Nonhuman Primate Models for Diabetic Ocular Neovascularization Using AAV2-Mediated Overexpression of Vascular Endothelial Growth Factor

The human VEGF cDNA (Genbank no. AF486837) was cloned from the Quick-Screen human cDNA library panel (Clontech, Palo Alto, CA) by PCR amplification using the following primers: forward, GCGGCCGCatgaactttctgctgtcttg ggtgcattgg; and reverse, AAGCTTtcaccgcctcggcttgtcacatctgcaagta.

The PCR product was cloned into the pCR2.1 plasmid using the TOPO cloning kit (Invitrogen, Carlsbad, CA) according to manufacturer instructions. After sequence verification, the hVEGF coding sequence was cut from pCR2.1 using NotI and HindIII and cloned into a pAAV2-cytomegalovirus (CMV) backbone. AAV vectors were produced by triple transfection followed by heparin purification (38). Genome copies were assessed by real-time PCR and infectious copies by the infectious center assay. Expression of VEGF encoded by AAV.VEGF was verified through analyses of the supernatant after transfection of 293 cells using a commercially available enzyme-linked immunosorbent assay kit (R & D Systems, Minneapolis, MN). As control vector, an AAV containing enhanced green fluorescent protein (EGFP) in reverse orientation, AAV.EGFP-RO, was used.

Animals used in this study were cared for in accordance with institutional and national regulations. Before treatments, baseline ophthalmic examinations, fundus photography and fluorescein angiographies were performed. Complete blood count, serum chemistry, and body weight were obtained before treatment and weekly posttreatment. Anterior chamber fluids were collected at designated time points before and after vector injection.

All procedures were performed using sterile conditions and complete anesthesia. Two monkeys (female rhesus; Covance, Denver, PA) were injected subretinally and two additional ones intravitreally as described (37). The left eyes of all animals were injected with AAV.VEGF and the right eyes with an equal dose of AAV.EGFP-RO. Injection volumes were the same in eyes of the same animals and ranged from 150 to 200 μl in all of the animals. Vector dosages are described in Table 1.

Animals were followed for up to 1 year after injection (Table 1) and, during the course of this time, received regular clinical examinations including ocular examinations. Changes in retinal and choroidal vasculature were documented through fluorescein angiograms. Photographs were taken with a Kowa Genesis camera (Keeler Instruments, Broomall, PA).

Changes in retinal microstructure of the experimental and control eyes were evaluated in vivo with optical coherence tomography (OCT3; Humphrey Instruments, San Leandro, CA). The principles of the method (39) and our techniques (40,41) have been previously published. Postacquisition processing of OCT data was performed with custom programs (MATLAB 6.5; MathWorks, Natick, MA). Retinal thickness at some loci in the VEGF-treated retina showing loss of signal contrast due to edema was specified manually. A full retinal thickness map was generated by bilinear interpolation and used a pseudocolor scale.

Eyes were surgically enucleated at designated time points while animals were under deep anesthesia (isoflurane). (Because the animals were simultaneously enrolled in another unrelated study they were not killed at the time of tissue harvest. In addition, only one eye was harvested per animal.) The eyes were fixed in 4% paraformaldehyde in PBS. The macula and surrounding (uninjected) retina was isolated using a trephine. One-half of the tissue was processed for cryosections, and the other was postfixed in 4% paraformaldehyde and 0.5% glutaraldehyde/PBS before processing for plastic sections. Additional regions in the retinal periphery were processed for cryosections or plastic sections. Cryoprotection was achieved by incubating samples in 30% sucrose/PBS before embedding and freezing in optimal cutting temperature compound (Fisher Scientific, Pittsburgh, PA). Serial sections were cut with a Reichert-Jung model 1800 cryostat (Leica Microsystems, Wetzler, Germany) at 10 μm.

For immunohistochemistry, antibodies against human VEGF (rabbit polyclonal no. A-20 and no. 147; Santa Cruz Biotechnology, Santa Cruz, CA) and 7G6 (mouse monoclonal antibody labeling cone photoreceptors; gift from P. Macleish) were used as primary and the respective fluorescently labeled rabbit and mouse antibodies as secondary (Alexa Fluor 488 and 594; Molecular Probes). Negative controls were performed by omission of the primary antibody. All immunohistochemistry sections were counterstained with 4′,6′-diamidino-2-phenylindole to stain cell nuclei (blue, 1 μg/ml; Molecular Probes).

Sections were evaluated on a Leica DMR microscope, and images were captured with a Hamamatsu charge-coupled device camera (Hamamatsu Photonics, Bridgewater, NJ) coupled to an Apple Macintosh G4 computer (Cupertino, CA) equipped with OpenLab 2.2 software (Improvision, Boston, MA).

