Changing Ligand Specificities of αvβ1 and αvβ3 Integrins by Swapping a Short Diverse Sequence of the β Subunit*

Integrins mediate signal transduction through interaction with multiple cellular or extracellular matrix ligands. Integrin αvβ3 recognizes fibrinogen, von Willebrand factor, and vitronectin, while αvβ1 does not. We studied the mechanisms for defining ligand specificity of these integrins by swapping the highly diverse sequences in the I domain-like structure of the β1 and β3 subunits. When the sequence CTSEQNC (residues 187–193) of β1 is replaced with the corresponding CYDMKTTC sequence of β3, the ligand specificity of αvβ1 is altered. The mutant (αvβ1–3-1), like αvβ3, recognizes fibrinogen, von Willebrand factor, and vitronectin (a gain-of-function effect). The αvβ1–3-1 mutant is recruited to focal contacts on fibrinogen and vitronectin, suggesting that the mutant transduces intracellular signals on adhesion. The reciprocal β3–1-3 mutation blocks binding of αvβ3 to these multiple ligands and to LM609, a function-blocking anti-αvβ3 antibody. These results suggest that the highly divergent sequence is a key determinant of integrin ligand specificity. Also, the data support a recent hypothetical model of the I domain of β, in which the sequence is located in the ligand binding site.

Integrins are a family of ␣/␤ heterodimers of cell adhesion receptors that mediate cell-extracellular matrix and cell-cell interactions (1)(2)(3)(4)(5). Integrin-ligand interactions are critically involved in the pathogenesis of many diseases in human and animal models. Although integrin-ligand interaction is a therapeutic target, we poorly understand at the molecular level how integrins recognize multiple ligands. Evidence suggests that the I or A domain, a set of inserted sequences consisting of about 200 amino acid residues, of several integrin ␣ subunits (␣M, ␣L, ␣1, ␣2) is important in ligand binding and receptor activation (reviewed in Ref. 6 and references therein). The presence of an I domain-like structure within the ␤ subunit has been suggested based on the similarity in hydropathy profiles between the I domain and part of the ␤ subunit (7). Interestingly, this region of ␤ has been reported to be critical for ligand binding and its regulation (reviewed in Ref. 8) (Fig. 1). The Asp-119 (␤3) (9) and Asp-130 (␤1) (10,11) and the correspond-ing residues in ␤2 and ␤6 are critical for ligand binding (12,13). A synthetic peptide of ␤3 (MDLSYSMKDDLWSI, residues 118 -131) has been shown to produce a ternary complex with cations and ligand (14). Also, the sequence DDLW (residues 126 -129 of ␤3) was shown to be critical for interaction with the RGD sequence using a phage display system (15). A synthetic peptide of ␤3, DAPEGGFDAIMQATV (residues 217-231 of ␤3), has been shown to bind to immobilized fibrinogen (Fg), 1 von Willebrand's factor (vWf), and fibronectin (Fn) (16,17). A synthetic peptide of ␤3, SVSRNRDAPEG (residues 211-221 of ␤3), has been reported to block binding of Fg to ␣IIb␤3 (18,19). We identified a small region of ␤1 (residues 207-218, a regulatory epitope) that is recognized by both activating and inhibiting anti-␤1 antibodies (20). These antibodies probably induce high or low affinity states, respectively, by changing the conformation of the ␤1 subunit through binding to the non-ligand binding site (20).
We and researchers at other laboratories have recently identified residues critical for ligand binding in the putative I domain-like structure of ␤1 (6), ␤2 (21), and ␤3 (22). In ␤1, eight critical oxygenated residues are located in several separate predicted loop structures, which probably constitute multiple ligand/cation binding sites within the I domain-like structure of the ␤ subunit. These critical oxygenic residues are conserved among integrin ␤ subunits, indicating that these residues are ubiquitously involved in ligand binding regardless of ligand and integrin species. We observed that a large predicted loop region (residues 176 -199 of ␤1) is diverse among the ␤ subunits ( Fig. 1). Furthermore, a recent structural model (23) and our preliminary model (not shown) of the I domain-like structure of ␤ suggest that the sequence is also on the same side of the domain as residues critical for ligand binding. We hypothesized that the predicted loop (especially the disulfide-linked short sequences, e.g. residues 187-193 of ␤1) is involved in ligand specificity of integrins. ␣v␤3 has been shown to recognize a wide variety of ligands, including Fn, Fg, vWf, and vitronectin (Vn); ␣v␤1 is specific to Fn. We designed experiments, using ␣v␤1 and ␣v␤3 integrins, to determine whether a diverse sequence in the predicted loop (e.g. residues 176 -199 in ␤1, residues 166 -190 in ␤3) is involved in ligand specificity of integrins.
