Enhanced Recombinant Adeno-Associated Virus-Mediated Vascular Endothelial Growth Factor Expression in the Adult Mouse Retina: A Potential Model for Diabetic Retinopathy

The construction of recombinant adeno-associated virus (rAAV)/green fluorescent protein (GFP) and rAAV.VEGF.GFP have been previously described (15). Following the appropriate validation of the constructs, large-scale production of the viruses was performed at the Vector Core Facility (University of North Carolina, Chapel Hill, NC). Titers of the virus were 4 × 10¹⁰ transducing units/ml and 1 × 10¹⁴ particles/ml for rAAV.GFP and rAAV.VEGF.GFP, respectively.

Animals were handled in accordance with The Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research. Five-week-old C57BL mice were anesthetized and subretinally injected with 1 μl solution containing 10¹¹ particles of rAAV.VEGF.GFP as previously described (15). Following subretinal injection, the presence of a localized retinal detachment (bleb) of ∼25–40% of the retina was confirmed by color fundus photography. Control eyes were injected with rAAV.GFP in the same way.

Photography was conducted on anesthetized animals at 2, 4, 8, and 12 weeks postinjection. Animals were anesthetized by intraperitoneal injection of ketamine (50 mg/kg body wt) and xylazine (8 mg/kg body wt) before pupil dilation with a drop of 2.5% phenylephrine hydrochloride and 1% mydriacyl. Artificial tears were used on all animals to prevent corneal drying. A small animal fundus camera (Kowa Genesis, Tokyo, Japan) was used to photograph mouse fundi using Kodak Elite 200 ISO film (Kodak, Rochester, NY). Fluorescence photography was used to image GFP expression using the small animal fundus camera (Kowa Genesis) with fluorescein filter and Kodak P1600 ISO film (Kodak). Fluorescein angiography was carried out on anesthetized animals following an intraperitoneal injection of 10% sodium fluorescein. The fluorescein angiography finding was recorded using the small animal fundus camera (Kowa Genesis). On the basis of the angiograms, eyes were categorized and leakiness was defined as minor, moderate, or severe according to the size of hyperfluorescent leakage.

Fluorescein angiograms taken at 4 weeks postinjection of rAAV.VEGF.GFP- and rAAV.GFP-injected eyes were scanned into PhotoShop (Adobe, San Jose, CA) then opened in Scion Image (Scion, Frederick, MD). Venous blood vessels were outlined and their area and length digitally calculated. Blood vessel area was divided by blood vessel length to give a mean diameter for blood vessels that could be compared across animals. Mean blood vessel diameters, measured in pixels, were averaged for rAAV.VEGF.GFP- and rAAV.GFP-injected eyes, and the significance of the difference was analyzed by a one-tailed Student’s t test.

Ten rAAV.VEGF.GFP-injected eyes were enucleated at 12 weeks postinjection for different histological studies. rAAV.GFP-injected eyes were enucleated at 4 and 12 weeks postinjection. The eyes were whole mounted, paraffin embedded, or snap frozen following routine protocols (16,17).

VEGF immunohistochemistry was conducted on whole mounts as previously described (16). VEGF immunohistochemistry on frozen sections was performed by incubating the sections in rabbit anti-VEGF primary antibody (1/100; Sigma, St. Louis, MO), followed by LINK secondary antibody (DAKO, Carpinteria, CA), streptavidin-horseradish peroxidase (HRP) (DAKO), then developed with AEC (3-amino-9-ethyl-carbazole; DAKO).

Von Wellibrand factor primary antibody was used to detect the endothelial cells surrounding capillaries, and proliferating cell nuclear antigen (PCNA) primary antibody was used to detect proliferating cells in four eyes. Paraffin sections were incubated in rabbit anti-human von Wellibrand factor primary antibody (1/100; DAKO) or mouse anti-human PCNA primary antibody (1/50; DAKO), washed, and then incubated in fluorescein isothiocyanate-conjugated goat anti-rabbit secondary antibody (1/50; Zymed, San Francisco, CA) or HRP-conjugated sheep anti-mouse secondary antibody (1/100; Amersham Biosciences, Buckingham, England), respectively. Paraffin sections stained with PCNA were developed with AEC (DAKO).

