The mechanism of interaction of chondrocytic cells with cartilage-specific type II collagen has been examined using HCS-2/8 human chondrosarcoma cells as a model system. By the criteria of specific collagen secretion and integrin expression profile, HCS-2/8 have a similar differentiated phenotype to normal chondrocytes and are therefore a good model system. HCS-2/8 cells were able to attach and spread on both native and heat-denatured pepsinised type II collagen, and assays using denatured cyanogen bromide fragments apparently localised the major cell binding site to the CB10 fragment. However, when they were used as soluble inhibitors, cyanogen bromide fragments were found to block adhesion to denatured collagen, but had no effect on either attachment or spreading on the native molecule. The inability of cyanogen bromide fragments to reproduce the cell binding site of native collagen demonstrated a strict dependence on collagen conformation. This was also reflected in the receptors that were employed by HCS-2/8 cells for binding to type II collagen: binding to native collagen was mediated by the integrin α2β1 while binding to denatured collagen was mediated by a novel α5β1-fibronectin bridge. The identification of this bridge adds to the mechanisms by which cells can bind to denatured collagens. The previously characterised KDGEA active site peptide from type I collagen was found to be inactive as an inhibitor of type II collagen-mediated adhesion. The implications of these findings for the strategies used to identify adhesive active sites within collagens are discussed. In particular, these data suggest that, unlike other integrin ligands, a synthetic peptide-based approach is not suitable for the identification of collagen active sites.

The interactions of collagens with cells are important in a variety of biological processes, including adhesion, migration, growth and differentiation. Cell adhesive activity has been demonstrated in vitro for a number of collagens including types I, II, III, IV and VI (Wayner and Carter, 1987; Aumailley et al., 1989; Kramer and Marks, 1989; reviewed by Tuckwell and Humphries, 1993) and the specificity of these interactions is illustrated by their sensitivity to collagen denaturation; e.g. for type I collagen (Santoro, 1986), type IV collagen (Aumailley and Timpl, 1986) and type VI collagen (Aumailley et al., 1989).

The integrins are a family of heterodimeric receptor proteins that mediate the interactions of many cell types with a wide range of extracellular matrix, cell membrane and soluble proteins (reviewed by Hynes, 1987, 1992; Humphries, 1990). Integrins α1β1, α2β1 and α3β1 have been reported to be the major cell surface receptors for collagens (Wayner and Carter, 1987; Takada et al., 1988; Kramer and Marks, 1989), although other non-integrin collagen receptors and collagen binding proteins have been documented (reviewed by Akiyama et al., 1990).

The molecular biology of integrin-collagen binding is at present poorly understood. While a common feature of most integrin ligand active sites is a critical aspartate residue within a short linear peptide sequence, e.g. RGD, LDV and QAGDV, collagen molecules appear to employ different mechanisms. Collagens are characterised by the presence of a triple-helical region or regions (reviewed by Kielty et al., 1993). The triple-helical structure constrains the conformation of any receptor binding site and, although collagen molecules contain a number of RGD sequences, cell binding to native collagens has been repeatedly demonstrated to be insensitive to RGD peptides (e.g. see Santoro, 1986; Kramer and Marks, 1989; Gullberg et al., 1992). A number of studies have attempted to localise integrin binding sites within collagen molecules. Integrins α1β1 and α2β1 can both act as receptors for type I and type IV collagen (Wayner and Carter, 1987; Kramer and Marks, 1989; Staatz et al., 1989; Gullberg et al., 1990). A cyanogen bromide (CNBr)-generated fragment of the α1 chain of type I collagen, α1(I)CB3, has been shown to support α2β1 binding (Staatz et al., 1990; Gullberg et al., 1990, 1992), and α1β1 binding sites have been found in α1(I)CB3 and α1(I)CB8 (Gullberg et al., 1992). α1β1 and α2β1 binding sites have also been found on a CNBr fragment of type IV collagen (Vandenberg et al., 1991). However, the molecular details of the integrin-collagen interaction are still unclear as the studies to date either fail to cover the full molecule (Gullberg et al., 1990, 1992; Vandenberg et al., 1991) or conflict with existing data on the role of collagen conformation (Staatz et al., 1990).

To date, detailed studies of cell-collagen interactions have concentrated on type I collagen binding to platelets (Santoro, 1986; Staatz et al., 1990, 1991) or hepatocytes (Rubin et al., 1981; Gullberg et al., 1990, 1992), and binding of HT1080 fibrosarcoma cells and other cell types to type IV collagen (Vandenberg et al., 1991). The interaction of chondrocytic cells with type II collagen, as occurs in cartilage, has received relatively little attention because of the difficulties associated with working on these cells, particularly from human sources. Recently, a novel cell line, HCS-2/8, derived from a human chondrosarcoma, has been described (Takigawa et al., 1989). These cells resemble chondrocytes in that they synthesise collagen types II, IX and XI, and cartilage-type proteoglycans (Takigawa et al., 1989; D. S. Tuckwell, A. Marriott, S. Ayad, M. J. Humphries, unpublished work), and possess a response profile to various vitamins and growth factors similar to that of normal chondrocytes (Enomoto and Takigawa, 1992). These cells are thus an ideal model for normal chondrocytes. We have therefore used HCS-2/8 cells together with pepsinised bovine type II collagen to study the potential cell-matrix interactions occurring in cartilage.

We have determined the integrin profile for the HCS-2/8 cells and defined those receptors used for binding to native and denatured collagen. The receptor for native collagen was found to be the integrin α β, while cell binding to denatured collagen was found to be mediated by a novel α5β1-fibronectin bridge. Using spreading and attachment assays, binding to collagen was found to be highly conformation-dependent. Denatured CNBr fragments were unable to reproduce the integrin binding site found on native collagen, thereby demonstrating that integrin-collagen binding requires an intact triple helix.

Materials

Anti-human integrin and other antibody reagents were obtained from the following sources: anti-α1 (rabbit polyclonal) and anti-β3 (rat polyclonal) from Chemicon International Inc, London, UK; anti-α2 (mouse monoclonals), P1E8 from Gibco, Paisley, Scotland; 5E8 from R. Brankert, Roswell Park Memorial Institute, Buffalo, NY, USA; HAS6 from F. Watt, ICRF, Lincoln’s Inn Fields, London, UK (Tenchini et al., 1993); and 1C11 (S. A. Weston and M. J. Humphries, unpublished); anti-α3 (mouse monoclonal) P1B5 from GIBCO, Paisley, Scotland; anti-α4 (mouse monoclonal) HP 2/1 from Serotec, Oxford, UK; anti-α5 (rat monoclonal) Mab16 and anti-β1 (rat mono-clonal) Mab13 from K. Yamada and S. Akiyama, NIDR, NIH, Bethesda, MD, USA; anti-αv (mouse monoclonal) 147 from Calbiochem, La Jolla, CA, USA; anti-αv (mouse monoclonal) LM142 and anti-αvβ3 (mouse monoclonal) LM609 from D. Cheresh, Scripps Clinic, La Jolla, CA, USA; anti-fibronectin (mouse monoclonal) IST-2 from Seralab, Crawley Down, Sussex, UK and (goat polyclonal) Sigma, Poole, Dorset, UK; non-immune mouse IgG from ICN, High Wycombe, Bucks, UK; control mouse ascites from A. Ager, NIMR, Mill Hill, London, UK and J. L. Brookman, University of Manchester, Manchester, UK; non-immune rat IgG, non-immune goat IgG, FITC-labelled rabbit anti-goat IgG and TRITC-labelled rabbit anti-rat IgG from Sigma, Poole, Dorset, UK; FITC-labelled rabbit anti-mouse IgG and FITC-labelled rabbit anti-rat IgG from Serotec, Oxford, UK. All peptides were synthesised on an Applied Biosystems 431A peptide synthesiser and purified as described by Humphries et al. (1987). Peptide sequences were verified by a combination of sequencing, using an Applied Biosystems 476A protein sequencer, and fab-mass spectrometry (SERC service in Department of Chemistry, University of Manchester, Manchester, UK).

Cell culture

HCS-2/8 human chondrosarcoma cells (Takigawa et al., 1989) were maintained in Minimum Essential Medium with Earle’s Salts (MEM), 20% (v/v) foetal calf serum (FCS), 2 mM glutamine, 10 μg/ml gentamycin in 5% CO2/95% air, 100% humidity. Cells were grown past confluency until they had become rounded, but before nodules of cells had formed. At this point, cells were detached with 0.05% (w/v) trypsin, 0.02% (w/v) EDTA in Puck’s saline, and subcultured at a ratio of 1:3 to 1:6.

