Exogenous Growth Factors Stimulate the Regeneration of Ganglion Cells in the Chicken Retina

Transcription

1 Developmental Biology 251, (2002) doi: /dbio Exogenous Growth Factors Stimulate the Regeneration of Ganglion Cells in the Chicken Retina Andy J. Fischer and Thomas A. Reh 1 Department of Biological Structure, University of Washington, Seattle, Washington Recent reports have found that the posthatch chicken retina has the capacity for neuronal regeneration. The purpose of this study was to test whether the types of cells destroyed by neurotoxic lesions influence the types of cells that are regenerated, and whether exogenous growth factors stimulate neural regeneration in the chicken retina. N-methyl-D-aspartate (NMDA) was used to destroy amacrine and bipolar cells; kainate was used to destroy bipolar, amacrine, and ganglion cells; colchicine was used to selectively destroy ganglion cells. Following toxin-induced damage, bromo-deoxyuridine was used to label proliferating cells. In some animals, growth factors were injected into the vitreous chamber of the eye. We found that the proliferation of cells within the retina was stimulated by toxin-induced cell loss, and by insulin and FGF2. After either kainate- or colchicine-induced retinal damage, some of the newly generated cells expressed markers and had the morphology of ganglion cells. The combination of insulin and FGF2 stimulated the regeneration of ganglion cells in kainate- and colchicine-treated retinas. We conclude that exogenous growth factors can be used to stimulate neural regeneration in the retina. We propose that the type of neuron destroyed in the retina may allow or promote the regeneration of that neuronal type Elsevier Science (USA) INTRODUCTION The discovery of neural stem cells in the adult central nervous system has implied that neural regeneration may be possible in birds and mammals. Several recent in vivo studies have demonstrated that under certain circumstances there is limited regeneration of neurons in the cortex of mice (Magavi et al., 2000), vocal nuclei of songbirds (Scharff et al., 2000), and retina of chickens (Fischer and Reh, 2001). While the numbers of newly generated neurons that are produced in response to damage are few, these studies provide hope that regenerative mechanisms might somehow be enhanced to increase the replacement of lost neurons. We have recently found that the combination of insulin and FGF2 injected into the vitreous chamber of the eye activates a neurogenic program in mature Müller glia of the retina (Fischer et al., 2002b). It is possible that these growth factors might stimulate neural regeneration when applied to damaged retinas. One purpose of this study 1 To whom correspondence should be addressed. Fax: (206) tomreh@u.washington.edu. was to test whether insulin and FGF2 stimulate the regeneration of neurons in toxin-damaged retinas. We have recently demonstrated that Müller glia in the chick retina are a potential source of neural regeneration (Fischer and Reh, 2001). In response to NMDA-induced excitotoxic damage, numerous Müller glia dedifferentiate, re-enter the cell-cycle, and express proteins found in embryonic retinal progenitors. The majority of cells generated by proliferating Müller glia remain as undifferentiated progenitor-like cells, while some differentiate into Müller glia and a few differentiate into amacrine or bipolar neurons. We found no evidence that photoreceptors or ganglion cells were regenerated following NMDA-induced excitotoxicity. It is possible that the types of neurons produced by Müller glia-derived progenitors are influenced by the types of cells that are destroyed, since NMDA destroys primarily amacrine and bipolar neurons, while photoreceptor and ganglion cells survive (Tung et al., 1990; Fischer et al., 1998). For example, while projection neurons in the cortex of adult rodents are not normally produced, selective destruction of projection neurons induces the regeneration of this cell type (Magavi et al., 2000). In addition, the destruction of dopaminergic amacrine cells stimulates the selec /02 $ Elsevier Science (USA) All rights reserved. 367

2 368 Fischer and Reh

3 Regeneration of Ganglion Cells in the Retina 369 tive regeneration of this cell type in the larval frog retina (Reh and Tully, 1986). To test whether the projection neurons of the retina could be regenerated, we selectively destroyed ganglion cells by using colchicine or by inducing more general damage that included the destruction of ganglion cells by using kainic acid (KA). KA is a potent neurotoxin in the chick retina, destroying primarily bipolar, amacrine, and ganglion cells (Ingham and Morgan, 1983; Dvorak and Morgan, 1983; Ehrlich et al., 1987). By comparison, when applied on the day of hatching, colchicine selectively destroys about 90% of ganglion cells (Morgan, 1981; Fischer et al., 1999) and nearly all dopamine- and glucagon-containing amacrine cells (Fischer et al., 1999). Here, we present data demonstrating that ganglion cells are regenerated in retinas where ganglion cells have been destroyed. Numbers of newly generated ganglion cells were not increased by insulin or FGF2 alone, but were increased when these factors were co-applied. These newly generated ganglion cells may have been produced by Müller gliaderived progenitors. MATERIALS AND METHODS Animals The use of animals in these experiments was in accordance with the guidelines established by the National Institutes of Health and the University of Washington. Newly