|Home | About | Journals | Submit | Contact Us | Français|
In pathological corneas, accumulation of fibrotic extracellular matrix is characterized by proteoglycans with altered glycosaminoglycans that contribute to the reduced transparency of scarred tissue. During wound healing, keratocytes in the corneal stroma transdifferentiate into fibroblasts and myofibroblasts. In this study, molecular markers were developed to identify keratocyte, fibroblast, and myofibroblast phenotypes in primary cultures of corneal stromal cells and the structure of glycosaminoglycans secreted by these cells was characterized. Quiescent primary keratocytes expressed abundant protein and mRNA for keratocan and aldehyde dehydrogenase class 3 and secreted proteoglycans containing macromolecular keratan sulfate. Expression of these marker compounds was reduced in fibroblasts and also in transforming growth factor-β-induced myofibroblasts, which expressed high levels of α-smooth muscle actin, biglycan, and the extra domain A (EDA or EIIIA) form of cellular fibronectin. Collagen types I and III mRNAs were elevated in both fibroblasts and in myofibroblasts. Expression of these molecular markers clearly distinguishes the phenotypic states of stromal cells in vitro. Glycosaminoglycans secreted by fibroblasts and myofibroblasts were qualitatively similar to and differed from those of keratocytes. Chondroitin/dermatan sulfate abundance, chain length, and sulfation were increased as keratocytes became fibroblasts and myofibroblasts. Fluorophore-assisted carbohydrate electrophoresis analysis demonstrated increased N-acetylgalactosamine sulfation at both 4- and 6-carbons. Hyaluronan, absent in keratocytes, was secreted by fibroblasts and myofibroblasts. Keratan sulfate biosynthesis, chain length, and sulfation were significantly reduced in both fibroblasts and myofibroblasts. The qualitatively similar expression of glycosaminoglycans shared by fibroblasts and myofibroblasts suggests a role for fibroblasts in deposition of non-transparent fibrotic tissue in pathological corneas.
The corneal stroma is a dense connective tissue with a highly organized extracellular matrix responsible for the remarkable strength and light transparency of the cornea. A notable feature of this matrix is its unique proteoglycan content consisting of proteins of the small leucine-rich proteoglycan (SLRP)1 family. Lumican, a SLRP protein abundant in the stroma, has been implicated in formation of the small and highly regular collagen fibrils required for corneal transparency (1). The glycosaminoglycans modifying SLRPs also appear to have a role in corneal transparency. Keratan sulfate in cornea is of higher polymer length and is at least an order of magnitude more abundant than the keratan sulfate found in other tissues (2). Conversely, corneal chondroitin/dermatan sulfate is low in abundance and sulfate content compared with the dermatan sulfate of skin and sclera (3). This unusual glycosaminoglycan composition has long been considered important in corneal transparency, a hypothesis consistent with several heritable disease conditions. Individuals with macular corneal dystrophy, for example, develop cloudy corneas as a result of an inability to produce keratan sulfate (4, 5). In Hurler’s and Scheie’s syndromes, a lack of glycosaminoglycan-degradative enzymes results in accumulation of highly sulfated dermatan sulfate in the cornea, causing corneal opacity at an early age (6, 7).
Corneal proteoglycans are also implicated in the pathology of corneal scarring. As a result of trauma or chronic corneal inflammation, the stroma develops fibrotic deposits that disrupt visual acuity. Such corneal scars are long-lasting and often constitute the cause for corneal transplantation. A number of early studies showed that corneal wound healing resulted in a reduction of keratan sulfate and in accumulation of highly sulfated chondroitin/dermatan sulfate in the scar (3, 8–14). More recent studies on scars developing during the chronic stress associated with keratoconus showed a glycosaminoglycan profile similar to that occurring in acute healing (15–19).
Corneal wound healing is associated with appearance in the stroma of cells with phenotypes clearly distinct from those of the normal tissue. In the normal cornea, keratocytes are flattened, quiescent, and neural crest-derived cells with a stellate morphology. Extensive cellular processes link adjacent cells via gap junctions (20). Filamentous actin is confined to the cortical region and is not organized into stress fibers (21). In response to wounding, keratocytes become motile and mitotic and develop actin cytoskeletal fibers associated with fibronectin in the extracellular matrix (21). These fibroblastic cells also secrete metalloproteinases that are thought to initiate tissue remodeling (22). In latter stages of healing, keratocyte-derived fibroblasts express α-smooth muscle actin incorporated into cytoplasmic stress fibers (23–27). These cells known as myofibroblasts exhibit reduced motility and cell division compared with the repair fibroblasts and may contribute to the contractile force involved in wound closure (28, 29). Myofibroblastic cells appear in response to transforming growth factor-β (TGF-β) and are associated with secretion of fibrotic extracellular matrix both in the cornea and in other tissues.