Immunoprecipitations

125I-labelled detergent-soluble cell extracts were prepared using the methods of Lebien et al. (1982) and Akiyama and Yamada (1987). Two 225 cm2 flasks of HCS-2/8 cells (approximately 0.5-1×108 cells) were used for each iodination. Flasks of cells were washed once with Hanks’ balanced salts solution (HBSS: 0.4 g/l KCl, 0.06 g/l KH2PO4, 8.0 g/l NaCl, 0.35 g/l NaHCO3, 0.09 g/l Na2HPO4.7H2O, 1.0 g/l D-glucose, 0.01 g/l Phenol Red) and the cells removed by treatment with 3 mM EDTA in HBSS. Cell suspensions were centrifuged (500 g for 5 minutes), washed with 30 ml ice-cold PBS without divalent cations (PBS: 0.2 g/l KCl, 0.2 g/l KH2PO4, 8.0 g/l NaCl, 2.16 g/l Na2HPO4.7H2O) and resuspended to 2×107 cells/ml in ice-cold PBS. Then 400 μl lactoperoxidase (1 mg/ml in PBS), 2 mCi Na125I (ICN, 2 1 High Wycombe, Bucks, UK) and 20 μl 0.12% (w/v) H2O2 in PBS. were added. The cell suspension was incubated on ice for 15 minutes with 20 μl 0.12% (w/v) H2O2 being added every 5 minutes. The iodination reaction was quenched by the addition of 10 ml PBS and the cells washed with 310 ml PBS. Cell pellets were then extracted with 2 ml extraction buffer (1% (w/v) Triton X-100, 0.13 M NaCl, 1 mM MnCl2, 1 mM MgCl2, 1 mM phenylmethylsulphonyl fluoride, 10 μg/ml leupeptin, 5 mg/ml ovalbumin, 20 mM Tris-HCl, pH 7.4) on ice for 30 minutes and insoluble material was removed by centrifugation (3000 g for 5 minutes). Remaining unincorporated iodide was removed by passing the cell extract down a PD10 column (Pharmacia, Knowlhill, Milton Keynes, UK) equilibrated in extraction buffer.

For immunoprecipitations, the 125I-labelled cell extract was first clarified by centrifugation at 14,000 g for 15 minutes, and then incubated at 4°C for 30 minutes with 10 μl 50% (v/v) Protein G-Sepharose (Pharmacia, Knowlhill, Milton Keynes, UK) in phosphate buffered saline with divalent cations (PBS+; 0.10 g/l CaCl2, 0.2 g/l KCl, 0.2 g/l KH2PO4, 0.10 g/l MgCl2.6H2O, 8.0 g/l NaCl, 2.16 g/l Na2HPO4.7H2O). The Protein G-Sepharose was removed by centrifugation at 14,000 g for 15 minutes, 50-100 μl samples of supernatant were transferred into tubes containing the same volume of antibody and the mixture was incubated on ice for 60 minutes. Final concentrations were 50 μg/ml for purified antibodies or 1:20 to 1:50 for ascites. An unlabelled cell extract was then prepared as above and incubated with Protein G-Sepharose, on ice for 30 minutes. This Protein G-Sepharose was then washed with 3×1 ml PBS+, 1% (w/v) Triton X-100 and 20 μl of a 50% (w/v) suspension in PBS+ was added to each immunoprecipitation sample. Mixtures were incubated on ice for 30 minutes and the Protein G-Sepharose was then washed with 5×1 ml PBS+, 1% (w/v) Triton X-100, suspended in 40 μl SDS-PAGE sample buffer (containing 2% mercaptoethanol if reduced samples were required) and boiled for 3 minutes.

SDS-PAGE was carried out using the method of Laemmli (1970). Gels were soaked in 10% (v/v) acetic acid, 25% (v/v) methanol and dried. Gels were then either exposed to X-ray film or analysed using a Molecular Dynamics phosphorimager.

Fluorescence activated cell scanning (FACS)

Flasks of HCS-2/8 cells were washed once with HBSS and cells detached with 3 mM EDTA in HBSS. Cell suspensions were centrifuged (500 g for 5 minutes) and resuspended at 107 cells/ml in MEM, 1% (v/v) FCS, 25 mM HEPES. For each antibody to be tested, 50 μl of cell suspension was mixed with 50 μl of the chosen antibody diluted in PBS+, 0.02% (w/v) sodium azide. Tubes were incubated on ice for 45-60 minutes, after which cells were washed with 3×300 μl PBS+, 1% (v/v) FCS in a Sorvall washing centrifuge. Cells were then incubated with an FITC-labelled second antibody (diluted 1:200 in PBS, 10% (v/v) human serum) for 30-45 minutes on ice. Cells were then washed as above with 2×300 μl PBS+, 1% FCS (v/v) followed by 1×300 μl PBS+ and were fixed by the sequential addition of 100 μl PBS, 100 μl 2% (w/v) formaldehyde in PBS, and 300 μl PBS+. Samples were read on a Becton Dickinson FACScan.

Preparation of type II collagen

Pepsinized type II collagen was prepared from foetal bovine articular cartilage as described by Ayad et al. (1981). Samples were at least 95% pure by SDS-PAGE. CNBr fragments of type II collagen were generated and purified as described by Piez (1968), Miller and Lunde (1973), and Miller and Rhodes (1982). No contaminating bands were visible in the fragment preparations on SDS-polyacrylamide gel electrophoresis.

Cell attachment and spreading assays

Type II collagen or its CNBr fragments were dissolved in 0.1 M acetic acid at a concentration of 2 mg/ml and diluted with PBS+ to the desired concentrations. Microtitre plates (Costar, Cambridge, MA, USA) were coated with 100 μl samples of collagens and incubated at room temperature for 1 hour. Denatured collagen was prepared by heating the collagen solution at 50°C for 30 minutes directly before coating, CNBr fragments were also heated before coating. The collagen solution was then removed and the wells blocked with 100 μl of heat-denatured BSA (10 mg/ml in PBS, filtered through a 0.22 μm filter and denatured by heating at 85°C for 12 minutes) at room temperature for 1 hour. Cells were removed from culture flasks with trypsin/EDTA, the trypsin was quenched with MEM, 20% (v/v) FCS, and the suspension centrifuged at 500 g for 5 minutes. Cell pellets were washed three times with MEM and resuspended to 2×105 cells/ml in MEM, 10 mg/ml BSA. To each well of the microtitre plate was first added 50 μl of antibody, CNBr fragment, or peptide inhibitor, dissolved in PBS+ (50 mM HEPES was also included in the case of CNBr peptides and peptide inhibitors, giving a final concentration of 25 mM), followed by 50 μl of cell suspension. Solutions of CNBr fragments were heated briefly at 50°C prior to addition to wells to assist their dissolution. The concentrations of all antibodies used were optimised to give the maximum inhibition. For those wells that were not to be treated, 50 μl of PBS+, with 50 mM HEPES if appropriate, were added. Plates were incubated under normal culture conditions for 3 hours. All treatments were carried out in duplicate or triplicate.

For attachment assays, wells were washed with 100 μl PBS+, and fixed by the addition of 10 μl 50% (w/v) glutaraldehyde. Wells were then washed with 3×100 μl distilled water, and stained with 0.1% (w/v) crystal violet in 200 mM MES, pH 6.0, for 20 minutes at room temperature. Unbound dye was washed off with 3×100 μl distilled water, and bound dye was solubilised with 100 μl 10% (w/v) acetic acid. Absorbance at 570 nm was read with an ELISA reader. To enable calibration, attachment assays included ‘blank’ wells, which were not coated with substratum and to which cells were not added, and ‘maximum’ wells, which were coated with 100 μg/ml concanavalin A to ensure that all cells adhered, and which were not washed prior to fixing. For spreading assays, wells were fixed by direct addition of 10 μl 50% (w/v) glutaraldehyde; 100 cells were counted per field of vision and the number of spread cells (phase-dark cells with a flattened or flattening morphology) was recorded.

For assays of antibody or peptide inhibition of attachment or spreading the percentage inhibition was calculated as:

[(uninhibited value − inhibited value)/(uninhibited value − BSA background)]×100.

Quantitation of collagen coating of microtitre plates

Type II collagen was labelled by reductive methylation of carboxyl groups using [3H]acetic anhydride (Amersham, Bucks., UK) as described by Gisslow and McBride (1975). Wells of microtitre plates were coated with the desired concentrations of native or denatured collagen and then blocked as for attachment/spreading assays above. The blocking solution was removed and wells were excised with a hot scalpel and placed in a tube containing scintillant. Wells were counted and the amount of protein attached was calculated from the known specific activity (cpm/mg protein) of the original labelled samples.

Immunofluorescence microscopy

Immunofluorescence experiments were designed to mimic conditions used for cell attachment and spreading assays and all reagents used were as for those assays. Coverslips were acid washed, rinsed, placed in 24-well tissue culture plates and coated with 0.5 ml native or heatdenatured collagen, at 50 μg/ml, for 1 hour. Coverslips were then blocked with heat-denatured BSA for 1 hour and HCS-2/8 cells, prepared as for cell attachment assays, were added to wells at 2-3 times the concentration used above to ensure sufficient cells for visualisation. Plates were incubated as for attachment assays, washed once with PBS and fixed in methanol at −20°C for 10 minutes.

