Glial Cell Line–Derived Neurotrophic Factor Induces Histologic and Functional Protection of Rod Photoreceptors in the rd/rd Mouse

Abstract

purpose. To evaluate the neuroprotective potential of glial cell line–derived neurotrophic factor (GDNF) in the retinal degeneration (rd/rd) mouse model of human retinitis pigmentosa.

Methods.

Subretinal injections of GDNF were made into rd/rd mice at 13 and 17 days of age and electroretinograms (ERGs) recorded at 22 days. Control mice received saline vehicle injections or underwent no procedure. At 23 days of age, retinas from treated and control mice were fixed and processed for wholemount immunohistochemistry using an anti-rod opsin antibody, and rod numbers were estimated using an unbiased stereological systematic random approach. Subsequent to counting, immunolabeled retinas were re-embedded and sectioned in a transverse plane and the numbers of rods recalculated.

Results.

Although ERGs could not be recorded from sham-operation or nonsurgical rd/rd mice at 22 days of age, detectable responses (both a- and b-waves) were observed in 4 of 10 GDNF-treated mice. Stereological assessment of immunolabeled rods at 23 days showed that control rd/rd retinas contained 41,880 ± 3,890 (mean ± SEM; n = 6), phosphate-buffered saline (PBS)–injected retinas contained 61,165 ± 4,932 (n= 10; P < 0.001 versus control retinas) and GDNF-injected retinas contained 89,232 ± 8,033 (n= 10; P < 0.001 versus control retinas, P < 0.002 versus PBS). This increase in rod numbers after GDNF treatment was confirmed by cell counts obtained from frozen sections.

Conclusions.

GDNF exerts both histologic and functional neuroprotective effects on rod photoreceptors in the rd/rd mouse. Thus rescue was demonstrated in an animal model of inherited retinal degeneration in which the gene defect was located within the rods themselves, similar to most forms of human retinitis pigmentosa. GDNF represents a candidate neurotrophic factor for palliating some forms of hereditary human blindness.

Retinitis pigmentosa (RP) is the term given to a group of inherited retinal dystrophies resulting in progressive photoreceptor loss and eventual extreme visual handicap. There is no known treatment for this condition, which can be caused by a large number of mutations in a variety of genes that code for proteins involved in phototransduction (rhodopsin, cyclic guanosine monophosphate[ cGMP]–dependent phosphodiesterase) ^{1 2} or in structural roles (peripherin/rds). ³ Several strategies are under examination for limiting or preventing the photoreceptor loss associated with this disease: gene therapy, ⁴ pharmacologic treatment (growth factor application ^{5 6} or vitamin supplementation ⁷ ), and retinal transplantation. ^{8 9} Among these approaches, the use of growth factors has been advocated since pioneering work by the group of Faktorovich et al. ¹⁰ showed that intraocular injections of basic fibroblast growth factor (FGF-2) delayed photoreceptor degeneration in an animal model of RP, the Royal College of Surgeons rat. FGF-2’s effects on photoreceptor survival were subsequently demonstrated by the same group in rat models of light-induced photoreceptor degeneration. ^{5 11} In addition, in vitro studies have demonstrated the differentiative effects of FGF-2 on newborn rat photoreceptors. ¹² Yet, photoreceptor degeneration in the RCS rat as well as after constant light damage does not correlate to any currently known form of human retinal degeneration. 

The retinal degeneration (rd/rd) mouse is a model of human RP that has been investigated for more than 70 years because photoreceptor cell degeneration follows a similar evolution to human RP. Most rod cells undergo apoptosis over the first 3 weeks, ¹³ followed by death of cone photoreceptors. ¹⁴ Furthermore, rod cell death has been found to result from a mutation in the β subunit of cGMP-dependent phosphodiesterase, ¹⁵ similar to some human families affected by the disease. ¹⁶ Recently, among a large variety of neurotrophic factors injected into rd/rd eyes, only ciliary neurotrophic factor (CNTF) reduced histologic photoreceptor cell loss. ⁶  

Glial cell line–derived neurotrophic factor (GDNF), which belongs to the transforming growth factor β superfamily, was first described as a stimulant of survival of dopaminergic neurons in vitro. ¹⁷ Subsequently, its protective effects were demonstrated in in vivo models of Parkinson disease ^{18 19 20} or on developing motoneurons. ²¹ Because GDNF is synthesized in the retina, ²² has been found to stimulate survival of newborn mice photoreceptors in vitro, ²³ and recently was shown to delay photoreceptor outer segment collapse in vitro, ²⁴ we conducted electrophysiological, histologic, immunohistochemical, morphometric, and molecular biologic studies to see whether it could stimulate photoreceptor survival in the rd/rd mouse after intraocular injections in vivo. We showed that such treatment leads to functional improvement, as indicated by the presence of recordable electroretinograms (ERGs) in some GDNF-injected retinas. Increased numbers of immunolabeled rods were detected using two methods of counting, and retinal glial cell physiology was also affected. In contrast, we could exclude the involvement of decreased levels of endogenous GDNF in rod photoreceptor cell death in the rd/rd mouse. 

