Molecular Vision 2005; 11:859-868 <>
Received 3 May 2005 | Accepted 12 October 2005 | Published 20 October 2005

Differential conversion of plasminogen to angiostatin by human corneal cell populations

Debra J. Warejcka,1,2 Kimberly A. Vaughan,1,2 Audrey M. Bernstein,3 Sally S. Twining1,2

Departments of 1Biochemistry and 2Ophthalmology, Medical College of Wisconsin, Milwaukee, WI; 3Department of Ophthalmology, Mount Sinai School of Medicine, New York, NY

Correspondence to: Sally S. Twining, PhD, Department of Biochemistry, Medical College of Wisconsin, 8701 Watertown Plank Road, Milwaukee, WI, 53226; Phone: (414) 456-8431; FAX: (414) 456-6510; email:


Purpose: Maintenance of avascularity of the normal cornea and control of neovascularization during wound healing depend on a balance of angiogenic and antiangiogenic factors. The purpose of this paper is to determine the ability of corneal cells to convert plasminogen to angiostatins and to compare these products with those made by intact corneas.

Methods: RT-PCR was performed using plasminogen specific primers and the generated cDNA was sequenced. The proteins in corneal extracts, cornea conditioned medium, and medium from corneal epithelial cells, stromal fibroblasts, and myofibroblasts incubated with plasminogen were separated by SDS-PAGE and electroblotted. Western blots used monoclonal antibodies to kringles 1-3 to detect plasminogen and angiostatins. Angiostatins were isolated and tested for activity in a vascular endothelial cell proliferation inhibition assay.

Results: Plasminogen, its mRNA and angiostatins were found in human corneal tissue extracts from the epithelial, stromal, and endothelial layers and from cornea conditioned medium, but not in medium from cultured epithelial cells, stromal fibroblasts, or myofibroblasts. However, cultures of corneal epithelial cells and stromal fibroblasts were able to convert exogenously added plasminogen to angiostatins, whereas cultured myofibroblasts did not. Angiostatins of 38 and 34 kDa were found under all angiostatin generating conditions; however other angiostatins differed in size. Further, the angiostatins isolated from fibroblast culture supernatants inhibited vascular endothelial cell proliferation.

Conclusions: Conversion of plasminogen to angiostatin is cell-type dependent. Because corneal cells generate angiostatins, use of human angiostatins may be a means of treating abnormal corneal neovascularization without the risk of side effects.


Control of neovascularization in the cornea is important for maintenance of transparency under both normal and wound healing conditions. In the last few years several antiangiogenic molecules have been identified in the cornea. These include pigment epithelium derived factor (PEDF), maspin, thrombospondin, in addition to endostatin, a proteolysis product of type XVIII collagen, and angiostatin, a proteolysis product of plasminogen [1-5]. The number and variety of anti-angiogenic molecules may in itself be a reflection of the importance of their function in the cornea.

Angiostatin molecules are a group of proteolytic products of plasminogen or plasmin containing at least one intact kringle domain. Plasminogen is composed of an NH2-terminal peptide, five kringle domains, and a protease domain. Each kringle contains 3 disulfide bonds characteristic of this protein motif. Plasminogen is converted to plasmin by cleavage of the N-terminal peptide at the K77-K78 bond and cleavage of the activation loop at R561-V562. Although they contain kringles 1-5, neither plasminogen nor plasmin are anti-angiogenic.

The first angiostatin molecule identified was isolated from the urine of mice with Lewis lung carcinoma and was shown to be composed of kringles 1-4 of plasminogen [6]. In subsequent studies, other plasminogen derived products, kringles 1-5, 1-3, 2-3, and individual kringles, were shown to possess anti-angiogenic activity [7-9]. The size of the individual angiostatins depends upon the cleavage specificity of the proteases involved in angiostatin generation. Production of these angiostatin molecules from plasminogen or plasmin involves at least two proteolytic cleavages, one to remove the NH2-terminal peptide and a second within the linker regions between kringles or within kringle 5 to remove the COOH-terminal portion of plasminogen including the protease domain. Kringle 5 is sensitive to reductases that reduce disulfide bonds allowing release of the angiostatin molecules following cleavage within kringle 5 [10]. Under nonreducing SDS-PAGE conditions, 25-35 kDa angiostatins contain kringles 1-3, 38-45 kDa angiostatins are composed of kringles 1-4 and angiostatins in the range of 47-60 kDa angiostatins include kringles 1-4 plus part of kringle 5.

The in vivo role of angiostatins in controlling neovascularization in the cornea was first suggested by Kao et al. [11] and Drew et al. [12] using plasminogen knockout mice. Following either an epithelial scrape injury or photorefractive surgery on corneas from plasminogen-deficient mice, fibrin was deposited within the cornea, an excessive inflammatory response occurred, and vessels grew from the limbus. Angiostatins prevented bFGF and VEGF induced corneal neovascularization and induced regression of newly formed corneal vessels [13].