Staining was carried out as follows: coverslips were washed three times with PBS, treated with 0.1% (v/v) Triton X-100 in PBS, washed three times with PBS and blocked with 3% (w/v) BSA in PBS (blocking buffer) for 1 hour. Coverslips were then incubated with first antibody (diluted in blocking buffer; P1E6 anti-α2, 1:200; goat anti-fibronectin, 1:100; Mab16 anti-α5, 1:50) for 1 hour, washed three times with PBS, incubated with second antibody (diluted in blocking buffer; FITC anti-mouse antibody, 1:50; TRITC anti-rat and FITC anti-goat, 1:100) for 1 hour, washed once in PBS, twice in water and mounted in PBS:glycerol (1:1, v/v), 1 mg/ml p-phenylenediamine. Coverslips were viewed on a Zeiss Axiophot microscope.

Ligand affinity chromatography

Native and denatured type II collagen were coupled to CNBr-activated Sepharose (Pharmacia, Knowlhill, Milton Keynes, UK) according to the manufacturer’s instructions. For denatured collagen, the solution was heated to 50°C for 30 minutes and then cooled on ice for 5 minutes before coupling.

Iodinated cell extracts, prepared as above, were centrifuged to remove debris (14,000 g for 15 minutes) and diluted 10-fold with extraction buffer without Triton X-100 or ovalbumin to reduce the Triton concentration to 0.1% (w/v). The diluted extract (typically 15-20 ml) was then incubated with 4 ml Sepharose 4B for 2 hours at 4°C to remove those proteins that bound non-specifically to Sepharose, after which the Sepharose was removed by centrifugation (500 g for 1 minute). The supernatant was then added to 5 ml of either the native collagen- or denatured collagen-Sepharose, pre-equilibrated in extraction buffer with 0.1% (w/v) Triton X-100 and without ovalbumin, and incubated at 4°C overnight. Each suspension was then poured into a column, washed with 10 volumes of extraction buffer with 0.1% Triton X-100 and without ovalbumin, and bound proteins were eluted with 0.1% (w/v) Triton X-100, 0.13 M NaCl, 10 mM EDTA, 1 mM phenylmethlysulphonyl fluoride, 10 μg/ml leupeptin, 20 mM Tris-HCl, pH 7.4. Divalent cations were restored to eluted fractions by the addition of MgCl2 to a final concentration of 12-15 mM, and Tris-HCl, pH 7.5, to 50 mM.

Samples of the column fractions were assayed for radioactivity using a γ-counter and the eluted iodinated protein peak was pooled and concentrated 5- to 10-fold using a Centricon 30 microconcentrator (Amicon, Beverley, MA, USA). Fractions were then analysed by SDS-PAGE or used for immunoprecipitations as above.

An integrin profile for HCS-2/8 cells

The integrin profile for HCS-2/8 cells was determined by immunoprecipitation and FACS analysis. From immunoprecipitations (Fig. 1), it was apparent that cells expressed α2β1 and α3β1, and both αvβ1 and αvβ3. Although immunoprecipitation with an anti-α1 antiserum gave no result, a weak high molecular mass band of approximately the correct molecular mass for α1 was seen in the β1 immunoprecipitate. The cells may therefore possibly express α1β1, but if so it is only present at a relatively low level. No bands were seen in immunoprecipitations with control non-immune antibodies.

Fig. 1.

SDS-PAGE of immunoprecipitation of HCS-2/8 cells with anti-integrin antibodies. Lane 1, anti-α1; lane 2, anti-α2 (1C11); lane 3, anti-α3; lane 4, anti-αv (147); lane 5, anti-β1; lane 6, anti-β3; lane 7, non-immune rabbit serum; lane 8, non-immune rat IgG; lane 9, non-immune mouse IgG.

Fig. 1.

SDS-PAGE of immunoprecipitation of HCS-2/8 cells with anti-integrin antibodies. Lane 1, anti-α1; lane 2, anti-α2 (1C11); lane 3, anti-α3; lane 4, anti-αv (147); lane 5, anti-β1; lane 6, anti-β3; lane 7, non-immune rabbit serum; lane 8, non-immune rat IgG; lane 9, non-immune mouse IgG.

FACS data (Fig. 2) also identified α2, αv and β1 integrins and, in addition, showed α5 to be present. Consistent with its restricted cell-type distribution, the α4 subunit was not detected. From FACS analysis, it was also apparent that fibronectin was associated with some, although not all, cells.

Fig. 2.

FACS analysis of HCS-2/8 cells. (A) Mouse anti-integrin antibodies; (—) anti-α2; (....) anti-α4;.(.....) anti-αv (147); (‐ ‐ ‐ ‐) anti-fibronectin (IST-2). (B) Mouse antibody control. (C) Rat anti-integrin antibodies; (—) anti-α5; (....) anti-β1; (.....) non-immune rat IgG, (‐ ‐ ‐ ‐) no first antibody.

Fig. 2.

FACS analysis of HCS-2/8 cells. (A) Mouse anti-integrin antibodies; (—) anti-α2; (....) anti-α4;.(.....) anti-αv (147); (‐ ‐ ‐ ‐) anti-fibronectin (IST-2). (B) Mouse antibody control. (C) Rat anti-integrin antibodies; (—) anti-α5; (....) anti-β1; (.....) non-immune rat IgG, (‐ ‐ ‐ ‐) no first antibody.

The interaction of HCS-2/8 cells with type II collagen

To determine the role of collagen structure in the binding of HCS-2/8 cells, attachment and spreading assays were carried out using both native and heat-denatured pepsinised bovine type II collagen. Cells attached to both native and denatured collagen with similar dose-response curves, each showing maximal levels of attachment of ∼60% (Fig. 3A). Spreading of cells on collagen, however, was affected by collagen conformation, with native collagen supporting ∼40% spreading compared to only ∼20% spreading on denatured collagen (Fig. 3B).

Fig. 3.

Adhesion of HCS-2/8 cells to native and denatured pepsinised bovine type II collagen. (A) Attachment of cells to native and denatured collagen (mean ± s.e.m, n=8 from 3 experiments). (B) Spreading of cells on native and denatured collagen (mean ± s.e.m, n=6 from 2 experiments). (C) Efficiency of coating of microtitre plates by native and denatured collagen (mean ± s.e.m, n=7 from 3 experiments).

Fig. 3.

Adhesion of HCS-2/8 cells to native and denatured pepsinised bovine type II collagen. (A) Attachment of cells to native and denatured collagen (mean ± s.e.m, n=8 from 3 experiments). (B) Spreading of cells on native and denatured collagen (mean ± s.e.m, n=6 from 2 experiments). (C) Efficiency of coating of microtitre plates by native and denatured collagen (mean ± s.e.m, n=7 from 3 experiments).

From the above data, it was apparent that collagen conformation was important in spreading of HCS-2/8 cells. Studies of the efficiency of collagen coating of the microtitre plates used in the assays showed that, although native and denatured collagen did not coat with identical efficiencies (Fig. 3C), the difference between them was insufficient to account for differences in experimental data.

The interaction of HCS-2/8 cells with cyanogen bromide (CNBr) fragments of type II collagen

In order to study further the role of collagen structure in cell binding and to localise cell binding sites within the collagen molecule, a series of CNBr fragments of type II collagen were prepared, which together covered 95% of the parent pepsinised molecule (Fig. 4A). These fragments were tested in attachment and spreading assays.

Fig. 4.

Adhesion of HCS-2/8 cells to CNBr fragments of type II collagen. (A) The CNBr fragments of pepsinised type II collagen. The CNBr fragments of the pepsinised α1 chain of type I collagen are shown for comparison. (B) Attachment to CNBr fragments (mean ± s.e.m, n=6 from 2 experiments). (C) Spreading on CNBr fragments (mean ± s.e.m, n=6 from 2 experiments); native and denatured collagen and CNBr fragments were coated at 12.5 μg/ml.

Fig. 4.

Adhesion of HCS-2/8 cells to CNBr fragments of type II collagen. (A) The CNBr fragments of pepsinised type II collagen. The CNBr fragments of the pepsinised α1 chain of type I collagen are shown for comparison. (B) Attachment to CNBr fragments (mean ± s.e.m, n=6 from 2 experiments). (C) Spreading on CNBr fragments (mean ± s.e.m, n=6 from 2 experiments); native and denatured collagen and CNBr fragments were coated at 12.5 μg/ml.

All of the CNBr fragments (which were heat-denatured prior to use) supported cell attachment (Fig. 4B), although the fragment CB10 showed the greatest activity, being approximately equivalent to native or denatured collagen. The other three fragment preparations, CB11, CB12 and the mixed preparation containing both CB8 and CB9,7 (CB8+9,7), showed reduced activity compared to the parent molecule, but still supported cell attachment. An unfractionated CNBr digest was found to support a level of attachment intermediate between that of CB10 and the other fragments, consistent with there being no major cell binding sites in the 5% of the molecule that was not recovered as CNBr fragments.

