Targeted Disruption of the FGF2 Gene Does Not Prevent Choroidal Neovascularization in a Murine Model

Abstract

Choroidal neovascularization (CNV) is the major cause of severe visual loss in patients with age-related macular degeneration. Laser treatment is helpful for a minority of patients with CNV, and development of new treatments is hampered by a poor understanding of the molecular signals involved. Several lines of evidence have suggested that basic fibroblast growth factor (FGF2) plays a role in stimulating CNV. In this study, we tested this hypothesis using mice with targeted disruption of the FGF2 gene in a newly developed murine model of laser-induced CNV. One week after krypton laser photocoagulation in C57BL/6J mice, 34 of 60 burns (57%) showed fluorescein leakage and 13 of 16 (81%) showed histopathological evidence of CNV. At 2 weeks, CNV was detected in 9 of 10 burns (90%) in which a bubble had been observed at the time of the laser treatment. Electron microscopy showed fenestrated vessels with large lumens within choroidal neovascular lesions. Two weeks after laser-induced rupture of Bruch’s membrane, 27 of 36 burns (75%) contained CNV in FGF2-deficient mice compared with 26 of 30 (87%) in wild-type control mice, a difference that is not statistically significant. This study demonstrates that FGF2 is not required for the development of CNV after laser-induced rupture of Bruch’s membrane and provides a new model to investigate molecular mechanisms and anti-angiogenic therapy in CNV.

Choroidal neovascularization (CNV) occurs in several disease processes and is a major cause of severe visual loss, particularly in patients with age-related macular degeneration. 1 Destruction of the entire CNV complex by laser photocoagulation provides some benefit to a small minority of patients in whom the fovea is not involved and the extent of the CNV can be clearly delineated, but is complicated by a high rate of recurrence and subsequent loss of vision. Surgical removal of CNV may help some patients, particularly those with ocular histoplasmosis, but it is also complicated by a high rate of recurrence and resultant loss of vision. 2 Development of pharmacological therapy for CNV would be a major advance. To achieve this goal it would be useful to have a better understanding of the pathogenesis of CNV and a model that is inexpensive and reliable.

Several models of CNV have been described, and each has provided useful information. The laser-induced model in primates has provided important knowledge concerning the natural history of CNV, the role of the retinal pigmented epithelium (RPE) in re-establishing the blood-retinal barrier (BRB), and the participation of the RPE in the scarring process; 3 it has also been used to investigate drug treatment, 4 although it is limited by its tremendous expense. Models in rats are less expensive and more feasible to use for investigating new therapies. 5,6 In addition, they have provided important information concerning altered expression of growth factors during the development of CNV. 7 Models in minipigs 8 and rabbits 9 have demonstrated that sustained elevation of basic fibroblast growth factor (FGF2) in the subretinal or suprachoroidal space is capable of stimulating CNV.

Although each of these models is useful, a model of CNV in mice would have an important additional benefit; manipulation of gene expression is more feasible in mice than in other animals, making it possible to investigate the effect of altered expression of individual genes in the retina and RPE on the development of CNV. In this study, we produced and characterized a model of CNV in mice and then applied it to mice with targeted disruption of the FGF2 gene to investigate its role in the development of CNV. A portion of this study has been published in abstract form. 10

Materials and Methods

Production and Characterization of Laser-Induced CNV in Mice

Mice used in this study were handled according to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Thirteen adult male C57BL/6J mice were anesthetized by intraperitoneal injection of 0.3 ml of ketamine hydrochloride diluted 1:10 with sterile water. The pupils were dilated with 1% tropicamide, and three burns of krypton laser photocoagulation (50-μm spot size; 0.05 seconds duration; 350 to 400 mW) were delivered to each retina using a slit lamp delivery system and a cover glass as a contact lens. Burns were performed in the 9, 12, and 3 o’clock positions of the posterior pole of the retina so that each burn could be identified postmortem and compared with respect to fluorescein angiographic and histopathological characteristics.