hatched leghorn chickens (Gallus gallus domesticus) were obtained from HandN Highline International (Seattle, WA) and kept on a cycle of 16 h light, 8 h dark (lights on at 6:00 a.m.). Chicks were housed in clear Nalgene cages at about 25 C. Chicks received water and Purina chick starter ad libitum. Injections Chicks were anesthetized and injected as described elsewhere (Fischer et al., 1998, 1999; Fischer and Reh, 2000). To destroy retinal cells, we injected 200 nmol (42.6 g) of KA or 0.5 nmol (0.2 g) of colchicine in 20 l of sterile saline into the left eye on the day of hatching, postnatal day 0 (P0). Growth factors were injected every second day for 10 days starting at P5. We injected insulin (2 g/dose) or FGF2 (100 ng/dose), or a combination of these factors, in 20 l sterile saline added with 2 g BrdU and 0.1 mg/ml bovine serum albumin as carrier. All drugs were obtained from Sigma, and all growth factors were obtained from R&D Systems. Fixation, Sectioning, Whole Mounts, and Immunocytochemistry Retinas were dissected, fixed, sectioned, and immunolabeled as described elsewhere (Fischer et al., 1998, 1999; Fischer and Reh 2000). For whole-mounted retinas, we following procedures described elsewhere (Fischer et al., 2002a). Working dilutions and sources of antibodies used in this study included: mouse anti-brdu at 1:50 (Developmental Studies Hybridoma Bank), rat anti-brdu at 1:80 (Accurate Chemicals), mouse anti-neurofilament at 1:2000 (RMO270; Zymed), rabbit anti-brn3.0 at 1:1000 (Dr. E. Turner, University of California San Diego), mouse anti-hu at 1:200 (Monoclonal Antibody Facility, University of Oregon), rabbit anti-calretinin at 1:1000 (Swant, Bellinzona, Switzerland), rabbit anti-visinin at 1:5000 (Dr. R.S. Polans, Dow Neurological Institute, Portland, OR), and rabbit anti-protein kinase C at 1:1000 (Research Diagnostics Inc.). Secondary antibodies included goat anti-rabbit Alexa568, goat anti-mouse Alexa568, and goat anti-rat Alexa488 (Molecular Probes Inc., Eugene, OR) diluted to 1:500 in PBS plus 0.3% Triton X-100. Histology Toluidine blue staining was done as described elsewhere (Fischer et al., 1998, 1999). Photography, Measurements, Cell Counts, and Statistical Analysis Photomicrographs were taken by using a Ziess Axioplan II microscope equipped with epifluorescence, FITC, and rhodamine filter combinations, and a Spot Slider-RT digital camera (Diagnostic Inc.). The data and confocal images in Fig. 3 were obtained by using a Ziess Pascal LSM. Images were optimized for color, brightness, and contrast, and double-labeled images overlaid by using Adobe Photoshop 5.5. Montage figures were made by overlaying the clearly focused portions of images from two or more fields of view. Cell counts were made on at least four whole mounts from different animals, and means and standard errors calculated on data sets from at least four individuals. Data from treated and control eyes were compared statistically with the appropriate Student s t test (StatView for Macintosh). RESULTS NMDA-, colchicine-, and KA-induced damage have been well characterized in central regions of the chick retina (Ingham and Morgan, 1983; Dvorak and Morgan, 1983; Ehrlich et al., 1987; Morgan, 1981; Fischer et al., 1998, FIG. 1. Colchicine- and KA-mediated destruction of ganglion cells is not uniform across the retina. (a d) Vertical sections of retina that were stained with toluidine blue. Retinas were obtained 16 days after a single injection of saline (a), NMDA (b), colchicine (c), or KA (d) on the day of hatching (P0). (e l) Retinal whole mounts that were labeled for neurofilament-immunoreactivity. Retinas were obtained 16 days after an injection of saline (e, i), NMDA (f, j), colchicine (g, k), or KA (h, l). Photographs were taken in dorsal peripheral regions (e h) or central regions (i l) of the retina. Abbreviations: ONL, outer nuclear layer; INL, inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer. The calibration bar (50 m) in (d) applies to (a d), and the bar in (l) applies to (e l). (m) Histogram illustrating the number of ganglion cells per mm 2 that survive treatment with NMDA, colchicine, or kainate.

4 370 Fischer and Reh FIG. 2. Damage and exogenous growth factors stimulate proliferation within the retina. (a h) Whole-mount preparations of the retina that were labeled for BrdU-immunoreactivity. Retinas were obtained from eyes that were treated with saline (a, c, e, g) or colchicine (b, d, f, h) on the day of hatching. Subsequently, eyes were injected with saline (a, b), FGF2 alone (c, d), insulin alone (e, f), or both insulin and FGF2 (g, h) at P3, P5, P7, P9, and P11. Retinas were processed for immunocytochemistry at P16. The calibration bar (50 m) in (h) applies to (a h). (i) Histogram illustrating the mean number of BrdU-labeled cells in dorsal regions of the retina (90,000 m 2 ) with various toxin and growth factor treatments. 