¹Department of Molecular Ophthalmology, Lions Eye Institute, Nedlands, Western Australia, Australia

²Centre for Ophthalmology and Visual Science, University of Western Australia, Nedlands, Western Australia, Australia

Correspondence to: P E Rakoczy, Correspondence: PE Rakoczy, Department of Molecular OphthalmologyLions Eye Institute, 2 Verdun Street, Nedlands 6009, Western Australia, Australia

Neovascularisation (NV) within the eye often results in visual loss. Vascular endothelial growth factor (VEGF) has been implicated in the development of ocular NV. Previous studies have shown that VEGF antagonists successfully suppressed retinal and choroidal NV in animal models. However, the systemic approach and transient nature of the delivery systems used in these studies hinder therapeutic application. To achieve stable and localised ocular anti-angiogenic therapy, we explored the use of recombinant adeno-associated virus (rAAV)-mediated secretion gene therapy (SGT). In this study, we generated a rAAV vector encoding soluble VEGF receptor 1, sFlt-1 (AAV-CMV.sflt) and determined its ability to inhibit cautery-induced corneal NV and laser-induced choroidal NV. Delivery of AAV-CMV.sflt into the anterior chamber resulted in transgene expression in the iris pigment epithelium and corneal endothelium, which reduced the development of corneal NV in the stroma of cauterised rats by 36% compared with cauterised control groups (P = 0.009). Subretinal delivery of AAV-CMV.sflt near the equator of the eye also suppressed choroidal NV at the laser lesions around the optic nerve by 19% (P = 0.002), indicating that there was diffusion of the secreted anti-angiogenic protein across the retina. Both results suggest that the long-term suppression of ocular NV is possible through the use of stable rAAV-mediated SGT.

Gene Therapy 2002 9, 804-813. DOI: 10.1038/sj/gt/3301695

secretion gene therapy; vascular endothelial growth factor; sFlt-1; adeno-associated virus; ocular neovascularisation; inhibition

Introduction

Neovascularisation (NV) or angiogenesis is the growth of new blood vessels from pre-existing vasculature, and is a highly regulated process that occurs mainly during embryonic development. In adults, there is very little turnover of blood vessels except during tissue repair and in the female reproductive system. Uncontrolled blood vessel growth is undesirable and plays a central role in the development of cancers, rheumatoid arthritis and psoriasis.¹ In the eye, NV is a serious pathologic condition that occurs in a number of ocular diseases including diabetic retinopathy, age-related macular degeneration, retinopathy of prematurity and retinal vein occlusion. Neovascular complications in the retina are the most common causes of visual loss among adults in Western countries.^2,3 The new blood vessels that form are leaky resulting in haemorrhage, exudates and accompanying fibrosis that causes blindness.

Similar to other angiogenic conditions,^4,5 vascular endothelial growth factor (VEGF) has been identified as the major mediator of NV in the eye. The up-regulation of VEGF and its receptors in human neovascular ocular tissues has been well documented.^{6,7,8,9,10,11} Several studies have further shown that overexpression of VEGF in the retina by adenoviral vectors or in transgenic animal models resulted in retinal and choroidal NV.^12,13,14 Current conventional treatments of retinal and choroidal NV such as laser photocoagulation and surgical intervention provide only symptomatic treatment of the disease without addressing the underlying cause. Consequently, use of therapeutic agents that directly inhibit the angiogenic stimuli may be able to provide a more effective and permanent treatment.

Adeno-associated virus (AAV)-mediated gene delivery enables the long-term expression of transgenes,^{15,16,17,18,19} thus providing a potential solution for the long-term delivery of anti-angiogenic agents in the eye. However, one of the difficulties of ocular anti-angiogenic therapies is that several types of cells such as retinal pigment epithelial (RPE) cells, Muller cells, ganglion cells, the inner nuclear layer, macrophages and choroidal vascular endothelial cells have all been identified as potential contributors to VEGF upregulation.^20,21,22 Some of these target cells are difficult to access, transduce or undesirable to disturb, especially within the macula, the central part of the retina responsible for 80% of human vision.²³

To fulfil the specific requirements of ocular anti-angiogenic therapy, we explored the use of secretion gene therapy (SGT). The significant advantage of SGT over traditional gene therapy is that the target cells of transduction are not necessarily those that are responsible for the production of angiogenic factors, but are cells that can be accessed and transduced with high efficiency. The distribution and delivery of the recombinant virally produced therapeutic agent, which is a secreted protein, can then be facilitated by the natural fluid flow and diffusion pathways in the eye. The movement of fluids within the anterior chamber and the posterior flow out from the subretinal space towards the choroid will allow circulation of the secreted proteins.^23,24 It is possible that the continuous production and secretion of these proteins might also advance the slow diffusion of molecules across the retina.²⁵

In this study, two animal models of ocular NV were used: cautery-induced corneal NV and laser photocoagulation-induced choroidal NV. Both models have been shown to induce VEGF expression resulting in extensive blood vessel growth in the normally avascular cornea and subretinal space, respectively.^22,26,27 Based upon the proven efficacy of soluble VEGF receptor 1, sFlt-1, as a potent anti-angiogenic agent,^28,29,30 we selected sFlt-1 as our therapeutic gene and investigated the efficacy of AAV-mediated delivery and expression of this protein as SGT against global corneal NV and choroidal NV.

Results

Characterisation of AAV-CMV.sflt in vitro

Western blot analysis of culture supernatant from AAV-CMV.sflt-infected cells confirmed the expression and secretion of sFlt-1 (Figure 1a This secreted protein encoded by the AAV vector was further confirmed to inhibit VEGF-induced endothelial cell proliferation in vitro (Figure 1b). VEGF-induced proliferation was reduced by 32% when 40 l of conditioned medium from AAV-CMV.sflt-transduced 293 cells were added (P < 0.02). Doubling the volume of AAV-CMV.sflt-conditioned medium completely inhibited VEGF activity (P < 0.005) with cell growth returning to basal levels similar to culture in starvation medium.

