Matrix cell adhesion activation by non-adhesion proteins

Cell-matrix adhesion is important in diverse biological processes, including development, repair and disease. Many extracellular matrix components have been identified that regulate cell adhesion through interactions with various cell surface receptors (Damsky and Ilic, 2002; Ekblom et al., 2003; Pupa et al., 2002). These interactions are highly specific and multiple active sites on the adhesion molecules have been identified using protein fragments and synthetic peptides. For example, fibronectin binds to a certain integrin receptor, α₅β₁, through its RGD sequence and a synergy site. Similarly, laminin binds multiple cell surface proteins, including integrins, through various active regions that have also been defined at the synthetic peptide level (Belkin and Stepp, 2000; Colognato and Yurchenco, 2000; Engbring and Kleinman, 2003; Pankov and Yamada, 2002; Watt, 2002).

A simple way to measure cell adhesion is to coat the adhesive molecule on a substrate and then add the cells for a short incubation time. The proteins are coated onto the substrates at various concentrations in a physiological buffer overnight at 4°C, 1-2 hours at room temperature or 37°C, or dried-on overnight in H₂O. A direct comparison between all of these various assays has not been reported. Most likely, these diverse methods of coating results in different presentations and concentrations of the protein bound to the substrate. In addition to the various protocols for coating, several types of glass and plastic substrates are used and can yield differing results. For example, fibronectin binds better to hydrophobic surfaces than hydrophilic surfaces at a low concentration; however, more antigenic sites of fibronectin are exposed when it is bound to the hydrophilic surfaces (Grinnell, 1987; Grinnell and Feld, 1981; Grinnell and Feld, 1982). As well, more cells bind to a hydrophilic than a hydrophobic surface coated with fibronectin even though more fibronectin is bound to the hydrophobic surface (Grinnell, 1987; Grinnell and Feld, 1982). These results suggest that the presentation of the molecule is very important. How the proteins adsorb to the solid surface can influence the ability of the cells to recognize the protein. Even the concentration of the protein will influence its binding efficiency to certain plastic plates (Grinnell, 1987; Horbett, 1984). For example, fibrinogen will adsorb with high efficiency from plasma at intermediate plasma concentrations but with poor efficiency at both low and high concentrations. BSA can increase the binding of fibronectin to tissue culture plates and, thus, increase its biological activity (Grinnell and Feld, 1982).

Here, we have further demonstrated the ability of non-adhesive proteins to `activate' fibronectin adsorbed to plastic surfaces and extended previous studies in this area. Our studies detected artifactual cell adhesion-promoting activity for osteonectin when it was coated on plastic culture plates. A commercial preparation of human platelet osteonectin was found to be active for cell adhesion that was dependent on the presence of a low (1%) level of fibronectin whereas preparations lacking fibronectin were inactive even at high concentrations. This amount of fibronectin was not high enough to promote cell adhesion when used alone on the plate. When fibronectin at low (inactive) levels was added to the inactive osteonectin preparation, cell adhesion was observed. We further extended these data to show that osteonectin will `activate' other matrix molecules, such as laminin, vitronectin and collagen I for cell adhesion, when the adhesion molecules are also used at low levels where cell adhesion is normally not observed. The effect is non-specific for osteonectin since several proteins, including BSA and cytochrome C, also `activate' these proteins for cell adhesion. This `activation' of matrix adhesion molecules is dose-dependent on the amount of non-adhesive (inactive) protein added. In addition, the level of fibronectin binding to the substrate is not influenced by the non-adhesive protein. Finally, we show that the non-adhesive protein, osteonectin, `exposes' integrin-binding sites on fibronectin. We propose that the non-adhesive protein modulates the conformation of fibronectin and of other matrix adhesion molecules into an active form for cell adhesion.

Human MDA-MB-231 breast carcinoma cells were maintained in DMEM/F-12 medium containing 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (Invitrogen, Carlsbad, CA). Human PC3 prostate carcinoma cells were maintained in DMEM medium containing 10% FBS and 1% penicillin/streptomycin. Both cell lines were grown in humidified incubators with 5% CO₂ at 37°C.

