A Peptide Model of Basement Membrane Collagen α 1(IV) 531- 543 Binds the α3β1 Integrin

Tumor cell adhesion to the triple-helical domain of basement membrane (type IV) collagen occurs at several different regions. Cellular recognition of the sequence spanning α1(IV)531-543 has been proposed to be independent of triple-helical conformation (Miles, A. J., Skubitz, A. P. N., Furcht, L. T., and Fields, G. B.(1994) J. Biol. Chem. 269, 30939-30945). In the present study, integrin interactions with a peptide analog of the α1(IV)531-543 sequence have been analyzed. Tumor cell adhesion (melanoma, ovarian carcinoma) to the α1(IV)531-543 chemically synthesized peptide was inhibited by a monoclonal antibody against the α3 integrin subunit, and to a lesser extent by monoclonal antibodies against the β1 and α2 integrin subunits. An anti-α5 monoclonal antibody and normal mouse IgG were ineffective as inhibitors of tumor cell adhesion to the peptide. Two cell surface proteins of 120 and 150 kDa bound to an α1(IV)531-543 peptide affinity column and were eluted with 20 mM EDTA. When the eluted proteins were incubated with monoclonal antibodies against either the α3 or β1 integrin subunit, proteins corresponding in molecular weight to α3 and β1 integrin subunits were precipitated. No proteins were immunoprecipated with monoclonal antibodies against the α2 or α5 integrin subunits. Thus, the α3β1 integrin from two tumor cell types has been shown to bind directly to the α1(IV)531-543 peptide. The α1(IV)531-543 peptide is the first collagen-like sequence that has been shown to bind the α3β1 integrin.

In the present study, integrin interactions with a peptide analog of the ␣1(IV)-531-543 sequence have been analyzed. Tumor cell adhesion (melanoma, ovarian carcinoma) to the ␣1(IV)531-543 chemically synthesized peptide was inhibited by a monoclonal antibody against the ␣ 3 integrin subunit, and to a lesser extent by monoclonal antibodies against the ␤ 1 and ␣ 2 integrin subunits. An anti-␣ 5 monoclonal antibody and normal mouse IgG were ineffective as inhibitors of tumor cell adhesion to the peptide. Two cell surface proteins of 120 and 150 kDa bound to an ␣1(IV)531-543 peptide affinity column and were eluted with 20 mM EDTA. When the eluted proteins were incubated with monoclonal antibodies against either the ␣ 3 or ␤ 1 integrin subunit, proteins corresponding in molecular weight to ␣ 3 and ␤ 1 integrin subunits were precipitated. No proteins were immunoprecipated with monoclonal antibodies against the ␣ 2 or ␣ 5 integrin subunits. Thus, the ␣ 3 ␤ 1 integrin from two tumor cell types has been shown to bind directly to the ␣1(IV)531-543 peptide. The ␣1(IV)531-543 peptide is the first collagen-like sequence that has been shown to bind the ␣ 3 ␤ 1 integrin.
Other collagen-derived sequences may also function as integrin binding sites. Peptide models of the ␣1(IV)1263-1277 region promote the adhesion, spreading, and migration of highly metastatic tumor cells (Chelberg et al., 1990;Mayo et al., 1991;Fields et al., 1993). A peptide incorporating the ␣1(IV)531-543 1 (Gly-Glu-Phe-Tyr-Phe-Asp-Leu-Arg-Leu-Lys-Gly-Asp-Lys-Tyr) sequence promotes keratinocyte, corneal epithelial, melanoma, ovarian carcinoma, and Jurkat cell adhesion (Wilke and Furcht, 1990;Cameron et al., 1991;Miles et al., 1994) and migration of corneal epithelial cells and keratinocytes (Cameron et al., 1991;Kim et al., 1994). Cellular recognition of the ␣1(IV)531-543 peptide is, in general, independent of substrate conformation and configuration (chirality) (Miles et al., 1994). Competition studies suggested that the L-and D-peptides incorporating ␣1(IV)531-543 are bound by the same receptor (Miles et al., 1994). Preliminary results indicated that the ␤ 1 integrin subunit was involved in mediating cell adhesion to this sequence (Miles et al., 1994). We have presently examined integrin binding to the ␣1(IV)531-543 sequence. Since an ␣1␤ 1 integrin binding site in type IV collagen has already been identified (see above), we have focused on the ␣2␤ 1 and ␣ 3 ␤ 1 integrins as possible receptors for ␣1(IV)531-543. In addition, the ␣ 5 ␤ 1 integrin has been considered as a potential receptor for ␣1(IV)531-543 in this study, as it can mediate cell adhesion to denatured collagen (Gullberg et al., 1992;Tuckwell et al., 1994). We have utilized a number of assays including inhibition of cell adhesion and affinity chromatography of solubilized cells to characterize the integrin(s) involved in cellular recognition of the ␣1(IV)531-543 sequence. Two different tumor cell lines have been utilized, melanoma and ovarian carcinoma, since both of these cell types adhere well to type IV collagen and the ␣1(IV)531-543 peptide (Miles et al., 1994).
