Interaction of link protein with collagen.

Link protein (Mr = 42,000) is an integral component of cartilage as well as of some noncartilagenous tissues. In cartilage, it forms a macromolecular complex with cartilage proteoglycan and hyaluronic acid, but its function in other tissues is unknown. We provide evidence here that the link protein of cartilage binds well to native collagen types I and III. The binding occurs only if both link protein and collagen are native. The binding of link protein to collagen type fibrils is higher than to monomeric collagen. Link protein binding to collagen fibrils is saturable and occurs at molar ratio of collagen to link protein of 7-13:1. These data suggest that the link protein binds to collagen and that the binding requires the collagen to be in its native triple helical structure. This interaction may play a role in collagen fibril formation.

Link protein (M, = 42,000) is an integral component of cartilage as well as of some noncartilagenous tissues. In cartilage, it forms a macromolecular complex with cartilage proteoglycan and hyaluronic acid, but its function in other tissues is unknown. We provide evidence here that the link protein of cartilage binds well to native collagen types I and 111. The binding occurs only if both link protein and collagen are native. The binding of link protein to collagen type fibrils is higher than to monomeric collagen. Link protein binding to collagen fibrils is saturable and occurs at molar ratio of collagen to link protein of 7-13:l. These data suggest that the link protein binds to collagen and that the binding requires the collagen to be in its native triple helical structure. This interaction may play a role in collagen fibril formation.
Link protein (M, = 42,000) was first discovered in the extracellular matrix of cartilage (1-5). It was originally found in reassociated dissociative extracts of cartilage and was shown to be present as a macromolecular complex along with the cartilage proteoglycan and hyaluronic acid (1-8). Subsequent studies indicated that link protein stabilizes the binding of the cartilage proteoglycan monomer to hyaluronic acid (7)(8)(9)(10)(11)(12). These aggregates, along with type I1 collagen fibers, form the principal components of the cartilage matrix.
The link protein is also present in noncartilagenous tissues including aorta and sclera (13,14). However, sclera lacks the cartilage type of proteoglycan. Further, the proteoglycans of sclera appear to associate through their dermatan sulfate side chains, and do not require the link protein to form aggregates (14,15). Such observations suggest that link protein may serve other as yet unrecognized functions in cartilage and in other tissues.
The interactions of matrix glycoproteins, as well as of proteoglycans, with various collagens regulate the formation of various extracellular matrices. Such interactions are important in establishing the architecture of the matrix and its interaction with the cells in the tissues (reviewed in Refs. 16 and 17). In this paper, we have examined the binding of cartilage link protein to collagen. We find that the link protein has an affinity for the native collagens of the types found in the sclera and in the aorta (types I and 111) and that the binding requires the native triple helical conformation of the collagen molecule.
* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ Visiting Postdoctoral Fellow supported by the Fogarty International Center for Health Sciences.

MATERIALS AND METHODS
Preparation of Collagens-Type I collagen was prepared from an acid extract of rat tail tendon (18); type I1 collagen was prepared from the Swarm chondrosarcoma grown in lathyritic rats (19); type 111 from a pepsin digest of fetal calf skin (20); type IV from a murine tumor, producing a basement membrane (21); and type V from calf skin (22). Ascaris collagen was a gift from C. Sullivan of the National Institute of Dental Research (23). The isolated collagens were examined for purity by SDS-PAGE' (24) and amino acid analyses.
Isolation of the Link Protein-Link protein was isolated from the Swarm chondrosarcoma grown in rats by a modification of a previously published procedure (25). The tumor tissue was extracted with an equal volume (weight/volume) of 1 M guanidine hydrochloride, p H 7.0, containing 0.01 M disodium ethylenediaminetetraacetic acid, 0.05 M sodium acetate, 0.005 M benzamidine hydrochloride, and 0.1 M 6aminohexanoic acid for 18 h at 4 "C. Insoluble material was removed by centrifugation (25,000 X g for 20 min). Cesium chloride was added to the supernatant fraction to a final concentration of 1 g/g of the extract. The extract was then centrifuged in a 50.2 T I rotor (Sorvall) at 33,000 rpm for 68 h a t 10 "C. The bottom one-fifth of the gradient (A1) was collected, diluted with an equal volume of 8 M guanidine hydrochloride, and recentrifuged as described above. The top onefourth of the gradient (A,DJ was collected, dialyzed against 8 M urea, and applied to a DEAE-cellulose column equilibrated with 8 M urea in 0.05 M Tris, pH 7.4 (26). The bound material was then eluted from the column with a linear gradient of NaCl (0-1.0 M NaCl in 0.05 M Tris, pH 7.4, containing 8 M urea). The elution was monitored spectrophotometrically at 280 nm, and the proteins in various fractions were examined by SDS-PAGE. The link protein was present in the unbound fraction and no link protein was present in the NaCIeluted fractions. This fraction was passed over a column of Sepharose CL-GB equilibrated with 4.0 M guanidine hydrochloride. A single major peak with a Kav of 0.59 was eluted by this procedure. This material migrated as a single band of M , = 42,000 when electrophoresed in an SDS-PAGE system after reduction (data not shown).
