THE JOURNAL OF COMPARATIVE NEUROLOGY 312:549-560 (1991)

Gliosis During Optic Fiber Regeneration in the Goldfish: An Immunohistochemical Study R.L. LEVINE Department of BioloD, McGill University, Montreal, Quebec H3A 1B1, Canada

ABSTRACT Antisera directed against the 48 kDa and 50 kDa cytoskeletal antigens were used to examine changes in the astroglial fabric of the goldfish visual pathways following optic nerve crush. Several major observations are described. First, an optic nerve crush lesion in these animals appears to be devoid of glial cells for at least the first month after surgery. As a corollary, regenerating axons that grow across the lesion may do so over an aglial substrate. Once the axons cross the lesion, their growth is confined to the astroglial domains of the proximal nerve stump. In the optic nerve, gliosis comprises hypertrophy of astrocytic processes such that the open framework characterizing the normal nerve is obscured. In addition, during regeneration, optic nerve glia express large amounts of the 50 kDa cytoskeletal protein, which they ordinarily express at only minimal levels. In the optic tract, gliosis is reflected in a markedly increased expression of the 50 kDa protein as well as an apparent increase in the number and complexity of glial processes. In addition, optic tract glia begin to express the 48 kDa antigen during regeneration. This protein is ordinarily confined for the most part to the optic nerve and is not seen in the tract glia. Finally, no obvious changes were seen in the glia of the optic tectum. These results demonstrate many points of similarity between gliosis in the goldfish and in mammals. However, in some particulars the two responses differ, and it is possible that these differences are related to the differing ability of central axons to regenerate in the two groups of organisms. Key words: astrocytes, cytoskeleton, antibodies, optic nerve, optic tract

Reactive astrogliosis refers t o the changes that astrocytes undergo in response to central nervous system (CNS) trauma. These include proliferation, hypertrophy of processes, and increased expression of glial fibrillary acidic protein (GFAP) (Eng, '88). To date, reactive gliosis has been studied almost exclusively in the mammalian brain, a system where damage leads to neuronal death and glial scarring (Reier et al., '83). However, in the anamniote vertebrates, CNS damage is often followed by axonal regeneration and functional recovery (urodeles-Levine and Cronly-Dillon, '74; anurans-Levine and Jacobson, '74; teleosts-Lowenger and Levine, '88) so that it would be of great interest to compare the glial response in these animals with that seen in the mammals. However, there have been only a few investigations of this problem to date. A number of workers have examined gliosis in the anamniotes using ultrastructural techniques and have demonstrated gliotic changes in the optic nerve of newts (Turner and Singer, '74; Stensaas and Feringa, '77), frogs (Reier and Webster, '74; Reier, "79; Bohn et al., '821, and fish (Wolburg, '81b; Giulian et al., '85). Other workers have

o 1991 WILEY-LISS. INC.

used immunohistochemical methods and antibodies against nonfish GFAP to examine the response of the fish CNS to a variety of lesions (Bignami et al., '74; Bignami and Dahl, '76; Anderson et al., '84). The results of these studies have been varied, with an increase in GFAP immunoreactivity being associated with spinal cord lesions but not those of the optic nerve. More recently, workers have generated antibodies to goldfish cytoskeletal antigens and shown them to be an excellent tool for describing the normal distribution of astrocytes in the visual system (Levine, '89; Nona et al., '89) as well as the response of these cells to wounding (Stafford et al., '90). In the present study a panel of these probes has been applied to the damaged CNS in order to examine astrogliosis and its relationship to axonal regeneration. The results indicate that gliosis in the goldfish and mammal are similar in many ways, i.e., there is an increased expression of GFAP Accepted July 2, 1991. Address reprint requests to Dr. R.L. Levine, Dept. of Biology, McGill University, 1205 Dr. Penfield Ave., Montrkal, Quebec, H3A 1B1, Canada

R.L. LEVINE

550

and an apparent increase in the amount of astrocytic processes in the tissue. However, reactive astrocytes in the goldfish also express unique cytoskeletal proteins, a response that is only minimally evident in mammalian reactive cells (Pixley and DeVellis, '84; Schiffer et al., ' 8 6 ) . Finally, unlike in mammals, reactive changes in goldfish astrocytes are viewed against a backdrop of axonal regeneration, and we may ask whether any of the noted reactive changes are a response to the regenerating axons, or conversely, whether any of these changes are necessary if regeneration is to succeed.

MATERIALS AND METHODS The experimental animals were comet goldfish (ABCee, Montreal), 6-7 cm from snout to base of tail. Animals were anaesthetized in 0.1% Tricaine Methane Sulphonate (MS222-Sigma) in phosphate buffer at pH 7.4. The right optic nerve was crushed using watchmaker forceps with 45" angle tips. The nerve was crushed twice for a count of 10, which produced a readily visible discontinuity in structure. Operated animals were kept in 24-gallon tanks at 20-22°C and allowed to survive for 1, 2, 4, 8, and 12 weeks after surgery. At each time point, between 5 and 10 animals were examined.

Histology Animals to be processed for histology were anaesthetized in MS222. Their unfixed brains and optic nerves were removed and embedded in 10% gelatin. The gelatin blocks were then rapidly frozen on dry ice and sectioned in a cryostat at - 12°C. Sections were mounted on gelatincoated (subbed) slides and allowed to air dry at room temperature for approximately 1hour. Slides that were not to be used immediately were stored at - 70 or -20°C. Processing of the mounted sections was initiated by fixing them first in absolute ethanol a t -20°C for 2 minutes and then in acetone, also at - 20"C, for 30 seconds. Sections were then blocked in 2% goat serum for 1hour, treated with primary antiserum (1:100) for 2 hours, washed, and finally processed with an FITC secondary antibody (1:100-Zymed, Dimension Labs, Canada) for another 2 hours. All reactions were carried out at room temperature. The sections were mounted in 90% glvcerol. DH 9, and examined with a Leitz Ultraphot micros&pe using fluorescence optics. Details of this immunohistochemical procedure may be found in Levine ('89). I