Confirmation of gene expression by AAV-CMV.sflt in vivo

RT-PCR analysis of chorioretinal tissue confirmed expression of human sFlt-1 mRNA in an AAV-CMV.sflt-injected eye (Figure 2, lane 1>), whereas eyes injected with AAV-CMV.gfp or uninjected controls showed no detectable signal (lanes 2 and 3). The lack of contaminating viral DNA was confirmed by absence of the PCR band in the corresponding reverse transcriptase negative controls (lane 4).

Suppression of cautery-induced corneal neovascularisation

Slit-lamp photography: Following AAV-CMV.gfp injection into the anterior chamber, strong GFP expression was evident in the iris flat mount (Figure 3a-i). There was also some GFP expression present in the corneal flat mount Figure 3a-ii, but the transduction efficiency was low compared with the iris. Cryosectioning localised the GFP signal to the iris pigment epithelium and to the corneal endothelium Figure 3a-iii. A few cells of the iris stroma were also positive with GFP, however expression was weak Figure 3a-iii.

Injection with AAV-CMV.sflt showed no histological changes to the cornea, the corneal endothelium was intact and no inflammation was visible Figure 3b. Cautery of AAV-CMV.gfp-injected eyes resulted in formation of scar tissue and growth of numerous blood vessels in the cornea Figure 3c. On the other hand, cautery of AAV-CMV.sflt-injected eyes resulted in scarring but there was an absence of corneal vessels Figure 3d.

The level of corneal NV at 4 days after cautery was scored by three independent observers. The mean score Figure 3efor eyes injected with AAV-CMV.sflt was significantly lower compared with eyes with AAV-CMV.gfp (P < 0.03), PBS vehicle (P < 0.05) and uninjected eyes (P < 0.004). There was no significant difference between the mean scores of the control groups (P > 0.2). In total, only 27% of corneas with AAV-CMV.sflt were vascularised (mean score 1) compared with 63% of the control groups collectively (P = 0.009; Table 1).

Histology: Vascularisation of the cornea observed in slit-lamp photography was confirmed by histology. Numerous blood vessels filled with erythrocytes were present in the stroma of the cauterised AAV-CMV.gfp-injected cornea (Figure 4a). Abundant blood vessels were also evident in cauterised PBS vehicle-injected and cauterised non-injected eyes (data not shown). No corneal blood vessels were observed in the majority of cauterised AAV-CMV.sflt-injected eyes Figure 4b. The lack of proliferating endothelial cells was further confirmed by immunohistochemistry, which failed to detect the presence of vWF in the cornea of the cauterised AAV-CMV.sflt-injected eye Figure 4f suggesting an absence of vascular endothelial cells. In contrast, the cauterised AAV-CMV.gfp cornea stained strongly for vWF (Figure 4e) indicating strong endothelial cell proliferation.

Many infiltrating cells were observed along the entire length of the corneal stroma of cauterised AAV-CMV.gfp-injected eyes Figure 4a. There was also significant oedema in the vascularised AAV-CMV.gfp eyes with the corneal stroma swelling to more than double the thickness of normal cornea. Cauterised AAV-CMV.sflt-injected eyes also showed some evidence of infiltration, although most inflammatory cells were concentrated at the cautery site Figure 4band very few were visible at the peripheral region Figure 4d. There was no evidence of significant tissue swelling in AAV-CMV.sflt-injected eyes.

Suppression of laser-induced choroidal neovascularisation

Fluorescein angiography: We and others have previously demonstrated that delivery of AAV vectors into the subretinal space of rat eyes resulted in efficient transgene expression in the RPE cell layer and photoreceptors.^16,19 In this study, AAV was delivered transclerally into the subretinal space near the equator (Figure 5a). Using fluorescence fundus photography, at least 60% of cells within the injection bleb, which covered an area of approximately one-quarter of the retina, were transduced by AAV-CMV.gfp Figure 5b. However, transduction was confined mainly within the injection area with little lateral spreading.

In this rat model of choroidal NV, 10 to 12 laser spots were delivered near the optic nerve Figure 5cat a distance from the injection site Figure 5d. The formation of choroidal NV was monitored by fluorescein angiography. The intensity of fluorescein leakage in each lesion was used as an indication of the amount of new blood vessels formed. Each laser spot was given a score of 0 to 3 based on reference angiograms (Figure 6). At 5 weeks post-laser photocoagulation, approximately 60% of the 103 laser spots delivered to AAV-CMV.gfp-injected eyes demonstrated leakage (Table 2, Figure 5f), consistent with our previous study.²² Rats treated with AAV-CMV.sflt showed leakage in 41% of the 106 laser spots delivered indicating a 19% reduction in percentage of leaky lesions (P = 0.002, Table 2; Figure 5e). This was also reflected in the mean score per eye, which was 22% lower in AAV-CMV.sflt-treated eyes (P = 0.001, Table 2), indicating less leakage on the whole in these treated eyes. At 16 weeks after laser photocoagulation, the mean score per eye treated with AAV-CMV.sflt remained significantly lower at 1.15 ± 0.26 compared with AAV-CMV.gfp-injected eyes with a score of 1.59 ± 0.17 (P = 0.02, n = 6; Figure 5g).

Histology: Examination of haematoxylin and eosin-stained sections of the area immediately adjacent to the injection site revealed a normal RPE, outer segments and outer nuclear layer suggesting lack of toxicity from sFlt-1 expression (data not shown). Electroretinograms also indicated normal functioning of AAV-CMV.sflt-injected eyes (data not shown).

Most of the laser lesions in eyes treated either with AAV-CMV.sflt or AAV-CMV.gfp developed subretinal cellular membrane (Figure 7a). Measurement of the thickness ratio (B/C) of 20 lesions from two eyes in each group indicated a 16% reduction of B/C in AAV-CMV.sflt-treated eyes (P = 0.006; Table 2) supporting the angiogram data. More importantly, there was a significant difference in the vascularisation of these lesions. Lesions in eyes treated with AAV-CMV.sflt had generally less proliferating endothelial cells as determined by vWF and lectin staining Figure 7ccompared with AAV-CMV.gfp-injected eyes Figure 7d.