Materials and Methods

Animals

C3H/He/J mice homozygous for the retinal degeneration (rd) gene (rd/rd mice) and C57/BL6/J control mice aged 12, 15, 17, 19, and 24 days were obtained from single breeding colonies maintained exclusively for our laboratory by Charles River Animal Suppliers (St. Aubin-les-Elbeuf, France), maintained in clear plastic cages, and subjected to standard light:dark cycles of 12 hours All procedures conformed to the ARVO Statement on the Use of Animals in Ophthalmic and Vision Research. 

In Vivo Application of GDNF in the rd/rd Mouse

GDNF was injected into C3H mice at 13 and 17 days of age (with the day of birth defined as day 0). Mice (n = 10) were anesthetized with an intraperitoneal injection (16.7 μl/g body weight) of etomidate (0.5 mg/ml) and midazolan (0.5 mg/ml). The pupils were dilated with 0.5% tropicamide, and the cornea was anesthetized with 0.5% topical proparacaine. GDNF (1 μl; recombinant human GDNF; Promega, Madison, WI) was injected at a dose of 330 ng/μl in 0.1 M sterile PBS with a Hamilton syringe and a 30-gauge needle 1 mm from the limbus into the subretinal space. The contralateral eye was injected with the same volume of PBS solution only. It was necessary to perform bilateral ocular surgery to distinguish between the rescue effect due to GDNF per se and that due to experimental manipulation. In rare cases, partial backflow was observed. After injection, fundoscopy was performed to confirm the induction of a local retinal detachment. This was observed in every case at the time of injection, and we observed that the detachment created by the initial injection was resorbed by the time the second was administered. Processing of fixed retinas after recording (description follows) indicated that retinas were correctly attached at the end of the experimental period. 

Recording ERGs in the rd Mouse

ERGs were picked up by a cotton wick connected to an Ag:AgCl electrode with saline solution directed to the apex of cornea by a micromanipulator. A stainless steel reference electrode was placed subcutaneously on top of the head. Responses were amplified at a gain of 10,000 and filtered with the low 0.1-Hz and high 1000-Hz cutoff filters from an amplifier (Universal; Gould, Ballainvilliers, France). Responses were digitized using a data acquisition labmaster board (Scientific Solutions, Solon, OH), mounted on an IBM-compatible personal computer. Experimental data were acquired and analyzed using the Patchit and Tack software packages (Scientific Solutions). ²⁵ After a period of 24 hours, dark-adapted mice were prepared under dim red illumination. They were anesthetized as above and placed in a Faraday cage, the head resting in a U-shaped holder. The upper and lower lids were retracted to hold open and proptose the eye. The light stimulus was obtained from a 150-watt quartz-xenon lamp bulb (Müller Instruments, Moosinning, Germany). The light beam was focused to infinity through a heat-absorbing filter onto a hole in the Faraday cage. The illumination intensity of the light was 2.9 log cd/m² as measured with a luxmeter at the level of the eye. A computer-controlled shutter was used to deliver flashes of 300-msec duration. For GDNF-injected mice, recordings from both eyes were obtained successively, and the investigator was not aware of which eye had received the GDNF injection. 

Histology of Injected Eyes

Mice were killed by anesthetic overdose at 23 days. Whole eyes from GDNF- and PBS-injected and noninjected rd/rd eyes (n = 2 for each group), were removed from mice at 23 days, fixed in 2.5% glutaraldehyde, and embedded in epoxy resin. Eyes were sectioned along a perpendicular plane close to the optic nerve head, and semithin sections were processed for histologic examination after toluidine blue counterstaining. 