Cells were passaged for 4 -5 weeks and then replaced from frozen stocks of early passage cells to minimize phenotypic drift. All cells were maintained at 37°C in a humidified incubator containing 5% CO 2 . All media reagents were purchased from Sigma except where noted.
Cell Adhesion-Melanoma and ovarian carcinoma cell adhesion and inhibition of cell adhesion assays were performed as described previously (Chelberg et al., 1990;Wilke and Furcht, 1990;Miles et al., 1994) using the ␣1(IV)531-543 peptide at a substrate concentration of 5.7 M. For inhibition assays, cells were preincubated for 1 h at 37°C with various concentrations (0.005-5 g/ml) of the anti-integrin subunit mAb; then the cells (50,000/ml), in the continued presence of the mAb, were added to the Immulon plate wells and allowed to adhere for another 1 h at 37°C.
Affinity Chromatography-The ␣1(IV)531-543 peptide was coupled to activated CH-Sepharose according to the manufacturer's instructions (Pharmacia Biotech Inc.). In addition, a mock-coupled column was made without peptide for use as a control. Briefly, 30 mg of high performance liquid chromatography-purified peptide was dissolved in 200 l of Me 2 SO and diluted to 5 ml with coupling buffer (15 mM sodium carbonate and 35 mM sodium bicarbonate, pH 8.6). The peptide solution was added to 3 ml of preswollen beads and mixed overnight at 4°C. Unbound peptide was removed by washing the beads with coupling buffer, and the remaining reactive groups were hydrolyzed at pH 8.0 with 0.1 M Tris-HCl for 2 h. Cells were surface labeled with 125 I as described , extracted in buffer (50 mM Tris-HCl, pH 7.4, 50 mM octyl-␤-D-glucopyranoside, 15 mM NaCl, 1 mM MgCl 2 , 1 mM MnCl 2 , 1 mM CaCl 2 , 1 mM N-ethylmaleimide, 100 g/ml soybean trypsin inhibitor, and 1 mM phenylmethylsulfonyl fluoride) for 30 min at 4°C, and ultracentrifuged at 36,500 ϫ g for 1 h at 4°C. The 125 Ilabeled cell lysates were precleared with the mock-coupled Sepharose beads as a slurry by constant mixing overnight at 4°C. Cleared lysates were then incubated with the peptide-Sepharose beads as a slurry with constant mixing overnight at 4°C. The peptide-Sepharose beads were packed into a column, and the column was washed with extraction buffer and eluted with 20 mM EDTA in extraction buffer lacking the cations. Eluates were incubated with anti-integrin subunit mAbs to immunoprecipitate specific integrins . 2-Mercaptoethanol (10%) was added to some samples (see "Results"), followed by heating at 100°C for 5 min to reduce disulfide bonds. Elution fractions were electrophoresed by 7.5% SDS-polyacrylamide gel electrophoresis and analyzed by autoradiography . Molecular mass standards (Sigma) were rabbit muscle myosin (205 kDa), Escherichia coli ␤-galactosidase (116 kDa), rabbit muscle phosphorylase b (97.4 kDa), bovine albumin (66 kDa), and chicken egg albumin (43 kDa). RESULTS We had shown previously that 40 -60% maximum cell adhesion is achieved with 5-10 M amounts of a peptide analog of the ␣1(IV)531-543 sequence (Miles et al., 1994). Inhibition of cell adhesion studies were thus performed at a peptide concentration of 5.7 M. Ovarian carcinoma cell adhesion to the peptide could be inhibited in a dose-dependent manner by mAbs against ␤ 1 , ␣2, or ␣ 3 integrin subunits, with maximum inhibition occurring at the highest mAb concentration tested (5 g/ ml) (Fig. 1A). The mAb against the ␣ 3 subunit was most effective, producing Ͼ60% inhibition of adhesion at a mAb concentration as low as 2.5 g/ml (data not shown). A similar level of inhibition (ϳ50%) was achieved by the anti-␤ 1 subunit mAb at a concentration of 5 g/ml (Fig. 1A). The anti-␣2 subunit mAb was less effective as an inhibitor, causing only ϳ35% inhibition at 5 g/ml (Fig. 1A). A mAb against the ␣ 5 integrin subunit and normal mouse IgG did not give dose-dependent inhibition of cell adhesion, even up to a concentration of 5 g/ml (Fig. 1A). In a similar fashion, melanoma cell adhesion to the peptide could be inhibited in a dose-dependent manner by mAbs against the ␤ 1 , ␣2, or ␣ 3 integrin subunits, with maximum inhibition occurring at a mAb concentration of 5 g/ml (Fig. 1B). The anti-␣ 3 subunit mAb was the most effective inhibitor, causing ϳ60% inhibition. A mAb against the ␣ 5 integrin subunit and normal mouse IgG did not give dose-dependent inhibition of melanoma cell adhesion up to a concentration of 5 g/ml (Fig. 1B).
