Retinal Degeneration Is Slowed in Transgenic Rats by AAV-Mediated Delivery of FGF-2

Abstract

purpose. We evaluated adeno-associated virus (AAV)–mediated gene transfer of basic fibroblast growth factor (FGF-2) as a therapy for photoreceptor degeneration in a transgenic rat model of retinitis pigmentosa.

methods. Recombinant adeno-associated virus vector (rAAV) incorporating a constitutive cytomegalovirus (CMV) promoter was used to transfer the bovine FGF-2 gene to photoreceptors. AAV was administered by subretinal injection to transgenic rats (TgN S334ter-4) at postnatal day 15 (P15). Control eyes were uninjected, injected with PBS, or AAV–LacZ. Eyes were examined by histopathology, morphometric analysis, and electroretinography at P60.

results. Expression of recombinant FGF-2 slowed the rate of photoreceptor degeneration. Morphologic studies demonstrated significantly more photoreceptors surviving in eyes injected with AAV–FGF-2 than in controls. Insignificant rescue effects were seen in retinas injected with buffer only. No significant inflammatory response or neovascularization was detected. Electroretinographic (ERG) responses of eyes injected with AAV–FGF-2 were increased compared with uninjected eyes; however, these amplitudes were not significantly larger than eyes receiving an AAV–LacZ control vector.

conclusions. Transduction of retinal cells with AAV–FGF-2 reduces the rate of photoreceptor degeneration in an S334ter-4 animal model. Despite the lack of significantly increased ERG amplitudes from eyes expressing FGF-2, a greater number of surviving photoreceptors was demonstrated. Delivery of FGF-2 using recombinant AAV has potential as a therapy for retinal degeneration.

Retinitis pigmentosa (RP) is a heterogeneous group of inherited eye disorders, characterized by a progressive degeneration of rod and cone photoreceptors. ^{1 2} Patients with RP initially develop night blindness with a progressive reduction in peripheral visual field and, subsequently, lose central vision. Patients often exhibit an abnormal or unrecordable electroretinogram (ERG). As of this writing, over 150 specific mutations in various genes have been linked to autosomal dominant RP, autosomal recessive RP, or X-linked RP. Genes include rhodopsin, peripherin/RDS, α- and β-subunits of rod cGMP-phosphodiesterase, ROM1, α-subunit of the rod cGMP-gated channel, RP GTPase regulator, cellular retinaldehyde binding protein, and RPE65. ³ Each identified gene accounts for a small percentage of RP patients, with rhodopsin mutations being the most prevalent in approximately 10% of all cases. Currently, there is no cure or effective treatment for RP. 

Various transgenic animal models have been developed to study the causes of the disease and to test potential therapies. ^{4 5 6 7 8} The present study makes use of a transgenic rat line (TgN S334ter-4), which expresses a mutated rhodopsin gene. ⁹ The opsin transgene contains a termination codon at residue 334, resulting in the expression of a rhodopsin protein lacking the 15 C-terminal amino acids. The C terminus is involved in rhodopsin localization to the outer segments, and its absence contributes to photoreceptor degeneration, by a caspase-3–dependent mechanism. ^{10 11} Because multiple mutations within the C terminus have been identified in patients afflicted with RP, the use of S334ter-4 rats allowed us to design therapies for an animal model that mimics the human disorder. Heterozygous S334ter-4 rats are born with a full complement of photoreceptors and develop normally. At postnatal day (P) 15 when degeneration begins, these animals have 8 to 10 rows of photoreceptor nuclei in the outer nuclear layer (ONL). The degeneration rate is biphasic, with a faster initial rate between P15 and P60 and a slower one afterward. The ONL degenerates to 2 to 4 rows of nuclei by P60 and to 1 to 2 rows by P120. ¹⁰  

Substantial effort in retinal degeneration research has focused on the therapeutic effect of neurotrophins as a protective strategy to slow the rate of retinal degeneration. There exists a significant genetic heterogeneity in RP and a large number of mutations in multiple retinal genes leading to the common pathway of photoreceptor cell death. Specific gene therapies, such as antisense or ribozymes, cannot readily treat a significant fraction of RP patients. As a result, there is interest in a generalized survival factor therapy, which does not target the mutant gene product but rather alters the photoreceptor environment in a manner that promotes cell survival. Steinberg and LaVail have tested a large number of different survival factors and combinations of factors in two models of photoreceptor degeneration, the Royal College of Surgeons (RCS) rat and constant light damage in albino rats. ^{12 13 14 15} They had success in ameliorating photoreceptor cell death with direct protein injections of different growth factors or neurotrophic agents, including basic fibroblast growth factor (bFGF, or FGF-2), ciliary neurotrophic factor (CNTF), and brain-derived neurotrophic factor (BDNF). 

In the present study, we chose to express FGF-2, a neurotrophin that had a significant protective effect because it is involved in a number of regenerative, proliferative, and survival-related events in the central nervous system. ¹⁶ FGF-2 is endogenously expressed in the retina during development and in the adult. ¹⁷ Expression of dominant negative FGF receptors in photoreceptors results in slow degeneration and strongly suggests that photoreceptors require FGFs as a survival factor. ¹⁸ FGF-2 becomes upregulated with laser injury, mechanical puncture, or light damage ^{19 20} and incurs a protective effect. ²¹ In addition to preserving photoreceptors, intraocular protein injection protects inner retinal neurons against ischemia. ^{22 23} FGF-2 delivery by adenovirus (Ad) ²⁴ and polymer-encapsulated cells secreting FGF-2 ²⁵ have also slowed photoreceptor degeneration in RCS rats. 

