Choroidal Neovascularization in the Rat Induced by Adenovirus Mediated Expression of Vascular Endothelial Growth Factor

Abstract

purpose. To determine the effects of an adenovirus vector encoding vascular endothelial growth factor₁₆₅ (Ad.VEGF) delivered to the subretinal space in the rat.

methods. An E1–deleted adenoviral vector encoding VEGF was injected into the subretinal space of Long–Evans rats. Immunohistochemistry identified VEGF expression. Histopathologic changes in the retina were determined by light and electron microscopy, immunohistochemistry, fluorescein angiography, and examination of wholemounts of choroid and retina.

results. Increased expression of VEGF only in the retinal pigment epithelium (RPE) was detected after Ad.VEGF injection. Histopathology of these eyes revealed minimal subretinal exudation at 1 week followed by the appearance of vascular structures in the subretinal space by week 2, which persisted up to 4 weeks. Shortening of photoreceptor outer segments and reduction of the outer nuclear layer were present overlying areas of neovascularization. Fluorescein angiography of animals injected with fluorescein–dextran revealed a deep complex of new vessels. Choroidal flatmounts showed new vessel formation, verified by detection of endothelial cells via immunohistochemistry, arising from the choroid with absence of change in the overlying retinal vasculature. Electron microscopy confirmed the presence of sub-RPE endothelial cells and pericytes and the loss of integrity of Bruch’s membrane, and serial sectioning demonstrated choroidal vascular growth through Bruch’s membrane.

conclusions. These results support the hypothesis that overexpression of VEGF from RPE cells is capable of inducing choroidal neovascularization in the rat and provide a framework for further examining angiogenic processes in the RPE–choroid complex.

Choroidal neovascularization (CNV) remains the leading cause of rapid severe vision loss in patients with age-related macular degeneration (AMD). ¹ Histopathologic studies of choroidal neovascular membranes from patients with AMD have demonstrated the presence of various angiogenic and growth factors, including fibroblast growth factor (FGF), ^{2 3} vascular endothelial growth factor (VEGF), ^{2 4 5 6 7} and transforming growth factorβ . ^{3 6} VEGF, which is normally expressed by the retinal pigment epithelium (RPE), ganglion cells, and the inner nuclear layer, ⁸ is thought to modulate retinal vascular permeability, ^{9 10} vasculogenesis, ¹¹ and neovascular proliferation. ¹² The protein is a secreted polypeptide that has five homodimeric species, which are formed as a result of alternative splicing. ^{13 14} Of the many splice alternatives the 165- and 121-kDa forms are most commonly expressed in the ischemic retina. ¹⁵  

Genetically altered animal models have been used to investigate clinically relevant stimuli for the development of CNV. A transgenic mouse expressing VEGF within the photoreceptors developed retinal neovascularization extending into the subretinal space. ¹⁶ However, these animals failed to produce CNV. ¹⁷ Gene transfer through viral transduction has been used successfully to alter protein expression within targeted ocular tissues. ^{18 19 20 21 22} Previous studies using adenovirus vectors have established that transduction can be localized to the RPE ^{23 24} after subretinal injection and that the vector can be administered in various doses to control the amount of transgene expression. ²⁴  

The purpose of the present study was to investigate the anatomic changes of rats injected subretinally with an adenovirus vector (Ad5.hCMV.VEGF₁₆₅) expressing the 165 isoform of VEGF. We sought to determine whether RPE-targeted VEGF expression can be accomplished, in the adult rat, and to analyze the subsequent anatomic changes that might occur. 