Cell adhesion was assayed in U-bottom 96-well plastic plates (Immulon 2 HB, Thermo Electron Co., Somerset, NJ) coated overnight at 4°C with various concentrations of a 50 μl volume of different proteins in Dulbecco's phosphate buffered saline (dPBS, Invitrogen). These proteins were either human platelet or bovine bone osteonectin (Hematological Technologies, Essex, VT), human plasma fibronectin (a kind gift from Kenneth Yamada), human cellular fibronectin (Upstate, Lake Placid, NY), human fibronectin alpha-chymotryptic fragment 120 kDa or 40 kDa (Chemicon, Temecula, CA), mouse laminin (Trevigen, Inc., Gaithersburg, MD), Vitrogen-100 bovine collagen (type I) (Cohesion Technologies, Palo Alto, CA), recombinant osteopontin (a kind gift from Larry Fisher), equine cytochrome C (Calbiochem, San Diego, CA), or BSA (Sigma, St Louis, MO). Human plasma fibronectin was isolated as described (Miekka et al., 1982). The wells were blocked with 1% heat-denatured BSA for 30 minutes at 37°C and then washed twice with dPBS. Cells were detached with Versene (0.02% EDTA in PBS, Invitrogen), resuspended (3×10⁵ cells ml^–1) in Hank's balanced salt solution (Invitrogen) containing 0.1% BSA and 0.1 mM manganese and then 100 μl of cells were added to each well. Inhibition studies were done by pre-incubating cells with various concentrations (0.1, 1 and 10 μg ml^–1) of either GRGDS or GRGES (control) peptides (Sigma), or 20 μg ml^–1 of anti-integrin blocking antibodies α₁ (FB12), α₃ (ASC-1), α₆ (GoH3), α_v (L230), α_vβ₃ (LM609), α_vβ₅ (15F11), α₅ (mAb16) and β₁ (mAb 13), or non-blocking (control) anti-integrin antibody α₅ (mAb 11) [(mAb13, mAb 16 and mAb 11, kind gifts from Kenneth Yamada) (FB12, ASC-1, GoH3, L230, LM609 and 15F11 from Chemicon)] for 30 minutes at 4°C before addition to the wells. Each sample was assayed in triplicate. The plate was incubated for 45 minutes at 37°C in 5% CO₂. After washing to remove the unattached cells, the attached cells were stained with 0.2% crystal violet in 20% methanol for 10 minutes and washed twice with water. The cells were lysed with 10% SDS and the optical density (620 nm) was measured. This assay was repeated at least three times. Mean and standard deviations were determined. Data were analyzed by ANOVA with Dunnett's multiple comparisons post-test.

Proteins were coated in U-bottom 96-well plastic plates (Immulon 2 HB) and blocked and washed as described above. Rabbit anti-human osteonectin (Chemicon), rabbit anti-human fibronectin (a kind gift from Kenneth Yamada) and sheep anti-human collagen I (prepared at NIDCR, NIH) antibodies were diluted 1:1000, 1:7,500 and 1:1000, respectively, in 1% BSA-dPBS and incubated in the wells for 30 minutes at 37°C. The wells were then washed six times with dPBS, incubated for 30 minutes at 37°C with 1:25,000 dilution of either horseradish peroxidase-labeled secondary goat anti-rabbit or rabbit anti-goat IgG (Pierce, Rockford, IL) in 1% BSA-dPBS and then washed six more times with dPBS. Reactive proteins were detected according to the manufacturer's instructions with 1 StepTurbo TMB-ELISA (Pierce). The reaction was stopped by adding 1 M H₂SO₄ and optical density (450 nm) was measured. Rat anti-human fibronectin antibodies 3B8, 13G12 and 16G3 (gifts from Kenneth Yamada) were diluted to 0.5 μg ml^–1 in 1% BSA-dPBS and incubated in the wells overnight at 4°C. The wells were washed as above and incubated overnight at 4°C with 1:12,500 dilution of horseradish peroxidase-labeled secondary goat anti-rat IgG (Pierce, Rockford, IL) in 1% BSA-dPBS. Reactive proteins were detected according to the manufacturer's instructions with 1 StepTurbo Ultra TMB-ELISA (Pierce). The reaction was stopped as described above. Each experiment was repeated at least three times. Mean and standard deviations were determined.

