The isolation of collagen-associated proteoglycan from bovine nasal cartilage and its preferential interaction with alpha2 chains of type I collagen.

A collagen complex from bovine nasal cartilage was prepared by extraction of the tissue with 3M-MgCl2 solutions, by using two different procedures. When it was compared with calf skin acid-soluble tropocollagen by polyacrylamide-gel electrophoresis, the 3M-MgCl2-soluble cartilage collagen in the complex appeared to be predominantly type I in nature, consisting of both alpha1 and alpha2 chains. The soluble cartilage collagens were digested with purified bacterial collagenase, and the soluble digests were fractionated on Sepharose 4B. Hydroxyproline-free proteoglycan was isolated in the excluded volume of the column eluate, and this was found to be an aggregate which could be dissociated to link proteins and proteoglycan subunit by equilibrium-density-gradient centrifugation in a CsCl-4M-guanidinium chloride gradient. Interaction with calf skin-soluble tropocollagen was studied by CM-cellulose chromatography. The link-protein system did not interact, but proteoglycan from the bottom of the gradient did interact. In addition, when proteoglycan subunit was allowed to interact with collagen, there was a preferential binding to the alpha2 and beta12 components, and this effect was also observed with the proteoglycan material obtained from the collagenase digests of 3M-MgCl2-soluble cartilage collagen complexes. However, specificity for alpha2 and beta12 chains was not exhibited by chondroitin sulphate glycosaminoglycan, and it is therefore concluded that preference for alpha2 and beta12 chains is a function of the intact proteoglycan structure.     

Specific inhibition of type I and type II collagen fibrillogenesis by the small proteoglycan of tendon.

The small dermatan sulphate proteoglycan of bovine tendon demonstrated a unique ability to inhibit fibrillogenesis of both type I and type II collagen from bovine tendon and cartilage respectively in an assay performed in vitro. None of the other proteoglycan populations from cartilage, tendon or aorta, even those similar in size and chemical structure, had this effect. Alkali treatment of the small proteoglycan of tendon eliminated its ability to inhibit fibrillogenesis, whereas chondroitinase digestion did not. This indicates that its interaction with collagen depends on the core protein. Fibrillogenesis of pepsin-digested collagens was affected similarly, indicating that interaction with the collagen telopeptides is not involved. The results suggest that interactions between collagen and proteoglycans may be quite specific both for the type of proteoglycan and its tissue of origin.     

A study of the interaction in vitro between type I collagen and a small dermatan sulphate proteoglycan.

The interaction between a small dermatan sulphate proteoglycan isolated from human uterine cervix and collagen type I from human and rat skin was investigated by collagen-fibrillogenesis experiments. Collagen fibrillogenesis was initiated by elevation of temperature and pH after addition of proteoglycan, chondroitinase-digested proteoglycan or isolated side chains, and monitored by turbidimetry. Collagen-associated and unbound proteoglycan was determined by enzyme-linked immunosorbent assay after aggregation was complete. (1) The binding of proteoglycan to collagen could be explained by the presence of two mutually non-interacting binding sites, with Ka1 = 1.3 x 10(8) M-1 and Ka2 = 1.3 x 10(6) M-1. The number of binding sites per tropocollagen molecule was n1 = 0.11 and n2 = 1.1. The 0.1 high-affinity binding site per tropocollagen molecule indicates that the strong interaction between proteoglycan and collagen results from a concerted action of tropocollagen molecules in fibrils. Digestion of the proteoglycan with chondroitinase ABC did not affect these binding characteristics. (2) Proteoglycan did not affect the rate of fibrillogenesis, but increased the steady-state A400 by up to 90%. This increase was directly proportional to the saturation of the high-affinity type of binding sites. Neither isolated core protein nor isolated side chains induced a similar high increase in steady-state A400. (3) Electron micrographs showed that the fibril diameter was affected only to a minor extent, if at all, by the proteoglycan, whereas bundles of laterally aligned fibrils were common in the presence of proteoglycan. (4) Results obtained with human and rat collagen were similar.     