However, protective effects from purified protein delivery and adenovirally expressed protein are only transient. Degeneration is slowed initially but continues to proceed rapidly at later stages, probably due to degradation of purified protein or reduction of expression from adenoviruses. To achieve stable expression, in the present study we used AAV to express FGF-2 in the retina. AAV is a nonpathogenic single-stranded human parvovirus ²⁶ that, when transduced in the absence of adenovirus, triggers a latent infection. Recombinant adeno-associated virus (rAAV) vectors have the ability to deliver the gene of interest to photoreceptors and allow for long-term gene expression without significant toxicity or immune response. In retinal photoreceptors, AAV-mediated expression of a reporter gene, GFP, persists at 2 years following injection (Hauswirth WW and Flannery JG, personal communication, June 1999). AAV has been used to deliver therapeutic genes to correct defects in animal models of various human disorders, such as hemophilia, ^{27 28} lactose intolerance, ²⁹ obesity, ³⁰ lysosomal storage disease, ³¹ Parkinson’s disease, ^{32 33} α-1-antitrypsin deficiency, ³⁴ muscular dystrophy, ³⁵ and cystic fibrosis. ³⁶ In the retina, AAV delivery of ribozymes, ³⁷ and the PDE-β subunit, ³⁸ have slowed retinal degeneration as well. We show that the use of AAV to express FGF-2 (AAV–FGF-2) in S334ter-4 photoreceptor cells elicits a protective effect on photoreceptors at both early and late stages of degeneration. 

Methods

Animals

The transgenic S334ter-4 rats were produced on a Sprague–Dawley background (Chrysalis DNX Transgenic Sciences, Princeton, NJ) and cared for in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and the guidelines of the University of California–Berkeley Committee on Animal Research. Rats heterozygous for the S334ter-4 transgene and wild-type Sprague–Dawley rats (Simonsen, Gilroy, CA) were used in experiments discussed. Rats were reared on a 12-hr light/12-hr dark schedule. 

Construction of rAAV Vector Expressing FGF-2

rAAV constructs were based on the pKm201CMV vector. pKm201CMV is an AAV cloning vector in which an expression cassette, consisting of a cytomegalovirus (CMV) immediate early promoter/enhancer and a bovine growth hormone (bGH) polyadenylation site, is flanked by inverted terminal repeat (ITR) sequences from AAV-2. Briefly, pKm201CMV was derived from pKm201, a modified AAV vector plasmid in which the ampicillin resistance gene of pEMBL–AAV–ITR ³⁹ had been replaced with the gene for kanamycin resistance. The expression cassette from pCMVlink, a derivative of pCMV6c ⁴⁰ in which the bGH polyA site has been substituted for the SV40 terminator, was inserted between the ITRs of pKm201 to generate pKm201CMV. pKm201bFGF-2 was constructed by cloning the following, in order, into the multiple cloning site of pKm201CMV: the encephalomyocarditis virus (EMCV) internal ribosome entry site (IRES), the bovine FGF-2 cDNA, and the human growth hormone polyadenylation sequence. The cDNA for FGF-2 has two mutations that change amino acid 121 from serine to threonine and amino acid 137 from proline to serine. The schematic of pKm201bFGF-2 is shown in Figure 1 . 

rAAV vector particles were produced by a triple transfection. Briefly, human embryonic kidney 293 cells, grown to 50% confluence in a 10-layer Nunclon cell factory (Nalge Nunc, Naperville, IL), were cotransfected with 400 μg of helper plasmid pKSrep/cap, ⁴¹ 400 μg of vector plasmid, and 800 μg of Ad plasmid pBHG10 (Microbix Biosystems, Toronto, Ontario, Canada) using the calcium phosphate coprecipitation method. Seventy-two hours after transfection, cells were harvested and resuspended in Tris-buffered saline (200 ml/cell factory) and lysed by three cycles of freezing and thawing. Cell debris was removed by centrifugation at 2000g for 20 minutes. Packaged rAAV was purified by two rounds of cesium chloride equilibrium density gradient centrifugation. 

To estimate total number of rAAV particles, the virus stock was treated with DNase I, and encapsidated DNA was extracted with phenol–chloroform and precipitated with ethanol. DNA dot blot analysis against a known standard was used to determine titer. ⁴²  

To assay for wild-type AAV, 293 cells were coinfected with serial dilutions of rAAV stocks and Ad dl312 at a multiplicity of infection (MOI) of 2. Three days later the cells were harvested, lysed by three cycles of freezing/thawing, and centrifuged to remove cell debris. The supernatant was heat-inactivated (56°C for 10 minutes), and fresh plates of 293 cells (6 × 10⁶) were infected in the presence of Ad dl312 at a MOI of 2. Forty-eight hours after infection, low-molecular-weight DNA was isolated, ⁴³ subjected to agarose gel electrophoresis, and transferred to a nylon membrane. The membrane was hybridized with a biotinylated oligonucleotide probe specific for the AAV capsid region. The wild-type AAV titer was defined as the highest dilution of rAAV vector stock demonstrating a positive hybridization signal. The rAAV preparations contained less than 1 wild-type AAV genome per 10⁹ rAAV genomes. 