Preparation of Antibody against Link Protein-Antiserum against link protein was raised in young rabbits (27,28). Link protein (0.2 mg) was injected subcutaneously with complete Freund's adjuvant. Four weeks after the initial immunization, two booster injections (0.1 mg) in incomplete Freund's adjuvant were given in 4-week intervals. Three weeks after the last injection, blood was drawn from the rabbits and the levels of antibody to link protein in the serum were established by an ELISA. The antiserum gave a single line of identity against link protein in the Ouchterlony test.
Specificity of the antiserum was also established by an ELISA in which we compared its reaction with link protein and with cartilage proteoglycan (29)(30)(31). Using a competitive binding assay, varying amounts (0-5 pg/ml) of either link protein or cartilage proteoglycan were mixed with a fixed amount of antibody to link protein in a round bottom microtiter plate (Linbro) at 4 "C for 16 h. The dilutions of antigens and antibodies were done into PBS plus 0.05% Tween 20 (v/v). Then the solutions were transferred to flat bottom microtiter plates that had been precoated with 2 pg of link protein and were incubated for 1 h a t room temperature. The solution was then removed, and the wells were rinsed and incubated for an additional .~ ' The abbreviations used are: SDS-PAGE, sodium dodecyl sulfate-hour with peroxidase-conjugated goat antibody to rabbit immunoglobulin (1:lOOO dilution in PBS-Tween). Subsequently, the plates were rinsed to remove the unbound material and then incubated with substrate (10 mg of o-phenylenediamine in 1.0 ml of methanol, 10 p1 of 30% H202, and 99 ml of distilled H,O). The amount of reaction product was estimated by measuring absorbance a t 492 nm using a Titertek Multiskanner (Flow).
The results presented in Fig. 1 demonstrate that the antiserum is specific for link protein. Cartilage proteoglycan, a possible contaminant of the link protein preparations, did not compete with link protein, a t proteoglycan concentration of 200 pg/well and at an antibody dilution of 1:20. In similar competition ELISA experiments not reported here, we have also observed that the link protein antibody does not recognize type I1 collagen, fibronectin, or chondronectin (up to 100 pg of the proteins and 1:20 dilution of the link protein antibody).
Binding of Link Protein to Natiue and Denatured Collagen-The binding of link protein to microtiter wells coated with various collagens was measured by a modification of the ELISA assay (29)(30)(31). In brief, flat bottom microtiter plates (Dynatech Laboratories) were coated with collagens (20 pg/well) overnight a t 4 "C, and then the coated wells were incubated with various amounts of the link protein dissolved in PBS plus 0.05% Tween 20 (v/v) at room temperature. After 1 h, the solution was removed, and the wells were rinsed (with PBS-Tween) and incubated for an additional hour with antibody to link protein. Subsequently, the ELISA was completed by adding peroxidase-conjugated goat antibodies against rabbit immunoglobulin and the appropriate substrate (see previous section). To test for competition with soluble collagen, link protein was mixed with native or denatured collagen in PBS-Tween in round bottom wells (Linbro) at 4 "C for 16 h, and then transferred to flat bottom microtiter wells coated with native collagen type I. The amount of link protein bound under these conditions was measured using antibody to link protein and the peroxidase-conjugated goat anti-rabbit antibody. As controls, the collagen-coated wells were incubated with either 1) link protein alone, 2) link protein plus antibody to link protein, 3) link protein plus peroxidase-conjugated second antibody, or 4) antibody to link protein plus peroxidase-conjugated second antibody, or 4) antibody to link protein and the second antibody. In each case, the absorbance did not exceed 0.24.
