Overexpression of Collagenase (MMP-1) and Stromelysin (MMP-3) by Pterygium Head Fibroblasts FREE

PTERYGIUM represents one of the most common external eye diseases in countries with relatively high exposure to UV irradiation.^{1 - 2} Patients with an early pterygium frequently complain of increased redness and irritation, and their vision is reduced when the pterygium advances into the cornea. Although UV irradiation is causatively linked with the formation of pterygium, the underlying mechanism leading to corneal invasion remains obscure. Clinically, pterygium is characterized by a wing-shaped overgrowth of conjunctival tissue into the cornea and can be grossly subdivided into 2 portions, ie, the head and the body. One histopathological feature of pterygium overgrowth is excessive fibrovascular proliferation.^{3 - 5} The extent of such fibrovascular proliferation has recently been regarded as a reliable morphologic index for predicting recurrence following surgical removal.⁶ The other pertinent feature is the loss of the basement membrane, the Bowman membrane, and the superficial corneal stroma at the area invaded by the fibrovascular tissue.⁷ It has thus been speculated that such a tissue loss is a result of destruction by the invading pterygium,⁸ although direct evidence to support such a hypothesis is lacking.

Matrix metalloproteinases (MMPs) are a family of enzymes that act to modify or degrade the extracellular matrix .^{9 - 11} These enzymes are synthesized and secreted by a variety of cell types including fibroblasts. At least 19 members of the MMP family have been identified and categorized into 5 groups: collagenases (MMP-1, MMP-8, and MMP-13), gelatinases(MMP-2 and MMP-9), stromelysins (MMP-3, MMP-10, and MMP-11), membrane-type MMPs, and others. Matrix metalloproteinases are normally coexpressed with a family of tissue inhibitors of metalloproteinases (TIMPs), which inhibit active forms of MMPs. At least 4 inhibitors, 28-kd TIMP-1, 21-kd TIMP-2, 23-kd TIMP-3 and 24-kd TIMP-4 have been characterized and are also produced by many cell types including fibroblasts.^{9 - 10} The balance between the activity of MMPs and that of TIMPs determines the extent of proteolysis linked with tissue remodeling or degradation.^{10 - 11} For example, an increased ratio of MMPs to TIMPs is correlated with how tumor cells invade the host stroma.¹² Besides MMPs and TIMPs, the other proteolytic cascade leading to tissue degradation and remodeling involves urokinase plasminogen activator (uPA), a serine protease. For example, overexpression of uPA is also correlated with the invasiveness of human cancer cells.¹³ Herein, we provide experimental evidence that pterygium head fibroblasts in culture overexpress MMP-1 and MMP-3 without changes in MMP-2, TIMP-1, TIMP-2, and uPA when compared with fibroblasts derived from other portions of the pterygium, the normal cornea, and conjunctiva.

Dulbecco-Eagle minimum essential medium (DMEM), fetal bovine serum (FBS), fungizone, phenol, DNA or RNA size marker, and random primers DNA labeling kit were purchased from GIBCO-BRL (Grand Island, NY). Cell culture dishes, 6-well plates, and 15-mL centrifuge tubes were from Becton Dickinson (Lincoln Park, NJ). BCA protein assay kit was from Pierce Chemical (Rockford, Ill). Zymogram-ready gels containing gelatin or casein, 4% to 15% tris-hydrochloride polyacrylamide gradient ready gel, sodium dodecyl sulfate, and electrophoresis equipment were from Bio-Rad (Hercules, Calif). Human MMP-1 and MMP-3 enzyme-linked immunosorbent assay (ELISA) kits and the monoclonal antibodies against human MMP-1, MMP-2, MMP-3, MMP-9, TIMP-1, and TIMP-2 were from Oncogene Research Products of Calbiochem (Cambridge, Mass). Vectastain Elite ABC peroxidase kit was from Vector Laboratories (Burlingame, Calif). Nitrocellulose membranes were from Scheicher and Schuell (Keene, NH). GeneAmp RNA–polymerase chain reaction (PCR) kit was from Perkin-Elmer Cetus (Norwalk, Conn). Wizard PCR Preps DNA purification kit was from Promega (Madison, Wis). α-Phosphorus 32 deoxycytidine triphosphate ([α-³²P]-dCTP)was from DuPont NEN (Boston, Mass). XAR-5 and BioMax MS-1 films and intensifying screens were from Eastman Kodak (Rochester, NY). All other reagents and chemicals were from Sigma-Aldrich (St Louis, Mo).

