Integrins are heterodimeric adhesion receptors composed of noncovalently associated and subunits. The integrin superfamily consists of at least 20 members that are composed of different combinations of nine and more than 15 subunits. The different combinations of and subunits produce receptors that often possess a distinct ligand recognition specificity. With regard to integrin ligands, a number of discrete sites recognized by integrins have been identified and high resolution structures have been obtained for a number of integrin ligands(1, 2, 3) . An emerging general theme from these structural studies is that integrins recognize protein ligands through interaction with short peptide sequences often presented on extended loops(1, 2, 3, 4, 5) .

There is much less precise information concerning the sites within integrins that recognize ligands. A number of potential ligand interactive sites have been identified in the integrin subunits. Chemical cross-linking, site-directed mutagenesis, and immunological approaches have implicated a highly conserved sequence in the subunit in the ligand binding function(6, 7, 8, 9, 10, 11, 12, 13, 14) . A second site in the same region has also been reported to be involved in ligand binding (15, 16) . Six of the integrin subunits contain an additional 200-residue inserted (I) ()domain, and compelling evidence supports a role for the I domain in ligand binding (17, 18, 19, 20) . Mutational evidence and sequence alignment indicates that the I domain and integrin subunits might utilize a similar mechanism for ligand recognition(10, 18, 21) . These data have led to the hypothesis that the I domain and the conserved subunit ligand recognition site are structurally related and may define a novel motif essential for integrin receptor function(10, 21) . A high resolution structure of a recombinant I domain (22) supports this hypothesis.

A combination of approaches have been utilized to investigate potential ligand binding sites in subunits that do not contain an I domain; however, the results have been inconsistent. Cross-linking studies have demonstrated that bound ligand was proximal to the four divalent cation binding sites in and (23, 24) . Synthetic peptides (25) as well as a recombinant fragment (26) from this region of have been reported to bind ligand. A homology scanning approach mapped the epitopes of antibodies that block ligand binding to to the NH terminus, but not to the cation binding motifs (27) . Finally, the minimal ligand binding fragments of lack the COOH-terminal portions of the receptor, but contain more than half of the entire subunit(28, 29) . Thus, the structures critical for ligand recognition by integrin subunits that lack an I domain remain to be elucidated.

A major difficulty in determining the role of integrin subunits in the regulation of ligand binding specificity is that the binding of most macromolecular ligands is activation-dependent, i.e. the binding of these ligands is highly regulated by the conformational state of the receptor(30, 31) . In contrast, the binding of small peptide ligand mimetics is often activation-independent(32, 33) . A limitation of previous studies aimed at identification of ligand binding sites was that a spectrum of both activation-dependent and -independent ligands were not analyzed. To gain further insight into the structures in the subunits that regulate ligand recognition specificity, we exploited the unique tools available for the integrins and . These two integrins share the common subunit, and the two subunits are 36% identical in primary sequence(34) . They recognize a number of common ligands as well as small peptides containing the Arg-Gly-Asp (RGD) sequence(35) . In addition, there exist highly specific small activation-independent ligands(36, 37, 38) . Moreover, true ligand mimetic monoclonal antibodies, PAC1 (39) and OPG2(40) , have been prepared against . The ligand mimetic property of both mAbs is linked to the tripeptide sequence RYD within the third complementarity-determining region that appears to mimic the RGD recognition sequence(4, 5) . The binding of both antibodies to is blocked by adhesive protein and small competitive peptide ligands(39, 40) . Neither antibody binds to ligand binding defective mutants of (10) . However, these two antibodies differ in that the binding of PAC1 is activation-dependent while the binding of OPG2 does not require prior receptor activation. Finally, ligand binding to these receptors can be assessed indirectly by the conformational changes reported by the exposure of LIBS epitopes(41) . Utilizing this integrin pair, we have defined the region of the subunit that regulates recognition specificity for both activation-dependent and -independent ligands. We report here that neither the cation binding repeats or the NH terminus alone is sufficient to control the ligand recognition specificity of this integrin pair. Ligand specificity requires both regions. A minimal sequence encompassing the amino-terminal one third of the subunit was required to transfer ligand recognition specificity.