The binding of link protein to collagen was also determined by an immunoblotting procedure (32). Protein solutions ( (0.01-5.0 pg/ ml) were mixed with a constant amount of antibody to link proteln (1:40) in PBS-Tween) and incubated at 4 "C for 16 h. The solutions were then transferred to a flat bottom microtiter plate precoated with link protein (2 pg/well), and the ELISA was completed as described.
collagen, denatured collagen, and BSA) a t four different concentrations (5, 2.5, 1.25, and 0.625 mg/ml in 0.5 M acetic acid) were applied as spots onto nitrocellulose paper and were incubated at 4 "C for 8-16 h. The paper strips containing the collagens or BSA were separately incubated with TBS-BSA a t 4 "C for 12-16 h with gentle rocking. Each paper strip was treated with 10 ml of link protein solution (5 pg/ml in TBS-BSA) and was kept rocking gently at room temperature for 90 min. Control strips coated with either native or denatured collagen or BSA were also incubated with TBS-BSA but without link protein under identical conditions. The paper was then washed five times with TBS and incubated for 90 min with antibody to link protein (1250 in TBS-BSA). Subsequently the paper strips were treated for 90 min with peroxidase-conjugated goat anti-rabbit immunoglobulin and finally with substrate (5.0 mg of diaminobenzidine, 10 p1 of H202 to a final volume of 40 ml) with distilled water. After 1 min, the paper strips were rinsed thoroughly in distilled water to stop the reaction.
In the ELISA procedures and in the nitrocellulose paper assays, it is possible that denatured collagen may not have bound to the solid substrate. In order to rule out this possibility, we have carried out studies in which we tested the binding of fibronectin to native and denatured collagen applied to microtiter plates as well as to nitrocellulose paper. These studies showed a greater binding of fibronectin to denatured collagen than to native collagen (data not presented), in confirmation of previous studies (33,34), and indicate that denatured collagen binds adequately to microtiter wells and nitrocellulose paper. Further ELISA studies using anti-type I collagen antibody established that equal amounts of native and denatured type I collagen bound to microtiter plates.2 I n Vitro Fibril Assays-Fibrils were prepared by incubating collagen in PBS, pH 7.2, at 37 "C for 30 min, either with or without link protein (35,36). These fibrils were centrifuged a t 25,000 X g and washed twice with PBS. The samples were solubilized in sample buffer containing Cleland's reagent and electrophoresed on SDS-PAGE (24). Similar studies were carried out with denatured collagen and link protein or with link protein alone. The amounts of proteins in the precipitates were estimated by scanning the negatives of the photographs (Quick Scan, Helena Laboratories) using gels with known quantities of link protein and collagen to standardize the assay. Scatchard analysis was done by standard procedures.
SDE-PAGE-The discontinuous system of Laemmli was used (24). Briefly, the gel consisted of a 3% polyacrylamide stacking gel and a 7.5% polyacrylamide separating gel. Samples were reduced with 7.5 mg/ml of dithiothreitol. Protein bands were made visible by staining with Coomassie blue R-250.

RESULTS
Binding of Link Protein to Collagen Films-The affinity of the link protein for various isotypes of collagen was tested by measuring the binding of link protein to collagen films in microtiter wells. In these experiments, an excess of collagen (20 pg) was absorbed onto the plastic surface and various amounts of link protein were added to the wells. A non-ionic detergent, Tween 20 (0.5%), was included in the solution of the link protein, to reduce nonspecific adsorption to the surface of the well. Under these conditions, link protein was found to bind best to type I11 collagen and somewhat less to type I collagen (Fig. 2). Link protein bound poorly to films of collagen types 11, IV, and V.
Comparison of Link Protein Binding to Native and Denatured Type Z Collagen-More detailed studies were carried out on the binding of link protein with type I collagen. The importance of the triple helical structure of collagen in the interaction with link protein was evaluated by the film assay. In these studies, the plastic wells were coated with type I collagen, preheated to various temperatures (0,25,40,56, and l0O'C). Much less binding of link protein was observed to the denatured collagen film (Fig. 3). Ascaris collagen, which resembles other collagens in composition but lacks a fibronectin binding site (37), and BSA did not bind link protein (data not shown). These data indicate that the link protein binds better H. K. Kleinman, unpublished observation. to native than to denatured collagen type I and suggest that specific interactions are involved.
Related studies were carried out by measuring whether the binding of link protein to the collagen-coated surface could be inhibited by prior mixing of soluble collagen with the link protein (Fig. 4). These studies showed that native collagen in solution inhibited the binding of link protein to the native collagen surface. Under identical conditions, denatured collagen in solution did not compete for the binding of link protein with a native collagen surface.