Methods

Adenovirus Injection

Recombinant adenovirus based on the human adenovirus serotype 5 lacking the E1 region and expressing either β-galactosidase (Ad.LacZ) or human VEGF₁₆₅ (Ad.VEGF; kind gift of I. Kovesdi, Genvec, Rockville, MD) driven by a CMV promoter were used throughout the study. Virus was generated to titers of 2 × 10¹⁰ to 2.5 × 10¹⁰ pfu/ml. For intraocular injections, viral stocks were diluted with sterile saline at various concentrations indicated in the text. All virus stocks used for the present study were free of contamination with wild type recombinant adenovirus as determined by three sequential passages on A549 cells (ATCC, Rockville, MD). ²⁵  

Animals

Long–Evans rats, 8 to 12 weeks of age, were obtained from NCI/DCT. Animals were handled in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Delivery of Adenovirus

Subretinal injection of adenovirus was performed as previously described ²⁴ except for the substitution of an automated injector (Hamilton, Reno, NV) to allow subretinal delivery of approximately 5 to 10 μl of the viral solution. A standard end point, the presence of a localized retinal detachment involving approximately 25% of the retina, was confirmed visually by fundoscopy. The retinal detachment spontaneously resolved within 3 to 5 days after injection. 

Immunostaining of the Retina

Animals were killed by CO₂ asphyxiation and transcardially perfused with saline, followed by 4% paraformaldehyde in phosphate-buffered saline (PBS). The eyes were enucleated and postfixed overnight. Tissue was embedded in paraffin, 5-μm-thick sections were cut, deparaffinized with xylene followed by rehydration with graded dilutions of ethanol, washed in PBS, and incubated in 1% bovine serum containing 0.6% Triton X-100. Slides were stained with a rabbit polyclonal antibody against factor VIII (DAKO, Glostrup, Denmark), VEGF (Santa Cruz Biotechnology, Santa Cruz, CA), or a mouse monoclonal anti-EMMPRIN (kind gift of E. Rodriguez–Boulan, Cornell University, New York, NY) and then exposed to an appropriate Cy3 or fluorescein isothiocyanate (FITC)–conjugated secondary antibody (Jackson ImmunoResearch Laboratories, West Grove, PA). Slides were washed with PBS, dried, mounted with anti-fade mounting medium (Vectashield, Vector Laboratories, Burlingame, CA), coverslipped, and viewed with an epifluorescence microscope (model BX50; Olympus Optical, Melville, NY) equipped with a cooled charge-coupled device camera. Images were digitally acquired using NIH Image 1.52 software and recompiled in Adobe Photoshop, version 5.0 (San Jose, CA). Sections stained with secondary antibody alone did not show reactivity (data not shown). 

Angiography/Choroidal and Retinal Flatmounts

Four animals were anesthetized and perfused intravenously with 4 ml of PBS containing 50 mg/ml of FITC-labeled dextran (Sigma, St. Louis, MO) as described previously. ²⁶ During the infusion, representative fluorescein angiograms were performed using a Kowa small animal fundus camera (Kowa–Optimed, Torrance, CA). The eyes were then marked for orientation, enucleated, and placed in 4% paraformaldehyde overnight. The anterior segment and the retina were removed, and the retina and choroid were cut to allow flatmounting with AquaPoly/Mount (Polysciences, Warrington, PA), coverslipped, and examined by fluorescence microscope as described above. 

Light and Electron Microscopy

Eyes were prepared as described above and embedded in methacrylate-JB4 (Polysciences). Three-micrometer-thick section were cut and counterstained with hematoxylin and eosin. For electron microscopy, eyes were prepared from four animals and fixed in 4% glutaraldehyde/PBS, postfixed with 1% osmium tetroxide-cacodylate buffer, dehydrated, and embedded in LADD LX-112 epoxy resin (Ladd Industries, Burlington, VT). One-micrometer-thick sections were stained with toluidine blue and examined with a light microscope (model BX50; Olympus Optical, Melville, NY). Ultrathin sections were stained with uranyl acetate and lead citrate and examined with a transmission electron microscope (JEOL-1010; Japan Electron Optic, Tokyo, Japan). 

Results

VEGF protein expression and localization 7 days after subretinal injection of either Ad.VEGF or Ad.LacZ were determined by immunohistochemistry. Eyes injected with Ad.VEGF generated strong VEGF immunopositivity from the RPE in areas adjacent to the injection site (Fig. 1) and extending laterally corresponding to the area of surgically induced retinal detachment, whereas eyes injected with an equivalent dose of Ad.LacZ produced only modest VEGF staining from the corresponding location (Fig. 1) . Weak staining for VEGF was noted in the inner nuclear and the ganglion cell layers, whereas an absence of photoreceptor staining was noted in both the Ad.VEGF transduced and Ad.LacZ-control injected eyes. 