Fibronectin was immunoprecipitated from human platelet osteonectin using a polyclonal rabbit anti-human fibronectin antibody. The antibody was diluted 1:10,000 in dPBS and 2.5 ml were added to 500 μl of GammaBind Plus Sepharose beads (Pharmacia Biotech, Piscataway, NJ) for 60 minutes at 4°C. Immunopure rabbit IgG (Pierce) was used as a control at a 1:10,000 dilution. The beads incubated with either rabbit anti-human fibronectin antibody or control rabbit IgG were washed six times with dPBS and then incubated with 42 μg of human platelet osteonectin in 2.0 ml dPBS for 2 hours at 4°C. After the 2 hour incubation, the beads containing the osteonectin were centrifuged and the supernatant was used for adhesion assays and ELISA.

All sedimentation experiments were performed with a Beckman Optima XL-A at the Keck Biophysics Facility at Northwestern University, Evanston, Illinois. Sedimentation velocity experiments of fibronectin and osteonectin alone and at molar ratios 1:5, 1:1 and 2:1 (osteonectin:fibronectin) were analyzed as described (van Holde and Weischet, 1978). Sedimentation distributions were interpreted according to the method of Demeler (Demeler et al., 1997). Fibronectin was also analyzed with the finite element method (Demeler and Saber, 1998). All analyses were performed with UltraScan software [Demeler. UltraScan 6.2 – An integrated data analysis software package for sedimentation experiments. (2004) University of Texas Health Science Center at San Antonio, Department of Biochemistry. http://www.ultrascan.uthscsa.edu]. Hydrodynamic corrections for buffer conditions were made according to data published (Laue et al., 1992). The partial specific volumes were estimated from the protein sequence according to the method of Cohn and Edsall (Cohn and Edsall, 1943) and found to be 0.708 ccm g^–1 for fibronectin and 0.728 ccm g^–1 for osteonectin. All samples were analyzed in dPBS (Invitrogen). Sedimentation velocity experiments were performed at 230 nm, 4°C and 262,000 g with aluminum 2-channel centerpieces in an AN 60 Ti rotor. Scans were taken without delay in continuous scanning mode with 3×10^–3 cm stepsize. Extinction coefficients for 230 nm were determined with UltraScan as previously described (Russell et al., 2004). Extinction coefficients for fibronectin were determined to be 708,953 OD mol^–1 cm^–1 at 230 nm and 159,470 OD mol^–1 cm^–1 at 280 nm. Extinction coefficients for osteonectin were determined to be 182,861 OD mol^–1 cm^–1 at 230 nm and 33,640 OD mol^–1 cm^–1 at 280 nm.

Purified human platelet and bovine bone osteonectin and human plasma fibronectin were subjected to SDS-PAGE using 4–20%-(w/v) gels (Bio-Rad). To quantify fibronectin amounts, serial dilutions of human plasma fibronectin (1.7, 3.3, 6.6, 13.3 and 26.6 ng) were loaded on the gels. The protein in these gels was transferred to nitrocellulose and then immunoblotted using 1 μg ml^–1 of mouse anti-human osteonectin monoclonal antibody (AON-5031, Haematologic Technologies Inc.), 1:30,000 dilution of rabbit anti-human fibronectin polyclonal antibody (a kind gift from Kenneth Yamada), 1:3000 dilution of mouse anti-human vitronectin monoclonal antibody (VIT-2, Sigma), or 1:500 dilution of rabbit anti-laminin polyclonal antibody (prepared at NIDCR, NIH) in 5% nonfat milk-PBS with 0.05% Tween-20 (T-PBS). Membranes were probed with horseradish peroxidase-labeled secondary antibodies (goat anti-mouse or anti-rabbit IgG) in 5% nonfat milk-T-PBS and reactive proteins were detected using SuperSignal^® West Dura extended duration substrate (Pierce). Chemiluminescence was detected using a Fuji LAS-1000 luminscent image analyzer `intelligent dark box' (Fujifilm Medical Systems USA, Inc., Standford, CT) using exposure times at subsaturation levels. The immunoblots were quantified using Fujifilm Science Lab 98 Image Gauge software V3.3 (Fuji Photo Film Co., LTD) as described (Larsen et al., 2003) using a standard curve that was generated from purified fibronectin loaded on the same gel (described above). The blots were stripped for 30 minutes at 37°C with Restore™ western blot stripping buffer (Pierce) and then washed with T-PBS before reprobing. Each experiment was repeated at least three times. Some gels were not transferred but were silver stained using SilverQuest silver staining kit (Invitrogen).