_

I

Antisera The polyclonal antisera used in this study were raised against antigens eluted from polyacrylamide protein gels. The protocols used have been described in detail elsewhere (Levine, '89) and are mentioned only briefly here. Cytoskeleta1 proteins were extracted using a protocol modified from Chiu and Norton ('82). These were solubilized in SDS sample buffer and run on 10%polyacrylamide gels. Proteins were visualized in the gels by SDS precipitation using 0.3 M CuC1, (Lee et al., '87). The bands of interest were at Mr 48,000 (48 Kd), an optic nerve specific glial cytoskeletal protein (Quitschke et al., '85; Levine, '89), which may be a keratin (Giordano et al., '89); 50 Kd, the brain glial cytoskeletal protein, which is probably homologous to GFAP (Quitschke et al., '85; Maggs and Scholes, '86; Levine, '89; Nona et al., '89); and 80 Kd, a neurofilament subunit protein (Autilio-Gambetti et al., '81; McQuarrie and Lasek, '81). These were visualized, cut out, and eluted

overnight into 0.1% SDS in 0.1 M phosphate buffer, pH 7.4, containing 0.75% sodium chloride (PBS). The SDS was washed out and the samples were reconstituted into PBS and used to immunize mice. It is important to note that rabbits were not used as a source of antisera because the advantages they afford in the quantity of serum available was more than outweighed by the observation that rabbit sera appear to have a strong, endogenous immune cross-reactivity with goldfish glial cytoskeletal proteins of Mrs 48 kDa and 58 kDa. This was true of both "preimmune" sera and antisera.

RESULTS Anti48 and anti80 sera were used to follow changes in the optic nerve glia and axonal regeneration concurrently. Two major results of crushing the optic nerve were immediately apparent. First, all optic fibers were severed by the surgery. This was seen in animals examined at 1 week or less postoperative, where anti80 serum demonstrated that intact axons extended to the crush and no farther (Figs. la, 2). Along the rest of the visual pathway, there was an immunolabeled granular material (which presumably represented axonal debris), as well as thick, degenerating axons (Figs. 2, 3). (The persistence of degenerating axons for several days following optic axon section in the goldfish has been documented by Levine and Dethier, '88.) The density and deployment of the debris varied along the pathway. Within the crush itself there was sparse debris, which, in some animals, was organized into longitudinal tracks that were continuous with the astroglial domains in the proximal nerve stump (Figs. l a , 2). In this latter location, debris was extremely dense and essentially confined to the astroglial domains (Figs. la, 3). The second major result of crushing the optic nerve was the virtually complete loss of glial antigens from the lesion site within 1 week after surgery (Fig. lb). This situation persisted for several weeks before repair was effected. Between 1 and 2 weeks postoperative, optic fibers first entered (Fig. l a ) and then grew across the lesion in profusion (Fig. 4a). At the same time, anti48 immunoreactivity was represented only at the distal and proximal edges of the lesion (Figs. lb, 4b). From 4 weeks postoperative onward, anti48 immunoreactive structures extended across the lesion (Figs. 5b, 6b), and it was apparent from adjacent sections that they were associated with fascicles of optic fibers that had already traversed the lesion (Fig. 5 ) . Once across the lesion, optic fibers were virtually confined within the astrocytic domains of the proximal nerve stump (Figs. 4-7). At this point the regenerating axons grew in thin fascicles, which frequently lay toward the center of the domains, away from their peripheral surface (Fig. 7). The growth of the regenerating axons in fascicles continued into the optic tract (Fig. 8). Optic fibers which grow through the lesion confront a rapidly changing landscape as the astroglia of the undamaged segment of the nerve undergo a variety of transformations. Under ordinary circumstances these cells form a myriad of parallel channels which contain the optic fibers: periodically, along their length, these channels are interrupted by latticework partitions set orthogonal to the long axis of the nerve (Fig. 9). Within the first postoperative week, these astroglial channels swelled and were at the same time filled by an apparent proliferation of fine processes throughout their lumen. By the following week, all

Fig. 1. Adjacent longitudinal sections of optic nerve at 1 week postoperative. The eye is toward the reader’s left. (a)Anti80 serum. Optic fibers may be seen growing into the lesion from the distal nerve segment (long stem arrows). The lesion itself is only weakly labeled (arrowhead); the proximal nerve segment contains densely labeled debris, which is confined to the astrocytic domains (short stem arrows). (b)Anti48 serum. Labeling is seen only in the distal and proximal nerve segments; the lesion is essentially devoid of labeled structures. The large arrows indicate the level to which the axons have penetrated the lesion (cf. a). The astrocytic domains are apparent in the proximal segment (small arrows). Bar: 300 Km. Fig. 2. Same animal as Figure 1 at 1 week postoperative, anti80 serum. Note that the astrocytic domains, defined by the dense debris in the proximal nerve segment (arrows), appear to continue into the lesion (arrowheads). Invading optic fibers are indicated by the open arrows. Bar: 100 pm. Fig. 3. Longitudinal section of optic nerve at 1week postoperative, anti80 serum. Densely distributed debris is evident in two astrocytic

domains (large arrows) of the proximal nerve segment. Note that degenerating axons are seen as well (small arrows). Bar:65 Km. Fig. 4. Adjacent longitudinal sections of optic nerve at 2 weeks postoperative. Both this figure and Figure 5 include the intraocular nerve segment as well as a portion of the retina (to the reader’s left). This stands in contrast to Figure 1where the nerve was removed for histology by cutting it immediately behind the eye. The approximate level of that incision is indicated in Figure 4 by open arrows. (a)Anti80 serum: Optic axons in retina and distal nerve segments are densely packed and parallel to one another. The fibers have grown through the lesion where they form a disorganized meshwork (large arrow). In the proximal stump, the regenerated axons are confined to the astrocytic domains (medium arrows). The small arrows indicate the boundaries of the lesion defined by the anti48 serum in (b). (b) Anti48 serum: Only the proximal and distal segments bind antibody while the lesion is essentially free of immunoreactivity (large arrow). The medium arrows indicate the astrocytic domains. Bar: 300 pm.

552

Fig. 5. Adjacent longitudinal sections of optic nerve at 4 weeks postoperative. (a) Anti80 serum: Optic axons have crossed the lesion and completely occupied the astrocytic domains of the proximal nerve segment (medium arrows). The small short stem arrows indicate the boundaries of the lesion as determined with anti48 serum. By reference to (b) below, it can be seen that only the axon fascicles indicated by small long stem arrows are invested by glial cells. (b)Anti48 serum: The lesion is only traversed by glial cells in a restricted area (long stem small arrows). The astrocytic domains of the proximal stump are evident (medium arrows). Bar: 300 &m. Fig. 6 . Adjacent longitudinal sections of optic nerve at 12 weeks postoperative. The area of the nerve containing the lesion site is shown.

