Determinants of Ligand Binding Specificity of the α1β1 and α2β1Integrins*

The α1β1 and α2β1 integrins are cell surface collagen receptors. Cells expressing the α1β1integrin preferentially adhere to collagen IV, whereas cells expressing the α2β1 integrin preferentially adhere to collagen I. Recombinant α1 and α2 integrin I domains exhibit the same collagen type preferences as the intact integrins. In addition, the α2 integrin I domain binds echovirus 1; the α1 I domain does not. To identify the structural components of the I domains responsible for the varying ligand specificities, we have engineered several α1/α2 integrin I domain chimeras and evaluated their virus and collagen binding activities. Initially, large secondary structural components of the α2 I domain were replaced with corresponding regions of the α1 I domain. Following analysis in echovirus 1 and collagen binding assays, chimeras with successively smaller regions of α1 I were constructed and analyzed. The chimeras were analyzed by ELISA with several different α2 integrin monoclonal antibodies to assess their proper folding. Three different regions of the α1 I domain, when present in the α2 I domain, conferred enhanced collagen IV binding activity upon the α2 I domain. These include the α3 and α5 helices and a portion of the α6 helix. Echovirus 1 binding was lost in a chimera containing the αC-α6 loop; higher resolution mapping identified Asn289 as playing a critical role in echovirus 1 binding. Asn289 had not been implicated in previous echovirus 1 binding studies. Taken together, these data reveal the existence of multiple determinants of ligand binding specificities within the α1 and α2 integrin I domains.

The integrins constitute a large family of cell adhesion molecules that are involved in both cell-cell adhesion as well as the adhesion of cells to the extracellular matrix (for review see Ref. 1). The integrins are involved in many important physiologic processes including development and differentiation, cell migration, wound healing, thrombosis and hemostasis, metastasis, and immune system function. Structurally, integrins are heterodimeric glycoproteins composed of two noncovalently associated integral membrane subunits. The ␣ subunits range in size from 120 to 180 kDa, and the ␤ subunits range in size from 90 to 110 kDa. Many integrin subunits associate with more than a single ␣ or ␤ subunit resulting in a large number of different integrins.
The ␤ 1 integrin subunit associates with at least 11 different ␣ subunits. Members of the resulting ␤ 1 subfamily of integrins are cell surface receptors for specific components of the extracellular matrix. The ␣ 1 ␤ 1 integrin is a cell surface receptor for several different collagens and laminin-1 (2). ␣ 1 ␤ 1 integrin-dependent adhesion of cells to collagens or laminin requires the presence of divalent cations (3). Similarly, the ␣ 2 ␤ 1 integrin also serves as a cell surface receptor for collagens and laminins (for review see Ref. 4), and the adhesion of cells via the ␣ 2 ␤ 1 integrin to these ligands also depends upon the presence of divalent cations (5). However, the substrate specificity of the ␣ 2 ␤ 1 integrin depends upon the cell type on which it is expressed. The ␣ 2 ␤ 1 integrin on platelets and fibroblasts is a collagen receptor (5); on endothelial and epithelial cells it is a receptor for both collagens and laminin-1 (6,7). The ␣ 2 ␤ 1 integrin also mediates echovirus 1 attachment and infection (8).
Near their amino termini, the ␣ 1 and the ␣ 2 integrin subunits share (along with approximately half of the integrin ␣ subunits) an autonomously folding domain of approximately 220 amino acids known as the I (inserted) domain (for review see Ref. 9). The ␣ subunit I domains are critical determinants for ligand recognition and binding of both the ␣ 1 ␤ 1 and the ␣ 2 ␤ 1 integrins. Function blocking antibodies directed against both of these integrins map to their ␣ subunit I domain, and several mutations in the I domains adversely affect ligand binding activity (10,11). Bacterially expressed I domains from both of these integrins bind collagens and laminin-1 in a manner that is saturable, blocked by function blocking antibodies against the parent integrin, and, like the parent integrins, dependent upon the presence of divalent cations (12)(13)(14).
The crystal structures of several different integrin I domains have been solved, providing much insight into I domain structure. I domain structures that have been solved include those of the ␣ M , ␣ L , and ␣ 2 subunits (15)(16)(17). The three-dimensional structures of these three I domains are very similar. Each has a core that consists of a largely parallel ␤ sheet structure, and the core is surrounded by several ␣ helices. All three contain a single divalent cation-binding site in a crevice near the top of the ␤ sheet. Unique to the ␣ 2 I domain, however, is an additional short ␣ helix at the top of the domain, in close proximity to the divalent cation-binding site (17).
Although cells expressing either the ␣ 1 ␤ 1 or the ␣ 2 ␤ 1 integrins adhere to several different collagens, they exhibit different relative affinities depending on the identity of the substrate. For instance, cells expressing the ␣ 1 ␤ 1 integrin preferentially adhere to collagen IV, whereas cells expressing the ␣ 2 ␤ 1 integrin preferentially bind to collagen I (18). Likewise, the binding of purified recombinant ␣ 1 and ␣ 2 integrin I domains to collagen types I and IV reflects the same relative affinities to these ligands as do the parent integrins (13). To identify the structural components of these I domains responsible for conferring the observed substrate specificities, we have engineered a series of ␣ 1 /␣ 2 integrin I domain chimeras and examined their ligand binding activities in collagen I and IV and echovirus 1 binding assays. Initially, with the aid of the crystal structure data mentioned above, relatively large structural components of the ␣ 2 integrin I domain were replaced with the corresponding region from the ␣ 1 integrin I domain. The chimeras were examined for both a gain of function (enhanced collagen IV binding activity) and a loss of function (echovirus 1 binding). Regions that satisfied one or both of these criteria were subjected to subsequent rounds of chimera construction, purification, and analysis to define further the critical structural and functional determinants.