Hydroxyproline content in human plasma and bovine bone osteonectin was determined by high-performance liquid chromatographic separation and fluorometric quantitation as described (Palmerini et al., 1985). This assay was performed by Fibrogen, Inc., South San Francisco, CA.

MDA-231 breast and PC-3 prostate carcinoma cells metastasize to bone in vivo. We coated high-binding 96-well plates with various bone matrix proteins, including osteonectin, osteopontin, laminin and fibronectin, to determin whether the bone-metastasizing MDA-231 and PC-3 cells bound to these proteins. Data shown in all figures are for MDA-231, but similar results were seen with the prostate cancer cells. Adhesion of these breast cancer cells to all of these proteins was observed (Fig. 1). Binding to human platelet osteonectin was manganese-dependent (data not shown) and, therefore, in all adhesion assays manganese was used. These results suggest that cell binding to osteonectin is mediated by integrin interactions with the human platelet osteonectin. Blocking antibodies against β₁-integrins significantly inhibited MDA-231 cell binding to human platelet osteonectin, osteopontin, laminin and fibronectin (Fig. 1A). In addition, anti-α₅- and α_v-integrin blocking antibodies as well as α_vβ₃ integrin blocking antibodies inhibited cell binding to human platelet osteonectin and fibronectin. These results suggest α₅- and α_v-integrins are involved in cell adhesion to fibronectin and to human platelet osteonectin.

α₅β₁, α_vβ₁ and α_vβ₃ integrins are known to bind to RGD sequences in proteins. Cell adhesion to fibronectin and to osteopontin was dependent on the RGD sequence (Fig. 1B). Unexpectedly, cell binding to human platelet osteonectin, which does not have an RGD sequence like fibronectin and osteopontin, was also inhibited by an RGD peptide and at a much lower level than that needed to inhibit attachment to either of the RGD-containing proteins fibronectin or osteopontin (Fig. 1B). Adhesion to laminin was not blocked by RGD as expected and RGE, a control peptide, did not block cell adhesion to any of these proteins. These results suggest that other non-RGD-sensitive integrins are mediating adhesion to fibronectin and osteopontin. Additionally, since osteonectin does not contain an RGD sequence but cell binding to this protein was very sensitive to inhibition by RGD peptide and the cell surface receptor appeared to involve the fibronectin receptors, α₅β₁, α_vβ₁ and α_vβ₃ integrins, we hypothesized that the active preparations of osteonectin might contain low levels of fibronectin.

There are two commercially available sources of osteonectin and we found that although the breast and prostate cancer cells bound human platelet osteonectin, no adhesion to bovine bone osteonectin was observed (Fig. 2). We could not detect any proteins other than osteonectin by protein staining of SDS-PAGE gels (data not shown); however, we found that the human platelet osteonectin preparations contained low amounts of fibronectin whereas the bovine bone preparation did not contain any detectable amount by immunoblotting (Fig. 3) and ELISA (Table 1). In the ELISA analysis (Table 1), we tested whether the osteonectin preparations contained the matrix proteins fibronectin, collagen I, vitronectin or laminin. We found only fibronectin and a small amount of collagen I in human platelet osteonectin but not in bovine bone osteonectin. Collagen I binds to both osteonectin and fibronectin (Engvall and Ruoslahti, 1977; Termine et al., 1981). Known amounts of fibronectin were added to the plates from 0.01-10.0 μg ml^–1 to test the sensitivity of the ELISA assay. Since the reaction was no longer linear past 0.2 μg ml^–1 of fibronectin coating, which I was estimated by hydroxyproline analysis to be 0.7% of the total protein in the human platelet osteonectin, but none was detected in the bovine bone osteonectin. By quantitating the immunoblots, the amount of fibronectin was estimated to be 1% of the protein in the human platelet osteonectin (Fig. 3). These findings suggest that the integrin binding and adhesion activity of the human platelet osteonectin were due to low levels of fibronectin and/or collagen I present in the preparation.

