Angiostatin Inhibits and Regresses Corneal Neovascularization FREE

CORNEAL neovascularization is a central feature in the pathogenesis of many blinding corneal disorders and a major sight-threatening complication in corneal infections, chemical injury, and following keratoplasty, in which neovascularization adversely affects corneal graft survival.¹ Antiangiogenic molecules have been shown to inhibit corneal neovascularization. However, no biologic agent has been shown to regress established corneal neovascularization, a more clinically relevant end point. Thermal laser or photodynamic therapy induces only temporary closure of new vessels² and does not address the underlying biological mechanisms of neovascularization.

Angiostatin, an endogenous cleavage fragment of plasminogen, is secreted by a form of Lewis lung carcinoma (LLC) and inhibits basic fibroblast growth factor–induced corneal neovascularization in mice.³ Its antiangiogenic potential has been verified in numerous tumor models.⁴ We investigated the potential of an angiostatin-producing tumor to inhibit and regress corneal neovascularization in a pathophysiologically relevant murine model. In a confirmatory experiment, we also investigated the ability of pure recombinant angiostatin delivered systemically via an osmotic pump to regress corneal neovascularization in the same model.

All animal experiments were approved by the Massachusetts Eye and Ear Infirmary (Boston) animal care committee and conformed to the Association for Research in Vision and Ophthalmology guidelines for animal use. Male C57BL/6J mice (Jackson Labs, Bar Harbor, Me) were anesthetized by intramuscular injection of 50 mg/kg of ketamine hydrochloride and 10 mg/kg of xylazine. Animals were killed by a lethal dose of pentobarbital (150 mg/kg).

Three groups of mice underwent corneal injury. One group (referred to hereafter as control) remained tumor free and the other 2 were implanted with LLC tumor cells (either the high-metastatic [HM] or low-metastatic [LM] clones) subcutaneously the day before, 2 weeks after, or 4 weeks after injury by an investigator masked to the clone type. Corneas were harvested 2 weeks after tumor implantation (at 2, 4, and 6 weeks after injury for tumor-free mice). In a separate regression experiment, angiostatin pumps were implanted 2 weeks after corneal injury. Corneas were harvested 2 weeks after implantation and compared with corneas from control mice (implanted with pumps containing phosphate-buffered saline [PBS] alone) harvested 4 weeks after corneal injury. Therefore, the time of treatment with angiostatin was 2 weeks in all experimental groups. Neovascularization was quantified in all corneas by immunostaining, using an image analyzer by a masked investigator.

Topical proparacaine and 2 µL of 0.15M sodium hydroxide were applied to the right cornea of each mouse. The corneal and limbal epithelia were removed using a Tooke corneal knife (Arista Surgical Supply, New York, NY) in a rotary motion parallel to the limbus. Erythromycin ophthalmic ointment was instilled immediately following epithelial denudation.

Low-metastatic LLC tumors were passaged in vivo³ and HM-LLC cells were passaged in vitro.⁵ Recipient mice had their backs shaved, and the recipient site was cleansed with povidone-iodine and ethanol. The subcutaneous dorsa of the mice in the proximal midline were injected with 0.1 mL containing 10⁶ cells of the respective tumor cell lines.

Murine angiostatin was produced as a fusion protein with the murine immunoglobulin 2a Fc fragment (mFc-m-angiostatin) in a murine myeloma cell line as described previously.⁶ The protein was diluted in PBS, filtered through a Millipore filter (Millipore, Bedford, Mass), and stored at −20°C until used. The purity of the mFc-m-angiostatin ranged between 90% and 95% (data not shown). The murine recombinant angiostatin was continuously delivered with a mini-osmotic pump. Mini-osmotic pumps with a pump rate of 1.0 µL per hour were implanted into the intraperitoneal cavity 2 weeks after limbal injury (n = 6 animals). The total delivered dose of angiostatin was 10 mg/kg per day. After 8 days, the pumps were exchanged and animals were killed 2 weeks after the first pump was implanted. Control animals received pumps delivering equal amounts of PBS (n = 7 animals).