The ␣1(IV)531-543 peptide was immobilized to CH-Sepharose, and affinity chromatography was performed with an 125 Ilabeled extract of the ovarian carcinoma cells. Following application of cells, the column was first washed with extraction buffer, then eluted by 20 mM EDTA. Eluants were incubated with 5 g/ml mAbs against ␣2, ␣ 3 , ␣ 5 , or ␤ 1 integrin subunits. Precipitated proteins were then analyzed by 7.5% SDS-polyacrylamide gel electrophoresis with detection by autoradiography. Immunoprecipitation of the EDTA eluant with the anti-␤ 1 integrin subunit mAb resulted in detection of a 135-kDa protein under reducing conditions (Fig. 2). Immunoprecipitation of the same eluant with the anti-␣ 3 integrin subunit mAb, followed by reduction with 2-mercaptoethanol, resulted in detec-

FIG. 1. Inhibition of ovarian carcinoma (A) or melanoma cell (B) adhesion to
Immulon I plate surfaces adsorbed with 5.7 M ␣1(IV)531-543 peptide by 5 g/ml of mAbs to the integrin subunits ␤ 1 , ␣ 2 , ␣ 3 , or ␣ 5 . Normal mouse (n.m.) IgG was used as a negative control. Cells were preincubated with the mAbs for 1 h, then added to the wells in the presence of the mAbs for a 1 h incubation. All assays were repeated at least in triplicate. Conditions are given under "Experimental Procedures." tion of a 135-kDa protein (Fig. 2). In other studies, the ␤ 1 and ␣ 3 integrin subunits have been shown to have similar apparent molecular weights under reducing conditions Sonnenberg, 1993). No proteins were seen following incubation of the EDTA eluant with anti-␣2 integrin subunit mAb or normal mouse IgG (Fig. 2). When the immunoprecipitants were not reduced, two proteins of 150 and 120 kDa were immunoprecipitated from the EDTA eluant by the anti-␤ 1 integrin subunit mAb (Fig. 3). The molecular weights correspond to the ␣ 3 and ␤ 1 integrin subunits, respectively (Sonnenberg, 1993). In similar fashion, two proteins of 150 and 120 kDa were immunoprecipitated from the EDTA eluant by the anti-␣ 3 integrin subunit mAb (Fig. 3). No proteins were seen using normal mouse IgG or mAbs against the ␣2 or ␣ 5 integrin subunits (Fig. 3).
Affinity chromatography and immunoprecipitation experiments were repeated using 125 I-labeled melanoma cells. Incubation of the EDTA eluant with the anti-␣ 3 integrin subunit mAb resulted in immunoprecipitation of a 135-kDa protein under reducing conditions (data not shown). Similarly, incubation of the eluant with the anti-␤ 1 integrin subunit mAb resulted in immunoprecipitation of a 135-kDa protein under reducing conditions (data not shown). Under non-reducing conditions, two proteins of 150 and 120 kDa were immunoprecipitated from the EDTA eluant by the anti-␣ 3 integrin subunit mAb (Fig. 4). The apparent molecular weights correspond to the ␣ 3 and ␤ 1 integrin subunits, respectively (Sonnenberg, 1993). No proteins were seen using normal mouse IgG (Fig. 4) or mAbs against the ␣2 (Fig. 4) or ␣ 5 integrin subunits (data not shown).