Discussion

The conventional approaches to treating retinal and choroidal NV, namely laser photocoagulation, cryotherapy and surgical excision, can be beneficial, but are limited to a small subset of patients. These procedures can also induce multiple side-effects due to their destructive nature to the neuroretina and RPE layer, and hence are unsuitable for treatment near the macula. A newer form of treatment, photodynamic therapy (PDT) has been found to be relatively effective at reducing the risk of visual loss from predominantly classic subfoveal choroidal NV with minimal effect on the retina.³¹ However, all these approaches do not address the underlying stimuli of NV and recurrences are a significant problem.^32,33,34 Therefore, new therapies aimed at blocking the underlying stimuli for new vessel growth are necessary to achieve better treatment for these patients.

Treatment with sFlt-1, a naturally-occurring truncated and secreted form of VEGF receptor Flt-1, has been shown to inhibit angiogenesis in a variety of disease models.^29,30,35 The anti-angiogenic action of sFlt-1 is due to its ability to directly sequester VEGF and to form heterodimers with membrane-bound Flt-1 and VEGF receptor 2.^28,36 It is widely believed that VEGF up-regulation also plays a major part in the development of ocular NV.^{6,7,8,9,20,21,22} Several studies have further shown that inhibition of VEGF activity by administration of VEGF-specific kinase inhibitors, VEGF antibodies, VEGF receptors and VEGF antisense oligonucleotides successfully suppressed the development of retinal and choroidal NV.^37,38,39,40 In particular, Honda et al⁴¹ showed that expression of adenovirus-mediated sFlt-1 suppressed choroidal NV. However, the adenoviral vector was injected into the rat femoral muscle resulting in systemic circulation of sFlt-1, which dropped to very low levels after 21 days. Similarly, the other inhibitory studies mentioned used systemic delivery and presence of the anti-VEGF agent was temporary. Systemic circulation of anti-angiogenic drugs may adversely affect other organs particularly when a higher dosage is required to adequately achieve therapeutic levels in the eye. It is also likely that the angiogenic stimuli causing ocular NV are present for an extended period of time, longer than the presence of these transient drugs.

rAAV has been shown to efficiently and stably transduce RPE and photoreceptors following subretinal injection.^{15,16,17,18,19} rAAV-mediated expression of GFP persists at least 3 years after subretinal injection without significant toxicity and immune response,⁴² and would therefore be a suitable delivery vector for long-term inhibition of ocular NV. In this present study, we demonstrated that the stable expression of rAAV-mediated sFlt-1 injected directly into the eye suppressed experimental corneal and choroidal NV.

There was a significant reduction in corneal vessels, oedema and infiltrating cells in cauterised AAV-CMV.sflt-injected eyes. The inhibitory effect on inflammation in this model by AAV-CMV.sflt was not unexpected as it has been established that cauterisation leads to the secretion of VEGF, initially by polymorphonuclear neutrophils and then by recruited macrophages, resulting in the formation of a chemotactic gradient, which induces proliferation of vascular endothelial cells from the limbal vascular plexus into the corneal stroma towards the cautery site.²⁶ Therefore, presence of sFlt-1 in the anterior chamber inhibited cautery-induced corneal NV through two mechanisms: blockade of VEGF activity as chemoattractant for monocytes and as mitogen for vascular endothelial cells. In the choroidal NV study, the presence of infiltrating cells and other inflammatory cytokines and molecules have also been observed at the site of laser photocoagulation.^21,22 It is probable that injection with AAV-CMV.sflt similarly suppressed the activity of VEGF as a stimulator of inflammation and angiogenesis in this model.

In the anterior chamber, AAV transduction was confined mainly to iris pigment epithelial and corneal endothelial cells, but VEGF up-regulation and subsequent new vessel development occurred at the limbus and in the corneal stroma. In the choroidal NV study, AAV was delivered to the subretinal space near the equator at a distance from the laser lesions around the optic nerve. Subretinal injections of AAV-CMV.gfp clearly showed that AAV transduction was confined to the injection area. However, injection with AAV-CMV.sflt was able to suppress choroidal NV in lesions distant from the injection area suggesting the diffusion of sFlt-1 throughout the retina.

Such ability to suppress angiogenesis in areas distant from the injection site is important, because retinal and choroidal NV often manifest at multiple sites within the retina. Unlike antisense oligonucleotides, ribozymes or cell-bound anti-angiogenic factors, the use of secreted inhibitory factors circumvents the need to deliver directly into specific cells at the neovascular site, which may be vascular endothelial cells, inflammatory cells, RPE cells or other retinal cells. Delivery into such a diverse and large number of cells is complex and would require multiple or large injections of cell-bound anti-angiogenic agents and development of a universal delivery system that can efficiently target all these various cell types. In this study, we showed that these technical difficulties could be overcome by use of rAAV-mediated SGT, whereby transgene delivery and expression are targeted to ocular cells particularly permissive to rAAV transduction. The secreted anti-angiogenic factor then distributes throughout the ocular chambers suppressing angiogenesis globally in the retina or anterior chamber. Use of SGT would be especially beneficial in suppressing NV at the central fovea thus avoiding direct injections, which might cause detachments and functional damage.

However, it should be noted that the level of suppression obtained with AAV-CMV.sflt was less when compared with that achieved by Ad.sflt. In our cautery-induced study of corneal NV, injection with Ad.sflt achieved a 70% decrease in the number of vascularised corneas,²⁷ while injection with AAV-CMV.sflt reduced NV only by 36%. This lower efficacy of AAV-CMV.sflt was probably due to a combination of lower titer and lower transduction efficiency of AAV in the anterior chamber, particularly in the corneal endothelium from which secreted sFlt-1 proteins have closer access to the cautery lesion.

In the choroidal NV study, AAV-CMV.sflt also failed to totally inhibit NV in all lesions but suppressed only 19% of lesions from forming new vessels compared with AAV-CMV.gfp-injected eyes. The molecular weight of sFlt-1 at approximately 110 kDa may restrict its ability to diffuse freely within the retina. Therefore, to achieve a higher rate of inhibition may require a high viral titer or use of an intravitreal route of injection that may improve diffusion of sFlt-1 through the retina. However, the lower efficiency of gene transfer and expression by AAV vectors following intravitreal injections^15,17 suggests that this approach may not be beneficial. Higher efficacy might also be achieved by using other smaller anti-angiogenic proteins, such as angiostatin and pigment epithelium-derived factor,^43,44 that can be more easily secreted and diffused throughout the retina.

