The Effect of Growth Factor Signaling on Keratocytes In Vitro and Its Relationship to the Phases of Stromal Wound Repair

Abstract

purpose. To determine the relationship between signaling by different growth factors and the phases of corneal stromal wound repair. The authors hypothesize that the process involves sequential signaling, resulting first in proliferation and then in extracellular matrix (ECM) synthesis.

methods. The effects of IGF-I, TGF-β1, FGF-2, and PDGF on proliferation and ECM production by primary cultured bovine keratocytes were evaluated. DNA synthesis was determined by ³H-thymidine incorporation, and maximal cell density was determined by measurement of DNA content. Relative levels of ECM components synthesized by keratocytes and secreted into the media were evaluated by ³H-glycine incorporation into total ECM protein and collagen, by ³H-glucosamine incorporation into chondroitin sulfate, keratan sulfate, and hyaluronan, and by Western blotting with antibodies specific to procollagen types Ι and ΙΙΙ.

results. FGF-2 stimulated the highest level of proliferation and the lowest level of glycosaminoglycan synthesis and inhibited the synthesis of collagen types Ι and ΙΙΙ. IGF-I, in contrast, stimulated the lowest level of proliferation and the highest levels of collagen synthesis. PDGF and TGF-β1 had intermediate effects on proliferation and collagen synthesis. Although FGF-2 inhibited collagen production, it could be restored by subsequent treatment with IGF-I, TGF-β1, and PDGF.

conclusions. The results of this study showed that the level of proliferation induced by the growth factors was inversely related to the levels of collagen production. The authors suggest that FGF-2 initiates the hypercellular phase of corneal wound healing and that IGF-I and PDGF are involved in the restoration of a normal ECM.

The mature corneal stroma consists of a distinctively organized and transparent extracellular matrix (ECM) composed primarily of fibrillar collagen types Ι and V and three small leucine-rich proteoglycans, one with chondroitin sulfate chains (decorin) and two with keratan sulfate chains (lumican and keratocan). Collagen is present in the corneal stroma as fibrils that are separated by small spaces containing proteoglycans and are arranged in plywoodlike layers. Collagens play a fundamental role in the structure and function of the cornea. The organization of fibrils and lamellae and the cross-linking of the collagen provide the cornea with its tensile strength, whereas the uniform diameter and spacing of the collagen fibrils provide the cornea with its ability to transmit light onto the retina (for reviews see Refs. ^{1 2 3} ). Collagen type V initiates fibril formation and acts in conjunction with lumican and keratocan and decorin to regulate fibril diameter growth. ^{4 5 6 7 8 9} Type ΙΙΙ collagen also has been reported to be normally present in bovine and rabbit corneas. ^{10 11} More recent studies in mice suggest collagen type ΙΙΙ may be developmentally regulated and expressed primarily in neonatal corneas. ¹²  

Embedded in this unique ECM are keratocytes, neural crest-derived cells that produce the proteoglycans and collagens. ¹³ Keratocytes are characterized by a dendritic morphology and form a three-dimensional network in contact with one another by their extensive processes. ^{13 14} Keratocytes proliferate and are biosynthetically active during corneal development. In adult corneas, they are considered quiescent because they are locked in phase G0 of the cell cycle ¹⁵ and have a low level of proliferation and matrix synthesis. ^{1 13 16} Keratocytes also are characterized by high levels of the crystallins aldehyde dehydrogenase and transketolase—proteins thought to provide cellular transparency—in their cytoplasm. ¹⁷  

The ability of the cornea to remain transparent can be compromised when the cornea encounters trauma, infection, ulceration, or chronic inflammation. On injury, damaged keratocytes and keratocytes flanking the wound undergo apoptosis. ^{16 18} Some of the remaining keratocytes are stimulated to lose their quiescence and transition into divergent or “activated” phenotypes that proliferate, resulting in regions of hypercellularity that then go on to produce the new matrix needed to repair the wound. ^{16 19 20 21} These divergent phenotypes can aid in the regeneration of a normal stromal matrix or can become scars by remaining hypercellular or producing a disorganized matrix. ^{22 23} It has been proposed that an altered ratio of collagen types and proteoglycans in the ECM or the loss of the “crystallins” from the keratocytes causes the opacity. ^{17 24 25 26}  