Fluorescein angiograms were done in some mice by taking serial fundus photographs with a TRC-50FT camera (Topcon, Paramus, NJ) after intraperitoneal injection of 0.3 ml of 1% fluorescein sodium (Alcon, Fort Worth, TX). At various times after laser treatment, mice were sacrificed and their eyes were enucleated and fixed in 2% paraformaldehyde/2% glutaraldehyde in 0.1 mol/L cacodylate buffer (pH 7.4) for 24 hours at 4°C or in the same buffer containing only 4% paraformaldehyde. The eyes fixed in paraformaldehyde were embedded in paraffin for another study. The eyes embedded in paraformaldehyde and glutaraldehyde were dissected, each burn was identified and isolated under a dissecting microscope, and the tissue strips were post-fixed with 2% osmium tetroxide/cacodylate buffer (pH 7.4). The tissues were dehydrated through a series of graded alcohols and embedded in Poly/Bed 812 resin (Polysciences, Warrington, PA), and 1-μm serial sections were cut with an ultramicrotome, stained with toluidine blue, and examined by light microscopy. To get an estimate of the incidence of CNV at various time points after laser treatment, burns were randomly selected: 16 at 1 week, 11 at 2 weeks, and 11 at 4 weeks. For each, 1-μm serial sections were cut through the entire lesion, stained, and examined by light microscopy. For some lesions, ultrathin sections were prepared, counterstained with uranyl acetate and lead citrate, and examined with a transmission electron microscope (JEOL 100CX).

Comparison of Laser-Induced CNV in FGF2-Deficient and Control Mice

Mice with targeted disruption of the FGF2 gene were produced by homologous recombination, and their characterization has been published elsewhere. 11 FGF2-deficient and control mice with the same genetic background were treated with krypton laser in each eye as described above. After 2 weeks, fluorescein angiography was done, the mice were sacrificed, and their eyes were removed and fixed in 4% glutaraldehyde in 0.1 mol/L cacodylate buffer (pH 7.4). Each burn was identified and isolated under a dissecting microscope, and the tissue was processed as described above. One-micron serial sections were cut through each burn in entirety, stained, and examined to determine the presence or absence of CNV.

Results

Production and Characterization of Laser-Induced CNV in C57BL/6J Mice

A bubble, suggesting rupture of Bruch’s membrane, was observed in association with 67 of 75 burns (87%) in 25 eyes (13 mice) treated with laser photocoagulation. This resulted in chalk-white retinal burns without subretinal hemorrhage (Figure 1) ▶ . Fluorescein angiography was done 1 week after laser treatment in 10 mice that had a total of 60 laser burns. Of the 60 burns, 34 (57%) showed evidence of dye leakage, judged by the presence of a hyperfluorescent spot that increased in size and developed blurred margins over time (Figure 2) ▶ . The range of leaking hyperfluorescent spots was one to three per eye, with most mice showing fluorescein leakage in two of three lesions. In 56 burns in which a bubble had been produced, 34 (61%) leaked fluorescein. Ten burns (18%) showed a hyperfluorescent spot that increased to become a large bright fluorescent spot, suggesting extensive leakage of dye. Fluorescein angiography 4 weeks after laser treatment in three mice showed leakage in 7 of 18 burns (39%); 2 burns (11%) showed extensive leakage.

Figure 1.

Fundus color picture immediately after laser treatment. After intense laser photocoagulation, a bubble without subretinal hemorrhage was seen and resulted in chalk white burns.

Figure 2.

Fluorescein angiography at 1 week after laser treatment. In the early stage of the angiogram (left, 2 minutes after intraperitoneal injection), there was a hyperfluorescent spot that increased in size and developed blurred margins in the late stage of the angiogram (right, 8 minutes after injection).

The time course of histopathological changes after laser treatment was investigated in at least three mice at each time point. Immediately after laser treatment, Bruch’s membrane was disrupted in the center of laser burns and there was some damage to the outer retina, but the inner retina was intact (Figure 3a) ▶ . All layers of the choroid were destroyed within the laser burn, and the surrounding choriocapillaris was collapsed. There was no subretinal hemorrhage. Three days after laser treatment, there was cellular tissue extending from the choroid through the rupture in Bruch’s membrane (Figure 3b) ▶ . In association with this tissue, there were many pigment-containing macrophage-like cells, and at the borders of the burn there were many retinal pigmented epithelial (RPE) cells. One week after laser treatment, newly formed vessels with wide lumens extended from the choroid into the subretinal space where they became partially enveloped by RPE cells (Figure 3c) ▶ . The mass of blood vessels was often accompanied by an overlying serous retinal detachment. Two weeks after laser treatment, large collections of vascular channels were seen in the subretinal space. In burns that had shown leakage of dye by fluorescein angiography, there was serous detachment overlying the vascular channels, and the channels were not completely covered by RPE cells (Figure 3d) ▶ . Burns that had shown little or no fluorescein leakage, usually still contained choroidal neovascularization, but the vessels were small with narrow lumens and there was prominent RPE proliferation along the margins of the vessels and no or minimal serous detachment (Figure 3e) ▶ . Four weeks after laser treatment, there were still many blood vessels with large lumens in the subretinal space; however, all lesions examined appeared to be completely engulfed by RPE cells (Figure 3f) ▶ . Electron microscopy showed fenestrated vessels with large lumens in areas of choroidal neovascularization (Figure 4) ▶ . As noted on light micrographs, RPE cells surrounded the neovascular tissue and appeared to engulf it.

