Tripeptide RGD-dependent adhesion of articular chondrocytes to synovial fibroblasts

Chondrocytes embedded in the cartilage tissue interact with the extracellular cartilage matrix through substratum adhesion molecules. Chondrocytes have been shown to have attachment to collagen types I, II, V, IX and XI (von der Mark et al. 1986). Of these, chondrocytes show a greater degree of attachment to type II collagen. A 34 kDa molecule, called anchorin CII, has been purified and cloned from chondrocytes derived from adult chicken xiphoid cartilage (Fernandez et al. 1988). Other chondrocyte surface components that are involved in matrix attachment include receptors for hyaluronan (McCarthy and Toole, 1989), which has the capacity to assemble and organize the extracellular components within the pericellular matrix (Knudson and Kuettner, 1990). In vivo studies have shown that chondrocytes do not express vitronectin receptors (Simpson and Horton, 1989). Chondrocytes synthesize matrix components such as collagen type II and glycosoaminoglycan (Stockwell and Meachim,1979), and also produce a chondrocyte attachment factor called chondronectin (Hewlitt et al. 1980). These studies suggest that substratum adhesion molecules that are present on chondrocytes and the matrix that these cells produce are important for anchoring and in the organization of cartilage tissue.

Chondrocytes are located singly or as a cluster of a few cells within the cartilage matrix tissue. This type of tissue organization, in developed cartilage, restricts the kinds of cell-cell interactions that are an essential part of normal organ development and the remodeling process. However, chondrocytes, during development and in the adult organism, do encounter cell-cell interactions that can have either positive or negative influences on chondrogenic differentiation (von der Mark, 1986). Heterologous cell interactions between embryonic epithelia such as the notocord, retinal pigmented epithelium and the otic vesicle trigger chondrogenic differentiation. Mesenchymal condensation in the center of the limb bud, which is a model of homologous cell-cell interaction, also triggers cartilage differentiation (Ede, 1983). Similarly, in adult endo-chondrodral bone formation, osteophyte development and neoplastic conditions like chondrosarcoma, chondrocytes are involved in homologous and heterologous cell interactions (Reddi, 1981). In the unscheduled remodelling of tissue seen in rheumatoid arthritis, cartilage matrix degradation frees chondrocytes, which then interact with infiltrating inflammatory cells (Hough and Sokoloff, 1989).

Chondrocytes embedded within the matrix show extremely low rates of cell proliferation (Stockwell and Meachim, 1979). Isolated chondrocytes, on the other hand, proliferate rapidly in response to various growth factors (Morales and Hascall, 1989; Seyedin and Rosen,1991). There is recent evidence that chondrocytes produce some of the growth factors that can have an autocrine or paracrine influence on chondrocyte proliferation (Seyedin and Rosen, 1991). Chondrocytes that are released from the articular cartilage matrix into the synovial cavity during arthritis could respond to cytokines and growth factors present in the synovial fluid (Lipsky et al. 1989). Thus, chondrocytes in the synovial cavity could proliferate and anchor to the synovial membrane. We developed an experimental system to investigate these interactions by studying the adhesion of chondrocytes to synovial fibroblasts. Our study shows that chondrocytes specifically attach to synovial fibroblasts through RGD (tripeptide: arginineglycine-aspartic acid)-dependent receptors. Part of this attachment is mediated through the interaction between fibronectin (Fn) and the receptor for fibronectin (FnR). Chondrocytes, as well as synovial fibroblasts, express cell-bound Fn and FnR. Using specific monoclonal antibodies, we were able to immunoprecipitate the VLA₅α and β chains of FnR in both chondrocytes and synovial fibroblasts. Northern blot studies demonstrated the presence of mRNA for the α and β chains of FnR in chondrocytes and fibroblasts. Further, it was observed that binding of chondrocytes to fibroblasts promotes cell spreading in chondrocytes.

NZW rabbits weighing 1–1.5 kg were obtained from a local vendor and served as the source of chondrocytes and synovial fibroblasts. Animals were housed in the Vivarium facilities of this institution and were killed by intravenous injection of Beuthanasia-D special (Schering Corp, NJ).