The isolation of glycoproteins from bovine achilles tendon and their interaction with collagen

Two glycoprotein fractions, A and B, were isolated from bovine achilles tendon. Glycoprotein A was prepared from a 0.2m-sodium chloride extract and glycoprotein B was isolated from a 3m-magnesium chloride extract. They were free from serum proteins. Glycoprotein A was essentially free of collagen, but glycoprotein B contained about 8% collagen. Both glycoproteins gave several bands on isoelectric focusing. This technique was also used to demonstrate that both glycoprotein fractions interacted strongly with acid-soluble calf skin tropocollagen. It is concluded that these fractions are true components of tendon, and that they may have some function in the macromolecular organization of the tissue.     

The distribution of different molecular species of collagen in fibrous, elastic and hyaline cartilages of the pig.

The distribution of type II collagen, considered to be characteristic of cartilaginous tissues, was determined in various specialized cartilages of the mature pig. The tissues examined were: (1) fibrocartilage of the semilunar meniscus of the knee; (2) elastic cartilage of the external ear; (3) hyaline cartilage of (a) the synovial joint (b) the thyroid plate of the larynx, and (c) the nasal septum. The predominant species of collagen in each tissue, whether type I or type II, was appraised semi-quantitatively by analysis of purified collagen solubilized by pepsin and of peptide fragments produced by cyanogen bromide. Cyanogen bromide-derived peptides were characterized by column chromatography on CM-cellulose and by electrophoresis in sodium dodecyl sulphate-polyacrylamide gels. The proportion of each type of collagen was determined precisely by isolating the homologous small peptides alpha1(II)CB6 [nomenclature of Miller (1973) Clin. Orthop. 92, 260-280], by column chromatography on phosphocellulose and determining their relative proportions by amino acid analysis. Thus collagen of the fibrocartilage of the meniscus proved to be all type I; type II was not detected. In contrast, collagen of elastic cartilage of the outer ear, after rigorous exclusion of perichondrium, was type II. Similarly, type II was the only collagen detected in all the mature hyalline cartilages examined.     

Immunochemistry of cartilage proteoglycan. Immunodiffusion and gel-electrophoretic studies

Cartilage proteoglycan is thought to be composed of subunits, core proteins with covalently attached sulphated polysaccharide side chains, which form aggregates by non-covalent association with a link protein. The new technique of non-disruptive extraction followed by fractionation in caesium chloride gradients provides a useful means of preparing relatively pure proteoglycan aggregate, subunit and link fractions. Immunological studies of these fractions led to the identification of an antigen associated with the proteoglycan subunit which was common to several species and to the demonstration of additional species-specific antigens in aggregate and link fractions derived from bovine nasal cartilage. Polyacrylamide-gel electrophoresis with sodium dodecyl sulphate of bovine proteoglycan aggregate and link fractions gave two protein bands in the gels and a protein-polysaccharide band at the origin; subunit fractions gave only the band at the origin. These results are consistent with the current concept of cartilage proteoglycan structure.     

The mammalian chitinase-like lectin, YKL-40, binds specifically to type I collagen and modulates the rate of type I collagen fibril formation

YKL-40 is expressed in arthritic cartilage and produced in large amounts by cultured chondrocytes, but its exact role is unclear, and the identities of its physiological ligands remain unknown. Purification of YKL-40 from resorbing bovine nasal cartilage and chondrocyte monolayers demonstrated the existence of three isoforms, a major and minor form from resorbing cartilage and a third species from chondrocytes. Affinity chromatography experiments with purified YKL-40 demonstrated specific binding of all three forms to collagen types I, II, and III, thus identifying collagens as potential YKL-40 ligands. Binding to immobilized type I collagen was inhibited by soluble native ligand, but not heat-denatured ligand, confirming a specific interaction. Binding of the chondrocyte-derived species to type I collagen was also demonstrated by surface plasmon resonance analysis, and the dissociation rate constant was calculated (3.42 x 10(-3) to 4.50 x 10(-3) s(-1)). The chondrocyte-derived species was found to prevent collagenolytic cleavage of type I collagen and to stimulate the rate of type I collagen fibril formation in a concentration-dependent manner. By contrast, the cartilage major form had an inhibitory effect on type I collagen fibrillogenesis. Digestion with N-glycosidase F, endoglycosidase H and lectin blotting did not reveal any difference in the carbohydrate component of these two YKL-40 species, indicating that this does not account for the opposing effects on fibril formation rate.     