Four eyes were enucleated at 12 weeks postinjection and were sonicated in 200 ìl of PBS/0.1% Triton X-100. After centrifugation at 1,500g for 10 min, the supernatant was collected and the amount of VEGF present was assayed using the enzyme-linked immunosorbent assay (ELISA) kit (R&D Systems, Minneapolis, MN).

von Willebrand factor (DAKO)-immunostained sections were photographed at 100× magnification, and the photographs from the same section were tiled together to form a picture of the entire section. In three rAAV.VEGF.GFP-injected and two rAAV.GFP-injected eyes, capillaries interior and exterior to the ganglion cell layer (GCL) were counted in two 5-μm thick paraffin sections that were ∼200 μm apart and contained retinas of equal length (18). Means and standard errors of capillary number were calculated for GCL and non-GCL areas in rAAV.VEGF.GFP-injected and rAAV.GFP-injected eyes. One-tailed Student’s t tests of capillary number were conducted between the treatment and control groups in GCL and in non-GCL areas.

All injected eyes demonstrated subretinal bleb formation (Fig. 1A). The subretinal fluid was absorbed by 5 days postinjection, with the disappearance of the retinal detachment. The fundus looked normal, and only a small scar remained visible at the injection site (data not shown). Fluorescence fundus photography proved to be an unreliable method for tracking rAAV-mediated GFP expression in vivo. Only 16% of rAAV.VEGF.GFP and 33% of rAAV.GFP-injected eyes (Fig. 1B) demonstrated GFP levels high enough to be detected by fluorescence fundus photography between 4 and 12 weeks postinjection. In contrast, 100% of whole-mounted rAAV.GFP and rAAV.VEGF.GFP-injected animals demonstrated GFP expression, as observed by fluorescence microscopy when sampled at 2 weeks postinjection and at different time points up to 12 weeks postinjection. GFP expression in injected eyes was specific to photoreceptor and retinal pigment epithelial (RPE) cell layers. Fluorescence microscopy of whole-mounted RPE-choroid layers and neural retinas demonstrated the expression of GFP in RPE cells (Fig. 1C) and photoreceptors (Fig. 1D), respectively. The expression of GFP in the RPE and photoreceptor cells was confirmed by fluorescence microscopy of frozen sections (Fig. 1E). Apart from the presence of scar tissue at the injection site, retinal morphology in injected eyes was normal and no morphologic changes or infiltrating cells were observed throughout the course of the study (Fig. 1F).

Retinal vasculature was imaged by fluorescein angiography. No vascular changes were recorded in uninjected control eyes (n = 5) during the study period (data not shown). Eyes injected with rAAV.GFP (n = 5) showed normal vasculature (Fig. 2A). Changes in retinal vasculature were observed in rAAV.VEGF.GFP-injected eyes (Fig. 2B–F). These changes were defined as venous dilatation, microaneurysm, or leakage. Analysis of blood vessel size in fluorescein angiograms showed dilatation of venous retinal vessels following rAAV.VEGF.GFP injection (Fig. 2B). Mean venous blood vessel diameter of rAAV.GFP-injected eyes was 19.74 ± 2.03 pixels and that of rAAV.VEGF.GFP-injected eyes was 24.48 ± 2.33 pixels (P < 0.05). This increase represented a 24% increase in vessel diameter in rAAV.VEGF.GFP-injected eyes (Fig. 2B) when compared with rAAV.GFP-injected eyes (Fig. 2A). Fluorescein angiography performed on rAAV.VEGF.GFP-injected eyes (n = 10) demonstrated vascular changes, described as leakages or microaneurysm in 90% of these eyes (Table 1), appearing at 2–4 weeks postinjection. Microaneurysms were visible in four eyes, while minor to moderate leakage was observed in six eyes (Table 2). Severe leakage was noted in one animal, and one animal did not demonstrate any vascular changes (Table 2). Microaneurysms appeared as pinpoint hyperfluorescent dots that faded in the later phases of fluorescein angiography (Fig. 2C, arrows). In eyes where leakage was observed, the extent of leakage was defined as minor (Fig. 2D, small arrows), moderate (Fig. 2E), or severe (Fig. 2F), based on the intensity of fluorescein leakage.