Human corneas from donors aged 31 to 50 years that were preserved for less than 72 hours were obtained from the Florida Lions Eye Bank. Normal conjunctiva and pterygial specimens were obtained from patients receiving cataract and pterygial removal, respectively, with patient's informed written consent and followed tenants of the Helsinki declaration. Using the method described below, a total of 6 strains of healthy conjunctival fibroblasts, 8 of pterygium head fibroblasts, and 6 of pterygium body fibroblasts were obtained. For this study, fibroblasts at the third or fourth passage were used. The technique of removing the pterygium and its surrounding subconjunctival fibrovascular tissue has previously been described.¹⁴ The pterygial specimen was further subdivided into the head and the body, of which the former comprised the tip area of 2 × 2 mm, while the latter included the remainder of the pterygial specimen (for anatomic designation see Figure 1). Normal corneal, limbal, and conjunctival fibroblasts and pterygial head and body fibroblasts were obtained from explant cultures using a technique identical to a previously described method.¹⁵ In brief, each tissue was cut into explants of approximately 2 × 2 mm² and placed onto 100-mm tissue culture dishes. Ten to 20 minutes later, each explant was covered with a drop of DMEM containing 10% FBS (DMEM-FBS), 50 µg/mL of gentamicin, and 1.25 µg/mL of fungizone and placed overnight in an incubator at 37°C under 95% humidity with 5% carbon dioxide. On the next day, 10 mL of the same media was added and the media were changed twice weekly thereafter. These fibroblasts were subcultured with 0.1% trypsin and 0.02% EDTA in calcium-free Eagle minimum essential medium at 80% to 90% confluence with 1:3 split for several passages.

For Northern blot analysis, fibroblasts were cultured for 7 to 9 days in 100-mm dishes containing DMEM-FBS until confluence before extraction of total RNA. For ELISA, Western blot analysis, zymography, and quantitative collagenase assay, the same passages of fibroblasts from different sources were seeded in triplicates at the same density (1.5 × 10⁵cells/well) in 6-well plates and grown for 7 to 9 days until confluence. After being washed 4 times with phosphate-buffered saline, these cultures were switched to the same volume (1 mL) of a serum-free DMEM, containing 5 µg/mL of insulin, 5 µg/mL of transferrin, and 5 ng/mL of sodium selenite (DMEM-ITS) and incubated for additional 24 and 48 hours. The conditioned media were then collected, and the adherent cells were trypsinized for cell counting and lysed in phosphate-buffered saline (pH 7.3) containing 1.5-mol/L sodium chloride and 0.039% Triton X-100 for BCA protein assay.

Five of the human DNA probes, including 185 base pair (bp) fragment of MMP-1, 480 bp of MMP-2, 155 bp of MMP-3, 551 bp of TIMP-1, and 590 bp of TIMP-2 were provided by Velidi H. Rao, PhD (University of Nebraska Medical Center, Omaha). Three complementary DNA (cDNA) probes, 640 bp of MMP-9, 519 bp of uPA, and 498 bp of glyceraldehyde-3-phosphate dehydrogenase were purified from reverse transcriptase–PCR products by electrophoresis through a 1.2% low melting agarose gel using a DNA purification kit according to the manufacturer's protocol (Wizard PCR Prep, Promega). The primers used for PCR were 1502-1531 (sense) and 2111-2140 (antisense) for MMP-9 (accession J05070), 487-506 (sense) and 982-1002 (antisense) for uPA (accession A18397), 541-561(sense) and 1018-1038 (antisense) for glyceraldehyde-3-phosphate dehydrogenase(accession M33197). The ³²P-labeled cDNA probes (1-2 × 10⁹ cpm/µg DNA) were prepared with [α-³²P]-dCTP(1.1 × 10¹⁴ Bq/mmol [3000 Ci/mmol]), using a random primers DNA labeling system.