A role for disulphide bridges in the protein core in the interaction of proteodermatan sulphate and collagen

Proteodermatan sulphate from bovine skin retarded precipitation of fibrils from solutions of purified acid-soluble bovine skin collagen. The isolated protein core was as effective as the intact proteoglycan. Thermal denaturation leading to almost complete loss of the native secondary structure, (determined by circular dichroism spectroscopy to consist of about 60% beta structure) did not diminish the effect unless accompanied by reduction of disulphides, of which there were shown to be three per molecule. The reduced and alkylated protein core was totally ineffective. Electron-microscopy revealed a D-periodic arrangement of glycosaminoglycan on the surfaces of collagen fibrils precipitated in the presence of proteodermatan sulphate. Dermatan sulphate (with attached small peptide) prepared from the proteoglycan, had no effect on the rate of fibrillogenesis and was apparently not bound to the fibrils.     

Cartilage proteoglycan aggregate and fibronectin can modulate the expression of type X collagen by embryonic chick chondrocytes cultured in collagen gels

Chick embryo sternal chondrocytes from the caudal and cephalic regions were cultured within type I collagen gels and type I collagen/proteoglycan aggregate composite gels in normal serum. Caudal region chondrocytes were also cultured within type I collagen gels in the presence of fibronectindepleted serum. There was a marked stimulation of type X collagen synthesis by the caudal region chondrocytes after 9 days in the presence of fibronectin-depleted serum and after 14 days in the presence of proteoglycan aggregate. These results provide evidence for the ability of chondrocytes from a zone of permanent cartilage to synthesise type X collagen and for the involvement of extracellular matrix components in the control of type X collagen gene expression.     

Scorbutic and fasted guinea pig sera contain an insulin-like growth factor I-reversible inhibitor of proteoglycan and collagen synthesis in chick embryo chondrocytes and adult human skin fibroblasts

Chick embryo chondrocytes cultured in sera from scorbutic and fasted guinea pigs exhibited decreases in collagen and proteoglycan production to about 30-50% of control values (I. Oyamada et al., 1988, Biochem. Biophys. Res. Commun. 152, 1490-1496). Here we show by pulse-chase labeling experiments that in the chondrocyte system, as in the cartilage of scorbutic and fasted guinea pigs, decreased incorporation of precursor into collagen was due to decreased synthesis rather than to increased degradation. There was a concomitant decrease in type II procollagen mRNA to about 32% of the control level. As in scorbutic cartilage, proteoglycan synthesis by chondrocytes in scorbutic serum was blocked at the stage of glycosaminoglycan chain initiation. Scorbutic and fasted guinea pig sera also caused a 50-60% decrease in the rates of collagen and proteoglycan synthesis in adult human skin fibroblasts, which synthesize mainly type I collagen. Decreased matrix synthesis in both cell types resulted from the presence of an inhibitor in scorbutic and fasted sera. Elevated cortisol levels in these sera were not responsible for inhibition, as determined by the addition of dexamethasone to chondrocytes cultured in normal serum. Insulin-like growth factor I (IGF-I, 300-350 ng/ml) reversed the inhibition of extracellular matrix synthesis by scorbutic and fasted guinea pig sera in both cell types and prevented the decrease in type II procollagen mRNA in chondrocytes. Therefore, in addition to its established role in proteoglycan metabolism, IGF-I also regulates the synthesis of several collagen types. An increase in the circulating inhibitor of IGF-I action thus could lead to the negative regulation of collagen and cartilage proteoglycan synthesis that occurs in ascorbate-deficient and fasted guinea pigs.     

Peptides of type II collagen can induce the cleavage of type II collagen and aggrecan in articular cartilage.