Gabison et al. [14] more directly demonstrated the important physiological role of angiostatin in the cornea using a mouse system. Injection of anti-angiostatin antibodies and injury to the cornea resulted in the ingrowth of vessels, in contrast to corneas wounded without antibody injection. Thus, inactivation of angiostatin in the wounded cornea tipped the balance in favor of angiogenesis resulting in neovascularization.

During the corneal wound healing process, two types of cells become motile. Epithelial cells migrate from the unwounded area to form a protective monolayer of cells over the wounded area. In addition, the normally quiescent stromal keratocytes are converted to migratory fibroblasts, which move to the injured area to effect a repair and healing phase [15]. The stromal cells can also become contractile myofibroblasts as part of the wound healing process. The epithelial cells and stromal fibroblasts are involved in the initial healing while the myofibroblasts are involved in long term remodeling of the cornea.

Previously, we demonstrated the presence of plasminogen in the human cornea and its synthesis by human corneas in organ culture [5]. The level of plasminogen in 24 h organ culture medium of human corneas, dissected away from all other sources of plasminogen, is nine times that of freshly dissected corneas. Metabolic labeling studies confirmed the synthesis of plasminogen by human corneas. This suggests that plasminogen synthesis occurs in the normal cornea, and is an ongoing process. The cytokine, IL-1β stimulated plasminogen synthesis at all time points up to 24 h, suggesting that plasminogen synthesis is controlled by this cytokine. The mRNA for plasminogen was detected by RT-PCR in the epithelial layer dissected from human corneas and was confirmed by sequencing.

The level of plasminogen in the human cornea (1.1 μg/cornea or 122 nM) is high enough to produce physiologically relevant levels of angiostatin molecules. The average IC50 for inhibition of vascular endothelial cell proliferation by angiostatins is in the range of 40 nM and even lower for angiostatins containing all or part of kringle 5 [16].

We report here that angiostatins are present in extracts of each of the three layers of human corneas and in human cornea organ culture supernatant fractions. Also, we report that viable human corneal epithelial and stromal fibroblastic cells convert plasminogen into multiple active angiostatins in culture, whereas myofibroblasts do not.


Human corneas

Human corneas were obtained from the Wisconsin Lions Eye Bank (Madison, WI) or the National Disease Research Interchange within 24-48 h of death. Experiments with corneas followed the tenets of the Declaration of Helsinki (1989) of the World Medical Association. The MCW Human Research and Review Committee at the Medical College of Wisconsin approved the use of donor corneas for these experiments. The donor corneas were used for direct extraction of RNA and protein from the separated layers, for corneal organ culture and as a source of epithelial and stromal cells for cell culture.

Corneal organ and cell culture

Whole human donor corneas were cultured as previously described [5]. Briefly, the corneas, cut into eight pieces, were cultured in MEM containing antibiotics (100 units/ml penicillin and 100 μg/ml streptomycin). The corneas were preincubated for 3 h in MEM plus 5% LPS free BSA and then cultured without or with IL-1β at 3 ng/ml. The conditioned medium was removed at 0, 2, 4, 8, or 24 h following addition of the cytokine.

The epithelial layer was released from the posterior portion of the cornea using an overnight dispase treatment at 4 °C. The epithelial layer was then treated with trypsin-EDTA at 37 °C with intermittent resuspension to obtain a single cell suspension. The cells were cultured in 24 well Primaria plates (Becton, Dickinson and Co., Franklin Lake, NJ) in Keratinocyte Medium containing defined keratinocyte serum free medium growth supplement (Invitrogen Corp., Grand Island, NY) until confluent.

For corneal stromal cells, the endothelial and epithelial layers were scraped from the donor corneas and the stromal cells were released by collagenase digestion as previously described by Taylor et al. [17]. The cultured stromal cells were grown in flasks (Costar, Cambridge, MA) and maintained in high glucose DMEM supplemented with 1% L-glutamine (Invitrogen) with 5% defined FBS (Hyclone, Logan UT), 0.1% MITO+ serum extender (Becton Dickinson) and 10 μg/ml ciprofloxacin (Bayer, Kankakee, IL) at 37 °C in 5% CO2. Fibroblastic and myofibroblastic phenotypes were generated from the stromal cell population as described by Jester and Ho-Chang [18]. Briefly, the stromal cells were seeded onto type I collagen-coated (Cohesion Technologies, Inc., Palo Alto, CA) wells of a 24 well Primaria plate (Becton Dickinson) in DMEM containing 1% RPMI vitamin mix, 100 μM nonessential amino acids, 1 mM pyruvate, and 100 μg/ml ascorbic acid. For fibroblasts, 10 ng/ml FGF-2 (Sigma Chemical Co., St. Louis, MO) was added to the media. For myofibroblasts, 1 ng/ml TGFβ1 (Sigma) was added to the media. Cells were maintained for seven days in growth factor-containing media before beginning any experiment. The conversion of stromal cells to fibroblasts and myofibroblasts was confirmed by immunohistochemistry using cells grown on coverslips and stained with TRITC labeled phalloidin for assembled f-actin, FITC-labeled anti-human α-smooth muscle actin and, Hoescht 33258 for nuclear staining (all reagents from Sigma). The addition of plasminogen to either the fibroblasts or the myofibroblasts did not alter the phenotype of the cells as judged by the absence or presence of assembled α-smooth muscle actin.