R.L. LEVINE

The eye is towards the reader’s left. (a)Anti80 serum: Note the complicated fascicular patterns of the regenerated axons as they cross the lesion. (b) Anti48 serum: The complex pattern of astrocytic domains that cross the lesion replicates the fascicular patterns seen in (a).Bar: 160 km. Fig. 7. Adjacent cross sections of optic nerve, several hundred microns cranial to the lesion, 4 weeks postoperative. Large arrows indicate the same domains in (a) and (b). (a)Anti80 serum: Note that optic axon fascicles (small arrows) are distributed across the width of the domains. (b)Anti48 serum. Bar: 65 km.

GOLDFISH OPTIC PATHWAY GLIOSIS

Fig. 8. Longitudinal section of optic tract at 4 weeks postoperative, anti80 serum. Regenerating axons tend to grow in bundles (arrows) rather than diffusely. Bar: 100 pm. Fig. 9. Longitudinal section of normal optic nerve, anti48 serum. Two astrocytic domains are shown. Note the spaces that tend to be oriented orthogonal to the long axis of the nerve (small arrows). Also, note that there is very little space between the two domains (arrow). Bar: 65 pm.

Fig. Longitudinal section of an experimental nerve at weeks postoperative, anti48 serum. several astrocyticdomains (arrows) are shown. Note the loss of the orthoeonal maces which amear to have been filled in by fine processes andpartitiins. Note, too,*iheincreased interdomain connective tissue spaces (id). Bar: 65 pm.

evidence of the orthogonal arrangement which characterizes the normal nerve had been lost (Fig. 10). In addition, by this time the connective tissue spaces of the nerve were swollen such that the cross-sectional area of the nerve was increased and the usual close apposition of the astrocytic domains was no longer seen (Figs. 4, 5). This situation persisted for the duration of this study. By 4 weeks after crushing, the astrocytic channels in the vicinity of the

553

Fig. 11. Longitudinal section of an experimental nerve at 4 weeks postoperative, anti48 serum. (a) Several astrocytic domains from the vicinity of the lesion in the proximal nerve segment are shown. Note the nearly complete loss of open spaces in the domains, apparently due to an increased density of astrocytic processes. (b)Astrocytic domains from the vicinity of the optic foramen. Note that the open meshwork characteristic of earlier postoperative times is still present. Bars: 65 pm.

Fie. 12. LonPitudinal section of an exDerimental nerve at 12 weeks postoperative, anti48 serum. Astrocytic iornains from the vicinity of the optic foramen. Note that the open meshwork of astrocyte processes has been Obliterated. Bar: 65 pm.

lesion had become filled with thick processes and no longer appear patent (Fig. l l a ) . At the same time, as one moved along the nerve toward the optic foramen, these alterations in the glial fabric diminished and immediately adjacent to the foramen, the astrocytes were arranged as they had been 2 weeks earlier (Fig. llb). However, within a month, virtually the entire intraorbital nerve had undergone this metamorphosis and patent astrocytic channels could no

R.L. LEVINE

554

longer be found anywhere along its length (Fig. 12). It should be underscored at this point that in all animals examined at 2 weeks or more after surgery, the entire intraorbital nerve had been occupied by regenerated optic fibers (cf. Fig. 4a), that is, all of the glial changes described above transpired in the presence of growing axons. The patterns of expression of the second glial antigen, 50kDa, also change during regeneration. In the intact visual system, 50 kDa immunoreactivlty is most pronounced in the retina (in Muller cells-Bignami, '84; Nona et al., '89) and in the brain (in radial glia-Levine, '89; Nona et al., '89). However, there is expression of this immunoreactivity in the optic nerve as well, although it has not been reported by previous workers (Maggs and Scholes, '86; Levine, '89; Stafford et al., '90) and so it is described here in some detail. Anti50 immunoreactivity in the normal nerve is distributed in a discontinuous fashion, which varies from one nerve fascicle to the next. Thus one may find a fascicle with moderately labeled glia next to one that is essentially unlabeled (Fig. 13). Where glial cells are labeled they present threadlike profiles which tend to be arrayed in a disorganized fashion within the fascicle or along its surface (Fig. 13). Thus the pattern of binding of the anti50 serum in the intact optic nerve is quite different from that seen with the anti48 serum (cf. Fig. 9). Within a week after a crush lesion, anti50 immunoreactivity was strongly enhanced and uniformly distributed throughout the nerve (Fig. 14). As in the intact nerve, the structures labeled with the anti50 serum were threadlike and extended throughout the astrocytic domains (identified with the anti-48 serum) to form a complicated meshwork (Figs. 14, 15). However, within this meshwork some order could be discerned. First, at postoperative times up to 4 weeks there was a clear bias for the processes to be arrayed orthogonal to the long axis of the nerve (Fig. 15). This generalization did not appear to hold at later times when the labeled processes were more disorganized and had a tendency to lie parallel to the long axis of the nerve (see Fig. 17). At the crush site, anti50 immunoreactive processes extended into the lesion from both sides as early as 1week after surgery and completely across it by the fourth postoperative week (see Fig. 17). Comparison with adjacent sections demonstrated that, in all instances, these processes were confined to astrocytic domains as defined with the anti48 serum. In spite of its early appearance in the lesion, anti50 immunoreactivity was less intense in this location than in the undamaged nerve segments on either side of the lesion up to at least 12 weeks after surgery (see Fig. 17). In addition to the obvious increase in anti50 immunoreactivity in the operated nerve, there appeared to be an increase in this response on the unoperated side as well. Because the response was not robust and resulted in only a slight enhancement of immunoreactivity, it was difficult to characterize it effectively. However, at early postoperative times, the labeled processes showed a bias to a transverse orientation in the nerve (Fig. 16). As in the normal nerve in an intact animal, the labeling was patchy and showed no overall pattern of distribution in the nerve. By 12 weeks after surgery, the response of the unoperated nerve had dropped t o background levels. The optic nerve astrocyte compartment (as defined with anti48 serum) normally extends beyond the optic chiasm to the point where the optic tracts join the diencephalon (Levine, '89). Immediately dorsal and caudal to this level, anti48 immunoreactive radial glia arise from the preoptic

