Characterization of Proteoglycans Synthesized by Cultured Corneal Fibroblasts in Response to Transforming Growth Factor β and Fetal Calf Serum*

A culture system was developed to analyze the relationship between proteoglycans and growth factors during corneal injury. Specifically, the effects of transforming growth factor β-1 (TGF-β1) and fetal calf serum on proteoglycan synthesis in corneal fibroblasts were examined. Glycosaminoglycan synthesis and sulfation were determined using selective polysaccharidases. Proteoglycan core proteins were analyzed using gel electrophoresis and Western blotting. Cells cultured in 10% dialyzed fetal calf serum exhibited decreased synthesis of more highly sulfated chondroitin sulfate and heparan sulfate compared with cells cultured in 1% dialyzed fetal calf serum. The amount and sulfation of the glycosaminoglycans was not significantly influenced by TGF-β1. The major proteoglycan species secreted into the media were decorin and perlecan. Decorin was glycanated with chondroitin sulfate. Perlecan was linked to either chondroitin sulfate, heparan sulfate, or both chondroitin sulfate and heparan sulfate. Decorin synthesis was reduced by either TGF-β1 or serum. At early time points, both TGF-β1 and serum induced substantial increases in perlecan bearing chondroitin sulfate and/or heparan sulfate chains. In contrast, after extended periods in culture, the amount of perlecan bearing heparan sulfate chains was unaffected by TGF-β1 and decreased by serum. The levels of perlecan bearing chondroitin sulfate chains were elevated with TGF-β1 treatment and were decreased with serum. Because both decorin and perlecan bind growth factors and are proposed to modulate their activity, changes in the expression of either of these proteoglycans could substantially affect the cellular response to injury.

The extracellular matrix (ECM) 1 of the corneal stroma is synthesized and maintained by keratocytes. The matrix is primarily composed of collagen fibrils stacked in orderly lamellae surrounded by proteoglycans. The organization of proteoglycans and collagen fibrils in the stroma may be responsible for the optical and structural properties of the tissue (1).
The corneal stroma contains two major classes of proteoglycans, one possessing keratan sulfate side chains and the other possessing chondroitin/dermatan sulfate side chains (2)(3)(4)(5). Three corneal keratan sulfate proteoglycans, lumican, keratocan, and mimecan, have been cloned and sequenced (6 -9). The gene for the corneal chondroitin/dermatan sulfate proteoglycan protein core has been cloned from chick corneas and identified as decorin (10). The deduced amino acid sequences of decorin, lumican, and keratocan identify them as members of a group of small leucine-rich proteoglycans (5,8).
The structural and biochemical properties of ECM molecules in the corneal stroma are altered upon injury. Corneal wounds contain collagen fibrils with abnormally large diameter and irregular interspacing (11,12). Disruption of the fibrillar organization of collagen fibrils in corneal wounds is thought to be attributed, in part, to alterations in the proportion and chemical characteristics of specific proteoglycans. Injured corneas contain unusually large chondroitin/dermatan sulfate proteoglycans possessing glycosaminoglycan (GAG) side chains with higher than normal sulfation and increased amounts of iduronic acid (13). Keratan sulfate (KS) chains in corneal scars have increased size and lower sulfation (14,15). The ratio of chondroitin/dermatan sulfate to keratan sulfate has been shown to increase after wounding, and heparan sulfate (HS) has been detected in corneal scars (13,16,17). Interestingly, both transforming growth factor-␤ (TGF-␤) and basic fibroblast growth factor are detected transiently in corneal wounds coincident with the expression of heparan sulfate proteoglycans (HSPGs) (16,18).
TGF-␤ has been implicated as a regulatory agent in numerous cellular and physiological processes, including proteoglycan expression (19). This influence appears to be at the level of core protein synthesis and GAG chain elongation (20). TGF-␤ has been detected in corneal wounds and in corneal fibroblast cultures, suggesting that it plays a role in regulating the synthesis of stromal ECM components (16,21,22). Although TGF-␤ has been detected in vivo and in vitro, the relationship between TGF-␤ and proteoglycan expression by corneal fibroblasts has not been fully elucidated.
Cultured corneal fibroblasts synthesize proteoglycans remarkably similar to those in wounded corneas. Early reports indicated cultures of rabbit corneal fibroblasts produce mainly chondroitin sulfate (CS) and HS, with only low levels of KS (23)(24)(25). Hassell et al. (26) reported human corneal fibroblasts in culture synthesize substantial amounts of decorin and perlecan (basement membrane HSPG). Schrecengost et al. (27) reported reduced levels of a keratan sulfate proteoglycan containing truncated unsulfated keratan chains in chick corneal fibroblasts. Funderburgh et al. (28) reported that bovine cor-neal fibroblasts in culture synthesize keratan sulfate proteoglycans with shorter KS chains and lower sulfation compared with those in normal corneas. The altered properties of KS and the increase in chondroitin sulfate proteoglycan (CSPG) and HSPG synthesis suggests that the conditions of cell culture may recapitulate some of the aspects of injury. To date, most studies of corneal proteoglycans produced in vitro have been based upon biochemical analysis of GAG chains, with only limited analysis of the protein cores. Additionally, each of these studies employed different culture techniques, making the results difficult to compare.