1999). However, the extent of damage in peripheral regions of the retina ( 45 o eccentricity) has not been described. Accordingly, we sought to characterize the loss of cells in peripheral regions of the retina that results from intraocular applications of NMDA, colchicine, or KA. We observed the loss of retinal cells in vertical sections of the retina that were stained with toluidine blue. Consistent with previous reports, NMDA, colchicine, and KA induced a loss of thickness of the inner retinal layers (Figs. 1a 1d) (Morgan, 1981; Fischer et al., 1999). The loss of retinal thickness was greatest for KA-treatment, compared with that observed with either NMDA or colchicine treatment (Figs. 1a 1d). These data suggest that KA destroyed more cells than NMDA or colchicine. To assay the loss of ganglion cells, we labeled retinal whole mounts with antibodies to neurofilament. In salinetreated eyes, we observed that the density of neurofilamentpositive axons in the NFL was not uniform across all regions of the retina. For example, the density of axons in the NFL was less in peripheral regions of the retina (Fig. 1e) compared with that of more central regions of the retina (Fig. 1i). While there was no obvious loss of neurofilamentimmunoreactive ganglion cells in retinas treated with NMDA (Figs. 1f and 1j), retinas treated with colchicine or KA suffered massive losses (Figs. 1g, 1h, 1k, and 1l). Colchicine- and KA-induced losses of neurofilamentexpressing ganglion cells were not uniform across the retina. In dorsal regions of retinas treated with colchicine or KA, within several mm of the injection site, the loss of neurofilament-immunoreactive ganglion cells was severe (Figs. 1g and 1h). Many surviving ganglion cells were large, with somata up to 15 m in diameter and dendritic arbors spanning as much as 400 m, while other surviving ganglion cells were smaller, with somata about 7 m in diameter and dendritic arbors about 150 m in diameter. Many more ganglion cells survived in central and ventral regions of the retina (Figs. 1k and 1l) compared with dorsal/peripheral regions of the retina (Figs. 1g and 1h). To determine the extent of ganglion cell loss following treatment with different neurotoxins, we made cell counts from whole-mount preparations. Consistent with previous reports (Chen and Naito, 1999), we found about 10,000 ganglion cells per mm 2 in central regions of the retina, and about 4,000 ganglion cells per mm 2 in dorsal peripheral regions of the retina (Fig. 1m). NMDA treatment did not significantly affect the number of

5 Regeneration of Ganglion Cells in the Retina 371 FIG. 3. Proliferating cells accumulate in different retinal layers depending on the type of damage and treatment with insulin and FGF2. (a d) Confocal images of whole-mounted retinas that were labeled with antibodies to BrdU. Optical sections were obtained from the GCL and NFL, INL, or ONL. Retinas were obtained from eyes that were treated with NMDA or KA and saline or insulin and FGF2. Abbreviations: ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer; NFL, nerve fiber layer; NMDA, N-methyl-Daspartate; KA, kainic acid. The calibration bar (100 m) in (d ) applies to panels (a d). (e) Histogram demonstrating the total number of BrdU-labeled cells in the ONL, INL, or GCL of retinas treated with NMDA or KA and saline or insulin and FGF2. The significance (*, P 0.005; **, P 0.01) of difference between the means of KA/insulin FGF2 and NMDA/insulin FGF2 was assessed by ANOVA followed by a posthoc Student s t test. (f) Histogram illustrating the relative percentages of BrdU-labeled cells in the GCL/NFL, INL, or ONL.

6 372 Fischer and Reh FIG. 4. Damage to the retina results in the regeneration of neurofilament-expressing ganglion cells. (a) Histogram illustrating numbers of BrdU/neurofilament-labeled cells in toxin-damaged retinas treated with saline, insulin alone, FGF2 alone, or insulin plus FGF2. Retinas were damaged by a single intraocular injection of NMDA, colchicine, or KA on the day of hatching. (b m) Retinal whole-mount preparations were double-labeled for BrdU (in green) and neurofilament (in red). Retinas were obtained from eyes that were treated with KA (b j) or colchicine (k m) on the day of hatching. The plane of focus of all images was in the ganglion cell layer. The cells in these images were found between 1 and 4 mm into the retina, away from the retinal margin. The significance (*, P 0.005) of difference between the means of insulin and insulin FGF2 was assessed by ANOVA followed by a posthoc Student s t test. Arrows indicate cells labeled for BrdU and neurofilament. The calibration bar (50 m) in (g) applies to (b g); the bar in (j) applies to (h j); and the bar in (m) applies to (k m).

7 Regeneration of Ganglion Cells in the Retina 373 FIG. 5. The destruction of ganglion cells and treatment with insulin and FGF2 induces the regeneration of Brn3.0-expressing ganglion cells. Whole-mount preparation of retinas that were treated with saline (a), colchicine (b d), or KA (e g) on the day of hatching. Tissues were obtained from eyes that received five consecutive bidaily injections of insulin and FGF2 from P3 to P11. Arrows indicate cells double-labeled for BrdU and Brn3.0. The calibration bar (50 m) in (a) applies to (a) alone, and the bar in (g) applies to (b g). ganglion cells in central and peripheral regions of the retina (Fig. 1m). By contrast, colchicine and KA caused massive losses of ganglion cells in both central and peripheral regions of the retina. In comparison, fewer ganglion cells survived exposure to KA compared with numbers of ganglion cells that survive exposure to colchicine (Fig. 1m) Toxin-Induced Cell Loss and Exogenous Growth Factors Stimulate the Proliferation of Cells in the Retina We have shown previously that NMDA-induced retinal damage causes Müller glia to go through only one round of division, entering S-phase 2 days after the toxin was applied (Fischer and Reh, 2001). To assay cell proliferation in retinas treated with colchicine and KA, we injected the thymidine analogue 5-bromo-2-deoxyuridine along with growth factors into the vitreous