Histologic evaluations were performed on Ad.VEGF- and Ad.LacZ-injected eyes at 1, 2, 3, and 4 weeks. At 1 week after viral injection of 2.5 × 10⁴ pfu of Ad.VEGF, most eyes demonstrated localized subretinal hemorrhagic exudation (Fig. 2A ) that appeared to correspond to areas of RPE–VEGF overexpression. The overlying RPE, outer segments, and outer nuclear layer appeared normal. No tubular structures suggestive of neovascularization were identified in the subretinal space. Ad.LacZ-injected controls rarely demonstrated subretinal blood (data not shown). The histologic changes observed at weeks 2 through 4 postinjection of Ad.VEGF were similar at the various times. Most eyes developed distinct regions of sub-RPE cellular tissue with well-formed lumens and occasional red blood cells (Figs. 2B 2C 2D , white arrows), typical for CNV. The associated RPE (arrowheads) was often morphologically abnormal, usually demonstrating flattening, migration, or separation from neighboring RPE. The overlying retina revealed extensive rod outer segment shortening and thinning of the outer nuclear layer (Figs. 2B 2C 2D) . 

A fluorescein–dextran angiogram was obtained in animals 4 weeks postinjection. Figure 3 demonstrates the typical appearance, showing a discrete zone of subretinal hyperfluorescence with a continuous margin and lacy center, consistent with neovascularization. The anatomic location appeared to be deep to an overlying retinal vessel. 

Histologic flatmounts of the retina and choroid, corresponding to the area seen by fluorescein angiography, were generated (Fig. 4) . Fluorescence microscopy of the choroidal flatmount, focused on the overlying lesion, revealed a well-defined network of interconnecting luminal structures characteristic of neovascularization emanating from the underlying choroid (Fig. 4A) . The corresponding retinal flatmount exhibited normal retinal vascular architecture without evidence of penetrating vascular communications to the subretinal neovascular complex (Fig. 4B) . 

To further study this neovascular complex, histologic sections (Fig. 5) were examined from the area illustrated in the previous choroidal flatmount (Fig. 4B) . A localized area of fibrovascular tissue with well-formed endothelial-lined channels (asterisks) associated with a moderate amount of extracellular matrix was seen. RPE cells (arrows) encompassed the complex with cells noted both above and below the lesion (Fig. 5A) . Factor VIII staining confirmed the identity of the endothelial cells (Fig. 5B) within the center of the complex, whereas an anti-EMMPRIN antibody, specific for RPE cells, ²⁷ confirmed that the overlying pigment-containing cells were RPE cells (Fig. 5C) . 

To confirm the localization and cellular components present in the subretinal space (Fig. 6A ), transmission electron microscopy was performed on areas of neovascularization in eyes 4 weeks postinjection. To more easily identify distinct cellular features a smaller area of CNV was chosen (Fig. 6A) . As can be seen in Figure 6B , endothelial cells and pericytes were identified in the space between the RPE and Bruch’s membrane consistent with sub-RPE neovascularization. Inspection of Bruch’s membrane neighboring these areas of neovascularization revealed a loss of integrity with areas of rupture (Fig. 6C) . 

Verification of the origin of the CNV was performed by examination of 3-μm-thick serial sections through the lesions. Figure 7 demonstrates an area of subretinal extension of a small choroidal vessel in a lesion seen 2 weeks after injection of Ad.VEGF. As can be seen in this light micrograph, vascular extension (asterisk) is taking place from the underlying choroid through a distorted Bruch’s membrane (white arrowheads) into the area underlying the RPE (dark arrowheads). 