EXPERIMENTAL PROCEDURES
Cloning and Mutagenesis of Integrin I Domain cDNAs-The cloning and expression of the ␣ 2 integrin I domain has been previously described (12). Briefly, cDNA encoding the I domain was amplified using the polymerase chain reaction (PCR) 1 with the full-length ␣ 2 integrin cDNA as the template. This cDNA encodes Ser 124 -Met 349 of the published ␣ 2 integrin sequence (19). In addition, a cDNA encoding a shorter I domain protein lacking the 35 amino-terminal amino acids was also prepared. The shorter protein, referred to as ⌬I, contains Trp 159 -Met 349 . This proteins lack the DXSXS portion of the metal ion dependent adhesion site (MIDAS) motif (15). The PCR primers were designed such that both of the amplification products would contain a BglII restriction site at their 5Ј ends and a stop codon followed by an XhoI restriction site at their 3Ј ends. The PCR products were digested with BglII and XhoI, purified in agarose gels, and cloned into BamHI-and XhoI-digested glutathione S-transferase (GST) fusion protein expression vector pGEX-5X-1 (Amersham Pharmacia Biotech).
cDNA encoding the ␣ 1 integrin I domain was amplified by PCR using the full-length human ␣ 1 cDNA (Dr. Eugene E. Marcantonio, Columbia University) as the template. Analogous to the ␣ 2 I domain construct, this cDNA encodes Ser 124 -Met 349 of the published ␣ 1 integrin sequence (20). As with the ␣ 2 integrin I domain, the PCR primers were designed to create a product that contained a BglII site at the 5Ј end and a stop codon followed by a XhoI site at the 3Ј end. The PCR product was digested, purified, and cloned into pGEX-5X-1 as described above.
To facilitate the construction of chimeric ␣ 1 /␣ 2 integrin I domain cDNAs encoding the ␣ 2 integrin I domain with an internal region replaced with ␣ 1 I domain sequence, the cDNAs encoding both the I domains were transferred into pBluescript KS Ϫ (Stratagene). A 661-bp EcoRI-XhoI fragment of pGEX-5X-1/␣ 1 I, encoding the entire ␣ 1 I domain with the exception of the first seven amino acids, was isolated and ligated into pBluescript KS Ϫ digested with EcoRI and XhoI. For the ␣ 2 I domain, a 1157-bp BalI-XhoI fragment of pGEX-5X-1/␣ 2 I, encoding all of the ␣ 2 I domain and containing approximately two-thirds of the GST coding sequence, was isolated and ligated into pBluescript KS Ϫ digested with SmaI and XhoI.
Chimeric ␣ 1 /␣ 2 integrin I domain cDNAs were prepared by one of two methods, depending upon the length of the internal portion of ␣ 2 to be replaced with ␣ 1 sequence. An example of the construction of one of each type of chimera will be described. Chimera A, an example of one of the chimeras with a longer replacement, consists of the ␣ 2 I domain with an internal region from the ␣ 1 I domain consisting of the ␣1 helix and the ␣ 1 -␤B loop. First, pBluescript/␣ 2 I was mutated using Kunkel's method (21) and oligo 1 as the mutagenic oligo (Table I). Oligo 1 introduced an Eco47III site into the ␣ 2 sequence at the amino-terminal end of the ␣1 helix to create p4. Simultaneously, pBluescript/␣ 1 I was mutated using oligo 2 as the mutagenic oligo. This introduced an EcoRV site at the amino-terminal end of the ␣1 helix (a position analogous to the position of the Eco47III site in p4) to create p3. Eco47III and EcoRV were chosen so that the proper amino acids, serine and isoleucine, would be encoded by the 3Ј end of the ␣ 2 fragment and the 5Ј end of the ␣ 1 fragment, respectively, upon restriction enzyme digestion. The 3482-bp XhoI-Eco47III fragment of p4 and the 591-bp EcoRV-XhoI fragment of p3 were isolated and ligated to create p5, encoding amino acids 124 -155 of ␣ 2 followed by amino acids 155-349 of ␣ 1 . (Note that amino acid 155 is present from both the ␣ 1 I and ␣ 2 I sequences. This is due to the fact that a one-amino acid gap must be introduced into the ␣ 1 protein sequence between Pro 135 and Val 136 to align the two protein sequences (see Fig. 1). Thus Ile 155 of ␣ 1 corresponds to Ile 156 of ␣ 2 .) p5 was mutated using oligo 3 to introduce a StuI site at the carboxylterminal end of the ␤B strand to create p6. Simultaneously, pBluescript/␣ 2 I was mutated using oligo 4 to introduce an Eco47III site at the carboxyl-terminal end of the ␤B strand to create p7. StuI and Eco47III were chosen so that the proper amino acids, glycine and leucine, would be encoded at the 3Ј end of the ␣ 1 portion of p6 and the 5Ј end of the ␣ 2 fragment, respectively, upon restriction enzyme digestion. The 3565-bp XhoI-StuI fragment of p6 and the 505-bp Eco47III-XhoI fragment of p7 were isolated and ligated to create p8, encoding amino acids 124 -155 of ␣ 2 followed by amino acids 155-182 of ␣ 1 , which is in turn followed by amino acids 184 -349 of ␣ 2 . It is important to note that although amino acids 155-182 of ␣ 1 are encoded in this chimeric I domain, only amino acids 160 -178 differ in the two I domains. This is due to the fact that amino acids 155-159 of ␣ 1 are identical to amino acids 156 -160 of ␣ 2 and that amino acids 179 -182 of ␣ 1 are identical to amino acids 180 -183 of ␣ 2 ( Figs. 1 and 3). Finally, the 966-bp Bst BI-XhoI fragment of p8 and the 4669-bp XhoI-Bst BI fragment of pGEX-5X-1/␣ 2 I were isolated and ligated to create the expression plasmid pGEX-5X-1/Ch A, capable of directing the expression of the GST-chimera A fusion protein.