We developed a culture system to examine the regulation of proteoglycan synthesis by corneal fibroblasts during injury. The major proteoglycans synthesized by corneal fibroblasts were characterized and identified after culture in a defined environment. Specifically, we evaluated the effects of TGF-␤1 and serum on the synthesis of specific GAGs and protein cores. We found that corneal fibroblasts synthesized predominantly CS and HS, with only trace amounts of an unsulfated form of keratan. The major proteoglycan species secreted into the medium were decorin and perlecan, and proteoglycan synthesis was mediated by TGF-␤1 and serum. This model will allow us to systematically examine the relationship between specific growth factors and proteoglycan expression using a defined culture system.

EXPERIMENTAL PROCEDURES
Materials-Chondroitinase ABC (protease-free), keratanase II, chondroitin sulfate B, keratan sulfate, and the mouse monoclonal antibody 3G10 directed against unsaturated uronic acid residues arising from heparinase digestion of heparan sulfate were purchased from Seikagaku America, Inc. (Ijamsville, MD). Endo-␤-galactosidase was purchased from Boehringer Mannheim. Heparan sulfate, heparinase I, heparinase III, phenylmethylsulfonyl fluoride, benzamidine, N-ethylmaleimide, and peroxidase-conjugated donkey anti-sheep IgG antibodies were from Sigma. Q-Sepharose came from Pharmacia Biotech Inc. (Uppsala, Sweden). Peroxidase-conjugated donkey anti-rat IgG antibodies were purchased from Amersham Pharmacia Biotech . Ultrapure urea, sodium chloride, Tween-20, Tris-HCl, bovine serum albumin, and EDTA were obtained from American Bioanalytical (Natick, MA). TGF-␤1 was obtained from R & D Systems (Minneapolis, MN). All cell culture reagents were purchased from Life Technologies, Inc. The mouse monoclonal antibody A7L6 directed against perlecan was obtained from Upstate Biotechnology Inc. (Lake Placid, NY). The polyclonal sheep antiserum directed against rabbit corneal decorin was a generous gift from Dr. Charles Cintron (Schepen Eye Research Institute, Boston, MA). The rabbit polyclonal antibody R36 that binds unsaturated uronic acid residues resulting from chondroitinase ABC treatment was a generous gift from Dr. John Couchman (University of Alabama, Birmingham, AL).
Corneal Fibroblast Isolation and Cell Culture-Corneas were excised from whole rabbit eyes purchased from Pel Freeze (Rogers, AR), and the epithelium and endothelium were removed as described previously (29). The corneas were washed two times with Dulbecco's modified Eagle's medium (DMEM) containing 1000 units/ml penicillin, 1.0 mg/ml streptomycin sulfate, and 20 units/ml nystatin. The corneas were minced with a sterile razor blade and subsequently digested with collagenase A (1.5 mg/ml) in DMEM containing 200 units/ml penicillin, 200 g/ml streptomycin, and 100 units/ml nystatin for 2-3 h with agitation at 37°C. The digests were centrifuged at 1840 ϫ g for 10 min, and the cells were suspended in DMEM supplemented with 100 units/ml penicillin, 100 g/ml streptomycin, 100 units/ml nystatin, nonessential amino acids, and 10% fetal calf serum (FCS). Cell were plated in 75-mm vented tissue culture flasks, and cultures were maintained in DMEM supplemented with 100 units/ml penicillin, 100 g/ml streptomycin, 100 units/ml nystatin, nonessential amino acids and 4% fetal calf serum. The cultures achieved confluency after 7-10 days, at which time cells were passaged 1:4 and cultured in 4% FCS for 3 days. All experiments were performed on confluent fibroblast cultures that had been passaged once.
Cell Treatment and Metabolic Labeling-The synthesis of sulfated glycosaminoglycans was followed by metabolically radiolabeling corneal fibroblasts with [ 3 H]glucosamine (18 Ci/ml) and/or [ 35 S]sulfate (36 Ci/ml). Proteoglycan core proteins were metabolically labeled with [ 35 S]cysteine/methionine (50 Ci/ml). Corneal fibroblasts in first passage were cultured until confluent (3 days) in 4% FCS. Upon confluence, corneal fibroblasts were treated as indicated in the figure legends. Radioisotopes were added immediately after the addition of TGF-␤1.
