Identification of Motifs for Cell Adhesion within the Repeated Domains of Transforming Growth Factor-β-induced Gene,βig-h3 *

βig-h3 is a transforming growth factor-β-inducible cell adhesion molecule that has four characteristic homologous repeated domains. We made recombinant βig-h3 proteins, which were highly active in mediating human corneal epithelial (HCE) cell adhesion and spreading. The 2nd and the 4th repeated domains were sufficient to mediate HCE cell adhesion. A sequence analysis showed that aspartic acid (Asp) and isoleucine (Ile) of the 2nd and the 4th domains are highly conserved in many fasciclin 1 homologous (fas-1) domains. Substitution mutational study identified these two amino acids are essential for cell adhesion. Synthetic peptides containing Asp and Ile, NKDIL and EPDIM derived from the 2nd and the 4th domains, respectively, almost completely blocked cell adhesion mediated by not only wild type βig-h3 but also each of the 2nd and the 4th domains. These peptides alone were fully active in mediating cell adhesion. In addition, we demonstrated the functional receptor for βig-h3 is α3β1integrin. These results, therefore, establish the essential motifs within the 2nd and the 4th domains of βig-h3, which interact with α3β1 integrin to mediate HCE cell adhesion to βig-h3 and suggest that other proteins containing Asp-Ile in their fas-1 domains could possibly function as cell adhesion molecules.

␤ig-h3 is a transforming growth factor-␤-inducible cell adhesion molecule that has four characteristic homologous repeated domains. We made recombinant ␤ig-h3 proteins, which were highly active in mediating human corneal epithelial (HCE) cell adhesion and spreading. The 2nd and the 4th repeated domains were sufficient to mediate HCE cell adhesion. A sequence analysis showed that aspartic acid (Asp) and isoleucine (Ile) of the 2nd and the 4th domains are highly conserved in many fasciclin 1 homologous (fas-1) domains. Substitution mutational study identified these two amino acids are essential for cell adhesion. Synthetic peptides containing Asp and Ile, NKDIL and EPDIM derived from the 2nd and the 4th domains, respectively, almost completely blocked cell adhesion mediated by not only wild type ␤ig-h3 but also each of the 2nd and the 4th domains. These peptides alone were fully active in mediating cell adhesion. In addition, we demonstrated the functional receptor for ␤ig-h3 is ␣ 3 ␤ 1 integrin. These results, therefore, establish the essential motifs within the 2nd and the 4th domains of ␤ig-h3, which interact with ␣ 3 ␤ 1 integrin to mediate HCE cell adhesion to ␤ig-h3 and suggest that other proteins containing Asp-Ile in their fas-1 domains could possibly function as cell adhesion molecules.
␤ig-h3 is an extracellular matrix protein that can be induced by transforming growth factor-␤ in several cell types, including human melanoma cells, mammary epithelial cells, keratinocytes, and lung fibroblasts (1). Several studies suggest ␤ig-h3 is involved in cell growth (1), cell differentiation (2,3), wound healing (4), and cell adhesion (5,6), although the underlying mechanisms for these effects are still unclear. In addition, some ␤ig-h3 missense mutations were identified in families affected with human autosomal dominant corneal dystrophies (7). Yet the exact role of mutant ␤ig-h3 proteins in developing corneal dystrophies is unidentified.
␤ig-h3 contains an RGD motif and four internal repeated domains, which have highly conserved sequences found in some secretory and membrane proteins of several species including mammals, insects, sea urchins, plants, yeast, and bac-teria (8). These proteins include periostin, fasciclin I, sea urchin HLC-2, alga Algal-CAM, and mycobacterium MPB70. The homologous domain (designated fas-1) 1 of these proteins is 110 -140 amino acids long and is characterized by two highly conserved stretches of about 10 amino acids (H1 and H2). Some proteins including ␤ig-h3, periostin, and fasciclin I have four sets of fas-1, HLC-2 has two sets, and MPB70 contains only one set. Although their biological functions are poorly characterized, some of them have been reported to function as cell adhesion molecules. ␤ig-h3, periostin, and fasciclin I have been reported to mediate attachment of fibroblasts (5), osteoblasts (9), and neuronal cells (10), respectively. Algal-CAM is also known to be a cell adhesion molecule in embryos of the alga Volvox (11).