Total RNA isolation and Northern hybridization were performed using a previously described method.¹⁵ Briefly, total RNA was isolated from various types of fibroblast cultures by acid guanidium thiocyanate-phenol-chloroform extraction. Total RNA at 25 µg per lane was electrophoresed through 1.2% agarose containing formaldehyde, transferred to nitrocellulose membranes, and hybridized with ³²P-labeled cDNA probes at 2 × 10⁶ to 4 × 10⁶ cpm per 3 to 8 ng/mL in the hybridization solution. After visualization of the hybridization product in the x-ray film, the ³²P label on the membrane was stripped by washing the membranes at 65°C for 1 hour twice in 5-mmol/L trishydrochloride(pH 8.0), 0.2-mmol/L EDTA, 0.05% sodium pyrophosphate, and 0.1 × Denhardt solution, and rehybridized with other ³²P-labeled probes. The relative amount of each messenger RNA (mRNA) of interest was determined by scanning its autoradiofluorogram with a laser scanner (Densitometer Model FB910; Fisher Scientific, Pittsburgh, Pa) and normalized as a ratio to that of the glyceraldehyde-3-phosphate dehydrogenase mRNA band.

Human MMP-1 or MMP-3 double-sandwiched ELISA was performed using commercial ELISA kits according to the manufacturer's protocol. In brief, 100 µL of standard dilutions of recombinant human MMP-1 or MMP-3 and experimental conditioned media were dispensed into a 96-well microtiter plate coated with mouse anti-MMP-1 or MMP-3 monoclonal antibody, respectively. The plate was sealed, incubated at room temperature for 2 hours or at 4°C for 1 hour, respectively, and washed 4 times with phosphate-buffered saline containing 0.033% Tween 20. After addition of 100 µL of diluted rabbit anti–MMP-1 serum into each well and incubation for 2 hours followed by 4 washes, 100 µL of diluted donkey antirabbit horseradish peroxidase conjugates was added and incubated for 1 hour. For MMP-3, 100 µL of diluted rabbit anti–MMP-3 horseradish peroxidase was added into each well and incubated at 4°C for 2 hours. Aliquots of 100 µL of the color reagent 3, 3', 5,5'-tetramethylbenzidine were then applied for 30 minutes to develop a blue color, and the reaction was stopped by adding 100 µL of 1-mol/L sulfuric acid. Absorbance was read at 450 nm by an automatic plate reader with a reference wavelength of 570 nm.

To identify MMP and TIMP proteins present in each fibroblast-conditioned medium, Western blot analysis was performed using their specific antibodies. Conditioned media from different fibroblast cultures were adjusted to a final volume of 20 to 25 µL to represent the same quantity of cellular protein(8.3 µg) or cell number (5000 cells) and electrophoresed under reducing condition at 4°C in a 4% to 15% gradient polyacrylamide gel. After electrophoretic transfer to a nitrocellulose membrane at 4°C, the membrane was immersed with 0.1% (vol/vol) Tween 20 in tris-buffered saline (100-mmol/L tris, 0.9% sodium chloride, pH 7.5) (TTBS) for 30 minutes with agitation. The primary antibody, ie, 1 µg/mL of mouse monoclonal antibody against human MMP-1, MMP-2, MMP-3, MMP-9, TIMP-1, or TIMP-2, in TTBS containing 1% horse serum was placed on each membrane and incubated at roomtemperature for 60 minutes with agitation. After being washed with 3 to 4 changes of TTBS over 15 minutes, each membrane was transferred to a 1:200 diluted solution of biotinylated second antibody (goat antimouse IgG from Vectastain Elite ABC kit) in TTBS containing 1% horse serum and incubated for 30 minutes. After 3 to 4 washes with the same solution, the membranes were incubated to 1:50 diluted Vectastain Elite ABC reagent conjugated with peroxidase for 30 minutes and processed for color development in 0.5 µg/mL of diaminobenzidine in 50-mmol/L tris-hydrochloride (pH 7.2) containing 0.05% hydrogen peroxide for 10 to 20 minutes.

To determine gelatinolytic and stromelysin activities of the various fibroblast cultures, zymography was performed using a method similar to that previously described.¹⁶ Each conditioned medium(12-15 µL), after being adjusted to represent the same quantity of cellular protein (5 µg) or cell number (3000 cells), was treated with sample buffer without boiling or reduction. Sodium dodecyl sulfate–polyacrylamide gel electrophoresis was performed using a 10% polyacrylamide gel containing 0.1% gelatin or a 12% gel containing 0.1% casein. The gels were soaked in 2.5% Triton X-100 for 30 minutes at room temperature to remove the sodium dodecyl sulfate and incubated in a reaction buffer (50-mmol/L tris-hydrochloride[pH 7.5], 200-mmol/L sodium chloride, 5-mmol/L calcium chloride, and 0.02% 23 lauryl ether [Brij-35]) at 37°C overnight to allow proteinase digestion of its substrate. Gels were rinsed again in distilled water, stained with 0.5% Coomassie brilliant blue R-250 in 40% methanol and 10% acetic acid for 1 hour, and destained with 40% methanol and 10% acetic acid. Proteolytic activities appeared as clear bands of lysis against a dark background of stained gelatin or casein. To verify that the detected gelatinolytic and caseinolytic activities were specifically derived from metalloproteinases, the gels were treated with the Triton X-100 solution and the tris–sodium chloride/calcium chloride reaction buffer containing 5-mmol/L phenylmethylsulfonyl fluoride with or without 10-mmol/L EDTA in the parallel experiments.