We next tested whether the activity in the human platelet osteonectin preparation could be due to the low levels of fibronectin. The low amount of fibronectin (0.1 μg ml^–1) present in the osteonectin that was used to coat the 96-well plate did not support cell adhesion when coated alone on the plate (Fig. 2). Previous studies have found that cells bind better to surfaces exposed to high concentrations of fibronectin versus those exposed to low concentrations of the protein (Grinnell, 1987). When this low level of `inactive' fibronectin was mixed with the inactive bovine bone osteonectin and then coated on the plate, significant cell adhesion was observed (Fig. 2). The cell adhesion to this mixture of bovine bone osteonectin plus fibronectin (0.1 μg ml<sup>–1</sup>) is similar to cell adhesion to human platelet osteonectin; however, cell adhesion to both human platelet osteonectin and the mixture of bovine bone osteonectin plus fibronectin (0.1 μg ml<sup>–1</sup>) is significantly lower than that observed to a concentration of fibronectin (1.25 μg ml<sup>–1</sup>) which promotes cell adhesion (Fig. 2). Furthermore, when bovine bone osteonectin was added to low `inactive' levels of either laminin or vitronectin, a similar significant level of `activation' of cell adhesion activity was observed (Fig. 2). Collagen I, which gave a low level of cell adhesion, also showed significantly increased activity in the presence of bovine bone osteonectin, but the effect was not as great as that observed with bovine bone osteonectin mixed with the other matrix proteins.

We immunoprecipitated the fibronectin from the human platelet osteonectin to demonstrate further that the fibronectin contained in the human platelet osteonectin was responsible for cell binding to the human platelet osteonectin. The MDA-231 cells did not bind human platelet osteonectin when the fibronectin was removed (Fig. 4); however, once the fibronectin was added back to the depleted preparation at only 0.1 μg ml^–1, cell binding to human platelet osteonectin could be restored. The amount of fibronectin added back is equivalent to the amount that was present before removal by immunoprecipitation as determined by immunoblotting (Fig. 3) to be 1% of the total human platelet osteonectin. Although some osteonectin was also removed in the immunoprecipitation experiment (Fig. 4B), this did not affect the binding since osteonectin was also lost in the control immunoprecipitation experiment and the cells still bound to this osteonectin (Fig. 4A). The loss of osteonectin was attributed to non-specific binding of osteonectin to the immunobeads used in this experiment. There was a significant increase in cell adhesion to the human platelet osteonectin that was immuno-depleted of fibronectin and then had fibronectin added back, compared with human platelet osteonectin before immunoprecipitation. Human platelet osteonectin contains fibronectin having the molecular weight of approximately 188-120 kDa while the human plasma fibronectin that was added back has higher molecular weight forms of fibronectin present (Fig. 3). These different molecular weight forms of fibronectin may have contributed to the increase in cell binding.

These data demonstrate that osteonectin is not an adhesion protein but rather the observed adhesion activity was due to low levels of fibronectin. Osteonectin increased cell spreading on low levels of fibronectin in addition to increasing cell adhesion (data not shown). We tested the effect of increasing concentrations of bovine bone osteonectin on adhesion to a fixed amount of fibronectin (0.1 μg ml^–1) and found that optimal adhesion was observed at 4.0 μg ml^–1 of osteonectin. At higher concentrations, less adhesion was observed (Fig. 5A). Using this optimal adhesion concentration (4.0 μg ml^–1) of the bovine bone osteonectin, we added increasing amounts of either fibronectin, laminin, or vitronectin starting at concentrations that promote little or no attachment to determine the optimal amount of adhesion protein required for maximal cell attachment activity (Fig. 5B). When increasing amounts of adhesion proteins were added to the bovine bone osteonectin and coated on the plate together, the level of adhesion was greater than with the adhesion protein alone for all adhesion molecules tested. Statistically significant differences in adhesion were reached at all concentrations of fibronectin and vitronectin tested with 4 μg ml^–1 osteonectin, but laminin was only activated at 0.4 g ml^–1 μ. No effect was observed on osteopontin adhesion with the addition of osteonectin (data not shown). These data demonstrate that the activity of different adhesion molecules can be potentated by specific concentrations of osteonectin.