Immunohistochemical Labeling of Retinas

Mice that underwent surgery (n = 10 for both GDNF-injected and sham-operation mice) were killed by anesthetic overdose at 23 days. Nonsurgical eyes were also sampled from a further six 23-day-old rd/rd mice. Eyecups were removed and fixed in 4% paraformaldehyde for 2 hours. Retinas were dissected from the posterior eyecup, washed three times in PBS, permeabilized in PBS containing 0.1% Triton X-100 for 5 minutes, and incubated in blocking buffer (PBS containing 0.1% bovine serum albumin, 0.1% Tween 20, and 0.1% sodium azide ([buffer A]) for 15 minutes. They were then incubated with rod photoreceptor-specific monoclonal antibody rho-4D2 ²⁶ (10 μg/ml in buffer A for 1 hour), washed and incubated in goat anti-mouse IgG-Texas red (10 μg/ml in buffer A for 1 hour; Molecular Probes, Portland, OR). Retinas were washed and flatmounted in PBS-glycerol (50:50 vol/vol), with the photoreceptor layer facing up, and examined under a photomicroscope (Optiphot 2; Nikon, Melville, NY) equipped with differential interference and fluorescence optics. 

After stereological counting of rod numbers (see below), five wholemounted immunolabeled retinas from each experimental group were processed further by re-embedding and cryostat sectioning in a transverse plane. Frozen sections were collected from near the central retina, mounted on microscope slides, and examined for rho-4D2 staining. After labeled rods were counted (see later description), sections were incubated with anti-glial fibrillar acidic protein (GFAP) polyclonal antibody (Dako, Trappes, France), diluted 1:400 in buffer A for 2 hours. Sections were washed and incubated in goat anti-rabbit IgG-Bodipy FL (Molecular Probes) and the nuclear dye 4′-6-diamidino-2-phenylindole (DAPI; Sigma–Aldrich, Saint Quentin Fallavier, France), both 10 μg/ml in buffer A for 1 hour. Slides were washed thoroughly and examined as above, with different filter sets allowing visualization of only rho-4D2–immunolabeled rods (single-band-pass, filter XF42), only anti-GFAP-immunolabeled retinal glia (single-band-pass, filter XF22), simultaneous rho-4D2 and anti-GFAP staining (double-band-pass, filter XF53) and DAPI-labeled nuclei (single-band-pass, filter XF03: all fluorescence filters from Omega Optical, division of Molecular Probes). For photography, picture series were taken with similar exposure times (HP5 film; Ilford, Basildon, UK) and printed using identical times. 

Cell Counts

For quantification of rods in flatmounted retinas we used a stereological approach permitting unbiased sampling. ²⁷ Immunolabeled cells were observed by microscope (Optiphot 2; Nikon) under epifluorescence illumination with a plan 40/070 differencial interference contrast (DIC) (160/0.17) objective. One hundred twenty fields of 8000 μm² were selected throughout the entire retinal surface (approximately 13 mm²) using a stage encoder and a systematic random procedure. ²⁸ Fields were viewed with a Sony Trinitron color graphic display camera and digitized software (Automator for Windows; Biocom, Lyon, France). In each of the 120 fields, cells were counted in two unbiased counting frames of 900 μm² generated by the software. The total number of rod cells in the entire retina was then estimated by normalizing these numbers to the entire retinal surface area, which was measured using an image analysis system (Optiscan; Macintosh LC II Ci; Apple Computer, Cupertino, CA). For counting of rods in transverse sections, a single section from each of five retinas for each treatment, passing through the optic nerve head and traversing the entire retinal width, was mounted on a microscope slide and stained with DAPI (10 μg/ml in PBS for 10 minutes). Viewing sections with the two appropriate filter sets allowed visualization of rho-4D2–labeled rods and total DAPI-labeled nuclei. Images were captured on the computer as described along the entire length of each section. 

Reverse Transcription–Coupled Polymerase Chain Reaction Analysis of GDNF mRNA

Total RNA was isolated with an RNA-DNA extraction kit (Qiagen; Valencia, CA) with standard protocol. RNA samples (1 μg) were primed with random hexamer and incubated 2 hours at 37°C with Moloney murine leukemia virus reverse transcriptase. cDNA (1:20; 2 μl) was amplified for 35 cycles with specific pairs of primers. Primer sequences 5′ to 3′were: rhodopsin, AAGCCGATGAGCAACTTCC, TCATC TCCCAGTGGATTCTT; GDNF, ACCAGATAAACAAGCGGCAG, TCAGATACATCCA CACCGTTTAG; and glucose-6-phosphate dehydrogenase (MG6PDH), GCAGTCACCAAGAACATTCAAG, CCCAAATTCATCAA AATAGCCC. Amplified products were visualized on agarose gels stained with ethidium bromide and digitized with a camera. Quantitative measurement of band intensity was made using commercial software (Phoretix International, Newcastle-upon-Tyne, UK). To quantify the amplification products, the exponential phase of the reaction was experimentally determined using different cycling numbers for each primer pair. Under these conditions the amount of the amplification product reflects the initial concentration of transcripts. ²⁹  