Chimera B2, an example of one of the chimeras with a shorter internal portion of the ␣ 1 integrin I domain, consists of the ␣ 2 I domain with an internal region from the ␣ 1 I domain consisting only of the ␣3 helix, amino acids 205-210 (VLVAAK). Amino acids 206 -211 (MI-VATS) of the ␣ 2 I domain sequence were replaced with amino acids 205-210 (VLVAAK) of the ␣ 1 I domain sequence in a single mutagenesis reaction using pBluescript/␣ 2 I and oligo 5 (Table I) to create pBluescript/Ch B2. The 966-bp Bst BI-XhoI fragment of pBluescript/Ch B2 and the 4669-bp XhoI-Bst BI fragment of pGEX-5X-1/␣ 2 I were isolated and ligated to create pGEX-5X-1/Ch B2. The sequences of the cDNAs used in this study, including all of the chimeras, were determined using the BigDye terminator cycle sequencing method (PE Applied Biosystems), and were compared with the published ␣ 1 and ␣ 2 integrin sequences (19,20).
Expression and Purification of GST-I Domain Fusion Proteins-Trial inductions were performed to determine whether the selected clones could direct the expression of appropriately sized GST fusion proteins. Escherichia coli DH5␣ containing each of the plasmid constructs was grown at 37°C in 5 ml of 2ϫ YT medium supplemented with 0.2% glucose and 100 g/ml ampicillin. Uninduced samples were removed from each culture after 1 h. Isopropylthiogalactoside was then added to a final concentration of 1 mM, and the cultures were returned to the incubator for 3 h to permit for accumulation of the expressed proteins. Cell lysates from the uninduced and induced samples were analyzed by SDS-polyacrylamide gel electrophoresis (22) followed by Coomassie Blue staining. Constructs that directed the expression of recombinant proteins of the expected size were used for large scale induction of protein expression.
For the purification of the fusion proteins, the inductions were performed as above except that the culture volume was increased to 500 ml and the expression was induced with 1 mM isopropylthiogalactoside when the A 550 reached 0.3-0.4. At the end of the induction period, the cells were recovered by centrifugation at 2600 ϫ g for 10 min, washed twice with 10 ml of ice-cold phosphate-buffered saline (10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , 140 mM NaCl, 2.7 mM KCl, pH 7.4), and stored at Ϫ70°C until needed. The GST-I domain fusion proteins were expressed, purified, and characterized as recently described (12).
Collagen Binding Assays-The wells of a 96-well microtiter plate (Immulon 2, Dynatech) were coated overnight at 4°C with 0.1 ml/well of 30 g/ml collagen I from calf skin (Sigma) or collagen IV from human placenta (Sigma) in 0.09% acetic acid. The wells were washed twice . The wells were washed once with 0.15 ml of the appropriate wash buffer, and then 0.1 ml of each recombinant protein was added and allowed to interact with the collagen substrate for 1.5 h at room temperature. Wells were then washed three times with 0.15 ml of the appropriate wash buffer and 0.1 ml of a 1:8,000 dilution of anti-GST antiserum (Amersham Pharmacia Biotech) in the appropriate wash buffer was added for 1 h at room temperature. Following this incubation, the wells were again washed three times, and then 0.1 ml of a 1:20,000 dilution of pig anti-goat IgG secondary antibody horseradish peroxidase conjugate (Roche Molecular Biochemicals) in the appropriate wash buffer was added per well for 1 h at room temperature. The wells were again washed three times, and 0.1 ml of tetramethylbenzidine dihydrochloride (Sigma) was prepared according to the manufacturer's directions was added per well. After 15 min of substrate conversion, reactions were stopped with 0.025 ml of 4 N H 2 SO 4 , and the plates were read at 450 nm using a Molecular Devices E max microplate reader. Each protein was expressed and purified twice, and both preparations were tested in collagen I and IV binding assays. All data are presented as means of triplicate determinations. Echovirus 1 Binding Assays-[ 35 S]Methionine-labeled echovirus 1, prepared as described (23), was added to purified fusion proteins immobilized on glutathione-Sepharose (10,000 cpm, 2 g of protein/aliquot) and then incubated for 1 h at room temperature with rocking. Beads were washed to remove unbound virus and then dissolved for liquid scintillation counting. Additional aliquots of immobilized fusion proteins were boiled in Laemmli sample buffer (22) and analyzed by SDS-polyacrylamide gel electrophoresis to confirm equal protein loading.
ELISA-I domain-containing proteins were diluted to 10 g/ml in TBS containing 2 mM MgCl 2 and used to coat the wells of a 96-well microtiter plate (Immulon 2, Dynatech). Coating was carried out overnight at 4°C with 0.1 ml of solution/well. The wells were washed twice with 0.15 ml of TBS containing 2 mM MgCl 2 and then blocked for 1 h at room temperature with 0.15 ml of TBS containing 300 g/ml bovine serum albumin and 2 mM MgCl 2 . Primary antibodies used include the human ␣ 2 integrin I domain monoclonal antibodies 6F1 (Dr. Barry S. Coller, Mt. Sinai Medical Center), 12F1 (Dr. Virgil L. Woods, Jr., University of California at San Diego), 5E8 (Dr. Richard B. Bankert, Roswell Park Cancer Institute), P1E6 (Chemicon), and AK7 (Chemicon). The monoclonal antibodies were diluted to 1 g/ml in wash buffer (TBS containing 0.05% Tween-20, 30 g/ml bovine serum albumin, and 2 mM MgCl 2 ). Following blocking the wells were washed once with 0.15 ml of wash buffer, and then 0.1 ml of primary antibody was added and allowed to interact for 1 h at room temperature. The wells were washed three times with 0.15 ml of wash buffer, then 0.1 ml of a 1:10,000 dilution of goat anti-mouse IgG secondary antibody horseradish peroxidase conjugate (Roche Molecular Biochemicals) in the appropriate wash buffer was added per well for 1 h at room temperature. The wells were again washed three times, and 0.1 ml of tetramethylbenzidine dihydrochloride (Sigma) prepared according to the manufacturer's directions was added per well. After 10 min of substrate conversion, reactions were stopped with 0.025 ml of 4 N H 2 SO 4 and the plates read at 450 nm using a Molecular Devices E max microplate reader. Each protein was expressed and purified twice, and both preparations were analyzed in the ELISA assays. All data represent means of triplicate determinations.