Immediately after subretinal injections of AAV, a “bleb” was observed corresponding to a detachment of the neural retinal from the underlying RPE (Fig. 1, arrowheads). The size of the bleb correlated with the injection volume (Table 1). The detachments resolved spontaneously and were no longer apparent within 2 days of the injection. Animals had normal visual behavior, and pupils were reactive to light and symmetrical from eye to eye.

The experimental and control retinas of the animal subretinally treated with the “lower dose” of virus (animal 1, Table 1) were unremarkable in appearance until 21 days after injection. At that point, abnormalities in retinal vasculature were apparent only in the virus-exposed region of the eyes injected with AAV.VEGF but not in the periphery of the same or the control eye (Fig. 1). Fluorescein angiography confirmed the presence of CNV in the experimental but not the control-treated eyes (Fig. 2A–D). Figure 2D shows leakage of fluorescein in the AAV-VEGF–exposed outer retina, while the inner retinal vessels are intact (appearing black, i.e., nonleaky, over the fluorescein-stained outer retina). The leakage occurs only in the region of subretinal injection (compare the circular fluorescein-stained region with the circular region showing retinal vessel changes in Fig. 1, top panel, day 23).

Similar results were observed in the animal treated subretinally with the higher dose of AAV.VEGF (animal 3, Table 1), except that retinal vasculature abnormalities and CNV were apparent earlier (7 days after treatment) and were more pronounced. Again, there were no abnormalities identified in the anterior segment structures of experimental or control-injected eyes.

In the other cohort, AAV was injected intravitreally directly above the optic nerve head. There were no immediate effects of the injection on the underlying retina/RPE. The retinas were well perfused immediately after the injection and throughout the course of the study.

By 3 weeks after injection of the low vector dose, iris neovascularization was apparent in the eye injected intravitreally with AAV.VEGF (animal 2, Table 1), which was confirmed by fluorescein angiography (Fig. 2E and F). The pathology slowly progressed, finally leading to a decreased pupillary light reflex and ectropion uveae 8 months after vector injection. There were no signs of retinal neovascularization or CNV in the eye treated with the lower dose of AAV.VEGF over the year that it was studied.

In contrast, local regions of retinal neovascularization were observed in the animal injected intravitreally with the higher titer of AAV.VEGF (animal 4, Table 1) within 7 days of treatment. Fluorescein angiograms documented this retinal neovascularization but had poor resolution due to leakage of fluorescein from the iris neovascularization and a vascularized pupillary membrane, which was present at the same time (Fig. 2F). There were no abnormalities identified in the anterior segment structures or retinas of the control eyes of these monkeys.

As seen in human diabetic retinopathy in all animals subretinally or intravitreally treated, pathological changes steadily increased over the study period.

Microstructural abnormalities in the retina and their spatial distribution were defined in vivo with OCT, an objective method used increasingly in the clinic to define and safely monitor ocular micropathology in human clinical trials. Scans performed 60 days after a subretinal injection of the low-dose AAV.VEGF preparation showed gross abnormalities in retinal microstructure in the experimental eye compared with the contralateral control eye (Fig. 3). Abnormalities included retinal thickening that was the greatest near the fovea (thickness ∼1.3 mm as opposed to ∼0.18 mm in the control eye) and extended from there in all meridians (Fig. 3A). The extent of abnormalities corresponded approximately to the size of the bleb formed during the subretinal injection. Intraretinal edema and cystic structures and abnormality of the reflections originating from photoreceptor and RPE layers were apparent (Fig. 3B). There were no detectable microstructural abnormalities in the control eye that was injected subretinally with the control virus.

Histopathological findings mirrored the results of OCT. Appearance of the regions of the retinas unexposed to AAV.VEGF (i.e., the regions of the retina peripheral to the subretinal injection) were indistinguishable from those of the untreated control monkey. In these regions, there were no alterations in the thicknesses of the various retinal layers or abnormalities in the retinal/choroidal vasculature. There was a sharp boundary between AAV.VEGF-unexposed and -exposed regions, where there were gross abnormalities consisting of large cystic structures and enlarged inner and outer retinal vessels distorting the retinal cell layers and increasing the retinal thickness. In addition, CNV, defined by blood vessels emanating from the choroidal space and traversing Bruch’s membrane, and the RPE, was found in numerous regions (Fig. 4). These changes are not apparent in the region outside of the subretinal injection site region.

All these changes were accompanied by a disturbance in the architecture of the photoreceptor cells as noted by immunofluorescence. The outer segments were shortened, and the cell bodies were stretched due to the presence of the cystic spaces. There was degeneration of photoreceptor cells as evidenced by decreases in the numbers of these cells, loss of lamination of the retina, and thinning of the outer nuclear, outer plexiform, and outer segment layers.