After the designated radiolabeling period, the medium was collected and immediately combined with two volumes of 10 M urea containing 50 mM Tris-HCl, 10 mM EDTA, pH 7.0. Cell monolayers were washed with phosphate buffered saline (pH 7.4) and ECM proteins were isolated by gently scraping cell monolayers in 1.0 M urea, 50 mM Tris-HCl, 50 mM EDTA, pH 7.0. The resulting suspension was centrifuged (5520 ϫ g) for 10 min, and the supernatant was collected and defined as the ECM fraction. The cell pellet was extracted with TUT (8 M urea, 50 mM Tris-HCl, 0.1% Triton X-100, pH 7.0). The extract was clarified by centrifugation, and the supernatant was collected and defined as the cell fraction (30). Cell number was determined by measuring acid phosphatase activity on a replicate set of cultures (31). Total radiolabeled protein present in [ 35 S]cysteine/methionine labeled samples was determined by performing trichloroacetic acid precipitation on aliquots of medium and cell fractions prior to proteoglycan purification and quantitating the radioactivity in a liquid scintillation counter (27).
Proteoglycan Purification-Medium, cell, or ECM fractions were mixed with 1.0 ml of a 70% Q-Sepharose suspension and rocked for 45 min. The slurries were poured into 5.0-ml disposable minicolumns (Pierce), and the unbound fractions were discarded. The columns were washed with 25 column volumes of TUE (8 M urea, 50 mM Tris-HCl, 50 mM EDTA, pH 7.0) and subsequently washed with 25 column volumes TUE containing 0.2 M NaCl. Columns were eluted with 7 column volumes of TUE containing 1.5 M NaCl. Salt fractions were exhaustively dialyzed against Milli-Q water, using membranes with a molecular weight cutoff of 25,000 (Spectropore, Laguna Hills, CA), and lyophilized. Samples were resuspended in 2 mM sodium phosphate, 30 mM NaCl, 1 mM phenylmethylsulfonyl fluoride, 10 mM N-ethylmaleimide, pH 7.4.
Selective Polysaccharidase Treatment-Selective polysaccharidases were used to identify and quantitate GAGs and proteoglycan core proteins. Digestion conditions were optimized for time, temperature, concentration, and pH. Enzymes were routinely tested for activity and specificity using highly purified GAG standards (Seikagaku, Tokyo, Japan) and the dimethyl methylene blue assay (32). Purified proteoglycans were subjected to digestion for 3 h at 37°C in 40 mM Tris-HCl. The pH of the digest was adjusted to the optimum for each enzyme: chondroitinase ABC (1.0 unit/ml, pH 8.0), a mixture of heparinase I and heparinase III (10 and 20 units/ml respectively, pH 7.3), and a mixture of keratanase II and endo-␤-galactosidase (both enzymes 0.01 unit/ml, pH 5.9).
Glycosaminoglycans Analysis-Specific GAGs were quantitated by measuring the low molecular weight digestion products released after polysaccharidase treatment. Purified GAGs co-radiolabeled with [ 3 H]glucosamine and [ 35 S]SO 4 were treated with chondroitinase ABC, a mixture of heparinase I and heparinase III, a mixture of keratanase II and endo-␤-galactosidase, or control with buffer lacking enzyme. Digests were subjected to ultrafiltration (Microcon, Millipore) to separate GAG digestion products from intact proteoglycans. The radioactivity in the filtrate was determined using liquid scintillation. A 10,000 molecular weight cut-off filter was used to recover CS and KS digestion products, and a 30,000 molecular weight cut-off filter was used to recover HS digestion products. The radioactivity in the filtrate from the undigested control was subtracted from the enzyme treated samples.
In experiments with a large number of samples, fractions containing [ 35 S]SO 4 -GAGs were analyzed, without any prior purification. [ 35 S]Glycosaminoglycans in medium, cell, and ECM fractions were quantitated by dot-blotting samples onto cationic nylon filters as described previously (30). Briefly, filters (Zeta-probe; Bio-Rad) were prehydrated in TBS (50 mM Tris-HCl, 0.15 M NaCl, pH 8.0). The filter was then placed into a Bio-Dot apparatus (Bio-Rad) and washed once by drawing TBS through each well with vacuum. Samples (100 l) were applied to each well and pulled through under vacuum and wells were washed with 0.6 ml of TUT. The filter was washed twice with TBS followed by two additional washes with Milli-Q water (Millipore, Bedford MA). The washed filter was briefly immersed in 95% ethanol, and the area of the filter containing each sample was removed and counted using liquid scintillation.
Heparan sulfate and CS were determined by treating a replicate filter with nitrous acid, which selectively degrades HS chains. After sample application, washed filters were treated twice with fresh nitrous acid (0.48 M sodium nitrite combined with 3.6 M acetic acid) for 90 min followed by a wash with TBS containing 0.65 M NaCl. The difference between the radioactivity measured on the non-acid-treated and acid-treated filters was defined as HS. The amount of radioactivity remaining after nitrous acid treatment was defined as CS.