Results

ERGs

Figure 1 illustrates the comparative evolution of ERGs from wild-type (C57) and rd/rd mice. In wild-type mice, the amplitude of a- and b-waves steadily increased from postnatal days 12 to 24. In rd/rd mice, the ERG was similar to the wild-type at postnatal day 12, but thereafter a- and b-waves steadily decreased from postnatal day 15 onward, becoming unrecordable by 22 days (Fig. 1) . The light intensity for triggering a signal was 0.3 log cd/m², which corresponds approximately to the threshold for evoking cone responses. 

To determine whether GDNF could improve visual functions in rd/rd mice, subretinal injections of this trophic factor were administered at 13 and 17 days, together with PBS vehicle injections into the contralateral eye to serve as control eyes. ERGs were recorded in animals at 22 postnatal days when ERG signals had disappeared completely in noninjected rd/rd mice. None of the eyes that received PBS exhibited a detectable ERG (n = 10). In contrast, ERGs were measurable in 4 of 10 GDNF-injected eyes (Fig. 2 A, numbers 3, 4, 7, and 9). It should be noted that two of the six mice in which no signal was detected (Fig. 2A , numbers 1 and 5) showed prominent cataracts in both eyes, and they were not considered in the functional scores. A typical ERG recorded from a GDNF-injected eye is shown in Figure 2B . The average amplitudes of a- and b-waves (±SEM) for the four GDNF-injected mice showing a response were 8.5 ± 2.8μ V and 15.5 ± 2.3 μV, respectively. These represent responses that would normally be obtained from 18.5-day-old rd/rd mice, or 22% of the maximal response observed in rd/rd mice (Fig. 1B) . These responses had implicit times (a-wave 120 ± 14 msec; b-wave 242 ± 43 msec; n = 4) similar to those observed in 15- to 16-day-old control rd animals (a-wave 144 ± 10 msec; b-wave 274 ± 23 msec; n = 6) but longer than those of control C57 mice at 24 days (a-wave 31 ± 3 msec; b-wave 110 ± 12 msec; n = 4). These data indicate that GDNF treatment prolonged visual function by approximately 30% relative to the period of massive photoreceptor death in this strain. 

Histologic, Immunohistochemical, and Morphometric Analyses

After ERGs were recorded, the animals that received GDNF and PBS injections (n = 10) and a second group of control mice of the same age (n = 6) were killed and the eyes fixed and dissected. Epon-embedded histologic sections of surgical or control eyes showed that injection of GDNF or PBS did not lead to complications such as neovascularization, retinal scarring, or macrophage invasion by 6 days after the final injection (Fig. 3) . 

Figure 4 shows a low-power view of GDNF-injected (Fig. 4A) and PBS-injected (Fig. 4B) wholemounted rd/rd retinas immunolabeled with rho-4D2. The overall greater numbers of surviving rod photoreceptors after GDNF treatment are apparent. Higher power photographs showed that the antibody stained the entire rod photoreceptor cell, including the cell body and vestigial outer segment (Fig. 5) . The stereological counting frame and the principle of the random systematic sampling procedure used are also shown in Figure 5 . Rod cell counts for individual animals are shown in Figure 6 A (paired GDNF-treated and PBS-injected rd/rd mice ^{1 2 3 4 5 6 7 8 9 10} and nonsurgical rd/rd mice ^{11 12 13 14 15 16} ). The variability between individual mice in opsin-immunolabeled rod numbers is evident, but in each case GDNF treatment led to greater cell counts. Correspondence between the degree of rescue and functionality was suggested by the fact that three of the four mice in which ERGs were detected (Fig. 6A , mice 3, 4, and 7) showed the highest relative differences in rod numbers between GDNF- and PBS-injected animals: 70% for the ERG-positive group (n= 4) and 34% for the ERG-negative group (n = 6). However, the differences between these two groups were not statistically significant. The average (±SEM) rod cell number found in GDNF-injected eyes (89,232 ± 8,033; n = 10) showed a 46% increase compared with the number in PBS-injected eyes (61,165 ± 3,259; n = 10) and a 113% increase compared with control noninjected eyes (41,880 ± 3,890; n = 6; Fig. 6B ). PBS injection also led to an increase (36% more) in rod numbers compared with the number in untreated eyes. These increases were all statistically significant: GDNF versus PBS, P < 0.01; GDNF versus control, P < 0.001; PBS versus control, P < 0.02. These data expressed relative to total rod numbers found in normal adult C57 mice (estimated by us as 4.13 × 10⁶; data not shown) yield 2.2%, 1.5%, and 1.0% for GDNF, PBS, and noninjected rd/rd retinas, respectively. 