Peptide growth factors—extracellular signal proteins (ligands) that promote the growth, proliferation, and differentiation of cells in animal tissues by interacting with cell-surface receptors—are partially credited for the regulation of collagen and proteoglycans expression during corneal wound healing. ²⁷ Growth factors are able to reach the wound site through tear film, aqueous fluid, and the epithelial layer of the cornea. ^{27 28 29} IGF-II has been shown to be present in the aqueous humor and in the cornea. ^{30 31} Both IGF-I and IGF-II, which act on the same receptors, have been shown to cause keratocytes to proliferate and maintain their dendritic morphology in vitro. ^{29 31} Transforming growth factor-beta (TGF-β), fibroblast growth factor-2 (FGF-2), and platelet-derived growth factor (PDGF) have all been found in tear fluid. ^{32 33} These growth factors can activate keratocytes to proliferate, differentiate into fibroblasts or myofibroblasts, and migrate, similar to the in vivo keratocyte response to injury. ^{19 29 34 35}  

Although a number of studies have reported the effect of growth factors on keratocyte morphology, crystallin content, α-smooth muscle actin expression, and proteoglycan synthesis, ^{29 31 34 36 37 38 39 40} comparatively little attention has been given to their effect on collagen synthesis, ^{41 42} which is the single most important structural element of the cornea. Consequently, in this study, we analyzed the effects of different growth factors (IGF-I, TGF-β1, FGF-2, and PDGF) on keratocyte proliferation and matrix synthesis to test our hypothesis that stromal wound healing involves sequential signaling, leading to keratocyte proliferation followed by ECM synthesis and deposition. 

Materials and Methods

Chemicals, growth factors, papain, and hyaluronate lyase were purchased from Sigma-Aldrich (St. Louis, MO) unless otherwise stated. Isotopes were purchased from Perkin Elmer (Boston, MA). Cell proliferation assay kits (CyQuant; Invitrogen, Carlsbad, CA), gels, reagents, and equipment used for protein separation and transfer were purchased from Invitrogen (Carlsbad, CA). Endo-β-galactosidase and chondroitinase ABC were purchased from Associates of Cape Cod (East Falmouth, MA). Collagenase for measuring ³H-glycine incorporation into collagen was purchased from Advance Biofactures Corp. (Collagenase Form III; Lynbrook, NY). 

Keratocyte Isolation and Culture

Bovine keratocytes were collagenase isolated, as previously described. ^{37 43} In brief, freshly harvested corneas were obtained from 12-month-old calves and sequentially digested. Initially, the tissue was subjected to a 45-minute collagenase digestion to remove the endothelium and epithelium, and then keratocytes were released from the stroma by a second 150-minute digestion. Keratocytes were plated in serum-free Dulbecco’s modified Eagle’s medium (DMEM)/F12 containing antibiotics and 1 mM 2-phospho-ascorbic acid at 20,000 cells/cm² to allow attachment (day 0) and were incubated at 37°C with 5% CO₂. Medium was changed the next day (day 1) and every third day thereafter with DMEM/F12 or DMEM/F12 supplemented with 10 ng/mL IGF-I, 2 ng/mL TGF-β1, 10 ng/mL FGF-2, or 10 ng/mL PDGF-BB. Cultures were harvested on days 1, 7, 10, and 13. 

DNA Content

DNA was extracted from the cell layers of each well of harvested cultures with the use of a cell proliferation assay buffer (CyQuant; Invitrogen). DNA in the extract was measured with a cell proliferation assay kit (CyQuant; Invitrogen) according to the manufacturer’s instructions and was expressed as nanogram of DNA per well. Calf thymus DNA was used as a standard. 

Radiolabeling of Keratocytes

Keratocytes were incubated with 10 μCi/mL ³H-thymidine for DNA synthesis or 25 μCi/mL ³H-glycine for protein/collagen synthesis, beginning on days 4, 7, and 10 and were harvested 3 days later. Keratocytes were also incubated with 25 μCi/mL ³H-glucosamine for glycosaminoglycan synthesis beginning on day 7 and were harvested on day 10. 

DNA Synthesis.

Incorporation of ³H-thymidine into DNA was determined as previously described. ³⁷ In summary, aliquots of cell proliferation assay buffer extract (CyQuant; Invitrogen) of the cell layer were adjusted to contain 10% trichloroacetic acid (TCA). Carrier protein bovine serum albumin (BSA) was added, and the mixture was held at 4°C overnight to precipitate DNA. The precipitate was collected onto glass filters, and the unincorporated isotope was removed by washing with ice-cold 5% TCA. Incorporation was measured by liquid scintillation counting and was expressed as counts per minute per nanogram of DNA. 

Protein/Collagen Synthesis.

