Corneal angiogenic privilege: angiogenic and antiangiogenic factors in corneal avascularity, vasculogenesis, and wound healing (an American Ophthalmological Society thesis)

Abstract

Purpose: To determine the molecular basis of corneal avascularity during wound healing and determine the role of angiogenic and antiangiogenic factors in corneal vasculogenesis.

Methods: The expression of proangiogenic factors (vascular endothelial growth factor [VEGF]; basic fibroblast growth factor [bFGF]; matrix metalloproteinase-2 [MMP-2]; and membrane-type 1-MMP [MT1-MMP]) and antiangiogenic factors (pigment epithelium-derived factor [PEDF]; angiostatin; restin; and endostatin) was analyzed in avascular corneas and in models of corneal neovascularization (bFGF pellet implantation, intrastromal injection of MT1-MMP cDNA, and surgically induced partial limbal deficiency).

Results: Immunohistochemistry demonstrated the presence of antiangiogenic factors (PEDF, angiostatin, restin, and endostatin) and proangiogenic molecules (VEGF, bFGF, MMP-2, and MT1-MMP) in the cornea after wounding. Proangiogenic MMPs were upregulated in stromal fibroblasts in the vicinity of invading vessels following bFGF pellet implantation. Corneal neovascularization (NV) was also induced by intrastromal injection of MT1-MMP naked cDNA in conjunction with de-epithelialization. Partial limbal deficiency (HLD-) resulted in corneal NV in MMP-7 and MMP-3 knockout mice but not in wild type controls.

Conclusions: Corneal angiogenic privilege is an active process involving the production of antiangiogenic factors to counterbalance the proangiogenic factors (which are upregulated after wound healing even in the absence of new vessels). Our finding that the potent antiangiogenic factors, angiostatin and endostatin, are colocalized with several MMPs during wound healing suggests that MMPs may be involved in the elaboration of these antiangiogenic molecules by proteolytic processing of substrates within the cornea.

Figures

FIGURE 1

Electron micrography of normal and vascularized mouse corneas. Normal and vascularized corneas were fixed in half-strength Karnovsky fixative (A, C). The normal cornea shows uniform arrangement of collagen fibers (B, D). The vascularized cornea shows irregular collagen distribution. (epi, epithelium; asterisk (*), Descemet’s membrane)

FIGURE 2

Characterization of bFGF and VEGF in mouse bFGF-induced corneas and neovascularization. Corneal epithelial cells were scraped and lysed in SDS sample buffer. A, Western blot analysis shows a reactive band at 13 kDa for bFGF (lane 1) and 25 kDa for VEGF (lane 2) when compared to commercially available bFGF and VEGF. B, Immunohistochemistry shows VEGF (keratocyte, a) and vascular endothelial cell marker (CD31, b) expression in corneal stroma and shows co-localization in the double staining (c).

FIGURE 3

Kinetics of bFGF-induced corneal neovascularization. Pellets containing bFGF (50 ng/pellet) were implanted into mouse corneal stromata. Slit-lamp photographs were taken of corneal vessels (A, day 1; B, day 3; C, day 4; D, day 7; E, day 10; F, day 14).

FIGURE 4

Comparison of potency of VEGF and bFGF for induction of corneal neovascularization. A wild-type mouse corneal stroma was implanted with 80 ng/pellet VEGF (A) and the same concentration of bFGF (B). On day 14, immunohistochemistry revealed vascular endothelial cells near the bFGF pellet (C) and in the limbal zone (D). The presence of VEGF in the cornea was not sufficient to induce corneal NV after pellet implantation. (* indicates pellet location)

FIGURE 5

Characterization of MMP-2 and MT1-MMP in the cornea and cultured keratocytes and fibroblasts. Bovine keratocytes (A, B) and fibroblasts (C, D) were isolated and immunostained with anti-MT1-MMP (A, C) and anti-MMP-2 antibodies (B, D). MT1-MMP is localized in an unwounded cornea (E; arrow) and MMP-2 is in the stroma of a wounded cornea (F; arrow). Although these MMPs are proangiogenic, they are present in the cornea in the absence of corneal NV. (Arrowheads show the margin of the epithelium; PI was used for nucleus staining.)

FIGURE 6

Colocalization of MMP-2 and MMP-14 in vascular endothelial cells. Vascularized corneal vessels were immunostained with anti-MT1-MMP(A) and CD31 antibodies (B) and double stained with both (C). They were also immunostained with anti–type IV collagen (D) and anti-MMP-2 antibodies (E) and double stained with both (F).