To compare these data more easily with previously published results on rod loss in the rd/rd mouse, we sectioned five wholemounted retinas from each experimental group and re-examined rod cell labeling. Figure 7 shows representative areas from each experimental group and clearly shows immunolabeled rods scattered among immunonegative cells within the vestigial outer nuclear layer (ONL). Noninjected control rd/rd retinas had only sparsely scattered rods, appearing as a single discontinuous layer along the scleral border (Figs. 7A 7B) . PBS-injected retinas had more immunolabeled rods, appearing as a single-to-double discontinuous layer (Figs. 7C 7D) . GDNF-injected retinas often contained two to three rows of immunolabeled rods, although areas of rod opsin–immunonegative cells within the ONL were still apparent (Figs. 7E 7F) . Sections were subsequently immunolabeled with anti-GFAP antibody and the two labels visualized simultaneously with a double-band-pass fluorescence filter. GFAP immunoreactivity was visible as bright labeling of astrocytes within the nerve fiber layer and faint transverse fibrous labeling of retinal Müller glia. In noninjected retinas, GFAP immunoreactivity was mostly confined to the astrocytes, with only rare GFAP-immunopositive fibers (Fig. 7A) . PBS-injected retinas showed increased numbers of such fibers (Fig. 7C) , and GDNF-injected retinas contained large numbers (Fig. 7E) . Quantitative analysis of rod numbers from transverse sections revealed an average (±SEM) of 7.70 ± 1.10 rods/120 μm (n = 5) for GDNF-injected, 5.30 ± 0.48 rods/120 μm (n = 5) for PBS-injected, and 3.14 ± 0.13 rods/120 μm (n= 5) for noninjected retinas (Fig. 8) . The difference in rod numbers between GDNF-treated, PBS-treated, and nonsurgical retinas was not significant between GDNF and PBS (P < 0.08), but was significant between GDNF and nonsurgical control eyes (P < 0.005) and between PBS-injected and nonsurgical control eyes (P < 0.005). 

Expression of GDNF mRNA in rd/rd Mice

To study the possible involvement of GDNF as an endogenous trophic factor for rods, we analyzed the expression of endogenous GDNF mRNA in mouse retina (Fig. 9 A). Lane 1 shows that GDNF mRNA was expressed in higher steady state amounts in retina compared with the remainder of the eye (ratio 6:4). Twelve-day-old rd/rd mice expressed rod-specific rhodopsin mRNA at levels comparable to 35-day-old C57 mice (Fig. 9B , lanes 1, 10, and 11). Large decreases in rhodopsin mRNA expression were observed in rd/rd mice between days 12 and 120 (lanes 1 through 9). During the whole period, levels of GDNF mRNA (lanes 12 through 20) and those of the housekeeping enzyme G6PDH (lanes 23 through 31) remained unchanged and were similar to their respective expression intensities in C57 mouse eye at 35 days (lanes 21, 22, 32, and 33). 

Discussion

Our data show a stimulatory effect of subretinal injections of the neurotrophic factor GDNF on rod photoreceptor survival in the rd/rd mouse. This effect was demonstrated through immunohistochemical and morphometric analyses and through measurement of light-evoked responses. To the best of our knowledge this is the first report in which injection of a neurotrophic growth factor has been shown to promote improvement of retinal function in an animal model of retinal degeneration closely resembling human RP. 