To characterize the reproducibility and natural history of the vascular lesions induced by Ad.VEGF, presence, total number of CNVs in all eyes and average cross-sectional distance of luminal vascular lesions found in the subretinal space were noted. Table 1 summarizes these results from 16 eyes of rats injected with Ad-VEGF. None of the control eyes injected with a comparable amount of Ad.LacZ demonstrated pathologic changes (data not shown). As can be seen, all animals developed CNV 2 to 3 weeks after injection, but the number and size of these lesions were quite variable. However, at 4 weeks, only 50% of eyes showed signs of subretinal vascular lesions. There was a small reduction in the mean size of lesions between 2 and 3 weeks, which then remained essentially unchanged through 4 weeks. 

Discussion

Adenoviral vectors have proven useful in transducing retinal tissues. ^{19 21 22 28} Although the use of tissue-specific promoters in adenovirus gene transfer may not be sufficient to limit expression, ^{29 30} targeted surgical delivery of adenoviral vectors appears to aid in directing retinal cell–specific expression. Intravitreal delivery of these vectors results in ganglion and Müller cell expression, ¹⁸ whereas subretinal delivery into adult animals appears to preferentially infect RPE cells. ^{23 24 31} This infection of adenovirus into RPE cells is thought to occur by binding and internalization of virus particles bound to αvβ5 integrin sites ³² found on the apical surface of the cell. ^{33 34} Protein expression after adenoviral transduction has been reported as early as 48 hours after injection and may continue for well over 60 days in some models. ^{19 24 35} This RPE-specific overexpression of VEGF generated by adenovirus transfer, in the present model, was detected by immunohistochemistry at 1 week after subretinal delivery of the viral vector, a pattern not seen in the Ad.LacZ virus controls. VEGF was undetectable in the photoreceptor layer of both the control and Ad.VEGF-injected eyes. No secretion of VEGF from the RPE into the subretinal space was noted, with some staining extending into the basal surface of the RPE–Bruch’s membrane complex, suggestive of RPE basal secretion of VEGF. This finding is consistent with surface-specific polarization of VEGF expression as has been shown in Ad.VEGF-infected cultured RPE cells. ³⁶ Additionally, normal expression of VEGF from the RPE has been shown to occur on the basolateral aspect. ³⁷ This targeted secretion is thought to allow for binding to Flt-1 and KDR receptors on the adjacent choriocapillaris, ⁸ presumably to aid in choroidal vascular integrity. 

The mitogenic potential of injected VEGF to cause retinal neovascularization in animal models has been well demonstrated. ^{38 39} Although there is considerable evidence from the histopathology of humans that the upregulation of VEGF in RPE cells may play a role in the evolution of CNV, ^{2 4 5 6 7} the direct demonstration of RPE cell expressed VEGF inducing CNV has not been achieved. 

During the first week after Ad.VEGF injection, eyes later destined to develop neovascularization developed moderate subretinal fluid and localized subretinal hemorrhage. This effect is consistent with the actions of VEGF as a vascular permeability factor. ¹⁰ By 2 weeks, most eyes had developed areas of CNV as detected by histopathology. Eyes with CNV demonstrated extensive photoreceptor injury characterized by loss of outer nuclear layer and rod outer segment shortening. Similar changes have been observed in other rat models expressing high doses of VEGF from the RPE after viral transduction (G. Byrnes, personal communication) and is unlikely to be a manifestation of the adenovirus itself, because previous reports using LacZ, ²⁴ green fluorescent protein, ²³ or cathepsin S ²³ encoding adenovirus vectors targeted to the RPE failed to show similar changes. VEGF may have direct photoreceptor toxic effects. One line of photoreceptor expressing VEGF transgenic mice also demonstrated photoreceptor degeneration. ¹⁶  