Cell spreading assays showed that, of the CNBr fragments, only CB10 supported cell spreading (Fig. 4C). Again, an unfractionated CNBr digest gave a value intermediate between CB10 and the other fragments. Since spreading is a co-operative process relying on combined high-affinity interactions, a low level of cell attachment, as seen for CB11, CB12 and CB8+9,7, might be of too low an affinity to progress to cell spreading, thus accounting for the differing behaviour of these fragments in the two assays. From the cell attachment and spreading assays it was apparent that the major cell binding site on type II collagen was in CB10, although cell binding sites were clearly present on the other fragments.

To establish the relationship between the mechanism or mechanisms used for cell binding to the CNBr fragments and those used for binding to native and denatured collagen, the CNBr fragments were tested for their ability to inhibit cell attachment and spreading. All the fragment preparations inhibited cell spreading on denatured collagen by 50%-80%, with CB10 and CB8+9,7 giving the greatest inhibition (Fig. 5A). None of the fragments, however, inhibited spreading on native collagen. Similar data were obtained for inhibition of attachment (Fig. 5B) and although CB11 did not inhibit attachment to denatured collagen, this is consistent with its being the least inhibitory CNBr fragment in spreading assays. An unfractionated CNBr digest showed the same inhibition of attachment as the CNBr fragments, again showing that no additional sites were missing from the purified fragments (data not shown).

Fig. 5.

Inhibition of adhesion of HCS-2/8 cells to native and denatured collagen by CNBr fragments. (A) Inhibition of spreading by CNBr fragments (mean ± s.e.m, n=6 from 2 experiments). (B) Inhibition of attachment by CNBr fragments (mean ± s.e.m, n=5 from 2 experiments). Wells were coated with 12.5 μg/ml collagen. Fragments were added at ∼0.1 mM, i.e. 3 mg/ml, for CB10, CB11 and CB8+9,7, 1 mg/ml for CB12.

Fig. 5.

Inhibition of adhesion of HCS-2/8 cells to native and denatured collagen by CNBr fragments. (A) Inhibition of spreading by CNBr fragments (mean ± s.e.m, n=6 from 2 experiments). (B) Inhibition of attachment by CNBr fragments (mean ± s.e.m, n=5 from 2 experiments). Wells were coated with 12.5 μg/ml collagen. Fragments were added at ∼0.1 mM, i.e. 3 mg/ml, for CB10, CB11 and CB8+9,7, 1 mg/ml for CB12.

Taken together, these results indicate that the mechanism used by HCS-2/8 cells to interact with CNBr fragments is identical to that used for cell binding to denatured collagen, while the cell binding site on native collagen is not represented in any of the fragment preparations.

The mechanism of HCS-2/8 binding to native collagen

A number of integrins have been recognised as collagen receptors (reviewed by Akiyama et al., 1990 and Tuckwell and Humphries, 1993), in particular α2β1 and α3β1, receptors for native collagen, and αvβ1 and αvβ3, receptors for denatured collagen, all of which are present on HCS-2/8 cells. α1β1 is also known to be a receptor for native collagen, but HCS-2/8 cells express little if any of this receptor (see above). The role of β1 integrins in the interaction of HCS-2/8 cells with collagen was investigated using an anti-functional anti-β1 monoclonal antibody. In attachment assays, this antibody completely inhibited attachment to both native and denatured collagen (Fig. 6), suggesting that β1 integrins are responsible for the interaction with both forms of collagen.

Fig. 6.

Inhibition of attachment of HCS-2/8 cells to native and denatured collagen by the anti-functional anti-β1 integrin antibody Mab13 (mean ± s.e.m, n=6 from 2 experiments). Wells were coated with 50 μg/ml collagen. Mab13 was added at 10 μg/ml, control rat IgG at 20 μg/ml.

Fig. 6.

Inhibition of attachment of HCS-2/8 cells to native and denatured collagen by the anti-functional anti-β1 integrin antibody Mab13 (mean ± s.e.m, n=6 from 2 experiments). Wells were coated with 50 μg/ml collagen. Mab13 was added at 10 μg/ml, control rat IgG at 20 μg/ml.

To establish whether or not integrin α2β1 was involved in binding to native collagen, an anti-functional anti-α2 mono-clonal antibody was tested in attachment assays. This antibody inhibited attachment to native collagen by ∼60% (Fig. 7A). Consistent with the observed difference between the mechanisms of attachment to native and denatured collagen, attachment to denatured collagen was only inhibited by ∼10%, and there was no significant effect on attachment to an unfractionated CNBr digest or to CB10. Inhibition of attachment to the other CNBr fragment preparations by the anti-functional anti-α2 monoclonal antibody was difficult to assess because of the low level of attachment that these fragments support. However, attachment to CB11 and CB12 was not significantly affected by the anti-α2 antibody although there did appear to be some inhibition of attachment to CB8+9,7 (data not shown). As spreading assays are more sensitive to inhibition than attachment assays, inhibition of spreading on native collagen was studied to examine the extent to which α2β1 was responsible for the interaction of HCS-2/8 cells with native collagen. Spreading on native collagen was inhibited by ∼80%, compared with only ∼20% inhibition of spreading on denatured collagen (Fig. 7B). The role of integrin α2β1 in cell adhesion to collagen was also investigated using immunofluorescence staining of HCS-2/8 cells plated onto collagen-coated coverslips (Fig. 8). α2-containing focal contacts were seen in cells plated on native collagen whereas far fewer α2-containing focal contacts were visible in cells plated on denatured collagen. These results indicate that α2β1 is the dominant functional receptor for native collagen on these cells.

Fig. 7.

Inhibition of HCS-2/8 adhesion to collagen by the anti-functional anti-α2 antibody HAS6. (A) Inhibition of attachment to native and denatured collagen and CNBr fragments by HAS6 (mean ± s.e.m, n=6 from 2 experiments). Wells were coated with 50 μg/ml substratum and antibodies were added at 5 μg/ml. (B) Inhibition of spreading on native and denatured collagen by HAS6. Wells were coated with 12.5 μg/ml substratum. HAS6 was added at 10 μg/ml (mean ± s.e.m, n=5 from 2 experiments) as was mouse IgG (mean ± s.e.m, n=2 from 1 experiment).

Fig. 7.

Inhibition of HCS-2/8 adhesion to collagen by the anti-functional anti-α2 antibody HAS6. (A) Inhibition of attachment to native and denatured collagen and CNBr fragments by HAS6 (mean ± s.e.m, n=6 from 2 experiments). Wells were coated with 50 μg/ml substratum and antibodies were added at 5 μg/ml. (B) Inhibition of spreading on native and denatured collagen by HAS6. Wells were coated with 12.5 μg/ml substratum. HAS6 was added at 10 μg/ml (mean ± s.e.m, n=5 from 2 experiments) as was mouse IgG (mean ± s.e.m, n=2 from 1 experiment).

Fig. 8.

Immunofluorescence staining of HCS-2/8 cells plated onto native (A,C,D,F) or denatured (B,E) collagen and stained with anti-α2 antibody (P1E6; A,B). No-first antibody control is shown in (C). Corresponding phase-contrast photographs are shown (D,E,F). Focal contacts are marked with arrowheads. Bar, 10 μm.

Fig. 8.

Immunofluorescence staining of HCS-2/8 cells plated onto native (A,C,D,F) or denatured (B,E) collagen and stained with anti-α2 antibody (P1E6; A,B). No-first antibody control is shown in (C). Corresponding phase-contrast photographs are shown (D,E,F). Focal contacts are marked with arrowheads. Bar, 10 μm.

In order to determine whether any other receptors were involved in binding to native collagen, ligand affinity chromatography was carried out using a native type II collagen-Sepharose column. Elution with EDTA yielded two protein bands on SDS-PAGE, molecular masses 137 kDa and 112 kDa (non-reduced), 151 kDa and 126 kDa (reduced), similar to that expected for α2β1 (Hemler, 1990; Fig. 9). Immunoprecipitation of the EDTA-eluted fraction with anti-integrin antibodies gave a precipitate (which comigrated with the original eluted proteins) with an anti-α2 antibody, but not with antibodies against α3 or αv (Fig. 9), confirming the identity of the eluted receptor as α2β1. In the HCS-2/8 cells the sole receptor for native collagen is therefore seen to be α2β1.

Fig. 9.

Ligand-affinity chromatography of HCS-2/8 cells using a native type II collagen-Sepharose column. Lane 1, EDTA-eluted fraction (non-reduced); lane 2, EDTA-eluted fraction (reduced); lane 3, immunoprecipitate with anti-α2 antibody (HAS6); lane 4, immunoprecipitate with anti-α3 antibody; lane 5, immunoprecipitate with anti-αv antibody (147); lane 6, immunoprecipitate with non-immune mouse IgG.

Fig. 9.