chamber. Five injections were made every other day starting at P5, and retinas were processed for immunocytochemistry at P21, 5 days after the final injection. In saline-treated eyes that received injections of BrdU in saline, we found a few BrdU-labeled cells (Figs. 2a and 2i). Most of these cells were in the NFL and GCL (data not shown). In toxin-damaged retinas that received injections of BrdU in saline, we found significantly more BrdU-labeled cells than in undamaged retinas (Figs. 2b and 2i). Undamaged and toxin-treated retinas that received injections of FGF2 alone did not contain increased numbers of BrdU-labeled cells (Figs. 2c, 2d, and 2i). In undamaged retinas that were injected with insulin alone, we found increased numbers of BrdU-labeled cells compared with numbers observed in undamaged, saline-treated eyes (Figs. 2e and 2i). The number of BrdU-labeled cells in retinas treated with insulin alone was increased in toxin-damaged eyes compared with numbers observed in undamaged, insulin-treated eyes (Figs. 2f and 2i). The greatest numbers of BrdU-labeled cells were observed in retinas treated with both insulin and FGF2, regardless of toxin treatment (Figs. 2g 2i). To determine whether toxin-induced damage or injections of growth factors influence where proliferating cells accumulate within the retina, we used laser-scanning confocal microscopy to quantify the distribution of BrdUlabeled cells within the cell layers of the retina. In retinas damaged by NMDA or KA and treated with BrdU in saline, the number of BrdU-labeled cells was not significantly different between toxin treatments, with most of the cells in the GCL/NFL, few in the INL, and none in the ONL (Figs. 3a, 3b, 3e, and 3f). By contrast, in undamaged retinas that received doses of insulin and FGF2, we found increased numbers of BrdU-labeled cells; most of these cells were in the INL and some were found in the ONL (Figs. 3c, 3e, and 3f). Numbers of BrdU-labeled cells in the GCL/NFL, INL, or ONL were not significantly different in retinas damaged by NMDA and treated with insulin and FGF2 compared with numbers observed in undamaged retinas treated with insulin and FGF2 (Figs. 3c, 3e, and 3f). In retinas damaged by KA and treated with insulin and FGF2, increased numbers of BrdU-labeled cells were in the GCL and decreased numbers

8 374 Fischer and Reh in the ONL, compared with numbers of BrdU-labeled cells observed in undamaged and NMDA-damaged retinas treated with insulin and FGF2 (Figs. 3c 3f). This finding suggests that the destruction of ganglion cells by KA and subsequent application of insulin and FGF2 increases the recruitment of BrdU-labeled cells to the GCL. Toxin-Induced Cell Loss and Insulin Plus FGF2 Stimulate the Regeneration of Ganglion Cells To assay for the production of ganglion cells, we used retinal whole-mount preparations that were immunolabeled for BrdU and neurofilament. We did not observe any BrdU/neurofilament-labeled cells in NMDA-treated retinas, where there was little or no loss of ganglion cells (Fig. 4a). In retinas treated with colchicine or KA, we found BrdU/neurofilament-positive cells within the retina. These cells were found primarily in peripheral regions of the retina, within 4 mm from the retinal margin. Most of these cells (78.5%) were found in dorsal regions of the retina, where losses of ganglion cells were greatest. Regardless of the type of toxin treatment, the abundance of newly generated neurofilament-immunoreactive cells within the retina was not increased significantly by treatment with insulin alone or FGF2 alone (Fig. 4a). By comparison, coinjection of insulin and FGF2 significantly (P 0.005) increased the number of BrdU/neurofilament-labeled cells in dorsal regions of retinas damaged by colchicine or KA (Fig. 4a). However, we never observed BrdU/neurofilament-labeled cells in NMDA-treated retinas that were treated with the combination of insulin and FGF2 (Fig. 4a). Regardless of growth factor treatment, there were significantly (P 0.005) more BrdU/neurofilament-labeled cells in retinas treated with KA compared with the numbers of these cells in retinas treated with NMDA or colchicine (Fig. 4a). Cells labeled for BrdU and neurofilament had a variety of morphologies. Some BrdU/neurofilament-labeled cells formed few local processes lacking well-defined dendritic arbors or axons (Figs. 4b 4d). These cells may have been in early stages of differentiation. Other cells appeared to be in an intermediate stage of differentiation with numerous short dendrite-like processes ( 30 m in length; Figs. 4e 4g). Most BrdU/neurofilament-positive cells were relatively small, with somata about 8 m in diameter, and formed numerous peripheral processes (Figs. 4h 4j). A few BrdU/neurofilament-positive cells were large, with somata up to 14 m in diameter, and formed extensive peripheral processes (Figs. 4k 4m). In the chicken retina, neurofilament is express by ganglion cells and efferent target cells (Bennett and DiLullo, 1985a,b; Pittack et al., 1997; Fischer and Stell, 1999). To further investigate the phenotype of BrdU/neurofilamentlabeled cells, we probed retinal whole mounts for BrdU- and Brn3.0-immunoreactivity. Brn3.0 is a homeotic transcription factor that contains a POU domain and is known to be required for the differentiation and survival of retinal ganglion cells (Xiang et al., 1993; Gan et al., 1996; Liu et al., 2001). In the postnatal chicken retina, Brn3.0 is expressed by only orthotopic ganglion cells (Fischer et al., 2002a). In whole-mount