Immunohistochemical staining for vascular endothelial cells was performed on corneal flat mounts. Fresh corneas were dissected, rinsed in PBS for 30 minutes, and fixed in 100% acetone for 20 minutes. After washing in PBS, nonspecific binding was blocked with 0.1M PBS, 2% albumin for 1 hour at room temperature. Incubation with fluorescein isothiocyanate conjugated–coupled monoclonal antimouse CD31 antibody at a concentration of 1:500 in 0.1 M of PBS, 2% albumin at 4°C overnight was followed by subsequent washes in PBS at room temperature. Corneas were mounted with Gelmount (Biomeda Inc; San Francisco, Calif), an antifading agent, and visualized with a fluorescent microscope.

Digital quantification of corneal neovascularization has been previously described.^{7 - 8} Images of the corneal vasculature were captured using a CD-330 charge-coupled device camera attached to a Leica MZ FLIII fluorescent microscope (Leica Microsystems Inc, Deerfield, Ill). The images were analyzed using Openlab software (Improvision Inc, Lexington, Mass), resolved at 624 × 480 pixels, and converted to tagged information file format files. The neovascularization was quantified by setting a threshold level of fluorescence above which only vessels were captured. The entire mounted cornea was analyzed to minimize sampling bias. The quantification of the neovascularization was performed in masked fashion. The total corneal area was outlined using the innermost vessel of the limbal arcade as the border. The total area of neovascularization was then normalized to the total corneal area and the percentage of the cornea covered by vessels calculated.

All euthanized animals that had been implanted with a tumor underwent autopsy to identify metastases. The weights of the animals were measured using an AE 200 microbalance (Mettler, Toledo, Ohio).

Differences in areas of corneal neovascularization were analyzed with the Wilcoxon (Mann-Whitney) test and the Kruskal-Wallis test.

There were no significant differences in corneal neovascularization between corresponding groups of control mice and mice implanted with HM-LLC(all P values >0.1). There was no significant spontaneous regression of corneal neovascularization in tumor-free mice (P = .87).

Mean percentages of neovascularized cornea area in mice 2 weeks after corneal scraping and implantation are shown in Figure 1. The mean percentages of neovascularized corneal area in mice 2 weeks after denudation in control mice, mice implanted with LM-LLC, and mice implanted with HM-LLC were 62.0%, 4.6%, and 45.4%, respectively. Thus, mice implanted with LM-LLC had 89.9% less neovascularized corneal area than those implanted with HM-LLC (P = .01) and 92.6% less neovascularized corneal area than control mice (P= .003).

Mean percentages of neovascularized corneal area in mice 2 weeks after tumor implantation and 4 or 6 weeks after corneal scraping are shown in Figure 2. The mean percentages of neovascularized corneal area in mice 4 weeks after denudation in control mice, mice implanted with LM-LLC, and mice implanted with HM-LLC were 68.9%, 3.7%, and 90.1%, respectively. Thus, mice implanted with LM-LLC had 95.9% less neovascularized corneal area than those implanted with HM-LLC (P = .03) and 94.6% less neovascularized corneal area than control mice (P= .03).

The mean percentages of neovascularized corneal area in mice 6 weeks after denudation in control mice, mice implanted with LM-LLC, and mice implanted with HM-LLC were 59.3%, 37.0%, and 80.3%, respectively. Thus, mice implanted with LM-LLC had 53.9% less neovascularized corneal area at 6 weeks after corneal scraping than those implanted with HM-LLC (P = .005) and 37.6% less neovascularized corneal area than control mice (P = .06). Representative photographs are shown in Figure 3.

Mean percentages of neovascularized corneal area in mice 2 weeks after pump implantation and 4 weeks after corneal scraping are shown in Figure 4. Mice with angiostatin pumps inserted had 37.7% less neovascularized corneal area than mice with sham pumps (P = .007).