Human fibronectin and vitronectin were purchased from Telios Inc. Gly-Arg-Gly-Asp-Ser-Pro (GRGDSP) and Gly-Arg-Gly-Glu-Ser-Pro (GRGESP) peptides were purchased from Peninsula Laboratories Inc (Belmont, CA) and Telios Inc (San Diego, CA). Purified rat monoclonal mAb 13 (against α chain of VLA integrin) and mAb 16 (against β chain of VLA₅) were kind gifts from Drs. S.K. Akiyama and K.M. Yamada (Akiyama et al. 1989). (We show that these antibodies prevent the adhesion of rabbit cells to matrix proteins.) Monoclonal BIE5 (against α chain of VLA₅) was a kind gift from Dr. C.H. Damsky (Werb et al. 1989). Antirabbit fibronectin was purchased from Organon Teknika (PA) and was the IgG fraction of goat antibody. ⁵¹Chromium and [³⁵S]methionine were purchased from ICN Biochemicals, Inc (Costa Mesa, CA). cDNA probes for the a and /3 chains of FnR were obtained from Telios Inc (San Diego, CA).

The synovial fibroblasts were propagated from explants of synovial membranes obtained from the infrapatellar fat pads of rabbit knee joints (Tiku et al. 1985b). The confluent monolayers were washed with Ca²⁺- and Mg²⁺-free HBSS (Gibco, NY), digested with 0.25% trypsin-EDTA (Gibco), and then split 1:3 or 1:4 into 100 mm plastic Petri dishes and cultured in 10% FBS-Dulbecco’s minimum essential media (DMEM). Synovial fibroblasts at second-eighth passages were used for binding assays. The synovial fibroblasts were also cultured in serum-free media, purchased from Ventrex Labs (Portland, Maine). Fibroblasts were passaged three times in serum-free media before being used in the binding assay. In certain experiments, the fibroblasts cultured in microtitre wells were fixed with 100 μl of 1% freshly prepared paraformaldehyde in HBBS, incubated at room temperature for 15 min and washed at least 4–5 times with HBSS before addition of chondrocytes.

The chondrocytes were isolated, as mentioned previously, by overnight enzymatic digestion of rabbit articular cartilage shavings obtained from the ends of long bones (Tiku et al. 1985a). The chondrocytes were washed in FBS-containing media and comprised a homogeneous population of cells with more than 98% viability. Primary articular chondrocytes were used in all experiments.

Each experiment was repeated at least 2–3 times and the data shown in each figure are from a representative experiment.

Fibronectin and vitronectin were added at a concentration of 5 μg/ml in a volume of 100 μl PBS per microtitre well (Hayman et al. 1985). After the plates were incubated for 2 h at room temperature, the solution was aspirated and replaced with DMEM containing 5 mg/ml of BSA. The plates were further incubated for 30–45 min to block the unbound sites on the plastic wells and washed with DMEM. The entire procedure was done under sterile conditions and plates were used on the same day. The binding assay between chondrocytes and purified matrix protein was done, as described above, in serum-free conditions.

Chondrocytes and enzyme-dispersed fibroblasts were reacted with Fn and FnR antibodies in PBS containing BSA for 30 min at 4°C in the presence of azide. The cells were washed and reacted with a second antibody, i.e. fluorescein isothiocyanate (FTTC) conjugated to goat or rat IgG. (Cappel). The cells were washed and fixed in 1% paraformaldehyde and examined using a Carl Zeiss Axiophot microscope. Background fluorescence was determined after the cells had been incubated with either rat or goat IgG (Sigma) and FITC-conjugated second antibody.

Articular chondrocytes (2 × 10⁷) and synovial fibroblasts (1 ×10⁷) were washed twice in methionine-free DMEM and resuspended in 10 ml of methionine-free DMEM containing 10% dialyzed FBS. The cells were labelled with [³⁵S]methionine (100 μCi/ml) for 16 h at 37°C in a CO₂ incubator. The cells were washed with cold PBS containing methionine and sodium azide, and lysed in 2 ml of extraction buffer containing 1% NP-40 with inhibitors, PMSF (50 μg/ml), pepstatin (1 gg/ml), aprotinin (2 μg/ml) and leupeptin (2 μg/ml). The lysates were spun at 27,000 g for 30 min at 4°C and the supernatants were subjected to immunoprecipitation.