Figure 3.

Time course of histopathological findings after laser treatment. a: Immediately after laser treatment, Bruch’s membrane was disrupted in the center of a laser burn (arrows). There was some damage to the outer retina, but the inner retina was intact. All layers of the choroid were destroyed within the laser burn, and the surrounding choriocapillaris was collapsed. There was no subretinal hemorrhage. b: Three days after laser treatment, there was cellular tissue extending from the choroid through the rupture in Bruch’s membrane. There were many pigment-containing macrophage-like cells (arrowheads), and at the borders of the burn there were many RPE cells (arrows). c: One week after laser treatment, newly formed vessels with wide lumens extended from the choroid into the subretinal space and were partially enveloped by RPE cells (arrowheads). The “feeder” vessel from the choroid (arrow) was seen. The mass of blood vessels was accompanied by an overlying serous retinal detachment. d: Two weeks after laser treatment in a burn that had shown leakage of dye by fluorescein angiography, large collections of vascular channels were seen in the subretinal space. There was serous retinal detachment overlying the vascular channels. Feeder vessels from the choroid (arrow) were seen. The channels were not completely covered by RPE cells (arrowheads). e: Two weeks after laser treatment in a burn that had not shown leakage of dye by fluorescein angiography, a small area of choroidal neovascularization was completely covered by RPE cells. There was no serous retinal detachment. f: Four weeks after laser treatment, there were still many blood vessels with large lumens in the subretinal space. RPE cells (arrowheads) completely engulfed the vessels.

Figure 4.

Ultrastructural findings 1 week after laser treatment in C57BI/6J mice. Neovascularization (NV) with large lumens (L) and many fenestrations (inset, arrowheads) was seen anterior to Bruch’s membrane (BM). Retinal pigment epithelial (RPE) cells partially surrounded the neovascular tissue. En, endothelial cell.

Randomly selected burns were serially sectioned through their entire extent to provide an estimate of the incidence of CNV at various time points. The number of lesions that contained CNV was 13 of 16 at 1 week (81.3%; 13 of 15, 86.7%, in which a bubble had been produced), 9 of 11 at 2 weeks (81.8%; 9 of 10, 90%, in which a bubble had been produced), and 11 of 11 at 4 weeks (100%, in all of which a bubble had been produced).

Comparison of Laser-Induced CNV in FGF-2-Deficient and Control Mice

Krypton laser photocoagulation was done in FGF2-deficient and wild-type control mice in the same manner as that described for C57BL/6J mice. Because in C57BL/6J mice it appeared that production of a bubble at the time of laser was an important factor in obtaining CNV, only mice in which a bubble was produced for all three burns were used (six FGF2-deficient and five controls). Two weeks after treatment, fluorescein angiography showed leakage in 63% of burns in control mice compared with 39% in FGF2-deficient mice, a difference that was statistically significant (Table 1) ▶ . Examination of serial sections showed that 87% of lesions in control mice contained CNV compared with 75% in FGF2-deficient mice, which was not a statistically significant difference. There was no identifiable difference in the light microscopic (not shown) or ultrastructural appearance (Figure 5) ▶ of lesions in FGF2-deficient mice versus controls.

Table 1.

Incidence of Fluorescein Leakage and Histopathological Evidence of CNV in FGF2-Deficient and Wild-Type Mice 2 Weeks after Laser-Induced Rupture of Bruch’s Membrane

Mice	n	Fluorescein leakage	Histopathological evidence of CNV
FGF2 deficient	36	14 (39%)*	27 (75%)†
Wild type	30	19 (63%)	26 (87%)

*P = 0.0453 for difference from wild type by a two-tailed test for proportions.

^†P = 0.2249 for difference from wild type by a two-tailed test for proportions.

Figure 5.

Ultrastructural findings 2 weeks after laser treatment in FGF2-deficient and control mice. A large lumen (L) is seen in new vessels containing fenestrated endothelial cells (insets) in both wild-type (a) and FGF2-deficient (b) mice. In both, the vessels are partially surrounded by RPE cells, and there is no discernible difference in their appearance.