In vitro, primary keratocytes can be maintained in serum-free or low mitogen serum-containing culture media in a quiescent state exhibiting a cellular morphology and matrix secretion similar to keratocytes in vivo (21, 30). When stromal cells are subjected to serial passage in media containing fetal bovine serum, they lose the dendritic morphology typical of keratocytes, develop actin stress fibers, and begin secretion of metal-loproteinases (31, 32). In response to endogenous or exogenous TGF-β, stromal fibroblasts become myofibroblasts, expressing α-smooth muscle actin (33, 34).
Cultures of quiescent primary keratocytes secrete proteoglycans similar to those found in vivo, including all three of the proteoglycans bearing keratan sulfate: lumican, keratocan, and mimecan (30, 31, 35). It has long been observed that keratan sulfate secretion is greatly reduced or absent in serially passaged corneal fibroblasts (36), and we recently demonstrated that freshly isolated primary bovine keratocytes exhibit a loss of sulfated keratan sulfate-proteoglycans and an increase in sulfated chondroitin/dermatan sulfate-containing proteoglycans during transdifferentiation from keratocytes to myofibroblasts (35). This previous study (35) shows that myofibroblasts exhibit reduced expression of keratocan, a keratan sulfate-linked proteoglycan, and up-regulate biglycan, a dermatan sulfate proteoglycan. These proteins, however, represent minor components of the total cellular proteoglycan, and the overall expression of core proteins modified by keratan sulfate and chondroitin/dermatan sulfate was not greatly altered in myofibroblasts compared with keratocytes. Incorporation of labeled sulfate into proteoglycans, however, did exhibit marked differences between the two phenotypes with chondroitin/dermatan sulfate increased and keratan sulfate decreased. This observation lead to the hypothesis that a major feature of the alteration in corneal proteoglycan profile during the phenotypic transition in wound healing arises via modulation of the structure of the glycosaminoglycan chains modifying the core proteins.
The present study addresses this hypothesis by characterization of keratan sulfate and chondroitin/dermatan sulfate chains modifying proteoglycans secreted by stromal cells of different phenotypes. Primary non-passaged keratocyte cultures were characterized using molecular markers to identify the keratocyte, fibroblast, and myofibroblast phenotypes. Structural analyses of glycosaminoglycans secreted by the three cell types demonstrated a marked increase in chain length and sulfation of chondroitin/dermatan sulfate in both fibroblasts and myofibroblasts and a reduction in both sulfation and chain length of the keratan sulfate secreted by fibroblasts and myofibroblasts. These results establish the key link between cells observed in pathological corneas and specific alterations in biosynthesis of corneal glycosaminoglycans.
Primary keratocytes were obtained from fresh bovine corneal stromae by collagenase digestion as described previously (35). The cells were diluted in serum-free Dulbecco’s modified Eagle’s/F12 medium containing antibiotics and cultured on tissue culture-treated plastic at 4 × 104 cells/cm2 (keratocytes) or 1 × 104/cm2 (fibroblasts and myofibroblasts) in a humidified atmosphere containing 5% CO2. Culture medium was changed after 24 h (day 1) to Dulbecco’s modified Eagle’s/F12 medium with antibiotics (35) for keratocytes or the same containing 2% fetal bovine serum for fibroblasts and 2% fetal bovine serum with 2 ng/ml recombinant human TGF-β1 (Sigma-Aldrich) to induce myofibroblast formation. These media were changed at day 4, and cultures were harvested at day 5 or 6 as noted in the figure legends.
Cultures were labeled with 100 μCi/ml carrier-free [35S]sulfate (PerkinElmer Life Sciences) added on day 5 and the medium collected on day 6. In some experiments, 0.5 mM 4-nitrophenyl-β-D-xylopyranoside (N2132, Sigma-Aldrich) was added 1 h before labeling. Proteoglycans in the culture medium were purified by ion-exchange chromatography, dialyzed against water, and lyophilized. For chondroitin/dermatan sulfate quantitation, glycosaminoglycans from six identical 9.5-cm2 cultures were dissolved and combined to make triplicate samples of 100 μl in 0.1 M Tris acetate, pH 8. These samples were digested with 2 μl of chondroitinase ABC (C3667, Sigma-Aldrich) at 10 units/ml, for 2 h at 37 °C. Digested products were recovered by ultra-filtration with Microcon YM-3 microfiltration devices (Millipore). Keratan sulfate was digested in a similar manner using a mixture of 0.2 milliunits of Escherichia freundii endo-β-galactosidase and 0.2 milliunits of keratanase II (Seikagaku) in 0.05 M sodium acetate, pH 6.5, at 37 °C overnight. The amount of labeled fragments liberated by digestion was determined by scintillation counting, corrected for non-digested controls, and normalized to protein content of the cells as described below.