area and sweep laterally across the optic fibers, to the pial surface of the optic tract. Beyond this point, in intact animals no further anti48 positive glial cells are seen. However, within 1 week following an optic nerve crush, anti48 immunoreactive filiform profiles appeared in the affected optic tract (Fig. 18).These processes tended to run in the direction of the optic fibers, and they became progressively more populous with time, persisting for at least 12 weeks after surgery (Fig. 19). At the earliest times investigated, the processes occupied the portion of the optic tract adjacent to the diencephalon (Fig. 18),whereas later they spread throughout its width (Fig. 19b). The origin of these processes was not clear, although similar processes which are anti50 immunoreactive appear to arise from the ventricular lining of the preoptic recess (Levine, '89). In addition, operated animals would occasionally show one or two short filiform processes in the intact optic tract when labeled with anti48 serum (Fig. 19a). The glial processes which may be visualized in the intact optic tract are exclusively anti50 positive radial glial cells which arise in the preoptic area and extend processes into the tract (Levine, '89; Nona et al., '89). Following optic nerve crush, the number and complexity of these radial glial processes increased dramatically. The first clear evidence for this effect was seen 2 weeks after surgery (Fig. 20). By 4 weeks the increase in labeled glial processes was maximal, and this state persisted unabated for at least 12 weeks postlesion (Fig. 21). Finally, no obvious changes in the optic tectum of experimental animals were seen using antiglial sera. In some instances, the thickened ependymal layer of the tectum showed an enhanced reactivity with the anti48 serum (with which it usually does not crossreact), but this could be seen on both sides of the brain and may, on occasion, be seen in intact animals as well. By contrast with the stability of the glial markers, optic fibers in the tectum could be clearly followed, using the anti80 serum, as they degenerated and were replaced by regenerating optic axons.

DISCUSSION Regeneration of optic fibers in the goldfish has been examined using immunohistochemical techniques with antisera against glial and neuronal cytoskeletal antigens. The observations indicate that glial cells of the visual paths undergo major changes in their structure and in their molecular constituents during the process of regeneration and that these changes persist well beyond the period usually associated with the completion of regeneration. Among the more pertinent observations made are: (1) following a crush lesion to the optic nerve, regenerating axons initially grow across a relatively large stretch of nerve that appears to be devoid of glial cells, (2) astrocytes in both the optic nerve and the optic tract show a hypertrophic response that persists for at least 3 months after surgery, and (3) during regeneration, both optic nerve and optic tract astrocytes begin to express cytoskeletal antigens, which they do not express in intact animals. These points are discussed in detail below. During the first several weeks after a crush lesion of the optic nerve, little or no immunoreactivity for either the 48 kDa or 50 W a glial cytokeletal antigen is seen in the damaged region. A similar observation has been made by Stafford et al. ('90) who used an antiserum against the 50 kDa antigen. These results may either indicate that there

GOLDFISH OPTIC PATHWAY GLIOSIS

Fig. 13. Longitudinal section of a normal nerve from an intact animal, anti50 serum. Note the weak labeling of filamentous profiles (arrows). Bar: 65 pm. Fig. 14. Longitudinal section of an operated nerve at 2' weeks postoperative, anti50 serum. Substantial increase in anti50 immunoreactivity has occurred. Many processes are oriented orthogonal to the long axis of the nerve (arrows). Bar: 65 pm. Fig. 15. Longitudinal section of an operated nerve at 12 weeks postoperative, anti50 serum. Labeled processes are no longer biased to the orthogonal arrangement seen earlier. Many labeled processes are now deployed along the long axis of the nerve (arrows).Bar: 65 Fm.

are no astrocytes in the lesion at this time or that they are present but not expressing gliofilamentantigens. To resolve this dilemma effectively,further immunohistochemical and ultrastructural studies will be necessary. However, these are reasons for thinking that the absence of immunoreactivity does indeed reflect an absence of glial cells in the lesion. First, in the proximal segment of the nerve, glial cells label robustly with both the anti48 and the anti50 sera. Thus being in the presence of degenerating and regenerating axons does not cause glial cells t o lose their immunoreactivity. Second, in one animal examined at 2 weeks postopera-

555

Fig. 16. Longitudinal section of the unoperated nerve from an experiment animal at 2 weeks postoperative, anti50 serum. A moderate increase in immunoreactivity is seen. The filamentous profiles are strongly biased to orthogonal orientation (arrows); compare with Figure 8. Bar: 65 pm. Fig. 17. Longitudinal section of an operated nerve at 4 weeks postoperative, anti50 serum. Two astrocytic domains are shown (large arrows) crossing the lesion (between the small arrows). Note that although anti50 labeling is present in the lesion, it is diminished compared to what is seen in the proximal and distal undamaged nerve segments. Bar: 125 pm.

tive, a single intact glial domain was seen crossing an otherwise unlabeled lesion. This was presumably the result of an incomplete nerve crush and demonstrates that astroglia can express their cytoskeletal proteins within a lesion. Third, it seems likely that a complete crush lesion kills most or all cells at the lesion site. Thus any astrocytes in the lesion at early postoperative times would have to migrate in from the margins of the wound, and no evidence for such migration was seen in the present study (although the cells might lose their immunoreactivity before beginning to migrate). Finally, the general response of astrocytes to

Fig. 18. Longitudinal section of optic tract at 2 weeks postoperative, anti48 serum. The midline is to the left. The large arrowheads indicate the boundary between the diencephalon (d) and the tract (ot). Note the presence of filiform profiles (arrows) adjacent to the diencephalon (d), whereas in the more lateral part of the tract, very few such processes are seen. The large profiles are blood vessels (small arrows), which are typically labeled with anti48 serum (see Levine, '89). Bar: 100 pm. Fig. 19. Longitudinal section of the optic tracts a t 12 weeks postoperative, anti48 serum. The midline lies between the two figures. Large arrowheads as in Figure 18. (a)Unoperated side of the brain. Labeled processes at the bottom of the figure (short stem arrows) arise from the preoptic area and are present in the normal animal. Two small threadlike profiles are present farther dorsally in the tract (long stem

arrows). No other labeled processes are seen. (b) Operated side. Note that the entire tract is filled with fine processes which run parallel to its longaxis. Bar: 100 pm. Fig. 20. Longitudinal sections of the optic tracts at 2 weeks postoperative, anti50 serum. Details as in Figure 19. (a)Unoperated side. (b) 0perat.ed side. Note the increased population of fine processes compared to the unoperated tract. Bar: 65 pm. Fig. 21. Longitudinal sections of the optic tracts at 12 weeks postoperative, anti50 serum. Details as in Figure 19. (a)Unoperated side. (b) Operated side. There is a substantial increase in labeled profiles throughout the operated tract. Bar: 100 km.

