The retina develops as an outgrowth of the neural tube known as the optic vesicle. Although there are many highly specialized cells in the retina, such as photoreceptors and pigmented epithelial cells, the basic mechanisms of neurogenesis are largely conserved with other regions of the central nervous system (CNS). The cells of the optic vesicle act as retinal founder cells, or “stem cells,” in that they have the potential to generate all of the various cells in the retina: five basic types of neurons and one type of intrinsic glial cell, the Müller glial cell. The classic studies carried out by Sidman (1961) first demonstrated that the neurons and Müller glia that make up the mature retina are generated in a specific sequence during histogenesis: ganglion cells, cone photoreceptors, and horizontal cells are generated early in development, while amacrine cells, rod photoreceptors, bipolar cells, and Müller glia are generated in a second cohort. The molecular basis for this change in the genesis of the different types of retinal cells probably reflects changes both within the progenitor cells and in their local microenvironment; however, several different lines of evidence indicate that Müller glia share a lineage and a precursor with retinal neurons (Turner and Cepko, 1987; Holt et al., 1988; Turner et al., 1990).
The developmental relationship between retinal neurons and Müller glia has led to speculation that proliferating glia could serve as a source for retinal regeneration. In the retina of teleost fish, for example, a robust regeneration after surgical lesions results in functional repair. Although it has been difficult to determine precisely which cells are responsible for generating the new retinal cells, the following three sources have been proposed: (1) an intrinsic progenitor cell, known as the rod precursor cell, which normally produces only rod photoreceptors during the growth of the fish eye (Raymond et al., 1988); (2) an intrinsic retinal stem cell that normally has a very long cell cycle and is activated after injury (Julian et al., 1998; Otteson et al., 2001); and (3) the dedifferentiation of Müller glial cells (Raymond and Hitchcock, 1997). In mammals, it has been known for some time that Müller cells can reenter the cell cycle and can be maintained in vitro after retinal damage (Sarthy, 1985; Lewis et al., 1992). More recently, several lines of evidence have indicated a relationship among neural progenitors, stem cells, and glia (see other articles in this issue). In this review, we highlight recent studies from our laboratory that show that Müller glia have the potential to acquire the characteristics of retinal progenitors, both phenotypically and functionally, and that they can serve as a source of retinal regeneration in the chicken.
In a series of recent experiments, we have tested the potential of the avian retina to regenerate after neurotoxic damage. In the first set of studies, we induced extensive cell death of inner retinal neurons, such as amacrine cells and bipolar cells, by making intraocular injections of toxic levels of N-methyl-D-aspartate (NMDA) (Fig. 1A). One day after the injection, we found widespread apoptosis of the inner retinal cells and activation of microglia, presumably acting as phagocytes (Fischer et al., 1998; Fischer and Reh, 2001). Two days after the NMDA injection, we were surprised to find large numbers of mitotically active cells in the retina as shown by BrdU incorporation and immunoreactivity for Phospho-Histone H3 and proliferating cell nuclear antigen (PCNA) (Fig. 1B). By double labeling for BrdU and glutamine synthetase, a marker for Müller glia, we found that 100% of the mitotically active cells were Müller glia 2 days after toxin treatment. However, by the next day, many of the BrdU-labeled cells began to express markers of embryonic retinal progenitors, including the neurogenic bHLH transcription factor CASH-1 and homeodomain transcription factors Pax6 and Chx10 (Fischer and Reh, 2001). In addition, these progenitor-like cells expressed neurofilament transiently at 2–4 days after toxin treatment. Similar to the transient expression of neurofilament, CASH-1 was expressed between 2 and 4 days after toxin treatment, with expression disappearing thereafter.
We followed the fate of the mitotically active cells over the next few weeks by making a single intraocular injection of BrdU 2 days after the NMDA treatment and allowing the animals to survive for varying lengths of time. We can make several conclusions from these experiments. First, similar to embryonic neural progenitors in the developing retina and cortex, Müller glia that reenter the cell cycle undergo interkinetic nuclear migration, entering S-phase with somata located in the center of the inner nuclear layer and continuing to M-phase with somata located in the outer nuclear layer. Second, after a single mitotic division, when the number of BrdU-labeled, Müller glia-derived cells approximately doubles, these cells do not continue to proliferate in vivo. Third, the newly generated cells become distributed throughout the inner and outer nuclear layers of the retina and survive for at least several weeks after damage. Fourth, the BrdU-labeled cells can adopt one of three different fates: a small percentage (less than 4%) of the Müller glia-derived cells differentiate into retinal neurons that express Hu, calretinin, or cellular retinoic acid-binding protein (CRABP), a greater percentage (about 20%) of the newly generated cells differentiate as Müller glia that express glutamine synthetase, and most (about 80%) remain undifferentiated with continued expression of Pax6 and Chx10.