Incorporation of ³H-glycine into collagen and total protein secreted into the media was measured as previously described. ⁴⁴ Briefly, unincorporated radioactivity was removed by chromatography on desalting columns (PD-10; Sigma-Aldrich) in 4 M guanidine-HCl. Guanidine was removed by dialysis against water. Aliquots of 400 μL were incubated with 1 mg papain in papain digestion solution (PBS containing 10 mM cysteine and 10 mM EDTA), then digested for 3 hours at 55°C and precipitated with 10% TCA, and the supernatant was counted to determine the incorporation of ³H-glycine into total protein. Additionally, 400-μL aliquots were digested with collagenase at 37°C for 3 hours in collagenase digestion solution (3 mM N-ethyl maleimide, 50 mM Tris, 150 mM NaCl, and 5 mM CaCl₂) and were processed as described for papain digests to determine incorporation into collagen. The incorporated radiolabeled radioactivity in collagen and in total protein was expressed as counts per minute per nanogram of DNA. 

Glycosaminoglycan Synthesis.

Incorporated ³H-glucosamine secreted in the medium was isolated by chromatography on desalting columns (PD-10; Sigma-Aldrich) in 4 M guanidine and was dialyzed against distilled water overnight. Aliquots were digested with H₂O (blank), chondroitinase ABC (0.05 U/μL), endo-β-galactosidase (0.5 mU/ μL), and hyaluronate lyase (ampule dissolved in 250 μL H₂O; used 5-μL/sample) for 3 hours at 37°C. Undigested GAGs were precipitated out by the addition of 2% dextran sulfate and 1% KAc in 95% ETOH at 4°C. The precipitate was removed by centrifugation, and the radioactivity of the supernatant was measured by liquid scintillation spectrophotometry. Chondroitinase ABC digests chondroitin sulfate and hyaluronan; therefore, the amount of hyaluronan was subtracted from the amount of chondroitin sulfate to determine chondroitin sulfate content. The incorporated radioactivity in chondroitin sulfate, keratan sulfate, and hyaluronan synthesis was expressed per microgram of DNA. 

SDS-PAGE/Coomassie Stain

Medium was collected from each condition and concentrated 10-fold using 10,000 MWt cutoff spin concentrators (Amicon Ultra Millipore Corp., Milford, MA). Samples were loaded based on DNA content and were electrophoresed under reducing conditions on 3% to 8% Tris-acetate acrylamide gels. Gels were stained with Coomassie blue (SimplyBlue SafeStain; Invitrogen) according to the manufacturer’s instructions. 

Pepsin Digestion

Eight milliliters of medium pooled from four wells was adjusted to 0.5 M acetic acid by the addition of 230 μL glacial acetic acid. Two milliliters of 0.5 M acetic acid was added to each of the four wells. A pepsin stock solution (4 mg/mL of 0.5 N acetic acid) was prepared. After the samples had chilled to 4°C, 100 μL stock was added to the medium, and 25 μL stock was added to each well. Samples were digested overnight at 4°C. Cell layer samples were combined, and 50 μL pepsin stock was added to the cell layer samples and digested for an extra 5 hours at 4°C. Samples were adjusted to neutrality with NaOH to inactivate pepsin. Samples were dialyzed against water, lyophilized, and reconstituted in 1× LDS sample buffer. Samples were loaded based on DNA content, determined on parallel cultures, to 3% to 8% Tris-acetate gels and were electrophoresed as described. 

Western Blotting

Proteoglycan levels, fibronectin levels, and procollagen levels in the media were determined by Western blot, as described previously. ³⁷ Medium was digested with endo-β-galactosidase to detect keratocan and lumican core proteins or with chondroitinase ABC to detect decorin and biglycan core proteins. All samples were loaded based on DNA content and electrophoresed as described. Proteins were transferred to a nitrocellulose membrane and were blocked with solution (I-Block; Tropix, Bedford, MA). The following primary antibodies were added at 4°C overnight: keratocan, ⁴³ lumican, ⁴³ decorin (DS1; Hybridoma Bank, University of Iowa), biglycan (gift from Larry Fisher), extra domain A (EDA) fibronectin (Accurate Chemical & Scientific Corp., Westbury, NY), fibronectin (Chemicon International, Temecula, CA), procollagen type Ι (SP1.D8; Hybridoma Bank, University of Iowa), and procollagen type ΙΙΙ (US Biological, Swampscott, MA). The corresponding secondary antibody was added, tagged with horseradish peroxidase, and detected with an ECL detection kit in accordance with the manufacturer’s instructions. 