According to a recent report (6), ␤ig-h3 enhanced the spreading of fibroblasts via integrin ␣ 1 ␤ 1 , and its RGD motif was not necessary for mediating cell spreading. They also reported that the conserved H1 and H2 peptides were not sufficient to affect ␤ig-h3-mediated cell adhesion. This suggests that the essential amino acids for the cell adhesion activity of ␤ig-h3 may exist elsewhere rather than H1 and H2 regions. A computer search based on homologies not only among the repeated fas-1 domains of ␤ig-h3 but also among fas-1 domains of other proteins revealed that there are a few highly conserved amino acids in addition to H1 and H2. Particularly, we focused on aspartic acid near H2 because aspartic acid is known to be an essential amino acid for interaction with several integrins (12). Based on the above findings, we hypothesized that a highly conserved sequence including aspartic acid may support ␤ig-h3-mediated cell adhesion. To prove our hypothesis, we have generated wild type and several mutant recombinant ␤ig-h3 proteins. By using these proteins, we found that ␤ig-h3 mediates human corneal epithelial (HCE) cell adhesion through ␣ 3 ␤ 1 integrin and that each of the 2nd fas-1 domain and the 4th fas-1 domain of ␤ig-h3 is sufficient to mediate cell adhesion via ␣ 3 ␤ 1 integrin. Then, we demonstrated that two conserved amino acids, Asp and Ile, near H2 in both repeated domains are essential and that synthetic peptides including Asp and Ile from the 2nd and the 4th domains are sufficient to mediate cell adhesion via ␣ 3 ␤ 1 integrin. These results, therefore, establish the essential amino acid residues within the fas-1 domains of ␤ig-h3, which interact with ␣ 3 ␤ 1 integrin to support ␤ig-h3-mediated HCE cell adhesion.
Cell Adhesion and Spreading Assay-The cell adhesion assay was performed as described previously (14). Briefly, 96-well microculture plates (Falcon, Becton-Dickinson, Mountain View, CA) were incubated with recombinant ␤ig-h3 proteins or other extracellular matrix proteins at 37°C for 1 h and then blocked with PBS containing 0.2% BSA for 1 h at 37°C. The coated extracellular matrix proteins used were as follows: human plasma vitronectin (Promega), purified human plasma fibronectin (pFN), chicken collagen types I and II (Chemicon International Inc., Temecula, CA), bovine collagen types IV and VI (Chemicon), mouse laminin (Chemicon), and bovine serum albumin (BSA) (Sigma). Cells were trypsinized and suspended in the culture media at a density of 2 ϫ 10 5 cells/ml, and 0.1 ml of the cell suspension was then added to each well of the plates. Cell attachment was analyzed as follows. After incubation for 1 h at 37°C, unattached cells were removed by rinsing twice with PBS. Attached cells were incubated for 1 h at 37°C in 50 mM citrate buffer, pH 5.0, containing 3.75 mM p-nitrophenyl-N-acetyl ␤-Dglucosaminide (hexosaminidase substrate) and 0.25% Triton X-100. Enzyme activity was blocked by the addition of 50 mM glycine buffer, pH 10.4, containing 5 mM EDTA, and the absorbance was measured at 405 nm in a Multiscan MCC/340 microplate reader (Titertek Instruments, Inc., Huntsville, AL). To determine cell area, 4ϫ10 4 cells were applied to substrates in 48-well culture plates. The attached cells were fixed with 8% glutaraldehyde (Sigma) and then stained with 0.25% Crystal Violet (Sigma) in 20% methanol (w/v). Cell area was measured using Image-Pro plus software (Media Cybernetics, Silver Spring, MD). Experiments were repeated in triplicate with 200 or 300 measurements per site for each experiment. Data are reported as the mean area at specific time points Ϯ S.E.
Inhibition Assay-Various reagents and synthetic peptides were examined for their ability to prevent cells from adhering to the prepared substrata. Synthetic peptides were synthesized on an automated multiple peptide synthesizer (PE/ABD 433, PE Corp., Norwalk, CT) using standard solid phase procedures. Peptides were purified by reverse phase high performance liquid chromatography. Cell adhesion assay was done as described above. To analyze the divalent cation sensitivity of ␤ig-h3-mediated adhesion, cells were suspended at 2 ϫ 10 5 cells/ml in Hepes-buffered saline (HBS), 150 mM NaCl, 25 mM Hepes, pH 7.4, and incubated at 37°C for 30 min. They were then washed twice in HBS and resuspended in the same buffer. Aliquots of cells (50 l) were then added to the microculture plate wells and incubated with 50-l aliquots of HBS containing twice the final concentration of divalent cations (MnCl 2 , MgCl 2 , or CaCl 2 ) for 30 min at 37°C in a humidified atmosphere of 5% CO 2 . They were then plated on ligand-coated dishes to perform the adhesion assays, as described above. To identify the receptor for ␤ig-h3, monoclonal antibodies to different types of integrins (Chemicon) were preincubated individually with HCE in 0.05 ml of incubation solution (2 ϫ 10 5 cells/ml) at 37°C for 30 min. The preincubated cells were transferred onto plates precoated with ␤ig-h3 proteins and then incubated further for 1 h at 37°C. Attached cells were then quantified as described above.
