Anti-semaphorin 3A Antibodies Rescue Retinal Ganglion Cells from Cell Death following Optic Nerve Axotomy*

Damage to the optic nerve in mammals induces retrograde degeneration and apoptosis of the retinal ganglion cell (RGC) bodies. The mechanisms that mediate the response of the neuronal cells to the axonal injury are still unknown. We have previously shown that semaphorins, axon guidance molecules with repulsive cues, are capable of mediating apoptosis in cultured neuronal cells (Shirvan, A., Ziv, I., Fleminger, G., Shina, R., He, Z., Brudo, I., Melamed, E., and Brazilai, A. (1999) J. Neurochem. 73, 961–971). In this study, we examined the involvement of semaphorins in an in vivo experimental animal model of complete axotomy of the rat optic nerve. We demonstrate that a marked induction of type III semaphorin proteins takes place in ipsilateral retinas at early stages following axotomy, well before any morphological signs of RGC apoptosis can be detected. Time course analysis revealed that a peak of expression occurred after 2–3 days and then declined. A small conserved peptide derived from semaphorin 3A that was previously shown to induce neuronal death in culture was capable of inducing RGC loss upon its intravitreous injection into the rat eye. Moreover, we demonstrate a marked inhibition of RGC loss when axotomized eyes were co-treated by intravitreous injection of function-blocking antibodies against the semaphorin 3A-derived peptide. Marked neuronal protection from degeneration was also observed when the antibodies were applied 24 h post-injury. We therefore suggest that semaphorins are key proteins that modulate the cell fate of axotomized RGC. Neutralization of the semaphorin repulsive function may serve as a promising new approach for treatment of traumatic injury in the adult mammalian central nervous system or of ophthalmologic diseases such as glaucoma and ischemic optic neuropathy that induce apoptotic RGC death.

Axon guidance molecules are a large family of secreted and membrane-bound proteins that participate in axonal pathway formation during development of the nervous system in a receptor-mediated process (reviewed in Refs. [1][2][3][4]. In particular, several members of the semaphorin family of proteins may play an important role in navigating axonal networks throughout neuronal development via their chemorepulsive activity on neuronal growth cones (5)(6)(7). The basic structure of sema includes a semaphorin domain of 500 bp as well as immunoglobulin-like and thrombospondin domains. Semaphorins are currently grouped into nine subclasses, which contain both secreted and transmembranous members (3,8). Inhibition of axonal outgrowth was demonstrated by some semaphorins; it occurs at the tip of the growth cone and is manifested by depolymerization and loss of F-actin (9,10). The downstream pathways by which semaphorins exert their actions are still unclear, but are known to include the neuropilin and plexin receptors (11,12) as well as intracellular CRMP-62 (collapsin response mediator protein) (13) and G-proteins (14). Interestingly, it has recently been shown that semaphorins are bifunctional molecules, and cyclic nucleotides can regulate their activity from repulsion to attraction (15,16). Involvement of semaphorin family members in other physiological phenomena was suggested, such as regulation of organ formation outside the nervous system (17), regulation of the immune response (18,19), and tumor-related cell survival mechanisms (20,21).
We have recently shown a functional involvement of semaphorins in neuronal apoptosis (22). In a model system of cultured chick sympathetic neurons undergoing dopamine-induced apoptosis, induction of both collapsin-1, now termed chick semaphorin 3A (Sema3A 1 ; a secreted and diffusible type III semaphorin with repulsive activity), and CRMP-62 was detected. This up-regulation was evident at the early stages following exposure to the apoptotic trigger, even before the emergence of the typical morphological signs of apoptosis. The up-regulation of Sema3A expression was coupled to the onset of the commitment time point of the neurons to the death process. Moreover, marked and prolonged protection of neurons from dopamine-induced apoptosis was achieved by co-treatment of cells with function-blocking anti-Sema3A antibodies. Antibodies against neuropilin-1, the putative semaphorin III/Sema3A receptor (23,24), also inhibited neuronal apoptosis. In addition, induction of apoptosis was evident by treatment of neurons with a recombinant human semaphorin III protein (22). These results demonstrated a correlation between the semaphorin death-inducing activity and the repulsive activity on axonal growth cones. We therefore proposed that repulsive diffusible cues, acting on appropriate receptor-bearing cells, could modulate neuronal cell fate (22). Further support for our studies was provided by the indication of involvement of semaphorins in apoptosis of sensory neurons (25) and neuronal progenitor cells (26).