Statistical Analysis

All DNA values were the mean of three or four determinations. ³H-Thymidine incorporation values were the mean of four determinations. ³H-Glucosamine incorporation values were the mean of three determinations. All other values shown were based on a single determination, but all experiments were conducted two or more times with different batches of primary cells. Results of one representative experiment are shown. Statistical software (StatView, version 5; SAS Institute, Cary, NC) was used for statistical comparisons. Samples were analyzed with a paired t-test. Standard error was used when n > 3, and standard deviation was used when n = 3. 

Results

Each growth factor stimulated a different rate of keratocyte proliferation. Cells were incubated with media containing ³H-thymidine for 3 days, beginning on days 4, 7, and 10, and incorporation per nanogram of DNA was determined on days 7, 10, and 13 (Fig. 1A) . All growth factors significantly stimulated higher levels of incorporation compared with the control for all three time points except IGF-I on day 10. At day 7, FGF-2 stimulated incorporation the most and IGF-I stimulated incorporation the least; TGF-β1 and PDGF stimulated intermediate levels of incorporation. Incorporation in the FGF-2–treated cultures decreased from day 7 to day 10 to levels lower than those of TGF-β1 and PDGF but still higher than those of IGF-I. Each of the growth factors except TGF-β1 stimulated the highest levels of incorporation on day 7; TGF-β1 stimulated the highest levels of incorporation on day 10. 

Growth factors differentially increased keratocyte culture density. DNA content of the keratocyte cultures was measured on days 1, 7, 10, and 13 to determine whether cultures treated with each of the growth factors increased the cell number (Fig. 1B) . DNA content of the control on day 1 was not significantly different from the control on day 7 (data for day 1 control not shown on figure). DNA content of the control and growth factor–treated cultures were significantly different on day 7. Compared with control, FGF-2 was 3.2-fold higher, PDGF was 2.0-fold higher, TGF-β1 was 1.8-fold higher, and IGF-I was 1.6-fold higher. DNA content for the control and growth factor–treated cultures on days 10 and 13 followed the same pattern as on day 7. 

Dermal fibroblasts and keratocytes in standard monolayer culture do not readily remove the N- and C-terminal globular domains on the procollagen they synthesize. Procollagen cannot form collagen fibrils that would associate with the cells. As a result, the procollagen is secreted into the media. ^{45 46} In the absence of collagen fibrils, proteoglycans and other components of the ECM that associate with collagen fibrils are also secreted into the media. Therefore, the medium was analyzed for the components of the ECM. Each growth factor stimulated a different level of total ECM protein and collagen synthesis. Keratocytes were cultured in media containing ³H-glycine for 3 days beginning on days 4, 7, and 10. The sensitivity of the incorporated radiolabel secreted into the media to papain and to collagenase was used to determine the amount of total ECM protein synthesis and collagen synthesis, respectively. Results show that compared with controls, IGF-I, TGF-β1, and PDGF stimulated incorporation into total ECM protein from as little as 4-fold for TGF-β1 on day 7 to as much as 15-fold for IGF-I on day 13. Keratocytes cultured in IGF-I synthesized the highest levels of total ECM protein, FGF-2 synthesized the lowest, and PDGF and TGF-β1 synthesized intermediate levels of total ECM protein at all three time points (Fig. 2A) . Levels of total ECM protein synthesized by the keratocytes cultured in IGF-I, TGF-β1, and PDGF increased from days 7 to 10 and from days 10 to13, and they increased most in the TGF-β1–treated cultures. Levels of collagen synthesis followed the same pattern as total ECM protein synthesis but were approximately 20% to 40% lower (Fig. 2B) . 

Proteins that were secreted into the media by keratocytes also were evaluated by SDS-PAGE/Coomassie blue staining (SimplyBlue SafeStain; Invitrogen). Analysis of DNA-normalized aliquots from day 10 cultures showed that each of the tested growth factors caused keratocytes to secrete proteins into the medium that produced a similar banding pattern, but the band intensity was growth factor dependent (Fig. 3) . IGF-I– and TGF-β1–treated cultures secreted the most proteins, FGF-2 secreted the least, and PDGF secreted intermediate levels of proteins. Compared with day 10 cultures, the band intensity of proteins in the medium was lower in samples from day 7 cultures and greater in samples from day 13 cultures, but the banding pattern for each of the growth factors was comparable at all three time points (data not shown). These results suggest that IGF-I, TGF-β1, PDGF, and FGF-2 stimulate the cells to secrete essentially the same matrix proteins into the media; however, the expression levels of the proteins are growth factor specific. 