Flow Cytometry-For flow cytometry analysis, cells at confluence were detached by gentle treatment with 0.25% trypsin, 0.05% EDTA in PBS, washed, and incubated with the antibodies for 1 h at 4°C. Cells were then incubated with 10 g/ml affinity purified fluorescein-labeled secondary antibodies for 1 h at 4°C and analyzed on the flow cytometer FACSCalibur system (Becton Dickinson, San Jose, CA) equipped with a 5-watt argon laser at 488 nm.
Immunoprecipitation-HCE cells or ␣ 3 -transfected HCE cells were solubilized in 200 mM n-octyl ␤-D-glucopyranoside, 1 mM phenylmethylsulfonyl fluoride, 100 mM Tris-HCl, pH 7.4. Immunoprecipitations were carried out by overnight incubation at 4°C of the immunoadsorbents (antibodies adsorbed onto protein A-Sepharose (Amersham Pharmacia Biotech) with samples of cell lysates. Precipitated proteins were separated on 10% SDS-polyacrylamide gel. After separation of precipitated proteins by SDS-PAGE and transfer to a nitrocellulose membrane (Schleicher & Schuell), blots were incubated for 2 h with either anti-␣ 3 or anti-␤ig-h3 polyclonal antibodies, then detected using horseradish peroxidase-conjugated anti-rabbit IgG antibodies (Sigma), followed by enhanced chemiluminescence (ECL) system (NEN Life Science Products).
Construction of ␣ 3 cDNA Expression Vector and Transfection-Fulllength cDNA for the human ␣ 3 subunit (clone 3.10) was purchased from the American Type Culture Collection. A fragment (3.47 kilobases) containing the entire cDNA excised by digestion with XbaI/SalI was cloned into the EcoRV site of the mammalian expression vector pcDNA 3.1ϩ (Invitrogen, Carlsbad, CA). The ␣ 3 integrin expression plasmid DNA (1 g) was transfected into HCE cells using the LipofectAMINE (Life Technologies, Inc.).

␤ig-h3 Supports Cell Adhesion and Spreading
Independent of the RGD Motif-For cell adhesion assay, we used two recombinant ␤ig-h3 proteins that have been described previously (3). We changed the nomenclature from His␤-c and His␤-d as used in the previous paper (3) to ␤igh3-WT and ␤igh3-⌬RGD, respectively (Fig. 1A). ␤ig-h3 was previously demonstrated to support the adhesion and spreading of fibroblasts (5,6), and the RGD motif was proposed not to be necessary for such activity (6). The numbers and surface areas of HCE cells that adhered to ␤igh3-WT were clearly greater than those attached to albumin and were comparable to those of cells that adhered to fibronectin (Fig. 1B). The cell adhesion and spreading activities of ␤ig-h3 were concentration-dependent ( Fig. 1, C and D). Similar results were also obtained with Chinese hamster ovary cells (data not shown). As expected, ␤igh3-⌬RGD lacking the RGD motif was almost equally effective at supporting cell adhesion and spreading (Fig. 2, A and B). These results confirm that ␤ig-h3 supports cell adhesion and spreading independent of the RGD motif.
Cell Adhesion to ␤ig-h3 Is Dependent on Integrin ␣ 3 ␤ 1 and Divalent Cations-To identify the nature of the cell surface receptor for ␤ig-h3, several reagents were used. Cell adhesion to ␤ig-h3 was significantly inhibited by ␤ig-h3 itself, RGD peptide, and EDTA, and it was partially inhibited by fibronectin and EGTA but not inhibited by RGE peptide. Cell adhesion to fibronectin was also significantly inhibited by fibronectin itself, RGD peptide, and EDTA and partially inhibited by ␤ig-h3 and EGTA but not by RGE peptide (Fig. 3A). Then we examined the effects of Mn 2ϩ , Mg 2ϩ , and Ca 2ϩ on ␤ig-h3mediated cell adhesion. Cell adhesion to ␤ig-h3 was strongly promoted by Mn 2ϩ , and to a lesser extent by Mg 2ϩ , but only marginally by Ca 2ϩ (Fig. 3B). These results suggest that the cell surface receptor for ␤ig-h3 could be one of the RGD-dependent integrins, which require divalent cations for interaction with ligands.