Collagenase activity was verified and quantified by incubation with soluble, telopeptide-free collagen extracted from rat skin and labeled with[³H]-acetic anhydride.¹⁷ For assay of collagenase with labeled substrate, it was necessary to destroy the TIMPs in media (500 µL) from various fibroblast cultures by reduction in 2-mmol/L dithiothreitol at 37°C for 30 minutes, followed by alkylation in 5-mmol/L iodoacetamide at 37°C for 30 minutes. This step also destroys any α₂-macroglobulin. The samples were chilled on ice and dialyzed against the assay buffer (50-mmol/L tris-hydrochloride, 200-mmol/L sodium chloride, 10-mmol/L calcium chloride, and 0.005% Brij-35, pH 7.5) for 4 hours before use. Each sample was then tested in triplicate at different volumes (10, 30, and 90 µL), and the assay was performed twice for accuracy. One of each triplicate sample was added with aminophenylmercuric acetate to a final concentration of 0.5 mmol/L to activate latent procollagenase. Another was added with 1,10-phenanthroline to a final concentration of 2.0 mmol/L in the presence of 0.5-mmol/L aminophenylmercuric acetate to chelate the zinc and inactivate the collagenase. The blanks were prepared by replacing the conditioned medium with an equivalent volume of the assay buffer. [³H]-acetic collagen (116 000 cpm per 16.6 µg/5 µL) was added to each sample and the final volume was adjusted to 110 µL with the assay buffer. The incubation was performed at 30°C for 18 hours, and the reaction was stopped by placing the tubes in an ice bath. After adding with 120 µL of the assay buffer containing 200 µg acid-soluble intact collagen as a cold carrier, 20 µg of trypsin, 20 µg of chymotrypsin, and 30-mmol/L EDTA, the second digestion was performed at 31.5°C for 90 minutes. The soluble digested products were separated from the undigested collagen by precipitating with an equal volume of ice-cold 20% trichloroacetic acid. After centrifugation at 13 000 rpm for 5 minutes, triplicate aliquots (100 µL each) of the supernatant (representing trichloroaceticacid–soluble peptides) were counted by liquid scintillation for 3 minutes. The collagenase activity was reported as units per milliliter (1 U of enzyme digests 1 µg of collagen per minute at 30°C).

The t test was used for statistical comparison for the data of Northern hybridization, ELISA, and collagenase activity assay.

Northern blot analysis showed that the transcripts of 3 MMPs, ie, 2.2 kilobase (kb) of MMP-1, 3.1 kb of MMP-2, and 1.9 kb of MMP-3, and of 2 TIMPs, ie, 0.9 kb of TIMP-1 and 3.5 kb of TIMP-2, were expressed by all 5 types of fibroblasts cultured from normal cornea, limbus, conjunctiva, pterygium head, and pterygium body (Figure 2). The sizes of these 5 transcripts were consistent with those previously reported.^{13 ,18 - 19} No MMP-9 transcript was detected in any of these 5 fibroblasts. Among the 3 MMPs expressed, the MMP-2 transcript did not show any difference in quantity among the 5 types of fibroblasts (Figure 2). In contrast, corneal fibroblasts expressed much more MMP-1 and MMP-3 transcripts than limbal fibroblasts, and a marked increase in the expression of MMP-1 and MMP-3 transcripts by pterygium head fibroblasts was noted (Figure 2). Transcripts of both TIMP-1 and TIMP-2 were expressed by all 5 fibroblasts. TIMP-1 transcript was more predominantly expressed by human corneal fibroblast. Expression of TIMP-1 by pterygium head fibroblasts was slightly lower than that by normal conjunctival fibroblasts. There was no significant difference in the expression of TIMP-2 transcript by these 5 types of fibroblasts (Figure 2).