Transfection of Mammalian Cells-Human ␣v and ␤3 cDNAs were provided by J. Loftus (Scripps). Ten g of wild-type human ␣v cDNA in pBJ-1 vector (33,34) was transfected into parental CHO-K1 cells (8 ϫ 10 6 cells) together with 1 g of pFneo plasmid containing a neomycinresistant gene by electroporation as described (35). After they were selected for G418 resistance, cells expressing ␣v were cloned by cell sorting in FACStar cell sorter (Becton-Dickinson) with mAb LM142 (the cloned cells are designated ␣v-CHO cells). Human ␤1 or ␤3 (WT/mutant) cDNA in pBJ-1 vector was transfected into ␣v-CHO cells together with 1 g of pCD-hygro plasmid with a hygromycin-resistant gene or into parent CHO cells together with 1 g of pFneo; cells were then selected with hygromycin (500 g/ml; Calbiochem) or G418 essentially as described above. Cells expressing human ␤1 or ␤3 were cloned by sorting with mAb A1A5 or 15 as described above. The flow cytometric analysis was carried out using FACScan (Becton-Dickinson).
Adhesion Assays-Wells of 96-well Immulon-2 microtiter plates (Dynatech Laboratories, Chantilly, VA) were coated with 100 l of PBS (10 mM phosphate buffer, 0.15 M NaCl, pH 7.4) containing Fg, vWf, Fn, and Vn at a concentration of 10 g/ml overnight at 4°C. The remaining protein binding sites were blocked by incubating with 1% bovine serum albumin (Calbiochem) for 1 h at room temperature. Cells (10 5 cells/well) in 100 l of Dulbecco's modified Eagle's medium containing 0.5 mg/ml bovine serum albumin were added to the wells and incubated at 37°C for 1 h. After gently rinsing the wells three times with PBS to remove unbound cells, bound cells were quantified using endogenous phosphatase activity (36).
Affinity Chromatography-Cells were harvested with 3.5 mM EDTA in PBS and washed with PBS. Cells (about 5 ϫ 10 6 ) were then surfacelabeled with 125 I by using IODO-GEN (Pierce) (37), washed three times with PBS, and solubilized in 1 ml of 100 mM octyl glucoside in 10 mM Tris-HCl, 0.15 M NaCl, pH 7.4 (TBS), containing 2.5 mM MnCl 2 , 1 mM phenylmethylsulfonyl fluoride (Sigma) at 4°C for 15 min. The insoluble materials were removed by centrifugation at 15,000 ϫ g for 10 min. The supernatant was then incubated with a small amount of underivatized Sepharose 4B at 4°C for 15 min to remove nonspecific binding material. The supernatant was incubated at 4°C for 1 h with 200 -500 l of packed Fg-, Vn-, Fn 110-kDa fragment-, or GRGDS-Sepharose that had been equilibrated with TBS containing 2.5 mM MnCl 2 , 1 mM phenylmethylsulfonyl fluoride, 25 mM octyl glucoside (washing buffer). The unbound materials were washed with a 20 ϫ column volume of washing buffer, and the bound materials were eluted with 20 mM EDTA instead of 1 mM MnCl 2 in washing buffer; and then 0.5-ml fractions were collected. Twenty-l aliquots from each fraction was analyzed by SDSpolyacrylamide gel electrophoresis using 7% polyacrylamide gel followed by autoradiography.
Immunostaining-Glass coverslips (Fisher) were treated with 10% KOH in methanol for 1 h at room temperature, washed three times with distilled H 2 O, and stored in ethanol. Etched coverslips were then coated with 50 g/ml Fg, 50 g/ml Fn, or 22 g/ml Vn in PBS overnight at 4°C and then blocked with 10 mg/ml heat-denatured bovine serum albumin (Calbiochem) in PBS for 10 min at room temperature. For plating experiments, cells were washed and then detached with 2.5 mM EDTA/ PBS. Detached cells were isolated, washed, resuspended in Dulbecco's modified Eagle's medium, and then replated on coated coverslips. Cells were allowed to attach and spread for 2 h. Prior to fixation, cells were chilled on ice for 5 min, washed with cold PBS, and then extracted with cold PIPES buffer (0.1 M PIPES, pH 6.8, 1 mM MgCl 2 , and 1 mM EGTA) containing 1% glycerol and 0.5% Nonidet P-40 for 1-2 min. Extracted cells were washed with cold PIPES buffer and then fixed with 3.7% methanol-free formaldehyde (Polysciences) in PIPES buffer for 20 min at room temperature. Following fixation, cells were washed with PBS and then blocked with 10% normal goat serum (Life Technologies, Inc.)/PBS for 20 min at 37°C. Human integrins were detected using FIG. 1. The diverse sequence within the putative I domain-like structure of the ␤ subunit. This region of the ␤ subunit contains eight predicted ␤strands and five predicted ␣-helices (6). Eight critical oxygenated residues (Asp-130, Ser-132, Asn-224, Asp-226, Glu-229, Asp-233, Asp-267, and Asp-295 in ␤1) are located in several separate predicted loop structures (or at the boundary between a predicted ␣-helix and a ␤-strand), which probably constitute multiple ligand/cation binding sites within the I domain-like structure of the ␤ subunit. Although most of the sequences of the predicted turn structures containing the critical oxygenated residues are conserved among integrin ␤ subunits, a large predicted loop region (residues 176 -199) is not conserved among the ␤ subunits. Particularly, the sequence and size of the disulfide-linked short sequence (e.g. residues 187-193 in ␤1, boxed area) is diverse. We hypothesized that the predicted loop region of the ␤ subunit is involved in ligand binding specificity.