The major proteins secreted into the media were identified with the use of Western blotting. We analyzed DNA-normalized aliquots of media harvested at day 10 for EDA fibronectin, an isoform of fibronectin found after wounding, ⁴⁷ and fibronectin. Antisera to EDA fibronectin reacted with a single 210-kDa band present in the media (Fig. 4A) . This protein was most abundant in the media of TGF-β1–treated cultures; lesser amounts were found in the media of cultures treated with the other growth factors, and only trace amounts were found in the media of controls. A Western blot using an antibody that recognizes all forms of fibronectin, including EDA fibronectin, indicated that the levels of fibronectin produced by the control were similar to those of the FGF-2–treated cultures (Fig. 4B) . The blue staining band at 210 kDa in Figure 3was likely to be fibronectin and was identified as such. 

Western blot analysis using an antibody specific for type Ι procollagen on DNA-normalized aliquots of medium from day 10 of culture demonstrated that type Ι procollagen was readily detected in the medium of keratocytes cultured in IGF-I, TGF-β1, and PDGF (Fig. 5A) . Compared with control, IGF-I, TGF-β1, and PDGF stimulated the production of type I procollagen, whereas FGF-2 inhibited the production of type I procollagen. The production of type ΙΙΙ procollagen by growth factor–treated cultures also was evaluated by Western blotting using an antibody specific for type ΙΙΙ procollagen. Compared with control, type ΙΙΙ procollagen was stimulated to the greatest extent in cultures treated with IGF-I and TGF-β1 and was inhibited in cultures treated with FGF-2 (Fig. 5B) . The major blue-staining bands that react with these antibodies are identified in Figure 3as proα1(Ι)(ΙΙΙ) and pNα(Ι)(ΙΙΙ). 

Fibril-forming collagen (types Ι and ΙΙΙ) content in the media and cell layer from day 10 cultures was determined by pepsin digestion and SDS-PAGE/Coomassie blue staining (SimplyBlue SafeStain [Invitrogen]; Fig. 6 ). Pepsin readily degrades globular proteins under acidic conditions, but those regions of collagen molecules that form a stable triple helix are resistant to this protease. Keratocytes cultured in IGF-I had the highest levels of fibrillar collagen in the media. TGF-β1 and PDGF induced the production of intermediate amounts of collagen deposited in the medium. Medium from FGF-2–treated cultures had levels of fibrillar collagen similar to those of the control. Fibrillar collagen associated with the cell layer was seen only in the TGF-β1–treated cultures. 

Glycosaminoglycan synthesis by cultured keratocytes in response to the different growth factors also was evaluated. Cells were radiolabeled with ³H-glucosamine from days 7 to 10, and the glycosaminoglycans secreted into the media were digested with chondroitinase ABC, endo-β-galactosidase, or hyaluronate lyase to determine the relative levels of chondroitin sulfate, keratan sulfate, and hyaluronan synthesis, respectively. Results show that compared with control, cultures treated with TGF-β1 and PDGF synthesized the highest levels of chondroitin sulfate and FGF-2 synthesized the least (Fig. 7A) . Furthermore, cultures treated with PDGF and IGF-I produced the highest levels of keratan sulfate synthesis, but FGF-2 did not stimulate keratan sulfate synthesis (Fig. 7B) . Hyaluronan synthesis was 20-fold higher than the control in keratocytes cultured in media containing TGF-β1 (Fig. 7C) . These data indicate that the levels of glycosaminoglycan synthesis stimulated by each of the growth factors generally follow the pattern of total ECM protein synthesis. 

Western blot analysis with antibodies to the core proteins of decorin, biglycan, keratocan, and lumican was performed. Aliquots of media harvested on day 10 from keratocytes cultured in each of the growth factors were normalized for DNA content, digested with chondroitinase ABC or endo-β-galactosidase, and analyzed (Fig. 8) . Decorin and keratocan production was stimulated by all the growth factors. Lumican production was stimulated only by IGF-I. Biglycan was detected only in media from keratocytes cultured in TGF-β1. 

FGF-2 induced proliferation (a hallmark of the initial phase of wound healing) the most but inhibited the synthesis of procollagen Ι, which is essential for corneal reconstruction. Therefore, we sought to determine whether keratocytes cultured in FGF-2 could still respond to other growth factors that stimulate collagen synthesis and upregulate procollagen type Ι synthesis. Keratocytes were treated with FGF-2 for 7 days; this was followed by 6 days of treatment with DMEM/F12, IGF-I, TGF-β1, FGF-2, or PDGF. DNA content of the cell layer and type Ι procollagen in the medium were determined (Fig. 9) . The DNA content of cultures continually treated with FGF-2 significantly increased from day 1 (C1) to day 7 (F7) and then declined a small amount from day 7 (F7) to day 13 (F-F). Compared with continuous culture in FGF-2 (F-F), cultures switched to PDGF (F-P) and IGF-I (F-I) had significantly higher DNA content (Fig. 9A) . Aliquots of medium normalized for DNA content were analyzed using Western blots with the antibody to type Ι procollagen (Fig. 9B) . Type Ι procollagen levels in the media of keratocytes treated with FGF-2 for 7 days (F7) and 13 days (F-F) were lower than in controls (C7 and C13), but cultures switched to IGF-I (F-I), TGF-β1 (F-T), and PDGF (F-P) had increased type Ι procollagen levels. 