Subretinal Injections

Rats were anesthetized by an intramuscular injection of ketamine–xylazine at P15. Eyes were dilated using of 2.5% phenylephrine hydrochloride and 1% atropine sulfate. All subretinal injections are performed using a stereomicroscope. A volume (2.5 μl) of either virus or phosphate-buffered solution (PBS) is injected through a blunt 32-gauge Hamilton syringe by a medial approach. The tip of the needle is inserted through the nasal sclera, choroid, retina, vitreous, and into the superior central retina to deliver a 2.5 μl volume into the subretinal space. We found that this approach was most successful in avoiding damage to the lens. 

Reverse Transcription–Polymerase Chain Reaction

Rat retina mRNA was extracted from AAV–FGF-2 and AAV–LacZ–injected retinas at 4 weeks following injection using the Qiagen RNeasy Kit. cDNAs were synthesized using Clontech Advantage RT-for-PCR Kit. Kits were used as instructed by the manufacturer. cDNAs were subjected to polymerase chain reaction (PCR) amplification using the following primers. The upstream primer, 5′-ATCCACGCTGTTTTGACCTC-3′, binds to sites within the 5′-untranslated region specific to the vector (917–936 bp) and the downstream primer, 5′-ATGTGTGGGTCGCTCTTCTC-3′, binds within the FGF-2 gene (2666–2685 bp; Fig. 1 ). Primers were designed to amplify a 950-bp band. This design ensured that the amplified FGF-2 band is from the recombinant construct. 

Immunohistochemistry

Rat eyes were enucleated from animals injected with AAV–FGF-2, AAV–LacZ, or PBS at 4, 8, or 15 weeks following injection. Eyecups were fixed in 4% formaldehyde in PBS for 1 hour at room temperature and washed in PBS three times. Eyes were cryoprotected in 30% sucrose overnight at 4°C and embedded in ornithine carbamoyltransferase for at least 2 hours at 0°C. Sections (25 μM-thick) were cut using a CM1850 cryostat (Leica, Nussloch, Germany) and allowed to dry overnight. Sections were incubated for 2 hours at room temperature using a monoclonal antibody to FGF-2 (Upstate Biotechnologies, Lake Placid, NY) diluted in 1% fetal calf serum, 1% bovine serum albumin, and 0.3% Triton X-100 in PBS (1:100). Bound antibodies were detected by incubating sections for 1 hour at room temperature with goat anti-mouse IgG antibodies conjugated to either fluorescein isothiocyanate or Cy3 (Sigma, St. Louis, MO). Images were acquired using the Applied Precision Deltavision deconvolution microscope (Applied Precision, Issaquah, WA). The antibody is specific to FGF-2, recognizing both recombinant (AAV-derived) FGF-2 and the endogenous FGF-2 produced by the retina. 

AAV Capsid Enzyme-Linked Immunosorbent Assay

Microtiter plates were coated overnight at 4°C with purified rAAV particles (1 × 10⁹/well) in PBS buffer. The coated plates were washed and then blocked with PBS containing 1% goat serum and 0.3% Tween-20 at 37°C for 30 minutes. Serial threefold dilutions of rat serum samples and control sera were loaded onto the plates. The positive rat serum control was used (from a previous experiment), showing a high titer of AAV antibody. The plates were then incubated at 37°C for 1 hour. Those plates were washed and incubated at 37°C for 30 minutes with goat anti-rat IgG H + L CH horseradish peroxidase at 1:10,000 (Bethyl, Montgomery, TX). o-Phenylenediamine substrate (Sigma) was used to develop the color. The plates were read at 492 nm with a cutoff of 0.2 OD. 

Electroretinography

Two S334ter-4 litters were treated as described above with AAV–FGF-2 delivered to one eye and either AAV–LacZ (n = 8) or no injection (n = 7) into the contralateral eye. At P60, these two litters and four wild-type rats were dark-adapted overnight and anesthetized by intramuscular injection of ketamine–xylazine at P60. A drop of 0.5% proparacaine hydrochloride, a local anesthetic, was applied to the cornea, and pupils were dilated with 2.5% phenylephrine. Wire loop electrodes were placed on the cornea and moistened with 1% methylcellulose. Reference electrodes were placed subcutaneously under each eye, and a ground electrode was inserted into the tail. Full-field scotopic ERGs were elicited with 10-usec flashes of white light, and responses were recorded using a UTAS-E 2000 Visual Electrodiagnostic System (LKC Technologies, Gaithersburg, MD). Stimuli were presented at intensities of 0.173,− 1.896, and −3.886 log candela seconds/m² at 1-minute, 30-second, and 15-second intervals, respectively. The a-wave amplitudes were measured from the prestimulus baseline to the first negative peak, and b-wave amplitudes were measured from the a-wave peak to the most positive peak. Four responses at each stimulus intensity were averaged to help reduce noise. Statistical significance of amplitudes was determined by paired Student’s t-test. Eyes were also processed for histochemical and morphologic analyses. 

