Because of the accessibility of the optic nerve, optic nerve regeneration (ONR) is often used for studies of central nervous system regeneration. In fish, typified by zebrafish, regeneration of the optic nerve after injury by crushing or transectioning is rapid with new neurites crossing the lesion and entering the optic tectum in as few as 7 days [1]. In mammals, typified by mice, regeneration does not take place in the absence of specific molecular interventions and suppression of astrocyte reactivity in the optic nerve [2, 3] (for a recent review, see [4]).

As part of an ongoing study of ONR in zebrafish [5], we examined intermediate filament (IF) expression of astrocytes in the zebrafish retina and optic nerve. Many previous studies have used the type III IF glial fibrillary acidic protein (GFAP) as a marker for retinal and optic nerve astrocytes, both in fish and mammals, even though it has been known for some time that astrocytes of optic nerves in many fish, including zebrafish, express cytokeratins rather than GFAP [6, 7]. A possible exception are astrocytes of goldfish optic nerve, which, as reported by Nona et al[9], appear GFAP positive both before and after optic nerve injury.

Sections of Tg(gfap:GFP) fish displayed prominent fluorescence that illuminated Müller glial cells in the retina and radial astrocytes in the brain. No GFP fluorescence was detected in the optic nerve, beginning at the optic nerve head and through the chiasma, until the optic tract (Figure 1). Anti-GFAP staining of wild-type ZDR retinae revealed strong localization of GFAP in the foot processes of Müller glia extending inward into the nerve fiber layer, and in the reticular layer formed in the synaptic zone between the photoreceptor and bipolar cells (Figure 2A). Anti-GFAP failed to label any structures in optic nerve sections (data not shown). Sections of Tg(gfap:GFP) retina displayed prominent GFP fluorescence in the Müller glial cells that illuminated their entire length from foot processes on the inner limiting membrane to the terminal outer limiting membrane that forms at the level of the photoreceptor ellipsoids; i.e., between the nuclei and inner segments of the photoreceptors where the intense labeling lateralizes and ends just outer to the nuclear layer (Figure 2B). GFP fluorescence also revealed fine details of the Müller cells including spines along the length of the cells and filamentous arborizations forming the foot processes (Figure 2D). Anti-cytokeratin labeling of sections of Tg(gfap:GFP) retina revealed antigen localization in cells making up the inner limiting membrane, in the bundled retinal ganglion cell (RGC) axon layer, and in scattered cells of the inner plexiform layer (Figure 2C, D; Figure 3A). While it is clear that in both ZDR and Tg(gfap:GFP) fish, Müller glia express GFAP and not cytokeratin, the differences between anti-GFAP staining in ZDR fish retina and GFP expression in the Tg(gfap:GFP) retina suggest that GFP localization in Tg(gfap:GFP) does not accurately reflect where the GFAP protein is localized in the Müller cells. This observation in turn suggests that GFAP in Müller glia contains cytoplasmic localization, or sorting, signals which GFP lacks, as suggested by Bernardos and Raymond [9], the creators of this transgenic fish.

The rapid regeneration of optic nerve in zebrafish as compared non-regeneration in mammals (who express GFAP in their optic nerves) suggests that GFAP itself is non-permissive to axonal regeneration. In mammals, reactive astrogliosis that includes upregulation of GFAP and vimentin provides a neuroprotective effect, particularly in a stroke model. Li et al. [8] found transection of the middle cerebral artery in Gfap(-/-) Vim(-/-) mice generated an infarct that was 210% to 350% larger than in wild type mice. They also report that Gfap(-/-) Vim(-/-) mice show attenuated reactive gliosis and improved post-traumatic regeneration as compared to wild type. In goldfish, Nona et al. [9] reported the presence of GFAP-positive astrocytes 7 days following an optic nerve crush injury on both proximal and distal to the lesion site; however, the injury site itself remained GFAP-negative, and astrocytes were excluded until after axonal regeneration was complete. Thus one could speculate that the absence of GFAP expression in fish optic nerve contributes to an environment that is permissive to nerve regeneration, but there seems to be no evidence that cytokeratins promote regeneration. In the present study, we found no evidence of increased cytokeratin expression in the injured optic nerve as compared to the uninjured. This result is consistent with previous studies of cytokeratin expression during optic nerve regeneration by Fuchs et al. (1994) in goldfish, where no changes in mRNA expression for the goldfish optic nerve cytokeratins GK48 and GK49 were found 10 days post-injury.

Anti-cytokeratin labeling of sections of optic nerve from Tg(gfap:GFP) fish revealed strong cytoplasmic labeling of cells consistent with reticular astrocytes, as described by Macdonald et al. [12] and earlier by Maggs and Scholes [13] (Figure 3B, D). As GFAP expression appears absent in zebrafish optic nerve, confirmation of these cells as astrocytes must depend on morphology, and the pattern of anti-cytokeratin staining seen in Figure 3B is most consistent with the Macdonald et al. [12] description of optic nerve astrocytes and cytokeratin distribution. Neurons would not be expected to label with anti-cytokeratin, and reports of cytokeratin expression in zebrafish oligodendrocytes are absent from the literature. Intermediate filaments of mammalian oligodendrocytes have been characterized as nestin and vimentin [14]. In contrast to the GFP expression observed in Müller glia in retina of Tg(gfap:GFP) fish, no GFP expression was observed in the optic nerves of the same fish. At the optic nerve head, a sharp boundary was present which appeared to exclude GFP-expressing Müller cells from the optic nerve. Cells showing cytokeratin labeling in their cytoplasm were found in the optic nerve head, and appeared to extend into the RGC nerve fiber layer, retinal inner plexiform layer, and may contribute to the inner limiting membrane (Figure 2D, 3C).

On the basis of these results, it appears that if Müller glial cells can be considered astrocytes, zebrafish have two populations of astrocytes in their retina, the GFAP expressing Müller cells, and the cytokeratin expressing reticular astrocytes that appear to extend into the retina from the optic nerve, forming the inner limiting membrane and contributing to the bundled nerve fiber and inner plexiform layers. According to Watanabe and Raff [15], a similar situation exists in mammalian retina with respect to non-Müller astrocytes entering the retina from the optic nerve along retinal vasculature, and in the mature retina, locating near the retinal vasculature and nerve fiber layer (although mammalian astrocytes do express GFAP and not cytokeratin). Because of the apparent absence of GFAP expression by any cell type in the zebrafish optic nerve - either injured or uninjured - studies of the role astrocytes may play during ONR in zebrafish cannot rely on GFAP as an marker for astrocytes or an indicator of reactivity. Future studies of ONR in zebrafish should include evaluation of changes in cytokeratin expression and localization in the optic nerve.