The molecular mechanisms responsible for transforming the repellent guidance cue from the damaged axon into a death signal that may affect the cell body are unknown. Therefore, it was important to further substantiate and extend our in vitro findings to an experimental in vivo model of neuronal apoptosis. One of the models used to study the fate of injured neurons in the adult central nervous system is the visual system, in which retinal apoptosis occurs following axonal injury to the optic nerve. In rats, transection of the optic nerve close to the cell bodies is a trigger for retrograde degeneration and delayed death of retinal ganglion cells (RGC) (27). The mechanism that initiates the loss of retinal cell bodies as a function of axonal injury is still unknown, and several hypotheses have been suggested (reviewed in Ref. 28). However, many lines of evidence indicate that apoptotic processes are responsible for the loss of RGC in the retina following axotomy of the optic nerve and for the loss of other axotomized or traumatized central nervous system neurons (27, 29 -31). This secondary death of the RGC also serves as an example of the correlation between neuronal cell survival and axonal integrity. Such a correlation is supported by studies indicating that when an axon is injured, the cell body goes through a series of changes that may culminate in the death of the axotomized neurons (32).
Following optic nerve axotomy in adult rats, initial reduction of the original population of RGC is observed 5-7 days after injury. A rapid process of cell death then takes place, and Ͼ80% of the RGC die within 2 weeks after the injury (27,(33)(34)(35). Because a relatively long interval precedes this massive death process, it can be speculated that a time window exists in which protective intervention with modulators that attenuate or prevent cell death can be employed. Indeed, partial and temporal rescue of RGC from axotomy-induced apoptosis was obtained by several approaches. These included early treatment with microglial inhibitors (36,37), glial or brain-derived neurotrophic factors (38,39), caspase inhibitors (40,41), implantation of stimulated macrophages (42), and overexpression of Bcl-2 (43,44). Therefore, we chose the visual system as an in vivo model in which neuronal apoptosis can be induced and, more importantly, that can also offer an opportunity to intervene and alter the fate of the injured RGC.
We now report that high levels of semaphorin expression are transiently induced in the rat retina at an early post-axotomy stage and before any RGC death is observed. Furthermore, we demonstrate that rescue of RGC from axotomy-induced degeneration can be achieved by treatment of the injured retina with intraocular administration of anti-Sema3A function-blocking antibodies.

Optic Nerve Transection in the Rat
Adult male Sprague-Dawley rats (8 -10 weeks old, 300 g) were deeply anesthetized (50 mg/kg xylazine and 35 mg/kg ketamine), and their left optic nerves were exposed by lateral canthotomy. The conjuctivae were incised lateral to the cornea, and the retractor bulbus muscles were separated. Through a small opening in the menings (ϳ200 m), the nerve fibers were completely transected at a distance of 2-3 mm from the globe. The procedure was performed without damage to the nerve vasculature and optic nerve blood supply and with minimal damage to the meninges by the use of a specially designed glass dissector with a 200-m tip and a smooth blunt edge (45). The injury was unilateral in all animals, and the other eye served as a control.

Retrograde Labeling of RGC
At day 8 following axotomy of the optic nerve, the rats were deeply anesthetized. Small crystals of the lipophilic neurotracer dye 4-(4-(didecylamino)styryl)-N-methylpyridinium iodide (4-Di-10-Asp; Molecular Probes, Inc.) were dissolved gently in incomplete Freund's adjuvant (at a concentration of 1 mM; Difco). This dye serves as a marker for living RGC because it is transmitted through the axonal network and stains the cell bodies of live neurons only, whereas nonviable neurons as well as other cell types such as endothelial cells remained unstained (42). The dye was applied to the transected nerve 0.5 mm from the proximal border of the transection site. The site of injury was visible by its grayish color in comparison with the rest of the nerve, which maintained its original color. Four days after dye application, retinas were excised, whole-mounted on Millipore filters, fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS), and viewed under a fluorescent microscope using a fluorescein isothiocyanate filter.

Evaluation of RGC Survival
Two independent and blinded researchers counted the number of labeled RGC in flat-mounted retinas by fluorescent microscopy. For each retina, 20 representative microscopic fields were evaluated: 10 fields from the peripheral area and 10 fields from the central area (each field covering an area of 0.069 mm 2 ). Central retinal areas were defined as located within two-thirds from the optic disc, and peripheral retinal areas were defined as located within one-third of the retinal radius. (These definitions were based on the apparent change in density of RGC between the two central and peripheral areas.) For comparison, we performed a previously described counting method (35) that divides each quadrant of the retina into three different areas and counts a total of two microscopic fields from each area of the retina (making a total of 12 fields). However, when the two counting protocols were compared (by counting the RGC in control retinas), we found that the numbers came out the same. Therefore, we saw no advantage in using the previously described method, and we used the above protocol for the entire analysis. Retinas from eyes that were subjected to axotomy contained less RGC and were compared with intact retinas.