Microscopy and Morphologic Analysis

Fifty-two rats were killed by carbon dioxide overdose, and cardiac perfusion using 2.5% glutaraldehyde and 2% formaldehyde in PBS was performed at P60 or P120. For light microscopy, eyecups were embedded in Epon–araldite resin, and 1-μm-thick sections were made along the vertical meridian in the same plane as the optic nerve. Tissue sections were aligned so that rod outer segments were continuous throughout the plane. Twenty-seven measurements of the ONL thickness were made around the inferior or superior regions separately using Bioquant 98 image analysis system (R&M Biometrics, Nashville, TN). ¹⁵ These measurements from each region were averaged to obtain the mean ONL thickness. These data were analyzed using one-way ANOVA, paired Student’s t-test, or the Bonferroni post-hoc test, as appropriate. For electron microscopy, eyecups were further fixed in 1% osmium tetroxide, dehydrated, and embedded in LR White resin. Ultrathin sections were stained with uranyl acetate and examined using a JEM 1200EXII microscope (JEOL, Tokyo, Japan). 

Results

Expression of FGF-2 Transgene

The AAV–FGF-2 vector contains a full-length bovine FGF-2 cDNA driven by a CMV immediate early promoter/enhancer element (Fig. 1) . The codon usage for this FGF cDNA was optimized for expression in human cells. We performed reverse transcription–polymerase chain reaction (RT–PCR) on cDNAs obtained from retinal tissue acquired 4 weeks following injection to evaluate expression of recombinant FGF-2 mRNA by retinal tissues. To ensure that we did not amplify endogenous FGF-2 message, our upstream primer was designed to bind a vector-specific region (Fig. 1) . Expression was present in retinas injected with AAV–FGF-2 and absent from retinas injected with AAV–LacZ (Fig. 2) . 

Expression of the FGF-2 protein was also evaluated by immunohistochemistry on cryosections. Photoreceptor cell bodies and inner segments stained intensely for FGF-2 in AAV–FGF-2–injected eyes (Fig. 3A ). Subretinal injection of AAV–FGF-2 also leads to intense FGF-2 expression in the retinal pigment epithelium (RPE; Fig. 3B ) and ganglion cells (not shown). This pattern of expression was observed at 4, 8, and 15 weeks following injection, and expression levels appeared consistent at all points. Retinal sections from eyes injected with AAV–LacZ or PBS did not express recombinant FGF-2 in photoreceptor cells, but we did see upregulation of endogenous FGF-2 in cell types within the inner nuclear layer (INL; Fig. 3C ). This pattern of endogenous FGF-2 upregulation in the INL was also observed in AAV–FGF-2–injected retinas (Fig. 3A) . Recombinant FGF-2 expression was found in up to half the area of the retina, predominantly in the superior region (Fig. 4) . The extent of expression of recombinant FGF-2 in an AAV–FGF-2–injected retina is similar to the LacZ expression pattern observed in an AAV–LacZ–injected retina (not shown). 

Morphologic Rescue

We quantified ONL thickness in treated and untreated S334ter-4 retinas to determine whether the expression of recombinant FGF-2 could prevent photoreceptor cell death. In control uninjected S334ter-4 retinas, the superior ONL was reduced to 2 to 3 rows of photoreceptor nuclei by P60, compared with 8 to 10 rows in wild-type rat retinas (Fig. 5) . Degeneration varied greatly between the inferior and superior regions of the eye in S334ter-4, so data from each region are presented separately. In the superior region, retinas injected with AAV–FGF-2 have an ONL that is 6 to 8 cells thick, compared with 4 to 6 cells thick in AAV–LacZ– and PBS-injected controls (Fig. 5) . In both the superior and inferior regions, AAV–FGF-2–injected eyes were thicker than the three controls (Fig. 6) . The mean ONL measurements from AAV–FGF-2–injected retinas were 23.9μ m compared with 18.3 μm in PBS-injected retinas in the superior region (P < 0.05; Fig. 6A , table) and 29.5 μm in AAV–FGF-2–injected retinas compared with 26.0 μm in PBS-injected retinas in the inferior region (P > 0.05; Fig. 6B , table). The mean ONL measurements from AAV–FGF-2–injected retinas were 23.9 μm compared with 16.4 μm in AAV–LacZ–injected retinas in the superior region (P < 0.001; Fig. 6A , table) and 29.5 μm in AAV–FGF-2 retinas compared with 23.3 μm in AAV–LacZ–injected retinas in the inferior region (P < 0.001; Fig. 6B , table). There were no statistically significant differences between any of the three controls. 

In addition to ONL thickness measurements, we performed transmission electron microscopy to examine photoreceptor inner and outer segment integrity in rescued and control retinas. Although rescued retinas have cell debris present in the subretinal space, many rod inner and outer segments are continuous and well organized (Fig. 7A ). Uninjected S334ter-4 retinas at P60 display disorganized photoreceptor inner and outer segments, representative of control retinas examined (Fig. 7B) . 

To evaluate FGF-2 potential for long-term rescue in S334ter-4 retinas, we also performed morphologic quantitation of ONL thickness at P120, 105 days following injection. The superior ONL in S334ter-4 rats degenerates to 1 row of photoreceptor nuclei by P120 (Fig. 8) . Superior region retinas injected with AAV–FGF-2 had an ONL that was 2 to 3 cells thick, compared with 1 cell thick in PBS-injected controls (Fig. 8) . In both superior and inferior regions, AAV–FGF-2–injected eyes were thicker than controls (Fig. 9) . The mean ONL measurements from AAV–FGF-2–injected retinas were 12.1μ m compared with PBS-injected 8.8 μm in the superior (P < 0.05; Fig. 9A , table) and 19.7 μm compared with 17.3 μm in the inferior region (P > 0.05; Fig. 9B , table). 