We next determined if the order of coating the proteins on the plate regulated the cell adhesive activity. `Activation' of breast cancer cell adhesion was observed either when fibronectin was added together with bovine bone osteonectin or when the bovine bone osteonectin was added to the plate first followed by the addition of fibronectin (Table 2). Cell adhesion was not supported when fibronectin was first added to the plate followed by the addition of bovine bone osteonectin. These data suggest that there may be some effect of osteonectin on fibronectin binding to the plate. When we tested this by ELISA, we found that adding the bovine bone osteonectin to the plate first had no effect on the binding of fibronectin to the plate (Table 3). The level of plate-bound fibronectin, when added alone or preceded by osteonectin, was identical. Some reduction in bound fibronectin levels was detected when fibronectin was coated first followed by osteonectin. These data demonstrate that the increase in cell adhesion when osteonectin is added to fibronectin was not due to an increase in the amount of fibronectin bound to the plate. Furthermore, we found no interaction of fibronectin with osteonectin either by solid phase assays (data not shown) or by sedimentation velocity centrifugation (Table 4). Finite element analysis of velocity experiments of fibronectin alone at 1 μM concentration indicated the presence of 7% dimer (8.1s) and 93% tetramer (13.4 s). This ratio remained unchanged within experimental error when osteonectin was added to the mixture, regardless of the osteonectin:fibronectin ratio examined. In addition, no new species was observed in the mixtures when osteonectin was added except the osteonectin itself. The observed ratios are listed in Table 4. We deduce a dimer-tetramer dissociation constant of 18.2 nM for fibronectin. These results suggest that there is no evidence of osteonectin/fibronectin interactions. The van Holde-Weischet G(s) distributions for fibronectin and osteonectin/fibronectin mixtures are shown in Fig. 6.

Since osteonectin does not bind to fibronectin and did not affect its ability to bind to the adhesion plates, we tested whether the effect of osteonectin was non-specific and could be duplicated by other molecules. We found that the effect on `activating' the cell adhesion activity of fibronectin, laminin, vitronectin and collagen I could also be observed with other proteins. Cytochrome C and BSA both `activated' low levels of these matrix proteins for adhesion (Fig. 7). These data demonstrate that multiple proteins can `promote/activate' the activity of adhesion proteins.