Proteoglycan Core Protein Analysis-Proteoglycan core proteins were identified and quantified using selective polysaccharidases in conjunction with SDS-PAGE and/or Western blotting. Proteoglycans radiolabeled with [ 35 S]cysteine/methionine were digested with either chondroitinase ABC, heparinases I and III, or a mixture of chondroitinase ABC and heparinases I and III. Digests were run on 5 or 10% SDS-PAGE gels under reducing conditions (33), and loading was normalized to total radioactive protein present in fractions prior to purification (trichloroacetic acid precipitation). Gels were either processed for autoradiography or electrophoretically transferred to polyvinylidene difluoride membrane (Millipore) using a semi-dry transfer apparatus (integrated separations systems) in 25 mM Tris, 192 mM glycine, 20% methanol. The membranes were blocked in 5% bovine serum albumin in TBS-T buffer (10 mM Tris, 100 mM NaCl, 0.1% Tween 20, pH 7.2) and were incubated with either mouse monoclonal anti-perlecan (2 g/ml), sheep polyclonal anti-rabbit corneal decorin (1:9000), mouse anti-HS stub (1:1000), or rabbit anti-CS stub in 5% bovine serum albumin in TBS-T at room temperature for 1 h. Blots were washed with 1% bovine serum albumin in TBS-T and incubated with appropriate secondary antibodies coupled to horseradish peroxidase (1:3000) for 1 h at room temperature. Proteins were visualized using chemiluminescence (NEN Life Science Products). To distinguish between chemiluminescence and radioactivity, a piece of transparent plastic was placed between the membrane and the film, which was demonstrated to block Ͼ95% of the radioactivity. After probing, the blots were treated with 10% sodium salicylate in methanol and dried. Radiolabeled proteins were detected using autoradiography.

Effects of TGF-␤1 and Fetal Calf Serum on GAG Synthesis-
Confluent corneal fibroblasts were cultured for 96 h in either 1 or 10% dialyzed fetal calf serum with or without daily treatments with TGF-␤1 (1 ng/ml). Cellular viability studies were performed, and no significant differences were observed after 6 days of culture in 0, 1, or 10% FCS (data not shown). To evaluate GAG synthesis and sulfation, cells were metabolically labeled with both [ 35 S]SO 4 and [ 3 H]glucosamine. Glycosaminoglycans in medium, cell, and ECM fractions were purified using anion exchange chromatography on Q-Sepharose. The chromatographic conditions were optimized to separate highly charged proteoglycans from weakly charged glycoproteins and hyaluronic acid (16). Specific GAGs were quantitated by selective digestion with polysaccharidases. Chondroitin sulfate was determined using chondroitinase ABC. Glycosaminoglycans susceptible to chondroitinase ABC are referred to as CS because this enzyme does not distinguish between polymers containing iduronate and glucuronate (34). Keratan sulfate was determined using a mixture of keratanase II and endo-␤-galactosidase, and HS was determined using a mixture of heparinases I and III. Sulfation was defined as the ratio of polysaccharidase-sensitive [ 35 (Table III). Moreover, this increase in GAG sulfation must result from some nonspecific alteration in GAG metabolism, as both extracellular and cellular populations of CS and HS exhibited this response. TGF-␤1 did not significantly influence the sulfation of GAGs synthesized by cells cultured in 1 or 10% dFCS.
The decrease in GAG synthesis observed with culture in 10% dFCS compared with 1% dFCS seems to be influenced by TGF-␤1. Corneal fibroblasts exhibited smaller serum-associated decreases in [ 3 H]CS and [ 3 H]HS isolated from cell and ECM fractions when TGF-␤1 was present (Table IV). To evaluate the extent of this response, GAGs synthesized by corneal fibroblasts cultured in 1 or 10% dFCS with or without TGF-␤1 (1 ng/ml) were monitored over a 96-h time course using [ 35 S]SO 4 incorporation. Fig. 1   H]glucosamine. Glycosaminoglycans were purified using anion exchange chromatography on Q-Sepharose. Purified GAGs were digested with selective polysaccharidases. Digests were subjected to ultrafiltration (Microcon, Millipore), and the radioactivity in the filtrate was counted. The radioactivity in the filtrate from an untreated control was subtracted from the enzyme-treated samples. Results are expressed as cpm per 1 ϫ 10 3 cell Ϯ S.E. (n ϭ 3). ing was normalized to total radioactive protein present in prior to purification on Q-Sepharose columns. Electrophoresis of proteoglycans from the medium on 10% SDS-PAGE gels without prior enzyme treatment resulted in poorly resolved smears migrating between M r 140,000 and 200,000. After digestion with chondroitinase ABC, these smears were no longer apparent, and bands with M r Ϸ 45,000, M r ϭ 60,600, and M r ϭ 119,100 were present (Fig. 2.). A core protein (M r ϭ 60, 600) was also released by heparinase treatment. Digestion with both chondroitinase ABC and heparinases I and III did not release any additional core proteins or substantially alter the electrophoretic mobility of the protein cores. A chondroitin/dermatan sulfate proteoglycan core (M r Ϸ 45,000) protein migrated as a doublet and was identified as decorin using Western blot analysis (Fig. 3.). After 24 h, decorin levels were 1.4-fold higher in cells cultured in 1% dFCS than in cells cultured in 10% dFCS. Cultures treated with TGF-␤1 had decorin levels similar to controls. After 96 h, decorin was 2.1fold higher in cells cultured in 1% dFCS than in cells cultured in 10%. TGF-␤1 decreased decorin by 32 and 19% in 1 and 10% dFCS respectively. The M r ϭ 60,600 and M r ϭ 119,100 core proteins detected after 96 h in culture migrated with M r consistent with those of syndecan-1 and betaglycan (35,36). Syn-decan-1 belongs to the class of transmembrane proteins that undergo proteolytic cleavage and release their ectodomains into the extracellular environment (37,38). Betaglycan has been reported to exist as a soluble form that is released by cells into the medium and is found in the extracellular matrices and serum (36,39).