The cell lysates were mixed with equal volumes of 4% nonfat dry milk (Blotto), precleared with Protein G/Sepharose (Pharmacia) for 30 min and incubated with antibodies, i.e. mAb 13, mAb 16 and BIE5. Immune complexes were precipitated with Protein G/Sepharose and washed 3 times in PBS containing 0.1% NP-40 and inhibitors. The precipitates were boiled in SDS sample buffer and electrophoresed in 7.5% SDS-PAGE under reducing and non-reducing conditions. Known M_r markers were included in the gel. The gels were fixed, stained with Coomasssie Blue, destained and embedded with Fluorohance (RPI, IL). Autoradiographs were obtained after 16–48 h.

Total cellular RNA was prepared from enzyme-isolated primary chondrocytes and cultured synovial fibroblasts using the guanidine isothiocynate/cesium chloride method (Chirgwin et al. 1979). A 2.5–5 μg sample/lane of total RNA was separated on a 1.5% agarose gel containing 2.2 M formaldehyde. The RNA was transferred to a nitrocellulose filter and hybridized with ³²P-labelled cDNA probes. The probes were radiolabelled by nick translation to a specific activity of approximately 10^{8 * 10} to 2 × 10⁸ cts/min per μg DNA (Maniatis et al. 1982). Hybridization was carried out overnight at 42°C in 50% formamide, 5 × SSPE (1 × SSPE is 0.15 M NaCl/0.01 M NaH₂P04/0.001 M Na EDTA), 1 × Denhardt’s solution (0.02% polyvinylprrolidone/0.02% Ficoll/0.02% BSA), 100 gg/ml denatured salmon sperm DNA, and 0.1% SDS. Filters were exposed to an X-ray film and an autoradiogram was obtained.

The fluorescent lipophilic dye (PKH-l)-labelled chondrocytes were used to study cell spreading on monolayer of fibroblasts (Horan and Slezak, 1989). PKH-1 is stably incorporated and retained in the plasma membrane of the cell. The dye has an excitation wavelength of 488 nm and emits at 525 nm. The chondrocytes were labelled with 5 μM solution of PKH-1 according to the instructions of the manufacturer (Zynaxis Cell Science, PA). The washed cells were set up in the binding assay with or without fibroblasts. The adherence of chondrocytes to the fibroblasts was monitored in an inverted Zeiss IM35 fluorescence microscope. Photographs of the cells were obtained at different times.

A fixed number of chondrocytes (10⁵/well) were added to quadruplicate sets of wells in which increasing numbers of synovial fibroblasts (0.5 ×10⁴ to 4 ×10⁴) had been allowed to adhere for overnight periods. Chondrocytes that bound to wells containing no fibroblasts represented the background binding. Replicate sets of the cultures were harvested after 1, 2, 3, 4 and 24 h of incubation and the percentage binding was measured. About 6-12% of chondrocytes bound to the plastic during the the first 4 h of incubation. When chondrocytes were cultured on the synovial monolayer there was 4- to 10-fold-enhanced chondrocyte adhesion, which peaked at 4 h of incubation (Fig. 1). Chondrocyte adhesion increased depending upon the number of fibroblasts that had been seeded in the wells. A total of 10⁴ synovial fibroblasts, which formed a monolayer in the wells, were therefore routinely used in a 4-h binding assay. Fig. 2 shows photographic representations of chondrocyte adhesion to fibroblasts before (A) and after the wells were washed (B and C). Residual chondrocytes are seen remaining on the plastic (B) and on a monolayer of fibroblasts (C) after the 4-h binding assay and washing step.

Chondrocytes (0.5 × 10⁴, 10⁵ and 2 × 10⁵) cultured on monolayers of fibroblasts (10⁴/well) resulted in 32, 37 and 50% binding, respectively. In order to prevent homotypic binding between chondrocytes, when increasing numbers of cells were used, the binding assay was done routinely with 10⁵ chondrocytes per well.