RESULTS
Binding of the ␣ 1 ␤ 1 and ␣ 2 ␤ 1 integrins to a variety of extracellular matrix molecules has been examined previously. The ␣ 1 ␤ 1 integrin preferentially binds to collagen IV, whereas the ␣ 2 ␤ 1 integrin preferentially binds collagen I (18). Analysis of the binding of the ␣ 1 and ␣ 2 integrin I domains to collagens I and IV using surface plasmon resonance and solid phase binding assays indicate that the substrate preferences of the ␣ 1 and ␣ 2 I domains reflect those of the parent integrins (13). We also have examined the binding of these I domains to collagens I and IV in solid phase substrate binding assays. In agreement with the above, we observed that the ␣ 1 I domain preferentially bound collagen IV, and the ␣ 2 I domain preferentially bound collagen I (Fig. 2A). The collagen type preferences of the ␣ 1 and ␣ 2 I domains were confirmed by more extensive studies over a wide range of I domain concentrations (see Fig. 7, left and middle panels).
The ␣ 2 ␤ 1 integrin is also a cell surface receptor for echovirus 1 (8). Binding of echovirus 1 to the integrin is mediated by the ␣ 2 I domain (24). We tested the binding of echovirus 1 to both the ␣ 1 and ␣ 2 integrin I domains. As expected, the virus bound to the ␣ 2 I domain. However, the virus failed to bind to the ␣ 1 I domain (Fig. 2B). To determine which regions of the ␣ 1 I domain were responsible for enhanced collagen IV binding activity and which regions of the ␣ 2 I domain were required for echovirus binding, we designed a series of ␣ 1 /␣ 2 integrin I domain chimeras and examined their collagen I and IV and echovirus 1 binding activities. As a guide for designing the chimeras, we used the recently published crystal structure of the ␣ 2 integrin I domain (17). Initially, relatively large portions of the ␣ 2 I domain were replaced with the analogous regions of the ␣ 1 I domain. Because loops and helices are exposed on the I domain surface, the initial round of chimeras contained loops and ␣ helices from the ␣ 1 I domain. Fig. 3 (A-F) shows the initial round of six ␣ 1 /␣ 2 I domain chimeras and indicates the regions of the ␣ 2 I domain that were replaced with ␣ 1 I domain sequence.
Each of the chimeras was tested in collagen I and IV binding assays and the echovirus 1 binding assay. In addition, to assess the folded state of the I domain chimeras, each was tested by ELISA using a panel of five ␣ 2 integrin I domain monoclonal antibodies. Because any given epitope could be lost because of replacement of the epitope with protein sequence from the ␣ 1 I domain, the chimeras were considered to be folded properly if a majority of the monoclonal antibodies bound. The results of these analyses are shown in Fig. 3. The collagen binding data are given as percentages Ϯ S.E. of control (binding of the ␣ 2 I domain to collagen I); the echovirus binding data are also expressed as the percentages Ϯ S.E. of control (binding of echovirus 1 to ␣ 2 I domain). Of the six chimeras, A and C failed to bind to collagen I, collagen IV, and echovirus 1. Furthermore, neither of these chimeras was effectively recognized by any of the ␣ 2 I domain monoclonal antibodies. Therefore, we assume chimeras A and C are incorrectly folded, and no conclusions can be drawn from their lack of ligand binding. The remaining four chimeras, B, D, E, and F, were recognized by a majority of the antibodies and bound at least one of either collagen I, collagen IV, or echovirus at a level comparable with or greater than that of the ␣ 2 I domain. Of these four chimeras, B, D, and E showed enhanced collagen IV binding activity. This represents a gain of function; these chimeras have become more ␣ 1 I domain-like with respect to their collagen binding activity. Also, of the four chimeras showing positive antibody reactivity, only chimera E lost the ability to bind echovirus 1. This finding indicated an important role in echovirus 1 binding for the region of the ␣ 2 I domain consisting of the ␣C helix, the ␣C-␣6 loop, and the ␣6 helix.
To further localize regions of the ␣ 1 I domain responsible for enhanced collagen IV binding activity and regions of the ␣ 2 I domain necessary for echovirus 1 binding, iterative rounds of chimera design, expression, purification, and analysis were undertaken. The design of each round was based on the results of the analysis of the previous round. Fig. 4 shows the resultant B series of chimeras. Because chimera B possessed enhanced collagen IV binding activity and contained both the ␤C-␣3 loop and the ␣3 helix from the ␣ 1 I domain, chimeras B1 and B2 contained just the ␤C-␣3 loop (and one residue from the aminoterminal end of the ␣3 helix), or the remainder of the ␣3 helix from the ␣ 1 I domain, respectively. The enhanced collagen IV binding activity of chimera B segregated completely into chimera B2 containing only the ␣3 helix from the ␣ 1 I domain. Chimera B1 reverted to the parental ␣ 2 I domain phenotype of preferential binding to collagen I. An attempt to further map the region of the ␣3 helix necessary for the enhancement (Fig.  4, chimeras B2a and B2b) resulted in reversion to the ␣ 2 I domain collagen binding properties, indicating the requirement of the entire ␣3 helix from the ␣ 1 I domain for the effect.
Chimera D, containing the ␣5 helix from the ␣ 1 I domain, as well as three amino-terminal flanking residues and one carboxyl-terminal flanking residue, had also shown enhanced collagen IV binding activity. To determine which region was required for the effect, chimeras D1 and D2 were prepared. These chimeras contained the mutations GSMLKA 3 NHRLKK (three residues from the ␤D-␣5 loop and the amino-terminal half of the ␣5 helix) and DQCNHD 3 QDCEDE (carboxylterminal half of the ␣5 helix and one residue from the ␣5-␤E loop). The results of collagen I and IV binding assays using these chimera are shown in Fig. 5. Chimera D1, containing the amino-terminal half of the ␣5 helix of the ␣ 1 I domain, bound to both collagens I and IV considerably less effectively than did chimera D. Chimera D2, containing the carboxyl-terminal half of the ␣5 helix of the ␣ 1 I domain, bound to collagen I as well as chimera D, but again the binding to collagen IV was diminished with respect to that of chimera D. Thus it appears that the entire ␣5 helix from the ␣ 1 I domain is required to observe the enhanced collagen IV binding activity.