preparations of undamaged retina, we found numerous Brn3.0-immunoreactive cells in the GCL (Fig. 5a). In eyes that were damaged by colchicine or KA and received injections of insulin and FGF2, we found cells that were labeled for BrdU and Brn3.0 (Figs. 5b 5g). This suggests that many of the newly generated neurofilamentlabeled cells were ganglion cells. These cells were observed up to 4 mm away from the retinal margin, toward central regions of the retina. To confirm findings in retinal whole mounts, we probed for newly generated neurons in retinal sections. In both normal and toxin-treated retinas, neurofilamentimmunoreactivity is weak in the somata of ganglion cells and is difficult to clearly localize. Therefore, we probed for newly generated neurons by using antibodies to Hu, an RNA-binding protein related to ELAV proteins of Drosophila (Marusich et al., 1994; Barami et al., 1995). In the postnatal chick retina, Hu is expressed by most, if not all, amacrine and ganglion cells (Fischer and Reh, 2000, 2001). In retinas damaged by colchicine or KA, we found Huimmunolabeled cells that were positive for BrdU, regardless of growth factor injections. Most of these cells (57 of 84 cells counted) were found near the GCL (Figs. 6a 6c), while a few of these cells were found in the INL (20/84; Figs. 6d 6f). Occasionally, we observed Hu/BrdU-labeled cells in the ONL (7/84; Figs. 6g 6i), where this cell type normally is not situated, suggesting that the distribution of newly generated neurons in toxin-damaged retinas may be abnormal. By comparison, we never observed neurofilament/ BrdU-positive cells in retinal layers other than the GCL. Hu/BrdU-positive cells were found up to 4 mm from the peripheral edge of the retina. We found many BrdU-labeled cells in the ONL and distal INL, where the cell bodies of photoreceptor, horizontal and bipolar cells, respectively, are known to reside. However, we did not find any BrdU-labeled cells that were colabeled with the photoreceptor marker visinin, the horizontal cell marker calretinin, or the bipolar cell marker protein kinase C. DISCUSSION Here, we report that neural regeneration in the chicken retina can include the regeneration of ganglion cells. This cell type was regenerated only in retinas where ganglion cells were destroyed and in regions of retina with significant losses of ganglion cells (Fig. 7). We also found that the combination of insulin and FGF2 stimulated the regeneration of ganglion cells, while these factors applied separately had no effect on the number of regenerated neurons. Some of the BrdU/neurofilament-positive cells in toxintreated retinas were likely to be newly generated ganglion cells. In normal chicken retinas, neurofilament is expressed selectively by ganglion cells (Torelli et al., 1989; Bradshaw et al., 1995; Pittack et al., 1997) and efferent target cells

9 Regeneration of Ganglion Cells in the Retina 375 (Fischer and Stell, 1999). However, neurofilament is expressed transiently by Müller glia in following toxininduced retinal damage (Fischer and Reh, 2001) or following intraocular injections of exogenous insulin and FGF2 (Fischer and Reh, unpublished observation). In both of these cases, the expression of neurofilament lasted only a few days and returned to normal by 4 days after treatment. In the current study, we probed for the expression of neurofilament at least 5 days after the final dose of growth factor. Therefore, levels of neurofilament expression should have fallen below detectable levels in Müller glia by the time we made our observations. Further, none of the BrdU/ neurofilament-positive cells had the morphology of Müller glia; instead, these cells had neuronal morphology (see Fig. 4). Further, we detected BrdU in the nuclei of cells that expressed the ganglion cell-specific markers Brn3.0 and Hu, confirming that at least some of the newly generated cells in retinas damaged by colchicine or KA are ganglion cells. In addition to the neurofilament/brdu-labeled cells, we observed many cells labeled with BrdU alone in saline- and insulin-treated retinas in the GCL or NFL. These cells may have been oligodendrocytes. In the chick retina, although astrocytes are confined to the optice nerve head (Schuck et al., 2000), oligodendrocytes migrate into the retina from the optic nerve (Huxlin et al., 1992; Ono et al., 1998) and reside in the GCL and NFL (Villegas, 1960; Nakazawa et al., 1993; Ono et al., 1998). Oligodendrocytes have been shown to proliferate within the chick retina at about E14 (Nakazawa et al., 1993). It is possible that insulin and FGF2 stimulated the proliferation of oligodendrocytes in the postnatal chick retina, and many of the BrdU-labeled cells that we observed in the GCL and NFL were oligodendrocytes. One potential source for the regenerated ganglion cells is the Müller glia. We have recently reported that Müller glia in the postnatal chicken retina are a source of neural progenitors (Fischer and Reh, 2001). We demonstrated that in response to sufficient damage elicited by NMDA, numerous Müller glia re-enter the cell cycle, dedifferentiate, express genes common to embryonic retinal progenitors, and produce some amacrine and bipolar neurons and some glia. In addition, we have recently reported that intraocular injections of insulin and FGF2 stimulated the production of new neurons from mature, postmitotic Müller glia in the absence of retinal damage (Fischer et al., 2002b). The combination of insulin and FGF2 was required to stimulate the proliferation and