The objective of this study was to determine whether a fragment(s) of type II collagen can induce cartilage degradation. Fragments generated by cyanogen bromide (CB) cleavage of purified bovine type II collagen were separated by HPLC. These fragments together with selected overlapping synthetic peptides were first analysed for their capacity to induce cleavage of type II collagen by collagenases in chondrocyte and explant cultures of healthy adult bovine articular cartilage. Collagen cleavage was measured by immunoassay and degradation of proteoglycan (mainly aggrecan) was determined by analysis of cleavage products of core protein by Western blotting. Gene expression of matrix metalloproteinases MMP-13 and MMP-1 was measured using Real-time PCR. Induction of denaturation of type II collagen in situ in cartilage matrix with exposure of the CB domain was identified with a polyclonal and monoclonal antibodies that only react with this domain in denatured but not native type II collagen. As well as the mixture of CB fragments and peptide CB12, a single synthetic peptide CB12-II (residues 195-218), but not synthetic peptide CB12-IV (residues 231-254), potently and consistently induced in explant cultures at 10 microM and 25 microM, in a time, cell and dose dependent manner, collagenase-induced cleavage of type II collagen accompanied by upregulation of MMP-13 expression but not MMP-1. In isolated chondrocyte cultures CB12-II induced very limited upregulation of MMP-13 as well as MMP-1 expression. Although this was accompanied by concomitant induction of cleavage of type II collagen by collagenases, this was not associated by aggrecan cleavage. Peptide CB12-IV, which had no effect on collagen cleavage, clearly induced aggrecanase specific cleavage of the core protein of this proteoglycan. Thus these events involving matrix molecule cleavage can importantly occur independently of each other, contrary to popular belief. Denaturation of type II collagen with exposure of the CB12-II domain was also shown to be much increased in osteoarthritic human cartilage compared to non-arthritic cartilage. These observations reveal that peptides of type II collagen, to which there is increased exposure in osteoarthritic cartilage, can when present in sufficient concentration induce cleavage of type II collagen (CB12-II) and aggrecan (CB12-IV) accompanied by increased expression of collagenases. Such increased concentrations of denatured collagen are present in adult and osteoarthritic cartilages and the exposure of chondrocytes to the sequences they encode, either in soluble or more likely insoluble form, may therefore play a role in the excessive resorption of matrix molecules that is seen in arthritis and development.     

The inability to prepare high-buoyant-density proteoglycan aggregates from extracts of normal adult human articular cartilage.

High-buoyant-density proteoglycan aggregates could not be prepared from extracts of adult human cartilage by associative CsCl-density-gradient centrifugation with a starting density of 1.68 g/ml, even though proteoglycan subunits, hyaluronic acid and link proteins were all present. In contrast, aggregates could be prepared when extracts of neonatal human cartilage or bovine nasal cartilage were subjected to the same procedure. This phenomenon did not appear to be due to a defect within the hyaluronic acid-binding region of the adult proteoglycan subunit, but rather to an interference in the stability of the interaction between the proteoglycan subunit and hyaluronic acid towards centrifugation. The factor responsible for this instability was shown to reside within the low-density cartilage protein preparation obtained by direct dissociative CsCl-density-gradient centrifugation of the adult cartilage extract.     

Synthesis of cartilage matrix by mammalian chondrocytes in vitro. II. Maintenance of collagen and proteoglycan phenotype

The in vitro phenotype of bovine articular chondrocytes is described. Chondrocytes plated at high density in roller-bottle and dish cultures were maintained in vitro. The major matrix macromolecules, collagen and proteoglycan, synthesized by these cells were characterized during the course of the culture period. The chondrocytes synthesized mainly Type II collagen, which was found predominantly in the cell-associated matrix. The media contained a mixture of Type II and Type III collagens. Type I collagen was detectable in neither the medium nor the cell-associated matrix. The proteoglycan monomers found in media and cell-associated matrix had the same hydrodynamic sizes as monomers synthesized by cartilage slices or those extracted from adult articular cartilage. The majority of proteoglycans synthesized by the cells were found in high molecular weight aggregates which were readily recovered from the media and were extractable from cell-associated matrix with low ionic strength buffers. The results demonstrate the long-term in vitro phenotypic stability of the bovine articular chondrocytes. The advantages of the in vitro system as a model for studying the effects of external agents, such as drugs and vitamins, are discussed.     

Biochemical and physiochemical characterization of pepsin-solubilized type-II collagen from bovine articular cartilage.

Solubilization of collagen from bovine articular with pepsin requires the preliminary extraction of proteoglycans from the ground substance. Biochemical and physiochemical properties of this pepsin-solubilized collagen are independent of the pretreatment (extraction with 1.5M-CaCl2, 5M-guanidinium chloride or 0.2M-NaOH) and of the age range (2-4-year-old and 2-month-old animals). Characterization of the de-natured components, of the CNBr peptides and of the amino acid and cross-link composition shows that the collagen of the hyaline cartilage is all type II. Electrical birefringence measurements showed the presence of tropocollagen molecules (length 280nm) and molecules whose length is slightly less than twice that of the tropocollagen molecules. This latter molecule may be a dimer composed of two monomers linked by intermolecular head-to-tail bonds and whose theoretical length (530nm), according to the quarter-stagger theory, is in good agreement with our measured values (510-530nm). We have verified that the beta-components of this collagen are formed of two alpha-chains linked by the stable intermolecular bond, dehydrodihydroxylysinonorleucine. These dimeric molecules are absent from solutions of skin collagen whose beta-components possess only aldol-type intramolecular cross-links. Although reconstituted fibres from solutions of skin and cartilage collagen are similar, the segment-long spacing crystallites formed with pepsin-solubilized cartilage collagen present a symmetrical and dimeric form corresponding to the lateral aggregation of two monomers with an overlap (90nm) of the C-terminal ends.     