Data Analysis
Histological parameters are presented as means Ϯ S.E. Mann-Whitney and analysis of variance tests were used to assess the results.

Treatment with Sema3A-derived Peptides
Sema3A-derived peptides (Pep1A, corresponding to amino acids 363-380 from chick Sema3A, and the control peptide Pep3A, corresponding to amino acids 240 -256 according to Ref. 46) were synthesized by Genemed Biotechnologies, Inc. (San Francisco, CA) and purified to homogeneity on a RP-18 column (Vydac, Hesperia, CA) using a Gilson Model 303 HPLC system. Prior to injection, each peptide was dissolved in PBS, and a total amount of 0.33 mg/kg in a volume of 2 l was injected into the vitreous body of one eye, and the opposite eye served as a control. Insertion was at the corneal limbal border behind the lens, over the optic nerve head area and close to the retinal surface.

Antibody Preparation
Polyclonal Anti-Sema3A Antibodies-Polyclonal anti-Sema3A antibodies were prepared in rabbits by immunization against Pep1A, a peptide corresponding to amino acids 363-380 from chick Sema3A (46). The sequence of this peptide is identical in chick Sema3A, human semaphorin III, and rat and mouse semaphorin D. These antibodies were therefore expected to react with the highly conserved Sema3A/ semaphorin III/semaphorin D proteins present in those species. Synthetic peptides were obtained and purified as described above. Peptide conjugates with keyhole limpet hemocyanin (KLH; Sigma) were prepared by incubation of the peptide (3 mg) with 1 ml of KLH containing 10 mg of protein and 1.25% glutaraldehyde. After 3-4 min at room temperature, the reaction was stopped by addition of 0.1 ml of 1 M glycine in water. Two rabbits were immunized against the peptide. Rabbits were immunized by subcutaneous injections of 0.1 mg of KLHconjugated peptide. Booster shots were injected 2, 5, and 8 weeks after immunization. The titer of the antibodies in the serum was 1:10 4 as determined by enzyme-linked immunosorbent assay and compared with preimmune serum. Both the immune and preimmune sera were fractionated using 40% saturated ammonium sulfate and dialyzed against PBS. These antibodies were previously characterized and shown to react with the chick Sema3A protein and to act as functionblocking antibodies that block the repellent activity of human sema-phorin III (22). These antibodies were used in this study for inhibiting the effect of axotomy.
Monoclonal Anti-semaphorin Antibodies-Balb/c mice were immunized subcutaneously with 50 g of KLH-conjugated Pep1A in complete Freund's adjuvant and boosted four times subcutaneously with 50 g of KLH-conjugated Pep1A in incomplete Freund's adjuvant at 3-week intervals. Three days before fusion, mice were boosted by intravenously injection of 50 g of KLH-conjugated Pep1A in PBS. One mouse that had a relatively high titer (as evidenced by enzyme-linked immunosorbent assay with bovine serum albumin-conjugated Pep1A) and that was also positive in Western blot analysis, immunohistochemistry, and immunoprecipitation was chosen for fusion. Hybridomas were prepared by fusion of spleen cells with mouse myeloma NS0 cells (donated by Dr. C. Milstein. Following the cloning procedure, positive clones were chosen for the preparation of ascitic fluids, which was carried out by intraperitoneal injection of 2-5 ϫ 10 6 hybridoma cells into pristane-primed male mice. The fluid was harvested starting at 7-10 days and then twice every 2nd day. Ascitic fluids were centrifuged at 800 ϫ g and kept frozen at Ϫ80°C. The antibodies were characterized by their ability to react with a single protein with the right molecular mass (110 kDa) in chick embryo retinas at embryonic day 18, by their ability to immunoprecipitate this protein, and by their ability to label the ganglion cell layer of embryonic day 18 retinal slices. These antibodies were used in this study for both immunohistochemistry and Western blot analysis.