Analysis of lyase-treated proteoglycans from the medium on 5% SDS-PAGE gels revealed the presence of three high M r proteoglycan core proteins (M r Ϸ 375,000, Ϸ 440,000, and Ϸ 480,000) (Fig. 4). The M r Ϸ 375,000 protein core was released by treatment with either chondroitinase ABC, heparinases I and III, or a mixture of both lyases, indicative of a proteoglycan bearing either CS, HS, or both CS and HS. The M r Ϸ 440,000 core protein could only be resolved after treatment with heparinases I and III or with both heparinases and chondroitinase ABC and was therefore synthesized either as an HSPG bearing only HS chains or a hybrid proteoglycan possessing both HS and CS chains. Both the M r Ϸ 375,000 and M r Ϸ 440,000 core proteins containing either HS and/or CS chains reacted with a monoclonal antibody directed against perlecan (Fig. 5). This heterogeneity observed in the size of the perlecan core protein is suggestive of alternative splicing, as has been reported in several species (40). Heparan sulfate proteoglycan and CSPG  4 . Glycosaminoglycans were purified using anion exchange chromatography on Q-Sepharose. Purified GAGs were digested with selective polysaccharidases. Digests were subjected to ultrafiltration (Microcon, Millipore), and the radioactivity in the filtrate was counted. The radioactivity in the filtrate from the undigested control was subtracted from the enzyme-treated samples. Results are expressed as cpm per 1 ϫ 10 3 cell Ϯ S.E. (n ϭ 3).  forms of perlecan have been reported in the Engelbreth-Holm-Swarm tumor matrix (41). The M r Ϸ 480,000 protein core was derived from a CSPG, as it was only observed after treatment with chondroitinase ABC or both chondroitinase ABC and heparinase treatment. The M r Ϸ 480,000, core protein did not react with the antibody directed against perlecan. The smear between M r Ϸ 400,000 and 460,000 in both the heparinasetreated and untreated control lanes was not present after treatment with chondroitinase ABC and likely represents an intact CSPG.
To determine the effects of TGF-␤1 and serum on the secreted high molecular weight core proteins, bands detected on autoradiographs after combined chondroitinase ABC and heparinases I and III treatment were compared by densitometry. A 1.6-fold increase in the amount of the M r Ϸ 375,000 perlecan secreted by TGF-␤1-treated corneal fibroblasts cultured in 1% dFCS was observed after 24 h, and a 2.0-fold increase was observed after 96 h. This stimulatory effect of TGF-␤1 did not depend on serum factors, as similar increases were observed when cells were cultured in 10% dFCS. After 96 h of culture in 10% dFCS, there was a decrease in the amount of both the M r Ϸ 375,000 and the M r Ϸ 440,000 forms of perlecan by 1.6-and 2.2-fold, respectively, compared with cells cultured in 1% dFCS. Similar serum associated decreases were seen when TGF-␤1 was present. After 96 h, the levels of the M r Ϸ 440,000 core proteins were not significantly affected by TGF-␤1 treatment. The pattern of expression of the M r Ϸ 480,000 protein core was similar to that of the M r Ϸ 375,000 with respect to TGF-␤1 treatment and serum. Densitometric analysis of the immunoblotted core proteins released with combined chondroitinase ABC and heparinase treatment (Fig. 5B) revealed substantial increases in both the M r Ϸ 440,000 (2.2-fold) and M r Ϸ 375,000 (2.1-fold) forms of perlecan with TGF-␤1 treatment when cells were cultured in 1% dFCS for 24 h. Furthermore, culture in 10% dFCS for 24 h increased the levels of the M r Ϸ 440,000 core by 2.3-fold with respect to 1% dFCS.