Faktorovich et al. ¹⁰ first showed the feasibility of using polypeptide growth factors to slow photoreceptor loss in inherited retinal degeneration. Using a rat model of RP they showed that intraocular injections of FGF-2 delayed rod photoreceptor loss. They also reported recently the effects of intraocular injection of multiple neurotrophic factors (CNTF, brain-derived neurotrophic factor, FGF-2, leukemia inhibitory factor, insulin-like growth factor II, neurotrophin-3 and -4, and nerve growth factor) on slowing rod photoreceptor loss in several mouse models of retinal degeneration including the rd/rd mouse. ⁶ These molecules were chosen because they are known to exhibit neuroprotective effects in other retinal models or elsewhere in the nervous system. ^{5 30 31} Of all these factors, only rat CNTF injected between postnatal days 7 and 10 had a protective effect 1 week later in rd/rd mice. In an independent study, CNTF delivered by adenoviral vectors showed histologic and functional protection of rod photoreceptors in another mutant, the retinal degeneration slow (rds) mouse. ³² Rescue effects of PBS injections alone have not been previously observed in mouse rd mutants, whereas in our study PBS injection alone induced a significant rescue effect at the histologic level, but did not lead to detectable functional improvement in any animal examined. Transient rescue effects of PBS injections have been previously reported in the rat ³³ and are thought to result from local release of endogenous trophic factors as a result of tissue damage. 

GDNF was one of the few neurotrophic factors not tested in the exhaustive recent trials by LaVail et al. ⁶ It was shown recently, when used at relatively high concentrations, to have protective effects against photoreceptor outer segment collapse in short-term in vitro assays. ²⁴ It is impossible to know the final concentration of the subretinally injected GDNF used in the present study, although the dose administered lies in the range normally used for intraocular growth factor injections. ^{6 10} Although it is possible that GDNF may constitute one of the few identified neurotrophic factors influencing rod photoreceptor survival in inherited retinal degenerations in which the gene defect is localized to the rod cell, other criteria such as timing or ocular site of injection and approach were probably critical. Our injections were performed at relatively late times, after the initial onset of rod loss in this strain. ¹⁴ These ages were chosen to permit injection into the subretinal space, which was not routinely possible in younger animals. The stereological method of counting, originally developed for assessment of neuronal numbers in brain regions, ^{27 34} allows an accurate unbiased estimation of total cell numbers, reducing greatly the variability induced by regional differences (superior versus inferior hemisphere, site of injection). Nevertheless, increased rod numbers were also reliably scored in transverse retinal sections as routinely used by other investigators. ^{6 10}  

GDNF may be a particularly interesting neuroprotective molecule, because unlike FGF-2 it is not reported to be angiogenic and thus should not lead to neovascular complications. ³⁵ Histologic observations in the present study also did not record the presence of new retinal blood vessels or invading macrophages. Although this distant member of the transforming growth factor-β superfamily was originally thought to be a specific survival factor for dopaminergic neurons, ¹⁷ GDNF is now known to stimulate a wide variety of autonomic, sensory, and motoneurons ^{21 36 37} and has an important role in kidney development. ³⁸ Previous studies have indicated the potential value of GDNF as a neuroprotective agent for the retina through the demonstration that subnanomolar concentrations of GDNF increase cell survival in enriched photoreceptor cultures prepared from newborn mice and that rat photoreceptor cultures possess high-affinity GDNF receptors. ²³ GDNF is expressed in the rat retina from embryonic day (E) 15 to E19, mostly in the innermost layer. ²² In the mouse embryo, it is expressed from E8.5 in the neuroectoderm surrounding the optic vesicle and later in the mesenchymal components of the developing eye. ³⁹ Another member of the GDNF subfamily, neurturin and its receptor components GFRα2 and Ret, were also recently detected in normal and rd mouse retina throughout the life span. ⁴⁰ In the present study, two histologic observations may be of particular importance. The first is that GDNF injection was effective at slowing rod loss even though treatment was begun after the initiation of rod degeneration in this strain. ⁴¹ Such findings may be of clinical significance, because they suggest growth factor treatment could be effective after the onset of photoreceptor death in RP patients. Enhanced photoreceptor survival after gene transfer of CNTF in rds mice, ³² gene transfer of the normal PDE gene in rd mice, ⁴² or targeted digestion of mutant opsin mRNA by adenovirally delivered ribozymes in transgenic retinal degeneration rats ⁴³ have all been obtained when treatment began long before photoreceptor loss. The second is that intraocular injection of GDNF (and to a lesser extent PBS) led to upregulation of GFAP expression, indicating that, similar to FGF-2, ⁴⁴ this growth factor may regulate phenotypic expression in retinal Müller glia and that its effects on rod survival may be mediated through an indirect pathway. ⁴⁵ We are currently exploring this hypothesis. 