Anti-integrin blocking antibodies, α₅, α_v, β₁ and α_vβ₃ used in Fig. 1 also block adhesion of MDA-231 breast carcinoma cells to bovine bone osteonectin plus fibronectin (data not shown). These results suggest that the adhesion of cells to fibronectin in the presence of a `non-adhesive' protein is a result of fibronectin binding to the dish in a conformation that `exposes' the integrin-binding site. The RGD integrin-binding site appears to be exposed in fibronectin when it is mixed with a `non-adhesive' protein and important for the cell adhesion since cell adhesion to human platelet osteonectin was very sensitive to inhibition by RGD peptide. Therefore, the `non-adhesive' proteins should not have an effect on the adhesion of cells to fragments of fibronectin that do not contain the integrin-binding sites. In the previous experiments, we used human plasma fibronectin. We also tested if human cellular fibronectin, another variant of fibronectin (Pankov and Yamada, 2002), or human fibronectin fragments (Fig. 8) could be activated in the same way as well. Two alpha-chymotryptic fragments of fibronectin were tested, a 40 kDa fragment, which contains heparin-binding domains of fibronectin, but not integrin-binding domains and a 120 kDa fragment, the classic cell-binding fragment, which contains only integrin-binding domains, including the `synergy', loop and RGD regions of fibronectin (Pankov and Yamada, 2002; Pierschbacher et al., 1981). Bovine bone osteonectin and cytochrome C both significantly enhanced attachment of cells to the cellular fibronectin as well as to the 120 kDa fragment of fibronectin, but not to the 40 kDa fragment (Fig. 8), evidence further suggesting that the integrin-binding sites are important in this cell adhesion. To further examine if the integrin-binding sites in fibronectin are exposed by the addition of the non-adhesive protein, we used three monoclonal anti-fibronectin antibodies, 3B8, 13G12 and 16G3, which bind to the `synergy' and loop region and near the RGD region of fibronectin, respectively (Nagai et al., 1991). These antibodies map to the 9th and 10th type III modules of fibronectin (Nagai et al., 1991; Pankov and Yamada, 2002). The `synergy' region is in the central cell-binding domain distinct from the RGD site and is important for cell adhesion to fibronectin via α<sub>5</sub>β<sub>1</sub> integrin (Nagai et al., 1991; Pankov and Yamada, 2002). The loop region is a domain between the synergy and RGD sites and is not essential for cell attachment (Nagai et al., 1991). The best-characterized cell-binding domain of fibronectin is the RGD sequence, which binds to α<sub>5</sub>β<sub>1</sub>, α<sub>v</sub>β<sub>1</sub> and α<sub>v</sub>β<sub>3</sub> integrins and to other integrins. All three antibodies detected fibronectin at a concentration that promotes cell attachment (1.25 μg ml<sup>–1</sup>), but only 13G12 (loop region) detected fibronectin in plates coated with a low level of fibronectin (0.1 μg ml<sup>–1</sup>) that does not promote cell attachment (Table 5). None of these antibodies detected fibronectin in bovine bone osteonectin; however, all three antibodies detected fibronectin in human platelet osteonectin and in the mixtures of fibronectin (0.1 μg ml<sup>–1</sup>) plus bovine bone osteonectin (Table 5). These results demonstrate that the presence of a non-adhesive protein, such as osteonectin or cytochrome C exposes the integrin `synergy' and RGD-binding sites on fibronectin.

Cell adhesion to proteins adsorbed on plastic substrates is often used to study the biological activity of various extracellular matrix components. Such adhesion assays are simple, quick and quantitative, and are commonly used as a first screen for protein biological activity. These assays are also used to identify active sites on adhesion molecules as well as cellular receptors and possible receptor antagonists. Fibronectin is an abundant adhesion protein found in serum and in many extracellular matrices. The binding of fibronectin to plastic substrates is well characterized as well as its biological activity after binding to the plastic substrates (Grinnell, 1987; Grinnell and Feld, 1981; Grinnell and Feld, 1982). Studies on the adsorption of adhesion molecules have found previously that different substrates can adsorb varying amounts of fibronectin, but the biological activity may not correlate with the amount of material adsorbed. For example, fibronectin binds better to hydrophobic surfaces than to hydrophilic surfaces, but cell spreading is better on hydrophilic surfaces (Grinnell and Feld, 1982). The concentration of the adhesion protein used and/or the presence of additional non-adhesive proteins can also regulate the amount of adhesion protein bound to the plate and the biological activity (Grinnell and Feld, 1981). In searching for the adhesion activity of osteonectin, we have confirmed here the finding that non-adhesive proteins can increase the biological activity of adhesion proteins. We find that this increase in activity occurs when the adhesion protein and the non-adhesive protein are added to the substrate together or when the non-adhesive protein is added first. In our studies, the amount of fibronectin added to the plate alone is too low to be active, but when added in the presence of osteonectin or other non-adhesive proteins, including BSA and cytochrome C, fibronectin can be `activated' for adhesion (Fig. 9). We extended these findings and found that similar results can also be obtained with other adhesion molecules, including collagen I, vitronectin and laminin; however, the activation of an adhesion protein by a `non-adhesion' protein was not universal to all `adhesion' proteins since osteopontin could not be `activated'. The concentrations of both the non-adhesive protein and the adhesion protein are critical for `activation' of adhesion. These data raise concerns about low levels of adhesion proteins in non-active protein preparations, which might give the artifactual result of promoting adhesion and binding certain cellular receptors. The amounts of adhesion protein that we used to achieve adhesion were well below the detectable level of a standard SDS-PAGE gel stained with Coomassie blue.