No animal that had been implanted with LM-LLC had visceral metastases, whereas all animals implanted with HM-LLC were noted to have liver and/or lung metastases. The mean ± SD weights of LM-LLC mice at 2, 4, and 6 weeks after implantation were 16.8 ± 0.6 g, 19.1 ± 0.3 g, and 21.8 ± 0.6 g, respectively. The mean ± SD weights of HM-LLC mice at 2, 4, and 6 weeks after implantation were 16.5 ± 0.5 g, 19.3± 0.3 g, and 22.0 ± 0.4 g, respectively (all P values >0.1).

These data demonstrate that the LM clone of LLC, an angiostatin-producing tumor, can inhibit and regress corneal neovascularization induced by corneal injury secondary to mechanical and alkali trauma. In contrast, HM-LLC, a spontaneous variant of the LM-LLC tumor that is unable to generate angiostatin, was unable to inhibit angiogenesis or regress neovascularization. These data were confirmed in experiments showing the regression of corneal neovascularization induced by recombinant angiostatin delivered systemically via an osmotic pump. To our knowledge, this is the first report of biologically induced regression of corneal neovascularization as well as the first direct demonstration in any tissue of regression of blood vessels likely mediated by angiostatin.

The results indicate near complete inhibition of corneal neovascularization by LM-LLC (in accordance with previous data) and near-complete regression of corneal neovascularization when LM-LLC was implanted 2 weeks after corneal injury. The regressive effect was substantially less when LM-LLC was implanted 4 weeks after injury. This may indicate that the more mature vessels are less amenable to the ability of angiostatin to induce regression or they may require further exposure to angiostatin to regress. This was not addressed in our study, as mice generally died of tumor burden from LM-LLC between 18 and 24 days, and supplies of recombinant angiostatin pumps did not permit longer-term or varied dose-effect experiments. The mechanisms of the regression of the new vessels are not fully characterized. Down-regulation of vascular endothelial growth factor can induce the regression of capillaries not covered by pericytes in the corpus luteum of ovaries, while increased levels of angiopoietin-2 destabilize mature capillaries and induce their regression.⁹ It would be valuable to determine to what extent angiostatin is capable of the 2 effects.

Angiostatin inhibits the development of basic fibroblast growth factor– and vascular endothelial growth factor–induced corneal neovascularization in mice.³ Other studies have shown that thalidomide, curcumin, integrin antagonists, cyclooxygenase inhibitors, prolactin, octreotide, herbal extracts, matrix metalloproteinase inhibitors, angiostatic steroids, thrombospondin-2, kringle 1-3, and beta-cyclodextrin tetradecasulfate all exert inhibitory effects on the development of corneal neovascularization induced by various methods.^{10 - 23} Our study demonstrates the ability of angiostatin to inhibit neovascularization in a clinically relevant model of mechanical and chemical corneal trauma. The method of neovascularization induction affects pharmacological efficacy, eg, integrin antagonists inhibit basic fibroblast growth factor–induced corneal neovascularization but not that caused by chemical injury.¹⁰ Moreover, this study demonstrates for the first time, to our knowledge, the biologically induced regression of corneal neovascularization. The only other interventions demonstrated to affect established corneal neovascularization have been laser photocoagulation, fine-needle diathermy, and photodynamic therapy.^{24 - 28} However, these methods do not induce biological vessel regression, as was observed with angiostatin. Angiostatin's ability to inhibit blood vessels on the basis of tumor regression and endothelial apoptosis has been demonstrated in culture^{4 ,28} ; our study provides a direct in vivo correlate.