VEGF expression was monitored by immunohistochemistry and quantified by ELISA from pooled retinas. No immune reactivity to the VEGF antibody was detected in whole mounts of rAAV.GFP-injected and noninjected eyes (Fig. 3A and B). Whole-mounted retinas showed a strong immune reaction to VEGF antibody around the injection site in rAAV.VEGF.GFP-injected retinas, covering up to 30% of the total retinal surface (Fig. 3C). Immunohistochemistry of paraffin-embedded sections of rAAV.GFP-injected eyes demonstrated the presence of low level VEGF in the GCL and in the inner nuclear layer (INL) (data not shown). In contrast, in paraffin-embedded sections of rAAV.VEGF.GFP-injected eyes, the majority of VEGF immunoreactivity was present in the outer nuclear layer (ONL), the inner segments of the photoreceptor cells (Fig. 3D, thick arrow), and to a lesser degree in the RPE cells (Fig. 3D, thin arrow). These observations were in agreement with the transduction specificity of rAAV and confirmed that the increased level of VEGF in these eyes was the result of rAAV-mediated gene delivery. Increased VEGF immunoreactivity was correlated with an increased amount of VEGF (10.37pg VEGF/ìg total protein), quantitated by ELISA, in rAAV.VEGF.GFP-injected retinas that demonstrated leakiness. Although the expression of endogenous VEGF in the retina has been previously reported (19,20), we were unable to detect endogenous VEGF in the retina of uninjected or rAAV.GFP-injected animals.

Histological examination of rAAV.VEGF.GFP-injected eyes (n = 10) demonstrated normal retinal morphology in nine eyes (Fig. 4A). In one eye significant fluorescein leakage was seen at 4 weeks postinjection, and retinal degeneration was observed by 12 weeks postinjection. The retinal degeneration coincided with the growth of blood vessels from the retinal vasculature into the subretinal and inner plexiform layers (Fig. 4B). There was no statistically significant difference (P = 0.225) in the number of blood vessels present in the GCL between rAAV.GFP-injected and rAAV.VEGF.GFP-injected animals (Fig. 5). However, there was a statistically significant increase (P < 0.03) in the number of retinal blood vessels of the INL (82.87 ± 5.09) in rAAV.VEGF.GFP-injected eyes showing microaneurysm, or minor to moderate leakage when compared with the control rAAV.GFP-injected eyes (65.00 ± 7.95) (Fig. 5). The change was most pronounced around the injection site, where the increased number of capillaries covered an area approximately the size of the retinal bleb (Fig. 4D). In these eyes the presence of PCNA-positive cells was also demonstrated in the INL (Fig. 4F). When compared with an area of the same size away from the injection site (Fig. 4C), the number of capillaries decreased to a level similar to that observed in rAAV.GFP-injected control eyes and there were no PCNA-positive cells present. Control rAAV.GFP-injected eyes did not have any PCNA-positive cells in the INL (Fig. 4E).

This work has demonstrated that long-term specific increase of VEGF expression in the adult retina could induce limited vascular changes resembling those of diabetic retinopathy. VEGF is a potent vascular permeability factor (21–23), and VEGF upregulation has been linked to neovascular eye diseases including diabetic retinopathy. VEGF-induced neovascular changes have previously been demonstrated on animal models based on increasing VEGF levels through implants (24), recombinant adenovirus-mediated VEGF expression (25,26), or transgenic technologies (27,28).

Studies in which VEGF protein is directly administered require the use of complicated delivery systems, as the half-life of the VEGF protein is relatively short (29). Recombinant adenovirus-mediated delivery has successfully overcome this problem by providing consistent VEGF production (25,26). However, recombinant adenoviral vectors elicit an immune response resulting in transient transgene expression, and in first-generation recombinant adenoviruses the transgene expression cannot be expected beyond 2 weeks. In addition, the inflammatory response may enhance the neovascular response, thus influencing the final outcome (30). The use of an adeno-associated viral vector could overcome the problem, as it could maintain long-term VEGF expression without evoking an immune response (31). Previous studies demonstrated GFP expression for up to 18 months (32,33) without any signs of retinal toxicity or inflammation. In this work, the simultaneous rAAV-mediated expression of GFP and VEGF was demonstrated for up to 3 months without any significant change in GFP signal intensity. One of the advantages of the binary virus complex (rAAV.VEGF.GFP) was the ability to monitor the expression of the biologically active transgene, VEGF, through noninvasive monitoring of GFP expression (34,35). Although, the sensitivity of our in vivo set up was not always sufficient to reflect the level of GFP expression, the system has provided valuable information for experimental follow-up.