Both enhanced collagen IV binding activity as well as complete loss of echovirus binding activity was observed for chimera E, containing the ␣C helix, the ␣C-␣6 loop, and the ␣6 helix. Chimeras E1, E2, and E3 were prepared, containing only the ␣C helix, the ␣C-␣6 loop, and the ␣6 helix, respectively (Fig.  6). Chimera E3 reverted to the collagen binding activity profile and the echovirus binding activity of the ␣ 2 I domain. Chimera E2 showed both enhanced collagen IV binding activity as well as significantly diminished echovirus binding activity. Chimera E1, which contained a two-residue mutation in the ␣C helix (YL 3 SY), bound to both collagens I and IV. However, as opposed to the binding of both the ␣ 1 and ␣ 2 I domains as well as chimera E, the binding of chimera E1 to both collagens I and IV was independent of divalent cations. The binding of chimera E1 to collagen IV was not as pronounced as the binding of chimeras E and E2 to collagen IV. Chimera E1a, which contained a single amino acid replacement (Y 3 S) within the ␣C helix, also bound both collagens I and IV in a divalent cationindependent manner. However, chimera E1R, in which the residues YL in the ␣C helix of the ␣ 2 I domain were reversed to LY, showed enhanced collagen IV binding activity and bound collagens I and IV in a divalent cation-dependent manner.
To localize the regions of E2 required for enhanced collagen IV binding activity and loss of echovirus binding, E2a, E2b, and E2c were prepared (Fig. 6). The amino acid replacements in these chimeras are NA 3 GN, D 3 S, and KN 3 EK, respectively. The enhanced collagen IV binding activity segregated to chimera E2c, containing residues EK from the ␣ 1 I domain at the amino-terminal end of the ␣6 helix. Chimera E2c1, containing only the point mutation K 3 E at the amino-terminal end of the ␣6 helix, reverted to the collagen binding properties of the ␣ 2 I domain, indicating the importance of K294 for the enhanced collagen IV binding activity of chimera E2c. The loss of echovirus binding activity segregated completely to E2a, in which NA in the ␣C-␣6 loop had been replaced. Chimeras E2a1 and E2a2, containing point mutations N 3 G and A 3 N in the ␣C-␣6 loop, were prepared to determine which of the two residues from the ␣ 1 I domain present in E2a was responsible for the loss of echovirus 1 binding activity. This analysis revealed that Asn 289 of the ␣ 2 I domain, within the ␣C-␣6 loop, is required for echovirus binding.
Because chimeras E2 and E2a containing amino acids 288 -294 and 288 -289 of the ␣ 1 I domain had severely diminished echovirus 1 binding activity (E2) or exhibited in a complete loss (E2a) of echovirus 1 binding activity, reciprocal I domain chimeras were constructed in which amino acids 288 -294 and 288 -289 of the ␣ 1 I domain were replaced with the corresponding residues from the ␣ 2 I domain to determine whether either of these replacements could confer echovirus 1 binding activity upon the ␣ 1 I domain. Neither of these two chimeras, E2Ј or E2aЈ, gained echovirus 1 binding activity (data not shown). However, these two chimeras also failed to bind collagens I and IV in a divalent cation-dependent manner and therefore may represent incorrectly folded species.
To gain a more complete understanding of the reciprocal collagen type specificities of the ␣ 1 I domain and the ␣ 2 I domain, as well as the cation-independent collagen binding activity of chimera E1, the binding of each of these proteins to collagens I and IV was measured over a wide range of I domain concentrations and in the presence of 1 mM EDTA or 2 mM Mg 2ϩ (Fig. 7). The preferential binding of the ␣ 1 I domain to collagen IV and of the ␣ 2 I domain to collagen I was evident over the entire range of I domain concentrations. In addition, the divalent cation-independent binding of chimera E1 to collagens I and IV was confirmed over the entire range of protein concentrations tested.
Chimera F, containing the ␣7 helix from the ␣ 1 I domain, was recognized by a majority of the ␣ 2 integrin I domain monoclonal antibodies and echovirus 1 but showed significantly impaired collagen (both I and IV) binding activity. Three more chimeras, F1, F2, and F3, were prepared to ascertain which region of the ␣7 helix from the ␣ 1 I domain was necessary for the loss of collagen binding activity. Chimeras F1, F2, and F3 contained the mutations A 3 L, LEKAG 3 VTIVK, and Q 3 R within the ␣7 helix, respectively. Chimera F1 reverted to the collagen binding activity profile of the ␣ 2 I domain (Fig. 8). Chimera F2 showed somewhat diminished collagen I binding activity (in comparison to the ␣ 2 I domain) and severely impaired collagen IV binding activity. Of the chimeras F1-F3, F3 showed the FIG. 6. Structure of E series of ␣ 1 /␣ 2 integrin I domain chimeras and ligand binding results. Chimera E contained the ␣C helix, the ␣C-␣6 loop, and the ␣6 helix from the ␣ 1 I domain. Chimeras E1, E2, and E3 contained just the ␣C helix, the ␣C-␣6 loop, or the ␣6 helix from the ␣ 1 I domain, respectively. Chimera E1a contained a point mutation (Y 3 S) within the ␣C helix; chimera E1R contained a reversal of two residues in the ␣C helix of the ␣ 2 I domain (YL 3 LY). Chimeras E2a, E2b, and E2c contain mutations in the ␣C-␣6 loop (E2a and E2b) or at the amino-terminal end of the ␣6 helix (E2c). These mutations are NA 3 GN, D 3 S, and KN 3 EK, respectively. Chimeras E2a1 and E2a2 contain mutations in the ␣C-␣6 loop, N 3 G and A 3 N, respectively. Chimera E2c1 contains a mutation in the amino-terminal end of the ␣6 helix, K 3 E. Each of these chimeras were tested in the collagen I and IV binding assays. The collagen binding data are given as percentages Ϯ S.E. of control (␣ 2 I domain binding to collagen I). Data marked with a lowercase letter a reflect divalent cation-independent binding to collagen. Many of these chimeras were also tested for echovirus 1 binding. Echovirus binding data are given as percentages Ϯ S.E. of control (echovirus 1 binding to the ␣ 2 integrin I domain).