Polysaccharidase-treated proteoglycans purified from cell fractions were analyzed by Western blot analysis. Blots were probed with antibodies that recognize CS stubs remaining after chondroitinase ABC treatment (R36) or HS stubs remaining after heparinase treatment (mAb 3G10). Western blot analysis with R38 revealed the presence of a M r ϭ 44,700 band that was released only with combined chondroitinase and heparinase treatment (Fig. 6A). Interestingly, this band did not react with mAb 3G10. It is conceivable that chondroitinase ABC may have modified the epitope for mAb 3G10. Heparinases I and III released a M r ϭ 51,600 protein core that reacted with mAb 3G10 (Fig. 6B). This band was not detected after combined chondroitinase ABC and heparinase treatment. The presence of both CS and HS chains and the electrophoretic mobility suggest that these proteoglycans may be members of the syndecan family (35). After 96 h, both the M r ϭ 44,700 and M r ϭ 51,600 protein cores were decreased in cells cultured in 10% dFCS compared with 1% dFCS but were unchanged by treatment with TGF-␤1.
Effects of TGF-␤1 on Proteoglycan Synthesis-To examine the effects of TGF-␤1 in the absence of serum factors, corneal fibroblast were cultured without dFCS for 18 h and subsequently treated with 0, 1, 5, or 10 ng/ml TGF-␤1. Proteoglycans were radiolabeled with [ 35 S]SO 4 for 0 -24, 24 -48, or 48 -72 h, and medium fractions were collected after the radiolabeling periods. Aliquots of medium were treated with chondroitinase ABC or left untreated, and resistant proteoglycans were isolated on Q-Sepharose columns. Chondroitin sulfate was defined as GAG susceptible to chondroitinase ABC. Heparan sulfate was defined as GAG resistant to chondroitinase ABC.
Corneal fibroblasts cultured in the absence of serum exhibited TGF-␤1-dependent changes in GAG synthesis. During the 0 -24 h labeling period, TGF-␤1 treatment resulted in a dosedependent increase in both CS (up to 2.0-fold) and HS (up to 2.3-fold) (Fig. 7, A and B). From 24 to 48 h, the overall synthesis of CS was not significantly affected by TGF-␤1 treatment. However, cells treated with TGF-␤1 continued to show increased HS (up to 3.0-fold) synthesis during the 24 -48-h labeling period. During the 48 -72-h labeling period, substantial decreases in GAG synthesis were observed after TGF-␤1 treatment. Chondroitin sulfate synthesis decreased by as much as 13.6-fold relative to control. During this period, HS synthesis, although modestly elevated after treatment with 1 ng/ml TGF-␤1, exhibited a 5.7-fold decrease relative to control after treatment with either 5 or 10 ng/ml TGF-␤1.
The TGF-␤1-dependent increase in GAG production seems to involve the induction of specific proteoglycans. Aliquots of medium from TGF-␤1-treated corneal fibroblasts radiolabeled with [ 35 S]SO 4 were analyzed on 5% SDS-PAGE gels. Bands exhibited broad size heterogeneity characteristic of proteoglycan migration on SDS-PAGE gels (Fig. 7C). Three major populations of proteoglycans were observed; a low M r proteoglycan (median M r Ϸ 175,000), a high M r proteoglycan (median M r Ϸ 420,000), and a second high M r proteoglycan that barely entered the gel. To quantitate TGF-␤1-induced changes in secreted proteoglycans, bands detected on autoradiographs were Radioactive proteins were detected using autoradiography. Sample loading was normalized to the amount total radioactive protein present in medium (trichloroacetic acid precipitation) prior to anion exchange chromatography. Data are representative of at least three experiments. FIG. 3. Analysis of decorin. Corneal fibroblasts were cultured for 24 h in 1 or 10% dFCS Ϯ TGF-␤1 (1 ng/ml). Radiolabeled ([ 35 S]Cys/Met) proteoglycans purified from the medium were digested with chondroitinase ABC or buffer lacking enzyme. Digests were run on 10% SDS-PAGE gels under reducing conditions. Proteins were electrophoretically transferred to polyvinylidene difluoride. A, blots were probed with polyclonal antiserum raised against rabbit corneal decorin. B, after probing, radioactive proteins were detected using autoradiography. Sample loading was normalized to the amount total radioactive protein present in medium (trichloroacetic acid precipitation) prior to Q-Sepharose chromatography. Data are representative of at least three experiments.