We investigated the temperature dependency of chondrocyte-fibroblast adhesion by doing binding assays at 4°C, 25°C and 37°C. Chondrocytes only bound to fibroblasts at 37°C; binding was not permissive at lower temperatures. On the other hand, chondrocyte binding to poly-L-lysine-coated wells was not affected by the temperature of incubation. The data suggest that the binding of chondrocytes to fibroblasts requires metabolic energy (Hynes, 1987; Ruoslahti and Piersch-bacher, 1987). Binding of chondrocytes to fibroblasts was abolished in the presence of 10 mM EDTA (data not shown). However, when an equimolar concentration of calcium was added to the assay wells, the binding returned to normal control values, suggesting that the adhesion of chondrocytes to fibroblasts is a calcium-dependent process that resembles the heterotypic binding described in other cell types (Hynes, 1987; Ruoslahti and Pierschbacher, 1987). When chondrocyte adhesion was measured in the presence of 10⁻⁴ M cytochalasin B there was no adhesion between chondrocytes and synovial fibroblasts (data not shown).

Cell-cell adhesion is a membrane-mediated event. Some membrane-initiated events such as antigen presentation to T lymphocytes are retained even after fixation of antigen-presenting cells by paraformaldehyde (Unanue and Allen, 1987). We investigated if chondrocytes would bind in a specific manner to fixed fibroblasts. After 4 h of incubation, the mean % specific binding of labelled chondrocytes to fixed fibroblasts was 19.0 ± 0.4 compared to 25.9 ±6.1 when the fibroblasts were alive (mean of 5 experiments). This suggested that the ligands on fibroblasts retained significant binding activity even after fixation and that the adherent fibroblasts need not be viable for binding to occur.

To identify the integrin-ligands involved in chondrocyte-fibroblast interaction, we performed the binding experiments in the presence of increasing concentrations of cell adhesion antagonists (Fig. 3). GRGDSP (1,10 and 50 μg/ml) is a synthetic peptide that competes with ligand-integrin through the arginine-glycine-aspartic acid (RGD) tripeptide sequence (Ruoslahti and Pierschbacher, 1987). RGD is the primary cell binding site of fibronectin for fibroblasts and is located in the central cell binding domain of the fibronectin molecule (Yamada, 1989). GRGESP (1, 10 and 50 Mg/ml) is the control peptide, which has no inhibitory effect. Antifibronectin (anti-FN) (20, 100 and 200 Mg/ml) is a polyclonal goat anti-rabbit fibronectin antibody; G-serum (2, 20 and 100 Mg/ml) is a goat anti-mouse Ig antibody (control). As shown in Fig. 3, the binding of chondrocytes to fibroblasts was completely inhibited by the peptide GRGDSP. The inhibition by GRGDSP was concentration-dependent and the control peptide, GRGESP, had no inhibitory effect. Compared to the other agents used, only anti-FN antibodies resulted in partial inhibition of chondrocyte adhesion and the inhibitory activity was dependent on the amount of antibody present in the wells. Almost complete inhibition by the RGD peptide indicates that most chondrocyte-fibroblast adhesion occurs through the recognition of the RGD sequence. Addition of anti-Fn antibodies resulted in partial inhibition, suggesting that fibronectin is one of the ligands involved in chondrocyte-fibroblast adhesion. Anti-Fn antibodies at a con centration of as much as 200 μg/well did not completely inhibit chondrocyte adhesion, indicating that fibronectin is only one of the ligands involved in chondrocyte adhesion. The inhibition was specific in that control antibodies Ag8 (plasmacytoma), L11/35 (which reacts to a rabbit T cell marker) and 2C4 (a monoclonal, which reacts to rabbit la antigen) had no inhibitory effect on the process of cell adhesion (data not shown). la antigen has been shown to be present on populations of normal rabbit articular chondrocytes (Tiku et al. 1985a).

Next we investigated the binding of chondrocytes to purified matrix proteins bound to plastic in serum-free conditions. Monoclonal antibodies were used in these assays to identify and establish specificity of inhibition. The antibodies used were BIE5, mAb 13 and mAb 16. BIE5 is a rat anti-FnR mAb specific against the α chain of FnR (Werb et al. 1989) (BIE5 in the form of hybridoma culture supernatant was used). mAb 13 and mAb 16 (50 μg/well) are rat monoclonal antibodies that react to the common chain of VLA integrins and to the α chain of VLA₅, respectively (Akiyama et al.1989). As shown in Fig. 4, there was more than 50% binding of chondrocytes to wells coated with fibronectin. The binding was completely inhibited by anti-Fn antibodies. Rat monoclonals BIE5, mAb 13 and mAb 16 caused more than 50% inhibition of chondrocyte adhesion. These observations suggest that chondrocytes bind to purified fibronectin and that adhesion to fibronectin can be inhibited by the mAb, which binds either to the common chain or to the a chain of FnR (i.e. VLA₅ or V₅ β1₁). Chondrocytes also bound to vitronectin and this binding was partially inhibited by anti-Fn, BIE5, mAbl3, mAbl6 and the GRDGSP peptide. Further studies are needed to investigate the interaction of chondrocytes to vitronectin.