Ligand-affinity chromatography of HCS-2/8 cells using a native type II collagen-Sepharose column. Lane 1, EDTA-eluted fraction (non-reduced); lane 2, EDTA-eluted fraction (reduced); lane 3, immunoprecipitate with anti-α2 antibody (HAS6); lane 4, immunoprecipitate with anti-α3 antibody; lane 5, immunoprecipitate with anti-αv antibody (147); lane 6, immunoprecipitate with non-immune mouse IgG.

In a recent study, Staatz et al. (1991) localised the α2β1 binding site on type I collagen to the peptide KDGEA. This peptide was therefore tested for its ability to inhibit the attachment of HCS-2/8 cells to native and denatured collagen (Fig. 10A), but neither this peptide nor the control peptide KAGEA, at concentrations up to 5 mM, had an effect on cell attachment to either substratum. This peptide sequence is therefore apparently not functional in the system studied here.

Fig. 10.

Inhibition of attachment to collagen by synthetic peptides. (A) Inhibition of attachment to native and denatured collagen by the peptide KDGEA and the control peptide KAGEA (mean ± s.e.m, n=6 from 2 experiments). (B) Inhibition of attachment to native and denatured collagen by the peptide GRGDS and the control peptide GRGES (mean s.e.m, n=6 from 2 experiments). Wells were coated with 50 μg/ml collagen.

Fig. 10.

Inhibition of attachment to collagen by synthetic peptides. (A) Inhibition of attachment to native and denatured collagen by the peptide KDGEA and the control peptide KAGEA (mean ± s.e.m, n=6 from 2 experiments). (B) Inhibition of attachment to native and denatured collagen by the peptide GRGDS and the control peptide GRGES (mean s.e.m, n=6 from 2 experiments). Wells were coated with 50 μg/ml collagen.

The mechanism of HCS-2/8 binding to denatured collagen

As described above, all CNBr fragment preparations supported cell binding and inhibited cell binding to denatured collagen. However, CB10 was the most potent fragment, and CB10 is also known to contain the principal fibronectin-binding site in type II collagen (Kleinman et al., 1978a). A possible role for fibronectin in cell binding to denatured collagen was therefore investigated. Cell attachment to denatured collagen was inhibited by ∼50% by an anti-fibronectin polyclonal antiserum, and attachment to the CNBr fragments and to the unfractionated CNBr digest was inhibited by 80-100% (Fig. 11A). As expected, attachment to native collagen was unaffected.

Fig. 11.

Inhibition of attachment to collagen by antibodies against fibronectin and integrin α5. (A) Inhibition of attachment to native and denatured collagen and CNBr fragments by an anti-fibronectin polyclonal antiserum (mean ± s.e.m, n=6 from 2 experiments). Wells were coated with 50 μg/ml collagen and anti-fibronectin antiserum was used at a dilution of 1:100. Inhibitions of greater than 100% were found for CB11, CB8+9,7 and CB12 because of the low percentage attachment to these fragments. These values have been adjusted to 100%. Inhibition of attachment to native and denatured collagen and CNBr fragments by the anti-functional anti-α5 integrin antibody Mab16 (mean ± s.e.m, n=6 from 2 experiments). Wells were coated with 50 μg/ml collagen, antibodies were added at 20 μg/ml. The high levels of control inhibition seen with CB11, CB8+9,7 and CB12 are a consequence of the relatively low level of attachment supported by these fragments, leading to an increased sensitivity to non-specific inhibition.

Fig. 11.

Inhibition of attachment to collagen by antibodies against fibronectin and integrin α5. (A) Inhibition of attachment to native and denatured collagen and CNBr fragments by an anti-fibronectin polyclonal antiserum (mean ± s.e.m, n=6 from 2 experiments). Wells were coated with 50 μg/ml collagen and anti-fibronectin antiserum was used at a dilution of 1:100. Inhibitions of greater than 100% were found for CB11, CB8+9,7 and CB12 because of the low percentage attachment to these fragments. These values have been adjusted to 100%. Inhibition of attachment to native and denatured collagen and CNBr fragments by the anti-functional anti-α5 integrin antibody Mab16 (mean ± s.e.m, n=6 from 2 experiments). Wells were coated with 50 μg/ml collagen, antibodies were added at 20 μg/ml. The high levels of control inhibition seen with CB11, CB8+9,7 and CB12 are a consequence of the relatively low level of attachment supported by these fragments, leading to an increased sensitivity to non-specific inhibition.

The role of β1 integrins in HCS-2/8 binding to denatured collagen has been demonstrated above (Fig. 6). Cell binding to fibronectin is known to be mediated by a number of integrins, in particular α5β1. It was therefore suggested that HCS-2/8 cells might interact via α5β1 with fibronectin, which itself binds to denatured collagen. An anti-functional anti-α5 mono-clonal antibody was found to inhibit attachment to denatured collagen, the unfractionated CNBr digest and CB10 by 50%-70% (Fig. 11B), while attachment to native collagen was unaffected. Attachment to CB11, CB12 and CB8+9,7 was not significantly affected by the antibody compared to control IgG. However, it should be stressed that these results are complicated by the low control levels of adhesion and a resulting high sensitivity to inhibition. In order to provide further evidence for this mechanism, immunoflourescence staining was carried out on cells plated on denatured collagen (Fig. 12). Colocalisation of fibronectin with α5 integrin in focal contacts could be seen, as well as large amounts of fibronectin associated with and deposited round the cells. These data suggested that the interaction of HCS-2/8 cells with denatured collagen was mediated, at least in part, by a novel α5β1-fibronectin bridge. This would also be the mechanism for attachment to CB10, and may also be the mechanism for attachment to the other fragments.

Fig. 12.

Immunofluorescence staining of HCS-2/8 cells plated onto denatured collagen and stained with anti-fibronectin antiserum (A,D) or anti-α5 monoclonal antibody (B,E). Corresponding phase-contrast photographs are shown in (C,F). No first antibody staining was negligible. Focal contacts containing colocalizations of fibronectin and α5 are marked with arrowheads. Bar, 10 μm.

Fig. 12.

Immunofluorescence staining of HCS-2/8 cells plated onto denatured collagen and stained with anti-fibronectin antiserum (A,D) or anti-α5 monoclonal antibody (B,E). Corresponding phase-contrast photographs are shown in (C,F). No first antibody staining was negligible. Focal contacts containing colocalizations of fibronectin and α5 are marked with arrowheads. Bar, 10 μm.

To test whether an α5β1-fibronectin bridge was the sole mechanism responsible for cell binding to denatured collagen, the effects of a series of anti-integrin antibodies, together with the anti-fibronectin antiserum described above, were tested for their effects on spreading. In particular, since type II collagen contains four RGD sequences, the αv integrins present on the HCS-2/8 cells, αvβ3 and αvβ1, might make a contribution to the recognition of the denatured molecule (Dedhar and Gray, 1990; Davis, 1992). The anti-fibronectin polyclonal antiserum inhibited spreading on denatured collagen by ∼70% and the anti-α5β1 antibody inhibited spreading by ∼80% (Fig. 13). Anti-functional monoclonal antibodies against αv and αvβ3 were, however, ineffective. An anti-functional anti-αv antibody was also unable to inhibit attachment to denatured collagen (data not shown). αv integrins therefore do not appear to be involved in HCS-2/8 binding to denatured collagen, and the major receptor for this ligand appears to be the α5β1-fibronectin bridge.

Fig. 13.

Inhibition of spreading of HCS-2/8 cells on denatured collagen by a polyclonal antiserum to fibronectin and anti-functional antibodies against integrins α5, αv (LM142) and αvβ3 (mean ± s.e.m, n=8 from 2 experiments). Wells were coated with 12.5 μg/ml collagen, anti-fibronectin antiserum was used at a dilution of 1/100, anti-α5 antibody was used at 20 μg/ml, and anti-αv and -αvβ3 ascites were added at dilutions of 1:1000 and 1:5000.

Fig. 13.

Inhibition of spreading of HCS-2/8 cells on denatured collagen by a polyclonal antiserum to fibronectin and anti-functional antibodies against integrins α5, αv (LM142) and αvβ3 (mean ± s.e.m, n=8 from 2 experiments). Wells were coated with 12.5 μg/ml collagen, anti-fibronectin antiserum was used at a dilution of 1/100, anti-α5 antibody was used at 20 μg/ml, and anti-αv and -αvβ3 ascites were added at dilutions of 1:1000 and 1:5000.

In order to ascertain whether any other receptors were involved in cell binding to denatured collagen, and to confirm the lack of involvement of αv integrins in this process, ligand affinity chromatography was carried out using a denatured collagen II-Sepharose column. Elution with EDTA revealed two bands, with molecular masses 144 kDa and 119 kDa (non-reduced), and 159 kDa and 126 kDa (reduced), which resembled α2β1 (data not shown). This result is probably explained by the binding of α2β1 to a small proportion of renatured collagen within the column. The absence of αv integrin, which has a characteristic decrease in molecular mass on reduction (Hemler, 1990), is consistent with the antibody data. The known low affinity of α5β1 for fibronectin accounts for its absence from the eluted fraction, as it is established that α5β1 cannot be purified by fibronectin affinity chromatography. We were therefore unable to identify any receptor for denatured collagen other than the α5β1-fibronectin bridge. The incomplete inhibition of attachment seen with anti-α5 and -β1 antibodies could indicate an additional receptor. However, these data may well be a consequence of weak antibody affinities.