either the anti-human ␤1 antibody P5D2 or the anti-human ␤3 antibody 15. Cells were immunostained for 1 h at 37°C, washed, and then stained with a fluorescein isothiocyanate-conjugated sheep anti-mouse IgG secondary antibody (Molecular Probes) for 30 min at 37°C; cells were also labeled with rhodamine phalloidin (Molecular Probes) to detect actin stress fibers. Stained cells were mounted in Fluoromount-G (Fisher) and photographed using a Nikon Diaphot inverted microscope.
Other Methods-Site-directed mutagenesis of the ␤1 and ␤3 cDNA in a pBJ-1 vector was carried out using unique restriction site elimination (38). The presence of mutations was confirmed by DNA sequencing. Immunoprecipitation was carried out as described previously (20).
The ␣v␤1-3-1 Mutant Is Recruited to Focal Contacts and Transduces Signals on Adhesion to Fg and Vn-Next we determined if the altered ligand specificity of the ␣v␤1-3-1 chi-mera affected intracellular signaling. Cells were plated on Fn, Vn, or Fg, and localization of the human integrin was determined by immunostaining with anti-human ␤1 (␣v␤1 and ␣v␤1-3-1) or anti-human ␤3 (␣v␤3). While all three receptors localized to focal adhesions in cells plated on Fn (Fig. 4, A, C, and F), only ␤1-3-1 and ␤3 localized to focal adhesions in cells on Vn; ␣v␤1-CHO cells did attach and spread on Vn due to endogenous ␣v␤5. However, ␣v␤1 exhibited a diffuse staining pattern. This result is consistent with the binding data and indicates that the ␣v␤1-3-1 chimera is able to generate intracellular signals. In addition, we found that the ␣v␤1-3-1 chimera, like ␣v␤3, induced cell spreading and focal adhesion formation in cells plated on Fg; ␣v␤1 cells did not adhere to Fg. Similar results were obtained with the ␤1-, ␤1-3-1-, and ␤3-CHO cells that express lower levels of the transfected integrins (data not shown). These results indicate that the ␣v␤1-3-1 chimera is a functional receptor and has the same signaling properties as ␣v␤3.
The immunoprecipitation of whole lysate of ␣v␤3-1-3-CHO cells using anti-␣v and anti-␤3 mAbs showed that anti-␣v and anti-␤3 co-precipitated ␤3-1-3 and ␣v subunits, respectively, suggesting that the ␤3-1-3 mutation does not affect the ␣-␤ association. However, the ␣v␤3-1-3 mutant did not react with LM609, a function-blocking anti-␣v␤3 mAb, upon immunoprecipitation (Fig. 6) and flow cytometric analysis (data not shown), suggesting that the ␤3-1-3 mutation destroyed the LM609 epitope and that the CYDMKTTC sequence of ␤3 is closely located to ligand binding sites of ␣v␤3. DISCUSSION We established that swapping the CTSEQNC sequence of ␤1 with the corresponding CYDMKTTC sequence of ␤3 induces significant changes in ligand specificity of ␣v␤1. The ␤1-3-1 mutation markedly increases affinity of ␣v␤1 to Fg, vWf, and Vn (a gain-of-function effect). Since the ␣v␤1-3-1 mutant is functional in cultured cells and transduces signals on adhesion to the ligands, the swapping did not induce a detectable ad-verse effect on the other receptor functions (e.g. ␣-␤ association and signal transduction). In reciprocal experiments, swapping a disulfide-linked CYDMKTTC sequence of ␤3 with the corresponding CTSEQNC sequence of ␤1 blocks the binding function of ␣v␤3 to Fg, vWf, and Vn. Taken together, the present study suggests that a small disulfide-linked CYMKTTC sequence of ␤3 (and the CTSEQNC sequence of ␤1 as well) defines a novel site of integrin ␤ critical for ligand specificity. Sequence diversity among ␤ subunits and localization within an I domain-like structure of ␤, close to putative ligand binding sites (see Introduction) is consistent with the proposed function of the sequence. In a preliminary study, we introduced mutations into the corresponding predicted loop of the ␤2 subunit. We found that these mutations showed profound effects on the ligand binding function of ␣L␤2 integrin, 2 indicating that the diverse predicted loops of the ␤ subunits are ubiquitously involved in the regulation of ligand binding functions.