To verify that pterygium head fibroblasts indeed carried the phenotype of MMP-1 and MMP-3 overexpression, normal conjunctival fibroblasts from 6 patients, pterygium head fibroblasts from 8 patients, and pterygium body fibroblasts from 6 patients were compared. As shown in Figure 3, the amount of the MMP-1 transcript expressed by 6 strains of normal conjunctival fibroblasts was low to undetectable. In contrast, the amount of the MMP-1 transcript was dramatically increased from 11 to 94 folds(mean ± SD, 38.6 ± 28.3 folds) in all 8 strains of pterygium head fibroblasts when compared with that of normal conjunctival fibroblasts(P<.01; n = 8). Except 1 (No. 6), the amount of the MMP-1 transcript expressed by 6 strains of pterygium body fibroblasts was also low to undetectable and was not significantly different from that of normal conjunctival fibroblasts (mean ± SD, 2.3 ± 2.2 folds, P>.1; n = 6). A similar trend was noted for the expression pattern of the MMP-3 transcript. The amount of the MMP-3 transcript was also dramatically increased in all 8 pterygium head fibroblasts except 1 (No. 3) from 1.6 to 37 folds (mean ± SD, 12.6 ± 13.6 folds) when compared with that of normal conjunctival fibroblasts (P<.05; n = 8). Although the amount of the MMP-3 transcript expressed by 2 of 6 pterygium body fibroblasts (Nos. 2 and 5) was higher than, but, as a group, was not statistically different from that of normal conjunctival fibroblasts (mean ± SD, 1.28 ± 1.33 folds; P>.5; n = 6). There was no significant difference in the transcript expression of MMP-2 and uPA (Figure 3) and TIMP-1 and TIMP-2 (not shown) among these 3 groups of fibroblasts.

The protein level of MMP-1 and MMP-3 was determined by its respective ELISA in their serum-free conditioned media (Figure 4). The amount of MMP-1 and MMP-3 in the conditioned media from corneal fibroblasts was 48.4 ± 16.2 and 370.4 ± 36.1 ng/mL(n = 3), respectively, which were 10.4-fold and 9.0-fold higher (P<.05 and P<.01; n = 3) than that (4.7 ± 3.7 and 41.2 ± 12.9 ng/mL; n = 3) from conjunctival fibroblasts, respectively. Similar to the aforementioned Northern data, the amount of MMP-1 and MMP-3 in the conditioned media from pterygium head fibroblasts was the highest among the 5 types of fibroblasts (332.0 ± 52.2 and 1448.5 ± 67.2 ng/mL, respectively; n = 3), which were 71-fold and 35-fold higher (P<.001 and P<.001; n = 3) than that from normal conjunctival fibroblasts, respectively. MMP-1 and MMP-3 protein levels decreased in order from pterygium head to body to subconjunctival fibroblasts, of which the latter two were slightly but not significantly higher(all P>.1; n = 3) than that of normal conjunctival fibroblasts, respectively.

Western blot was performed to identify and compare the protein expression of MMPs and TIMPs in serum-free conditioned media derived from normal conjunctiva and pterygium head, body, and subconjunctival fibroblasts using their specific monoclonal antibodies. Monoclonal antibodies to MMPs used in this study recognize both latent and active forms. As shown in Figure 5, the protein amount of each MMP and TIMP expressed by the first 3 fibroblasts was consistent with their mRNA expression. In other words, the amounts of a 54-kd MMP-1 band and a 57-kd MMP-3 band secreted by pterygium head fibroblasts were markedly increased when compared with those from normal conjunctival fibroblasts and pterygium body and subconjunctival fibroblasts. In normal conjunctival fibroblasts, MMP-1 and MMP-3 proteins were secreted much less than MMP-2. In pterygium head fibroblasts, these 2 proteins were secreted much more than MMP-2. In pterygium body fibroblasts, MMP-1 and MMP-3 were secreted at a level similar to MMP-2. In contrast, MMP-1 and MMP-3 proteins secreted by pterygium subconjunctival fibroblasts were less than MMP-2 protein, a pattern similar to that of normal conjunctival fibroblasts. Because the level of 72-kd MMP-2 protein was similar among all 4 fibroblasts, it is concluded that the ratios of MMP-1 and MMP-3 to MMP-2 increased from subconjunctival fibroblasts to body fibroblasts and head fibroblasts. No MMP-9 was detected by Western blot, similar to the result of Northern hybridization. The protein levels of 28-kd TIMP-1 and 21-kd TIMP-2 did not show any notable difference in the conditioned media secreted by these 4 types of fibroblasts.