The α5β1-fibronectin interaction is partly mediated by the sequence GRGDS in fibronectin. The ability of GRGDS peptide to inhibit the interaction of HCS-2/8 cells with denatured collagen was therefore investigated. Consistent with the above data, GRGDS peptide (0.5 mM) inhibited cell attachment to denatured collagen by 40% while the control peptide GRGES had no effect. In common with the findings of many other workers, neither peptide inhibited attachment to native collagen (Fig. 10B).

In this study, we have elucidated the mechanisms used by HCS-2/8 human chondrosarcoma cells to bind to native and denatured type II collagen. Attachment and spreading assays showed a strict dependence on collagen conformation, and adhesion to native collagen was mediated by a different mechanism to that used for adhesion to denatured collagen or to denatured CNBr fragments: HCS-2/8 cells bind native collagen via α2β1, and bind denatured collagen via a novel α5β1-fibronectin bridge. Integrin binding to native collagens therefore requires an intact triple helix. These studies suggest that the classical approach of identifying ligand active sites by truncation may not be always applicable to collagens.

HCS-2/8 chondrosarcoma cells

Few studies of cell attachment to collagen have addressed the integrin-mediated interaction of chondrocytes with type II collagen, for technical and practical reasons (for exceptions, see Mollenhauer and von der Mark, 1983; Enomoto et al., 1993). Using the HCS-2/8 cell line, we have been able to determine those mechanisms used by chondrocytic cells in their interactions with collagen. These data confirm and extend those of Enomoto et al. (1993), who showed that chick chondrocytes interact with type II collagen via β1 integrins. Tumour cells can often differ in integrin profile from their non-tumour progenitors (Marshall et al., 1991; Nip et al., 1992; Verfaille et al., 1992). However, HCS-2/8 cells resemble normal chondrocytes in synthesising collagen types II, IX and XI rather than type I (Takigawa et al., 1989; D. S. Tuckwell, A. Marriott,S. Ayad and M. J. Humphries, unpublished work), and the integrin profile of HCS-2/8 cells resembles that of chick chondrocytes, which have been shown to express β1, α3, α5 and an integrin subunit resembling α2, but not α1 (Enomoto et al., 1993). A study of human articular chondrocytes gave slightly different results, with these cells expressing α1β1, α3β1 and α5β1, but not α4β1 or α2β1 (Salter et al., 1992). These differences may be partly explained by the techniques used: Salter et al. determined integrin profiles by immunohistochemistry of frozen sections, in which antigen masking is a potential problem, whereas FACS or immunoprecipitations were used here and by Enomoto et al. Taking these data together, our characterisation of the HCS-2/8 cell line suggests that it is a good model for normal chondrocytes.

Mechanisms for cell attachment to collagen

Attachment and spreading assays demonstrated that adhesion to native collagen was mediated by one mechanism while adhesion to denatured collagen and denatured CNBr fragments was mediated by another. Ligand affinity chromatography and the use of an anti-functional anti-α2 antibody showed that the sole receptor for native collagen on HCS-2/8 cells was the integrin α2β1, This integrin is well established as a receptor for collagen types I, II, III, IV and VI (Wayner and Carter, 1987; Kramer and Marks, 1989; Staatz et al., 1989). Adhesion to native collagen was found to be RGD-peptide-independent, in line with the findings of many others (reviewed by Tuckwell and Humphries, 1993). However, there may be some role for RGD-like sequences in the interaction of collagens with integrins, as Cardarelli et al. (1992) have shown that a cyclic RGD analogue inhibits α2β1 binding to collagen, and Eble et al. (1993) have convincingly demonstrated a role for arginine and aspartate side-chains in the interaction of α1β1 with collagen.

Cell adhesion to both denatured collagen and to its CNBr fragments were mediated by the same mechanism. The observation that the major cell binding site was localised to the CNBr fragment CB10, coupled with the published evidence for this fragment containing the site at which fibronectin binds to denatured collagen (Kleinman et al., 1978a), suggested that HCS-2/8 binding to denatured collagen might be mediated by fibronectin. Antibody experiments and fluorescence microscopy data indicated that a novel α5β1−fibronectin bridge mediated HCS-2/8 cell binding to denatured collagen, with the fibronectin binding to the denatured collagen and α5β1 binding in turn to the fibronectin. Primary chondrocytes have been demonstrated to synthesise fibronectin in vitro after release from the cartilage matrix (Dessau et al., 1978) and it is apparent from FACS and immunofluorescence data that fibronectin is associated with HCS-2/8 cells in this study. Consistent with this mechanism, addition of fibronectin to HCS-2/8 attachment assays increased cell attachment to denatured collagen by approximately 100% while increasing attachment to native collagen by less than 10% (data not shown).

Fibronectin was recognised some time ago as an exogenous or endogenous factor enabling cells to bind to denatured collagen (Grinnell, 1978; Kleinman et al., 1978b; Yamada and Olden, 1978; Grinnell and Feld, 1979, 1980). However, it is a novel observation that the cell receptor involved in fibronectin-mediated attachment to denatured collagen is the integrin α5β1, The involvement of integrins is consistent with the requirement for divalent cations and the expenditure of metabolic energy associated with fibronectin-mediated adhesion to denatured collagen (reviewed by Yamada and Olden, 1978). A precedent for the use of a protein bridge to mediate cell-collagen interactions is the αIIbβ3-von Willebrand factor bridge proposed to mediate platelet adhesion to native collagen (Coller et al., 1989; Fressinaud et al., 1990). The major fibronectin binding site in denatured type II collagen is present within CB10, and a second site has been reported in CB12 (Guidry et al., 1990). In this study, however, attachment of HCS-2/8 cells to all of the CNBr fragments other than CB10 was similar and was inhibited by the anti-fibronectin antiserum. This suggests that, under our experimental conditions, all denatured collagen fragments exhibit a low-affinity interaction with fibronectin that is sufficient to produce some cell attachment.

The α5β1-fibronectin bridge adds to the mechanisms by which cells can bind, either directly or indirectly, to denatured collagen. αvβ1 and αvβ3 have previously been recognised as receptors for denatured collagen, apparently binding to RGD sequences exposed on denaturation (Dedhar and Gray, 1990; Davis, 1992). However, although HCS-2/8 cells express both αvβ1 and αvβ3, it is to be noted that these integrins account for little if any of the adhesion to denatured collagen. Since many mesenchymal cells synthesise high levels of fibronectin in vivo, it may be that the fibronectin–α5β1 bridge is at least as important as αv-mediated binding in those situations where collagen denaturation takes place.

The role of collagen structure in cell attachment

The importance of collagen structure in cell attachment has been recognised for some time, e.g. for collagens type I, IV and VI (Schor and Court, 1979; Rubin et al., 1981; Aumailley and Timpl, 1986; Santoro, 1986; Aumailley et al., 1989; reviewed by Tuckwell and Humphries, 1993). A number of studies have employed CNBr cleavage of collagens as an approach for identifying cell binding sites. The use of these fragments poses potential problems as a decrease in helix length leads to a decrease in helix melting point (reviewed by Traub and Piez, 1971), which could in turn lead to denaturation under experimental conditions. In addition, the preparation of CNBr fragments requires denaturing conditions. The structure of the CNBr fragments is therefore unclear. In order to use CNBr fragments of known structure, we have diverged from other studies and employed heat-denatured CNBr fragments of type II collagen to investigate the importance of collagen structure in cell binding. When denatured CNBr fragments were tested in spreading and attachment assays, the major cell binding site was found to be within the fragment CB10, although the other fragment preparations were able to support a low level of cell attachment. When the denatured fragments were used to inhibit spreading, all fragments were able to inhibit cell spreading on denatured collagen, but none inhibited spreading on native collagen (similar results were obtained from attachment assays). From this it can be seen that binding to denatured collagen and CNBr fragments was mediated by the same mechanism while none of the fragments reproduced the binding site found on native collagen. This strict conformation dependence was also reflected in the use of different receptors for binding to native or denatured collagen. A binding site for the HCS-2/8 native collagen receptor, α2β1, was therefore not present in any of the fragment preparations. These studies demonstrate that integrin binding to collagen requires an intact triple-helical binding site and that disruption of the triple-helical structure abolishes or at least severely reduces specific binding. Other studies of cell-collagen interactions have employed type I collagen, which shows considerable homology to type II collagen. In common with other integrin ligands, binding of platelets to type I collagen requires divalent cations, and is supported by magnesium ions, but only poorly by calcium ions (Santoro, 1986). Staatz et al. (1990) studied the magnesium-dependent adhesion of platelets to collagen and found not only that heat-denatured collagen supported adhesion, but that the cell attachment activity of the collagen molecule, mediated by α2β1, was almost all contained within the CNBr fragmentα1(I)CB3.