There was no evidence of changes in thickness or number of blood vessels in anterior segment structures, including the iris and cornea.

Iris neovascularization limited the ability to perform detailed in vivo retinal testing (such as fluorescein angiography, OCT) in the animal intravitreally injected with the high dose of AAV.VEGF. Therefore, the eye was enucleated for histological evaluation. Focal hemorrhages were identified in the iris of this animal in addition to a few blood vessels in the normally avascular cornea. Neovascularization of the optic disc was appreciated. In addition, focal hemorrhagic blister-like regions were observed in the subretinal space of the peripheral retina (Fig. 5). Inflammatory cells were present in high numbers around the retinal blood vessels and in the hemorrhagic regions (data not shown). Despite these changes, there was no retinal degeneration apparent in eyes injected intravitreally with AAV.VEGF.

After treatment, significantly elevated levels of VEGF were detected in the anterior chamber fluid of every AAV.VEGF-treated eye as compared with titers in the contralateral control eyes using a commercially available enzyme-linked immunosorbent assay kit for human VEGF. Approximately 5- to 20-fold higher VEGF levels were detected in eyes injected intravitreally than in eyes injected subretinally with the same virus/dose of AAV.VEGF. Treatment with the higher dose of vector resulted in a sharp increase in VEGF levels followed by a marked drop in levels thereafter, whereas treatment with the lower dose of vector resulted in lower but more stable levels of VEGF, which only slowly declined after a period of several months (Fig. 6).

Immunofluorescence was used to identify cells containing VEGF. VEGF is present endogenously in the ganglion cell layer and in outer segments of photoreceptors in the untreated retina (Fig. 7A), and therefore immunofluorescence could only identify qualitative changes in the distribution of VEGF. Twenty-one days after high-dose (Fig. 7B) or low-dose subretinal (Fig. 7C) AAV.VEGF, VEGF became distributed over the entire photoreceptor cell body and increased relative to baseline in the retinal pigment epithelium. Relative VEGF levels were increased in the ganglion cell layer 21 days after low-dose intravitreal injection of AAV.VEGF (Fig. 7D), while high levels of VEGF persisted in photoreceptor outer segments. Relative VEGF levels in photoreceptor outer segments decreased after subretinal injection (Fig. 7B and C), as many of the expressing cells degenerated and became stretched due to cystoid macular edema.

All animals maintained normal activity and maintained their weights throughout the study. Serum clinical chemistry and hematology remained within normal limits. There was no detectable increases in serum VEGF levels in animals treated intraocularly with the lower dose of AAV.VEGF. A slight increase was observed in animals treated with the higher dose of AAV.VEGF, but this was well below levels measured in ocular fluids (data not shown).

In this study, ocular neovascularization was induced in nonhuman primates after injection of an AAV2/2 carrying the human VEGF cDNA under control of a constitutive promoter. Two different vector doses and injection routes were chosen to determine the best model imitating the features of diabetic ocular neovascularization as seen in humans.

At a low dose, VEGF gene transfer induced two very different phenotypes, depending on the injection route: iris neovascularization after intravitreous or subretinal neovascularization and cystoid macular edema after subretinal injection. This is surprising because VEGF is a secreted protein and one might expect it to affect multiple locations in the eye through diffusion. However, in other studies where VEGF protein was injected intravitreally, only inner retinal neovascularization (and not subretinal neovascularization) was found (29).

Related to this, we measured close to 20-fold lower levels of VEGF in the anterior chamber fluid after subretinal than after intravitreal injection of the same vector dose. This suggests that VEGF produced after subretinal injection does not readily diffuse across the multilayered retina either because it is too large or because it becomes bound to receptors that are abundant in the outer retina. Interestingly, significantly elevated VEGF protein levels are found in intraocular fluid samples from individuals with inner (but not outer) retinal neovascularization (22–27).

Other potential explanations for the lower VEGF levels we found in anterior chamber fluid samples after subretinal rather than after intravitreal injection would be more efficient transduction or a higher VEGF production level of transduced cells after intravitreal injection due to their own intrinsic activity.

Delivery of high-dose AAV.VEGF induced more rapid and pronounced effects than delivery of low-dose vector. Intravitreal injection led to both iris neovascularization and inner retinal neovascularization within 21 days after injection. Subretinal injection, on the other hand, induced extensive neovascular changes throughout all layers of the retina and the choroid but not in the iris. As seen in the low-vector-dose group, anterior chamber VEGF levels were significantly lower after subretinal than after intravitreal injection. VEGF levels in our study ranged from 0.13 ng/ml to >12.0 ng/ml (shown as pg/ml in Fig. 6) after intravitreal injection of AAV.VEGF and were thus highly comparable with those noted in humans with proliferative retinopathy, reported to range between 0.065, 1.28, and 168,000 ng/ml (22–27).