FIG. 7. Binding of integrin I domain proteins to collagens I and IV.
The binding of the ␣ 1 I domain, the ␣ 2 I domain, and chimeric I domain E1 to collagens I and IV was measured in a solid phase binding assay over a range of I domain concentrations from 1.6 to 400 nM. Binding was determined in the presence of either 1 mM EDTA or 2 mM Mg 2ϩ . most impaired collagen I binding activity, only marginally higher than chimera F. Thus it appears that the loss of collagen binding activity seen in chimera F is a result of a combination of effects from the F2 region (LEKAG) of the ␣7 helix as well as the Q 3 R mutation near the carboxyl terminus of the ␣7 helix.
As mentioned earlier, five ␣ 2 integrin I domain monoclonal antibodies recognizing complex conformational epitopes were used as probes in ELISAs to assess the folded state of chimeric I domain molecules. The chimeras were considered to be properly folded if a majority of the antibodies were able to bind. All of the chimeras presented in this study were deemed properly folded with the exception of chimeras A and C in the first round and E2Ј and E2aЈ.
New epitope mapping information for three of the five ␣ 2 integrin I domain monoclonal antibodies was obtained in this study from the analysis of their binding to the ␣ 1/ ␣ 2 integrin chimeras (Table II). These antibodies were 6F1, 12F1, and 5E8. These antibodies had previously been mapped to the same relatively large region (residues 173-259) of the I domain using human/bovine chimeric ␣ 2 I domain proteins (11). In our experiments, 6F1 binding was abrogated by mutation of the sequence TS (residues 210 and 211) at the carboxyl-terminal end of the ␣3 helix (chimera B2b). The binding of 12F1 required the sequence TYKTK (residues 199 -203) within the ␤C-␣3 loop (chimera B1). P1E6 recognition required DQCNHD (residues 268 -273), the carboxyl-terminal half of the ␣5 helix (chimera D2). With the exception of the P1E6-binding site, these regions are all within the range of amino acids previously specified as required for binding using the human/bovine chimeras. Amino acids 268 -273, required for P1E6 binding in this study, are close in the primary structure to the region previously shown to be important for P1E6 recognition (11); it appears that P1E6 binding may require amino acids within both regions. DISCUSSION Although purified ␣ 1 ␤ 1 and ␣ 2 ␤ 1 integrins each bind both collagen I and collagen IV, the ␣ 1 ␤ 1 integrin exhibits a greater affinity for collagen IV than for collagen I, whereas the ␣ 2 ␤ 1 integrin preferentially binds collagen I (18). BIAcore and solid phase binding studies using isolated ␣ subunit I domains from these two integrins demonstrate that the collagen type specificities of the I domains are those of the parent integrins (13). The ␣ 2 ␤ 1 integrin has been shown to be a cell surface receptor for echovirus 1 (8). The ␣ 2 I domain was shown to contain the echovirus 1-binding site (24). Murine ␣ 2 I domain fails to bind the virus (24) and human/murine ␣ 2 I domain chimeras have been used to identify some regions of the human ␣ 2 I domain required for echovirus binding (25).
To determine which regions of the ␣ 1 integrin I domain are responsible for enhanced collagen IV binding activity, and to identify additional regions of the ␣ 2 integrin I domain required for echovirus 1 binding, we undertook a systematic study in-volving the preparation and analysis of human ␣ 1 /␣ 2 integrin I domain chimeras. Initially, a series of six chimeras were constructed that contained primarily ␣ 2 sequence, each with a single internal region that had been replaced with the analogous residues from the ␣ 1 I domain. The design of the initial round of chimeras was undertaken with the aid of the recently published crystal structure of the ␣ 2 integrin I domain (17). For the most part, chimeras in the initial round (Fig. 3, A-F) contained a combination of secondary structural elements from the ␣ 1 I domain, for instance a helix and a loop, as in chimera A. Helices and loops were chosen, as opposed to strands, because of their surface exposure. Two of the chimeras (D and F) contained only a helix and chimera E contained two helices and a loop. Following the identification of chimeras with enhanced collagen IV binding activity or impaired echovirus binding activity, additional chimeras were constructed that had progressively smaller regions of replacement by ␣ 1 sequence. The pattern of chimera design, expression, purification, and analysis was repeated until the chimeras reverted to the ␣ 2 phenotype with regard to ligand binding or until the effect was isolated to a single amino acid.
The only chimera in the first round to lose echovirus 1 binding activity was chimera E. This chimera contained the ␣C helix, the ␣C-␣6 loop and the ␣6 helix from the ␣ 1 I domain. Through iterative rounds of expression and analysis, the loss or severe impairment of echovirus binding was traced through chimera E2 (␣C-␣6 loop) and chimera E2a (residues NA in the loop) and finally to chimera E2a1 (Asn 289 ). The fact that the loss of echovirus 1 binding can be traced through these four chimeras in the E series provides compelling evidence for the involvement of this region and Asn 289 in the interaction of echovirus 1 with the ␣ 2 I domain.