The lack of total inhibition by both Ad.sflt and AAV-CMV.sflt could also point to the complexity of the angiogenic process in the eye. Inhibition of VEGF activity alone might not be sufficient to totally prevent development of retinal and choroidal NV. There are conflicting data as to the effectiveness of VEGF antagonists in treating ocular NV. Previous attempts using VEGF antisense oligonucleotides, soluble Flt-1-chimeric proteins and adenovirus-mediated sFlt-1 also failed to completely block retinal or choroidal NV, achieving only a maximum inhibition of 50%.^37,38,40,41 In contrast, administration of kinase inhibitors that blocked phosphorylation by VEGF and platelet-derived growth factor (PDGF) receptors completely prevented the development of retinal NV, but kinase inhibitors of PDGF receptors alone had no effect.³⁹ Although this implies that blockade of VEGF activity alone is sufficient for the complete inhibition of NV, a synergy might exist between VEGF and PDGF, or other angiogenic factors, which all need to be blocked concurrently for complete inhibition.

Although no obvious abnormalities were seen in the retina despite expression of sFlt-1 up to 16 weeks, thorough and long-term toxicity studies examining the effect of prolonged blockade of VEGF activity to the retina, to the normal vasculature particularly normal choroidal vessels,⁴⁵ and to non-ocular tissues are needed. The application of AAV-CMV.sflt in treating retinal and choroidal NV may require more investigation, but the concept of AAV-mediated SGT appears to be an attractive approach to suppress new blood vessel growth in the eye.

Materials and methods

Cell cultures

Human umbilical vein endothelial cells (HUVECs) and human embryonic kidney 293 cells were obtained from the American Type Culture Collection (Rockville, MD, USA). The human RPE cell line D407 was kindly provided by Dr Richard C Hunt (University of South Carolina Medical School, Columbia, SC, USA). The 293 and RPE cells were cultured in Dulbecco's modified essential medium (DMEM; Gibco, Grand Island, NY, USA) supplemented with 10% foetal bovine serum (FBS, Gibco), 100 units/ml penicillin (Gibco) and 100 g/ml streptomycin (Gibco). HUVECs were maintained in the same medium supplemented with 25 g/ml bovine endothelial growth factor (bEGF, Roche Diagnostics, Indianapolis, IN, USA) and 100 g/ml heparin. All cultures were maintained at 37°C and 5% CO₂ in a humidified atmosphere.

AAV vector construction and production

The human sFlt-1 gene was obtained from the Ad.sflt construct kindly provided by Dr Imre Kovesdi (GenVec, Gaithersburg, MD, USA). The sFlt-1 gene was subcloned into plasmid pCI (Promega, Madison, WI, USA), which contains the cytomegalovirus (CMV) major immediate-early gene enhancer/promoter, a chimeric intron and a simian virus 40 polyadenylation (SV40 polyA) signal. The resultant CMV-sFlt-1-SV40 polyA construct was then cloned into AAV plasmid SSV9 to generate pAAV-CMV.sflt vector. The construction of plasmid pAAV-CMV.gfp has been described previously.⁴⁶

AAV vector particles were purified by heparin affinity chromatography.⁴⁷ The preparation of Ad.sflt has been described.²⁷ AAV vector titers were determined using a DNase-resistant physical particle assay.⁴⁷ Titers for viruses were 4 ´ 10¹¹ particles/ml of AAV-CMV.sflt, 8 ´ 10¹¹ particles/ml of AAV-CMV.gfp and 10¹¹ plaque-forming units (p.f.u.)/ml of Ad.sflt.

Western analysis

D407 and 293 cells were infected with AAV vectors at a multiplicity of infection (MOI; particles or p.f.u./cell) of 100 in DMEM medium supplemented with 2% FBS. As positive control, D407 cells were also infected with Ad.sflt at MOI of 20. After 2 days, conditioned media were collected and sFlt-1 proteins present were concentrated with heparin sepharose CL-6B beads (Pharmacia Biotech, Uppsala, Sweden) for 4 h at 4°C. The beads were collected, washed with PBS and then boiled in reducing sample buffer. An aliquot was electrophoresed on an 8% polyacrylamide gel and then transferred to a nitrocellulose membrane (Hybond ECL; Amersham, Buckinghamshire, UK). Presence of sFlt-1 was detected by a monoclonal anti-Flt-1 antibody (1:500 Flt-11; Sigma, St Louis, MO, USA) followed by incubation with horseradish peroxidase-conjugated sheep anti-mouse IgG (Amersham). Labelling was detected with the enhanced chemiluminescence system (Amersham).

HUVEC proliferation assay

The biological activity of sFlt-1 produced from AAV-CMV.sflt was assessed on the basis of its ability to suppress VEGF-induced proliferation of HUVECs. The protocol used has been described previously.²⁷ In this case, conditioned media from AAV-CMV.sflt-transduced and AAV-CMV.gfp-transduced 293 cells were added to the VEGF-treated HUVECs in increasing dilutions. A control of starvation medium (normal HUVEC growth medium without bEGF) only was also included. Heparin was added to each well at 100 g/ml. The relative VEGF-induced proliferation of HUVECs treated with VEGF and the different conditioned media was assayed by addition of 25 l of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT, 5 mg/ml; Sigma) to each well for 4 h at 37°C. The cells were lysed with 100 l of extraction buffer (20% SDS in 50% dimethylformamide, pH 4.7) at 37°C overnight. The plates were read on an ELISA plate reader (Dynatech Medica, Guernsey, UK) at a wavelength of 570 nm. The experiment was done in triplicates and the relative proliferation of the HUVECs was compared on the basis of the OD₅₇₀ value.

Injections

All animal experiments were performed in accordance with the ARVO Statement on the Use of Animals in Vision and Ophthalmic Research. Before injection, animals were anaesthetized with a mixture of ketamine (50 mg/kg body weight) and xylazine (8 mg/kg body weight). For anterior chamber injections, topical application of proparacaine hydrochloride was applied. Eyes that were injected subretinally were further treated with topical amethocaine drops and the pupils dilated with 1% tropicamide and 2.5% phenylephrine hydrochloride drops.