Protective Antibody Treatment
For protective antibody treatment, the axotomized animals were grouped into four groups of four to six rats each. Two groups were injected with the partially purified polyclonal anti-Sema3A antibody (at a single dose of 30 g in a volume of 2 l). One of the groups was injected immediately after optic nerve transection (time 0), and the second group was injected with the antibody 24 h post-injury. The other two groups served as controls: one group was injected with preimmune serum (30 g in a volume of 2 l, taken from the same rabbit used for immunization), and the second with PBS. The anti-Sema3A antibody was injected into the vitreous body of the axotomized eye. Injection was performed using a glass pipette that was inserted into the eye globe. Insertion was at the corneal limbic border behind the lens, over the optic nerve head area and close to the retinal surface. Injection of the anti-Sema3A antibody did not cause any signs of inflammation or irritation or any other changes that may indicate that this treatment is toxic. Clinical examination of the eyes was performed using a slit lamp, and the eyes were monitored for any signs of hyperemia, edema, discharge, fibrin, and other inflammatory parameters.

Western Blot Analysis
Western blot analysis was performed as described by Harlow and Lane (65) using 8% polyacrylamide gels. Each lane was loaded with an equal amount of protein extracts (175 g), which, following electrophoresis, were transferred to an Immobilon (polyvinylidene disulfide) membrane for 1.5 h. Blots were stained with Ponceau to verify equal loading and transfer of proteins. Membranes were then probed with monoclonal anti-semaphorin antibodies (1:1000 dilution). The intensity of the signal was determined using the ECL-Plus detection system (Amersham Biosciences, Buckinghamshire, UK).

Immunohistochemistry
Tissue Preparation-Dissected retinas from optic nerve-axotomized rat eyes and from non-injured control were from optic nerve-axotomized rat eyes, and controls were fixed in 3.5% paraformaldehyde in PBS for 1-2 h and then infiltrated for cryoprotection with 4% sucrose for 2 h at 4°C, followed by 20% sucrose ϩ 5% glycerol in PBS at the same temperature, overnight. Fixed tissue was quickly frozen in liquid nitrogen. Cross-sections (10 m) were placed on subbed slides (1% gelatin containing 0.3% chromium potassium sulfate) and stored at Ϫ20°C.
Labeling-Sections were washed with PBS and incubated in PBS and 1% bovine serum albumin for 30 min at room temperature. The sections were incubated overnight with the primary monoclonal antibody against Sema3A at 4°C. The slides were then washed three times with PBS and incubated with fluorescein isothiocyanate-conjugated goat anti-rabbit IgG for 1 h at room temperature. After washing three times with PBS, the sections were mounted with galvanol. Observations were made and photography was carried out with a Zeiss Axioplan microscope with an appropriate filter. As a control, the sections were reacted either with preimmune rabbit serum or with the secondary antibody only.

Hoechst Staining
Cross-sections were stained with a cell-permeable form of bisbenzimide (Hoechst 33342) at a concentration of 4 g/ml for 30 min at 37°C. Cells were visualized by UV light microscopy, and the relative number of apoptotic nuclei with condensed or fragmented chromatin was then evaluated and compared with non-apoptotic nuclei, showing a pale and diffuse staining. Photomicrographs using a UV filter were taken for documentation.

Experimental Model of Complete Axotomy of the Rat Optic
Nerve-In the adult rat, injury to the optic nerve leads to delayed apoptotic death of the RGC. To establish the experimental model, rats were subjected to intrameningeal complete axotomy of the optic nerve in one eye, whereas the second eye served as a control. To demonstrate the effect of axotomy on RGC in this experimental system, two approaches were used. One method estimated the number of live RGC that were still detected 12 days following the injury. The RGC from one eye that was subjected to optic nerve axotomy were retrogradelabeled by application (at day 8) of the neurotracer dye 4-Di-10-Asp, and the retina was fixed and flat-mounted 4 days later. Non-axotomized eyes taken from different normal rats (that were not subjected to optic nerve axotomy) served as controls. The retinas were then observed under a fluorescence microscope (Fig. 1). The axotomized retinas appeared to be depleted, with very few live RGC compared with intact retinas, in which multiple cells were observed, and the large RGC exhibited long neuritic processes. The number of stained (live) RGC was counted, and the results are presented in Table I. At 12 days after the injury (8 days after axotomy ϩ 4 days of staining), a massive loss of RGC was observed in both the central and peripheral retinal areas. The survival rates were 35 and 52%, respectively. The higher death rate of the central RGC may be explained by the proximity to the injury site, which was close to the cell bodies of these neurons. The overall RGC death in our experimental system was compatible with results from previously published protocols following 8 days of axotomy (35) and was ϳ60%.