A similar approach has been previously used to map the echovirus 1-binding site on the ␣ 2 I domain (25). However, chimeras in the earlier study were human/murine ␣ 2 integrin I domain chimeras. Murine ␣ 2 integrin I domain fails to support echovirus 1 binding (24). Two regions of the I domain were FIG. 8. Structure of F series of ␣ 1 /␣ 2 integrin I domain chimeras, and ligand binding results. Chimera F contained the ␣7 helix from the ␣ 1 I domain. Chimeras F1, F2, and F3 contained residues from the amino-terminal third (A 3 L), the middle third (LEKAG 3 VTIVK), and the carboxyl-terminal third (Q 3 R) of the ␣7 helix of the ␣ 1 I domain, respectively. Each of these chimeras were tested in the collagen I and IV binding assays. The binding data are given as percentages Ϯ S.E. of control (␣ 2 I domain binding to collagen I). shown to be required for maximal echovirus 1 binding activity. These regions consisted of amino acids 199 -201 in the ␤C-␣3 loop and amino acids 212, 214, and 216 in the ␣3-␣4 loop. In a reciprocal experiment, replacement of the analogous residues within the mouse ␣ 2 I domain with human residues 199 -216 conferred echovirus 1 binding upon the mouse ␣ 2 I domain. This provided unambiguous evidence for the role of this region of the ␣ 2 I domain in echovirus binding. A similar attempt was undertaken in this study involving the ␣C-␣6 loop. Two chimeras were prepared in which portions of the ␣ 1 I domain were replaced with the analogous regions of the ␣ 2 I domain. These chimeras were analogous to E2 and E2a and thus are referred to as E2Ј and E2aЈ. Although neither of these two chimeras bound echovirus, they also failed to bind collagens I and IV, indicating that they may be incorrectly folded. We conclude that Asn 289 is necessary for echovirus 1 binding; it may not be sufficient.
The mouse/human ␣ 2 I domain chimera study did not implicate Asn 289 or the region around the ␣C-␣6 loop in echovirus binding. A comparison of the mouse and human ␣ 2 I domain sequences (25) provides a possible explanation for this. Not only is Asn 289 conserved in these two I domains, but the entire region containing the ␣C helix, the ␣C-␣6 loop, and the ␣6 helix as well as some flanking residues is completely conserved between the human and mouse ␣ 2 I domain structures. Thus it would have been impossible to ascertain the importance of this region for echovirus 1 binding using mouse/human chimeras. Fig. 9 shows the residues involved in echovirus binding by the ␣ 2 integrin I domain in green. Residues implicated in the previous study (25) (residues 199 -201, 212, 214, and 216) and in this study (Asn 289 ) are shown. It has been hypothesized that the ␣ 2 integrin I domain may bind to echovirus 1 within depressions that exist on the viral surface at both the 5-and 2-fold symmetry axes (25,26). The topography of these depressions may explain how residues that are quite distant from each other on the I domain, such as Asn 289 , and residues 199 -201, could be involved in virus binding by the I domain.
Chimeras E1 and E1a bound collagens I and IV in a divalent cation-independent manner. The mutation present in E1a is a point mutation; Tyr 285 of the ␣C helix of the I domain was changed to a serine. A model of a collagen-I domain complex has been proposed (17). In this model, the collagen triple helix has been fitted into a groove near the top of the I domain, in close proximity to the MIDAS site. The ␣C helix and Tyr 285 are also in close proximity to collagen in the model of the collagen-I domain complex. We have previously presented evidence for ligand binding-induced divalent cation displacement from the ␣ 2 I domain (27). In this model, divalent cation is required for initiation of ligand binding, and an intermediate ternary complex is formed containing the I domain, divalent cation, and ligand. The divalent cation is subsequently displaced, leading to the formation of a divalent cation-free I domain-ligand complex. Chimeras E1 and E1a bound collagen in the absence of divalent cations; this may be because they are fixed in a conformation similar to that of the I domain in the metal ion-free ligand complex.
Three of the chimeric ␣ 1 /␣ 2 integrin I domains (B, D, and E) in the first round of chimeras showed enhanced collagen IV binding activity, i.e. they exhibited a gain of function, becoming ␣ 1 I domain-like with respect to collagen IV binding. Chimera B contained the ␤C-␣3 loop and the ␣3 helix from the ␣ 1 I domain. Chimera D contained the ␣5 helix, and chimera E contained the ␣C helix, the ␣C-␣6 loop, and the ␣6 helix from the ␣ 1 I domain. As described above for the echovirus 1 binding studies, iterations of chimera design and analysis were undertaken to determine the minimal amount of ␣ 1 I domain sequence that would still confer the gain of function upon the ␣ 2 I domain, for each of these three regions.
For chimera B, the minimal sequence replacement that still conferred enhanced collagen IV binding activity was residues 205-210 of the ␣ 1 I domain sequence. This region contains the ␣3 helix and is the sequence present in chimera B2. Attempts FIG. 9. Location of residues affecting the ligand binding specificity of the ␣ 2 integrin I domain. Shown are two representations of the ␣ 2 integrin I domain structure (17). In the space-filling model (left) and the main-chain schematic (right) residues affecting the collagen type specificity of the I domain are shown in purple, and residues required for echovirus 1 binding to the I domain are shown in green. The replacement of any of three groups of residues including 206 -211 (␣3 helix), 260 -273 (␣5 helix with some flanking residues), and 294 -295 in the ␣6 helix with the analogous residues from the ␣ 1 I domain results in chimeras with enhanced collagen IV binding activity, an ␣ 1 I domain-like quality. Tyr 285 , in the ␣C helix, when mutated to serine resulted in enhanced collagen IV binding activity and divalent cation-independent binding to both collagens I and IV. Residues 199 -201 and 212-216 have previously been shown to be required for echovirus 1 binding using human/mouse ␣ 2 I domain chimeras (25). Based on this study, Asn 289 is also required for echovirus 1 binding. In the figure on the right, the secondary structural elements are labeled (␣ helices 1, 3-7, and ␣C are shown in red, and ␤ strands A-F are shown in blue).
to further localize the portion of the ␣5 helix necessary for the enhanced collagen IV binding activity in chimera D led to the construction of chimeras D1 and D2. However, neither of these chimeras possessed collagen IV binding activity comparable with that of chimera D, and D1 also had significantly diminished collagen I binding activity. Thus, the enhanced collagen IV binding activity effect of chimera D required the entire ␣5 helix. Significant progress was made, however, at localizing the region within chimera E that was responsible for the enhanced collagen IV binding activity of this protein. Enhanced cationdependent collagen IV binding activity segregated to chimera E2, containing only the ␣C-␣6 loop from the ␣ 1 I domain (residues 288 -294). The effect was further localized in chimera E2c, in which the sequence KN at the amino terminus of the ␣6 helix was mutated to EK, the analogous sequence from the ␣ 1 I domain. Somewhat surprisingly, chimera E2a1 (Asn 289 3 G), also exhibited increased collagen IV binding activity, although its immediate parent in the chimera series E2a (NA 3 GN) did not. These findings clearly identify an important role for residues of the ␣C-␣6 loop and the amino-terminal portion of the ␣6 helix in establishing collagen type specificity.