Myofibroblasts, fibroblasts, or epithelial cells were cultured in 24 well plates in serum free Keratinocyte Medium (Invitrogen) for epithelial cells, or serum-free medium containing either FGF-2 for stromal fibroblasts or TGF-β for stromal myofibroblasts [18]. Glu-plasminogen (Haematologic Technologies Inc., Essex Junction, VT) was added to the culture medium at a final concentration of 33 nM. At 24, 48, and 72 h, media from triplicate wells was collected and centrifuged at 13,000x g for 10 min and frozen at -20 °C. In some studies, 24 h fibroblast conditioned medium was substituted for the fibroblasts.

Western blots of corneal proteins

Corneas were dissected by cutting away the limbus and scraping the epithelial and endothelial layers from the stroma with a Gill knife. The stromal layer was minced with scissors. The scraped epithelia and endothelial layers and the minced stromal layer were individually broken apart using a Dounce homogenizer or Lysing Matrix D Beads (Qbiogene, Irvine, CA) in 100 mM Tris buffer, pH 7.6 containing 154 mM NaCl and a protease inhibitor cocktail (2 mM PMSF, 1 mM EDTA, 1 mM iodoacetamide and 5 μg/ml leupeptin from Sigma Chemical Co., St. Louis, MO). After centrifugation at 13,000x g for ten min, protein concentrations of the supernatant fractions were determined using the BioRad Protein Reagent (BioRad, Hercules, CA).

The proteins present in the supernatant fractions from extracts of the three corneal layers or conditioned medium from stromal and epithelial cell or corneal organ cultures were separated on SDS polyacrylamide gels under nonreducing conditions. The proteins were transferred to nitrocellulose membranes (BioRad). The blots were blocked with 10% (w/v) nonfat dry milk (Nestle USA, Inc., Solon, OH), washed and probed as previously reported [19]. The primary mouse monoclonal antibodies were specific for the K1-3 or the K4 portion of plasminogen (American Diagnostica, Greenwich CT and Enzyme Research Laboratories, South Bend IN). The secondary antibody was horseradish peroxide conjugated goat antimouse IgG (BioRad). The specific bands were visualized using a luminol based chemiluminesence system (ECL, Amersham, Arlington Heights IL). Relative molecular weights were calculated relative to molecular weight standards (BioRad).

Because the proteins were separated under nonreducing conditions, the assigned apparent molecular weights based on a molecular weight standard are not the same as those obtained under reducing conditions.


Samples were separated on polyacrylamide gels containing 0.2% casein (for plasmin detection), 0.2% casein and 16 μg/ml plasminogen (for plasminogen activator detection) or 1 mg/ml gelatin (for MMP detection). Following electrophoresis, gels were soaked in 2.5% Triton X-100 for one h, rinsed in water then incubated in casein buffer (50 mM Tris pH 8, 5 mM CaCl2, 0.02% azide) or gelatin buffer (50 mM Tris pH 8.0, 150 mM NaCl, 10 mM CaCl2, 1 μM ZnCl2, 0.02% azide) at 37 °C for 5 h. The buffer was poured off and the gels were stained with Coomassie Blue, and then destained to visualize the cleared bands where the substrate was cleaved by proteases.

RT-PCR of corneal mRNA for plasminogen

Cultured human corneal epithelial cells, stromal fibroblasts, stromal myofibroblasts, and the three separated layers of the cornea were individually placed in Trizol (Invitrogen) and the tissues were disrupted using a Dounce homogenizer. RNA extractions were performed using the manufacturer's protocol with the high proteoglycan content modification. RT was carried out using random hexamers and Superscript II (Invitrogen). The touchdown PCR method was performed with Taq Polymerase HiFi (Invitrogen) using annealing temperatures ranging from 56 to 45 °C. Plasminogen specific primers that bind to exon 16 (5'-GAA TCT CGA ACC GCA TGT-3') and exon 19 (5'-AAG ACC CCA AGA AGT GAC-3') were used to amplify a 418 bp product. PCR products were sequenced by the Protein and Nucleic Acid Facility at the Medical College of Wisconsin (Milwaukee, WI).