In addition, we found that the proliferation and dedifferentiation of Müller glia in response to acute damage is not uniform across the retina but rather occurs in specific regions of the retina and changes as the animal ages (Fischer and Reh, unpublished observations) (Fig. 2). For example, toxin treatment within the first week after hatching results in proliferating Müller glia in central regions of the retina, while toxin treatment after the first post-hatch week results in proliferating Müller glia in more peripheral regions of the retina. As the animal ages, the region in which proliferating glia are found in response to toxin treatment becomes increasingly confined to peripheral regions of the retina.
The regenerative response of Müller glia to neurotoxic damage has raised many questions: (1) Are the dedifferentiated Müller glia capable of regenerating all retinal cell types, or are they instead acting as fate-restricted “late progenitors”? (2) Are the types of cells regenerated after neurotoxin damage dependent on the types of cells destroyed by the toxin? (3) What molecules activate the proliferation and dedifferentiation of Müller glia after damage? (4) Can Müller cells be stimulated to proliferate in the absence of retinal damage? (5) What is the basis for the regional nature of this response? (6) Why do the glia in central retina ultimately lose their ability to proliferate? (7) Why do so many of the cells generated in response to damage fail to differentiate into neurons or new Müller glia?
To address the first two questions, we tested the Müller glial response to two other toxins: (1) kainate, which destroys bipolar, amacrine, and ganglion cells; and (2) colchicine, which selectively destroys ganglion cells (Fischer and Reh, 2002). We found that intraocular injections of either kainate or colchicine stimulated the proliferation of Müller glia in a manner similar to that which we observed after NMDA injection. In animals treated with either kainate or colchicine, we found that some of the newly generated cells differentiated into cells that expressed markers (Brn3; Islet 1; neurofilament) and had the morphology of ganglion cells (Fig. 3). Because the regeneration of this cell type was never observed after NMDA-induced retinal damage, which does not destroy ganglion cells, the results suggest that the type of neuron destroyed in the retina may allow or promote the regeneration of that neuronal type by the Müller glia-derived progenitor cells. In addition, these results indicate that the Müller glia-derived progenitors can generate a cell type, the retinal ganglion cell, that is normally generated early in retinal development. This suggests that these dedifferentiated Müller cells may have a potential to generate all retinal cell types, not just the late-generated neuronal types. However, we have yet to define conditions that promote photoreceptor production by these cells.
What molecules are responsible for activating Müller glial proliferation after damage? Can the glial cells be stimulated to proliferate in the absence of damage? In response to damage, previous studies have shown that growth factors, including fibroblast growth factors (FGFs), are produced by retinal cells (Kostyk et al., 1994; Wen et al., 1995; Valter et al., 1998; Walsh et al., 2001; Cao et al., 2001). Therefore, it is possible that FGFs produced by damaged retinal cells cause Müller glia to dedifferentiate, proliferate and become progenitor-like cells in toxin-damaged chick retina. Like FGFs, insulin and insulin-like growth factor (IGF) may be involved in the Müller glial response to injury. During embryonic retinal development, IGF-I is transiently expressed by Müller glia and pigmented epithelial cells (Hansson et al., 1989; de la Rosa et al., 1994). To test whether these growth factors are sufficient to activate the regenerative response of Müller glia, we made intraocular injections of FGF2 and insulin, either alone or in combination, and assayed for changes in glial phenotype and proliferation within the retina (Fischer et al., 2002). While insulin or FGF2 alone had no effect, the combination of insulin and FGF2 stimulated the proliferation in Müller cells similar to the response we observed in neurotoxin-damaged retinas. In addition, the progeny of the Müller glia-derived progenitor cells went on to the same fates as those observed after NMDA-induced damage. As there was no evidence of retinal damage in eyes treated with insulin and FGF2, these growth factors are sufficient to initiate a response in Müller glia similar to that observed with neurotoxic damage, and these factors may be normally involved in the damage response. Furthermore, these findings imply that exogenous growth factors might be used to stimulate endogenous glial cells to regenerate neurons throughout the CNS. For example, the combination of insulin and FGF2 stimulated the regeneration of ganglion cells in kainate- and colchicine-treated retinas (Fischer and Reh, 2002). Some of our findings are summarized in Figure 4.