Discussion

In this study, we have described a murine model of laser-induced CNV that has several features that are advantageous for investigations of the pathogenesis and/or treatment of CNV. 1) The time course of events is appropriate for studies intended to determine the impact of a particular intervention on the development of CNV. The CNV occurs rapidly (within 1 week), is well established within 2 weeks, and does not show evidence of spontaneous regression as judged by histopathology at 1 month. 2) The model is highly reliable in that CNV can be identified in over 80% of lesions at 1 or 2 weeks. This is sufficiently high to allow detection of a treatment effect with a relatively small number of mice. 3) Mice are relatively inexpensive, and the manipulation required to induce CNV is well standardized and can be done rapidly. Therefore, it is possible to generate a large number of lesions, allowing statistical analysis of the effect of interventions. 4) The small size of mice allows for use of smaller amounts of drugs. This is an important consideration for testing new agents that are not mass produced and are available in limited quantities.

The murine model of CNV compares favorably with previously published rat models with respect to the features listed above. CNV occurred in 6 of 14 rat eyes 1 to 3 months after multiple moderate krypton laser burns 5 and in 60 to 78% of lesions 1 month after intense laser burns meant to rupture Bruch’s membrane. 6 Thirty-eight percent of rats immunized with interphotoreceptor retinol-binding protein synthetic peptides developed CNV after 45 days. 12 However, the major benefit of the mouse model is that, unlike other animals in which CNV has been produced, genetic manipulations are feasible in mice; therefore, the model can be used to investigate the impact of overexpression or underexpression of individual genes on the development of CNV.

We selected FGF2 for investigation because several lines of evidence have suggested that it may be an important stimulus for CNV. Cultured RPE cells produce FGF2, and it stimulates proliferation, migration, and tube formation by choroidal endothelial cells. 13 Surgically removed CNV membranes show immunohistochemical staining for FGF2, including staining within RPE cells. 14-17 In a rat model of laser-induced CNV, an increase in FGF2 mRNA was seen in RPE-like cells, choroidal vascular endothelial cells, and fibroblast-like cells in the lesions. 18 Sustained release of FGF2 in the subretinal space of minipigs 8 or rabbits 9 results in CNV.

Our data demonstrate that the incidence of CNV after laser-induced rupture of Bruch’s membrane is not statistically different in FGF2-deficient and control mice. This means that despite the fact that exogenous administration of FGF2 is capable of stimulating CNV, endogenous FGF2 is not required for CNV to occur. However, this does not mean that FGF2 does not contribute to the development of CNV. There was significantly less fluorescein leakage associated with laser-induced lesions in FGF2-deficient mice compared with wild type mice, and we have noted that small areas of CNV are less likely to leak fluorescein than large lesions. Therefore, it is possible that the areas of CNV were smaller in FGF2-deficient mice, but we were unable to obtain a direct quantitative assessment of the size of the lesions and we cannot be sure there was a difference.

Although our data do not permit a determination as to whether or not FGF2 contributes to CNV, they clearly demonstrate that it is not necessary for its occurrence, indicating that, in the absence of FGF2, other growth factors may be involved. One possibility is that another member of the FGF family could compensate for the absence of FGF2. A reasonable candidate is FGF1, which has similar activity and characteristics and is present in retina and RPE; generation of mice deficient in FGF1 will help to address this issue. Growth factors outside the FGF family may also participate. Vascular endothelial growth factor (VEGF) is an angiogenic factor that has been localized to CNV membranes and could play a role. 16,17,19 We have recently demonstrated that transgenic mice that overexpress VEGF in photoreceptors develop neovascularization originating from retinal blood vessels but not from choroidal vessels. 20 This could mean that choroidal vessels do not respond to VEGF, but it more likely means that disruption of Bruch’s membrane, a component of our model, is necessary for CNV to develop. Tumor necrosis factor is an angiogenic factor produced by macrophages, 21 and macrophage-like cells are seen in our model and in some patients with CNV due to age-related macular degeneration. 22

To our knowledge, this is the first study to use targeted gene disruption to investigate the involvement of a particular gene in a model of CNV. We have recently used this approach to demonstrate that FGF2 does not play a major role in stimulating retinal neovascularization. 23 This is a potentially useful strategy to identify genes that either stimulate or inhibit the development of ocular neovascularization and may help guide attempts to develop new treatments.