Isolation of angiostatin molecules from conditioned medium of cells incubated with exogenous plasminogen

Medium (10 ml) from corneal stromal cells incubated in the presence of 15 μg/ml plasminogen in DMEM plus 1% glutamine for four days was used for angiostatin isolation. The medium was passed over a 1 ml Lysine-Sepharose column (Amersham Biosciences, Uppsala, Sweden). The column was washed sequentially with 10 ml 50 mM phosphate buffer, pH 7.4, 5 ml of the same buffer plus 50 mM NaCl, 5 ml of the same buffer plus 500 mM NaCl and 5 ml of the initial buffer. The specifically bound products were eluted with three 1 ml aliquots of 200 mM ε-aminocaproic acid (EACA). Plasminogen and plasmin were removed by a 50 kDa spin concentrator and EACA was removed by dialysis with 50 mM phosphate buffer and then with DMEM plus 1% L-glutamine. The solution was passed through a 0.22 μm filter. Protein content was determined using the BioRad Protein Assay (BioRad, Hercules, CA).

Vascular endothelial cell proliferation assay

Human umbilical vascular endothelial cells were obtained from BioWhittaker (San Diego, CA). The cells were trypsinized and replated in gelatinized wells at 2,000 cells/well in high-glucose DMEM with 1% L-glutamine and 5% FBS in 48 well plates. At 24 h, the initial medium was exchanged for the same medium with or without fibroblast growth factor-2 (FGF-2) at 60 pM and in the presence or absence of corneal cell produced angiostatin at 32, 320, or 3,200 ng/ml. The cells were cultured for 72 h and the number of cells was quantified using CYQUANT GR according to the manufacturer's protocol (Molecular Probes, Inc., Eugene, OR). The results were confirmed in an independent experiment. Statistical significance was determined using a one way ANOVA for overall differences and the Tukey Test for individual comparisons with the program SigmaStat (SPSS, Chicago, IL).


The sizes of angiostatins varies with the donor tissue and the corneal layer being examined

To determine whether the three layers of the human cornea contain plasminogen and angiostatins, protein extracts were isolated in the presence of protease inhibitors and analyzed on western blots. Figure 1 shows the results of a typical western blot probed with a mixture of two monoclonal antibodies that react specifically with the K1-3 portion of the human plasminogen molecule. A band running at 87 kDa represents full-length plasminogen and was observed in all epithelial, endothelial, and stromal extracts. Smaller molecular weight bands running at 38-49 kDa and 28-34 kDa are consistent with the reactivity and size of molecules containing K1-4 and K1-3, respectively. The levels of individual angiostatin molecules depended upon the individual donor and the layer of the cornea being examined. All layers contained bands at 49, 38, and 34 kDa. In the epithelial and endothelial layers, the 49 kDa angiostatin was the major band, while in the stromal extracts, the 34 kDa band represented the major angiostatin. The 45 kDa angiostatin band was observed in the stromal extracts and weakly in the endothelial extracts. The endothelial extracts contained a 30 kDa form of angiostatin.

Corneal plasminogen is converted to angiostatin in organ culture

Since the limbal area of the cornea contains blood vessels that may be a source of plasminogen, we isolated the cornea in an organ culture system to determine if plasminogen produced by the cornea could be converted into angiostatin. To simulate a wound healing system, the cornea was cut into eight pieces and then placed in organ culture. Conditioned medium taken from human corneal organ cultures were separated on SDS-PAGE, electroblotted, and the membranes probed with antibodies specific for the K1-3 portion of the plasminogen molecule. The blot in Figure 2A shows that the corneal culture medium contains both plasminogen and multiple angiostatins which accumulate in the media between 3 and 27 h. The angiostatins produced were in the range of 28-55 kDa. These results were confirmed in cultures of corneas from two additional donors.

Our previous work showed IL-1β increased secreted plasminogen levels in corneal organ culture medium relative to the control [5]. To determine whether plasminogen conversion to angiostatins is also increased, western blots of conditioned medium from corneas incubated without or with IL-1β treatment were probed with antibodies to K1-3 (Figure 2B,C). In the presence of IL-1β, the 34-45 kDa angiostatins decreased while the 28 kDa angiostatin increased at 5 h (Figure 2C). In contrast, at earlier times, 0 and 2 h, and later times, 8 and 24 h, no differences were noted in the angiostatins present or their amounts in the organ culture medium. This would suggest that one or more proteases are specifically upregulated at 5 h in the presence of IL-1β. Neither the level of matrix metalloproteinases (Figure 3A) nor urinary-type plasminogen activator (uPA, Figure 3B) differed significantly at 5 h between the control and the IL-1β treated corneas. Although plasmin activity was increased at 5 h with IL-1β treatment relative to the control, it was also increased at 8 and 24 h (Figure 3B,C) where no significant differences were noted in the angiostatin patterns (Figure 2B,C). The same results were observed for organ cultures of corneas from 4 other donors. These results indicate that the protease which converts larger angiostatins to the 28 kDa form is not a matrix metalloproteinase as observed by gelatin zymography, a plasminogen activator as observed by casein/plasminogen zymography, nor plasmin or other caseinase as observed by casein zymography.