GOLDFISH OPTIC PATHWAY GLIOSIS trauma is an upregulation of gliofilament expression, not its cessation (Eng, '88). In view of these considerations, it seems likely that the absence of immunoreactivity for the 48 and 50 kDa antigens in a crush lesion of the optic nerve reflects a true absence of astrocytes from the lesion. In spite of the apparent early absence of glial cells, regenerating optic axons grow through the lesion in profusion. By 4 weeks postoperative, immunoreactive glial cells may be seen in the damaged area, and in this situation they are always associated with the axon fascicles that preceded them. Based on these observations, one may hypothesize that following optic nerve crush, regenerating optic axons in the goldfish grow across a nonglial substrate in order to gain access to the proximal nerve segment. Only later in regeneration do glial cells migrate out along the axons to invest them and reinstate continuity of the glial domains across the lesion. This hypothesis is supported by the observations of Stensaas and Feringa ('77) who showed that optic axons were able to regenerate across a lesion produced by a freezing crush of the optic nerve in the newt, T. pyrrhogaster, although it was at first acellular and then populated exclusively by macrophages. Astrocytes in these lesions were only represented by fine processes which extended from the cells at the margins of the wound, and at no time did they appear to be acting as substrates for the growing axons (which grew along the basal lamina of the glia limitans). Once across the lesion, the axons grew among the astrocytes of the cranial nerve stump within which they were confined, as was the case in the present study. The suggestion that the first axons to traverse the lesion are growing over a nonglial substrate is somewhat heterodox since many workers have shown that astrocytes constitute an excellent substrate for central axon growth, both in vitro and in vivo (Noble et al., '84; Fallon, '85; Smith et al., '86; Kliot et al., '88). Indeed, we have shown in the present work that once axons have crossed the aglial lesion, they appeared to be constrained to grow within the astrocytic domains of the proximal nerve stump. Perhaps goldfish optic axons express a hierarchy of substrate preferences, with astrocytesbeing a preferred substrate, whereas nonglial substrates are less suitable but still acceptable for growth. Similar conclusions were reached by Bohn et al. ('82) who examined regeneration following optic nerve transection in Xenopus and found that axons grew randomly in the orbital connective tissue but rapidly became more organized and grew directly toward the brain when they entered the proximal nerve stump. If glial cells do not form the substrate over which the regenerating axons initially grow, what might the axons use for this purpose? One observation that is pertinent to this question is that in some animals, at early postoperative times, the lesion appeared to be traversed by remnants of the astrocytic domains. These could be visualized because they constrained axonal debris in channels that were continuous with the intact domains of the proximal nerve segment. In addition, in these cases, immunohistochemistry with antilaminin serum has shown basal lamina-like profiles around the putative domain remnants in the lesion (Levine-unpublishedobservations). Since laminin has been implicated as a growth substrate in the goldfish visual system (Hopkins et al., '85; Liesi, '85) and is a major constituent of the basal lamina (Timpl et al., '79), it is possible that where domain remnants are present, their basal lamina serves as a growth substrate for early regener-

557 ating axons. Further support for this contention comes from the work of Stensaas and Feringa ('77) who showed that following a freezing lesion to the optic nerve of the newt, the earliest regenerating optic axons that traversed the wound grew along the persisting basal lamina of the glia limitans. However, in many animals, axons that grew across the lesion formed a chaotic tangle (cf. Fig. 4a), indicating that effective guidance was not always offered to the growing fibers. In fact, it may be that the unavoidable variability of lesions produced by crushing the nerve generates a range of results varying from lesions which totally spare domain basal lamina and thus offer extensive guidance for growing axons to those which completely destroy the lamina and so offer none. There are also arguments that militate directly against the possibility that basal laminae are used for growth and guidance by regenerating axons in this system. First, we have shown in the present work that optic axons invade the glial domains of the proximal segment as fascicleswhich are as likely to lie deep in the domain as they are to lie at the surface in proximity to the basal lamina. Several other workers (Wolburg, '81a; Murray, '82; Easter, '87) have also reported that regenerating optic axons tend to fasciculate with one another and become associated with astrocytes rather than growing along the basal lamina of the glial domains. Second, optic axons regenerate perfectly well in the optic tract where basal laminae are only found at the glia limitans of blood vessels and the surface of the brain (Peters et al., '76). Since the axons show no predeliction to grow in these locations, they clearly do not have an obligatory dependence on basal laminae as growth substrates. These considerations suggest that even if basal laminae are involved in mediating or guiding the growth of regenerating axons in this system, other factors must be involved as well. Gliosis in the optic nerve transforms the open astrocytic strutwork seen in the intact nerve into a densely packed meshwork of coarse astrocytic processes. This transformation is likely to have functional significance at at least two stages of regeneration. First, when the regenerating axons first invade the proximal nerve stump, they encounter a welter of astrocytic processes in the first phase of the gliotic reaction. It is known from other studies that astrocytes can form a growth-promoting substrate in embryonic or juvenile mammals (Noble et al., '84; Fallon, '85; Smith et al., '86; Kliot et al., '88) and it may be that they perform the same role in the goldfish optic nerve. Certainly, it is clear that once the regenerating axons reach the glial domains of the proximal nerve stump, they are constrained to grow within them, suggesting that the most favorable growth substrate resides therein. Again, it is worth emphasizing that in this situation, the axons do not appear to be following the glial domain basal lamina. The second phase of the gliotic reaction, which encompasses a thickening of the astrocytic processes with consequent obliteration of the open spaces seen in nerve sections, occurs during the period starting at 4 weeks postoperative, when the optic axons and their terminals are maturing (Schmidt et al., '83; Northmore and Masino, '84; Rankin and Cook, '86; Matsumoto et al., '87). This is also the time when myelination begins (Wolburg, '81b) and it is of interest that Maggs and Scholes ('90) have recently shown that astrocytes in the intact optic nerve of cichlid fish appear to have a unique spatial association with the nodes of Ranvier of the optic axons. That is, they find that astrocytes form transverse partitions