The data presented above suggested that Fn and FnR are involved in chondrocyte-fibroblast adhesion. Compared to the effect of GRGDSP, which resulted in almost complete inhibition of chondrocyte adhesion, antibodies to Fn only had a partial inhibitory effect (19% binding) on cell binding (Fig. 5). Fig. 5 also shows that mAb 13, against the common β1₁ chain of VLA integrins, which should abrogate the binding initiated with all the possible a chains of β1₁ integrins, only resulted in 17% binding. We reasoned that a combination of antibodies would identify the ligand-receptor involved in chondrocyte adhesion to fibroblasts. As shown in Fig. 5, the combination of anti-Fn with three different monoclonals, i.e. mAb 13, mAb 16 and BIE5, had no significant additive effect on the inhibition of chondrocyte binding to fibroblasts. If VLA integrins other than VLA₅ were involved in chondrocytefibroblast adhesion, the combination of anti-Fn and anti-β1₁ common chain should have resulted in a greater inhibitory effect than that observed by a combination of anti-Fn and VLA5 a-chain-specific antibodies such as mAb 16 and BIE5. Since no additive inhibitory effect was observed, it is possible that only VLA₅ integrin of the β1₁ family is utilized in chondrocyte-fibroblast cell adhesion. The additional molecules involved in cell adhesion, however, are recognized through the presence of RGD.

In the experiments described above, it was not clear if the antibodies were reacting with fibroblasts, chondrocytes or both cell types. Since chondrocytes could bind to paraformaldehyde-fixed fibroblasts, we studied the adherence of chondrocytes to fixed synovial fibroblasts that had been pretreated with antibodies. As shown in Fig. 6, fibroblasts were treated with various antibodies for 30 min at 4°C. The cells were washed twice, fixed with paraformaldehyde and chondrocyte binding was measured. The experimental protocol presumes that the antibodies would bind and block Fn and/or FnR and that the receptors and ligands on the cells would not be modulated during the binding assay following fixation. No free antibodies were present in the culture system during the adhesion assay. As shown in Fig. 6 A, blocking of Fn, FnR or both molecules on the fibroblasts did not significantly inhibit chondrocyte adhesion, suggesting that there may be other ligandreceptors present on the fibroblasts to which chondrocytes bind. When chondrocytes were pretreated with antibodies for 30 min at room temperature and then transferred to fixed fibroblasts along with antibodies the degree of inhibition was greater than that observed in pretreated fixed fibroblasts (Fig. 6 B). This suggests that blocking of Fn and FnR or both on fibroblasts was not sufficient to prevent chondrocyte adhesion but that the antibodies added in suspension must be interacting with the chondrocytes as well as the fibroblasts to result in the greater degree of inhibition observed. Therefore both these cells must express the ligands Fn and FnR, which may be interacting with each other, resulting in bidirectional adhesion.

Chondrocytes and fibroblasts were reacted with appropriate antibodies as described in Materials and methods. As shown in Fig. 7, both chondrocytes and fibroblasts reacted with antibodies to Fn, suggesting that this matrix protein was intimately associated with the cell membranes. Fibronectin was also detected in the cell lysates of fibroblasts and chondrocytes in western blot studies (data not shown). The monoclonal antibodies BIE5, mAb 13 and mAbl6 reacted with chondrocytes and fibroblasts. Immunological reactivity against these monoclonals was specific in that the control rat IgG showed no binding to the cells (Fig. 7). The immunofluorescent reactivity to mAb 16 was more prononunced in the synovial fibroblasts than in the chondrocytes, suggesting that fibroblasts have more FnR than chondrocytes. Thus expression of αand β integrin chains was demonstrated by immunofluorescence techniques in rabbit articular chondrocytes and fibroblasts.