Discussion

When the cornea stroma is wounded, injured keratocytes and keratocytes surrounding the wound undergo apoptosis. ^{16 48} Proliferation by some or all the remaining cells is essential to replace the cells lost because of apoptosis. The results of this study show IGF-I, TGF-β1, FGF-2, and PDGF stimulate different rates of DNA synthesis and different levels of maximal cell density. FGF-2–treated keratocytes synthesized DNA at the highest rate at the day 7 time point and reached the highest level of maximal cell density. PDGF-treated keratocytes synthesized DNA at a lower rate at day 7 time point but then exceeded FGF-2 at days 10 and 13 and eventually achieved nearly the same cell density as cells treated with FGF-2. Keratocytes treated with TGF-β1 synthesized DNA at a rate similar to that of PDGF and achieved the third highest level of maximal cell density. IGF-I–treated keratocytes had the lowest rate of DNA synthesis and the lowest level of maximal cell density. These results indicate that for these growth factors, the level of maximal cell density achieved is related to the rate of DNA synthesis induced by that growth factor. All these growth factors, however, are potential candidates for regulatory roles in controlling proliferation involved in the hypercellular phase of corneal stromal wound healing. 

After proliferation, the synthesis of an ECM-containing collagen is essential for restoring the integrity of the corneal stroma after wounding. Results of this study show that IGF-I, TGF-β1, and PDGF stimulate significantly different levels of collagen synthesis; IGF-I stimulated higher levels of collagen synthesis than did TGF-β1 and PDGF. FGF-2 did not stimulate collagen synthesis. This same pattern was observed in ³H-glycine labeling for total ECM proteins and collagenous proteins, Coomassie blue staining of matrix proteins, Western blotting with antibodies to types Ι and ΙΙΙ procollagen, and levels of collagenous proteins revealed by pepsin digestion of the medium. The level of collagen synthesis induced by each of the growth factors is, then, inversely related to the level of confluence achieved by keratocytes cultured in that growth factor. 

In addition to collagens, proteoglycans and glycoproteins play a major role during wound healing. Keratan sulfate proteoglycans are normally present in the corneal stroma at high levels and function to direct collagen assembly. In contrast, hyaluronan is present only in mature corneas during wound healing and may reduce transparency by disrupting the spacing of collagen fibrils. ^{20 49} We found that though all the growth factors stimulated chondroitin sulfate synthesis, FGF-2 stimulated its synthesis the least. FGF-2 was also the only growth factor that did not stimulate keratan sulfate synthesis. TGF-β1 treatment, however, also stimulated the synthesis of biglycan and hyaluronan, known markers of ECM scarring. ^{20 50 51} In addition, TGF-β1 stimulated the highest levels of EDA fibronectin and the greatest amount of collagen associated with the cells. Fibronectin is known to assist in the formation of ECM by enhancing the conversion of procollagen to collagen. ⁵² These results suggest that in addition to stimulating the synthesis of collagens, IGF-I and PDGF also stimulate the synthesis of the collagens, proteoglycans, and glycoproteins that more closely resemble normal corneal stromal composition. Therefore, the regeneration of normal stroma after injury is likely to involve these growth factors, whereas the production of a scar-type ECM would involve TGF-β1. 

Although FGF-2 stimulated keratocytes to proliferate at high levels and stimulated the production of decorin and chondroitin sulfate, it did not stimulate the synthesis of collagenous proteins. Western blot analysis, however, indicated that FGF-2 actually inhibited the synthesis of types Ι and ΙΙΙ procollagen. The ECM produced by the action of FGF-2 on keratocytes would essentially consist of proteoglycans and fibronectin; without collagen, it would lack tensile strength. We also showed that the FGF-2–mediated inhibition of type Ι procollagen synthesis could be reversed by subsequent culture in media containing the growth factors that stimulate collagen synthesis, IGF-I, and PDGF, growth factors that also stimulate the synthesis of the normal proteoglycans. This indicates that the inhibition of type Ι procollagen synthesis by FGF-2 was not a permanent phenotypic change. 