Covalent structure of collagen. Amino acid sequence of an arthritogenic cyanogen bromide peptide from type II collagen of bovine cartilage

Bovine articular type II collagen was prepared by limited pepsin digestion, differential salt fractionation and carboxymethylcellulose chromatography. Cyanogen bromide digestion of purified type II collagen alpha chains yielded twelve distinct peptides designated CB1-12. The peptide alpha 1(II)-CB11 was isolated by carboxymethylcellulose chromatography and Sephadex G-75S gel filtration. Automated Edman degradation together with chymotrypsin, thermolysin and trypsin digestion enabled identification of its complete amino acid sequence. Compared with type I and type III collagen, the data show similarity with alpha 1(I)-CB8 and alpha 1(III)-CB6-1-8-10-2 peptides, respectively. The peptide is located within residues 124-402 of the alpha 1(II) collagen chain and with its identification, now extends the known amino acid sequence of bovine type II cartilage collagen to 660 amino acid residues including alpha 1(II)-CB1-2-6-12-11-8-10 (partial). This corresponds to alpha 1(I)-CB0-1-2-4-5-8-3-7 (partial; 1-660) and alpha 1(III)-CB3A-3B-3C-7-6-1-8-10-2-4-5 (partial; 1-660) of bovine alpha 1(I) and alpha 1(III) collagen chains.     

The electrophoretic heterogeneity of bovine nasal cartilage proteoglycans.

1. Proteoglycans were extracted from bovine nasal cartilage with 2.0M-CaC2 or with 0.15M-KCl followed by 2.0M-CaC2.. Proteoglycan fractions were prepared from the extracts by density-gradient centrifugation in CsCl under 'associative' and 'dissociative' conditions. 2. The heterogeneity of the proteoglycan fractions was investigated by large-pore-gel electrophoresis. It was concluded that extracts made with 2.0M-CaCl2 or sequential 2.0M-CaCl2 contain two major species of proteoglycan 'subunit' of different hydrodynamic size, together with proteoglycan aggregates. Both 'subunits' have mobilities that are greater than those of proteoglycans obtained from pig articular cartilage McDevitt & Muir (1971) Anal. Biochem. 44, 612-622] and are therefore probably smaller in size than the latter. 3. Proteoglycan fractions isolated from cartilage extracted lith 0.15M-KCl separated into two main components on large-pore-gel electrophoresis with mobilities greater than those of proteoglycans extracted with 2.0M-CaCl2. Proteoglycans extracted at low ionic strength from bovine nasal cartilage are of similar hydrodynamic size to those extracted from pig articular cartilage under the same conditions [McDevitt & Muir (1971) Anal. Biochem. 44, 612-622]. 4. The role of endogenous proteolytic enzymes in producing proteoglycan heterogeneity, particularly in low-ionic-strength cartilage extracts is discussed. 5. Hyaluronic acid and 'link proteins' were present in the proteoglycan fraction separated from KCl extracts as well as in the fraction separated from CaCl2 extracts. Hyaluronic acid can only be identified in proteoglycan fractions by large-pore-gel electrophoresis after proteolysis and further purification of the fraction. 6. Collagen was extracted by both salt solutions and was tentatively identified as type II. Small amounts of collagen appear to be associated with the proteoglycan-aggregate fraction from the high-ionic-strength extract but not with the corresponding fraction from the KCl extract.     