The α and β chains of the FnR were immunoprecipitated from [³⁵S]methionine-labelled chondrocytes and fibroblasts. Fig. 8 is a SDS-PAGE autoradiogram of immunoprecipitates under non-reducing and reducing conditions. No specific reactivity was observed in the cell lysates of fibroblasts and chondrocytes when control rat IgG was used (lane 1). As shown in lanes 2,3 and 4, monoclonal antibodies BIE5, mAb 13 and mAb 16 immunoprecipitated two molecules at 140 kDa and 110 kDa, respectively, when analyzed under nonreducing conditions. The 140 kDa and 110 kDa bands represent the αand β subunits of integrins, respectively. Immunoprecipitates that were analyzed on SDS-PAGE under reduced conditions showed a single chain at 140 kDa. The comigration of the two integrin chains to a single chain at 140 kDa under reduced conditions is consistent with the biochemical characteristics of the integrins (Ruoslathi and Pierschbacher, 1987). The intensity of the immunoprecipitates in fibroblasts was greater than in chondrocytes, suggesting that fibroblasts express more FnR than chondrocytes, which is consistent with observations in the immunofluorescence studies.

We next used northern blot analysis to identify FnR α and β transcripts in RNA extracted from rabbit synovial fibroblasts and chondrocytes. As shown in Fig. 9, the mRNA transcript for the common fi chain of FnR was present in both fibroblasts and chondrocytes, with larger amounts present in fibroblasts. Also the mRNA transcript for VLA₅α chain was present in both cell types, with larger amounts present in chondrocytes than in fibroblasts. These observations are consistant with data from immunoflourescence and immunoprecipitation studies presented above. It should be noted that equal amounts of total RNA from fibroblasts and chondrocytes were loaded per well as seen by ethidium bromide staining (data not shown).

It has previously been shown that when cells adhere to matrix proteins through specific integrin-receptors spreading of the cells occurs, and that this may have many biological consequences such as migration and differentiation of the cells (Akiyama at al. 1989). It has been observed that, on adhesion to Fn, chondrocytes show morphological changes and appear fibroblast-like (Pennypacker et al. 1979; West et al. 1979). These changes could depict dedifferentiation of chondrocytes (von der Mark, 1986). We investigated if chondrocytes in vitro change shape on binding to fibroblasts. As described in Materials and methods, chondrocytes were labelled with the fluorescent lipophilic dye PKH-1 and shape changes in the chondrocytes were monitored by fluorescence microscopy after 4, 24, 48 and 72 h of in vitro culture. Chondrocytes were also cultured alone on plastic as a control. As shown in Fig. 10, chondrocytes grown on fibroblasts gradually changed shape when compared to the control chondrocytes cultured on plastic (compare Fig. 9 B, F and J with D,H and L). The chondrocytes spread out, became spindle-shaped and grew on the fibroblast monolayers. After a period, chondrocytes and fibroblasts viewed under phasecontrast showed no morphological difference. Under these coculture conditions, chondrocytes imperceptibility blended into the monolayer of fibroblasts, as seen in the photomicrograph (Fig. 9 K). The chondrocytic nature of the cells could be identified only by the presence of the fluorescent dye PHK-1. On the other hand, chondrocytes that were cultured on plastic remained rounded and no significant changes in the shape of these cells were observed even after 72 h of culture (Fig. 9 B, F and J). These studies suggest that chondrocytes acquire altered shape when cultured on fibroblasts. This change in shape could represent early signs of dedifferentiation (von der Mark, 1986).

We have studied the heterotypic cell-cell interactions between articular chondrocytes and synovial fibroblasts. These studies demonstrate that chondrocytes attach to synovial fibroblasts through RGD-dependent receptors. Extracellular matrix proteins with RGD recognition sites include fibronectin, vitronectin, collagen, osteopontin, thrombospondin, fibrinogen, von Willebrand factor, entactin and laminin (Yamada,1991). Since synovial fibroblasts are not likely to produce laminin, von Willebrand factor, osteopontin, thrombospondin, fibrinogen or vitronectin, we inferred that chondrocytes may be binding to either fibronectin, collagen or some other adhesive protein recognized through the RGD primary cell binding site of fibronectin (Yamada, 1989). Investigations using anti-Fn antibodies revealed that only a part of the binding by chondrocytes was mediated by the recognition of cell-associated Fn in fibroblasts. Inhibition of cell adhesion by antibodies to receptors of matrix proteins is a powerful investigative tool to determine the role of these receptors in cell-cell and cell-matrix interaction (Springer, 1990; Albelda and Buck, 1990). Monoclonal antibodies against the α and β chains of the FnR partially inhibited chondrocyte adhesion to fibroblasts, suggesting that only part of the adhesion process was mediated through the interaction between fibronectin and FnR. When a combination of antibodies was used in the adhesion assay, no additive inhibitory effect was observed, indicating that perhaps only the VLA₅ integrin of the β1₁family is utilized in chondrocyte-fibroblast cell adhesion, i.e. that other a chains of VLA integrins play little part in chondrocyte adhesion. On the other hand, other RGD-dependent ligand-cell adhesion molecules are likely to be involved in the adhesion of chondrocytes to synovial fibroblasts.