DISCUSSION
In an attempt to further refine our knowledge of cellular receptors for type IV collagen, melanoma and ovarian carcinoma cell adhesion to the ␣1(IV)531-543 peptide was examined in the presence of anti-integrin subunit mAbs. For both cell types, adhesion to this peptide was most effectively inhibited by the anti-␣ 3 integrin subunit mAb, followed by the anti-␤ 1 and anti-␣2 integrin subunit mAbs. Cell surface proteins with molecular masses of 120 and 150 kDa bound to a ␣1(IV)531-543 peptide affinity column in an EDTA-dependent fashion. These proteins could be immunoprecipitated with mAbs against either the ␣ 3 or ␤ 1 integrin subunit, and the protein molecular weights corresponded to ␣ 3 and ␤ 1 integrin subunits. Thus, it would appear that the ␣ 3 ␤ 1 integrin binds directly to the ␣1(IV)531-543 peptide.
The ␣1(IV)531-543 peptide is the first collagen-like sequence identified as a specific ␣ 3 ␤ 1 integrin/ligand binding site. Two different tumor cell types, melanoma and ovarian carcinoma, bind this site via the ␣ 3 ␤ 1 integrin. At present, it is unclear as to what role this integrin plays during tumor cell invasion, where extravasating cells have contact with type IV collagen. The level of ␣ 3 ␤ 1 expression varies amongst tumor cell types. Metastatic melanoma cells up-regulate the ␣ 3 ␤ 1 integrin compared with primary melanoma cells (Yoshinaga et al., 1993), while highly invasive prostate carcinoma cells have decreased expression of the ␣ 3 ␤ 1 integrin compared to the parental cell line . Transformed fibroblasts retain the same level of ␣ 3 ␤ 1 as non-transformed cells while decreasing levels of other integrins (Plantefaber and Hynes, 1989).
The level of ␣ 3 ␤ 1 integrin expression may correlate to the utility of this integrin for cell migration. mAbs to the ␣ 3 subunit inhibit melanocyte and melanoma cell motility (Morelli et al., 1993;Yoshinaga et al., 1993;Melchiori et al., 1995) and dysplastic nevus cell spreading and migration (Vink et al., 1994) on type IV collagen. Tumorigenic cell types such as metastatic melanoma cells may have increased levels of ␣ 3 ␤ 1 , which enhances motility on the basement membrane or basement membrane molecules. Other cell types may use different receptors for migration. For example, keratinocytes use ␣2␤ 1 integrins for migration on type IV collagen (Chen et al., 1993;Kim et al., 1994). There are suggestions that cellular interactions with the basement membrane via ␣ 3 ␤ 1 integrins may also lead to basement membrane degradation. Antibodies to ␣ 3 ␤ 1 integrin stimulate the expression of matrix metalloproteinase-9 (92 kDa type IV collagenase) (Larjava et al., 1993). Matrix metalloproteinase-9 can be induced in transformed cells (Wilhelm et al., 1989) or by direct contact with tumorigenic cells (Himelstein et al., 1994) and has been localized to the invasion front of oral squamous cell carcinoma (Kawahara et al., 1993). Matrix metalloproteinase-9 efficiently degrades type IV collagen (Morodomi et al., 1992).
The ␣2␤ 1 integrin may have a role in cellular recognition of the ␣1(IV)531-543 sequence based on the inhibition of cell adhesion assays. ␣2␤ 1 integrin binding to the anti-␣2 integrin FIG. 2. Immunoprecipitation analysis of ovarian carcinoma cell surface proteins eluted from the ␣1(IV)531-543 peptide affinity column. The proteins eluted by EDTA from the peptide column were immunoprecipitated with anti-integrin mAbs and reduced. Immunoprecipitation of a 135-kDa protein(s) was seen using mAbs against either the ␤ 1 or ␣ 3 integrin subunit. No proteins were immunoprecipitated when an anti-␣2 mAb or normal mouse (n.m.) IgG were used.
FIG. 3. Immunoprecipitation analysis of ovarian carcinoma cell surface proteins eluted from the ␣1(IV)531-543 peptide affinity column. The proteins eluted by EDTA from the peptide column were immunoprecipitated with either an anti-␤ 1 or anti-␣ 3 integrin subunit mAb. Two proteins of 120 and 150 kDa, corresponding to the ␤ 1 and ␣ 3 integrin subunits, respectively, were immunoprecipitated. No proteins were immunoprecipitated when an anti-␣2 mAb, an anti-␣ 5 mAb, or normal mouse (n.m.) IgG were used.
There is also a greater expression of the ␣ 3 ␤ 1 integrin than the ␣2␤ 1 integrin for the ovarian carcinoma 2 and melanoma 3 cell types studied here, possibly contributing to more ␣ 3 ␤ 1 receptorligand binding events and a more avid binding overall.