To identify the ␤ig-h3 receptor, the effects of function-blocking monoclonal antibodies to integrin subunits were examined on the adhesion of HCE cells to the surface coated with ␤ig-h3. Adhesion to the ␤ig-h3-coated surface was specifically inhibited by antibody to ␣ 3 subunit but not by antibodies to other ␣ subunits (Fig. 3C). Because the integrin ␣ 3 subunit is known to couple with the integrin ␤ 1 subunit, anti-␤ 1 antibody was also expected to inhibit cell adhesion to ␤ig-h3. As expected, anti-␤ 1 antibody significantly blocked cell adhesion (Fig. 3C). To determine whether ␤ig-h3 interacts with ␣ 3 integrin, we carried out co-immunoprecipitation assays. Immunoblotting with anti-␤ig-h3 antiserum showed that ␤ig-h3 was detected in immunoprecipitates formed by anti-␣ 3 integrin antibody (Fig. 3D, lanes  2-4). Conversely, ␣ 3 integrin was also detected in immunoprecipitates formed by anti-␤ig-h3 antiserum (Fig. 3D, lanes 6 -8). ␣ 3 integrin was detected in all four cell lysates (Fig. 3D, lanes 1-4) among which ␣ 3 overexpressed cell lysate showed the highest amount of ␣ 3 integrin in immune complex. Because the basal expression level of ␤ig-h3 in cell lysate is very low, it was barely detected in immune complex precipitated by anti-␤ig-h3 antiserum (Fig. 3D, lane 5). However, when we added recombinant ␤ig-h3 protein to the culture medium (Fig. 3D, lanes 6 and 8) or to cell lysate (Fig. 3D, lane 7), it was detected in immunoprecipitates. Nothing was detected in immunoprecipitates with nonimmune rabbit serum (data not shown). To determine which integrins are expressed in HCE cell, we performed fluorescence-activated cell sorter analysis. Fig. 4 shows that all the integrins tested were detected on the HCE cell surface although their expression levels varied. The expression level of ␣ 3 integrin was relatively high, whereas that of ␣ 5 ␤ 1 was low. All the others had similar expression levels. Taken together, these results suggest integrin ␣ 3 ␤ 1 is a specific functional receptor for ␤ig-h3 in HCE cells.
Each of the 2nd Fas-1 Domain and the 4th Fas-1 Domain Is Sufficient to Mediate Cell Adhesion-In an attempt to identify essential amino acid residues conferring cell adhesion activity of ␤ig-h3, we first tested whether each repeated domain is capable of mediating cell adhesion. We made four recombinant proteins corresponding to each repeated domain (Fig. 5, A and  B) and tested their cell adhesion activities. We found that the 2nd fas-1 domain and the 4th fas-1 domain were equally active compared with the wild type ␤ig-h3, whereas the 1st fas-1 domain was moderately active and the 3rd fas-1 domain was very weakly active (Fig. 5C). Both 2nd fas-1 and 4th fas-1 domain-mediated cell adhesions were almost blocked by antibodies to ␣ 3 and ␤ 1 integrin subunits (Fig. 5D) suggesting that both 2nd fas-1 and 4th fas-1 domains have essential amino acid residues for interacting with integrin. These results also support that neither H1 nor H2 is mediating cell adhesion activity of ␤ig-h3 because the 1st and the 3rd domains are not active in cell adhesion although they have H1 or H2.
Two Conserved Amino Acids, Aspartic Acid and Isoleucine, Are Essential for Cell Adhesion-To identify cell adhesion motifs within the 2nd and the 4th domains, we performed a computer search using Prodom Release 99.2 based on homologies not only among the repeated fas-1 domains of ␤ig-h3 but also among fas-1 domains of other proteins. In many fas-1 domains including the 2nd and the 4th fas-1 domains of ␤ig-h3, two amino acid residues, aspartic acid and isoleucine near H2, are highly conserved, whereas in some cases including the 1st fas-1 domain of ␤ig-h3, only aspartic acid is conserved (Fig. 6).