transdifferentiation of Müller glia, while insulin alone or FGF2 alone had no effect. Similarly, in the present study, we found that the combination of insulin and FGF2 increased the number of newly generated ganglion cells, while insulin alone or FGF2 alone had no effect. Taken together, these findings suggest that injections of insulin and FGF2-stimulated Müller glia to regenerate ganglion cells in toxin-damaged retinas. An alternative source of the regenerated ganglion cells is the progenitors at the retinal margin. The progeny of these cells could have migrated laterally into the retina and, subsequently, differentiated into ganglion cells. In this study, we observed newly generated neurons up to 4 mm away from the retinal margin (Fig. 6). Given that the first application of BrdU was about 2 weeks prior to observation, ample time elapsed for a migrating cell to travel 4 mm. However, long-distance lateral migration of neuroblastic cells within the retina has not been previously observed. A final possibility is that regenerated ganglion cells were derived from quiescent stem cells that are seed within the retina. For example, a population of quiescent stem cells has been described in the adult teleost retina (Julian et al., 1998; Otteson et al., 2001). Quiescent neural stem cells may also exist in the postnatal chicken retina, but there is currently no evidence to support this hypothesis. The mechanisms underlying the synergistic activity of insulin and FGF2 remain uncertain. We report here that insulin or FGF2 alone did not increase numbers of regenerated ganglion cells, while the combination of these two factors had a significant effect. Insulin and FGF have been shown to act in synergy in a variety of biological systems (Chamberlain et al., 1991; Werther et al., 1993; Frodin and Gammeltoft, 1994; Reape et al., 1996; Liu et al., 1996). We have reported elsewhere that in undamaged retinas the combination of insulin and FGF2 stimulates the production of ganglion cells by progenitors at the retinal margin (Fischer et al., 2002a) and activates a neurogenic program in postmitotic Müller glia in the absence of retinal damage (Fischer et al., 2002b). Receptors for FGF2 and insulin/igf are likely to be coexpressed by Müller glia. Receptors for FGF2 and insulin/igf are expressed throughout the chick retina (Waldbillig et al., 1991; de la Rosa et al., 1994; Rohrer et al., 1997; unpublished observations). Activation of receptors for FGFs or insulin results in the activation of a MAP kinase signaling cascade (Boulton et al., 1991; reviewed by Szebenyi and Fallon, 1999). In the retina, activated MAP kinase signaling is commonly observed in Müller glia (Peng et al., 1998; Wahlin et al., 2000; Kinkl et al., 2001; Takeda et al., 2002). It is possible that the cumulative activation of the MAP kinase pathway by both insulin and FGF2 is required to activate a neurogenic program in Müller glia in the chick retina, and thus increase the number of regenerated ganglion cells in toxin-damaged retinas. Our findings suggest that the loss of ganglion cells somehow allows or promotes their regeneration. Alternatively, the types of neurons that survive toxin treatment in the chick retina may regulate the types of cells that are regenerated. We have reported previously that bipolar and amacrine neurons are regenerated in retina damaged by a single toxic dose of NMDA (Fischer and Reh, 2001). In NMDA-treated retinas, we did not observe the regeneration of any cells in the GCL, including displaced cholinergic amacrine cells (Fischer and Reh, 2001). NMDA destroys primarily amacrine and bipolar cells, while leaving photoreceptor and ganglion cells intact (Tung et al., 1990; Fischer et al., 1998). By comparison, we observed regenerated ganglion cells in retinas that were damaged by colchicine or KA. We found that these toxins destroyed numerous ganglion cells, consistent with the reports of others (Ingham

10 376 Fischer and Reh FIG. 6. Damage to the retina results in the regeneration of Hu-expressing retinal neurons. Vertical retinal sections were double-labeled for BrdU (in green) and Hu (in red). Retinas were obtained from eyes that were treated with colchicine on the day of hatching, and received doses of insulin plus FGF2 every second day for 10 days starting at P3. Retinas were processed for immunocytochemistry at P16. Hu/BrdU-labeled cells were found in GCL (a c), INL (d f), or ONL (g i). The calibration bar (50 m) in (f) applies to (d f), and the bar in (i) applies to (a c) and (g i). Abbreviations: ONL, outer nuclear layer; INL, inner nuclear layer; IPL, inner plexiform layer. and Morgan, 1983; Dvorak and Morgan, 1983; Morgan, 1981; Fischer et al., 1999). It is possible that the destruction of ganglion cells allows some of the progeny of proliferating Müller glia to differentiate as new ganglion cells. Consistent with this hypothesis, we observed increased numbers of regenerated neurons where more cell loss had occurred. For example, more ganglion cells were regenerated in KAtreated retinas compared with colchicine-treated retinas, and KA caused greater ganglion cell loss than colchicine. In addition, the regenerative response was greater in dorsal retinal regions near the site of toxin injection. Consistent with these findings, the selective destruction of dopaminergic amacrine cells in the larval tadpole retina results in the selective regeneration of this cell type at the retinal margin (Reh and Tully, 1986), and KA-induced retinal cell loss causes in an increase in the production of inner retinal neurons from stem cells at the retinal margin (Reh, 1987). Further, Magavi et al. (2000) reported that selective ablation of projection neurons in the cortex of adult mice selectively stimulated the regeneration of this cell type. It is possible