Many questions remain unresolved regarding the response of Müller glia to retinal damage. As noted above, toxin treatment within the first postnatal week results in proliferating Müller in central regions of the retina, while toxin treatment at as late as one month of age results in proliferating Müller glia in more peripheral regions of the retina. We do not know what underlies the regional nature of this response. Why do the glia in central retina ultimately lose their ability to regenerate? It is possible that Müller cells are capable of undergoing dedifferentiation for only a limited period of time after their genesis. The production of Müller glia, like the generation of other retinal cell types, proceeds in a central-to-peripheral wave during development (Prada et al., 1991). For example, there are 7 days between the time the first Müller cells are generated in central retina and when the last ones are generated in the periphery (excluding those produced at the retinal margin). Thus, differences in the timing of cell genesis between central and peripheral Müller cells might underlie the difference in when these cells are capable of proliferating in response to damage. Since the peripherally located Müller cells are capable of regeneration for a much longer time than those in the central retina, perhaps these peripheral cells can undergo this response indefinitely. In addition to differences in birth dates, Müller cells have distinct morphological differences, depending on where they are isolated within the retina (Anezary et al., 2001). Immediately after hatching, many Müller cells have only a single ventricular process, but these cells become progressively more complex in morphology in the central retina. By 1 month after hatching, nearly all these glia in the central retina have more complex branching of the ventricular processes. By contrast, Müller glia in the peripheral retina lag behind in morphological maturation. It is possible that the failure of all Müller glia in the peripheral retina to undergo this stage in maturation is related to their ability to dedifferentiate in response to neurotoxic damage or growth factor injections.
In addition to differences between Müller glia in different regions of the retina, there may be heterogeneity between neighboring Müller glia in the same region of retina. For example, in regions of retina where proliferation occurs in response to NMDA-induced damage, a maximum of 65% of the Müller glia reenter the cell cycle, while the remaining 35% of glia do not (A.J. Fischer and T.A. Reh, unpublished observations). We have not found any molecular differences between proliferating and nonproliferating Müller glia within the same region of retina that might regulate reentry into the cell cycle. However, we have found that Müller glia that increase their expression of glial fibrillary acidic protein (GFAP) in response to damage do not reenter the cell cycle, while the glia that fail to increase their expression of GFAP incorporate BrdU (A.J. Fischer and T.A. Reh, unpublished observations). In addition, some genes normally restricted to the nonpigmented epithelium of the ciliary body (quiescent neurogenic cells) are upregulated in some Müller glia following neurotoxic damage (R. Kubota, C. McGuire, B.D. Dierks, and T.A. Reh, unpublished observations).
We also do not understand why most of the cells generated in response to damage fail to differentiate into either neurons or new Müller glia. There could be environmental restrictions, with the mature retinal environment lacking necessary factors to direct progenitors to neural or glial fates. Alternatively, key cell-intrinsic factors, like proneural transcription factors, might also be absent. Evidence for the first possibility comes from preliminary experiments that are currently under way. We are testing the ability of embryonic retinal progenitor cells to differentiate into various types of neurons when transplanted to the toxin-treated postnatal chicken retina. Our results suggest that the postnatal chicken retina does not support the neuronal differentiation of embryonic progenitor cells; thus, the local environment may also suppress the differentiation of the Müller glial-derived progenitors (Fischer and Reh, unpublished observations). In addition, results of similar experiments in mice show that few brain-derived progenitors differentiate into neurons in the degenerating retina of the retinal degeneration (rd) mouse (Lu et al., 2002). We have also been testing whether there may be intrinsic limits to the potential of the dedifferentiated Müller cells. While these cells transiently express the proneural bHLH transcription factor CASH-1 and homeodomain transcription factors consistent with neural differentiation (e.g., Pax6), the expression of these genes is not sufficient for neurogenesis in all cells. Indeed, overexpression of the proneural transcription factors NeuroD1 or Neurogenin fails to drive most of these cells into neuronal fates in vitro (A.J. Fischer and T.A. Reh, unpublished observations).
While many questions remain unanswered concerning the response of Müller cells to damage in the chicken retina, it is clear that, under normal conditions, Müller glia are not neural stem cells; dedifferentiation and proliferation are required for Müller glia to become progenitors. In chicken, as in all vertebrate species, Müller glia are differentiated support cells that express a complement of cell-specific functional proteins. The relationships among stem cells, progenitors, and glia have undergone considerable revision in recent years, as highlighted in the work we have reviewed in the avian retina. Understanding of the molecular basis of the relationship among these cell types may improve our ability to control the process of neurogenesis and may lead to the exciting possibility of promoting retinal regeneration in higher vertebrates, like ourselves.