MATERIALS AND METHODS

Monoclonal Antibodies and Reagents

Generation of / Chimeric Subunits

Cell Transfection and Flow Cytometry

Binding Assay

Affinity Chromatography and Immunoprecipitation

RESULTS

The Divalent Cation Repeats and the NH-terminal Region of Are Required for an Ligand Binding Specificity

Figure 1: Schematic representation of the / chimeric subunits. Each chimera consists of the backbone of (solid line) from which the indicated portions have been removed and replaced with the homologous region of (shaded boxes). The location of the three silent restriction sites introduced into the wild type sequence to facilitate exchanges and the endogenous SphI site are indicated. Solid black rectangles indicate the position of the four divalent cation binding repeats present in each subunit. The position of residues that delineate chimeras are indicated.

Figure 2: Flow cytometric analysis of stably transfected cell lines expressing / chimeric subunits. CHO cells co-transfected with wild type and the indicated subunit and were examined for receptor expression by flow cytometry. Cells transfected with wild type or the indicated / chimeric subunit were stained by indirect immunofluorescence with the anti- mAb LM142. Cells transfected with were stained with the -specific mAb D57. Results are depicted as histograms with the log of the fluorescence intensity on the abscissa and the cell number on the ordinate.

Since several of the substitutions resulted in reactivity with -specific mAbs, the binding of the ligand mimetic mAb PAC1 was examined utilizing flow cytometry. The binding of mAb PAC1 is activation-dependent(39) . Therefore, while resting exhibited low reactivity with mAb PAC1, activation with the mAb anti-LIBS2 significantly increased the binding of mAb PAC1 (Fig. 3). mAb anti-LIBS2 acts directly upon , provoking high affinity ligand binding function(30) . The binding of mAb PAC1 was specific since it was completely blocked by GRGDSP peptide. Similarly, cells expressing the chimeras 2b(L1-Q459) or 2b(L1-P334) specifically bound mAb PAC1 in the presence of activating mAb anti-LIBS2. In contrast, cells expressing wild type , 2b(L1-F223), or 2b(1-4C) failed to bind mAb PAC1 after activation with the mAb anti-LIBS2. The lack of mAb PAC1 binding to these chimeras was not due to the failure of anti-LIBS2 to bind to the chimeric receptor as the epitope was present on each of these receptors as assayed by flow cytometry (data not shown). These data suggest that the chimeras 2b(L1-Q459) and 2b(L1-P334) have a ligand binding pocket very similar to .

Figure 3: Binding of mAb PAC1 to cells stably transfected with , , or chimeric / receptors. The binding of the activation-specific mAb PAC1 to CHO cells stably transfected with and the indicated subunit was examined by flow cytometry. Results are depicted as histograms of cell number versus fluorescence intensity. Transfected cells were incubated (activated) in the presence of 8 µM purified IgG anti-LIBS2 for 30 min followed by the addition of mAb PAC1 (IgM). Cells were washed, stained with fluorescein-conjugated goat anti-mouse IgM for 30 min, and analyzed. The binding of mAb PAC1 was analyzed in the presence (open histogram) or absence (solid histogram) of 1 mM GRGDSP peptide.

Interaction of these / chimeras with another ligand mimetic mAb, OPG2, was also examined by flow cytometry (Fig. 4). OPG2 inhibits the binding of adhesive proteins to and its binding is blocked by RGD peptides(40) . However, unlike mAb PAC1, the binding of mAb OPG2 to is activation-independent. Cells expressing wild type stained brightly with mAb OPG2. Consistent with the results obtained with mAb PAC1, mAb OPG2 bound to cells expressing 2b(L1-Q459) or 2b(L1-P334). No specific mAb OPG2 staining was observed with cells expressing wild type , 2b(L1-F223), or 2b(1-4C). The fact that neither mAb bound to the chimera 2b(L1-F223) indicates that the NH-terminal region alone does not control ligand binding specificity.