In conclusion, we suggest that stromal wound healing occurs as a result of the sequential action of growth factors on keratocytes (Fig. 10) . Quiescent keratocytes are activated to proliferate on wounding to create regions of hypercellularity that consist of densely packed cells with a sparse, or provisional, matrix that contains keratocan. ^{21 23 53} We found that though FGF-2 stimulated proliferation and chondroitin sulfate synthesis, it did not stimulate keratan sulfate synthesis, and it inhibited procollagen types Ι and ΙΙΙ synthesis. This low level of total ECM synthesis by FGF-2–activated keratocytes would result in a sparse matrix. Thus, we speculate that the regions of hypercellularity seen during the initial phases of wound healing are keratocytes that have been activated by FGF-2. In vivo, the provisional matrix in regions of hypercellularity can be replaced by a fibrillar matrix. We also observed that collagen synthesis can be restored to FGF-2–activated keratocytes by subsequent culture in media containing IGF-I, TGF-β1, and PDGF. Thus, we further speculate that IGF-I or PDGF could have acted on the hypercellular keratocytes to make them become collagenous keratocytes. These collagenous keratocytes replaced the provisional matrix with a normal fibrillar matrix that restored stromal integrity, and the keratocytes later become quiescent again. TGF-β1 also restored collagen synthesis to FGF-2–activated keratocytes, but this growth factor also stimulated the synthesis of biglycan and hyaluronan, markers of fibrosis. We propose, therefore, that TGF-β1 causes hypercellular keratocytes to become fibrocollagenous keratocytes and to produce a scar-type fibrillar matrix. TGF-β1 has also been shown to cause keratocytes to become myofibroblasts, based on the appearance of α-smooth muscle actin. ³⁸ Keratocytes made myofibroblastic by TGF-β1 would be synonymous with fibrocollagenous keratocytes. The absence of any subsequent growth factor would result in a hypercellular scar. The targeted modulation of these signaling molecules has considerable clinical relevance. 

 Supported by National Institutes of Health Grants EY08104 (JH) and 5R01EY008104 (LE).

 Submitted for publication October 29, 2008; revised December 19, 2008, and January 16, 2009; accepted April 13, 2009.

 Disclosure: L. Etheredge, None; B.P. Kane, None; J.R. Hassell, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Keratocyte DNA synthesis and cell density. Keratocytes were cultured in DMEM/F12 (Control) or DMEM/F12 supplemented with IGF-I, TGF-β1, FGF-2, or PDGF. (A) Keratocyte DNA synthesis. Cultures were radiolabeled with ³H-thymidine on days 4 to 7, 7 to 10, and 10 to 13. Incorporation was determined and expressed per nanogram of DNA. Compared with control, IGF-I, TGF-β1, FGF-2, and PDGF significantly (P < 0.011) increased ³H-thymidine incorporation at all three time points except for IGF-I on day 10. (B) Keratocyte density. DNA content of keratocyte cultures measured on days 7, 10, and 13. Keratocytes cultured with IGF-I, TGF-β1, FGF-2, and PDGF contained significantly (P < 0.044) more DNA than the control at all three time points.

Figure 2.

 

Protein and collagen synthesis by keratocytes in culture. Keratocytes were cultured with DMEM/F12 (Control) or DMEM/F12 supplemented with IGF-I, TGF-β1, FGF-2, or PDGF and were radiolabeled with ³H-glycine on days 4 to 7, 7 to 10, and 10 to 13. Sensitivity of the incorporated radiolabel secreted in the media to papain (A, total protein) and to collagenase (B, collagenous protein) was determined and expressed per nanogram of DNA.

Figure 3.

 

Coomassie blue–stainable proteins secreted into the medium by keratocytes in culture. Keratocytes were cultured with DMEM/F12 (Control) or DMEM/F12 supplemented with IGF-I, TGF-β1, FGF-2, or PDGF. Medium from days 7 to 10 was harvested and analyzed with SDS-PAGE. The bands later identified by Western blot in Figures 4 and 5are labeled fibronectin (FN), procollagen I (proα1(Ι) and pNα1(Ι)), and procollagen ΙΙΙ (proα1(ΙΙΙ) and pNα1(ΙΙΙ)).

Figure 4.

 

Western blot analysis of media from day 10 keratocyte cultures using antibodies to EDA fibronectin for EDA fibronectin (A) and to fibronectin for total fibronectin (B). Keratocytes were cultured with DMEM/F12 (Control) or DMEM/F12 supplemented with IGF-I, TGF-β1, FGF-2, or PDGF. The bands detected by these antibodies correspond to the bands marked FN on Figure 3 .