The detection of an ERG in GDNF-injected eyes is an especially important finding in the present study, indicating that visual function was extended in treated rd/rd retina. Our data on ERG decrease in young rd/rd mice are in agreement with several previous studies. ^{46 47 48} The persistence of recordable ERG correlated in three of four cases, with the largest relative increases in rod cell numbers. However, because ERG were absent in several rd/rd retinas containing at least as many rods as one animal in which signals were detected, recording was probably also influenced by anesthesia and cataract formation. Experimental constraints (dark adaptation) prevented monitoring of ocular status before recording, but overt cataract formation was noted in both eyes of two animals that did not produce a detectable ERG. In addition, under the experimental conditions used, rods and cones would both have been stimulated and would have contributed to the signal. It should be stressed that ERG recordings and stereological assessment were performed by a naive observer unaware of which eye had received GDNF. 

We showed previously in the rd/rd mouse model the protective effects of rod photoreceptor transplantation on host cone cell survival and suggested that this effect was mediated by a diffusible factor released by rods. ⁴⁹ Additional evidence for the existence of such diffusible signals produced by normal retina and influencing cone survival was obtained using coculture models. ⁵⁰ Although exogenous GDNF protects rod photoreceptors, it is unlikely to represent this endogenous signal for several reasons. GDNF is expressed in the rd/rd mouse retina even after complete loss of rods, suggesting that GDNF is not expressed in these cells (although it is possible that its expression may be upregulated in inner retinal cells during photoreceptor degeneration as has been shown to occur for FGF-2 and CNTF during retinal injury ⁵¹ ). Because the cone-survival–promoting effect observed in our previous coculture studies was dependent on the presence of rods, ⁵⁰ GDNF is unlikely to constitute this diffusible molecule. Endogenous GDNF is not efficient in counteracting the apoptotic process in mutant rods, and it may have functions other than photoreceptor cell survival in the mouse retina. 

In conclusion, although the endogenous expression of GDNF was not sufficient to prevent photoreceptor loss in the rd/rd mouse, intraocular injections were able to delay such losses significantly. Administration of GDNF using injections or gene delivery systems may be very useful to protect against photoreceptor degeneration in humans. 

 Supported by British Retinitis Pigmentosa Society, Fédération Nationale des Aveugles et Handicapés Visuels, La Fondation de l’Avenir, Retina-France, the A. and M. Suchert Foundation, Université Louis Pasteur, Association pour le Developpement de la Recherche sur la Regeneration Retinienne et sa Transplantation, L’Etablissement Français des Greffes, l’Institut Electricité Santé, and ESSILOR International.

 Submitted for publication July 30, 1998; revised January 20 and April 27, 1999; accepted June 15, 1999.

 Commercial relationships policy: N.

 

Figure 1.

 

Evolution of rd/rd and wild-type mice ERG on light stimulation. (A) ERG recordings at different postnatal ages for rd/rd and wild-type (C57) mice. Postnatal days are indicated on the left of the ERG recordings. (B) Evolution of a- and b-wave amplitude from 12 to 24 postnatal days. Note the continuous decrease toward complete disappearance of the a- and b-waves in rd/rd mice from postnatal days 15 to 24 in contrast to the regular increase observed in wild-type mice. The plot for each age represents data from 5 animals, except for the latest time point for the rd/rd mouse (24 days), which represents 10 animals.

Figure 2.

 

Preservation of recordable ERG in GDNF- compared with PBS-injected 22-day rd/rd mouse. (A) Histogram showing amplitudes of ERG a- and b-waves from individual animals. (B) ERGs recorded from GDNF- and PBS-injected eye. The recordings were made from mouse 7 in (A) and represent average traces from three separate recordings.

Figure 3.

 

Histologic sections of whole eyes from 23-day rd/rd mice injected with (A) GDNF or (B) PBS or (C) noninjected. Semithin epoxy sections were stained with toluidine blue. There was no detectable new blood vessel formation within the injected eyes or visible gliosis or macrophage invasion into the subretinal or vitreal spaces. All micrographs were taken in proximity to the optic nerve head and exposed for the same times. cc, choriocapillaris; rpe, retinal pigmented epithelium; onl, outer nuclear layer; inl, inner nuclear layer; ipl, inner plexiform layer; gcl, ganglion cell layer. Scale bar, 20 μm.

Figure 4.

 

Low-power view of GDNF-injected (A) and PBS-injected (B) wholemounted rd/rd retinas immunolabeled with rho-4D2. The greater numbers of surviving rod photoreceptors after GDNF treatment are apparent. This example represents the mouse showing the greatest difference between the GDNF-treated eye and the PBS-injected contralateral eye. Scale bar, 100 μm.

Figure 5.