Mechanisms by which the disulfide-linked sequences in a predicted loop within the I domain-like structure of the ␤ subunits define ligand specificity of integrins have yet to be studied. In preliminary studies, we did not obtain evidence that the ␤1-3-1 mutation induces constitutive activation of ␤1 integrins or induces drastic conformational changes. We determined the reactivity of the ␤1-3-1 mutant to an activation-dependent anti-␤1 mAb 15/7, which recognizes the highly activated form of ␤1 integrin (30). The binding profiles of 15/7 to the ␤1-3-1 mutant and wild-type ␤1 were identical; binding of 15/7 was dependent on activation in both cases (data not shown). The epitope for 15/7 has been localized within the residues 354 -425 of ␤1 (in the non-ligand binding region outside the I domain-like region) (42). Therefore, there is a possibility that the effect of the ␤1-3-1 mutation on conformation remains local (e.g. within the I domain-like structure of ␤1) and 15/7 does not detect it. The amino acid residues surrounding the tripeptide RGD of the ligands have been reported to be critical for receptor specificity of snake venom disintegrins (43)(44)(45). One possible mechanism is that the predicted loop structures of ␤3 or ␤1 interact with the residues surrounding the tripeptide RGD of ligands, if we assume that the predicted loop structure of ␤ is close to the ligand binding site of ␣v␤3 or ␣v␤1. Another possibility is that the predicted loops regulate the access of a group of ligands (in the case of ␣v␤3, Fg, Vn, and vWf) to the ligand binding site.
( Fig. 7), and this model is similar to our preliminary model (not shown). All of the residues critical for ligand binding (e.g. Asp-130 and Glu-229 of ␤1) (6, 10) are located in the upper face of the model (predicted as the ligand binding site). Also, the regulatory epitope (residues 207-218 of ␤1) (20), which is recognized by both activating and inhibiting anti-␤1 mAbs, is located in the non-ligand binding site (in the lower face) of the domain. Interestingly, a diverse sequence in the predicted loop (e.g. residues 176 -199 in ␤1, residues 166 -190 in ␤3), which is involved in ligand specificity of integrins in the present study, is located in the upper face of the domain in this model. The finding that the ␤3-1-3 mutation blocked binding of the function-blocking anti-␣v␤3 antibody LM609 supports the idea that the predicted loop structure is close to the ligand binding site of ␣v␤3. Taken together, the present and previous mutagenesis data strongly support this model. Recently, Collins Tozer et al. (22) published an interesting atomic model of the putative I domain of ␤3, which is based on the crystal structure of the ␣M I domain (7). However, our mutagenesis data do not fit in very well with their model, since 1) the sequence CYDMKTTC of ␤3, which is critically involved in ligand binding to ␣v␤3, is not close to the MIDAS site (apparently in a non-ligand binding site) in their model, and 2) although Thr-197 of ␤3 is located in the MIDAS site of ␤3 in this model, the corresponding residue of ␤1 (Thr-206) is very close to the regulatory epitope. This epitope is probably located in a non-ligand binding site of ␤1 because 1) binding of some mAbs actually activates, instead of inactivating, the ␤1 integrins, and 2) this epitope has recently been shown to be an allosteric effector site of ␤1 (46), since the binding of an inhibitory anti-␤1 mAb 13 to the regulatory epitope is also dramatically attenuated by ligands (Fn fragments or the GRGDS peptide). Further biochemical and struc-tural characterization of this region of the ␤ subunit may be required to substantiate these models.
␣v␤3 has been shown to be involved in the progression of melanoma and induction of neovascularization by tumor cells. ␣v␤3 antagonists promote tumor regression by inducing apoptosis of angiogenic blood vessels (47,48). We identified a critical region for ligand binding and specificity of integrins using a gain of function mutant of the ␤ subunit. The predicted loop sequence of the integrin ␤3 subunit is a new potential target for designing inhibitors of ligand binding functions of ␣v␤3.