Figure 5.

 

Western blot analysis of media from day 10 keratocyte cultures using antibodies to procollagen type Ι (A) and procollagen type ΙΙΙ (B). Keratocytes were cultured with DMEM/F12 (Control) or DMEM/F12 supplemented with IGF-I, TGF-β1, FGF-2, or PDGF. The bands marked proα1(Ι), proα1(ΙΙΙ), pNα1(Ι), and pNα1(ΙΙΙ) correspond to the migration position of the band marked proα1(Ι)(ΙΙΙ) and pNα1(Ι)(ΙΙΙ) on Figure 3 . The migration position of pepsin-digested collagen types Ι and ΙΙΙ are marked α1(Ι), α2(Ι), and α1(ΙΙΙ), respectively.

Figure 6.

 

Fibrillar collagen present in the medium and cell layer of day 10 keratocyte cultures. Keratocytes were cultured with DMEM/F12 (Control) or DMEM/F12 supplemented with IGF-I, TGF-β1, FGF-2, or PDGF. Cultures were harvested on day 10, and samples were digested with pepsin before electrophoresis. A pepsin alone control was added (Pepsin). The migration position of the α1(Ι), α2(Ι), and α1(ΙΙΙ) chains of collagen are marked.

Figure 7.

 

Glycosaminoglycan synthesis by keratocyte cultures. Keratocytes were cultured with DMEM/F12 (Control) or DMEM/F12 supplemented with IGF-I, TGF-β1, FGF-2, or PDGF and radiolabeled with ³H-glucosamine on days 7 to 10. The incorporated radiolabel secreted in the media was then characterized by its sensitivity to chondroitinase ABC for chondroitin sulfate (A), to endo-β-galactosidase for keratan sulfate (B), and to hyaluronan lyase for hyaluronan (C). Compared with control, the synthesis of chondroitin sulfate (A) was significantly (P < 0.008) higher in cultures treated with each of the growth factors. Keratan sulfate (B) was significantly (P < 0.012) higher in cultures treated with TGF-β1, FGF-2, and PDGF. Hyaluronan (C) was significantly (P < 0.05) higher in cultures treated with TGF-β1, FGF-2, and PDGF.

Figure 8.

 

Western blot analysis of medium from day 10 keratocyte cultures using antibodies to decorin, biglycan, keratocan, and lumican. Keratocytes were cultured with DMEM/F12 (Control) or DMEM/F12 supplemented with IGF-I, TGF-β1, FGF-2, or PDGF. Media were digested with chondroitinase ABC for decorin and biglycan or with endo-β-galactosidase for keratocan and lumican before electrophoresis.

Figure 9.

 

DNA content and procollagen type Ι production of keratocytes cultured first with FGF-2 and then changed to each of the growth factors. Keratocytes were cultured with FGF-2 for 7 days (F7), and medium was changed to DMEM/F12 (F-C) or DMEM/F12 supplemented with IGF-I (F-I), TGF-β1 (F-T), FGF-2 (F-F), or PDGF (F-P), and the keratocytes were cultured for 6 additional days. Keratocytes cultured with DMEM/F12 on days 1 (C1), 7 (C7), and 13 (C13) were used as controls. (A) Keratocyte density. DNA content of keratocytes cultured with FGF-2 for 7 days was significantly (P < 0.0001) higher than the day 7 control. Keratocytes cultured in FGF-2 for 7 days and switched to PDGF (F-P) or to IGF-I (F-I) had significantly (P < 0.025) higher DNA content than keratocytes continuously cultured in FGF-2. (B) Procollagen type Ι. Western blot using an antibody to procollagen type Ι of media harvested on days 7 and 13 of culture.

Figure 10.

 

Schematic representation of possible growth factor–mediated wound healing of the corneal stroma. On wounding, quiescent keratocytes are activated by FGF-2 to become hypercellular keratocytes that proliferate and synthesize a provisional matrix that lacks collagen but contains proteoglycans and fibronectin. Hypercellular keratocytes can be stimulated by IGF-I or PDGF to become collagenous keratocytes and synthesize a normal collagenous matrix abundant in collagen type Ι and keratocan. In this case, collagenous keratocytes eventually became quiescent. Hypercellular keratocytes can also be stimulated by TGF-β1 to become fibrocollagenous keratocytes and produce an ECM scar, rich in EDA fibronectin, biglycan, and hyaluronan. In addition, without subsequent activation by other growth factors, hypercellular keratocytes can remain hypercellular, which gives rise to a hypercellular scar.