., and Scarborough, R. M.
23920
Ligand Binding Specificity of avb3
(1991) J. Biol. Chem. 266, 1415–1421
46. Tozer, E. C., Liddington, R. C., Sutcliffe, M. J., Smeeton, A. H., and Loftus,
J. C. (1996) J. Biol. Chem. 271, 21978 –21984
47. Puzon-McLaughlin, W., and Takada, Y. (1996) J. Biol. Chem. 271,
20438 –20443
48. Goodman, T. G., and Bajt, M. L. (1996) J. Biol. Chem. 271, 23729 –23736
49. Mariuzza, R. A., Phillips, S. E., and Poljak, R. J. (1987) Annu. Rev. Biophys.
Biophys. Chem. 16, 139 –159
50. Claverie, J. M., Prochnicka-Chalufour, A., and Bougueleret, L. (1989)
Immunol. Today 10, 10 –14
51. Segel, I. H. (1976) Biochemical Calculations, 2nd Ed., p. 152, John Wiley &
Sons, Inc., New York
52. Schulz, G. E., and Schirmer, R. H. (1979) in Principles of Protein Structure
(Cantor, C. R., ed) p. 28, Springer-Verlag, Inc., New York
53. Barbas, C. F., Languino, L. R., and Smith, J. W. (1993) Proc. Natl. Acad. Sci.
U. S. A. 90, 10003–10007
54. Marguerie, G. A., Edgington, T. S., and Plow, E. F. (1980) J. Biol. Chem. 255,
154 –161
55. McKay, B. S., Annis, D. S., Honda, S., Christie, D., and Kunicki, T. J. (1996)
J. Biol. Chem. 271, 30544 –30547
56. Thompson, J. D., Higgins, D. G., and Gibson, T. J. (1994) Nucleic Acids Res. 22,
4673– 4680