Although not shown here, some fas-1 domains such as the 3rd fas-1 domain of ␤ig-h3 do not have aspartic acid near H2. In order to examine the sequence containing aspartic acid that is needed for cell adhesion, we generated mutated 4th fas-1 proteins of ␤ig-h3 where each 616 proline, 617 aspartic acid, and 618 isoleucine was replaced with serine, alanine, and serine, respectively (Fig. 7, A and B). D617A (␤igh3 D-IV-PaI) and I618S (␤igh3 D-IV-PDs) mutations significantly blocked cell adhesion whereas P616S (␤igh3 D-IV-sDI) mutation did not affect cell adhesion. Consequently, three amino acids mutation, P616S/D617A/I618S (␤ig-h3 D-IV-sas) also blocked cell adhesion (Fig. 7C). These results support our hypothesis that the 617 aspartic acid is essential for cell adhesion and indicate that the 618 isoleucine is also important for cell adhesion.
Synthetic Peptides, NKDIL and EPDIM, from the 2nd Fas-1 and the 4th Fas-1 Domains Are Sufficient to Mediate HCE Cell Adhesion via ␣ 3 ␤ 1 Integrin-To confirm further aspartic acid and isoleucine are essential for cell adhesion, four synthetic peptides were generated. As shown in Fig. 8A, the first three peptides, KADHH (amino acids 219 -223), NKDIL (amino acids 354 -358), and EPDIM (amino acids 615-619), correspond to each conserved sequence of the 1st, the 2nd, and the 4th fas-1 domains of ␤ig-h3, respectively. The last one, DEMPI is from the 4th fas-1 but is scrambled and used as a control peptide. We tested whether these four peptides could inhibit ␤ig-h3-mediated cell adhesion. As shown in Fig. 8B, NKDIL from the 2nd fas-1 and EPDIM from the 4th fas-1 were capable of blocking cell adhesion to ␤igh3-WT, whereas KADHH from the 1st fas-1 weakly inhibited cell adhesion, and control peptide DEMPI did not affect cell adhesion. Similar results were obtained when we used the 2nd fas-1 and the 4th fas-1 proteins as cell substrata (Fig. 8C). Then we tested whether each peptide itself is capable of mediating cell adhesion. Several different concentrations of each peptide were used as cell substrata and tested for cell adhesion activity. As shown in Fig. 9A, NKDIL and EPDIM were capable of mediating cell adhesion in a dose-dependent manner. KADHH was also capable of mediating cell adhesion in a dose-dependent manner, but the activities were relatively weak. The control peptide was not active in cell adhesion. In the next experiment, we examined whether peptide-mediated cell adhesion was also mediated via ␣ 3 ␤ 1 integrin. Fig. 9B showed that cell adhesion to NKDIL or EPDIL was blocked by antibodies to ␣ 3 and ␤ 1 integrin subunits. These results suggest that the conserved aspartic acid and isoleucine in the 2nd fas-1 and the 4th fas-1 domains are essential for ␤ig-h3-mediated cell adhesion through ␣ 3 ␤ 1 integrin. To examine whether inhibition of cell adhesion by these two peptides was specific to ␤ig-h3, we tested the effects of peptides on cell adhesion to other adhesion molecules. As is shown in Fig. 10A, NKDIL and EPDIM efficiently blocked cell adhesion not only to ␤ig-h3 but also to laminin, whereas they moderately inhibited cell adhesion to fibronectin and did not affect cell adhesion at all to collagen type I, type II, and vitronectin. These inhibitory effects were dose-dependent (Fig. 10B). These results suggest that NKDIL and EPDIM specifically compete with ␣ 3 ␤ 1 integrininteracting molecules.

DISCUSSION
Although ␤ig-h3 has been considered to promote cell adhesion and spreading, the interacting cell receptor and the specific motifs of ␤ig-h3 for cell adhesion have not been characterized. In this report, we identified that the functional receptor for ␤ig-h3 is ␣ 3 ␤ 1 integrin and the sequences, NKDIL of the 2nd and EPDIM of the 4th fas-1 domains, are active sites and sufficient to induce cell adhesion through ␣ 3 ␤ 1 integrin. In addition, aspartic acid and isoleucine turned out to be the essential amino acid residues of these motifs.