The antiangiogenic effect observed in animals implanted with LM-LLC, as opposed to HM-LLC, has been localized to angiostatin. O'Reilly et al³ purified the serum and urine of LM-LLC and HM-LLC mice with reverse-phase chromatography and gel electrophoresis and assayed different fractions for inhibition of endothelial proliferation; inhibitory activity corresponded to angiostatin. Dong et al²⁹ performed fast protein liquid chromatography and Western blot analysis and found that the antiangiogenic activity of serum from LM-LLC mice localized to angiostatin and also found that a monoclonal antibody against angiostatin blocked the antiangiogenic effect (measured by an assay of endothelial cell proliferation) of LM-LLC serum in culture. O'Reilly et al³⁰ found that blocking the production of angiostatin by LM-LLC cells in culture using antibodies to gelatinase A (required for angiostatin production) eliminated the antiangiogenic effect of these cells, and that only fractions generated by gelatinase A that contained angiostatin had antiangiogenic effects (measured by ability to inhibit endothelial cell proliferation).

The level of regression in tumor-implanted mice was far greater than that induced by the angiostatin pump. This may be owing to several factors. The concentration of angiostatin produced by the tumor may be much higher than that delivered by the pump, although it is not possible to quantitate serum or urine levels of angiostatin because of lack of commercial assays. The concentration of angiostatin in the tumor model is also variable and likely steadily rises as the tumor grows, and this variability may enhance the regressive effect.

In our study, the tumor burden was not significantly different between the LM-LLC and the HM-LLC groups. The HM-LLC clone arose as a spontaneous variant of the LM-LLC clone in mice and it did not produce angiostatin in vitro or in vivo. Therefore, the antiangiogenic effects observed herein can very likely be attributed to angiostatin. The implantable pump study results are consistent with this conclusion.

Digital quantification of corneal neovascularization, described previously^{7 - 8} using flat-mounted corneas, can potentially be confounded by the 3-dimensional nature of specimens and the density of blood vessels. The dehydration of corneas prior to analysis and the use of a microscope with a good depth of field minimize the effect of 3-dimensionality. This is especially true in the very thin mouse cornea. The density of blood vessels in corneas with a high degree of neovascularization is potentially a problem because individual vessels sometimes cannot be discerned, leading to an overestimation of vascularization as areas of clear cornea in between are not resolvable. This is countered by the effect of underestimation of overlapping vessels. Previous investigators^{7 - 8} have found the magnitude of these errors to be small and not to affect the consistency among samples.

In summary, angiostatin inhibits and regresses the corneal neovascularization induced by mechanical and alkali corneal injury. This finding could advance the management of blinding disorders, such as Stevens-Johnson syndrome, cicatricial pemphigoid, corneal allograft rejection, and corneal injury from infection, trauma, or alkali. The potential of topical or sustained-release forms of angiostatin with distribution constrained to the eye should be a next line of investigation. Topical forms of angiostatin have already been used in the rat and rabbit.^{31 - 32} Gene transfer methods generating angiostatin have been developed to halt angiogenesis in mouse tumor models,^{33 - 34} and naked plasmids have been shown to be taken up by corneal epithelium.³⁵ Future research should also determine the effect of recombinant angiostatin in this model and the molecular interactions among angiostatin, inflammatory phenomena, and neovascularization. This model can also be extended to neovascularization in other ocular tissues, including the iris, retina, and choroid.

Submitted for publication October 16, 2001; final revision received April 1, 2002; accepted April 24, 2002.

This work was supported by the American Ophthalmological Society-Knapp Testimonial Fund, Chicago, Ill (Dr J. Ambati), a Foundation Fighting Blindness Career Development Award, Owings Mills, Md (Dr J. Ambati), Deutsche Forschungsgemeinschaft Jo 324/2-1, Bonn, Germany, the Massachusetts Lions Eye Research Fund, Boston(Dr Adamis), and the Roberta Siegel Research Fund, Boston (Dr Adamis).

Corresponding author and reprints: Anthony P. Adamis, MD, Department of Ophthalmology, Massachusetts Eye & Ear Infirmary, 243 Charles St, Boston, MA 02114 (e-mail: tony_adamis@meei.harvard.edu).