Histopathology

We inspected plastic sections for macrophages or other cellular infiltrates. No significant inflammation was observed in the eyes injected with either AAV–FGF-2 or AAV–LacZ when compared with PBS-injected controls. Neovascularization was not observed in plastic sections from AAV–FGF-2–injected retinas or any of the controls. In a representative litter with AAV-injected eyes, serum was collected and assayed for antibody to AAV viral capsid protein. There was a positive response in 8 of 9 rats injected with virus (Table 1) . 

Physiological Rescue

ERG recordings were made to determine whether an increase in physiological function accompanied morphologic rescue. Scotopic ERG recordings were performed on P60 rats that received an injection of the AAV–FGF-2 vector in one eye, whereas the contralateral control eye was either injected with AAV–LacZ (Fig. 10A ) or uninjected (Fig. 10B) . Control eyes yielded rod-mediated ERGs with severely impaired responses, whereas AAV–FGF-2–injected eyes exhibited mildly rescued ERG responses. a- and b-wave mean amplitudes from treated eyes were larger than amplitudes from control contralateral eyes (Table 2) . To compare treated transgenic animals with wild-type rats, electroretinography was also performed at P60 (Fig. 10C) , and a- and b-wave mean amplitudes were measured at 265 ± 50 and 541 ± 71 μV, respectively. a- and b-wave mean amplitudes from all AAV–FGF-2–injected eyes retained 20% and 55%, respectively, of wild-type amplitudes. 

Discussion

Growth factor therapy holds promise as a therapeutic strategy for these slowly progressing retinal degenerations. Our hypothesis is that the intraocular expression of one or more trophic factors will delay the time course of the retinal degeneration by protecting the photoreceptors from injury or cell death. We are testing this hypothesis in transgenic rats with mutations in the opsin gene that are similar to those known to cause autosomal dominant RP in humans. ^{44 45 46}  

Previous studies have shown that growth factors, neurotrophins, and cytokines can act as “survival factors,” protecting rat photoreceptors from injury and cell death for a brief period after injection into the eye. ^{12 13 14} In these experiments, investigators delayed the inherited degeneration in the RCS rat and prevented photoreceptor degeneration caused by light damage in the albino Sprague–Dawley rat. Photoreceptor cell death could be ameliorated with a number of different growth factors or neurotrophins, including FGF-2, ciliary neurotrophic factor, and BDNF. The initial results in the two rat models provided proof of principle that these survival factors can protect rat photoreceptors from injury and death. However, neither of these animal models is thought to be a good animal model for human RP. In addition, the protective effect of the protein in the vitreous was limited to several weeks postinjection. In more recent studies, gradual expression and secretion of a survival factor by lens epithelium appeared to be more effective at promoting photoreceptor survival than sporadic delivery by injection. ⁴⁷  

We extended these studies by inducing long-term expression of FGF-2 directly within photoreceptors and the RPE by means of viral-mediated delivery of the gene. We tested this approach with an animal model of retinal degeneration that more closely mimics human RP, a transgenic rat model in which apoptotic photoreceptor cell death occurs as a consequence of expression of a mutant opsin. ⁴⁸  

Trophic Effects of FGF-2 Expression in Photoreceptors and RPE

Growth factors act to inhibit the induction of apoptosis by photoreceptors. Apoptosis has been shown to be the mechanism of cell death in many retinal diseases. ^{48 49 50 51 52 53 54 55} There is an extensive literature demonstrating the ability of survival factors to attenuate neuronal apoptosis. ^{56 57} The antiapoptotic effects of survival factors in many neuronal systems suggest that they are applicable to the treatment of a number of retinal diseases. After a detailed anatomic characterization, we found that expression of FGF-2 in photoreceptors and RPE significantly reduced the rate of photoreceptor cell death in the rat model. In several of the most effectively rescued eyes, the ONL retained 90% of the thickness of the wild-type retina at P60, and the surviving photoreceptors retained better outer and inner segment structure by ultrastructural analysis. 

Viral Tropism

This viral construct directs expression of FGF-2 in several cell types that borders the subretinal space. Immunocytochemical localization of FGF-2 demonstrated expression in photoreceptors and RPE cells. Expression was observed in a small number of ganglion cells, apparently as a consequence of the route of injection through the retina. Other classes of retinal neuronal or glial cells are not transduced because they do not border the subretinal space. In addition, direct injection of AAV into the hippocampus results in uptake by neurons but not glial cells, ^{58 59} implying that retinal Müller cells and other glia may not express the appropriate receptors (bFGF and heparan sulfate) for transduction by AAV. ^{60 61 62}  

Variability and Extent of Rescue

Morphologic rescue is more apparent in the superior hemisphere than the inferior hemisphere. Two contributing factors may be as follows: the site of the viral injection and differences in baseline ONL thickness between the two regions. In the uninjected S334ter-4 rat eye, degeneration is more extensive in the superior region, because of higher levels of mutant opsin transgene. Therefore, we directed injections toward the superior region. AAV particles introduced into the superior region become distributed and spread to the inferior region. As a result, injection into the superior region would also lead to transduction of cells in the inferior region, albeit to a lesser extent. We estimate that greater than 95% of the photoreceptor and RPE cells around the injection site express FGF-2. This higher expression could account for the greater degree of rescue in the superior region when compared with the inferior region. 