FIGURE 7

Distribution of MT1-MMP, VEGF, and CD31 in mouse corneas 4 days following bFGF pellet implantation. FGF-2 pellets (50 ng) were implanted into mouse corneas. The corneas were harvested at 4 days postoperatively. The central avascular zone adjacent to the pellets and the peripheral neovascularization zone were examined. Individual sections were immunostained with MT1-MMP (A, B), VEGF (C, D), and CD31 (E, F), a vascular endothelial cell marker.

FIGURE 8

Distribution of MT1-MMP, VEGF, and CD31 in mouse corneas 14 days after bFGF pellet implantation. FGF-2 pellets (50 ng) were implanted into mouse corneas. The corneas were harvested at 14 days postoperatively. The central avascular zone adjacent to the pellets and the peripheral neovascularization zone were examined. Individual sections were immunostained with MT1-MMP (A, B), VEGF (C, D), and CD31 (E, F), a vascular endothelial cell marker.

FIGURE 9

Injection of MT1-MMP naked DNA into mouse corneas induced corneal NV and VEGF expression. LacZ mice were injected with MT1-MMP DNA at 3 and 8 days (A, B, respectively). Arrows denote VEGF expression (β-galactosidase activity). Asterisks denote the area close to the injection sites.

FIGURE 10

MT1-MMP DNA injection tips the balance towards corneal neovascularization during wound healing. Effects of limbal injury (A, D) and MT1-MMP DNA injection (B, E) and combination of limbal injury and MT1-MMP DNA injection (C, F). Combined hemilimbal injury and MT1-MMP DNA transfection resulted in corneal neovascularization.

FIGURE 11

Hemilimbal deficiency: a model for injury-induced corneal NV. Diagrams depict limbal injury (A; green) and limbal plus epithelial removal (D; purple). The nasal limbi of wild-type mouse corneas were removed and photographed at day 7 after surgery (B, C). The nasal limbus and the epithelium of WT mouse corneas were removed and the corneas photographed at day 7 after surgery (E, F). Vascularized vessels were immunostained with anti-type IV collagen (G), anti-CD31 antibodies (H), and double staining (I).

FIGURE 12

Development of the hemilimbal deficiency (HLD) models. No vessels were induced when the epithelium was removed (A). In HDL models, hemilimbal injury is combined with removal of half of the epithelium (B). In HDL+ models, hemilimbal injury is combined with entire epithelium removal (C). HLD+ models with steroid treatment (D). WT = wild-type.

FIGURE 13

Hemilimbal injury (HLD-) of corneas in wild-type, MMP-2−/−, and MMP12−/− mice. Hemilimbal injury was applied to mouse cornea in MMP-2−/− (A, B, and C), MMP12−/− (D, E, F), and wild-type mice (G, H, I) at day 7 after wounding. Corneal vessels were visualized with type IV collagen (J) and CD31 (K) and then the images were merged (L).

FIGURE 14

Hemilimbal deficiency model (HLD-) in MMP-3−/− and MMP-7−/− mice. Hemilimbal injury was applied to mouse corneas of MMP-2−/− (A, B, C) and MMP-7−/− (D, E, F) at day 7 after injury. Corneal vessels were visualized by slit-lamp photography.

FIGURE 15

Characterization of corneal NV in hemilimbal deficiency model (HLD+) in MMP-7−/− mice. Mouse corneas were injured and corneal sections were immunostained with type IV collagen (A, B). No vessels were found in the control (C, D). Vascularized MMP-7−/− mouse cornea (D, I). Corneal vessels were further characterized using an electron microscope (F, G, H). Corneal thickness was determined by ultrasound biomicrography (J).

FIGURE 16

Characterization of PEDF in corneas. Diagram of PEDF and its primary structure (amino acids) (A). A peptide based on the C-terminal region of PEDF was synthesized and used for antibody production. Immunostaining of PEDF revealed that PEDF is localized to the mouse corneal epithelium (B). Wild-type negative control (anti-PEDF antibody omitted) (E). Propidium iodine was used for nuclear staining control (C, F). Double staining (D, merging of B and C; G, merging of E and F). Recombinant PEDF was incubated with various MMPs and assayed for cleavage (H).