The order in which the adhesion protein and non-adhesive protein were added to the plate was very important for activity. If the two proteins were added together or the non-adhesive protein was added first, then an `activation' of cell attachment was observed. This activation was at concentrations of the adhesion protein that do not normally promote cell attachment. The amount of fibronectin bound to the plate under either condition was not the determinant of this activation. In both cases, we found that equal amounts of fibronectin were bound to the plate compared with fibronectin added alone, indicating that the non-adhesive molecule at the concentrations used did not block, increase, or compete with fibronectin binding to the plate. Since fibronectin alone was at a concentration too low to be active, we assume that its conformation on the plate, and not the amount bound, was important for its biological activity. Grinnell suggested that molecules adsorbed on plates at low density have the space to orient differently from molecules adsorbed on the plates at high density (Grinnell, 1987). Interestingly, we found that if we coated first with fibronectin and then with osteonectin, less fibronectin bound to the plate and no cell attachment activity was observed. We believe that the attachment was not observable because either fibronectin was bound first alone in the wrong conformation and/or the reduced amount resulted in no adhesion. To confirm that the conformation was important, we mapped the integrin-binding domain of fibronectin coated at a low level with and without osteonectin. We found that in the presence of osteonectin, the `synergy' and RGD sites are exposed along with the loop region of fibronectin. In addition, cells did adhere to the mixture of osteonectin or cytochrome C with a low level of a 120 kDa fragment of fibronectin but not to the mixture with a 40 kDa fragment of fibronectin, which does not contain the integrin-binding site. These results together suggest that the `non-active' protein exposes the integrin-binding sites of the active protein and this is important for cell adhesion. In addition, these results indicate that the RGD-binding site is exposed in fibronectin when it is mixed with osteonectin. Since the amount of fibronectin is at a very low concentration, only 1 μg ml^–1 of RGD peptide was necessary to inhibit cell binding to human platelet osteonectin. In support of this, Yamada and Kennedy have shown that the competitive effect of RGD peptide is diminished in a dose-dependent manner as the amount of fibronectin coated onto a substrate increases (Yamada and Kennedy, 1984).

Our finding of biological activity and receptor binding due to fibronectin in the human platelet osteonectin was unexpected. Determining the potentiation of fibronectin activity by osteonectin and by other non-adhesive proteins raises the concern that other researchers should search for fibronectin or other known adhesion molecules in their preparations of putative `adhesion' molecules. Additionally, researchers need rigorously to confirm the purity of their molecules when studying ligand/receptor interactions as well. Since extracellular matrix proteins generally interact with each other, there are likely to be multiple contaminants in many preparations. For example, our finding of both fibronectin and collagen in some of the human platelet osteonectin is understandable since osteonectin binds collagen and collagen binds fibronectin (Engvall and Ruoslahti, 1977; Termine et al., 1981). Such multiple direct and indirect binding molecules may be functional proteins in both commercially prepared and custom-purified preparations of adhesion molecules.

The in vivo significance of these findings is difficult to assess. Unlike cell culture where plastic surfaces bind single proteins in potentially non-physiological conformations, in vivo extracellular matrices are more complex. One might speculate that in the context of the tissue matrices where there are multiple interacting proteins there may be some conformational changes or `activation' of adhesion proteins similar to what we have observed in vitro.

This work was supported in part by National Research Service Award Grant CA-91572 (to J. E. Koblinski) from the National Cancer Institute, NIH. The development of the UltraScan software is supported by the National Science Foundation through grant DBI-9974819 to B. Demeler. We gratefully acknowledge Katherine Clark, Harold Erickson, Matthew Hoffman, Peter Nielsen and Kenneth Yamada for helpful discussions and critical review of the manuscript.