Plasminogen mRNA is present in corneal tissue extracts but not in cultured corneal cells

To explore whether cells in separated layers of normal human corneas and cells cultured from the cornea synthesize mRNA for plasminogen, RNA was prepared directly from the three layers of human donor corneas and from cultured primary epithelial cells, stromal fibroblasts, and myofibroblasts. RT-PCR was performed on the extracts using plasminogen specific primers from exons 16 and 19. In addition, protein was extracted from the three layers of the cornea and from cultured corneal cells. The plasminogen message and plasminogen protein were found in all three layers of the cornea (Figure 4). The sequences of the cDNA products were identical to that for human plasminogen. In contrast, no RT-PCR products were obtained using the RNA from cultured cells (Figure 4A). In neither was the plasminogen protein found in the conditioned medium (Figure 4B), indicating that once in culture corneal cells lose their ability to synthesize plasminogen.

Human corneal epithelial cells in culture convert exogenously added plasminogen to multiple forms of angiostatin

Because plasminogen is a secreted molecule, conversion to angiostatin probably occurs either on the cell membrane or in the extracellular matrix. Although plasminogen is not synthesized by corneal cells in culture (Figure 4), we investigated whether or not the enzymes required to process plasminogen were still expressed. Exogenous human plasminogen was added at a final concentration of 33 nM. This concentration is within physiological levels found in the normal cornea (122 nM) [5].

Intact primary human corneal epithelial cells in culture generated multiple angiostatin-like molecules from plasminogen that react with the K1-3 antibodies (Figure 5). By 24 h, the exogenously added plasminogen was nearly gone from the culture supernatants, and at 48 h, all of the plasminogen was converted to smaller molecular weight fragments (Figure 5A). In the absence of cells, plasminogen remained a full-length molecule (Figure 5A; control, far right lane). Five anti-K1-3 reactive products were observed in the molecular weight range of 34-52 kDa representing K1-4 and K1-3 forms of angiostatin. The concentrations of the 47 and 52 kDa angiostatin forms decreased at 48 and 72 h (Figure 5B). The smaller molecular weight K1-3 bands at 38 and 34 kDa accumulated over time (Figure 5A). Thus, cultured epithelial cells still express the proteases needed to convert plasminogen into angiostatins.

Human corneal stromal fibroblasts convert plasminogen to multiple forms of angiostatin

Because we detected significant amounts of plasminogen and angiostatin in the extracts of human corneal stroma layers (Figure 1), we wondered whether corneal stromal cells in culture would convert plasminogen to angiostatin. Human corneal stromal cells in the wound healing fibroblastic phenotype were maintained on a type I collagen-coated surface in a serum-free media containing FGF-2. These cells contained the characteristic large actin fibers that bound FITC-phalloidin, but were nonreactive with antibodies to α-smooth muscle actin (data not shown). Upon addition of plasminogen (33 nM final) to these cultures, the cells converted plasminogen to angiostatin-like fragments (Figure 6; FIB lanes). By 48 h, only trace amounts of full-length plasminogen were detected in the supernatants of the cultures. Fragments corresponding to K1-4 and K1-3 were generated. By 72 h, the amount of K1-4 angiostatins at 47 and 52 kDa had decreased while the amounts of 43 kDa K1-4 and the 34 kDa K1-3 angiostatins had increased. Not only was plasminogen converted to angiostatins in the presence of fibroblasts, 24 h stromal fibroblast conditioned medium converted plasminogen to angiostatins of the same size as those produced in the presence of the fibroblasts plus an additional 32 kDa form (Figure 7). This indicates that proteases secreted by the fibroblasts into the medium are sufficient to produce angiostatins.

Human corneal myofibroblasts do not readily convert plasminogen to angiostatin

During wound healing some corneal fibroblasts are converted to myofibroblasts, which contract to pull the wound edges together during the healing process [20]. These α-smooth muscle actin-expressing cells were generated by growing corneal stromal cells on a type I collagen-coated surface in serum-free medium in the presence of TGFβ1. After seven days, these corneal stromal cells expressed α-smooth muscle actin and assembled actin fibers (data not shown). Plasminogen (33 nM final) was added to these cultures and supernatants were sampled at 24, 48, and 72 h. The vast majority of the exogenously added plasminogen remained full length in these cultures (Figure 6, MYO lanes), showing no smaller angiostatin-like fragments until 72 h where only a faint band at 47 kDa was observed.

Human stromal cell generated angiostatin inhibits human umbilical cord vascular endothelial cell proliferation

Inhibition of vascular endothelial cell proliferation was chosen to test whether the corneal stromal fibroblast cell-derived plasminogen proteolysis products act as angiostatin molecules. Angiostatin-like molecules were isolated by lysine-Sepharose chromatography from the conditioned medium of corneal stromal cells incubated with plasminogen. The isolated angiostatin molecules inhibited FGF-2 stimulated cell proliferation of human umbilical vascular endothelial cells in a dose dependent manner (Figure 8). This inhibition was statistically significant for all concentrations of corneal stromal fibroblast generated angiostatins (p<0.01).