The specific binding of link protein to native, but not to denatured, collagen was further established by an immunoblotting procedure (32). In this assay, different concentrations of native and denatured collagens or BSA dissolved in 0.5 M acetic acid were applied as spots onto nitrocellulose papers Binding of link protein to native and denatured collagens-immunoblotting procedure. Native or denatured collagen or BSA solutions were applied as spots (at the indicated amounts) onto nitrocellulose paper and were allowed to dry a t 4 "C for 8 h. The paper strips were treated with 5 pg/ml of link protein and the bound link protein was determined by ELISA.
and allowed to bind to the paper at 4 "C for 8 h. The paper strips were then treated with 3% BSA solution to block the remaining sites. The proteins on the paper were then treated with link protein in the presence of 3% BSA to prevent nonspecific binding. The amount of link protein bound was determined using antibody against link protein. This study again demonstrates that link protein binds only to native collagens (Fig. 5) and is dependent on the amount of collagen. Little or no binding of link protein to denatured collagen (Fig.  5) or BSA to (Fig. 5) was observed.
Binding of Native versus Denatured Link Protein to Collagen-We further confirmed the specificity of the interaction of link protein with collagen by testing if link protein must be in its native conformation. Comparison of the binding of native and denatured (preheated to 100 "C for 5 min) link protein to native type I collagen (Fig. 6) shows that only the native form of link protein is capable of interacting with collagen. Thus, a specific conformation of link protein is also necessary for link protein-collagen interactions.
Binding of Link Protein to Collagen Fibrils-The binding of link protein to collagen was also measured under more physiological conditions. Here, solutions of collagen and a purified preparation of link protein were mixed and then incubated a t 37 "C under conditions which allow fibrils to form. The fibrils that formed were isolated by centrifugation, rinsed, dissolved in electrophoresis buffer, and electrophoresed. The results indicate (Fig. 7) that, under the conditions used, most of the collagen precipitated (Fig. 7, lane I), but none of the link protein precipitated when each protein was incubated alone (Fig. 7, lane 3 ) . However, when mixed together, link protein bound to the collagen fibrils and was present in the precipitate (Fig. 7, lane 2). No precipitate formed when denatured collagen was incubated alone (Fig. 7, lane 4 ) , or when denatured collagen and link protein were incubated together (Fig. 7, lane  5). Further, there was no change in the amount of link protein which bound to the collagen fibrils when denatured collagen was present in the reaction (Fig. 7 lane 6). These data show that link protein binds to collagen fibrils, and that the native helical conformation of collagen is necessary for the interaction. The link protein is not present in the fibrillar precipitate due to nonspecific trapping. When control experiments were carried out with link protein added to a mixture of rabbit preimmune antibody (used here as an antigen) and anti-rabbit antibody (used here as the antibody), under conditions where complexes precipitate, the link protein was not present in the immunoprecipitate (data not shown).
Relative Binding of Link Protein to Monomeric and Fibrillar Collagen-Collagen fibrils were produced by incubating an aliquot of a solution of collagen in PBS at pH 7.2 at 37 "C for 30 min. The suspension of fibrils was transferred to a water bath at 26 "C and incubated for 30 min either with or without link protein. In addition, another aliquot of the original collagen solution was also incubated at 26 "C for 30 min. This temperature (26 "C) was chosen because no fibrils formed from the collagen solution in 30 min and none of the fibrils which had formed at 37 "C dissolved a t 26 "C (data not shown). After 30 min, solid NaCl (20% w/v) was added to both samples. The precipitates that formed were collected by centrifugation and then examined by electrophoresis. Some link protein was precipitated by the 20% NaCl when incubated alone (Table I). However, all the collagen added to the samples was precipitated by this procedure. More link protein was present in the fibrillar rather than in the nonfibrillar collagen precipitate, indicating higher binding of link protein to fibrillar collagen than to monomeric collagen.
Molar Ratio and Strength of Link Protein-Collagen Binding-These studies indicate that less than 1 molecule of link protein is bound to each collagen molecule. A more precise estimate was obtained (Fig. 8) by measuring the amount of link protein that precipitated in the presence of increasing amounts of collagen. The amount of link protein that precip- . Link protein binding to collagen-effect of link protein concentration. A fixed amount of collagen (250 pg) was mixed with the indicated amounts of link protein and the amount of link bound to collagen fibrils was determined as described above (Fig. 7).

Binding of link protein to native collagen and to collagen fibrils
itated was directly proportional to the amount of collagen added. About 5 pg of link protein were precipitated by 450 pg of collagen. Given the ratio of their molecular weights as 6.8, this would indicate that 1 molecule of link protein was binding to 13 molecules of collagen.