Isolation of heparan sulfate proteoglycan from beneath the monolayers of rat hepatocytes and its binding to type IV collagen

Primary cultures of rat hepatocytes maintained as monolayer in a serum-free medium synthesise and secrete sulphated proteoglycans. Nearly 5% of the total 35(S)-sulphated material was obtained in a soluble form from beneath the cell layer. A shift in gel filtration pattern on beta-elimination with alkali suggested that it is a sulphated proteoglycan. On ion exchange chromatography over Dowex AG 1 x 2, the major fraction was eluted with 1.25 M NaCl. Further, nearly 80% of the 35(S)-labeled material was susceptible to nitrous acid degradation and more than 90% of the material was resistant to chondroitinase ABC digestion suggesting that it is predominantly a heparan sulphate proteoglycan (HSPG). Since HSPG is a major component of basement membrane, its binding with collagen was studied by a solid phase binding assay. About 75% of the 35(S) HSPG bound to wells coated with type IV collagen whereas only about 20% bound to type I collagen at physiological pH. Binding to collagen IV was reduced by about 50% when free GAG chains were used indicating that the protein core is also involved in interaction with the collagen. These results indicate the possible role of this basal extracellular heparan sulphate proteoglycan in the basal lamina formation.     

Cartilage type IX collagen is cross-linked by hydroxypyridinium residues

Type IX collagen, a recently discovered, unusual protein of cartilage, has a segmented triple-helical structure containing interchain disulfides. Its polymeric form and function are unknown. When prepared by pepsin from bovine articular cartilage, type IX collagen was found to contain a high concentration of hydroxypyridinium cross-links, similar to that in type II collagen. Fluorescence spectroscopy located the hydroxylysyl pyridinoline and lysyl pyridinoline cross-linking residues exclusively in the high-molecular-weight collagen fraction, from which they were recovered predominantly in a single CNBr-derived peptide. The results point to a structural role for type IX collagen in cartilage matrix, possibly as an adhesion material to type II collagen fibrils.     

Collagen-binding proteins in collagenase-released matrix vesicles from cartilage. Interaction between matrix vesicle proteins and different types of collagen.

Recent evidence indicates that matrix vesicles (MV) interact with cartilage-specific collagens and other matrix proteins. Both type II and X collagens bind to and cosediment with MV. Our companion study shows that MV also are tightly coupled to proteoglycan link proteins (LP) and hyaluronic acid-binding region (HABR) in cartilage matrix. Here we sought to identify proteins responsible for the nexus between MV and matrix collagens using affinity chromatography with types I, II, and X collagen-Sepharose columns. Elution with NaCl step-gradients in the presence of nonionic detergent was used to assess the affinity between the MV proteins and the covalently attached collagens. Several MV proteins were found to bind to native type I, II, and X collagens but none bound to denatured type I collagen. Alkaline phosphatase, proteoglycan LP and HABR, and the 33- and 67-kDa annexins, bound with varying affinities to the native type I, II and X columns. In particular, LP and HABR, the 67-kDa annexin, and alkaline phosphatase bound with high affinity to the cartilage-specific collagens, although LP, HABR, and a 37-kDa protein also bound less tightly to native type I collagen. Thus, several MV proteins bind specifically to native type II and X collagens and should promote interaction between MV and the extracellular matrix. Such interactions may be important in MV formation, or in MV-mediated mineralization.     

Synthesis of type III collagen by fibroblasts from the embryonic chick cornea

Synthesis of collagen types I, II, III, and IV in cells from the embryonic chick cornea was studied using specific antibodies and immunofluorescence. Synthesis of radioactively labeled collagen types I and III was followed by fluorographic detection of cyanogen bromide peptides on polyacrylamide slab gels and by carboxymethylcellulose chromatography followed by disc gel electrophoresis. Type III collagen had been detected previously by indirect immunofluorescence in the corneal epithelial cells at Hamburger-Hamilton stages 20--30 but not in the stroma at any age. Intact corneas from embryos older than stage 30 contain and synthesize type I collagen but no detectable type III collagen. However, whole stromata subjected to collagenase treatment and scraping (to remove epithelium and endothelium) and stromal fibroblasts from such corneas inoculated in vitro begin synthesis of type III collagen within a few hours while continuing to synthesize type I collagen. As demonstrated by double-antibody staining, most corneal fibroblasts contain collagen types I and III simultaneously. Collagen type III was identified biochemically in cell layers and media after chromatography on carboxymethylcellulose be detection of disulfide-linked alpha l (III)3 by SDS gel electrophoresis. The conditions under which the corneal fibroblasts gain the ability to synthesize type III collagen are the same as those under which they lose the ability to synthesize the specific proteoglycan of the cornea: the presence of corneal-type keratan sulfate.