The second method evaluated whether apoptotic cells can be detected in retinal cross-sections following 12 days of axotomy. For this purpose, retinal sections were stained with Hoechst 33342, after which the RGC were examined for the presence of chromatin changes. The relative number of cells in the axotomized eye with condensed or fragmented chromatin, indicating apoptosis, was evaluated and compared with that in control retinas of non-axotomized eyes, which showed only a pale and diffuse staining. Fig. 2 shows that, as a result of axotomy, there was a massive loss of the ganglion cells in the injured eye compared with the non-injured eye. Twelve days after the damage, the estimated density of ganglion cells/mm 2 in this layer was dramatically decreased. Moreover, a significant number of apoptotic cells could be identified using the Hoechst nuclear staining method (Fig. 2), which identifies pyknotic nuclei and chromatin fragmentation as hallmarks of apoptosis. In comparison, no apoptotic nuclei could be detected in the non-axotomized eye. Although the number of nuclei showing apoptotic morphology in these retinas is small, it is consistent with previous estimates that cells undergoing death in vivo are eliminated from the tissue within 3 h or less (47).
Induction of Semaphorin Expression in the Retina following Axotomy of the Optic Nerve-To examine whether secreted semaphorins participate in the degeneration process of RGC, we followed the retinal expression pattern of semaphorins in the above model system of optic nerve axotomy. Following axotomy in one eye, the retinas from both eyes were removed at different time intervals and used for immunohistochemical analysis with the anti-Sema3A antibody. As controls, we used retinas from the non-axotomized eyes of the same rats as well as retinas from eyes of different rats that were not subjected to axotomy. Four to six rats were analyzed at each time point.
Temporal analysis of semaphorin expression as a function of advancement of the death process was performed. Retinas at 1-4, 10, 12, and 16 days following axotomy were analyzed, and data from several time points are presented in Fig. 3. Axotomy initiated a gradual increase in semaphorin expression, starting 24 h post-injury and peaking at 48 -72 h. At 96 h, staining declined, indicating decreased expression. No additional wave of semaphorin expression could be detected at days 10, 12, and 16, at which time the RGC death process was fully manifested (data not shown).
The up-regulation of semaphorins reflects induction of their expression in the RGC layer and the inner nuclear layer, where the amacrine and bipolar cells reside. Some label was also detected in the outer nuclear layer, where the horizontal cells are found. It is possible that the induction of semaphorin expression is transmitted through a transsynaptic pathway, which may explain why the labeling of semaphorins could be detected in cell structures of neural origin.
Western blot analysis of retinas prepared at different time points revealed two specific bands that reacted with the antisemaphorin antibodies and showed elevated levels following axotomy (Fig. 4). One protein has a molecular mass of 110 kDa and corresponds to the full-length semaphorin protein (46). A lower molecular mass protein of 36 kDa appeared to be upregulated and may represent cleavage of a degradation product of the full-length protein. Furthermore, at 72 h post-axotomy, another protein band of 38 kDa could be detected, which appeared only at this time point; and its relation to the other two bands has yet to be investigated. Quantitative analysis of the results showed that, whereas the higher band peaked at 48 h  I Effect of anti-Sema3A antibody on RGC survival following axotomy Shown are RGC numbers 8 days after optic nerve axotomy either with no treatment or with intraocular injection of preimmune serum (immediately following injury) or anti-Sema3A antibodies. The numbers in the Survival columns show percentage of change in the number of RGC/treatment. All values (central and peripheral areas and average) were tested for statistical significance using the Mann-Whitney nonparametric test. RGC numbers differed significantly between control and axotomized and between axotomized versus antibody-treated injured eyes (see also Fig. 6C). These results indicate that the up-regulation of semaphorins in the retina peaks shortly after induction of damage, at a time point that precedes the manifestation of the actual death process of the RGC. High expression of semaphorins in neurons following nerve injury could have a potentially inhibitory effect on the outgrowth of nerve sprouts, thereby preventing regeneration processes.
The Sema3A-derived Peptide Is Capable of Inducing RGC Loss-Our previous studies indicate that exposure of neuronal cells to the full-length Sema3A protein can induce apoptotic death in cultured neurons, and a similar process was observed when neurons were exposed to a small conserved peptide derived from Sema3A (Pep1A) (48). The apoptotic process, accompanied by shrinkage of the axonal network was observed in both cases, suggesting that Pep1A can mimic the death-inducing activity of the full-length protein (48). Based on the evidence of involvement of semaphorins in RGC loss following axotomy, it was of interest to test whether direct application of the semaphorin-derived peptide to healthy animals has an effect on survival of RGC. As a control, a different Sema3Aderived peptide (Pep3A) was applied to retinas of healthy rats. Pep3A was previously shown to be nontoxic in neuronal cultures (48) and therefore was tested as a control peptide. Injection of Pep1A into healthy eyes resulted in 57% neuronal loss compared with only 9% RGC death in Pep3A-treated rats, indicating that the death-inducing activity is specific to Pep1A (Fig. 5). These results suggest that Pep1A has toxic effects when applied both in vitro (48) and in vivo and may imply that the exposure of RGC to secreted Sema3A can propagate the death process within neighboring RGC.