For size determination, chondroitin/dermatan sulfate proteoglycans were separated from total 35S-labeled proteoglycans by selective alcohol precipitation without enzymatic digestion of keratan sulfate (37). Protein was hydrolyzed with 20 μg/ml proteinase K twice for 30 min at 45 °C in 0.1 M Tris-HCl, pH 7.4, containing 0.1% SDS. Chondroitin/dermatan sulfate from xyloside-treated cultures were analyzed in the same manner. Keratan sulfate chains were obtained from total [35S]proteoglycans by treatment with chondroitinase ABC (as above), dialysis, lyophilization, and proteinase K digestion in a similar manner. The protein-free [35S]glycosaminoglycan chains were subjected to SDS-PAGE on 4–20% gels (chondroitin/dermatan sulfate) or 10–20% gels (keratan sulfate), electrotransferred in buffer without methanol to Genescreen Plus-charged nylon membranes (PerkinElmer Life Sciences), and subjected to autoradiography as described previously (35).
Non-labeled proteoglycans were purified from media of triplicate 75-cm2 cultures conditioned on days 4–6 as described above. They were digested with chondroitinase ABC in 0.1 M ammonium acetate, pH 7.5, or with combined keratanase II and endo-β-galactosidase in 0.1 M ammonium acetate, pH 6.5, followed by collection of the products by ultrafiltration as described above. Dried aliquots of the digestion products were fluorescently labeled with 5 μl of 0.1 M 2-aminoacridone in 3:17 acetic acid:dimethyl sulfoxide for 15 min followed by 5 μl of freshly dissolved cyanoborohydride 1 M at 37° overnight (38). Borohydride was quenched with 30 μl of 25% glycerol and 2 μl of 1 mg/ml bromphenol blue. The derivitized digestion products were separated on 8 × 10 × 0.05-cm gels of 28.5% acrylamide, 0.76% bisacrylamide containing 0.1 M Tris-HCl, pH 8. The running buffer was 0.089 M Tris borate, 2 mM EDTA, pH 8.4, chilled to 4 °C. This formulation duplicates the separation previously described for gels from Glyco Corp. (38) that are no longer available. Electrophoresis was carried out on ice at 5 watts of constant power/gel. Fluorescent bands were immediately photographed using a 12-bit Bio-Rad FluorS Max imaging system, and quantification was accomplished with Bio-Rad Quantity One software. FACE bands generated by chondroitinase were identified by co-electrophoresis with purified standards of fragments from chondroitin sulfate and hyaluronan (Sigma-Aldrich). Monosaccharide standards for keratan sulfate analysis were purchased from Sigma. Disaccharide standards were produced by digestion of purified bovine corneal keratan sulfate (Seikagaku) with endo-β-galactosidase or keratanase II, purified by size exclusion chromatography on a Superdex-Peptide column (Amersham Biosciences), and identified by FACE analysis as described previously (38). Molecular mass of the disaccharide keratan sulfate standards was confirmed by matrix-assisted laser desorption ionization time-of-flight mass spectroscopy (39).
Cellular morphology was observed after 2 days in cultures fixed in 100% methanol for 20 min and then stained with 1% crystal violet in 20% ethanol for 30 min followed by destaining in water. The cells were photographed by Bright field optics with a ×20 objective. For cytoskeletal analysis, cells after 2 days culture were fixed in a room temperature paraformaldehyde (35) and double-stained with Alexa-488 labeled phalloidin (Molecular Probes) and with anti-vinculin clone hVIN-1 (Sigma-Aldrich) followed by goat anti-mouse labeled with Alexa-546 (Molecular Probes) using procedures described previously (35). 6-day cultures were similarly fixed and stained for α-smooth muscle actin with anti-smooth muscle actin (clone asm-1, Sigma-Aldrich) followed by Alexa-546 goat anti-mouse antibody in a similar manner. Cytoskeletal photographs were acquired on a Bio-Rad laser-scanning confocal microscope using a ×60 oil objective.