Studies in development biology have shown that cell cell interactions play an important role in the differentiation of chondrocytes (von der Mark, 1986). For example, heterologous cellular interactions between somite mesenchyme and notochord epithelium give rise to vertebral cartilage while homologous cellular interactions result in cartilage formation in the limb bud. It has been suggested that the final step of chondrogenesis may be triggered by interactions with the extracellular matrix components (von der Mark, 1986). A number of studies support the notion that type II collagen might trigger chondrogenic differentiation. Fn, type I collagen and other adhesive substances help in the condensation of cells in the chondrogenic zone (Dessau et al. 1980). With overt differentiation of chondroblasts, type I collagen and Fn disappear while hyaluronic acid is drastically reduced (Dessau at al. 1980; Singley and Solursh, 1981). In vitro studies on the attachment of chondrocytes to Fn have shown that these cells lose the capacity to differentiate and become fibroblast-like (Pennypacker et al. 1979; West at al. 1979). We have shown that chondrocytes bind to purified fibronectin and vitronectin. Anti-Fn antibodies caused almost complete inhibition of chondrocyte binding to fibronectin while anti-FnR antibodies resulted in only partial inhibition, suggesting that during the adhesion assay chondrocytes were secreting other adhesive factors such as chondronectin (Hewlitt at al. 1980), which could bind to the fibronectin matrix. When fibronectin in the wells‘ was blocked by anti-Fn, there was complete inhibition because neither the cells nor the adhesive factors synthesized by the chondrocytes could attach to plasticbound fibronectin and therefore under these conditions there was complete inhibition of chondrocyte adhesion. Anti-FnR would have only a partially inhibitory effect because it would react specifically with FnR and chondrocytes would still be able to bind through other ligand-receptors.

Antibodies to the common ft chain of VLA integrin did not show increased inhibition of chondrocyte adhesion compared to antibodies to just the VLA₅α chain when binding to either purified fibronectin or to fibroblasts was studied. This suggested that perhaps only high-affinity FnR (VLA_S) from the VLA family of integrins was involved in chondrocyte adhesion to fibroblasts. FnR is thought to be mainly involved in cellmatrix interaction (Yamada, 1989). Recent studies suggest that this receptor may also be involved in cell-cell interactions (Adams and Watt, 1990; Peltonen et al. 1989). FnR-Fn is poorly expressed in most tissues and it is therefore interesting to note that FnR mediates heterotypic cell-cell adhesion between chondrocytes and synovial fibroblasts (Albelda and Buck, 1990). Our studies suggest that there may be other families of integrins involved in chondrocyte adhesion and that these additional integrins recognize a common RGD motif on some unidentified ligands.

In vivo, studies have shown that chondrocytes do not express vitronectin receptor (Simpson and Horton, 1989). Expression of other integrins such as β1₁ on chondrocytes in vivo or in vitro cultured cells has not been studied in detail. The pattern of integrin expression by cultured cells is not always identical to that expressed by the same cell type in vivo, which suggest that there are regulatory constraints placed on integrin expression in situ that are different from those in culture. Chondrocytes bind to collagen (von der Mark, 1986) but the biochemical nature of receptors that facilitate this binding remains to be investigated. Recent studies on hyaluronan receptors have revealed the important role that these receptors play in cell anchoring and organization of the pericellular matrix (Knudson and Knudson, 1991). Integrin expression in tissue fibroblasts and fibroblast cell lines has been more extensively studied (Wayner and Carter, 1987; Kahari et al. 1991; Peltonen et al. 1989). The attachment of chondrocytes to fibroblasts that were either pretreated with antibodies and paraformaldehyde-fixed or to fixed fibroblasts, in the presence of antibodies, suggests that the adhesive interactions between chondrocytes and fibroblasts is a 2-way process involving receptors on chondrocytes reacting with ligands on fibroblasts and vice versa. The interactive adhesive action between these two cells perhaps results in tighter binding.