To further evaluate the areas of CNV induced in this model, a fluorescein–dextran angiogram was performed. As is seen in human eyes with CNV, a network of vascular channels was detected in a location that appeared deep to the larger retinal vessels. Unlike photoreceptor-expressing VEGF transgenic mice, ^{16 17} which generated extensive areas of neovascularization extending from the retinal vascular bed, the Ad.VEGF-injected animals demonstrated primary vascular changes emanating from the choroid. Further inspection confirmed the vascular composition. The proliferation of RPE cells, seen to be encasing the lesion, has also been shown in laser-induced monkey ^{40 41} and FGF–microsphere rabbit ⁴² models of CNV. The presence of endothelial cells and pericytes, as is seen in human CNV, ⁴³ in the subretinal space was also noted in this model. Fragmentation of Bruch’s membrane is felt to be a prerequisite to allow for access of choroidal cells into the subretinal space, ⁴⁴ a finding that was also detected in this model in areas underlying neovascular tissue. Further demonstration of the source of these new vessels was achieved by examination of serial sections through the vascular complexes, a technique that is often used for detection of invading choroidal vessels in human pathology cases. ⁴⁴ As is typically seen in the eyes of humans with CNV, ^{44 45 46} the neovascularization invaded the subretinal space through a break in Bruch’s membrane. 

The natural history of these lesions, as detailed in Table 1 , suggests that, in the Long–Evans strain, CNV is induced within 2 weeks after VEGF overexpression. However, the number and size of lesions are quite variable. The number and average size of lesions are reduced at 3 weeks, with the number and average size of the lesion then appearing essentially unchanged at 4 weeks. The absence of lesions in 2 of the 4 eyes at 4 weeks suggests regression of some of the smaller neovascular complexes seen at weeks 2 and 3. Although the expression of VEGF in eyes with CNV was not measured, it is fair to assume, based on quantitative studies with β-galactosidase, ²⁴ that expression starts to decline after 2 weeks. In the present model this suggests that a transient expression of pathologic levels of VEGF is adequate to induce CNV but that once that level falls below a certain threshold for receptor activation, no further growth is noted. Alternatively, continuous production of VEGF may result in a loss of vascular proliferative effect, a finding demonstrated in experiments with continuous VEGF infusions in brain. ⁴⁷ Interestingly, though, is that most larger lesions remain stable without spontaneous involution. One could speculate that for CNV to progress and differentiate, other pathophysiological processes ⁴⁸ such as the release of other angiogenic factors, like angiopoietin, ⁴⁹ may be required. 

When VEGF is secreted into the subretinal space and outer retina, as produced by the transgenic mouse model, ^{16 17} stimulation of neovascularization occurs from the retinal vasculature but not from the choroid. However, when RPE-targeted expression of VEGF occurs, as in the present model, neovascularization appears to arise from the choroid. Several explanations exist for this finding. Adenoviral-targeted RPE expression of VEGF appears to result in directed secretion of VEGF into the choriocapillaris, a phenomenon that appears essential for binding to VEGF receptors in the choriocapillaris ⁸ and possible stimulation of neovascularization. Additionally, adenoviral vectors are known to induce an immune response and the recruitment of macrophages. ⁵⁰ Although no inflammatory cells have been seen in cross sections of eyes injected subretinally with adenoviral vectors, an immune activation is thought to occur. ³¹ The role of inflammatory cells in this model is now under investigation because VEGF itself may recruit macrophages expressing the Flt-1 receptor ⁵¹ and because macrophages have been shown to be associated with CNV in AMD. ^{52 53 54}  

Previous reports have established the ability of adenoviral vectors encoding VEGF to induce angiogenesis in the skin, ⁵⁵ muscle, ^{56 57} and myocardium. ⁵⁸ A report of the use of an adenovirus vector encoding murine VEGF₁₆₄ ⁵⁹ demonstrated intense leakage on fluorescein angiography after subretinal injection of 2 × 10⁸ pfu, an amount that was approximately 10,000 more than was injected in this model and appeared to have a predominant permeability rather than mitogenic effect on the vascular tissues. This pleiotropic effect of VEGF ⁶⁰ and other growth factors on cell proliferation is dependent on growth factor concentration. Several reports have demonstrated downregulation of epidermal growth factor receptor and growth inhibition after the addition of high doses of growth factor. ^{61 62 63} Thus, the ability to titrate the amount of protein expression using an adenoviral vector ²⁴ may play an important role in generating neovascularization in this model. 