The data reported here clearly differ from those reported by Staatz et al. (1990) in the matters of conformation dependence and the number of fragments supporting adhesion. To eliminate the possibility that any of these differences were due to choice of cations in the assays, the effects of magnesium and calcium ions on attachment to native and denatured collagen and to the CNBr fragment preparations were tested: while magnesium ions were found to support a greater percentage of cell attachment than calcium ions, no change in the overall profiles, compared to attachment in the presence of both cations, was seen (data not shown), therefore eliminating this possibility. In a later study, Staatz et al. (1991) localised the α2β1-binding site within α1(I)CB3 to the peptide KDGEA, which was able to inhibit the platelet-collagen interaction, albeit at high concentration. Type II collagen shows considerable homology to type I collagen, and the corresponding sequence in type II collagen is KDGET. Consistent with the differences between our work and that of Staatz et al. discussed above, the KDGEA peptide did not inhibit α2β1-mediated HCS-2/8 cell-collagen binding. The inability of KDGEA to inhibit cell-collagen binding has also been reported elsewhere (Cardarelli et al., 1992). It is therefore apparent that the magnesium-dependent attachment of platelets to type I collagen differs from the HCS-2/8 cell-type II collagen system and other systems in a number of important respects, and may not fully reflect the molecular events of integrin-collagen binding.

In a study of hepatocyte adhesion to type I collagen, Rubin et al. (1981) observed that adhesion to native and denatured collagen were not equivalent. Studies of the adhesive function of a range of CNBr fragments from α1(I) and α2(I), covering most of the former and over half of the latter, showed that all fragments supported cell attachment although the receptors involved were not identified. In later studies of α1(I)CB3 and α1(I)CB8, α1β1 binding to both fragments and α2β1 binding to α1(I)CB3, but not α1(I)CB8, was observed (Gullberg et al., 1990, 1992), in agreement with the report of Staatz et al. (1990).

The data reported here are in agreement with those of Rubin et al. (1981), since collagen denaturation affected adhesion and all CNBr fragments supported cell attachment. In the study by Gullberg et al. (1992), attachment assays were carried out using α1(I)CB3 and α1(I)CB8 stabilised by glutaraldehyde cross-linking to BSA-coated dishes. Because of the denaturing conditions employed in the preparation of collagen CNBr fragments, Gullberg et al. interpreted their data by proposing that the integrin binding sites on type I collagen are conformation-dependent and suggested that their CNBr fragments had renatured. Our conclusions therefore agree with and extend the data of Gullberg et al. in demonstrating conformation dependence in integrin binding to collagen and its CNBr fragments. Future work must now aim to test rigorously whether integrin binding sites can be localised to particular regions of the fibrillar collagen helix and, if so, to define the precise sequences involved. Crucially, the demonstration of conformation dependence suggests that linear peptides, including both CNBr fragments and short peptides such as KDGEA, may be unsuitable as inhibitors of cell-collagen interactions and implies that more sophisticated methods may be necessary for future studies.

This work was supported by the Wellcome Trust. The authors acknowledge the assistance of the following: Fiona Watt, David Cheresh, Richard Brankert, Steve Akiyama, Ken Yamada (gifts of antibodies); Adrian Shuttleworth (protein chemistry); Linda Berry (peptide sequencing); Phil Cheeseman (figures); Andy Povey (phosphorimager facility); Keith Matthews, Iain Hagan (immunofluorescence microscopy).