Whole corneas and cultured cells produce angiostatins of 34 and 38 kDa plus other angiostatins that differ in sizes

Angiostatins of 38 (K1-4) and 34 (K1-3) kDa were observed in the epithelial, stromal, and endothelial layer extracts, the corneal organ culture medium, and the supernatants of cultured epithelial cells and stromal fibroblasts (Figure 9). Three larger K1-4 containing bands at 52, 47, and 43 kDa produced by the corneal epithelial cells and stromal fibroblasts were about 2 kDa smaller than the three corresponding angiostatins found in the organ culture medium and the corneal layer extracts at 55, 49, and 45. Use of K1-3 and K4 antibodies confirmed the presence of kringles 1-4 in the 38-55 kDa forms of angiostatin (data not shown). The 34 and 30 kDa bands did not react with the K4 antibody.


Plasminogen mRNA and protein are synthesized by all three layers of nonwounded corneas and yield angiostatin derivatives. The particular angiostatins present and their relative levels in a given sample depended upon the corneal layer and the individual donor. This probably reflects the presence of different sets of active proteases at different relative levels in the various corneal layers and donors. The presence of mRNA for plasminogen in all three layers of the cornea suggests that the source of this protein for angiostatin production can be the corneal cells themselves. Because maintenance of avascularity in the cornea is of primary importance, it is not surprising to find angiostatin-like molecules already formed in the normal cornea. Other anti-angiogenic molecules such as endostatin, thrombospondin, PEDF, and maspin have also been found in the nonwounded cornea reaffirming the importance of these molecules for maintaining clarity in the cornea [1-4].

It is crucial in the wounded cornea to offset a tendency toward angiogenesis that could result once the corneal fibroblasts increase their synthesis and activation of angiogenic proteases and growth factors [11,21,22]. In contrast to normal corneas, wounding of the corneas of thrombospondin- and plasminogen-deficient mice results in persistent inflammation and neovascularization [11,21], indicating that the anti-angiogenic potential of multiple molecules in the cornea is required during this time to prevent neovascularization.

Inactivation of plasminogen derived angiostatins by injection of antibodies to kringles K1-3 or K1-5 of plasminogen into the mouse cornea after laser injury results in neovascularization [14]. In contrast, injection of antibodies to the protease domain of plasminogen does not lead to neovascularization. This is probably due to an alteration in the balance of antiangiogenic molecules relative to angiogenic molecules in the cornea by anti-K1-3 and K1-5 antibodies.

Although corneal cells in the intact cornea synthesize plasminogen, once they are in cell culture, they lose their ability to synthesize this protein. This may be an artifact of the culture system due to the loss of paracrine factors produced by cells of different layers of the cornea. Alternatively, it may be a control mechanism to limit plasminogen concentrations forcing the cells to use plasminogen deposited in the extracellular matrix for plasmin and angiostatin production.

Primary cultures of human corneal epithelial cells are capable of converting plasminogen to angiostatin molecules, and these molecules accumulate in culture during the 72 h incubation time tested. Also with time, larger angiostatins are further cleaved to smaller products. These results for human epithelial cells differ from previous studies by Gabison et al. [14] using scraped mouse corneal epithelial cells. The scraped mouse epithelial cells were shown to convert plasminogen to only one angiostatin product, a 50 kDa form which did not increase over time unless the plasmin inhibitor, ε-aminocaproic acid was present. Since the epithelium in the study by Gabison et al. [14] was scraped, intracellular proteases may have degraded the angiostatins.

Human corneal stromal fibroblasts and fibroblast conditioned medium convert plasminogen to angiostatin molecules of similar sizes to those produced by the human epithelial cells. This indicates that epithelial and stromal cells synthesize and activate similar proteases in culture. Further, the fibroblasts likely secrete proteases responsible for conversion of plasminogen to angiostatins.

It is not surprising that corneal cells express the proteases necessary to generate angiostatins, as epithelial cells and stromal fibroblasts are migratory cells that express multiple proteases including matrix metalloproteinases and plasminogen activators after injury [23]. Because these same proteinases can induce angiogenesis [24], the ability of these proteases to convert plasminogen to angiostatin may be a feedback mechanism to control angiogenesis. Thus conversion of endogenous plasminogen to angiostatin-like molecules by epithelial and stromal fibroblasts after wounding, may be critical.

Our data show that the less motile corneal myofibroblasts do not process plasminogen into forms of angiostatin. TGFβ induces the conversion of the fibroblast to the myofibroblast phenotype. TGFβ has been shown to downregulate the level of two matrix metalloproteinases, collagenase, and stromelysin [25], which may be involved in processing plasminogen to angiostatin. Thus, the finding that neither myofibroblasts (Figure 6) nor their conditioned media (data not shown) show any evidence of conversion of plasminogen to plasmin activity is consistent with the current protease data for the myofibroblast phenotype. Plasmin can generate angiostatin from plasminogen [7]. Furthermore, the urokinase type plasminogen activator receptor (uPAR) is not targeted to the cell surface of myofibroblasts [26] and thus uPA is not localized to the cell surface where it is most active in the conversion of plasminogen to plasmin. Thus, reduced levels of matrix metalloproteinases and the inability of these cells to convert plasminogen to plasmin are potentially why myofibroblasts do not produce angiostatins.