558

R.L. LEVINE

in the nerve (which are clearly the same as the latticework kDa antigen. The possible association of these cells with plates we identified in the goldfish-Levine, '891, which lie latticework plates, and thus nodes of Ranvier (Maggs and in register with the nodes. In addition, electron microscopic Scholes, 'go), suggests that they may form a functional analysis shows that nodes of Ranvier are usually en- subclass of astrocytes which interact in specific ways with wrapped in astrocytic processes. These observations sug- myelinated axons. The consequences of the gliotic changes gest that the astrocytic partitions described by Maggs and they undergo might then be reflected in the process of Scholes ('90) are somehow involved in maintaining the remyelination and the functioning of the reniyelinated structure or the proper functioning of myelinated axons in axons. A change in cytoskeletal protein expression also occurs in the fish optic nerve. The complete disruption of these structures, seen during the late stages of optic fiber regener- the optic tract during regeneration, but it is the opposite of ation in the goldfish at just the time when myelination is that seen in the nerve, i.e., it is the 48 kDa antigen, which is occurring and nodes of Ranvier are being formed (Wolburg, switched on over a background of the 50 kDa antigen. It is %la), is therefore likely to be reflected in functional not clear whether the cells expressing the new antigen are consequences for the visual system as a whole. This would coexpressing it with GFAP or whether there is a population mean that 3 months after a crush lesion, at a time when of optic tract glia which express only the 48 kDa antigen and regeneration is widely considered to be completed, the optic do so only in response to a lesion. There are, however, nerve would be both structurally and functionally compro- examples in the literature of glial cells coexpressing a new mised. Whether this state ever reverts to something ap- cytoskeletal antigen-vimentin-as part of their gliotic proaching normalcy is a question that can be answered only response (Pixley and DeVellis, '84; Schiffer et al., '86). In by the examination of animals at longer postoperative addition, in both the developing mammalian brain (Dahl, '81; Dahl et al., '81; Schnitzer et al., '81; Bignami et al., '82) times. The hypertrophic response of astrocytes in the optic tract and in certain locations in the adult brain (Schnitzer et al., differs in many ways from that seen in the nerve, since the '81; Shaw et al., '81) vimentin is ordinarily coexpressed tract glia are filiform cells which express a 50 kDa cytoskel- with GFAP in astrocytes. However, in the present instance, eta1 protein (probably the teleost equivalent of GFAPit will be necessary to use double-labeling protocols on Nona et al., '89). In intact animals, it is difficult to visualize single sections to determine whether or not coexpression is these cells (Levine, '89). However, starting at 2 weeks occurring. postoperative, increasing numbers of anti50 positive threadThe reactive astrocytic processes in the optic tract are like processes of increasing complexity appear in the tract. aligned parallel to the paths of growth followed by regenerThese changes mirror those seen in mammals, where ating optic axons. This suggests that they could be involved protoplasmic astrocytes or Muller cells in the resting state in guiding the axons in much the same fashion that radial express minimal amounts of GFAP, which are substantially glia guide migrating immature neurons during primate increased in response to trauma (Bignami and Dahl, '74a,b, cortical development (Rakic, '85; Hatten, '90). The growth '79; Bjorkland et al., '85; Matthewson and Berry, '85). In of the invading axons in fascicles is also consistent with this this respect, gliosis in the goldfish brain is similar to that interpretation. More pertinent is the fact that the 48 kDa seen in the mammalian CNS. However, it must be remem- antigen in glial processes first appears in precisely the bered that the gliotic changes reported here are a response location where the first axons to reach the tract grow, i.e., to deneurotization, not to trauma, and that they transpire adjacent to the diencephalon, (Lowenger and Levine, '88). in the presence of actively regenerating axons, a fact that Thus the appearance of the 48 kDa antigen in optic tract raises the possibility that a t least some of the gliotic glia may be correlated with a change in these cells, which responses in this system are triggered by interactions of makes them preferred substrates for regenerating optic glial cells with the invading axons (see, e.g., Johns et al., fibers. This is a particularly interesting possibility in view of '77; Stevenson and Yoon, '78; Giulian et al., '85). the recent suggestion that the 48 kDa antigen may be a The final major observation in this study is that both type I1 keratin of the sort usually associated with developoptic nerve and optic tract glia express unusual cytoskeletal ing systems (Giordano et al., '89). Thus the tract glia may proteins during optic fiber regeneration. In the nerve, the revert to a more embryonic state during regeneration and 50 kDa antigen, which is expressed at low basal levels in the so reprise an earlier role as substrates for axonal growth intact nerve, becomes strongly upregulated during regener- and guidance (see, e.g., Silver, '84; Silver and Rutishauser, ation. A similar observation has been made by Stafford et al. '84). ('90) who did not, however, detect anti50 immunoreactivity In summary, we have described the astrocytic response to in intact nerves. Stafford et al. ('90) also reported that the a crush lesion of the optic nerve in the goldfish. In the lesion increased anti50 immunoreactivity in the nerve had com- itself, astrocytes are not apparent and axons grow across pletely abated by approximately 20 weeks postoperative. what appears to be an aglial substrate. In both the cranial The possible significance of this increased expression of the segment of the optic nerve and in the optic tract, there is 50 kDa antigen during regeneration has been discussed at hypertrophy of astrocytic processes, which continues to length by Stafford et al. ('90). To their comments, the develop for at least a month after lesioning. In addition, in following points may be added. In undamaged nerves both locations unique glial cytoskeletal antigens are ex(either in intact or operated animals), there is a bias for the pressed during the course of regeneration. All of the anti50 immunoreactive processes to lie orthogonal to the changes described persisted for at least 3 months after long axis of the nerve. In fact, comparison of Figures 9 and surgery. However, there is a period of tectal synaptic 16 in this work indicates that these processes may be reorganization and optic fiber loss which continues for up to associated with the latticework plates revealed with anti48 36 weeks after surgery in this system (Murray, '82; Hayes serum. This and the paucity of these processes compared to and Meyer, '89), and it will be of interest to learn whether anti48-positive structures indicate that only a subpopula- the gliosis described here is eventually resolved in parallel tion of the optic nerve glia express and upregulate the 50 with these processes or whether it persists indefinitely. For

GOLDFISH OPTIC PATHWAY GLIOSIS the future it will be essential to examine certain aspects of this system in more detail. In particular, the nature of the cells in the early lesion is enigmatic but of great interest since it is across this region that the regenerating s o n s must grow to gain access to the brain. Both ultrastructural studies and further immunohistochemical investigations using antibodies against cells other than astrocytes will serve to clarify this issue. In addition, an ultrastructural examination of regenerating axons in the optic tract will be a first step toward determining whether radial glia may serve a role in axon guidance in this system. The answers to these questions will enhance our understanding of central axon regeneration in anamniotes and may, by extension, elucidate some of the factors responsible for the failure of central regeneration in mammals.