Characterization of the ligand-receptor molecules showed that both chondrocytes and fibroblasts produce fibronectin and that this adhesive protein is intimately associated with the cell surface. Synovial fibroblasts have been shown to produce fibronectin (Carson et al. 1989). We have observed, by western blot analysis, that rabbit articular chondrocytes in vitro produce appreciable amounts of fibronectin (M. L. Tiku, preliminary observation). We were able to immunoprecipitate the 140 kDa α subunit and the 110 kDa β subunit of FnR from fibroblasts and chondrocytes under non-reducing conditions; under reducing conditions these chains comigrated as a single 140 kDa chain. These biochemical characteristics are consistent with the integrin nature of the molecules (Ruoslathi and Pierschbacher, 1987). Rabbit synovial fibroblasts have been shown to express three members of the ft family of integrins, VLAj, VLA₃ and VLA₅ (Werb et al. 1989). Compared to fibroblasts, chondrocytes express less VLA₅ as confirmed by immunofluorescence and immunoprecipitation studies. Northern blot studies confirmed that there is a transcriptional message for α and the common β chain of VLA₅ integrin in chondrocytes. Fibroblasts contain more transcriptional message for β1₁ integrin than chondrocytes. Fibroblasts have smaller amounts of VLA₅α message than chondrocytes, suggesting that there may be other α chains that associate with the β1₁ translational product, consistent with the fact that these cells also express VLA₁, VLA₃ and VLA₅ (Werb et al. 1989). However, most of the β1₁ transcriptinal message in chondrocytes is equivalant to message for α chain VLA₅, suggesting that these cells may be only or mostly expressing VLA₅.

Cell-matrix interactions has been shown to regulate cell differentiation (Albelda and Buck, 1990; Adams and Watt, 1989). The attachment of chondrocytes to fibronectin has a profound effect on chondrocyte differentiation; namely, these cells become fibroblast like and lose differentiation functions (Pennypacker et al. 1979; West et al. 1979). Articular chondrocytes obtained from different zones of cartilage show morphological and biochemical characteristics that suggest the presence of distinct types of chondrocytes. For example, a subpopulation of upper zone pig articular chondrocytes that stained weakly with MZ15 (a monoclonal antibody to keratan sulfate) showed less attachment and cell spreading (Zanetti et al. 1985). Recently it was shown that terminal differentiation of kératinocytes involves loss of adhesion to fibronectin, laminin and collagen types I and IV. Kératinocyte adhesion to fibronectin is mediated by VLA₅ integrin and decreased adhesion of intact cells to fibronectin correlates with loss of integrin function (Adams and Watt, 1990). Fibronectin has been shown to inhibit differentiation of human kératinocytes (Adams and Watt, 1989). We have shown that chondrocytes cultured on fibroblasts begin to spread out like fibroblasts. In vitro, usually the cell spreading on purified matrix occurs in a matter of hours (Akiyma et al. 1989). Chondrocytes growing on fibroblasts show gradual changes in cell shape and chondrocytes that are normaly rounded in the first passage become completely fibroblast-like by about 72 h of culture. The expression of chondrogenic function is strictly associated with a spherical cell shape and the corresponding structure of the cytoskeleton (Solursh, 1989; von der Mark, 1986). Therefore changes in the cell shape of chondrocytes that adhere to fibroblasts signal a functional loss of differentiation in these cells (von der Mark, 1986). In vitro, it has been shown that a change in chondrocyte cell shape results in alteration of matrix production (Solursh, 1989). These studies point out that the signals that the cell receives through the engagement of integrin receptors can influence cell function. For example, the engagement of & receptors on synovial fibroblasts results in the induction of metalloproteinase secretion (Werb et al. 1989). In chondrocytes, the engagement of ft receptors may have a negative effect on chondrocytic differentation function. The underlying mechanism by which integrin signals dedifferentation in chondrocytes remains to be investigated.

This work was supported by a grant from the NIH (AR38971-02).