Transgenic mouse models with high-level VEGF expression were reported to develop leaky blood vessels and intraretinal endothelial cell proliferation (27,28,36). Subsequently, these animals rapidly developed severe proliferative retinopathy and retinal detachment. In our model, the process of new vessel development was relatively more gradual and less pronounced in rAAV.VEGF.GFP-injected adult mice than in any of the known VEGF transgenic models. This difference closely correlated with the lower level of ocular VEGF detected in rAAV.VEGF.GFP-injected animals. In the majority of our animals, VEGF protein expression was sufficient to manifest features of background retinopathy but not those of proliferative retinopathy. Considering the long-term presence of increased VEGF concentration, the subtle angiogenic changes induced in 80% of injected eyes were surprising. Although the venous blood vessels were dilatated with a 24% (P < 0.03) increase in vessel diameter in rAAV.GFP.VEGF-injected eyes and there was a statistically significant increase in the number of retinal blood vessels in the deep capillary of the retina, the majority of eyes showed no signs of histological or morphologic changes in the retina. These observations suggest that although a high level of VEGF would induce complete angiogenesis, VEGF is only one of the angiogenic factors participating in the neovascular response. Like all biological processes, angiogenesis results from subtle and complex interactions between regulatory and effector molecules. VEGF is one of the regulatory cytokines along with IGF, basic fibroblast growth factor (bFGF), and platelet-derived growth factor (13,37–39). Extracellular proteinases and their inhibitors, such as matrix metalloproteinases (MMPs) and plasminogen activators (PAs), are affected by these regulatory proteins (40). MMPs and PAs are responsible for the degradation of the extracellular matrix, an essential process initiating endothelial cell proliferation, thus enhancing the neovascularization process (41).

From a circulatory point of view, diabetes is a disease characterized by microvascular abnormalities resulting in hypoxia. In the eye, one of the earliest signs of diabetic retinopathy is retinal capillary occlusion, blocking blood flow and generating capillary-free areas (42). Hypoxic conditions could develop in these capillary-free areas, and this in turn could induce the upregulation of angiogenic factor production, such as VEGF, bFGF, and intercellular adhesion molecules (43,44). The increased concentration of angiogenic factors would then cause vascular changes including vascular dilatations, tortuous blood vessels, microaneurysms, and endothelial cell proliferation. Subsequently, over an extended period of time, these changes could result in the development of poorly matured leaky vessels (23,45,46). The most characteristic features of rAAV.VEGF.GFP-injected eyes were dilatation of venous blood vessels, microaneurysms, and minor to moderate leakages. Dilatation of the retinal venous blood vessels is among the most common and earliest manifestations of diabetic retinopathy and can be present in the absence of other manifestations of the disease (47). Fluorescein angiograms of 8 of 10 rAAV.VEGF.GFP-injected eyes demonstrated microaneurysms and/or minor to moderate fluorescein leakage. Previous histological studies have demonstrated a strong correlation between endothelial cell proliferation, pericyte loss, and the development of microaneurysm (48). Incidentally, VEGF, which is a known factor of endothelial cell proliferation, has also been shown to promote pericyte detachment and loss (49). Although, our present understanding of microaneurysm development remains incomplete, VEGF has recently been directly linked to the formation of these vascular structures. Similar to the mouse model described here, thin-walled hypercellular structures resembling diabetic microaneurysms were described in a monkey model of VEGF-induced retinopathy (50). Although by 12 weeks postinjection there was a statistically significant increase in the number of retinal vessels, these blood vessels were present only at the area corresponding to the injection or VEGF expression site, and even within this region they were limited to the INL. The lack of new vessel formation in the GCL might be due to the uneven distribution of VEGF across the retina. Retinal fluid flow, dominated by the movement of material from the GCL toward the RPE, is expected to reduce secreted VEGF concentration in the GCL layer. Spatially, with increasing distance from the injection site, changes were limited to dilatation and microaneurysm. Considering that rAAV-mediated gene delivery can last for the lifetime of the animal (32), rAAV-mediated gene delivery can potentially provide a suitable mouse model to study the long-term effect of early vascular changes on retinal function.

It is interesting to note that 1 of 10 eyes did not demonstrate an angiogenic response. In contrast, another eye showed a strong neovascular response with severe fluorescein leakage. The GFP signal in vivo and in whole mounts was similar in these eyes, suggesting that the efficacy of gene delivery was comparable with the other eight eyes. Histological examination of the eye demonstrating severe leakage showed extensive endothelial cell proliferation and retinal neovascularization resulting in retinal detachment and the development of a neovascular membrane. This variation in angiogenic response could be due to some experimental complications, such as bleeding, that might induce an initial inflammatory response enhancing the effect of VEGF. Alternatively, it might be due to the response to VEGF. Further studies using larger numbers of animals will be required to investigate this possibility.

In summary, this work has confirmed the essential role of VEGF expression in diabetic retinopathy and demonstrated that the development of different stages of diabetic retinopathy is closely correlated to the presence of VEGF.