 

Illustration of the systematic random sampling approach used (A) and high-power view of 23-day-old rd/rd retina showing aspect of immunolabeled rod cells and superimposed counting frame (B). The antibody stains the entire cell body and vestigial outer segment. Scale bar (B), 10 μm.

Figure 6.

 

Immunolabeled rod cell counts in GDNF-, PBS- and noninjected 23-day rd/rd mice. (A) Total surviving rod numbers for each mouse (x-axis) injected with GDNF in one eye (black bars) and PBS vehicle in the other eye (dark gray bars). The consistent beneficial effect of growth factor treatment compared with vehicle-injected eyes can be observed. Those mice in which ERGs were successfully recorded were numbers 3, 4, 7, and 9 (arrowheads). Total surviving rod numbers for nonsurgical rd/rd retinas from an additional five mice are shown by bars 11 to 15 (light gray). It can be seen that surgery itself led to variation in rod numbers compared with nonsurgical eyes and that PBS injection produced a smaller but statistically significant increment in rod numbers. (B) Average (mean ± SEM) rod cell numbers calculated from (A). The GDNF-induced increase in rod numbers compared with PBS-treated and untreated control eyes is visible. Probabilities indicated above GDNF are comparisons with PBS treatment (1: P < 0.01) and nonsurgical control eyes (2: P < 0.001) and above PBS is comparison with nonsurgical control eyes (3: P < 0.02).

Figure 7.

 

Transverse sections taken from wholemounts of 23-day rd/rd retinas prelabeled with rho-4D2 and relabeled with anti-GFAP. (A) and (B) are the same section from noninjected, (C) and (D) from PBS-injected, and (E) and (F) from GDNF-injected retinas. (A, C, E) Simultaneous visualization of rod opsin and GFAP immunoreactivity; (B, D, F) DAPI labeling of all nuclei. The rho-4D2–immunopositive rods are visible as brightly labeled, small, round cells lying along the vestigial ONL (stars in A, C, E, with the corresponding cell indicated by stars in B, D, F). Rho-4D2–immunonegative cells can be seen as unlabeled areas in the ONL (asterisks in A, C, and E, with corresponding cells indicated by asterisks in B, D, and F). The different number of immunolabeled rods after no injection (n = 16), PBS alone (n = 22), or GDNF treatment (n = ∼40) can be clearly seen (A, C, E, respectively). Also, the increasing intensity of radial Müller glial fibers labeled by anti-GFAP antibody can be seen in the order GDNF-injected > PBS-injected > noninjected (E, C, and A, respectively, arrows), whereas astrocyte staining by the same antibody remains approximately constant along the vitreal border. All micrographs were taken in proximity to the optic nerve head and exposed for the same times. ONL, outer nuclear layer; INL, inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer. Scale bar, 23 μm.

Figure 8.

 

Rod numbers determined from transverse sections of GDNF-, PBS-, and noninjected rd/rd retinas. Average numbers (mean ± SEM) of immunolabeled rods were calculated from the entire width of a single section passing through the optic nerve head from retinas of each experimental group (n = 5). Rod numbers were not statistically different between GDNF- and PBS-treated retinas (1: P < 0.08). Rod numbers were significantly higher in GDNF- compared with noninjected retinas (2: P < 0.005) and in PBS- compared with noninjected retinas (3: P < 0.005).

Figure 9.

 

GDNF mRNA expression was not altered during photoreceptor degeneration in the rd/rd mouse. GDNF mRNA expression was studied by semiquantitative RT-PCR. (A) GDNF expression was compared between retina and the remainder of the eye. Total RNA was isolated from rd/rd retina (lanes 1, 3) and the remainder of the eye (lanes 2, 4) and analyzed with primers specific for GDNF ^{1 2} and G6PDH. ^{3 4} (B) Total RNA isolated from whole eyes of 12- (lanes 1, 12, 23), 22- (lanes 2, 3, 13, 14, 24, 25), 28- (lanes 4, 5, 15, 16, 26, 27), 35- (lanes 6, 7, 17, 18, 28, 29) and 120- (lanes 8, 9, 19, 20, 30, 31) postnatal day rd/rd mice, and from 35-day postnatal normal mice (C57Bl6) (lanes 10, 11, 21, 22, 32, 33) was analyzed with primers specific for rhodopsin (RHO, upper row), GDNF (middle row), and the housekeeping gene glucose-6-phosphate dehydrogenase (G6PDH, lower row). The reduction in opsin mRNA levels during the course of rod degeneration and the uniform profile of steady state mRNA levels for GDNF and G6PDH can be seen.