ACKNOWLEDGMENTS Garry Kessler's able assistance in this work is gratefully acknowledged. Guy L'Heureux is responsible for the expertise evident in the figures. This work was supported by NSERC grant OGPIN 001 and a grant from the Department of Biology at McGill.

LITERATURE CITED Anderson, M.J., K.A. Swanson, S.G. Waxman, and L.F. Eng (1984) Glial fibrillary acidic protein in regenerating teleost spinal cord. J. Histochem. Cytochem. 31r1099-1106. Autilio-Gambetti, U., M.E. Velasco, J. Sipple, and P. Gambetti (1981) Immunochemical characterization of antisera to rat neurofilament subunits. J. Neurochem. 37: 12260-1265. Bignami, A. (1984) Glial fibrillary acidic protein in Mueller glia: Immunofluorescence study of the goldfish retina. Brain Res. 3OOr175-178. Bignami, A,, and D. Dahl (1974a) Astrocyte-specific protein and neuroglial differentiation. An immunofluorescence study with antibodies to the glial fibrillary acidic protein. J. Comp. Neurol. 153:27-38. Bignami, A., and D. Dahl(1974b) Astrocyte-specific protein and radial glial in the cerebral cortex of newborn rat. Nature 253255-56, Bignami, A,, and D. Dahl (1976) The astroglial response to stabbing. Immunofluorescence studies with antibodies to astrocyte-specific protein (GFA) in mammalian and submammalian vertebrates. Neuropath. Appl. Neurobiol. 299-1 11. Bignami, A., and D. Dahl (1979) The radial glia of Muller in the rat retina and their response to injury. An immunofluorescence study with antibodies to the glial fibrillary acidic (GFA) protein. Exp. Eye Res. 28:63-69. Bignami, A., L. Forno, and D. Dahl(1974) The neuroglial response to injury following spinal cord transection in the goldfish. Exp. Neurol. 44r60-70. Bignami, A., T. Raju, and D. Dahl (1982) Localization of vimentin, the non-specificintermediate filament protein in embryonal glia and in early differentiating neurons. Dev. Biol. 91r286-295. Bjorkland, H., A. Bignami, and D. Dahl (1985) Immunohistochemical demonstration of glial fibrillary acidic protein in normal rat Muller glia and retinal astrocytes. Neurosci. Lett. 54:363-368. Bohn, R.C., P.J. Reier, and E.B. Sourbeer (1982) Axonal interactions with connective tissue and glial substrata during optic nerve regeneration in Xenopus larvae and adults. Amer. J. Anat. 165r397-419. Chiu, F.C., and W.T. Norton (1982) Bulk preparation of CNS cytoskeleton and the seuaration of individual neurofilament uroteins bv " eel - filtration: Dye-binding characteristics and amino acid compositions. J. Neurochem. 39,1252-1260. Dahl, D. (1981) The vimentin-GFA protein transition in rat neuroglia cytoskeleton occurs at the time of myelination. J. Neurosci. Res. 6.5'41-748. Dahl, D., C.J. Crosby, and A. Bignami (1981) Filament proteins in rat optic nerves undergoing Wallerian degeneration. Exp. Neurol. 71r421-430. Easter, S.S. (1987) Retinal axons and the basal lamina. I n J.R. Wolff, J. Sievers, and M. Berry (eds): Mesenchymal-Epithelial Interactions in Neural Development. Berlin: Springer-Verlag, pp. 385-396.

559 Eng, L. (1988) Astrocytic Response to Injury. In P. Reier, R. Bunge, and F. Seil (eds): Current Issues in Neuronal Regeneration Research. New York: Alan R. Liss, Inc., pp. 247-255. Fallon, J.R. (1985) Preferential outgrowth of central nervous system neurites on astrocytes and Schwann cells as compared with nonglial cells in vitro. J. Cell Biol. 100:198-207. Giordano, S., E. Glasgow, P. Tesser, and N. Schechter (1989) A type I1 keratin is expressed in glial cells of the goldfish visual pathway. Neuron 21507-1516. Giulian, D., V. Iwanij, and H. Stuckenbrok (1985) The response of optic tract glia during regeneration of the goldfish visual system. 1. Biosynthetic activity within different populations after transection of retinal ganglion cell axons. Brain Res. 339r87-96. Hatten, M.E. (1990) Riding the glial monorail A common mechanism for glial-guided neuronal migration in different regions of the developing mammalian brain. Trends Neurosci 13:179-184. Hayes, W.P., and R.L. Meyer (1989) Normal numbers of retinotectal synapses during the activity sensitive period of optic regeneration in goldfish: HRP-EM evidence implicating synapse rearrangement and collateral elimination during map refinement. J. Neurosci. 9:1400-1413. Hopkins, J.M., T.S. Ford-Holevinski, J.P. McCoy, and B.W. Agranoff (1985) Laminin and optic nerve regeneration in the goldfish. J. Neurosci. 5r3030-3038. Johns, P.R., S.S. Easter, J.A. Burnham, and M.K. Powers (1977) Regenerating optic fibers stimulate tectal cell proliferation in the goldfish. Anat. Rec. 187r615-616. Kliot, M., G.M. Smith, J. Siegal, S. Tyrell, and J. Silver (1988) Induced regeneration of dorsal root fibers into the adult mammalian spinal cord. I n P.J. Reier, R.P. Bunge, and F.J. Seil (eds): Current Issues in Neural Regeneration Research, Vol. 48. New York Alan R. Liss, Inc., pp. 311-328. Lee, C., A. Levin, and D. Branton (1987) Copper staining: A five-minute protein stain for sodium dodecyl sulfate-polyacrylamide gels. Anal. Biochem. 166:308. Levine, R.L. (1989) Organization of astrocytes in the visual pathways of the goldfish: An immunohistochemical study. J. Comp. Neurol. 285231245. Levine, R.L., and J.R. Cronly-Dillon (1974) Specification of regenerating retinal ganglion cells in the adult newt, Triturus cristutus. Brain Res. 68.3 19-329. Levine, R.L., and S. Dethier (1988) Anomalous retrograde labeling of brain cells from the eye i n goldfish: Evidence for long distance growth of sprouted neurites. Exp. Neurol. 102:153-166. Levine, R., and M. Jacobson (1974) Deployment of optic nerve fibers is determined by positional markers in the frog's tectum. Exp. Neurol. 43r527-538. Liesi, P. (1985) Laminin-immunoreactive glia distinguish regenerative adult CNS systems from non-regenerative ones. EMBO 4:2505-2511. Lowenger, E., and R.L. Levine (1988) Studies of the early stages of optic axon regeneration in the goldfish. J. Comp. Neurol. 27Ir319-330. Maggs, A., and J. Scholes (1986) Glial domains and nerve fiber patterns in the fish retinotectal pathway. J. Neurosci. 6:424-438. Maggs, A., and J. Scholes (1990) Reticular astrocytes in the fish optic nerve: Macroglia with epithelial characteristics form an axially repeated lacework pattern, to which nodes of Ranvier are apposed. J. Neurosci. 10:1600-1614. Matsumoto, N., M. Kometani, and K. Nagano (1987) Regenerating retinal fibers of the goldfish make temporary and unspecific but functional synapses before forming the final retinotopic projection. Neurosci. 22:1103-1110. Matthewson, A.J., and M. Berry (1985) Observations on the astrocyte response to a cerebral stab wound in adult rats. Brain Res. 327:61-69. McQuarrie, I.G., and Lasek, R.J. (1981) Axonal transport of labelled neurofilament proteins in goldfish optic axons. J. Cell Biol. 91:234a. Murray, M. (1982) A quantitative study of regenerative sprouting by optic axons in goldfish. 3. Comp. Neurol. 109:363-373. Noble, M., J. Fok-Seang, and J. Cohen (1984) Glia are a unique substrate for the in vitro growth of central nervous system neurons. J. Neurosci. 4:1892-1903. Nona, S.M., S.A.S. Shehab, C.A. Stafford, and J.R. Cronly-Dillon (1989)Glial fibrillary acidic protein (GFAP) from the goldfish Its localization in the visual pathways. GliaZ: 189-200. Northmore, D.P.M., and T. Masino (1984) Recovery of vision in fish after optic nerve crush: A behavioral and electophysiological study. Exp. Neurol. 84r109-125.