The three layers of the cornea and the corneal organ culture medium contain similar angiostatins, some of which are the same size as those produced by corneal epithelial cells and stromal fibroblasts in culture. The different angiostatins may reflect the presence of distinct sets of proteases at differing concentrations in the corneal systems, one associated with normal corneas and one associated with activated cells during wound healing. In the organ culture system, the set of proteases present probably reflect a mixture of those present in the normal cornea and those present during wound healing. The epithelial cells and the stromal fibroblasts probably generate a set of proteases observed during wound healing. Angiostatins prevent neovascularization in several ocular systems including models of corneal and choroid neovascularization and proliferative retinopathy [13,27-29]. In the cornea, angiostatin inhibits angiogenin, FGF-2, and VEGF induced angiogenesis and angiogenesis induced by sodium hydroxide [13,30]. Angiostatin molecules not only inhibit the formation of vessels, they induce regression of vessels formed in the cornea in response to injury [30]. The mechanism of action of these molecules on vascular endothelial cells involves inhibition of proliferation and migration and induction of apoptosis through binding to the angiostatin receptor, ATP synthase and activating caspases [31]. Because the cornea produces angiostatins endogenously, use of human angiostatins may be a way to treat abnormal corneal neovascularization without risk of side effects.


This work was supported by RO1 EY12731 and P30 EY01931 from the National Eye Institute of the National Institutes of Health and an unrestricted grant from Research to Prevent Blindness, Inc. The authors have no financial interest in this project. Presented in part at ARVO 2005, E-2187.


1. Zhang M, Volpert O, Shi YH, Bouck N. Maspin is an angiogenesis inhibitor. Nat Med 2000; 6:196-9.

2. Dawson DW, Volpert OV, Gillis P, Crawford SE, Xu H, Benedict W, Bouck NP. Pigment epithelium-derived factor: a potent inhibitor of angiogenesis. Science 1999; 285:245-8.

3. Hiscott P, Seitz B, Schlotzer-Schrehardt U, Naumann GO. Immunolocalisation of thrombospondin 1 in human, bovine and rabbit cornea. Cell Tissue Res 1997; 289:307-10.

4. Lin HC, Chang JH, Jain S, Gabison EE, Kure T, Kato T, Fukai N, Azar DT. Matrilysin cleavage of corneal collagen type XVIII NC1 domain and generation of a 28-kDa fragment. Invest Ophthalmol Vis Sci 2001; 42:2517-24.

5. Twining SS, Wilson PM, Ngamkitidechakul C. Extrahepatic synthesis of plasminogen in the human cornea is up-regulated by interleukins-1alpha and -1beta. Biochem J 1999; 339:705-12.

6. O'Reilly MS, Holmgren L, Shing Y, Chen C, Rosenthal RA, Moses M, Lane WS, Cao Y, Sage EH, Folkman J. Angiostatin: a novel angiogenesis inhibitor that mediates the suppression of metastases by a Lewis lung carcinoma. Cell 1994; 79:315-28.

7. Cao R, Wu HL, Veitonmaki N, Linden P, Farnebo J, Shi GY, Cao Y. Suppression of angiogenesis and tumor growth by the inhibitor K1-5 generated by plasmin-mediated proteolysis. Proc Natl Acad Sci U S A 1999; 96:5728-33.

8. Cao Y, Ji RW, Davidson D, Schaller J, Marti D, Sohndel S, McCance SG, O'Reilly MS, Llinas M, Folkman J. Kringle domains of human angiostatin. Characterization of the anti-proliferative activity on endothelial cells. J Biol Chem 1996; 271:29461-7.

9. Cao Y, Chen A, An SS, Ji RW, Davidson D, Llinas M. Kringle 5 of plasminogen is a novel inhibitor of endothelial cell growth. J Biol Chem 1997; 272:22924-8.

10. Lay AJ, Jiang XM, Kisker O, Flynn E, Underwood A, Condron R, Hogg PJ. Phosphoglycerate kinase acts in tumour angiogenesis as a disulphide reductase. Nature 2000; 408:869-73.

11. Kao WW, Kao CW, Kaufman AH, Kombrinck KW, Converse RL, Good WV, Bugge TH, Degen JL. Healing of corneal epithelial defects in plasminogen- and fibrinogen-deficient mice. Invest Ophthalmol Vis Sci 1998; 39:502-8.