␤ig-h3 was first identified by differential screening of a cDNA library made from A549 human lung adenocarcinoma cells treated with transforming growth factor-␤ (16). Its cell adhesion activity was first reported with human dermal fibroblasts (5) and then with chondrocytes, peritoneal fibroblasts, and human MRC5 fibroblasts (6). Because ␤ig-h3 has an RGD motif at the carboxyl terminus, ␤ig-h3 was thought to mediate cell adhesion through its RGD motif. Ohno et al. (6), however, reported that the RGD motif at the carboxyl terminus of ␤ig-h3 was not necessary for enhancing the spreading of chondrocytes. This result was predictable because the RGD motif was not present in the mature ␤ig-h3 protein as a result of carboxylterminal processing, and the mature form was able to inhibit cell adhesion when it was added to the culture medium (1). Our result also supports the RGD motif is not necessary for mediating cell adhesion activity of ␤ig-h3.
Several efforts were made to identify a cell surface receptor for ␤ig-h3. The fact that ␤ig-h3-mediated cell adhesion was blocked by an RGD peptide and EDTA suggests that the surface receptor for ␤ig-h3 could be one of RGD-dependent integrins of which activity requires divalent cations such as Mn 2ϩ and Mg 2ϩ . The result of assay using function-blocking antibodies to ␣ 1 , ␣ 2 , ␣ 3 , ␣ 4 , ␣ 5 , ␣ 6 , ␣ v , and ␤ 1 integrins suggests that the specific integrin interacting with ␤ig-h3 is ␣ 3 ␤ 1 integrin, which has been known to belong to the RGD-and divalent cation-dependent integrins (12). Indeed, we showed that ␣ 3 integrin was co-immunoprecipitated by anti-␤ig-h3 antiserum, and conversely, ␤ig-h3 was also co-immunoprecipitated by anti-␣ 3 integrin antibody indicating that ␣ 3 ␤ 1 integrin is a specific functional receptor for ␤ig-h3 in HCE cells.
fas-1 domain is found in several proteins including ␤ig-h3, periostin, fasciclin, HLC-2, and algal-CAM, all of which are known as cell adhesion molecules, but they have different numbers of fas-1 domain (8). It suggests that ␤ig-h3 may not require all four fas-1 domains to mediate cell adhesion and even a single fas-1 domain could mediate cell adhesion. This hypothesis was proved by demonstrating that each of the 2nd and the 4th fas-1 domains was sufficient for cell adhesion activity. However, why were the 1st and the 3rd fas-1 domains not active in mediating cell adhesion? Complete lack of cell adhesion activity of the 3rd fas-1 domain may be due to its less homology with other three domains. Particularly, the region around H2 homologous sequence of the 3rd fas-1 domain is not well conserved. In contrast, although the 1st fas-1 domain is relatively highly homologous with the 2nd fas-1 and the 4th fas-1 domains, it is not as active as the 2nd and the 4th domains in mediating cell adhesion. These findings suggest that cell adhesion motifs in the 2nd and the 4th domains might be lacking or altered in the 1st domain and the 3rd domain. To answer this question, we first analyzed sequences of each domain of ␤ig-h3 and several other fas-1 domains looking for conserved amino acids rather than H1 and H2. In particular, we focused on aspartic acid, which is known to be essential for interacting with integrins. A sequence analysis uncovered that aspartic acid and isoleucine near H2 are highly conserved in many fas-1 domains. It is noteworthy that these two amino acids are not found in the 3rd domain and isoleucine is replaced by histidine in the 1st domain. In fact, mutation of either aspartic acid or isoleucine almost completely blocked the 4th fas-1 domain-mediated cell adhesion. In addition, synthetic peptides, NKDIL and EPDIM from the 2nd and 4th domains, respectively, were efficiently able to block cell adhesion mediated not only by each domain but also by wild type ␤ig-h3, whereas the synthetic peptide KADHH derived from the 1st domain was less efficient to block cell adhesion. These results indicate that both aspartic acid and isoleucine are required for cell adhesion and also give us an answer why the 1st domain is weakly active and the 3rd domain is not active in mediating cell adhesion.