Zymography was performed to verify the gelatinolytic and caseinolytic activities of MMPs detected by Western blot in serum-free conditioned media. For comparison, all 6 types of fibroblasts were subcultured at the same density until confluence and switched to the same volume of serum-free DMEM-ITS for 24 or 48 hours. The 48-hour media contained slightly higher MMP activity than that of 24-hour media. As shown in Figure 6A, the strong gelatinolytic activity of a 72-kd clear band noted on the gelatin zymogram corresponded to MMP-2. This band was similarly detected in 48-hour conditioned media from all 6 fibroblasts. Both latent (predominantly) and active forms of MMP-2 existed. This gelatinolytic activity of 72-kd MMP-2 was completely inhibited by incubating the gel with solutions containing 10-mmol/L EDTA (not shown). The gelatinolytic activity of 92-kd MMP-9 was not detected in any of these conditioned media.

The casein zymogram also disclosed a strong caseinolytic activity of 57-kd MMP-3 and a 70-kd proteinase in these conditioned media (Figure 6B and Figure 7A). With respect to MMP-3, the enhanced activity was primarily produced by pterygium head fibroblasts (Figure 6B), which was significantly more than body fibroblasts. In contrast, both subconjunctival fibroblasts and normal conjunctival fibroblasts produced little caseinolytic activity of MMP-3 (Figure 7A). The 2 bands of MMP-3 with molecular weights close to each other might be of the glycosylated and unglycosylated forms of MMP-3. For comparison, the caseinolytic activity of a 70-kd proteinase band was produced in a similar amount by these 4 fibroblasts. Its identity might be a serine proteinase, because its caseinolytic activity disappeared after treatment with 5-mmol/L phenylmethylsulfonyl fluoride, one of the serine proteinase inhibitors (Figure 7B). In contrast, the caseinolytic activity of 57-kd MMP-3 withstood such a treatment (Figure 7B) but disappeared after incubating the gel with the solution containing 10-mmol/L EDTA, one of the metalloproteinase inhibitors (Figure 7C). These results also confirmed that the markedly increased caseinolytic activity by pterygium head fibroblasts was indeed that of a metalloproteinase.

The collagenolytic activity of MMP-1 was investigated with a soluble, telopeptide-free collagen extracted from rat skin and labeled with [³H]-acetic anhydride. We noted that such an activity was mainly present in a latent form in the serum-free conditioned media of different fibroblasts and could be activated by p-aminophenylmercuric acetate (APMA). As shown in Figure 8, the activity after activation produced by pterygium head fibroblasts was 0.27 ± 0.01 U/mL, ie, 4.8-fold higher than that produced by normal conjunctival fibroblasts (0.056 ± 0.007 U/mL; P<.005; n = 3). The activities produced by pterygium body and subconjunctival fibroblasts were 0.04 ± 0.01 and 0.07 ± 0.002 U/mL, respectively, both of which were similar to that of normal conjunctival fibroblasts (P>.05; n= 3). This finding was consistent with the transcript and protein expression of MMP-1, indicating that pterygium head fibroblasts produced more MMP-1 than other fibroblasts.

Matrix metalloproteinases and their inhibitors, TIMPs, play a vital role in connective tissue degradation and remodeling.^{10 - 11} In the human eye, studies of MMPs and TIMPs have been performed in the aqueous humor,^{20 - 21} vitreous,^{22 - 23} keratoconus corneas,^{24 - 25} myopic sclera,²⁶ and corneas during wound healing.^{27 - 28} These studies have focused primarily on gelatinase A (MMP-2), gelatinase B (MMP-9), TIMP-1, and TIMP-2. Our study concludes that under the same culture condition with respect to the cell passage, seeding density, and presence or absence of FBS, the expression and activity of MMP-1 and MMP-3, but not those of MMP-2 or MMP-9, are significantly increased in pterygium head fibroblasts, while the expression of TIMP-1 and TIMP-2 is not changed compared with fibroblasts from other regions of the pterygium and from the normal bulbar conjunctiva. These data provide for the first time direct evidence supporting the notion that invasion of the pterygium may in part be facilitated by an unusual population of head fibroblasts. Although uPA, a serine protease, is found to be responsible for tissue degradation and tumor cell invasion,¹³ its expression is not changed among normal conjunctival and pterygial head and body fibroblasts.