Figure 4: Expression of the OPG2 epitope on recombinant wild type , , or chimeric / receptors. The binding of the complex-specific mAb OPG2 to CHO cells stably transfected with and the indicated subunit was examined by flow cytometry. Results are depicted as fluorescence-activated cell sorting histograms. In each panel, the binding of mAb OPG2 (solid histogram) is superimposed on the binding of the anti- mAb 142 (open histogram).

The Divalent Cation Binding Repeats Alone Do Not Determine Ligand Binding Phenotype

To determine the capacity of the chimeras containing substitutions of the divalent cation repeats to bind small activation-independent ligands specific for , we examined the capacity of the -selective peptidomimetic Ro 43-5054 (38, 51) to increase the binding of mAb anti-LIBS1 by flow cytometry. Since the mAb anti-LIBS1 binds preferentially to the occupied conformation of the receptor(41) , increased binding of mAb LIBS1 is evidence of receptor-ligand interaction. In the presence of Ro 43-5054, there was an increase in the binding of anti-LIBS1 to cells expressing but not to cells expressing (Fig. 5). Similarly, Ro 43-5054 failed to stimulate the binding of mAb anti-LIBS1 to cells expressing the chimeras 2b(2+3C) or 2b(3+4C), indicating lack of binding to the receptor. Unexpectedly, mAb anti-LIBS1 bound maximally to cells expressing the chimeras 2b(1+2C) or 2b(1-4C) even in the absence of ligand. This result suggested that these two chimeras possessed a structure that is slightly altered from that of the wild type receptors. Although the anti-LIBS1 epitope was exposed on these two chimeras, additional data (see below) indicate that their ability to bind ligand was not impaired.

Figure 5: Flow cytometric analysis of the capacity of transfected cells to bind an -specific peptidomimetic. The binding of the -selective peptidomimetic Ro 43-5054 (38) to cells stably transfected with and the indicated subunits was examined with the mAb anti-LIBS1. For LIBS1 binding analysis, cells were incubated with or without Ro 43-5054 (5 µM) and LIBS1 mAb (0.1 µM) for 30 min on ice. Cells were washed and incubated with fluorescein-conjugated goat anti-mouse Ig. Results are expressed as histograms of cell number versus fluorescence intensity. In each panel, the binding of LIBS1 in the presence of Ro 43-5054 (solid histogram) is overlaid on the binding of LIBS1 in the absence of Ro 43-5054 (open histogram).

To test whether the chimeric receptors containing substitutions of the cation binding repeats possessed an intact RGD ligand recognition function and to test their capacity to distinguish between the RGD and fibrinogen chain sequence, the ligand binding function of the recombinant receptors was analyzed by affinity chromatography (Fig. 6). Detergent lysates of radiolabeled, transfected cells were applied to an RGD affinity column and eluted with the fibrinogen chain peptide K16, followed by elution with EDTA. The eluted fractions were then immunoprecipitated with an anti- mAb. Eluted fractions of the control -expressing cells were immunoprecipitated with anti- antiserum(30) . Precipitated proteins were then resolved by SDS-polyacrylamide gel electrophoresis. Consistent with previous reports(48) , wild type was poorly eluted by the K16 peptide (data not shown). While 64% of the bound wild type was eluted from the affinity matrix by the chain peptide K16, the chimeras 2b(1-4C) (8.4%), 2b(1+2C) (3.4%), 2b(2+3C) (7.5%), or 2b(3+4C) (13%) were poorly eluted by the K16 peptide from the RGD affinity matrix (Fig. 6). Each of these receptors bound to the RGD matrix and was readily eluted from the matrix by EDTA. Both wild type and receptors and all the chimeras were readily eluted from the affinity matrix by RGD peptide (data not shown). The fact that the chimeras 2b(1-4C) and 2b(1+2C) bound to the RGD affinity matrix and were specifically eluted by EDTA or RGD peptide indicates that the alteration in structure reported by anti-LIBS1 did not affect the ligand binding function of these receptors. These data show that all chimeras containing substitutions of the cation binding motifs can recognize the RGD sequence, but that substitution of the divalent cation binding regions with the corresponding regions from was not sufficient to change the ligand binding specificity of to that of .