Rescue of RGC from Axotomy-induced Degeneration by Anti-Sema3A Antibodies-To examine whether the function-block- ing antibodies raised against Sema3A are capable of attenuating the axotomy-induced degeneration of RGC in the rat model, antibodies were injected into the axotomized eyes, and RGC viability was monitored. We chose intravitreous injections because, in this closed system, injected compounds remain in the vitreous body for a long period (49). Moreover, the close proximity of the vitreous body to the cell bodies of the RGC layer facilitates the delivery of molecules to their destined targets. Indeed, intravitreous injection has been successfully used in multiple studies as an effective mode of administration of compounds to the retina (50 -55). Such administration was also suggested as a pharmacological tool for treatment of the retina (51,52).
For the neuroprotection experiments, rats that were axotomized in one eye were divided randomly into four treatment groups (four to six animals/group). In the first group, immediately after conducting the axotomy, PBS solution was administered to the vitreous body of the damaged eye. In the second group, 30 g of the IgG fraction from preimmune serum (originating from the same rabbit used for preparation of anti-Sema3A antibodies) was injected into the vitreous body of the damaged eye. In the third group, an IgG fraction from anti-Sema3A antibodies at a single dose of 30 g (0.1 mg/kg) was administered. In the fourth group, identical quantities of anti-Sema3A antibodies were injected 24 h post-axotomy. Eight days later, the RGC were retrograde-labeled with the neurotracer dye 4-Di-10-Asp (which is specifically stains live neurons). The retinas were removed 4 days following the staining (12 days post-injury), and the number of stained neurons was counted. In addition, retinas from untreated animals served as a reference for evaluating the protection level achieved by the different treatments. The retinas were analyzed by fluorescence microscopy, and the results are presented in Fig. 6 (A-C). A massive loss of RGC was observed in retinas from axotomized eyes treated with PBS, and these retinas appeared to be very depleted. Similar retinal RGC depletion was obtained when the injured eyes were injected with preimmune serum (data not shown). However, remarkably higher RGC densities comparable to those in axotomized retinas were displayed in retinas treated with anti-Sema3A antibodies (Fig. 6A). Furthermore, the surviving large neuronal cells appeared to have long processes of axons, indicating that the axonal network remained intact (Fig. 6B). For quantitative analysis of the results, we evaluated the distribution of the surviving retinal neurons. The number of live (stained) RGC in each experimental group was further estimated by counting the RGC in the central and peripheral retinal areas (for specific location of the central and peripheral areas, see "Materials and Methods"). The results are shown in Fig. 6C and Table I. In untreated animals, ϳ65% (central area) and 50% (peripheral area) of the RGC population had degenerated 8 days following optic nerve axotomy (comparable to the results obtained by Isenman et al. (35)). However, treatment of the axotomized retinas with anti-Sema3A antibodies either at the time of axotomy (designated time 0) or 24 h FIG. 4. Up-regulation of semaphorin extracted from the rat retina following axotomy. A, retinas from axotomized eyes were dissected at different intervals (0 (control (C)), 24, 48, 72, and 96 h) after lesion, and total retinal proteins were extracted, separated on 12.5% polyacrylamide gel, and blotted onto polyvinylidene disulfide membrane. The membrane was stained with Ponceau, after which the blot was reacted with the monoclonal anti-semaphorin antibody (1:1000 dilution). The secondary antibody was horseradish peroxidase-conjugated goat anti-mouse IgG (1:25,000 dilution). The blots were developed using the ECL system. To verify that equal amounts of protein were transferred, the blots were stripped and reacted with the mouse monoclonal anti-tubulin antibody (1:25,000 dilution). The secondary antibody was horseradish peroxidase-conjugated goat anti-mouse IgG (1: 25,000 dilution). B, quantitative analysis of bands was performed using TINA software, and data are normalized to the corresponding tubulin. The results are represented as percent of control untreated animals. Three retinas isolated from three different animals were used per time point.