FIGURE 17

Characterization of corneal plasminogen, angiostatin-like molecules, and their function. The diagram depicts the structure and domain organization of plasminogen, including kringle domains and B chain (A). Six antibodies against various domains of plasminogen were generated. Using a combination of these antibodies and commercially available antibodies, expressions of plasminogen and angiostatin-like molecules in rabbit ocular tissues were detected in the aqueous humor (A), cornea (C), iris (I), lens (L), vitreous (V), retina (R), retina pigment epithelium (R’), choroid (Ch), and sclera (S).

FIGURE 18

Characterization of antiangiostatin and B chain antibodies. Plasminogen was isolated by lysine Sepharose and eluted with aminocaproic acid (A). A 90-kDa plasminogen from liver lysates was determined by Western blot analysis by anti-K1 antibody, and 50 kDa angiostatin-like fragments were visualized in liver lysate and corneal lysate. Plasminogen and angiostatin-like molecules were isolated by lysine Sepharose and eluted with various concentrations of aminocaproic acid (0.5, 0.1 M) (B). Plasminogen from liver lysate did not bind to Butesin agarose.

FIGURE 19

Induction of angiostatin-like molecules by co-culture of corneal keratocyte and epithelial cells. Lysine Sepharose isolated proteins from co-culture of corneal epithelial cells, and keratocytes were immunoblotted with anti-K1 or B-chain antibodies. The greatest level of protein identified by the anti-K1 antibody was observed in the epithelial/keratocyte co-culture control (A; lane 2). The antibody against the B chain of plasminogen did not identify the higher-molecular-weight proteins (A; lane 5). Supernatants and cell lysates of epithelial/keratocyte co-culture grown in either 0.5% or 10% FCS were purified with lysine Sepharose. The pellets were eluted with aminocaproic acid. The eluted fractions (B; lanes 1–4) and pellets (B; lanes 5–8) were blotted with anti-K1 antibody. The actively growing co-culture (10% FCS) appears to increase the level of protein that is recognized by the anti-K1 antibody (B; lanes 5, 6).

FIGURE 20

Characterization of mouse corneal collagen XV. The diagram depicts collagen XV and its C-terminal NC1 domain. The peptide based on the C-terminal region of collagen XV NC1 domain was synthesized and used for anti-collagen XV antibody production. Affinity-purified antibodies were used to localize collagen XV in the mouse corneal epithelium (B). Control with omitted primary antibody (C). Recombinant GST-XV NC1 was purified and incubated with various MMPs (D; MMP-1, -2, -3, -7, -9).

FIGURE 21

Characterization of corneal collagen XVIII. Diagram depicts collagen XVIII and its C-terminal NC1 domain. Two peptides corresponding to the C-terminal NC1 fragment were generated and used for antibody production. Immunolocalization of collagen XVIII in human corneas (B-E) and mouse corneas (F-I) using anti-NC1 antibody (B, F), antiendostatin antibody (C, G), and antihinge antibody (D, H) was performed. Negative control for collagen XVIII immunostaining was performed without primary antibody. Human cornea (E), mouse cornea (I). cDNAs of collagen XVIII and G3PDH were amplified by polymerase chain reaction, and the corresponding fragments were visualized by agarose gel electrophoresis (J). Competitive reverse transcriptase polymerase chain reaction was used to quantify the level of collagen XVIII expression (K).

FIGURE 22

Characterization of anticollagen XVIII antibodies. (A) Corneal collagen XVIII was immunoprecipitated and cleaved with active MMP-7 and analyzed by Western blot analysis. Corneal cellular lysate was immunoprecipitated with anti-NC1 antibody and cleaved without (lane 1) and with (lane 2) active MMP-7 and blotted with antiendostatin antibody. Similarly, corneal cellular lysate was immunoprecipitated with antihinge antibody, and cleaved without (lane 3) and with (lane 4) active MMP-7 and blotted with anti-NC1 antibody. A 28-kDa band was seen with MMP-7 cleavage. Heparin-isolated, 293-overexpressed NC1 fragment was blotted with antiendostatin (lane 5) and antiendostatin antibodies (lane 6), respectively. (B) Isolation of glutathione S-transferase (GST), GST-endostatin, and GST-hinge domain of collagen XVIII. GST (lane 1), recombinant GST-endostatin (lane 2) and GST-hinge of collagen XVIII NC1 fragments (lane 3) were isolated and subjected to Western blot analysis with antiendostatin antibodies.