Synthetic peptides, NKDIL and EPDIM, themselves were found to be good substrates for cell adhesion. Like wild type ␤ig-h3, cell adhesion mediated by the 2nd domain, the 4th domain, and two synthetic peptides were also specifically FIG. 3. Identification of HCE cell surface receptor for ␤ig-h3. A, plastic culture dishes were coated with 10 g/ml of each protein, i.e. pFN, ␤igh3-WT, or ␤igh3-⌬RGD. HCE cells were preincubated for 30 min in medium in the absence or presence of 5 mM EDTA, 5 mM EGTA, 100 g/ml ␤igh3-WT, 100 g/ml ␤igh3-⌬RGD, 1 mM GRGDSP, 1 mM GRGESP, or 100 g/ml pFN in tubes. B, HCE cells were preincubated for 30 min in Hepes buffer, pH 7.4, containing 150 mM NaCl in the absence or presence of 5 mM Ca 2ϩ , Mg 2ϩ , or Mn 2ϩ and then transferred to ␤igh3-WT-coated (10 g/ml) dishes for cell adhesion assay. C, HCE cells were preincubated with the following function-blocking monoclonal antibodies to integrin subunits at a concentration of 5 g/ml for 30 min at 37°C and then added to the precoated wells with 10 g/ml ␤igh3-WT: ␣ 1 , anti-integrin ␣ 1 subunit antibody (FB12); ␣ 2 , anti-integrin ␣ 2 subunit antibody (P1E6); ␣ 3 , anti-integrin ␣ 3 subunit antibody (P1B5); ␣ 4 , anti-integrin ␣ 4 subunit antibody (P1H4); ␣ 5 , anti-integrin ␣ 5 subunit antibody (P1D6); ␣ 6 , anti-integrin ␣ 6 subunit antibody (CLB701); ␣ v , anti-integrin ␣v subunit antibody (P3G8); ␤ 1 , anti-integrin ␤ 1 subunit antibody (6C6). After 1 h incubation, cells attached to the substrates were quantified by hexosaminidase assay as described under "Experimental Procedures." The values are expressed as percentages of the number of cells adhering in the absence of monoclonal antibodies. Each column represents the mean of triplicate assays. D, co-immunoprecipitation of ␣ 3 integrin and ␤ig-h3. HCE cells and transfectants expressing ␣ 3 were immunoprecipitated with antibodies to either ␣ 3 integrin or ␤ig-h3. The immunoprecipitated proteins were separated on 10% SDS-polyacrylamide gels, transferred, and immunoblotted with antibodies to either ␤ig-h3 or ␣ 3 integrin as described under "Experimental Procedures." HCE cell extract (lanes 1 and 5), cell extract from HCE cells treated with 2 g/ml of ␤ig-h3 proteins in the culture medium (lanes 2 and 6), HCE cell extracts added with 2 g/ml of ␤ig-h3 proteins (lanes 3 and 7), and cell extract from HCE cells expressing ␣ 3 treated with 2 g/ml of ␤ig-h3 proteins in the culture medium (lanes 4 and 8) were analyzed. blocked by antibodies to ␣ 3 and ␤ 1 subunits. This implies that ␤ig-h3 interacts with ␣ 3 ␤ 1 integrin via two major motifs, where one resides in the 2nd fas-1 domain and the other resides in the 4th fas-1 domain. This finding is different from what Ohno et al. (6) reported. They reported that the cell surface receptor for ␤ig-h3 was ␣ 1 ␤ 1 integrin in MRC5 fibroblasts. This discrepancy may not be due to simple differences in surface integrin profiles because we found that both HCE cells and MRC5 cells have not only ␣ 3 integrin but also ␣ 1 integrin on their cell surfaces. Actually, MRC5 has more ␣ 3 integrin than ␣ 1 integrin (data not shown). Our experiments with MRC5 fibroblasts revealed that their adhesion to ␤ig-h3 was incompletely blocked by antibody to ␣ 1 integrin but, interestingly, was almost completely blocked by ␣ v integrin antibody which Ohno et al. (6) have not tried in their experiment (data not shown). Furthermore, all mutant forms of the 4th fas-1 domain, which failed to mediate corneal epithelial cell adhesion in our experiments, still retain cell adhesion activities for MRC5 fibroblasts (data not shown). These results strongly suggest that ␤ig-h3 can mediate cell adhesion through different integrins depending on cell types, and that its interacting domains could be different.
A number of studies have defined multiple ligands for ␣ 3 ␤ 1 integrin, including laminin (17), certain types of collagen (18), fibronectin (18), and nidogen (19). Although there are some conflicting reports that some of these proteins do not support ␣ 3 ␤ 1 -mediated cell adhesion (19,20) despite apparent P1B5 blocking effects on some of them, ␣ 3 ␤ 1 integrin is considered to respond to a broad spectrum of extracellular ligands (21). There seems to be no conserved binding motif for ␣ 3 ␤ 1 integrin because no apparent sequence homology is observed among active peptides from thrombospondin (22), laminin (23), and type IV collagen (24), which have been suggested to interact with ␣ 3 ␤ 1 . Our peptides from ␤ig-h3 also do not share any sequence homology with the active peptides mentioned above. However, it is interesting to note that the ␣ 3 ␤ 1 -interacting peptide corresponding to ␣ 1 (IV)-(531-543) has two Asp residues, both of which are important for cell adhesion activity and that the first Asp is flanked by Leu (24,25). This Asp-Leu residue is reminiscent of our peptides, NKDIL and EPDIM, where Ile replaces Leu. Both Ile and Leu are hydrophobic and have bulky side chains which, together with Asp, are known to be important for interacting with integrins (12). Further investigations, however, are required to identify and prove ␣ 3 ␤ 1 integrin-interacting motifs more precisely from several ligands.