11 Regeneration of Ganglion Cells in the Retina 377 FIG. 7. Schematic diagram of the eye that demonstrates the region of the peripheral retina (in gray) where we found regenerated neurons. This region is posterior to CMZ (in light gray) and is anterior to central retinal regions (in dark gray). The diagram represents an axial section through the dorsal hemisphere of the eye. The diagram is drawn to scale in the axial dimension, but the thickness of the retina is exaggerated to increase clarity. Abbreviation: CMZ, ciliary marginal zone. that destroying a particular cell type allows for the replacement of that cell type. This process may occur via a feedback mechanism (Hinds and Hinds, 1974; Reh, 1987). For example, the production and/or survival of ganglion cells may be regulated by a feedback mechanism during embryonic development (Waid and McLoon, 1998; Gonzalez-Hoyuela et al., 2000). We conclude that the types of cells regenerated in the retina are influenced by the types of cells that are destroyed. We further conclude that the numbers of regenerated neurons can be increased by exogenous growth factors. We propose that exogenous growth factors may used to stimulate glial cells to generate neurons in the retina and other regions of the central nervous system. ACKNOWLEDGMENTS We thank Blair Dierks and Christopher McGuire for providing expert technical assistance. This work was supported by fellowships from the Alberta Heritage Foundation for Medical Research and the Canadian Institutes of Health Research (to A.J.F.); and by NIH RO1 EY13475, NSF IBN , and a grant from the Foundation Fighting Blindness (to T.A.R.). REFERENCES Barami, K., Iversen, F., Furneaux, H., and Goldman, S. A. (1995). Hu protein as an early marker of neuronal phenotypic differentiation by subependymal zone cells of the adult songbird forebrain. J. Neurobiol. 28, Bennett, G. S., and DiLullo, C. (1985a). Expression of a neurofilament protein by the precursors of a subpopulation of ventral spinal cord neurons. Dev. Biol. 107, Bennett, G. S., and DiLullo, C. (1985b). Transient expression of a neurofilament protein by replicating neuroepithelial cells of the embryonic chick brain. Dev. Biol. 107, Boulton, T. G., Nye, S. H., Robbins, D. J., Ip, N. Y., Radziejewska, E., Morgenbesser, S. D., DePinho, R. A., Panayotatos, N., Cobb, M. H., and Yancopoulos, G. D. (1991). ERKs: A family of protein-serine/threonine kinases that are activated and tyrosine phosphorylated in response to insulin and NGF. Cell 65, Bradshaw, A. D., McNagny, K. M., Gervin, D. B., Cann, G. M., Graf, T., and Clegg, D. O. (1995). Integrin alpha 2 beta 1 mediates interactions between developing embryonic retinal cells and collagen. Development 121, Chamberlain, C. G., McAvoy, J. W., and Richardson, N. A. (1991). The effects of insulin and basic fibroblast growth factor on fibre differentiation in rat lens epithelial explants. Growth Factors 4, Chen, Y., and Naito, J. (1999). A quantitative analysis of cells in the ganglion cell layer of the chick retina. Brain Behav. Evol. 53, de la Rosa, E. J., Bondy, C. A., Hernandez-Snachez, C., Wu, X., Zhou, J., Lopez-Carranza, A., Scavo, L. M., and de Pablo, F. (1994). Insulin and insulin-like growth factor system components gene expression in the chicken retina from early neurogenesis until late development and their effect on neuroepithelial cells. Eur. J. Neurosci. 6,

12 378 Fischer and Reh Dvorak, D. R. and Morgan, I. G. (1983). Intravitreal kainic acid permanently eliminates off-pathways from chicken retina. Neurosci. Lett. 36, Ehrlich, D., Teuchert, G., and Morgan, I. G. (1987). Specific ganglion cell death induced by intravitreal kainic acid in the chicken retina. Brain Res. 415, Fischer, A. J., Dierks, B. D., and Reh, T. A. (2002a). Exogenous growth factors induce the production of ganglion cells at the retinal margin. Development 129, Fischer, A. J., Dierks, B. D., McGuire, C., and Reh, T. A. (2002b). Insulin and FGF2 activate a neurogenic program in Müller glia. J. Neurosci., in press. Fischer, A. J., Morgan, I. G., and Stell, W. K. (1999). Colchicine induces excessive ocular growth and myopia in chicks. Vis. Res. 39, Fischer, A. J. and Reh, T. A. (2001). Müller glia are a potential source of neural regeneration in the post-natal chicken retina. Nat. Neurosci. 4, Fischer, A. J., and Reh, T. A. (2000). Identification of a proliferating marginal zone of retinal progenitors in the postnatal chicken. Dev. Biol. 220, Fischer, A. J., Seltner, R. L. P., Poon, J., and Stell, W. K. (1998). Immunocytochemical characterization of NMDA and QAinduced excitotoxicity in the retina of chicks. J. Comp. Neurol. 393, Fischer, A. J., and Stell, W. K. (1999). Nitric oxide synthasecontaining cells in the retina, pigmented epithelium, choroid and sclera of the chick eye. J. Comp. Neurol. 