The variability we see in the rescue effects has been observed in other photoreceptor degeneration models. ^{63 64} There are large differences in the ONL thickness within the AAV–FGF-2–injected groups at both times and in both inferior and superior regions. We attribute this variability in rescue to differences in FGF-2 expression from eye to eye. The injection technique itself may introduce varying amounts of virus to the subretinal space. In addition, it is unclear from our studies whether a minimal threshold of therapeutic FGF-2 is required for rescue. Perhaps the ONL measurements from retinas injected with AAV-FGF-2, which do not differ from controls (Figs. 6 9) , have not attained the threshold level required for rescue. 

This variability in rescue is reflected in the ERG analysis. The physiological rescue from an entire litter of animals appears minimal (Table 2) , and b-wave mean amplitudes between AAV–FGF-2– and AAV–LacZ–injected retinas border on significance (P = 0.11). However, the ERG is a global response of the entire retina, and we anticipate various levels of FGF-2 expression in up to 60% of the area of the retina. Pairwise examination of ERG responses within individual animals shows that AAV–FGF-2–injected eyes exhibit larger a- and b-waves than AAV–LacZ–injected (6/8 animals) and uninjected (6/7 animals) eyes. There was a significant difference in a-wave amplitudes between AAV–FGF-2– and AAV–LacZ–injected eyes (Table 2 , litter A) but not in b-wave amplitudes. These results are unexpected, because differences in the a-wave are normally amplified in the b-wave. A recent study from Stone et al. ⁶⁵ suggests that there is some suppression of b-wave with protein injections of FGF-2, despite increased a-wave amplitude. The mechanism for this effect has yet to be determined. 

In the present work, we used a line of transgenic rats (S334ter-4), whose ONL have 8 to 10 rows of photoreceptor nuclei at P15, degenerating at a rate of approximately 1 to 2 rows of nuclei per week until P60. Subretinal injection, performed earlier in the lifetime of a rat ensuring high expression before degeneration begins, may produce more significant rescue. However, this procedure before P15 may interfere with normal retinal development caused by physical injury. It is also unclear what adverse effects exogenous FGF-2 expression may trigger during normal retinal development. 

Assessment of Retinal Function

In previous studies to test therapies in this rat model, ^{37 66 67} photoreceptor rescue as measured by ONL thickness and preserved photoreceptor structure was correlated with an increase in ERG a- and b-wave amplitudes. To our surprise, the survival of more photoreceptors with better preserved structure in retinas transduced by AAV–FGF-2 only resulted in slight functional improvements as measured by ERG. ERG a- and b-waves recorded in AAV–FGF-2–injected eyes were significantly greater than those in uninjected retinas but were not significantly greater than those achieved with AAV that delivered only a reporter gene. 

The increased ERG amplitude recorded from eyes injected with AAV–LacZ when compared with uninjected controls is likely to be a response to injury from the subretinal injection and retinal detachment. In previous studies, rats have been shown to generate a localized rescue response about a wound after the simple insertion of a needle that does not contain any survival factors. ⁶⁸ The fact that eyes treated with AAV–FGF-2 retain better retinal structure but do not retain significantly increased function is unclear. Expression of FGF-2 by photoreceptor and RPE cells may act to suppress the amplitude of the ERG. A similar result was observed by Impleman and Copenhagen (personal communication, June 1999). They found that intravitreal injection of bFGF reduced the ERG responses of normal animals. We have also observed this effect after viral-mediated expression of other FGFs in the photoreceptor and RPE. ⁶⁹ We are conducting additional studies to examine the mechanism of this effect, and its implications for preservation of retinal function in this degeneration. 

Mechanism of Rescue

The novel secretion pattern of FGF-2 contributes a level of complexity to the mechanism of rescue mediated by FGF. Various isoforms of FGF-2 can be localized intracellularly in the cytoplasm and nucleus as well as secreted extracellularly. ⁷⁰ Staining of retinal sections suggests that expression remains cytosolic in photoreceptor and RPE cells (Fig. 3) . In the absence of a signal sequence, FGF-2 should remain in the cell. However, FGF-2 is slowly secreted in cultured cells through an unusual non-vesicular mechanism. ^{71 72} FGF-2 surrounding retinal cells may also be regulated by the presence of soluble FGF receptors that have been identified in ocular tissues. ^{73 74} We hypothesize that FGF-2 expressed in the retina may activate a variety of receptors, which could be found in the nucleus or cytoplasm or on the plasma membrane. 

FGF receptors (FGFR-1 and FGFR-2) have been localized to ganglion cells, photoreceptor outer segments, and RPE cells. ^{75 76 77 78} It is likely that the retina expresses other isotypes of FGF receptors as well. FGF-2 is capable of activating various isotypes from these four FGF receptor gene families. ⁷⁹ This suggests that there are multiple target receptors that could be responsible for mediating a survival response in the retina. In addition, multiple retinal cell types can be infected by AAV and a CMV promoter ubiquitously drives transgene expression, so the source of FGF-2 could arise from photoreceptor, ganglion, or RPE cells. The mechanism of photoreceptor survival may originate from different parts of the retina and activate one or a combination of receptors. 