Cells were collected by centrifugation after scraping into cold saline, and RNA was isolated using RNeasy mini kit (Qiagen). RNA was treated with DNase I (Ambion) according to supplier’s protocol and then concentrated by alcohol precipitation in the presence of GlycoBlue (Ambion). RNA was quantified by fluorimetry using RiboGreen (Molecular Probes).
RNA (400 ng) was transcribed to cDNA in a 100-μl reaction containing 1× PCR II buffer (Roche Applied Science), 5 mM MgCl2, 800 μM dNTP mixture (Roche Applied Science), 2.5 μM random hexamers (Invitrogen), 0.4 units of RNase inhibitor, and 125 units of SuperScript II reverse transcriptase (Invitrogen). PCR was carried out for 40 cycles of 15 min at 95° and 60 min at 60° after an initial incubation at 95° for 10 min in an ABI7700 thermocycler. Reaction volume was 50 μl containing 1× TaqMan Buffer A (Applied Biosystems), 5 mM MgCl2, 300 μM each dNTP, 0.025 units/ml AmpliTaq Gold polymerase, and 5 μl of cDNA. Forward and reverse primers and fluorescent internal hybridization probes for each gene, as shown in Table I, were used at optimized concentrations. Sequences for these genes were obtained from Gen-Bank™ with the exception of that of the extra domain A form of bovine fibronectin. This information was obtained by direct sequencing of reverse transcriptase-PCR amplification products obtained from myofibroblast cDNA using primers based on published flanking sequence data. The bovine extra domain A sequence thus obtained was deposited in GenBank™ with accession number AY221633. Amplification efficiency for each of the primer pairs shown was determined to be >90%.
For each gene/cDNA combination, amplifications without reverse transcriptase were carried out as negative controls. Amplification of 18 S ribosomal RNA was carried out for each cDNA (in triplicate) for normalization of RNA content. Threshold cycle number (Ct) of amplification in each sample was determined by ABI software. Relative mRNA abundance was calculated as the Ct for amplification of a gene-specific cDNA minus the average Ct for 18 S expressed as a power of 2, i.e. 2ΔCt. Three individual gene-specific values thus calculated were averaged to obtain mean ± S.E.
Proteoglycans from culture media collected at days 4 –6 were digested with chondroitinase ABC or keratanase II and endo-β-galactosidase as described above. Samples from the digests normalized for cell number (by total cell DNA content) were separated on 10% SDS-PAGE gels and transferred to polyvinylidene difluoride membranes subjected to immunoblotting as previously described. Keratan sulfate, biglycan, and keratocan were detected in proteoglycans pooled from 3–4 individual cultures. Biglycan was examined in chondroitinase digests and keratocan after digestion of keratan sulfate. Cell layers were lysed directly in 1% SDS sample buffer. Protein was determined using the Micro BCA assay (Pierce) and DNA by fluorimetry with PicoGreen (Molecular Probes). Equal amounts of protein were separated by SDS-PAGE, 10% gels for ALDH and α-smooth muscle actin and 4 –20% for keratan sulfate and fibronectin. Proteins were either stained with Coomassie Blue (40) or alternately electrotransferred to polyvinylidene difluoride membranes and subjected to immunodetection (35) with antibodies to cellular fibronectin (clone IST-9, Accurate Chemical), ALDH (41), antibody J36 against keratan sulfate (37), a peptide antibody to biglycan (42), or monoclonal antibody to α-smooth muscle actin (Clone 1A4, Sigma-Aldrich). Keratocan was detected using an antibody generated against a mixture of 10 synthetic peptides, each encoding a unique amino acid sequence of bovine keratocan linked to keyhole limpet hemocyanin carrier. This antiserum was prepared and affinity-purified as described for anti-lumican peptide antibodies (43).
Primary bovine keratocytes isolated from fresh stroma by collagenase digestion and cultured in absence of serum exhibited a dendritic (stellate) morphology (Fig. 1A) with multiple extended processes interconnecting individual cells. Phalloidin staining (Fig. 1D) revealed filamentous actin in the cortical region and associated with the cell-cell contacts at the intersection of the cell processes. Vinculin staining was weak, diffuse, and mostly perinuclear in localization. When cells prepared in a similar manner were exposed to 2% fetal bovine serum for 2 days, the cells became larger and flattened with a reduction in processes (Fig. 1B). Many cells were polarized with pseudopodial extensions (arrows) indicating motility. In these cells, filamentous actin formed stress fibers traversing the cell body (Fig. 1E). Vinculin was focally localized at the terminus of the actin fibers as is typical for matrix-adherent fibroblasts. Keratocyte cultures exposed to both fetal bovine serum and TGF-β1 contained larger, less refractile cells with a polygonal appearance. Fewer obviously motile cells were observed (Fig. 1C). Filamentous actin fibers were thicker and fewer in number than in fibroblastic cells (Fig. 1F). Vinculin accumulation in focal adhesion was denser and larger than in fibroblasts. After 5 days of culture, numerous cells were observed in which actin fibers were stained with antibodies to α-smooth muscle actin (Fig. 1G). Cells in serum-free medium (keratocyte phenotype) or grown in the fibroblastic phenotype did not exhibit α-smooth muscle actin staining (data not shown).