Anterior chamber injections were performed on 5- to 11-week-old nonpigmented RCS/rdy⁺ rats. Two microliters of AAV-CMV.sflt, AAV-CMV.gfp or vehicle (phosphate buffered saline; PBS) were injected into the anterior chamber of each right eye via the temporal limbus using a 32-gauge needle attached to a 5-l Hamilton syringe.

Subretinal injections were performed on 10, 8- to 9-week-old, pigmented RCS/rdy⁺ rats. The injection technique used has been described previously.¹² Briefly, the conjunctiva was cut close to the limbus to expose the sclera, which was then punctured with a 30-gauge needle. A 32-gauge needle was passed through this hole in a tangential direction under an operating microscope. Two microliters of virus were delivered into the subretinal space of each eye. AAV-CMV.sflt was injected into the left eye and the contralateral right eye was injected with AAV-CMV.gfp. The needle was kept in the subretinal space for 1 min, withdrawn gently, and antibiotic ointment applied to the wound site.

RT-PCR analysis

At 3 months following subretinal injection, an animal injected with either AAV-CMV.sflt, AAV-CMV.gfp or uninjected was killed and the eyes were enucleated. Each posterior eye cup was separated from the anterior segment and vitreous. RNA was extracted using RNeasy mini-columns (Qiagen, Valencia, CA, USA). Two micrograms of purified RNA were treated with RQ1 DNase I (Promega) and ethanol precipitated. cDNA was synthesised using Moloney murine leukaemia virus reverse transcriptase (Promega) and a human sFlt-1-specific primer (5'- GTTCAAGGGAGTGGTAGCAGTACAA-3'). PCR amplification was performed using two other gene-specific primers, 5'-GAAATGGTGAGTAAGGAAAGCG-3' and 5'-TACTGTCCCAGATTATGCGTTTT-3'. The presence and quality of RNA was confirmed for all samples by positive amplification of a housekeeping gene.

Induction of corneal angiogenesis by silver nitrate/potassium nitrate cauterisation

Three weeks after injection, the anterior chamber-injected rats were anaesthetized by Fluothane (Zeneca, Macclesfield, UK) and proparacaine hydrochoride. A 75% silver nitrate/25% potassium nitrate applicator stick (Graham-Field Surgical, New Hyde Park, NY, USA) was applied to the central cornea of each right eye for 4 s to produce a 1-mm diameter circular cauterisation. A total of 22 AAV-CMV.sflt-injected rats, 13 AAV-CMV.gfp-injected rats, 11 PBS-injected rats and 11 uninjected rats were cauterised.

Image analysis of corneal angiogenesis

The corneas of rat eyes were examined by slit-lamp photography at 4 days after cautery. The extent of corneal NV in these eyes was scored in a masked fashion by three examiners using the grading system previously established:²⁷ negative (0), no new vessels present; mild (1), dense new blood vessels present at limbus; moderate (2), blood vessels extending halfway from limbus into the cornea; and severe (3), blood vessels extending all the way from the limbus to the cautery site. The NV score for each eye was averaged from the three observations.

Induction of choroidal neovascularisation by krypton laser photocoagulation

At 1 month post-injection, 10 subretinally injected rats were anaesthetized and their pupils dilated as described above. Krypton laser irradiation (647.1 nm, Coherent Radiation System, CA, USA) was delivered to both the left AAV-CMV.sflt-injected eye and the contralateral AAV-CMV.gfp-injected eye of each animal through a Zeiss slit lamp with a hand-held coverslip serving as a contact lens. A total of 10 to 12 laser burns were applied in each eye surrounding the optic nerve at the posterior pole at a setting of 100 m diameter, 0.1 s duration and 150 mW intensity.

Fluorescence microscopy and fluorescence fundus photography

To view GFP expression in the anterior segment, the eyes injected with AAV-CMV.gfp into the anterior chamber were fixed in 4% paraformaldehyde for 30 min and then dissected. The cornea and iris were separated into flat mounts and examined under fluorescence microscopy (BX60; Olympus, Japan). The flat mounts were further embedded in optical cutting temperature compound (Sakara Finetek, Torrance, CA, USA), frozen and sectioned at 8-12 m. Real-time monitoring of GFP expression following subretinal injection was achieved by fluorescence fundus photography.¹⁶ The percentage GFP-positive area within the injection area was determined from the fluorescence fundus photograph using image analysis software based on the Matrox Imaging Library (Canada).¹⁶

Histology and immunohistochemistry

Respective animals were killed 4 days after cautery and 16 weeks after laser photocoagulation. The eyes were enucleated, fixed in 10% neutral buffered formalin for 4 h and dissected to remove the lens. Tissues were then dehydrated, embedded in paraffin and serially sectioned at 5 m. Sections were stained with haematoxylin and eosin, or immunostained for von Willebrand factor (vWF) and lectin to detect presence of vascular endothelial cells. Stained sections were examined with a compound fluorescence microscope (BX60; Olympus).

Sections that were immunostained for vWF were pre-treated with 20 g/ml proteinase K for 15 min, 3% hydrogen peroxide for 5 min and incubated 2 h with 1:200 rabbit anti-human vWF antibody (DAKO, Carpinteria, CA, USA). The sections were incubated with reagents from a horseradish peroxidase-conjugated streptavidin biotin kit (LSAB2; DAKO) and the signal was detected using 3-amino-9-ethyl-carbazole (AEC) as substrate. All washes were done with Tris-buffered saline (TBS, pH 7.6). Some sections were counterstained with haematoxylin.

Sections that were histochemically stained with fluorescein-conjugated Griffonia simplicifolia lectin I (GSA, Vector Laboratories, Burlingame, CA, USA) were also pre-treated with proteinase K as above and then incubated for 2 h with GSA (1:100). All washes were done with TBS.

Quantitation of choroidal neovascularisation

The extent of choroidal NV in laser lesions was evaluated using two different techniques: scoring of neovascular lesions by fluorescein angiography, and measurement of lesion thickness on serial sections.