Mechanical injury causes an upregulation of endogenous FGF-2 in inner retinal neurons in PBS-injected, AAV–LacZ–injected, and AAV–FGF-2–injected retinas (Fig. 3) , and we believe that this expression contributes to photoreceptor rescue. This observation also suggests that there are indirect rescue mechanisms originating from the inner retina. BDNF overexpression in Müller cells, for example, has been shown to rescue photoreceptors from constant light damage in rodents. ⁸⁰ Bennett et al. ⁸¹ have shown that delivery of wild-type copies of the β-PDE gene can rescue the rd mouse from retinal degeneration. Ad-mediated gene transfer of CNTF has been shown to prevent death and increase physiological functioning in the rds mouse. ⁸² Whether cell survival in the S334ter-4 model of photoreceptor degeneration benefits from FGF-2 by way of autocrine or paracrine mechanisms still needs to be resolved. 

Because of the ubiquitous localization of FGF-2 and its receptors in the retina, a wide variety of retinal degenerations may benefit from FGF-2 delivery. Its mechanism of function will not be specific to the mutation, but the possibility that FGF-2 functions through multiple pathways and mechanisms in the retina may be well suited for its use as a general survival factor against apoptosis. 

Advantages of AAV Gene Therapy for Treating Retinal Degeneration

Extended expression of transgenes by AAV-mediated delivery has been demonstrated in various tissues, such as liver, ⁸³ muscle, ⁸⁴ and retina. ⁸⁵ This particular characteristic allows us to evaluate the extent of photoreceptor survival at stages much later than would be possible with other viral delivery mechanisms. Rescue persists at P120, with the superior ONL approximately 35% thicker than controls and the inferior ONL approximately 12% thicker than controls. It has yet to be determined whether the increased ONL thickness seen at P120 is caused by actual prolonged rescue or is simply an increased number of cells surviving at early timepoints; this is a topic of further research. Long-term rescue becomes a particularly important issue when treating a chronic, slowly progressing genetic condition like RP. To the best of our knowledge our study evaluated photoreceptor rescue much later than any other potential therapeutic treatments. 

We did not see pronounced inflammation in the eyes injected with either AAV–FGF-2 or AAV–LacZ when compared with controls. The eye is an immune-privileged site within the body, ⁸⁶ but immunosuppressive agents must be administered before introduction of Ad to the eye. AAV infection compared with Ad does not stimulate significant cell-mediated response as others have observed with Ad injection into ocular tissues. ⁸⁷ In addition, Bennett’s group demonstrated that the presence of neutralizing antibodies to viral coat does not inhibit subsequent transduction events with readministration of AAV. 

Neovascularization was not observed in AAV–FGF-2–injected retinas, suggesting that its primary role in the retina is neurotrophic. FGF-2 has been shown to stimulate angiogenesis, ⁸⁸ but this may not be the case in the retina. Perhaps the low level and slow rise of FGF-2 expression derived from AAV-mediated delivery are not sufficient to stimulate neovascularization. Overexpression of FGF-2 in photoreceptors does not stimulate neovascularization, ⁸⁹ reinforcing the idea that FGF-2 alone is insufficient to produce new blood vessel growth in the eye. Another growth factor, such as vascular endothelial growth factor (VEGF), may be a better candidate for mediating ocular neovascularization, because its overexpression does stimulate vascularization. ⁹⁰  

Trophic factors, such as platelet-derived growth factor, transforming growth factors, insulin-like growth factors, pigment epithelium derived factor, VEGF, CNTF, and FGF-2, are produced by the retina and may aid in survival of injured neurons. ⁹¹ FGF-2 and CNTF have great potential for preventing or delaying photoreceptor degeneration in the retina because they are upregulated in response to injury. ⁹² Expression of CNTF has been previously shown to slow retinal degeneration. ^{82 93} The use of ribozymes to destroy the mutant gene product in transgenic P23H rats succeeded in preserving ONL thicknesses 30% to 40% greater than uninjected controls. ⁹⁴ We show similar rescue with average ONL thicknesses that are 63% and 13% thicker than controls in the superior and inferior regions, respectively. Ultimately, a combination of growth factors or ribozymes may provide the most beneficial therapy for retinal degeneration. 

⁴  These authors contributed equally to the work presented here and should, therefore, be regarded as equivalent senior authors.

 Supported by National Eye Institute Grant EY11123; NIH Training Grants T32 EY07043 and T32 GM07048; That Man May See; Chiron Corporation; the Hellman Family Fund; and The Foundation Fighting Blindness.

 Submitted for publication October 12, 1999; revised December 29, 1999 and March 3 and June 20, 2000; accepted June 23, 2000.

 Commercial relationships policy: F, P (DL, LHM, JGF); E, P (SZ, KGR, WCM, JAE).

 

Figure 1.

 

rAAV–FGF-2 construct contains a CMV immediate early enhancer/promoter (CMV IE ENH/PRO) and a bGH polyadenylation site, flanked by ITR sequences from AAV-2. It was constructed by the insertion of the EMCV IRES, the humanized bovine FGF-2 cDNA, and the human growth hormone (hGH) polyadenylation sequence. Vertical arrows indicate internal splice sites, creating an intron of 780 bp. Horizontal arrows indicate location of primers for RT–PCR.

Figure 2.

 

RT–PCR of retinal samples at 4 weeks following injection. Expression of recombinant FGF-2 is specific to AAV–FGF-2–injected retinas and absent from AAV–LacZ–injected retinas. Primers were designed such that the upstream primer binds to a site specific to the vector and spans an intron, whereas the downstream primer binds within the FGF-2 gene amplifying a band of 950 bp (lanes 2 to 4). This design amplified only the recombinant FGF-2. cDNA contamination in RT–PCR preps was ruled out because plasmid or viral DNA contains an intron and will amplify a larger fragment of 1.7 kb. Lane 1, DNA ladder. Lanes 2 to 4, AAV–FGF-2–injected retinal cDNA template. Lane 5, AAV–LacZ–injected retinal cDNA template. Lane 6, PKm201bFGF-2 plasmid DNA template.