A surprising finding was the drop in VEGF transgene levels 3 weeks after delivery of a high dose of AAV.VEGF after both intravitreal and subretinal injection. There are several possible explanations for this drop.

First, the very high levels of expression in these eyes may have induced an immune response against the human transgene product. Due to the limited availability of anterior chamber fluid, we were not able to evaluate antibody/cytokine generation. Inflammatory cells were apparent, however, in retinas exposed to the higher doses of AAV.VEGF.

Second, many of the cells targeted by AAV.VEGF may have degenerated due to toxicity of the high concentration of this protein. In support of this, photoreceptor degeneration was apparent in photoreceptors exposed to AAV.VEGF by subretinal injection. Photoreceptors are known to be susceptible to degeneration due to overexpression of proteins, even if those proteins are normally found in these cells (42).

Finally, photoreceptors may have died secondarily to the high levels of serum proteins/erythrocytes that appeared in the retina in response to overexpression of VEGF. Blood cells/serum proteins are known to be toxic to photoreceptors and can induce degeneration within 1 day of exposure (43).

Another unexpected finding in these studies was the high baseline level of VEGF protein in ganglion and photoreceptor cells in untreated monkey retina. Immunohistochemically detectable VEGF levels in these cells were significantly higher than those in RPE cells, retinal vascular endothelial cells, choroidal vessel endothelium, and the inner nuclear layer, in all of which VEGF RNA has been detected previously (44,45). VEGF mRNA is produced by ganglion cells (45); therefore, the presence of VEGF protein in these cells is not surprising. The high levels of VEGF that we observed in photoreceptor outer segments may be due to binding of VEGF to these structures through association with cell-surface proteoglycans. A challenge imposed by the normal VEGF localization pattern in the primate retina is to identify changes in this pattern induced by delivery of AAV.VEGF. Because AAV2 targets the exact same cell types in which VEGF protein is normally present (ganglion cells and photoreceptors) and because our transgene did not contain a biological marker, differences in VEGF levels in our study can only be evaluated by measurements in ocular fluid and alterations in the normal immunohistochemical pattern of VEGF localization. In this study, we observed qualitative increases in VEGF levels in ganglion cells of eyes injected intravitreally and in RPE cells exposed subretinally to AAV.VEGF. Increases of VEGF protein in photoreceptors could not be appreciated due to the severe pathological changes (including cystoid macular degeneration and photoreceptor degeneration) that had taken place in the regions of retina exposed to AAV.VEGF by the time of termination of the study.

In conclusion, we successfully established nonhuman primate models for inner retinal neovascularization, subretinal neovascularization/cystoid macular edema, and iris neovascularization. Iris neovascularization is known to occur in diabetes and can lead to blindness through closure of the irido-corneal angle and concomitant increase in intraocular pressure.

The characteristics and extent of pathology were related directly to the site of AAV.VEGF delivery (and the cells transduced by this virus) and the level of VEGF produced by this treatment. Neovascularization induced via the intravitreal route of delivery of AAV.VEGF simulated diabetic retinopathy, whereas CNV and disturbance of photoreceptor architecture resulted from subretinal injection. Future studies using the intravitreal delivery approach will allow delineation of the pathological mechanisms leading to diabetic retinopathy. In addition, the availability of nonhuman primate models by AAV gene transfer will provide the ability to obtain long-term safety data in an animal model with a macula with respect to efficacy of novel therapeutic compounds designed to treat blinding human retinal neovascular diseases.

This study was supported by the Juvenile Diabetes Research Foundation (to J.M.W. and J.B.), grants R01 EY10820 and EY12156 (to J.B.), grant EY13385 (to S.G.J.), grant EY13203 (to A.V.C.), grant 5-P30-DK-47747-10 from Vector Core Facility (to J.W.), the Foundation Fighting Blindness (to J.B., S.G.J., A.V.C., and T.S.A.), the Lois Pope LIFE (Leaders in Furthering Education) Foundation (to J.B.), the Macula Vision Research Foundation (to S.G.J. and J.B.), The Lew R. Wasserman Award (to J.B.), The William and Mary Greve International Research Scholar Award (to A.V.C.), Research to Prevent Blindness, the Ruth and Milton Steinbach Fund (to J.B.), the Paul and Evanina Mackall Trust, and the F.M. Kirby Foundation.

We thank Joe McLaughlin and Brian Murphy for expert technical assistance and Stuart Fine for review of fluorescein angiograms.