Transfection of ␤ig-h3 expression plasmids into Chinese hamster ovary cells led to marked decreases in cell growth and the ability of these cells to form tumors in nude mice (1). We previously reported ␤ig-h3 could inhibit osteoblast differentiation (3), and its expression is down-regulated in bone marrow stem cells treated with dexamethasone (2). Other studies imply The composition of amino acids of control peptide is same as peptide from the 4th fas-1 domain, but its order is changed. B, inhibition of HCE cell adhesion by synthetic peptides. Plastic culture dishes were coated with 10 g/ml BSA or 10 g/ml ␤igh3-WT. HCE cells were preincubated for 30 min in medium in the absence or presence of 100 M each synthetic peptide (DEMPI, KADHH, NKDIL, and EPDIM). C, inhibition of HCE cell adhesion to ␤igh3 D-II or ␤igh3 D-IV proteins by synthetic ␤ig-h3 peptides. Plastic culture dishes were coated with 10 g/ml each protein (BSA, ␤igh3 D-II, and ␤igh3 D-IV). HCE cells were preincubated for 30 min in medium in the absence or presence of 100 M each synthetic peptide (DEMPI, KADHH, NKDIL, and EPDIM). After 1 h incubation, cells attached to the substrates were quantified as described under "Experimental Procedures." a role for ␤ig-h3 in wound healing in corneal and vascular tissues (4,26). A role for disturbed morphogenesis has been assigned to ␤ig-h3 missense mutations in families affected with human autosomal dominant corneal dystrophies (7). Given that ␤ig-h3 is highly induced by transforming growth factor-␤ in several cells and that it functions as a cell adhesion molecule, together with all above reports, we suggest that ␤ig-h3 may play an important role in the regulation of morphogenesis and the maintenance of several tissues in normal and pathological conditions.
In conclusion, we have demonstrated that two motifs, NK-DIL and EPDIM within the 2nd and the 4th fas-1 domains of ␤ig-h3, are sufficient to mediate human corneal epithelial cell adhesion through ␣ 3 ␤ 1 integrin. We have also identified that two amino acid residues, Asp and Ile, in the context of motifs are essential for cell adhesion activity. These results, therefore, establish the mechanism of ␤ig-h3-mediated cell adhesion and suggest that other proteins containing Asp-Ile near H2 in their fas-1 domains could function as cell adhesion molecules. FIG. 9. Adhesion of HCE cells onto surfaces coated with synthetic peptides. A, HCE cell adhesion to synthetic ␤ig-h3 peptides. 96-Well microculture plates were coated with the indicated concentrations of each synthetic peptide and incubated with HCE cells at 37°C for 1 h. After incubation, cells attached to the substrates were quantified by hexosaminidase assay as described under "Experimental Procedures." B, ␣ 3 ␤ 1 integrin mediates HCE cell adhesion to EPDIM peptide and NKDIL peptide. HCE cells were preincubated with function-blocking monoclonal antibodies to integrin subunits at a concentration of 5 g/ml for 30 min at 37°C and then seeded to the precoated wells with 100 M EPDIM or 100 M NKDIL. After 1 h incubation, cells attached to the substrates were quantified by hexosaminidase assay as described under "Experimental Procedures." FIG. 10. Effect of synthetic peptides on HCE cell adhesion to several extracellular matrix proteins. A, inhibition of HCE cell adhesion by synthetic peptides onto surfaces coated with matrix proteins. HCE cells were preincubated in medium in the absence or presence of 100 M each synthetic ␤ig-h3 peptide and seeded onto surfaces coated with 10 g/ml of each protein, i.e. ␤igh3-WT, FN, Col I, Col II, LM, or VN. After 1 h of incubation, cells attached to the substrates were quantified by hexosaminidase assay as described under "Experimental Procedures." B, dose-dependent inhibition of HCE cell adhesion to surfaces coated with 10 g/ml ␤igh3-WT, FN, or LM was measured in the presence of the indicated concentrations of EPDIM.