Up-regulation of MMP-1 and MMP-3 transcript and protein expression by pterygium head fibroblasts was demonstrated by Northern hybridization (Figure 2 and Figure 3), ELISA (Figure 4), and Western blot analysis (Figure 5), respectively. For comparison, although transcripts of TIMP-1 was slightly down-regulated (Figure 2), their protein levels were not changed (Figure 5). Taken together, these data suggest that the ratio between MMPs and TIMPs produced by pterygium head fibroblasts is higher than those produced by other fibroblasts tested. This notion is further supported by a higher caseinolytic activity of MMP-3 (Figure 6B and Figure 7) and a significantly higher collagenolytic activity of MMP-1 produced by pterygium head fibroblasts(Figure 8). The nature of their being a genuine MMP was further confirmed by enzyme inhibitors (Figure 7). Expression of MMP-1 by cultured fibroblasts is mediated by activating an interleukin 1α autocrine feedback loop.²⁹ Therefore, future studies are needed to address whether overexpression of MMP-1 and MMP-3 by pterygium head fibroblasts is indeed a phenotype in vivo or an artifact created by cell culture. Furthermore, it may be informative to investigate whether pterygium head fibroblasts may have a defect leading to the activation of this interleukin 1α autocrine feedback loop.

MMP-1, an interstitial collagenase, can degrade native fibrillar collagen types I, II, III, IX, and XI.^{10 - 11 ,30} MMP-3, or stromelysin-1, has a broad substrate specificity that includes casein, proteoglycans, fibronectin, elastin, laminin, as well as collagen types III, IV, V, IX, and IX.^{9 - 11 ,31} Cooperative actions of MMP-1 and MMP-3 further augment the final proteolytic action. Therefore, it is conceivable that overproduction of MMP-1 and MMP-3 relative to their TIMPs facilitates the invasion of head fibroblasts into the cornea by degrading the basement membrane, the Bowman membrane, and the superficial corneal stroma, ie, histopathological findings well recognized in pterygia.⁷ Such higher expression of MMP-1 and MMP-3 with no change in TIMPs has also been proposed as the basis for tumor cell metastasis,¹² where cell invasion is a common feature.^{32 - 33}

Compared with head pterygium fibroblasts, body fibroblasts and subconjunctival fibroblasts adjacent to the pterygium expressed gradual decreases of MMP-1 and MMP-3 protein levels (Figure 4and Figure 5) and activities (Figure 7) without notable changes in TIMP-1 and TIMP-2. It has been reported that MMP-1 and TIMP-1 proteins are expressed by human conjunctival fibroblasts in in vivo wound healing and in culture.³⁴ Herein we noted that normal human conjunctival fibroblasts express additional MMP-3 and TIMP-2. In addition, both MMP-1 and MMP-3 were also expressed by cultured corneal fibroblasts, and the levels of their transcripts and proteins were higher than those of conjunctival fibroblasts (Figure 2 and Figure 4). It has been reported that MMP-1 and MMP-3 proteins are produced by explant cultures of human corneas³⁵ and rabbit corneal fibroblasts,³⁶ and that MMP-1 and MMP-3 are not expressed by normal uninjured corneal fibroblasts but are expressed during remodeling of corneal stroma wounds in rabbits.³⁷ Because there exists a gradient of decreasing MMP-1 and MMP-3 expression from the head to the normal conjunctiva, resembling that from the repairing corneal stroma,³⁷ we speculate that the primary abnormality of pterygium exists in the head region (also see Figure 1). Because the level of such expression by pterygium subconjunctival fibroblasts was similar to that produced by normal conjunctival fibroblasts, we also speculate that the mass built up in adjacent subconjunctival fibroblasts of a pterygium is a secondary effect. Future studies are needed to understand how the pterygium itself attracts and drags the surrounding normal conjunctiva into the well-known"wing-shaped" growth during the process of corneal invasion.

Expression of MMP-2 has been reported in human,^{24 ,35} rabbit,^{27 ,36} and rat ²⁸ corneal fibroblasts. Herein, we noted that the expression of MMP-2 transcript and protein and an as yet unknown 70-kd serine proteinase was not changed among different types of fibroblasts (Figure 2, Figure 3, and Figure 7). The finding the MMP-2 expression was unaltered is consistent with the view that MMP-2 expression tends to be constitutive and is thought to perform a surveillance function.²⁷ This unique feature is due to the unusual promoter structure of MMP-2, which does not have a TATA box or AP-1 elements commonly found and critical for gene activation in the promoters of MMP-1, MMP-3, and other inducible MMPs.⁹ With participation of the constitutively expressed serine proteinase and MMP-2, ie, gelatinase A, which can digest types IV, V, and VII collagens and denatured fibrillar collagens,^{10 - 11} corneal invasion by pterygium head fibroblasts can be enhanced.