FIG. 4. Analysis of high M r core proteins. Corneal fibroblasts were cultured for 96 h in 1 or 10% dFCS Ϯ TGF-␤1 (1 ng/ml), and proteins were radiolabeled with [ 35 S]Cys/Met. Proteoglycans purified from the medium were digested with chondroitinase ABC, both chondroitinase ABC and heparinases I and III, or buffer lacking enzyme. Digests were run under reducing conditions on 5% SDS-PAGE gels. Sample loading was normalized to total radioactive protein present in medium (trichloroacetic acid precipitation) prior to purification. Radioactive proteins were detected using autoradiography. Data are representative of at least three experiments. compared by densitometry. The second high M r proteoglycan was not sufficiently resolved to be accurately quantitated. TGF-␤1 treatment resulted in dose dependent increases in the high M r proteoglycans from 0 to 24 h (up to 4.7-fold) and from 24 to 48 h (up to 3.3-fold). In contrast, the low M r decreased by as much as 47% during the 0 -24 h labeling period and by as much as 70% during the 48 -72 h labeling period. During the 48 -72 h labeling period, TGF-␤1 treatment resulted in substantial decreases in both the low and high M r proteoglycans. The low M r proteoglycan decreased by 76% in 1 ng/ml TGF-␤1, relative to control, and was not detected at higher concentrations of TGF-␤1. The high M r was not significantly changed by 1 ng/ml TGF-␤1, and was not detected at higher concentrations of TGF-␤1. These TGF-␤1-dependent decreases in proteoglycan synthesis from 48 to 72 h after treatment were not the result of decreased cell viability as cells treated with TGF-␤1 exhibited similar levels of [ 3 H]thymidine incorporation into DNA during this period. DISCUSSION The current study was initiated to identify the major core proteins and GAG chains synthesized by rabbit corneal fibroblasts in culture. A defined culture system will allow the systematic examination of the relationship between specific growth factors and proteoglycans within the injured cornea in vitro. This system should provide a useful model of corneal injury. Because TGF-␤ has been detected in the corneal stroma after injury, we examined the effects of TGF-␤1 and serum on proteoglycan synthesis by corneal fibroblasts (16,21). The results of these studies showed that the synthetic profile of proteoglycans produced by corneal fibroblasts in culture, although significantly different from those of normal corneas, were remarkably similar to those found in wounded corneas. Approximately 60% of the GAG in the normal corneal stroma is KS and 40% is chondroitin/dermatan sulfate (2)(3)(4). Wounded corneal stromas synthesize increased quantities of both CS and HS and have reduced KS content (13)(14)(15)(16)(17). Corneal fibroblasts, in our culture system, synthesize substantial quantities of CS and HS with negligible amounts of an unsulfated form of kera-tan. Hassell et al. (13) reported that wounded corneas synthesize unusually large CSPGs (13). We detected substantial quantities of large proteoglycans bearing CS and HS chains secreted into the culture medium. Several reports have documented increased sulfation of CS and HS and decreased sulfation of KS after wounding (14 -15, 17). Keratan was not sulfated in our system, whereas the sulfation of CS and HS was significantly increased with increased serum. These results suggest that conventional cell culture and injury induce similar phenotypic changes and that altered proteoglycan expression is a reflection of these changes.
In addition to the GAG chains, we extended the analysis of proteoglycan production by characterizing and identifying core proteins secreted into the medium and associated with the cells. Four major species of protein core were observed in the medium of cultured corneal fibroblasts. The smallest core protein (M r Ϸ 45,000) possessed only CS chains and was identified as decorin. Two large core proteins were identified as perlecan by immunoblotting (M r Ϸ 375,000 and Ϸ 440,000). An additional CSPG containing a M r Ϸ 480,000 protein core was detected that did not react with the antibody directed against perlecan. Furthermore, we demonstrated that serum and TGF-␤1 influenced both the expression and A, immunoblot analysis of proteoglycans digested with chondroitinase ABC, chondroitinase ABC and heparinase I and III, or a buffer lacking enzyme. Blots were probed with antibody R36 directed against the CS stub remaining after chondroitinase ABC digestion. B, immunoblot analysis of proteoglycans digested with heparinase I and III, chondroitinase ABC and heparinase I and III, or a buffer lacking enzyme. Blots were probed with mAb 3G10 directed against the HS stub remaining after heparinase treatment. Sample loading was normalized to total radioactive protein present (trichloroacetic acid precipitation) prior to purification. Data are representative of at least three experiments. glycanation of these core proteins and altered the sulfation of their associated GAG chains.