Chimera F, containing the ␣7 helix from the ␣ 1 I domain, failed to bind either collagen I or IV. It was, however, recognized by all five of the ␣ 2 I domain monoclonal antibodies and was as effective as the wild-type ␣ 2 I domain at binding echovirus. Dissection of F revealed that two regions within the ␣7 helix contributed to the loss of collagen I binding activity. These regions, replaced in chimeras F2 and F3, consist of the amino acids LEKAG (residues 322-326) and Gln 331 of the ␣ 2 I domain. Gln 331 is most critical for collagen I binding by the ␣ 2 I domain. Mutation of this glutamine to arginine (chimera F3) resulted in a 50% reduction in collagen I binding activity.
Chimeras B2, D, and E2c each exhibited enhanced collagen IV binding activity. It is unlikely that amino acid replacements in these regions (␣3 helix, ␣5 helix, and EK in the ␣6 helix) could each exert their effect through a direct contact with a rod-shaped molecule like collagen. It is more likely that the amino acid replacements within these regions induce the same or a highly similar conformational change at the site of collagen binding, leading to the enhanced collagen IV binding activity seen in chimeras B2, D, and E2c.
There are several lines of evidence supporting the existence of multiple I domain conformations revealing significant structural differences, some of which may lead to differences in ligand binding activities. For example, the ␣ M I domain has been crystallized in the presence of both Mg 2ϩ and Mn 2ϩ (28). These divalent cation-dependent differences in conformation of the ␣ M I domain may be responsible for observed differences in the in vitro ligand binding activity of this I domain in the presence of Mg 2ϩ or Mn 2ϩ . Although the identity of the divalent cation has no effect on the binding of the ␣ M I domain to neutrophil inhibitory factor (29), binding to iC3b in the presence of Mn 2ϩ is twice that observed in the presence of physiologic levels of Mg 2ϩ and Ca 2ϩ (30). Similarly, the extent of binding of the ␣ 2 I domain to collagen is identical in the presence of Mg 2ϩ or Mn 2ϩ (12), yet laminin binding is enhanced in the presence of Mn 2ϩ as opposed to Mg 2ϩ (14). Thus it is possible, perhaps even likely, that the enhancement of collagen IV binding activity observed with chimeras B2, D, and E2c is due to a conformational change in the I domain that is a direct result of the replacement of residues in the ␣3, ␣5, and ␣6 helices rather than to the alteration of critical residues that directly interact with collagen.
It is, however, also possible that some of the amino acids that were replaced in these chimeras do come into contact with collagen. In the model of the collagen-I domain complex (17), the collagen molecule is very near the carboxyl terminus of the ␣6 helix. It is possible, therefore, that residues 294 and 295 ( Fig. 9), those replaced in chimera E2c, exert their effect through a direct contact with collagen.
In addition to the amino acid replacements in the ␣3, ␣5, and ␣6 helices, mutation of Tyr 285 to either a serine (chimeras E1 and E1a) or to leucine (chimera E1R), also resulted in enhanced collagen IV binding activity of the ␣ 2 I domain. The Tyr 3 Ser mutation was the only change in chimera E1a. The presence or absence of this tyrosine in the ␣ 1 I domain, the ␣ 2 I domain, and chimeras E1, E1a and E1R correlates perfectly either with low or enhanced collagen IV binding activity. This tyrosine is present within the ␣C helix of the ␣ 2 I domain and is positioned near the groove at the top of the I domain predicted to be the collagen-binding site (17). Fig. 10 shows a schematic of the model of the ␣ 2 I domain-collagen complex, looking directly down the collagen molecule. Side chains are included, and Tyr 285 is shown in a space-filling representation (yellow). The tyrosine side chain and in particular the hydroxyl group come into extremely close contact with collagen. Our data predict that Tyr 285 performs a "gatekeeper" function, composing part of the collagen-binding site on the ␣ 2 I domain that leads to preferential binding of collagen I over collagen IV. Additional evidence for such a role comes from analysis of chimera E1R, in which residues YL in the ␣C helix were reversed to LY. Chimera E1R showed enhanced binding to collagen IV, while retaining the requirement for divalent cations. Thus, it is possible that replacement of amino acids in chimeras B2, D, and E2c leads to enhanced collagen IV binding activity by causing an alteration in the positioning of the side chain of Tyr 285 .
In summary, we have identified critical amino acid residues and structural motifs that determine the collagen type specificity of the ␣ 1 and ␣ 2 integrin I domains. It appears likely that  (17). Here Tyr 285 is shown in a spacefilling representation (yellow). From this study, mutation of Tyr 285 to serine (chimeras E1 and E1a) or leucine (chimera E1R) results in I domains with enhanced collagen IV binding activity while retaining strong collagen I binding activity. In addition, chimeras E1 and E1a, each containing the Y 3 S mutation, bound collagens I and IV in a divalent cation-independent manner. collagen type specificity is established not only by residues that interact directly with collagen but also by more distant residues and motifs that interact with other structural elements of the I domain to indirectly, presumably by subtle alteration of the topography of the collagen-binding site, control selectivity. We have furthermore identified determinants that are essential for the echovirus 1 binding activity of the ␣ 2 integrin I domain. It is noteworthy that collagen type specificity and echovirus binding activity are determined by distinctly different structural features. It is also notable that several of the motifs that determine collagen type specificity presumably represent relatively large interactive surfaces or faces because they cannot be successfully reduced to single amino acid residues prior to reversion by the chimera to the parental I domain phenotype. This finding indicates the importance, indeed the necessity, of approaches such as that described here rather than the more traditional alanine mutagenesis approach.