Primary stromal cells in conditions similar to those in Fig. 1 exhibited differential expression of a number of marker molecules. α-Smooth muscle actin, cellular fibronectin, and biglycan are associated with myofibroblasts in vitro and in vivo. Immunoblotting showed a marked abundance of these three proteins in TGF-β-induced myofibroblasts compared with keratocyte and fibroblast cultures (Fig. 2, A–C). Accumulation of ALDH was recently reported to be a distinguishing feature of keratocytes in vivo (44). This protein described as a corneal crystallin represents one of the major soluble proteins in keratocytes but is reduced in fibroblasts populating healing wounds. ALDH was detected in all of the cultured bovine stromal cells, but its concentration was markedly elevated in cells maintained in the keratocyte morphology (Fig. 2D). The immunostained ALDH band corresponded to a major protein of ~54 kDa visualized by Coomassie Blue staining, prominent in keratocyte cell lysates but not apparent in lysates from fibroblasts and myofibroblasts (Fig. 2E).
Keratan sulfate glycosaminoglycan chains and keratocan, a SLRP core of corneal keratan sulfate proteoglycan, are extracellular products highly enriched in the corneal stroma. Immunoblotting using monoclonal antibody J36 to keratan sulfate revealed heterogeneous high molecular weight keratan sulfate in proteoglycans isolated from keratocyte culture medium (Fig. 2F). In fibroblasts, J36 epitopes were reduced in molecular size to a band of 50 –60 kDa. In myofibroblasts, the J36 keratan sulfate epitope was not detected. Keratan sulfate-linked proteins secreted by keratocytes also contained abundant keratocan in the proteoglycans isolated from quiescent cultures of keratocytes (Fig. 2G). Keratocan was decreased in fibroblasts and almost undetected in myofibroblast cultures.
Real-time quantitative reverse transcriptase-PCR analysis assays were designed to detect mRNA for the five proteins identified in Fig. 2. Relative abundance of the transcript pools for these five proteins (Table II) showed that the protein expression levels detected by Western blotting was consistent with differences in mRNA pools for these proteins. Pools for α-smooth muscle actin, biglycan, and cellular fibronectin were increased 12–39-fold in myofibroblasts compared with keratocytes. Fibroblasts, however, had little increase in these mRNAs compared with keratocytes. Keratocan transcripts were decreased 15-fold in fibroblasts and 50-fold in myofibroblasts as compared with keratocytes. Similarly, ALDH transcript abundance was 700- and 2000-fold lower in fibroblasts and myofibroblasts, respectively, compared with keratocytes. These assays link gene expression associated with in vivo cell phenotypes with the cell culture model.
Collagen type I represents the major fibrillar collagen of the cornea, but synthetic levels of collagen I are low in adult non-wounded corneas (45). Collagen III is a fibrillar cornea present in fetal and wounded cornea but only a very minor component of adult corneal stroma (45). We previously found that mRNA and protein for collagen I and III were up-regulated in myofibroblasts compared with keratocytes (35). Real-time PCR analysis of the mRNA pools for these collagens (Table II) confirmed these increases in myofibroblasts. These assays also showed that, unlike other myofibroblastic markers, mRNA pools for collagens are up-regulated in fibroblasts as well as myofibroblasts.
Proteoglycans were metabolically labeled for 18 h with [35S]sulfate and isolated from culture media by ion-exchange chromatography. In initial experiments, greater than 95% of sulfated glycosaminoglycan isolated from the media of the cultures was determined to be keratan sulfate and chondroitin/dermatan sulfate (data not shown). Thus heparan sulfate does not constitute a significant fraction of the soluble glycosaminoglycan secreted by these cultures. Keratan sulfate in the labeled proteoglycans, determined by sensitivity to endo-β-galactosidase and keratanase II, was reduced by ~40% in fibroblasts and ~ 60% in myofibroblasts compared with keratocytes (Fig. 3A). Conversely, 35S-labeled chondroitin/dermatan sulfate measured by sensitivity to chondroitinase ABC was increased 3–3.5-fold in fibroblasts and myofibroblasts compared with keratocytes. In the presence of nitrophenyl-β-D-xyloside, a synthetic initiator of chondroitin polymerization, chondroitin/dermatan sulfate biosynthesis was increased >5-fold in all of the cultures as compared with cultures without this initiator (data not shown), suggesting that (as with many cell types) chain initiation represents a rate-limiting step in chondroitin and dermatan sulfate synthesis. In the presence of β-xyloside, fibroblasts continued to incorporate an approximate 3-fold more sulfate than keratocytes (Fig. 3C) but myofibroblasts increased the relative biosynthesis to almost 6-fold than that of keratocytes.