In the second experiment ( Fig. 9), varying amounts of link protein were added to 250 pg of collagen and allowed to precipitate. A maximum of 5 pg of link protein bound to the 250 pg of collagen fibrils under these conditions, indicating that there are a fixed number of binding sites in the collagen fibrils. This corresponds to a ratio of 1 molecule of link protein to 7 molecules of collagen. Determination of the affinity of link protein-collagen interactions by Scatchard analysis (inset, Fig. 9) indicates that the equilibrium constant of the reaction is 3.8 X M and the number of equal and independent binding sites on the collagen molecule is 0.26.

DISCUSSION
While the link protein of cartilage is known to be an important component of the aggregate structure of cartilage proteoglycan, recent studies suggest that it may have additional functions. For example, link protein is present in noncartilagenous tissues such as sclera which lack the cartilage type of proteoglycan (13,14,15). Further, even in cartilage, a significant proportion of the link protein is extractable without the denaturing solvents required to dissociate the proteoglycan aggregate.3 In addition, recent studies indicate that in cartilage link protein may be associated with collagen fibrils (38). Our studies reported here show that the link protein purified from a chondrosarcoma binds to collagen films, particularly to those formed of types I and I11 collagen. We have also shown that the binding of link protein requires that both the collagen and the link protein be in their native form. Less than 1 molecule of link protein binds per molecule of collagen.
These data also indicate that link protein binds better to fibrillar collagen than to "monomeric" collagen. Conditions were chosen under which collagen molecules and collagen fibrils were stable as such. We find that almost twice as much link protein binds to the fibrillar than to the monomeric collagen preparation. In other systems, fibrillar collagen is more active than monomeric collagen in inducing platelet aggregation (39,40). Similar structural factors could be involved in the interaction of link protein with fibrillar collagen. Further, studies were carried out with the fibrillar collagen to determine the maximum binding of link protein to collagen. These studies suggest that about 1 molecule of link protein binds to 7-13 molecules of collagen. The binding to fibrils is S. Chandrasekhar, unpublished observation. saturable and is inhibited by native collagen, but not by denatured collagen. The Keg of link protein-collagen interaction as determined by Scatchard analyses (Fig. 9) was 3.8 X lo7 M. In a similar study, the Keg of fibronectin-collagen was found to be M, (41). Our data suggest that although the link protein-collagen interactions are not as strong as that of fibronectin-collagen, they are nevertheless significant.
Link protein binds to collagen fibrils which consist of aggregates of monomeric collagens. The ratio of link protein binding to collagen is 1:7-13. This value is probably an underestimate, because a molecular weight of 300,000 for collagen fibrils is too low. The fibrils are known to be formed of collagen monomers in multiples of 5 (42). Alternatively, it is possible that a contaminant in the collagen preparation is actually binding the link protein. We feel that this is unlikely. The collagen used for study here, as well as the link protein, has been chromatographed on a column of DEAE-cellulose to remove anionic materials, such as hyaluronic acid, proteoglycans, or glycosaminoglycans. Analyses of link protein and collagen indicate that they contain less than 1 pg of uronic acid/0.25 mg of material (not shown). Further, no crossreaction is observed between link protein and cartilage proteoglycan using antibody to either link protein or cartilage proteoglycan. Such results would indicate that the type of material (proteoglycan and hyaluronic acid) known to bind link protein is not involved in the reactions studied here. Also, neither the collagen (Fig. 7) nor the link protein (data not presented) when overloaded onto SDS gels, show any trace of contaminating proteins. Coomassie blue stain is known to detect even 1 pg of protein band (43). In addition, the binding occurs even in the presence of non-ionic detergents such as Tween 20 (ELISA) or in the presence of other proteins such as 3% BSA (immunoblotting). Finally, link protein does not bind to an antigen-antibody precipitate, which was used to measure "trapping" of link by the precipitation. These experiments demonstrate that the binding is specific and is not due to any detectable contamination.
It is not clear why the cartilage link protein binds poorly to type I1 (cartilage) collagen film. Since link protein is present in both cartilage and noncartilage tissues, it is possible that link protein may have tissue-specific interactions. Alternatively, in tissues, the binding interactions may be modulated by additional factors such as proteoglycans and hyaluronic acid.
The functional significance of link protein-collagen interaction is not known. It is generally agreed that link protein binds to proteoglycans and hyaluronic acid and is responsible for the stabilization of the interaction between those macromolecules. The higher binding of link protein to fibrillar collagen than to native collagen indicates that, in uiuo, an opportunity exists for link protein to bind to collagen. Whether that binding occurs either directly or through other components, such as proteoglycans, remains to be studied. Through such binding, link protein may be involved in r e wlating the formation and the morphology of collagen fibrils.

Link Protein Binding
to Collagen