Decorin is a member of the gene family of small leucine-rich proteoglycans and a normal constituent of the corneal stroma (10). There is increasing evidence that decorin is an important regulator of a number of important physiological processes. Several studies suggest an inhibitory role for decorin on cell proliferation through TGF-␤-dependent and TGF-␤-independent mechanisms. Overexpression of decorin inhibits cell growth in a number of different cell types (42)(43)(44)(45). Addition of exogenous recombinant decorin to cultures of several tumor cell lines suppresses cell growth (43). Decorin suppresses cell growth by activating the epidermal growth factor receptor and elevating cytosolic Ca 2ϩ in A431 squamous carcinoma cells (46 -47). In addition, Yamaguchi et al. (48) demonstrated that decorin specifically binds TGF-␤ and suppressed its growth stimulatory activity in Chinese hamster ovary cells. The ability of decorin to modulate the bioactivity of TGF-␤1 has been demonstrated in a number of different systems (49 -51). The effects of TGF-␤ on decorin expression vary widely among different fibroblast types. TGF-␤ down-regulates decorin expression in dermal (52) and gingival fibroblasts (53), whereas TGF-␤ up-regulates decorin expression in lung fibroblasts (54) and myocardial fibroblasts (55). TGF-␤ has been shown to stimulate proliferation of corneal fibroblasts through a mechanism that may involve the induction of basic fibroblast growth factor (56). In our model, both TGF-␤ treatment and culture in high serum decreased decorin production ( Figs. 2 and 3). This is consistent with studies showing a substantial induction of decorin expression upon quiescence in a variety of fibroblasts (52,57,58). The observations that decorin is an inducer of quiescence and that decorin is induced upon quiescence are suggestive of an autocrine mechanism of cell growth control possibly involving TGF-␤. It is conceivable that the substantial quantities of decorin in the normal corneal stroma limits TGF-␤ activity. In this manner, the proteolytic degradation of decorin likely to occur within a corneal wound might remove the restriction on TGF-␤ activity within this localized region. In addition to decorin, corneal fibroblasts in our system synthesized two large core proteins (M r Ϸ 375,000 and Ϸ 440,000) that reacted with a mAb directed against perlecan. The M r Ϸ 440,000 isoform was synthesized as either an HSPG or a hybrid possessing both CS and HS. The M r Ϸ 375,000 isoform was primarily glycanated with CS; however, forms bearing HS or potentially both CS and HS were detected. Perlecan, initially identified as an HSPG (59), has also been shown to be glycanated with CS or both CS and HS in a number of different tissues and cell types (26, 41, 60 -64). Perlecan isolated from the culture media in our system has shown heterogeneity not only in GAG chain substitution but also in the size of the core proteins. Size variants of perlecan bearing HS chains have been detected in the Engelbreth-Holm-Swarm tumor matrix (41). Several studies suggest that perlecan variants may be generated by alternative splicing in human (65)(66) and mouse (67). The significance of this heterogeneity of perlecan is not fully understood. However, when considering the potential importance of HS and its interactions with growth factors (68), it is conceivable that alternate glycanation could have a substantial impact on the biological activity of perlecan. We find that both TGF-␤1 and serum induce substantial increases in perlecan bearing both CS and HS chains at the early time point (Fig. 5). In contrast, after extended periods in culture the amount of perlecan bearing HS chains in the medium was unaffected by TGF-␤1, and decreased by serum. The levels of perlecan bearing CS chains were elevated with TGF-␤1 treatment and were decreased with serum (Fig. 6). These apparent differences with respect to culture duration and perlecan expression are indicative of an indirect response. The fact that TGF-␤ can induce secondary effectors, such as growth factors FIG. 7. Analysis of proteoglycans synthesized in response TGF-␤1. Corneal fibroblast were cultured without dFCS for 18 h and subsequently treated with 0, 1, 5, or 10 ng/ml TGF-␤1. Proteoglycans were radiolabeled with [ 35 S]SO 4 for 0-24, 24-48, or 48-72 h and medium fractions were collected after the radiolabeling periods. Aliquots of medium were treated Ϯ chondroitinase ABC and proteoglycans were isolated on Q-Sepharose columns. Chondroitin sulfate was defined as GAG susceptible to chondroitinase ABC. Heparan sulfate was defined as GAG resistant to chondroitinase ABC. Chondroitin sulfate synthesis (A) and HS synthesis (B) are expressed as CPM per 1 ϫ 10 6 cell Ϯ S.E. (n ϭ 3). Open columns, control; stippled columns, 1 ng/ml; hatched columns, 5 ng/ml; filled columns, 10 ng/ml. Aliquots of medium from TGF-␤1-treated corneal fibroblasts radiolabeled with [ 35 S]SO 4 were applied to 5% SDS-PAGE gels. Labeled proteoglycans and proteins from 1.5 ϫ 10 4 cells were visualized by autoradiography (C). Autoradiograph is representative of three gels. and matrix molecules, introduces further levels of complexity to the mechanism through which TGF-␤ may influence proteoglycan synthesis.
Corneal fibroblasts secreted an additional large core protein linked to CS (M r Ϸ 480,000) that did not react with a mAb directed against perlecan. It is unlikely that the M r Ϸ 480,000 protein core was a hybrid possessing both CS and HS chains, as the band intensities of the cores released with both chondroitinase ABC and heparinase were not significantly higher than those released with chondroitinase ABC alone. This core protein may be a perlecan variant lacking the epitope recognized by mAb A7L6 or may represent a novel CSPG. Although the identity of this proteoglycan is unclear, it may be important during wound healing as it responded to both TGF-␤1 and serum in culture.
Our results indicate that corneal fibroblasts in culture synthesize predominantly CS and HS with trace amounts of an unsulfated keratan. The major proteoglycan species secreted into the medium were decorin and perlecan. Our analysis indicates that TGF-␤1 and serum modulate the GAG chains and protein cores of both of these proteoglycans. In light of the influence that both decorin and perlecan exert on growth factor activity, changes in the expression of either of these proteoglycans could have important consequences on the cellular response to injury. A well characterized model system should allow the cooperation between proteoglycans and growth factors to be analyzed in detail. These studies could ultimately provide important insight into the mechanisms that control tissue remodeling after corneal stromal injury.