The size of the 35S-labeled glycosaminoglycan chains was determined by polyacrylamide gel electrophoresis after proteolytic removal of the core proteins. Keratan sulfate produced by fibroblasts and myofibroblasts decreased compared with that of keratocytes, whereas chondroitin/dermatan sulfate chain length increased (Fig. 4, A and B). Chondroitin/dermatan sulfate made in the presence of β-xyloside was smaller than that without this initiator but did not increase in fibroblasts and myofibroblasts (Fig. 4C). These results suggest a relationship between rate of chain initiation and final chain length in chondroitin/dermatan sulfate.
Analysis of non-labeled chondroitin/dermatan sulfate secreted by the keratocyte cultures was carried out by FACE analysis after chondroitinase digestion. As shown in Fig. 5A, keratocyte cultures contained sulfated and non-sulfated disaccharides in approximately a 3:2 ratio. Sulfation was primarily on the 4 position of the N-acetylgalactosamine. In fibroblasts, the non-sulfated component was significantly lower and both 4-O- and 6-O-sulfation increased. In myofibroblasts, 4-O-sulfation represented the majority of the moieties and unsulfated chondroitin disaccharide was reduced to <5% of the total. Quantitation of chondroitin disaccharides is depicted in Fig. 5B. Hyaluronan was also detected in this analysis, and quantitation of hyaluronan secreted by the different cultures is shown in Fig. 5C. As shown, hyaluronan was not detected in keratocyte culture media but hyaluronan represented 1.5 and 4.5% of the chondroitinase-sensitive glycosaminoglycan in fibroblast and myofibroblast cultures.
A large number of fragments is generated by enzymatic depolymerization of keratan sulfate (46 –48). Characterization of these fragments has employed a variety of analytical approaches including FACE, a technique that can be used to quantitate major components of corneal keratan sulfate (38, 49). Digestion of keratan sulfate from keratocyte culture media with mixed keratanase II and endo-β-galactosidase generated eleven major bands visualized on FACE (Fig. 6A). Of these, monosaccharides and disaccharides involved in keratan sulfate chain extension constituted ~60% of fragments secreted by keratocyte cultures (Fig. 6B, black bars). The abundance of this set of fragments dropped ~5-fold in the media from fibroblast and myofibroblast cultures. The abundance of these chain extension fragments as a proportion of the total fragments was also reduced in the fibroblasts and myofibroblasts. Based on previous studies of keratan sulfate structure, it seems likely that most of the unidentified bands (Fig. 6B, gray bars) released by enzyme digestion represent moieties capping the non-reducing terminus of keratan sulfate. A variety of such capping structures has been documented in corneal keratan sulfate by NMR, and these components also are present in FACE analysis of keratan sulfate (38). These components showed no significant decrease in fibroblasts and myofibroblasts compared with keratocytes (Fig. 6B). Reduction of keratan sulfate chain length would reduce the ratio of chain extension moieties to capping fragments. Thus, the altered ratio of chain extension moieties to total degradation products in fibroblasts and myofibroblasts shown in Fig. 6B is consistent with a reduced keratan sulfate chain length as seen in Fig. 4.
For more than half a century, the unique glycosaminoglycan composition of the cornea has been thought to be important to corneal transparency. Studies of pathological corneas, hereditary diseases, and knock-out mouse mutations have helped confirm this hypothesis. During the last decade, studies have identified distinct phenotypes of stromal cells present in healing wounds (50). In the current study, we set out to manipulate primary cultures of stromal cells to reproduce these phenotypic characteristics observed in vivo and to characterize their glycosaminoglycan biosynthesis. Although there are numerous previous studies of glycosaminoglycan biosynthesis in cultured corneal cells, an important aspect of this study is the use of primary cells without subculture and the linking of cultured cells to in vivo phenotypes using molecular markers. Previous studies have not employed such stringent criteria, thus comparisons of extracellular matrix biosynthesis in our model system are likely to reflect the pathological process more accurately than earlier studies.