405, Frodin, M., and Gammeltoft, S. (1994). Insulin-like growth factor acts synergistically with basic fibroblast growth factor and nerve growth factor to promote chromaffin cell proliferation. Proc. Natl. Acad. Sci. USA 91, Gan, L., Xiang, M., Zhou, L., Wagner, D. S., Klien, W. H., and Nathans, J. (1996). POU domain factor Brn-3b is required for the development of a large set of retinal ganglion cells. Proc. Natl. Acad. Sci. USA 93, Gonzalez-Hoyuela, M., Barbas, J. A., and Rodrigues-Tebar, A. (2000). The autoregulation of retinal ganglion cell number. Development 128, Hinds, J. W., and Hinds, P. L. (1974). Early ganglion cell differentiation in the mouse retina: an electron microscopic analysis utilizing serial sections. Dev. Biol. 37, Huxlin, K. R., Sefton, A. J., and Furby, J. H. (1992). The origin and development of retinal astrocytes in the mouse. J. Neurocytol. 21, Ingham, C. A., and Morgan, I. G. (1983). Dose-dependent effects of intravitreal kainic acid on specific cell types in chicken retina. Neuroscience 9, Julian, D., Ennis, K., and Korenbrot, J. I. (1998). Birth and fate of proliferative cells in the inner nuclear layer of the mature fish retina. J. Comp. Neurol. 394, Kinkl, N., Sahel, J., and Hicks, D. (2001). Alternate FGF2-ERK1/2 signaling pathways in retinal photoreceptor and glial cells in vitro. J. Biol. Chem. 276, Liu, J., Chamberlain, C. G., and McAvoy, J. W. (1996). IGF enhancement of FGF-induced fibre differentiation and DNA synthesis in lens explants. Exp. Eye Res. 63, Liu, W., Mo, Z., and Xiang, M. (2001). The Ath5 proneural genes function upstream of Brn3 POU domain transcription factor genes to promote retinal ganglion cell development. Proc. Natl. Acad. Sci. USA 98, Magavi, S. S., Leavitt, B. R., and Macklis, J. D. (2000). Induction of neurogenesis in the neocortex of adult mice. Nature 405, Marusich, M. F., Furneaux, H. M., Henion, P. D., and Weston, J. A. (1994). Hu neuronal proteins are expressed in proliferating neurogenic cells. J. Neurobiol. 25, Morgan, I. G. (1981). Intraocular colchicine selectively destroys immature ganglion cells in the chicken retina. Neurosci. Lett. 24, Nakazawa, T., Tachi, S., Aikawa, E., and Ihnuma, M. (1993). Formation of the myelinated nerve fiber layer in the chicken retina. Glia 8, Ono, K., Tsumori, T., Kishi, T., Yokota, S., and Yasui, Y. (1998). Developmental appearance of oligodendrocytes in the embryonic chick retina. J. Comp. Neurol. 398, Otteson, D. C., D Costa, A. R., and Hitchcock, P. F. (2001). Putative stem cells and the lineage of rod photoreceptors in the mature retina of the goldfish. Dev. Biol. 232, Peng, M., Li, Y., Luo, Z., Liu, C., Laties, A. M., and Wen, R. (1998). Alpha2-adrenergic agonists selectively activate extracellular signal-regulated kinases in Muller cells in vivo. Invest. Ophthalmol. Vis. Sci. 39, Pittack, C., Grunwald, G. B., and Reh, T. A. (1997). Fibroblast growth factors are necessary for neural retina but not pigmented epithelium differentiation in chick embryos. Development 124, Reape, T. J., Kanczler, J. M., Ward, J. P., and Thomas, C. R. (1996). IGF-I increases bfgf-induced mitogenesis and upregulates FGFR-1 in rabbit vascular smooth muscle cells. Am. J. Physiol. 270, Reh, T. A. (1987). Cell-specific regulation of neuronal production in the larval frog retina. J. Neurosci. 7, Rohrer, B., Tao, J., and Stell, W. K. (1997). Basic fibroblast growth factor, its high- and low-affinity receptors, and their relationship to form-deprivation myopia in the chick. Neuroscience 79, Reh, T. A., and Tully, T. (1986). Regulation of tyrosine hydroxylase-containing amacrine cell number in larval frog retina. Dev. Biol. 114, Scharff, C., Kirn, J. R., Grossman, M., Macklis, J. D., and Nottebohm, F. (2000). Targeted neuronal death affects neuronal replacement and vocal behavior in adult songbirds. Neuron 25, Schuck, J., Holger, G., and Wolburg, H. (2000). The peripapillary glia of the optic nerve head in the chicken retina. Anat. Rec. 259, Szebenyi, G., and Fallon, J. F. (1999). Fibroblast growth factors as multifunctional signaling factors. Int. Rev. Cytol. 185, Takeda, M., Takamiya, A., Yoshida, A., and Kiyama, H. (2002). Extracellular signal-related kinase activation predominantly in Muller cells of retina with endotoxin-induced uveitis. Invest. Ophthalmol. Vis. Sci. 43, Torelli, S., Sogos, V., Marilli, M. A., D Atri, M., and Gremo, F. (1989). Developmental expression of intermediate filament proteins in the chick embryo retina: In vivo and in vitro comparison. Exp. Biol. 48, Tung, N. N., Morgan, I. G., and Ehrlich, D. (1990). A quantitative analysis of the effects of excitatory neurotoxins on retinal ganglion cells in the chick. Vis. Neurosci. 4, Villegas, G. M. (1960). Electron microscopic study of the vertebrate retina. J. Gen. Physiol. 43,

13 Regeneration of Ganglion Cells in the Retina 379 Waid, D. K., and McLoon, S. C. (1998). Ganglion cells influence the fate of dividing retinal cells in culture. Development 125, Wahlin, K. J., Campochiaro, P. A., Zack, D. J., and Adler, R. (2000). Neurotrophic factors cause activation of intracellular signaling pathways in Muller cells and other cells of the inner retina, but not photoreceptors. Invest. Ophthalmol. Vis. Sci. 41, Waldbillig, R. J., Arnold, D. R., Fletcher, R. T., and Chader, G. J. (1991). Insulin and IGF-1 binding in developing chick neural retina and pigmented epithelium: A characterization of binding and structural differences. Exp. Eye Res. 53, Werther, G. A., Cheesman, H., and Russo, V. (1993). Olfactory bulb organ culture is supported by combined insulin-like growth factor-i and basic fibroblast growth factor. Brain Res. 617, Xiang, M., Zhou, L., Peng, Y. W., Eddy, R. L., Shows, T. B., and Nathans, J. (1993). Brn-3b: A POU domain gene expressed in a subset of retinal ganglion cells. Neuron 11, Received for publication February 19, 2002 Revised August 6, 2002 Accepted August 7, 2002 Published online October 10, 2002