12. Drew AF, Schiman HL, Kombrinck KW, Bugge TH, Degen JL, Kaufman AH. Persistent corneal haze after excimer laser photokeratectomy in plasminogen-deficient mice. Invest Ophthalmol Vis Sci 2000; 41:67-72.

13. Kim JH, Kim JC, Shin SH, Chang SI, Lee HS, Chung SI. The inhibitory effects of recombinant plasminogen kringle 1-3 on the neovascularization of rabbit cornea induced by angiogenin, bFGF, and VEGF. Exp Mol Med 1999; 31:203-9.

14. Gabison E, Chang JH, Hernandez-Quintela E, Javier J, Lu PC, Ye H, Kure T, Kato T, Azar DT. Anti-angiogenic role of angiostatin during corneal wound healing. Exp Eye Res 2004; 78:579-89.

15. Mohan RR, Hutcheon AE, Choi R, Hong J, Lee J, Mohan RR, Ambrosio R Jr, Zieske JD, Wilson SE. Apoptosis, necrosis, proliferation, and myofibroblast generation in the stroma following LASIK and PRK. Exp Eye Res 2003; 76:71-87.

16. Soff GA. Angiostatin and angiostatin-related proteins. Cancer Metastasis Rev 2000; 19:97-107.

17. Taylor JL, O'Brien WJ. Interferon production and sensitivity of rabbit corneal epithelial and stromal cells. Invest Ophthalmol Vis Sci 1985; 26:1502-8.

18. Jester JV, Ho-Chang J. Modulation of cultured corneal keratocyte phenotype by growth factors/cytokines control in vitro contractility and extracellular matrix contraction. Exp Eye Res 2003; 77:581-92.

19. Twining SS, Fukuchi T, Yue BY, Wilson PM, Zhou X, Loushin G. Alpha 2-macroglobulin is present in and synthesized by the cornea. Invest Ophthalmol Vis Sci 1994; 35:3226-33.

20. Jester JV, Petroll WM, Cavanagh HD. Corneal stromal wound healing in refractive surgery: the role of myofibroblasts. Prog Retin Eye Res 1999; 18:311-56.

21. Cursiefen C, Masli S, Ng TF, Dana MR, Bornstein P, Lawler J, Streilein JW. Roles of thrombospondin-1 and -2 in regulating corneal and iris angiogenesis. Invest Ophthalmol Vis Sci 2004; 45:1117-24.

22. Wilson SE, Mohan RR, Mohan RR, Ambrosio R Jr, Hong J, Lee J. The corneal wound healing response: cytokine-mediated interaction of the epithelium, stroma, and inflammatory cells. Prog Retin Eye Res 2001; 20:625-37.

23. Fini ME, Cook JR, Mohan R. Proteolytic mechanisms in corneal ulceration and repair. Arch Dermatol Res 1998; 290:S12-23.

24. Oh CW, Hoover-Plow J, Plow EF. The role of plasminogen in angiogenesis in vivo. J Thromb Haemost 2003; 1:1683-7.

25. Fini ME, Girard MT, Matsubara M, Bartlett JD. Unique regulation of the matrix metalloproteinase, gelatinase B. Invest Ophthalmol Vis Sci 1995; 36:622-33.

26. Bernstein AM, Twining SS, Vaughan KA, Masur SK. Urokinase Pathway Regulation in Corneal Fibroblasts and Myofibroblasts. ARVO Annual Meeting; 2005 May 1-5; Fort Lauderdale (FL).

27. Meneses PI, Hajjar KA, Berns KI, Duvoisin RM. Recombinant angiostatin prevents retinal neovascularization in a murine proliferative retinopathy model. Gene Ther 2001; 8:646-8.

28. Lai CC, Wu WC, Chen SL, Xiao X, Tsai TC, Huan SJ, Chen TL, Tsai RJ, Tsao YP. Suppression of choroidal neovascularization by adeno-associated virus vector expressing angiostatin. Invest Ophthalmol Vis Sci 2001; 42:2401-7.

29. Drixler TA, Borel Rinkes IH, Ritchie ED, Treffers FW, van Vroonhoven TJ, Gebbink MF, Voest EE. Angiostatin inhibits pathological but not physiological retinal angiogenesis. Invest Ophthalmol Vis Sci 2001; 42:3325-30.

30. Ambati BK, Joussen AM, Ambati J, Moromizato Y, Guha C, Javaherian K, Gillies S, O'Reilly MS, Adamis AP. Angiostatin inhibits and regresses corneal neovascularization. Arch Ophthalmol 2002; 120:1063-8.

31. Veitonmaki N, Cao R, Wu LH, Moser TL, Li B, Pizzo SV, Zhivotovsky B, Cao Y. Endothelial cell surface ATP synthase-triggered caspase-apoptotic pathway is essential for k1-5-induced antiangiogenesis. Cancer Res 2004; 64:3679-86.

Warejcka, Mol Vis 2005; 11:859-868 <>
©2005 Molecular Vision <>
ISSN 1090-0535