Figure 6: Fibrinogen chain peptide K16 does not displace or the / divalent cation binding repeat chimeras from a RGD affinity matrix. CHO cells stably expressing the wild type or chimeric / receptors were radioiodinated and lysed, and the extract was applied an GRGDSPK-Sepharose 4B column. After incubation and washing, the bound proteins were sequentially eluted with 1.5 mM K16, followed by 5 mM EDTA. The eluted fractions were immunoprecipitated with the anti- mAb LM142. The immunoprecipitated proteins were resolved by SDS-polyacrylamide gel electrophoresis on 7% nonreducing acrylamide gels and detected by autoradiography. Lanes 1, immunoprecipitate of K16-eluted material; lanes 2, immunoprecipitate of subsequent EDTA-eluted material.

Direct Binding of an -selective Peptidomimetic to Chimeric Integrins

Figure 7: Direct binding of an -selective peptidomimetic. The binding of the -specific peptidomimetic SC52012 (37) to stably transfected cell lines expressing , , or the indicated chimeric / receptor was determined by incubating transfected cells with [H]SC52012 (500 nM) at room temperature. After 40 min, bound ligand was separated from free ligand by centrifugation through 20% sucrose. The pellet associated counts were determined by liquid scintillation spectrometry. Background binding was measured in the presence of 5 mM EDTA. Shown are representative results of three separate assays. Results shown are mean ± S.D. of triplicates.

DISCUSSION

The major findings of the present study are as follows. 1) Ligand recognition specificity of integrins is regulated by the amino-terminal one-third of the subunit. Substitution of the amino-terminal portion of with the corresponding 334 amino acid residues of switched the ligand recognition specificity of to that of . This change in ligand specificity was observed with an activation-dependent ligand mimetic antibody, an activation-independent ligand mimetic antibody, and small activation-independent ligands. 2) Neither the amino-terminal region or the cation binding repeats alone is sufficient to control ligand specificity. Chimeras that omit the amino-terminal 140 residues or first two divalent cation binding repeats of fail to change ligand specificity. Thus, the ligand binding pocket of is a structure that contains elements of both the and subunits.

Previous studies have suggested that the regions that control ligand binding to reside in the amino-terminal portion of , but the minimal structures identified in these studies encompassed more than one half of constituent subunits(28, 29) . In the present study, we have mapped the regions that regulate ligand specificity to a smaller region of . The chimera designated 2b(L1-Q459) contained the amino-terminal portion and all four divalent cation repeats of and reacted with several complex-specific mAbs. In addition, this chimera specifically bound small activation-independent -specific peptidomimetics and both activation-dependent (PAC1) and activation-independent (OPG2) ligand mimetic mAbs. The chimera 2b(L1-P334) retains the amino-terminal portion of but contains only the first two divalent cation repeats of . This chimera also exhibited a ligand binding phenotype consistent with that of in that it bound specific peptidomimetics, the ligand mimetic mAbs PAC1 and OPG2, and several -specific mAbs. These results indicate that the ligand specificity of can be reconstituted with the first 334 amino acid residues of and does not require the third or fourth divalent cation repeats of .

Chimeras that omit the 140 amino-terminal residues or the first two divalent cation motifs of fail to change the ligand specificity of to that of . The chimera 2b(1-4C) contains a substitution of the entire divalent cation repeat region of with the corresponding region of . This chimera was expressed on the cell surface and could bind ligand as demonstrated by its ability to bind to an RGD affinity matrix. However, this chimera was poorly displaced from the matrix by a fibrinogen chain peptide and did not bind the ligand mimetic mAbs PAC1 and OPG2 or an -specific peptidomimetic. These data indicate that substitution of the divalent cation repeats alone is not sufficient to change the ligand binding specificity. Similarly, the chimera