FIG. 5. Intraretinal injection of Sema3A-derived peptide
Pep1A leads to RGC loss. Both peptides Pep1A and Pep3A were injected at a concentration of 0.33 mg/kg; and 8 days later, survival of the RGC was monitored following the application of the neurotracer dye as described under "Materials and Methods." A, whole-mounted retinas showing 4-Di-10-Asp-labeled RGC cells that were exposed to PBS, Pep1A, and Pep3A. Excessive RGC loss was evident when Pep1A (but not Pep3A) was applied. Bar ϭ 25 m. B, quantitative analysis of Pep1A-and Pep3A-treated RGC. Results are presented as means Ϯ S.E. of live cells/mm 2 . Statistical analyses were performed with the Mann-Whitney non-parametric test. ***, p Ͻ 0.01. later resulted in full protection, and retinas were indistinguishable from those of the control animals ( Fig. 6C and Table I). Treatment of retinas at time 0 with preimmune serum had no protective effect (Fig. 6, A and C). The RGC from the central retinal areas compared with the peripheral areas were more sensitive to the apoptotic trigger, showing lower survival rates, and were also less responsive to the antibody treatment. Overall, these results indicate that treatment of axotomized eyes with the anti-Sema3A antibody leads to a marked increase in the number of surviving RGC in the central as well as peripheral areas of the retina, which, with no other treatment, would otherwise suffer a massive loss of RGC. These findings further substantiate the concept that axon guidance molecules with repulsive cues play a role in degeneration of RGC in response to axotomy. The ability for neuroprotection at time points of up to 24 h following the injury indicates the existence of a therapeutic time window during which RGC death is still reversible and can be modulated. DISCUSSION Our study shows that a peak of induction of semaphorin expression in the RGC and in the inner plexiform layer at the early stages following complete axotomy of the optic nerve. This up-regulation was detected in the RGC and in the inner plexiform layer of the axotomized eyes and preceded the actual death process of the ganglion cells, which is the inevitable outcome of such optic nerve injury in mammalians. Moreover, we have shown that, by neutralizing semaphorin repulsive activity using a single intravitreous injection of function-blocking anti-Sema3A antibodies, we could prolong the survival of RGC and obtain dramatic protection from axotomy-induced degeneration. These results are consistent with our previous in vitro experiments on cultured neuronal cells exposed to an apoptosis-inducing challenge (22). In these studies, we have shown that a peak of induction of Sema3A expression occurred just prior to the commitment of sympathetic neurons to apoptosis, and prevention of cell death was achieved using the same anti-Sema3A antibodies (22). Although the in vitro experimental system utilized peripheral neuronal cells harvested from post-mitotic chick embryos, and the animal model described in this study focused on central nervous system ganglion cells from the rat retina, taken together, these data indicate the existence of a common pathway. The correlation between the pivotal roles played by semaphorins as mediators of neuronal fate in both the in vitro and in vivo models suggests that diffusible chemorepulsants may participate in a crucial step of the apoptotic cascade. Indeed, a diffusible apoptosispromoting activity that is heat-labile was isolated from R28 retinal cell-conditioned medium (56). The heat lability of the compound may indicate that it is likely to be a secreted protein. This apoptosis-inducing activity has not yet been character- ized, but it is supportive of our findings suggesting that semaphorins are candidates for serving as diffusible apoptosis-promoting elements.
More importantly, we have shown that apoptosis can be blocked during its early stages by antibodies that inhibit the axonal collapse activity. Indeed, recent lines of evidence support the view that repulsive cues are not restricted to the early stages of neurodevelopment, but may also play a role in the response of axons to their environment during adulthood. For instance, adult sensory neurons were shown to retain the ability to respond to semaphorin III (57). In addition, expression of semaphorin III was recently shown in selective adult brain tissue (58). It is also likely that semaphorins are involved in the pathogenesis of human neuronal degenerative diseases. A further example of the role of semaphorins in pathological conditions was reported by Fujita et al. (59). They found transient up-regulation of neuropilin-1, neuropilin-2, and semaphorin 3A in rat brain after occlusion of the cerebral artery distal to striate branches. They suggest that neuropilin-1, neuropilin-2, and semaphorin 3A are involved in microglial/neuron interactions to phagocytose the dying neurons. Increased levels of CRMP-62, a protein involved in the intracellular response to Sema3A, are associated with the pathological process of neurofibrillary tangle formation in Alzheimer's disease brains (60). Another study described an altered expression pattern of semaphorin IV in Alzheimer's disease brains (61). Although semaphorin IV expression does not co-localize with CRMP-2 expression in Alzheimer's disease brains (61), these findings may still link its accumulation with some of the pathological changes accompanying nerve degeneration. The emerging concept therefore suggests that, as a result of injury, neurons may react by up-regulation of proteins with a repulsive signal.