The phenotype of the cultured cells was clearly distinguishable by the molecular markers they expressed. The ALDH family of proteins is highly expressed in corneal epithelium and stroma and may serve a non-enzymatic function (44, 51). ALDH is down-regulated during wound healing, making it a marker for the quiescent keratocyte in vivo (44, 52). In our study, both ALDH protein and mRNA were dramatically down-regulated as quiescent keratocytes were activated by serum to become fibroblastic. Keratocan, a SLRP protein highly expressed in the corneal stroma, served as a second marker of the keratocyte phenotype. Both protein and mRNA pools for this protein were reduced in fibroblasts and myofibroblasts, suggesting regulation of expression at the nucleic acid level. A third marker of importance is the use of a monoclonal antibody to keratan sulfate. Although many such antibodies have been described, none yet has proved useful for detection of corneal keratan sulfate made in vitro. The finding that antibody J36 can serve such a function provides an important tool for non-disruptive screening of cultured keratocytes. It should be noted that expression of the J36 epitope does not correlate with the total abundance of keratan sulfate chains as determined in Figs. 3, ,4,4, and and6.6. As with previously described monoclonal antibodies (53), J36 probably recognizes a series of sulfated disaccharides in the some keratan sulfate chains. In the shorter, less highly sulfated chains these structures may be absent. Thus, the J36 antibody is valuable as a qualitative but not quantitative assessment of keratan sulfate expression.
Fibroblasts were readily distinguished from keratocytes by the development of actin cytoskeleton, focal adhesions, and the loss of keratocyte gene marker expression. Myofibroblasts share these characteristics with fibroblasts but in addition express protein and mRNA for α-smooth muscle actin. Such expression serves as a de facto definition of myofibroblasts. The alternately spliced form of cellular fibronectin that contains the type III extra domain A (EDA or EIIIA) is associated with healing wounds and fibrosis in cornea and other tissues (54 –56). Expression of this matrix molecule serves as a marker of fibrotic extracellular matrix that is closely linked to intracellular α-smooth muscle actin expression in granulation tissue myofibroblasts (57). Biglycan, a SLRP protein that is modified with chondroitin/dermatan sulfate, similarly is associated with tissue fibrosis and corneal scars and was previously identified as a product of corneal myofibroblasts (19, 35, 58, 59). The combination of α-smooth muscle actin, biglycan, and cellular fibronectin provides a powerful set of tools for distinguishing myofibroblasts from fibroblasts.
The availability of these three well characterized phenotypes of primary cells from corneal stroma allows us to pose important questions regarding extracellular matrix synthesis by these cells. A long time observation regarding healing corneal wound and corneal scar tissue is the reduction or disappearance of stromal proteoglycans containing keratan sulfate. This change may be key to corneal transparency in view of recent studies linking the loss of a keratan sulfate-specific sulfotransferase to macular corneal dystrophy (60). Our previous work has demonstrated that the corneal SLRP proteins to which keratan sulfate is attached continue to be expressed by keratocytes both in vivo and in vitro. Despite pronounced changes in keratan sulfate, total keratan sulfate-linked protein does not vary much as keratocytes become myofibroblasts (35), suggesting that the observed changes occur in the keratan sulfate chains themselves. Earlier studies typically expressed keratan sulfate biosynthesis as a proportion of the total glycosaminoglycan biosynthesis. Our current data document that keratan sulfate and chondroitin/dermatan biosynthetic rates are independent and altered in opposite directions. These results are consistent with the data showing these glycosaminoglycans to be synthesized by different glycosyltransferases and sulfotransferases and imply that activity of the enzymes is regulated independently.
Metabolic labeling with sulfate and Western blotting with anti-keratan sulfate antibodies suggested that keratan sulfate chains produced by fibroblasts and myofibroblasts are shorter and contained less sulfate than the keratan sulfate made by keratocytes. FACE analysis supported these conclusions. Fig. 6 shows a reduction in the ratio of sulfated disaccharides involved in chain elongation and components associated with non-reducing terminus of the chains. This ratio is consistent with shorter keratan sulfate chains observed directly by electrophoresis in Fig. 4. Keratan sulfate-linked SLRP proteins are not greatly reduced in myofibroblasts nor are the compounds in the FACE gels in Fig. 6, representing non-reducing termini of these chains. The conclusion from these observations is that alteration of keratan sulfate in fibroblasts and myofibroblasts (and by implication, in corneal scars) is due almost entirely to a shortening of the keratan sulfate length and not a reduction in the number of chains. Corneal keratan sulfate biosynthesis exhibits a strong link between glucosamine sulfation and chain elongation (38, 60). Chick stromal cells in culture exhibit a loss in chain elongation associated with decreased sulfotransferase activity (61). Our results are consistent with a similar alteration in bovine keratocytes as they become fibroblasts.