Expression of MMP-9 transcript and protein was not detected in all the types of cultured fibroblasts tested (not shown). In rabbits, MMP-9 is expressed by corneal stromal fibroblasts and epithelial cells after wounding.²⁷ In rats, however, MMP-9 is expressed by migrating basal epithelial cells but not by corneal fibroblasts.²⁸ Therefore, our human data resemble those of rats but not those of rabbits. Recently, a preliminary report showed that protein expression of MMP-2 and MMP-9 is higher in pterygial tissues.³⁸ The discrepancy in MMP-9 expression of the latter finding may be explained by their inclusion of epithelial cells in homogenized pterygium samples.

Damage to skin collagen and elastin, leading to the pathologic sign of "elastotic degeneration," is the hallmark of long-term exposure to solar UV irradiation in photo-aged skin.³⁹ A similar finding is also observed in the pterygium.⁴⁰ Such damage in the dermis has long been regarded as a result of proteolytic actions on the connective tissue extracellular matrix. This hypothesis has been supported by experimental data showing that cultured normal dermal fibroblasts increase expression of MMP-1^{41 - 44} and MMP-3⁴⁴ after UV irradiation. Because normal cultured dermal fibroblasts were used in these dermatological studies, overexpression of MMP-1 and MMP-3 has been regarded as a phenotype extrinsically induced by UV. Extending from these dermatological data, our report reveals for the first time that such an altered phenotype is still retained in fibroblasts grown out of the pterygium head and to a lesser extent out of the pterygium body, suggesting that chronic UV irradiation may have induced an intrinsic genotypic change. Future studies to explore the mechanism leading to such intrinsic overexpression of MMP-1 and MMP-3 should allow us to map pathways linking long-term UV exposure with the pterygium formation.

The histopathological feature of pterygium resembles that seen in corneal diseases with limbal epithelial stem cell deficiency,^{45 - 46} leading one to speculate that limbal stem cells may have been destroyed in pterygium. This hypothetical view has been suggested by theoretical calculation of the albedo (indirect) light projected from the temporal sclera. This light source is concentrated and focused at the nasal limbus.⁴⁷ Supporting this view are the findings that pterygial-basal epithelial cells expressed an oncogene, p53,^{48 - 49} and vimentin,⁵⁰ which is normally found in mesenchymal cells and migrating epithelial cells, and that pterygia frequently coexist with conjunctival intraepithelial neoplasia or carcinoma.⁴⁹ Recently, overexpression of MMP-1 and MMP-2 has been noted in pterygial epithelial cells.⁵¹ This new piece of data, along with those reported herein, suggests that both epithelial cells and stromal fibroblasts may contribute to the dissolution of the basement membrane and the Bowman layer. Future studies are also needed to discern the relative contribution between pterygial epithelial cells and fibroblasts and whether interactions between these 2 cell types may actually activate fibrovascular invasion into the cornea.

Although not directly relevant to this work, we have previously reported that expression of a smooth muscle actin, shown by Northern hybridization, is higher in serum-free condition (DMEM + ITS) than in serum-containing medium(DMEM + 10% FBS), and is dramatically up-regulated by transforming growth factor-β1, but suppressed by amniotic membrane in cultured human corneal and limbal fibroblasts.⁵² A similar finding was also noted in cultured human conjunctival and pterygium body fibroblasts.⁵³

Accepted for publication June 22, 2000.

This work was supported in part by Public Health Service Research Grant#EY 06819 (SCGT) from Department of Health and Human Services, National Eye Institute, National Institutes of Health, Bethesda, Md, and in part by an unrestricted grant from Research to Prevent Blindness, Inc, New York, NY.

Presented in part as an abstract at the annual meeting of the Association for Research in Vision and Ophthalmology, May 10, 1999, Fort Lauderdale, Fla.

Corresponding author and reprints: Scheffer C. G. Tseng, MD, PhD, Bascom Palmer Eye Institute, William L. McKnight Vision Research Center, 1638 NW 10th Ave, Miami, FL 33136 (e-mail: stseng@bpei.med.miami.edu).