At 5 weeks after laser photocoagulation, all 10 animals were examined by fluorescein angiography. At 16 weeks after photocoagulation, only six animals remained and were followed. The animals were injected intraperitoneally with 0.3 to 0.4 ml of 10% sodium fluorescein and the eyes were photographed after 5 min. The level of fluorescein leakage for each laser spot was graded by three independent observers using reference angiograms. Laser spots were graded as follows: 0, no leakage; 1, slight leakage; 2, moderate leakage; 3, strong leakage. The NV score for each eye was averaged from the three observations.

Serial paraffin sections were stained with haematoxylin and eosin. Microscopic images were acquired via a video camera (HV-C20M; Hitachi, Japan) and digitized. The section containing the largest area of subretinal membrane for each lesion was identified. Using a measurement technique that accounts for the position and angle of sections,⁴⁸ the thickness from the bottom of the choroidal layer to the top of the subretinal membrane was recorded using Scion Image software (Beta 4.0.2 version, Scion Corporation, Frederick, MD, USA). The level of NV was determined by the B/C ratio, where B refers to the maximum thickness of the lesion and C is the thickness of the normal RPE-choroidal layers adjacent to the lesion.

Statistical analysis

All in vivo results are presented as mean ± s.e.m. The score per lesion and the percentage of lesions scored >1 were calculated for each animal and then averaged to obtain mean score per eye and mean percentage leakage per eye. The mean score and percentage leakage per eye between paired AAV-CMV.sflt- and AAV-CMV.gfp-injected eyes were compared by paired Student's t tests. The mean B/C ratios and the mean scores in the corneal NV study were compared by independent-sample t tests. The incidences of corneal NV (score 1) in AAV-CMV.sflt-injected and control groups were compared by a 2 ´ 2 chi-square test. Differences were considered significant at P < 0.05.

Acknowledgements

The authors thank Tammy Zaknich for animal photography, Hieu Van Nguyen for preparation of paraffin-embedded sections, and Dr Dru Daniels and Dr Kevin Chee for assisting with grading of eyes. We are also grateful to Dr Anthony Kicic and Pamela Slobe for proofreading the manuscript. This work was supported by grants from the Juvenile Diabetes Research Foundation, USA, and the Lions Eye Institute, Western Australia.

1 Folkman J. Angiogenesis in cancer, vascular, rheumatoid and other disease. Nat Med 1995; 1: 27-31. MEDLINE

2 Macular Photocoagulation Study Group. Five-year follow-up of fellow eyes of patients with age-related macular degeneration and unilateral extrafoveal choroidal neovascularisation. Arch Ophthalmol 1993; 111: 1189-1199.

3 Lee P, Wang CC, Adamis AP. Ocular neovascularisation: an epidemiologic review. Surv Ophthalmol 1998; 43: 245-269. Article MEDLINE

4 Plate KH, Weich HA, Breier G, Risau W. Vascular endothelial growth factor is a potential tumour angiogenesis factor in human gliomas in vitro. Nature 1992; 359: 845-848. MEDLINE

5 Koch AEet al. . Vascular endothelial growth factor. A cytokine modulating endothelial function in rheumatoid arthritis. J Immunol 1994; 142: 4149-4156.

6 Kvanta A, Algvere PV, Berglin L, Seregard S. Subfoveal fibrovascular membranes in age-related macular degeneration express vascular endothelial growth factor. Invest Ophthalmol Vis Sci 1996; 37: 1929-1934. MEDLINE

7 Lopez Pet al. . Transdifferentiated retinal pigment epithelial cells are immunoreactive for vascular endothelial growth factor in surgically excised age-related macular degeneration-related choroidal neovascular membranes. Invest Ophthalmol Vis Sci 1996; 37: 855-868. MEDLINE

8 Kliffen Met al. . Increased expression of angiogenic growth factors in age-related maculopathy. Br J Ophthalmol 1997; 81: 154-162. MEDLINE

9 Boulton M, Foreman D, Williams G, McLeod D. VEGF localisation in diabetic retinopathy. Br J Ophthalmol 1998; 82: 561-568. MEDLINE

10 Smith Get al. . Immunolocalisation of the VEGF receptors FLT-1, KDR and FLT-4 in diabetic retinopathy. Br J Ophthalmol 1999; 83: 486-494. MEDLINE

11 Philipp W, Speicher L, Humpel C. Expression of vascular endothelial growth factor and its receptors in inflamed and vascularised human corneas. Invest Ophthalmol Vis Sci 2000; 41: 2514-2522. MEDLINE

12 Spilsbury Ket al. . Overexpression of vascular endothelial growth factor (VEGF) in the retinal pigment epithelium leads to the development of choroidal neovascularisation. Am J Pathol 2000; 157: 135-144. MEDLINE

13 Okamoto Net al. . Transgenic mice with increased expression of vascular endothelial growth factor in the retina. Am J Pathol 1997; 151: 281-291. MEDLINE

14 Schwesinger Cet al. . Intrachoroidal neovascularisation in transgenic mice overexpressing vascular endothelial growth factor in the retinal pigment epithelium. Am J Pathol 2001; 158: 1161-1172. MEDLINE

15 Ali RRet al. . Adeno-associated virus gene transfer to mouse retina. Hum Gene Ther 1998; 9: 81-86. MEDLINE

16 Rolling Fet al. . Evaluation of adeno-associated virus-mediated gene transfer into the rat retina by clinical fluorescence photography. Hum Gene Ther 1999; 10: 641-648. Article MEDLINE

17 Dudus Let al. . Persistent transgene product in retina, optic nerve and brain after intraocular injection of rAAV. Vision Res 1999; 38: 2545-2553.

18 Bennett J, Duan D, Engelhardt JF, Maguire A. Real-time, noninvasive in vivo assessment of adeno-associated virus-mediated retinal transduction. Invest Ophthalmol Vis Sci 1997; 38: 2857-2963. MEDLINE

19 Flannery Jet al. . Efficient photoreceptor-targeted gene expression in vivo by recombinant adeno-associated virus. Proc Natl Acad Sci USA 1997; 94: 6916-6921. Article MEDLINE

20 Pierce Eet al. . Vascular endothelial growth factor/vascular permeability factor expression in a mouse model of retinal neovascularisation. Proc Natl Acad Sci USA 1995; 92: 905-909. MEDLINE

21 Yi XVet al. . Vascular endothelial growth factor expression in choroidal neovascularisation in rats. Graefe's Arch Clin Exp Ophthalmol 1997; 2235: 313-319.