Furthermore, we have shown that, by applying the functionblocking anti-Sema3A antibodies to rats following optic nerve injury, we were able to protect the RGC. Several possible explanations can be suggested for the inhibitory mechanism displayed by neutralizing repulsive signals. The first is based on the apoptosis-inducing capability of semaphorins. When adult neurons are exposed to unscheduled high levels of semaphorins, death signals are conveyed to the cell bodies, leading to neuronal apoptosis (22,25). Evidently, neutralizing the activity of these proteins can prevent their toxic effects. A second possible explanation is that the increase in repulsive signals can contribute to unfavorable conditions leading to the failure of regenerative processes, thus culminating in neuronal cell death. Indeed, decreased Sema3A mRNA expression was recently reported as a temporary event accompanying peripheral nerve crush (62). This down-regulation was interpreted as an important mechanism enabling regenerative attempts that occur after nerve injury (62). In accordance with this concept, a transient up-regulation of semaphorin III mRNA was indicated at the lesion site following axotomy of olfactory axons and in neural scar tissue after injury to the adult central nervous system (62,63). It was even speculated that, by inhibiting the activity of repulsive cues, promotion of regeneration processes could take place (62). A third possible explanation is that anti-Sema3A antibodies interfere with some other regulatory step of axotomy-induced apoptosis that is non-relevant to inhibition of repulsive signals. Obviously, more studies are needed to gain a better understanding of the molecular events that underlie the rescue of RGC from axotomy-induced death. It is still unresolved whether neuropilin-1, the downstream receptor of semaphorin III (23,24), is involved in conveying the repulsive cue into a death signal in axotomized RGC. Although neuropilin-1 is known to be down-regulated in adult rat RGC (64), it appears from our study that the RGC retain their responsiveness to chemorepulsive signals following optic nerve injury. Further support for a possible involvement of neuropilin-1 in this process comes from our studies in cultured neuronal cells, in which function-blocking anti-neuropilin-1 antibodies were able to inhibit the death process (22).
The induction of semaphorins in the rat retina occurred during the early stages of the time window that separates the retrograde death of the RGC from the onset of the injury (i.e. axotomy). We have taken advantage of this time lag and of the temporal up-regulation of semaphorin expression to treat the damaged eye with a neuroprotective antibody. The delayed administration of the antibody did not impair its neuroprotective ability, suggesting that a high percentage of the neurons are still in the reversible stage and that it is therefore still possible to rescue RGC even 24 h after the injury and to achieve a high survival rate. It may be speculated that, in the case of optic nerve injury in humans, physicians will have a time window of at least 24 h to evaluate the clinical situation prior to treating the patients with neuroprotective treatment or antiapoptotic drugs. There are many eye conditions in which neuroprotective treatment can rescue RGC until the primary cause of injury to the optic nerve and retina can be abolished. Among them are acute increase in intraocular pressure, sudden occlusion of the central retinal artery, ischemic optic neuropathy, retinal vein occlusion, and others (Refs. 66 -68; for review, see Ref. 69). It is tempting to speculate that this neuroprotective approach may be applied not only to the visual system, but also to injuries to the nervous system function. Obviously, further studies need to be employed to test this theory. Our successful results obtained in this study raise the possibility that blockage of axon guidance molecules with chemorepulsive activity can serve (either alone or in combination with other neuroprotective agents) as an important novel strategy for modulation of neurodegeneration and of traumatic injury in the central nervous system.
It is of interest that the kinetics of semaphorin induction following optic nerve axotomy are similar to those of c-Jun after optic nerve crush (30). Both are rapidly up-regulated during the early days post-injury; they reach a maximum of labeling intensity after 3 days and then decline. Therefore, the list of immediate-early genes whose expression is up-regulated in response to apoptotic triggers may also be extended to include the semaphorins, which can be used as early predictors of retinal degeneration.
In summary, we have described a model in which RGC survival in the retina following optic nerve injury was enhanced by inhibiting repulsive cues. Our results suggest that diffusible chemorepulsive signals may have a direct effect on the survival of injured neurons and that a therapeutic benefit can be obtained by inhibiting their function. In our experiments, we did not include functional studies of the surviving neurons, and further studies of their ability to receive and integrate synaptic inputs are still needed. In addition, future studies should also examine whether longer-term survival can be obtained using this strategy.