Previous rat models of CNV have relied on the use laser thermal injury, ^{64 65 66 67 68} a stimulus that is not related to the pathogenesis of CNV associated with AMD. This study demonstrates that overexpression of VEGF within RPE cells after viral transduction in the rat eye reliably induces CNV, as demonstrated by fluorescein angiography, flatmount analysis, histopathology, and immunostaining. These findings further support the hypothesis that RPE-derived VEGF can induce CNV and may represent a more relevant model to study the process of CNV associated with AMD. 

 Submitted for publication April 3, 2000; revised June 13, 2000; accepted June 16, 2000.

 Commercial relationships policy: N.

 

Figure 1.

 

Cy-3-immunohistochemical detection of VEGF expression in vertical sections of rat retinas 1 week after subretinal injection of Ad.LacZ (A) or Ad.VEGF (B) using an antibody specific for VEGF, showing increased expression in the RPE layer after Ad.VEGF injection. GL, ganglion layer; INL: inner nuclear layer. Magnification, ×300.

Figure 2.

 

Light micrographs of plastic embedded sections at 1 week (A), 2 weeks (B), 3 weeks (C), and 4 weeks (D) after subretinal injection of Ad.VEGF showing subretinal exudation (dark arrow) at 1 week and development of CNV (white arrows) at weeks 2 through 4 with migration and flattening of the RPE (dark arrowheads). Note reduction of photoreceptor outer segments and the outer nuclear layer overlying areas of neovascularization (hematoxylin and eosin; magnification,× 400).

Figure 3.

 

Fluorescein–dextran angiogram of an eye 4 weeks after subretinal injection of Ad.VEGF showing vascular complex (asterisk) deep to an overlying retinal vessel (arrow).

Figure 4.

 

Flatmounts of choroid (A) and corresponding retina (B) of animals 4 weeks after injection with Ad.VEGF following perfusion with fluorescein–dextran. (A) Focus plane is on an area of an elevated neovascular complex (arrow) attached to the choroid (asterisks). (B) Intact retinal vascular filling can be seen in the region (asterisk) overlying the area of neovascularization seen in (A). Magnification, ×100.

Figure 5.

 

Morphology (A) of vertical sections cut from choroidal flatmounts from eyes 4 weeks after subretinal injection with Ad.VEGF showing lumen (asterisk) and matrix filled tissue overlying the choroid with surrounding migrated RPE cells (arrows). Corresponding immunohistochemical stain with Cy3 labeling of factor VIII–positive cells surrounding the luminal structures (asterisks; B) or FITC-labeled anti-EMMPRIN detection of surrounding RPE cells (C). Hematoxylin and eosin staining (A). Magnification, ×600.

Figure 6.

 

Ultrastuctural features of blood vessels in the subretinal space 4 weeks after subretinal injection of Ad.VEGF. (A) Toluidine blue staining demonstrates luminal structures under RPE layer and outer segments of the photoreceptors (OS) and above the choriocapillaris (CC). Electron microscopic view (B) of the area outlined in (A) shows endothelial cells (En) of newly formed vessels (L) with surrounding pericytes (P). Higher power view (C) of Bruch’s membrane (Br) underlying the areas of vessel formation demonstrated multiple breaks (arrowheads) and disruption. Magnification, (A) ×630; (B)× 6,000; (C) ×12,000.

Figure 7.

 

Light micrograph of a plastic embedded section obtained from 3-μm-thick serial sectioning of a neovascular lesion at 2 weeks after subretinal injection of Ad.VEGF showing growth of a small choroidal vessel (asterisk) through a disrupted Bruch’s membrane (white arrowheads) with migration and flattening of the RPE (dark arrowheads). Hematoxylin and eosin; magnification, ×630.

Table 1.

 

Incidence, Number, and Size (Mean ± SEM) of CNV in Eyes at Various Points after Injection of 2.5 × 10⁴ pfu of Ad.VEGF

Time Point, wks	Eyes with CNV	Total No. of CNV	Average Size of CNVM/Eye
1	0/4	—	—
2	4/4	16	234 ± 35 μm
3	4/4	10	141 ± 25 μm
4	2/4	11	133 ± 12 μm

The authors thank Mary Alice Crawford and Joseph Hackett for tissue preparation, sectioning, and electron microscopy and Scott Cousins for critical review of the manuscript.