Akiyama
,
S. K.
and
Yamada
,
K. M.
(
1987
).
Biosynthesis and acquisition of biological activity of the fibronectin receptor
.
J. Biol. Chem
.
262
,
17536
17542
.
Akiyama
,
S. K.
,
Nagata
,
K.
and
Yamada
,
K. M.
(
1990
).
Cell surface receptors for extracellular matrix components
.
Biochim. Biophys. Acta
1031
,
91
110
.
Aumailley
,
M.
and
Timpl
,
R.
(
1986
).
Attachment of cells to basement membrane collagen type VI
.
J. Cell Biol
.
103
,
1569
1575
.
Aumailley
,
M.
,
Mann
,
K.
,
von der Mark
,
H.
and
Timpl
,
R.
(
1989
).
Cell attachment properties of collagen type VI and arg-gly-asp dependent binding to its α2(VI) and α3(VI) chains
.
Exp. Cell Res
.
181
,
463
474
.
Ayad
,
S.
,
Abedin
,
M. Z.
,
Grundy
,
S.
and
Weiss
,
J. B.
(
1981
).
Isolation and characterisation of an unusual collagen from hyaline cartilage and intervertebral disc
.
FEBS Lett
.
123
,
195
199
.
Cardarelli
,
P. M.
,
Yamagata
,
S.
,
Taguchi
,
I.
,
Gorcsan
,
F.
,
Chiang
,
S. L.
and
Lobl
,
T.
(
1992
).
The collagen receptor α2β1? from MG-63 and HT1080 cells, interacts with a cyclic RGD peptide
.
J. Biol. Chem
.
267
,
23159
23164
.
Coller
,
B. S.
,
Beer
,
J. H.
,
Scudder
,
L. E.
and
Steinberg
,
M. H.
(
1989
).
Collagen-platelet interactions: evidence for a direct interaction of collagen with platelet GPIa/IIa and an indirect interaction with platelet GPIIa/IIIa mediated by adhesive proteins
.
Blood
74
,
182
192
.
Davis
,
G. E.
(
1992
).
Affinity of integrins for damaged extracellular matrix: αvβ3 binds to denatured collagen type I through RGD sites
.
Biochem. Biophys. Res. Commun
.
182
,
1025
1031
.
Dedhar
,
S
and
Gray
,
V.
(
1990
).
Isolation of a novel integrin receptor mediating Arg-Gly-Asp-directed cell adhesion to fibronectin and type I collagen from human neuroblastoma cells; association of a novel β1-related subunit with αv
.
J. Cell Biol
.
110
,
2185
2193
.
Dessau
,
W.
,
Sasse
,
J.
,
Timpl
,
R.
and
von der Mark
,
K.
(
1978
).
Roles of fibronectin and collagen types I and II in chondrocytic differentiation in vitro
.
Ann. NY Acad. Sci
.
312
,
404
405
.
Eble
,
J. A.
,
Golbik
,
R.
,
Mann
,
K.
and
Kühn
,
K.
(
1993
).
The α1β1 integrin recognition site of the basement membrane collagen molecule [α1(IV)]2α2(IV)
.
EMBO J
.
12
,
4795
4802
.
Enomoto
,
M. I.
, and
Takigawa
,
M.
(
1992
).
Biological Regulation of Chondrocytes
, pp.
321
338
.
CRC Press
,
Boca Raton, FL
.
Enomoto
,
M.
,
Leboy
,
P. S.
,
Menko
,
A. S.
and
Boettiger
,
D.
(
1993
).
β1 integrins mediate chondrocyte interaction with type I collagen, type II collagen, and fibronectin
.
Exp. Cell Res
.
205
,
276
285
.
Fressinaud
,
E.
,
Girma
,
J. P.
,
Sadler
,
J. E.
,
Baumgartner
,
H. R.
and
Meyer
,
D.
(
1990
).
Synthetic RGDS-containing peptides of von Willebrand factor inhibit platelet adhesion to collagen
.
Thromb. Haemost
.
64
,
589
593
.
Gisslow
,
M. T.
and
McBride
,
B. C.
(
1975
).
A rapid sensitive collagenase assay
.
Anal. Biochem
.
68
,
70
78
.
Grinnell
,
F.
(
1978
).
Cellular adhesiveness and extracellular substrata
.
Int. Rev. Cytol
.
53
,
65
144
.
Grinnell
,
F.
and
Feld
,
M. K.
(
1979
).
Initial adhesion of human fibroblasts in serum-free medium: possible role of secreted fibronectin
.
Cell
17
,
117
129
.
Grinnell
,
F.
and
Feld
,
M. K.
(
1980
).
Spreading of human fibroblasts in serum-free medium: inhibition by dithiothreitol and the effect of cold insoluble globulin (plasma fibronectin)
.
J. Cell. Physiol
.
104
,
321
334
.
Guidry
,
C.
,
Miller
,
E. J.
and
Höök
,
M.
(
1990
).
A second fibronectin-binding region is present in collagen α chains
.
J. Biol. Chem
.
265
,
19230
19236
.
Gullberg
,
D.
,
Turner
,
D. C.
,
Borg
,
T. K.
,
Terracio
,
L.
and
Rubin
,
K.
(
1990
).
Different β1-integrin collagen receptors on rat hepatocytes and cardiac fibroblasts
.
Exp. Cell Res
.
190
,
254
264
.
Gullberg
,
D.
,
Gehlsen
,
K. R.
,
Turner
,
D. C.
, Å
hlén
,
K.
,
Zijenah
,
L. S.
,
Barnes
,
M. J.
and
Rubin
,
K.
(
1992
).
Analysis of α1β1? α2β1 and α3β1 integrins in cell-collagen interactions: identification of conformation dependent α1β1 binding sites in collagen type I
.
EMBO J
.
11
,
3865
3873
.
Hemler
,
M. E.
(
1990
).
VLA proteins in the integrin family: structures, functions and their roles on leukocytes
.
Annu. Rev. Immunol
.
8
,
365
400
.
Humphries
,
M. J.
,
Komoriya
,
A.
,
Akiyama
,
S.K.
,
Olden
,
K.
and
Yamada
,
K.M.
(
1987
).
Identification of two distinct regions of the type III connecting segment of human plasma fibronectin that promote cell type-specific adhesion
.
J. Biol. Chem
.
262
,
6886
6892
.
Humphries
,
M. J
(
1990
).
The molecular basis and specificity of integrin-ligand interactions
.
J. Cell Sci
.
97
,
585
592
.
Hynes
,
R. O.
(
1987
).
Integrins: a family of cell surface receptors
.
Cell
48
,
549
554
.
Hynes
,
R. O.
(
1992
).
Integrins: Versatility, modulation, and signalling in cell adhesion
.
Cell
69
,
11
25
.
Kielty
,
C. M.
,
Hopkinson
,
I.
and
Grant
,
M. E.
(
1993
).
Collagen: The collagen family: structure, assembly, and organization in the extracellular matrix
.
In Connective Tissue and its Heritable Disorders
(ed.
P. M.
Royce
and
B.
Steinmann
), pp.
103
149
.
Wiley-Liss
,
New York
.
Kleinman
,
H. K.
,
McGoodwin
,
E. B.
,
Martin
,
G. R.
,
Klebe
,
R. J.
,
Fietzek
,
P. P.
and
Wooley
,
D. E.
(
1978a
).
Localization of the binding site for cell attachment in the α1(I) chain of collagen
.
J. Biol. Chem
.
253
,
5642
5646
.
Kleinman
,
H. K.
,
Murray
,
J. C.
,
McGoodwin
,
E. B.
and
Martin
,
G. R.
(
1978b
).
Connective tissue structure: cell binding to collagen
.
J. Invest. Derm
.
71
,
9
11
.
Kramer
,
R. H.
and
Marks
,
N.
(
1989
).
Identification of integrin collagen receptors on human melanoma cells
.
J. Biol. Chem
.
264
,
4684
4688
.
Laemmli
,
U. K.
(
1970
).
Cleavage of structural protein during the assembly of the head bacteriophage T4
.
Nature
227
,
680
685
.
Lebien
,
T. W.
,
Bove
,
D. R.
,
Bradley
,
J. G.
and
Kersey
,
J. H.
(
1982
).
Antibody affinity may influence antigenic modulation of the common acute lymphoblastic leukemia antigen in vitro
.
J. Immunol
.
129
,
2287
2292
.
Marshall
,
J. F.
,
Nesbitt
,
S. A.
,
Helfrich
,
M. H.
,
Horton
,
M. A.
,
Polakova
,
K.
and
Hart
,
I. R.
(
1991
).
Integrin expression on human melanoma cell lines: Heterogeneity of vitronectin receptor composition and function
.
Int. J. Cancer
49
,
924
931
.
Miller
,
E. J.
and
Lunde
,
L. G.
(
1973
).
Isolation and characterisation of the cyanogen bromide peptides from the α1(II) chain of bovine and human cartilage collagen
.
Biochemistry
12
,
3153
3159
.
Miller
,
E. J.
and
Rhodes
,
R. K.
(
1982
).
Preparation and characterisation of the different types of collagen
.
Meth. Enzymol
.
82
,
33
64
.
Mollenhauer
,
J.
and
von der Mark
,
K.
(
1983
).
Isolation and characterization of a collagen-binding glycoprotein from chondrocyte membranes
.
EMBO J
.
2
,
45
50
.
Nip
,
J.
,
Shibata
,
H.
,
Loskutoff
,
D. J.
,
Cheresh
,
D. A.
and
Brodt
,
P.
(
1992
).
Human melanoma cells derived from lymphatic metastases use integrin αvβ3 to adhere to lymph node vitronectin
.
J. Clin. Invest
.
90
,
1406
1413
.
Piez
,
K.
(
1968
).
Molecular weight determination of random coil polypeptides from collagen by molecular sieve chromatography
.
Anal. Biochem
.
26
,
305
312
.
Rubin
,
K.
,
Höök
,
M.
,
Öbrink
,
B.
and
Timpl
,
R.
(
1981
).
Substrate adhesion of rat hepatocytes: Mechanism of attachment to collagen substrates
.
Cell
24
,
463
470
.
Salter
,
D. M.
,
Hughes
,
D. E.
,
Simpson
,
R.
and
Gardner
,
D. L.
(
1992
).
Integrin expression by human articular chondrocytes
.
Br. J. Rheum
.
31
,
231
234
.
Santoro
,
S. A.
(
1986
).
Identification of a 160,000 dalton platelet membrane protein that mediates the initial divalent cation-dependent adhesion of platelets to collagen
.
Cell
46
,
913
920
.
Schor
,
S. L.
and
Court
,
J.
(
1979
).
Different mechanisms in the attachment of cells to native and denatured collagen
.
J. Cell Sci
.
38
,
267
281
.
Staatz
,
W. D.
,
Rajpara
,
S. M.
,
Wayner
,
E. A.
,
Carter
,
W. G.
and
Santoro
,
S. A.
(
1989
).
The membrane glycoprotein Ia-IIa (VLA-2) complex mediates the magnesium-dependent adhesion of platelets to collagen
.
J. Cell Biol
.
108
,
1917
1924
.
Staatz
,
W. D.
,
Walsh
,
J. J.
,
Pexton
,
T.
and
Santoro
,
S. A.
(
1990
).
The α2β1 integrin cell surface collagen receptor binds to the α1(I)-CB3 peptide of collagen
.
J. Biol. Chem
.
265
,
4778
4781
.
Staatz
,
W. D.
,
Fok
,
K. F.
,
Zutter
,
M. M.
,
Adams
,
S. P.
,
Rodriguez
,
B. A.
and
Santoro
,
S. A.
(
1991
).
Identification of a tetrapeptide recognition sequence for the α2β1 integrin in collagen
.
J. Biol. Chem
.
266
,
7363
7367
.
Takada
,
Y.
,
Wayner
,
E. A.
,
Carter
,
W. G.
and
Hemler
,
M. E.
(
1988
).
Extracellular matrix receptors, ECMRII and ECMRI, for collagen and fibronectin correspond to VLA-2 and VLA-3 in the VLA family of heterodimers
.
J. Cell. Biochem
.
37
,
385
393
.
Takigawa
,
M.
,
Tajima
,
K.
,
Pan
,
H.-O.
,
Enomoto
,
M.
,
Kinoshita
,
A.
,
Suzuki
,
F.
,
Takano
,
Y.
and
Mori
,
Y.
(
1989
).
Establishment of a clonal human chondrosarcoma cell line with cartilage phenotypes
.
Cancer Res
.
49
,
3996
4002
.
Tenchini
,
M. L.
,
Adams
,
J. C.
,
Gilbert
,
C.
,
Steel
,
J.
,
Hudson
,
D. L.
,
Malcovati
,
M.
and
Watt
,
F. M.
(
1993
).
Evidence against a major role for integrins in calcium-dependent intercellular adhesion of epidermal keratinocytes
.
Cell Adhesion Commun
.
1
,
55
66
.
Traub
,
W.
and
Piez
,
K. A.
(
1971
).
The chemistry and structure of collagen
.
Advan. Protein Chem
.
25
,
243
352
.
Tuckwell
,
D. S.
and
Humphries
,
M. J.
(
1993
).
The molecular and cellular biology of integrins
.
Crit. Rev. Oncol./Hematol
.
15
,
149
171
.
Vandenberg
,
P.
,
Kern
,
A.
,
Ries
,
A.
,
Luckenbill-Edds
,
L.
,
Mann
,
K.
and
Kühn
,
K.
(
1991
).
Characterization of a type IV collagen major cell binding site with affinity to the α1β1 and the α2β1 integrins
.
J. Cell Biol
.
113
,
1475
1483
.
Verfaille
,
C.
,
McCarthy
,
J. C.
and
McGlave
,
P. B.
(
1992
).
Mechanisms underlying abnormal trafficking of malignant progenitors in chronic myelogenous leukemia
.
J. Clin. Invest
.
90
,
1232
1241
.
Wayner
,
E. A.
and
Carter
,
W. G.
(
1987
).
Identification of multiple cell adhesion receptors for collagen and fibronectin in human fibrosarcoma cells possessing unique α and common β subunits
.
J. Cell Biol
.
105
,
1873
1884
.
Yamada
,
K. M.
and
Olden
,
K.
(
1978
).
Fibronectins - adhesive glycoproteins of cell surface and blood
.
Nature
275
,
179
184
.