Figure 3.

 

Immunohistochemical detection of FGF-2 expression in AAV–FGF-2–injected wild-type and control retinas at 4 weeks following injection. The FGF-2 transgene is predominantly expressed in photoreceptors (A) and RPE cells (B). Upregulation of endogenous FGF-2 expression is seen in the INL of AAV–LacZ–injected controls (C). This upregulation occurred in all injured retinas, including AAV–FGF-2–injected (A) and PBS-injected (not shown) retinas. Uninjected control sections displayed no recombinant FGF-2 expression (D). Arrowheads point to endogenous FGF-2 expression in blood vessels within the inner nuclear layer. CHR, choroid. Scale bar, 20μ m.

Figure 4.

 

Immunohistochemical detection of FGF-2 expression across an eyecup. Expression of recombinant FGF-2 extends over half the section, indicated between the arrowheads. Inset comes from superior central region of the retina. Sup, superior region; Inf, inferior region.

Figure 5.

 

Morphologic rescue of degenerated S334ter-4 superior retinas at P60. In the superior region of uninjected S334ter-4 retinas, the photoreceptors degenerated from 9 to 10 cells thick, as seen in wild-type retinas, to 2 to 3 cells thick. Retinas injected with AAV–FGF-2 had an ONL that was significantly thicker than uninjected, AAV–LacZ–injected, or PBS-injected retinas. ROS, rod outer segments; RIS, rod inner segments; OLM, outer limiting membrane; GCL, ganglion cell layer. Scale bar, 20μ m.

Figure 6.

 

Mean ONL thickness measurements in superior (A) and inferior (B) regions of AAV–FGF-2–injected, AAV–LacZ–injected, PBS-injected, and uninjected retinas at P60. Each data point represents the mean of 27 measurements from the superior or inferior region of an eye. Dots may overlay. Probability values compare the significance of AAV–FGF-2 mean ONL thickness to the three controls. Controls were not statistically significantly different from one another (superior region[ :] PBS and AAV–LacZ–injected P > 0.05, PBS and uninjected P > 0.05, AAV–LacZ and uninjected P > 0.05; inferior region [:] PBS and AAV–LacZ–injected P > 0.05, PBS and uninjected P > 0.05, AAV–LacZ and uninjected P > 0.05). n, No. of eyes.

Figure 7.

 

Transmission electronmicrograph of photoreceptor inner and outer segments in AAV–FGF-2–injected (A) and uninjected (B) S334ter-4 retinas at P60. ROS, rod outer segments; RIS, rod inner segments; OLM, outer limiting membrane. Scale bar, 2 μm.

Figure 8.

 

Morphologic rescue of degenerated S334ter-4 superior retinas at P120. The photoreceptors degenerated from 8 to 10 cells thick in wild-type ONL to 1 cell thick in uninjected S334ter-4 retinas. Retinas injected with AAV–FGF-2 had an ONL that was significantly thicker than uninjected or PBS-injected retinas. ROS, rod outer segments; RIS, rod inner segments; OLM, outer limiting membrane; GCL, ganglion cell layer. Scale bar, 20 μm.

Figure 9.

 

Mean ONL thickness measurements in superior (A) and inferior (B) regions of AAV–FGF-2–injected, PBS-injected, and uninjected retinas at P120. Controls were not statistically significantly different from one another. Each data point represents the mean of 27 measurements from the superior or inferior region of an eye. Probability values compare the significance of AAV–FGF-2 mean ONL thickness to the three controls. Dots may overlay. n, No. of eyes.

Table 1.

 

Detection of AAV Capsid Antibody in Rat Serum

Rat	Anti-AAV Titer
Control Uninjected	<5
1	5
2	447
3	724
4	289
5	69
6	<5
7	558
8	67
9	239

Figure 10.

 

Representative ERGs from individual S334ter-4 animals with one eye injected with AAV–FGF-2 and the contralateral eye injected with AAV–LacZ (A) or uninjected (B) at P60. Stimuli were presented at intensity of 0.173 log candela s/m². AAV–FGF-2–injected eyes show increased a- and b-wave amplitudes compared with controls. An ERG from a wild-type rat is presented for comparison (C). Note differences in scale bars.

Table 2.

 

Average a- and b-Wave Amplitudes from ERG Analysis of Two Independent Litters

Litter	n	Treatment	a-Wave (μV)	P	b-Wave (μV)	P
A	8	AAV–FGF-2–inj.	55.6 ± 7.7	<0.04	338.3 ± 55.4	>0.05
		AAV–LacZ–inj.	54.0 ± 11.1		303.5 ± 60.7
B	7	AAV–FGF-2–inj.	48.4 ± 14.3	<0.05	257.8 ± 74.6	<0.02
		Uninjected	31.6 ± 9.5		153.2 ± 30.9

 inj, injected.

The authors thank Ranjana Srivastava for conducting in vitro expression studies of AAV constructs and Matthew M. LaVail, Douglas Yasumura, Michael D. Menz, and Eric S. Green for their valuable advice throughout this project.