22 Shen WYet al. . Expression of cell adhesion molecules and vascular endothelial growth factor in experimental choroidal neovascularisation in the rat. Br J Ophthalmol 1998; 82: 1063-1071. MEDLINE

23 Hart WM Jr. Mosby Year Book Inc, 1992.

24 Marmor MF. Control of subretinal fluid: experimental and clinical studies. Eye 1990; 4: 340-344. MEDLINE

25 Marmor MF, Negi A, Maurice DM. Kinetics of macromolecules injected into the subretinal space. Exp Eye Res 1985; 40: 687-696. MEDLINE

26 Edelman JL, Castro MR, Wen Y. Correlation of VEGF expression by leukocytes with the growth and regression of blood vessels in the rat cornea. Invest Ophthalmol Vis Sci 1999; 40: 1112-1123. MEDLINE

27 Lai CMet al. . Inhibition of angiogenesis by adenovirus-mediated sFlt-1 expression in a rat model of corneal neovascularisation. Hum Gene Ther 2001; 12: 1299-1310. Article MEDLINE

28 Kendall RL, Thomas KA. Inhibition of vascular endothelial cell growth factor activity by an endogenously encoded soluble receptor. Proc Natl Acad Sci USA 1993; 90: 10705-10709. MEDLINE

29 Goldman CKet al. . Paracrine expression of a native soluble vascular endothelial growth factor receptor inhibits tumor growth, metastasis, and mortality rate. Proc Natl Acad Sci USA 1998; 95: 8795-8880. Article MEDLINE

30 Kong HLet al. . Regional suppression of tumor growth by in vivo transfer of a cDNA encoding a secreted form of the extracellular domain of the flt-1 vascular endothelial growth factor receptor. Hum Gene Ther 1998; 9: 823-833. MEDLINE

31 Husain Det al. . Effects of photodynamic therapy using verteporfin on experimental choroidal neovascularisation and normal retina and choroid up to 7 weeks after treatment. Invest Ophthalmol Vis Sci 1999; 40: 2322-2331. MEDLINE

32 Macular Photocoagulation Study Group. Macular Photocoagulation Study Group. Persistent and recurrent neovascularisation after laser photocoagulation for subfoveal choroidal neovascularisation of age-related macular degeneration. Arch Ophthalmol 1994; 112: 489-499.

33 Thomas MAet al. . Visual results after surgical removal of subfoveal choroidal neovascular membranes. Ophthalmol 1994; 101: 1384-1396.

34 Schnurrbusch UEKet al. . Histological findings of surgically excised choroidal neovascular membranes after photodynamic therapy. Br J Ophthalmol 2001; 85: 1086-1091. MEDLINE

35 Mori Aet al. . Soluble Flt-1 gene therapy for peritoneal metastases using HVJ-cationic liposomes. Gene Therapy 2000; 7: 1027-1033. MEDLINE

36 Kendall RL, Wang G, Thomas KA. Identification of a natural soluble form of the vascular endothelial growth factor receptor, FLT-1, and its heterodimerisation with KDR. Biochem Biophys Res Commun 1996; 226: 324-328. Article MEDLINE

37 Robinson GSet al. . Oligodeoxynucleotides inhibit retinal neovascularisation in a murine model of proliferative retinopathy. Proc Natl Acad Sci USA 1996; 93: 4851-4856. Article MEDLINE

38 Aiello Let al. . Suppression of retinal neovascularisation in vivo by inhibition of vascular endothelial growth factor (VEGF) using soluble VEGF-receptor chimeric proteins. Proc Natl Acad Sci USA 1995; 92: 10457-10461. MEDLINE

39 Ozaki Het al. . Blockade of vascular endothelial cell growth factor receptor signaling is sufficient to completely prevent retinal neovascularisation. Am J Pathol 2000; 156: 697-707. MEDLINE

40 Shen WYet al. . Pre-clinical evaluation of a phosphorothioate oligonucleotide in the retina of rhesus monkey. Lab Invest 2002; 82: 167-182. MEDLINE

41 Honda Met al. . Experimental subretinal neovascularisation is inhibited by adenovirus-mediated soluble VEGF/flt-1 receptor gene transfection: a role of VEGF and possible treatment for SRN in age-related macular degeneration. Gene Therapy 2000; 7: 978-985. MEDLINE

42 Hauswirth WW, Beaufrere LO. Ocular gene therapy: quo vadis? Invest Ophthalmol Vis Sci 2000; 41: 2821-2826. MEDLINE

43 Mori Aet al. . Pigment epithelium-derived factor inhibits retinal and choroidal neovascularisation. J Cell Physiol 2001; 188: 253-263. Article MEDLINE

44 Lai CCet al. . Suppression of choroidal neovascularisation by adeno-associated virus vector expressing angiostatin. Invest Ophthalmol Vis Sci 2001; 42: 2401-2407. MEDLINE

45 Blaauwgeers HGTet al. . Polarised vascular endothelial growth factor secretion by human retinal pigment epithelium and localisation of vascular endothelial growth factor receptors on the inner choriocapillaris. Am J Pathol 1999; 155: 421-428. MEDLINE

46 Lai YKY, Rolling F, Baker E, Rakoczy PE. Kinetics of efficient recombinant adeno-associated virus transduction in retinal pigment epithelial cells. Exp Cell Res 2001; 267: 184-192. Article MEDLINE

47 Lai YKY, Rolling F, Baker E, Rakoczy PE. Kinetics of efficient recombinant adeno-associated virus transduction in retinal pigment epithelial cells. Haberman RA, Kroner-Lux G, Samulski RJ. Production of recombinant adeno-associated viral vectors. In: Dracopoli NC et al. (ed.). Current Protocols in Human Genetics. John Wiley: New York, 1994, pp 12.9.1-12.9.17.