Increases in chondroitin/dermatan sulfate have been reported in corneal scar tissue, a change that appears to be stable for extended periods of time beyond active wound healing (10 –12, 19). Here we observed increases in chondroitin/dermatan sulfation and chain length in both fibroblasts and myofibroblasts. The differential in sulfate incorporation was maintained in the presence of saturating levels of β-D-xyloside, suggesting that differences among the cell phenotypes result from an altered biosynthetic capacity in the fibroblastic and myofibroblastic cells rather than an increase in the availability of core protein initiation sites. The fact that differences in chondroitin/dermatan sulfate molecular size were eliminated in the presence of xyloside suggests that the chain length may be a function of both chain elongation capacity and the abundance of initiation sites.
Relative sulfation of the chondroitin/dermatan chains increased in addition to the chain length. The ratio between 4-O-and 6-O- sulfation was not altered, and there was no detection of disulfated disaccharides in the chondroitin/dermatan sulfate from fibroblastic and myofibroblastic cells. Simultaneous sulfation of 4-O- and 6-O- moieties in chondroitin/dermatan sulfate is unusual (62). Our current data do not distinguish whether the 4-O- and 6-O-sulfation is in same molecule or of a mixture of chains modified only on one site. The relative sulfation was increased in fibroblasts and myofibroblasts in both untreated and xyloside-treated cultures (data not shown). Thus, unlike keratan sulfate, chain extension and sulfation in chondroitin/dermatan may be regulated independently.
Increased amount and sulfation of chondroitin/dermatan sulfate in corneal scars have been reported in several studies, but the finding of increased molecular size is novel. The presence of larger chondroitin/dermatan sulfate molecules in scar tissue is consistent with the appearance of exceptionally large chondroitinase-sensitive cuprolinic blue-stained filaments in the interfibrillar space fibrotic regions of pathological corneas (63). Because chondroitin/dermatan proteoglycans bind water more tightly than keratan sulfate, an accumulation of these large, more highly sulfated molecules could disrupt critical stromal collagen spacing because of their hydrodynamic volume.
Hyaluronan has been characterized in healing corneas, but the source has not been identified (3, 10, 14, 64). This study suggests that keratocytes activated into the fibroblastic or myofibroblastic phenotypes could be a source of the wound-healing hyaluronan. The identification of diverse biological effects of hyaluronan, including stimulation of cell motility, lends a potential importance of this observation to cellular events in healing wounds.
Overall both fibroblasts and myofibroblasts exhibited a qualitatively similar alteration in glycosaminoglycan biosynthesis compared with keratocytes. Keratan sulfate was reduced in amount, chain length, and sulfation, whereas chondroitin/dermatan sulfate was increased in abundance, chain length, and sulfation. The differences between fibroblasts and myofibroblasts were quantitative rather than qualitative. This pattern was similar to that observed with collagen mRNA pool. This observation is significant in terms of the concept of the myofibroblast as a fibrogenic phenotype. Transforming growth factor-β and the myofibroblastic cells that appear in response to this cytokine are generally recognized to be associated with connective tissue deposition, scar tissue formation, and fibrosis (65, 66). Conversely, fibroblasts have been associated with metalloproteinase secretion and tissue remodeling (22, 50). Our results suggest that neither myofibroblasts nor TGF-β is required for stromal cells to secrete glycosaminoglycans and collagens similar to those of scar tissue. Thus, myofibroblastic cells may not be the sole source of all of the molecular components of scar tissue.
We appreciate the advice and collaboration of both Dr. Anna Plaas in development of the FACE analysis experiments and Dr. R. Lindahl for the gift of antibodies to ALDH.
*This work was supported by National Institutes of Health Grants EY09368 (to J. L. F.), EY003263 (to N. S.), and 30-EY08098 (University of Pittsburgh, Department of Ophthalmology Core Grant), Research to Prevent Blindness, and Eye and Ear Foundation of Pittsburgh.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank™/EBI Data Bank with accession number(s) AY221633.
1The abbreviations used are: SLRP, small leucine-rich proteoglycan; Gal, galactose; Gn, N-acetylglucosamine; TGF-β, transforming growth factor-β; FACE, fluorophore-assisted carbohydrate electrophoresis; ALDH, aldehyde-3-dehydrogenase; β-D-xyloside, 4-nitrophenyl-β-D-xylopyranoside.