560 Peters, A,, S.L. Palay, and H. deF. Webster (1976) The Fine Structure of the Nervous System: The Neurons and Supporting Cells. Philadelphia: Saunders. Pixley, S.K.R., and J. DeVellis (1984) Transition between immature radial glia and mature astrocytes studied with a monoclonal antibody to vimentin. Dev. Brain Res. 15:201-209. Quitschke, W., P.S. Jones, and N. Schechter (1985) Survey of intermediate filament proteins in the optic nerve and spinal cord. J. Neurochem. 44:1465-1476. Rakic, P. (1985) Contact regulation of neuronal migration. I n G.M. Edelman and J.P. Thiery (eds): The Cell in Contact: Adhesions and Junctions as Morphogenetic Determinants. New York Neuroscience Research Foundation, pp. 67-81. Rankin, E.C.C., and J.E. Cook (1986) Topographic refinement of the regenerating retinotectal projection of the goldfish in standard laboratory conditions: A quantitative WGA-HRP study. Exp. Brain Res. 63:409-420. Reier, P.J. (1979) Penetration of grafted astrocytic scars by regenerating optic nerve axons inXenopus tadpoles. Brain Res. 164:61-68. Reier, P.J., and H.DeF. Webster (1974) Regeneration and remyelination of Xenopus tadpole optic nerve fibers following transection or crush. J. Neurocytol. 3:59 1-6 18. Reier, P.J., L.J. Stensaas, and L. Guth (1983) The astrocytic scar as an impediment to regeneration in the central nervous system. In C.C. Kao, R.P. Bunge, and P.J. Reier (eds): Spinal Cord Reconstruction. New York: Raven Press, pp. 163-195. Schiffer, D., M.T. Giordana, A. Migheli, G. Giaccone, S. Pezzotta, and A. Mauro (1986)Glial fibrillary acidic protein and vimentin in the experimental glial reaction to the rat brain. Brain Res. 374t110-118. Schmidt, J.T., D.L. Edwards, and C. Stuermer (1983) The re-establishment of synaptic transmission by regenerating optic axons in goldfish Time course and effects of blocking activity by intraocular injection of tetrodotoxin. Brain Res. 269t15-27. Schnitzer, J., W.W. Franke, and M. Schachner (1981) Immunocytochemical

R.L. LEVINE demonstration of vimentin in astrocytes and ependymal cells of developing and adult mouse nervous system. J. Cell Biol. 90:435447. Shaw, G., M. Osborn, and K. Weber (1981) An immunofluorescence microscopical study of the neurofilament triplet proteins, vimentin and glial fibrillary acidic protein within the adult rat brain. Eur. J. Cell Biol. 26:68-82. Silver, J. (1984) Studies on the factors that govern directionality of axon growth in the embryonic optic nerve and a t the chiasm of mice. J. Comp. Neurol. 223.238-251. Silver, J., and U. Rutishauser (1984) Guidance of optic axons in vivo by a preformed adhesive pathway on neuroepithelial endfeet. Develop. Biol. 106:485499. Smith, G.M., R.H. Miller, and J. Silver (1986) Changing role of forebrain astrocytes duringdevelopment, regenerative failure and induced regeneration upon transplantation. J. Comp. Neurol. 2 5 1 ~ 2 3 4 3 . Stafford, C.A., S.A.S. Shehab, S.N. Nona, and J.R. Cronly-Dillon (1990) Expression of glial fibrillary acidic protein (GFAP) in goldfish optic nerve following injury. Glia 3:33-42. Stensaas, L.J., and E.R. Feringa (1977) Axon regeneration across the site of injury in the optic nerve of newt Triturus pyrrhogaster. Cell Tiss. Res. 179:50 1-516. Stevenson, J.A., and M.G. Yoon (1978) Regeneration of optic nerve fibers enhances cell proliferation in the goldfish optic tectum. Brain Res. 153t345-351. Timpl, R., H. Rohde, P.G. Robey, S.I. Rennard, J.-M. Foidart, and G.R. Martin (1979) Laminin-A glycoprotein from basement membranes. J. Biol. Chem. 254:9933-9937. Turner, J.E., and M. Singer (1974)The ultrastructure of regeneration in the severed newt optic nerve. J. Exp. Zool. 190:249-268. Wolburg, H. (1981a) Myelination and remyelination in the regenerating visual system of the goldfish. Exp. Brain Res. 43t199-206. Wolburg, H. (1981b)Axonal transport, degeneration and regeneration in the visual system of the goldfish. Adv. Anat. Embryol. CellBiol. 67:l-94.