48 Lambert VI. Influence of plasminogen activator inhibitor type on choroidal neovascularisation. FASEB J 2001; 15: 1021-1027. MEDLINE

Figure 1 Expression, secretion and biological activity of sFlt-1 from AAV-CMV.sflt-transduced cells. (a) Western blot analysis of conditioned media from Ad.sflt-transduced 293 cells (lane 1), AAV-CMV.sflt-transduced D407 cells (lane 2), AAV-CMV.sflt-transduced 293 cells (lane 3), and AAV-CMV.gfp-transduced D407 cells (lane 4). (b) Inhibition of VEGF-induced HUVEC proliferation by conditioned media from AAV-CMV.sflt-transduced cells. HUVECs were cultured in starvation medium (), in medium containing recombinant VEGF (), in medium containing VEGF and 40 l conditioned medium from AAV-CMV.sflt-transduced 293 cells (), in medium containing VEGF and 80 l conditioned medium from AAV-CMV.sflt-transduced 293 cells (), and in medium containing VEGF and 80 l conditioned medium from AAV-CMV. gfp-transduced 293 cells (). *P < 0.02, **P < 0.005 for differences between AAV-CMV.sflt plus VEGF, and VEGF only.

Figure 2 In vivo expression of sFlt-1 following subretinal injection of AAV-CMV.sflt into rat eyes. Expression of human sFlt-1 mRNA at 3 months after injection was assessed by RT-PCR of chorioretinal tissue from AAV-CMV.sflt-injected eyes (372 bp, lane 1), an AAV-CMV.gfp-injected eye (lane 2), an uninjected control eye (lane 3) and negative RT control of AAV-CMV.sflt-injected eye (lane 4). M, molecular weight marker.

Figure 3 Inhibition of cautery-induced corneal NV by AAV-CMV.sflt injection into anterior chamber. (a) GFP expression following AAV-CMV.gfp injection into the anterior chamber was present in (i) the iris flat mount; (ii) the corneal flat mount; (iii) iris pigment epithelial cells (arrows) and occasional iris stromal cells (arrowheads); and (iv) corneal endothelial cells (inset). Box in haematoxylin-eosin stained section indicates position of GFP signal (inset). Original magnification: ´20 (i, iii and iv) and ´40 (ii and iv inset). (b) Paraffin-embedded section showing normal morphology of cornea following AAV-CMV.sflt injection. (c) Slit-lamp photograph of silver nitrate-cauterised eye injected with AAV-CMV.gfp. (d) Slit-lamp photograph of cauterised AAV-CMV.sflt-treated eye. (e) Mean NV score of cauterised eyes injected with AAV-CMV.sflt, AAV-CMV.gfp, PBS vehicle and uninjected. *P < 0.03, AAV-CMV.sflt-injected versus AAV-CMV.gfp-injected group.

Figure 4 Histopathology of corneas from AAV-CMV.gfp-injected (a, c, e) and AAV-CMV.sflt-treated eyes (b, d, f) at 4 days after cautery. (a) Presence of numerous blood vessels (black arrows) and infiltrating cells (arrowheads) at cautery site of AAV-CMV.gfp-injected eye. (b) Absence of blood vessels at the cautery site of AAV-CMV.sflt-injected eye. Black silver deposits were present at the cautery site (white arrows) and a few infiltrating cells (arrowheads). (c) Presence of some blood vessels (black arrows) and infiltrating cells (arrowheads) in peripheral cornea of AAV-CMV.gfp-injected eye. (d) Few infiltrating cells and no blood vessels in peripheral cornea of AAV-CMV.sflt-injected eye. (e) Positive immunostaining for vWF (open arrows) of AAV-CMV.gfp-injected eye. (f) Absence of vWF immunostaining in AAV-CMV.sflt-treated eye. Black silver deposits mark cautery site (white arrow). Counterstained with haematoxylin and eosin, original magnification ´20 (a-d); counterstained with haematoxylin, original magnification ´10 (e-f).

Figure 5 Inhibition of laser-induced choroidal NV by subretinal injection of AAV-CMV.sflt. (a) Fundus photograph of AAV-CMV.gfp injection site near equator (black arrow). (b) Corresponding fluorescence fundus photograph showing GFP expression confined within injection area. (c) Fundus photograph showing location of laser lesions (white arrows) near the optic nerve (asterisk), but away from the injection site (direction indicated by arrowheads). (d) Fundus photograph of same eye showing the position of injection site within the equator (black arrow) in relation to the optic nerve (asterisk). (e-h) Fluorescein angiograms of representative eyes at 5 and 16 weeks after laser photocoagulation. Less fluorescein leakage was detected in lesions of AAV-CMV.sflt-injected eyes at (e) 5 weeks, and (g) 16 weeks after photocoagulation. Stronger fluorescein leakage was detected in lesions from eyes injected with AAV-CMV.gfp at (f) 5 weeks, and (h) 16 weeks after photocoagulation.

Figure 6 Representative fluorescein angiograms of laser lesions. The intensity of fluorescein leakage of other lesions was graded according to this panel: (a) score 0, no leakage; (b) score 1, slight leakage; (c) score 2, moderate leakage; and (d) score 3, strong leakage.

Figure 7 Histological analysis of AAV-CMV.sflt-injected (a, c, e) and AAV-CMV.gfp-injected (b, d, f) eyes after laser treatment. Haematoxylin-eosin staining of a representative area of laser-induced lesion in AAV-CMV.sflt (a), and AAV-CMV.gfp (b) eyes. The B/C ratio was determined from maximum thickness of lesion (B) and thickness of adjacent normal RPE-choroid (C). Immunostaining of vascular endothelial cells by anti-vWF antibody (c and d) and fluorescein-conjugated GSA lectin (e and f). Staining was absent in lesions, but present in normal choroidal vessels (arrowheads) of AAV-CMV.sflt-injected eyes (c and e). Extensive NV detected by immunostaining (black arrows) in lesions of AAV-CMV.gfp-injected eyes (d and f), in addition to the normal choroidal vessels (arrowheads).

Table1 Incidence of corneal neovascularisation (score 1) in cauterized eyes injected with AAV-CMV.sflt, AAV-CMV.gfp, PBS and non-injected

Table2 Incidence of choroidal neovascularisation in AAV-CMV.sflt-injected and AAV-CMV.gfp-injected eyes after laser photocoagulation