Patterns of cell division during visual streak formation in the frog Limnodynastes dorsalis

The distribution of cells in the retinal ganglion cell layer of Anura changes after metamorphosis. In 1981 Dunlop & Beazley reported for Heleioporus eyrei that the radial density gradient with high peripheral values characteristic of tadpoles evolved to a horizontally aligned high-density visual streak in adults. Similarly, Bousfield & Pessoa (1980) reported an accentuation of the area centralis in postmetamorphic Hyla raniceps. Although less dramatic changes were found in Xenopus laevis, densities in nasal and in particular temporal peripheries were found to exceed other regions only after metamorphosis (Dunlop & Beazley, 1984). In each species these changes in cell distribution were accompanied by increases in both total cell number and retinal area.

Several hypotheses can be put forward to explain the events underlying visual streak formation. Cell division might take place within the ganglion cell layer although this is unlikely since at premetamorphic stages in amphibia, mitosis is largely confined to the ciliary margin (Glücksmann, 1940; Hollyfield, 1968; Straznicky & Gaze, 1971). As an exception Hollyfield (1971) observed some mitosis in the inner nuclear layer of late larval and metamorphosing Xenopus. However, if mitosis was limited to the ciliary margin, then the extent of cell division could be greater nasally and temporally than elsewhere. Alternatively, patterns of mitosis could be similar around the retina with cell death or migration, both known to occur widely in developing neural systems (Rakic, 1977; Lamb, 1977), playing roles in shaping the streak.

Recently Tay, Hiscock & Straznicky (1982) injected Xenopus with [³H]-thymidine at metamorphic climax and examined retinae 2 months later. They found a greater addition of cells to temporal retina than elsewhere, a result which could underlie increased cell densities found in this region of wholemounts (Dunlop & Beazley, 1984). However to begin to understand visual streak formation it is necessary to examine the retina of a species such as Limnodynastes dorsalis which has more pronounced density gradients than Xenopus. In this species we have analysed patterns of cell division to determine whether they could explain changes in cell distribution. Our findings have been published in abstract form (Coleman, Dunlop & Beazley, 1983).

We examined sectioned retinae from tadpoles, animals at metamorphic climax and juveniles in four conditions:-group (i) untreated, group (ii) treated with colchicine 1–3 days before sacrifice, group (iii) injected with [³H]thymidine and sacrificed within 1 day, group (iv) injected with [³H]thymidine as tadpoles or animals at metamorphic climax and sacrificed after several weeks. Since mitosis comprises only a small part of the cell cycle, colchicine was used to arrest cells in metaphase allowing a better estimate of the size of the dividing cell population. [³H]thymidine, which is taken up during the S phase of cell division, enabled us to identify cells dividing within a few hours of injection (Beach & Jacobson, 1979). By examining the distribution of mitotic figures (groups i and ii) and of labelled cells (group iii) the site of cell division could be established. Furthermore the extents of cell division could be compared between nasal, temporal, dorsal and ventral retina from counts of mitotic figures (groups i and ii) and of labelled cells (groups iii and iv). Group (iv) was included because it was the only series in which it was possible to consider solely those cells which finally formed part of the ganglion cell layer.

We also prepared retinal wholemounts for tadpoles, animals at metamorphic climax, juveniles and adults to document changes in the number and distribution of cells in the ganglion cell layer and retinal area. In addition, axon totals were estimated in a tadpole, juvenile and an adult.

Egg masses were collected from Kings Park, Perth in June/July and adults from the metropolitan area. Tadpoles were fed Biorell fish food and kept at a density of approximately 20/5 litres of spring water. At metamorphic climax animals were transferred to shallow water and gravel. Postmetamorphic animals were fed mealworms twice weekly ad libitum or by hand. A temperature of 22° ± 2°C and a 12h light/dark cycle were maintained throughout.

Tadpoles and animals at metamorphic climax were examined at stages equivalent to 53–54 and 61–65 respectively in Xenopus (Nieuwkoop & Faber, 1956); juveniles were approximately 2 months postmetamorphosis and adults at least 1-year old (Fig. 4). For injections, anaesthesia was by immersion in 0-1 % MS222. Animals were sacrificed by decapitation while deeply anaesthetized with MS222 (1 %) except for adults which received intraperitoneal injections of Nembutal (0·7ml/gm body weight).

Colchicine (10 or 100 HIM) was administered intraperitoneally injecting 20 μl for tadpoles and animals at metamorphic climax and 50 pl for juveniles (group ii). Tadpoles and juveniles were sacrificed after 24, 48 or 72h whereas animals at metamorphic climax proved sensitive and survival was limited to a maximum of 15 h.

[³H]thymidine was injected intraperitoneally (10 μCi in 10 μl, Amersham specific activity 888GBq/mmol) and some animals at each stage were sacrificed 4 or 24 h later (group iii). Other injected tadpoles were killed either at metamorphic climax or as juveniles whilst some animals injected at metamorphic climax were examined as juveniles (group iv).

Eyes were removed with surrounding tissue to assist orientation and fixed in 10 % buffered formalin (pH 7· 4) before being wax embedded, serially sectioned and stained with haematoxylin and eosin. Right eyes were sectioned nasotemporally to reveal dorsal and ventral retina while dorsoventral sections of left eyes allowed us to examine nasal and temporal retina. Slides for autoradiography were dipped in NTB emulsion (Kodak), exposed for 10– 14 days at 4°C and developed in D19 before staining.

To examine the site of cell division we noted the position of mitotic figures in one out of every three sections of group (i) and in every section containing the centrally placed optic nerve head (Fig. 5) of group (ii); the position of labelled cells was observed in all sections of groups (iii) and (iv). To ensure that comparable sections were used for estimating numbers of dividing cells in nasal, temporal, dorsal and ventral retina, we analysed only those sections containing the optic nerve head; equal numbers of sections for both eyes of each animal were examined. In group (iv) sample grain counts were made and only the most heavily labelled cells, considered to be generated within a few hours of injection (Hollyfield, 1968; Meyer, 1978), were recorded. For statistical analysis, mean numbers of mitotic figures or labelled cells were calculated for nasal, temporal, dorsal and ventral retina. Results from individual animals were treated separately and an F test applied to determine if there were significant differences between means. If so, differences were further analysed using a series of planned orthogonal comparisons testing between poles of the same eye (i.e. nasal versus temporal and dorsal versus ventral) as well as between eyes of the same animal (i.e. nasal and temporal versus dorsal and ventral).

Eyes were removed from normal animals and fixed for at least 1 week in 10 % buffered formalin (pH 7·4). Retinae were dissected and radial cuts made before drying down, ganglion cell layer uppermost, on gelatinized slides and staining with cresyl violet (Dunlop & Beazley, 1981). Retinae were oriented by the ventral entry of the hyaloid artery. Pre- and post-staining areas were estimated from tracings using a MOP-3 image analyser (Zeiss). Shrinkage ranged from 0–12 % and by reference to features such as the vascular tree was considered to be confined largely to cut edges. The distribution of the total population in the ganglion cell layer was determined by counting cells/ (100 μm)² sample area at X1000 final magnification. Retinae were sampled systematically analysing 12·5 % of area for tadpoles and animals at metamorphic climax, 8 % and 5 % of area for juveniles and adults respectively. Total cell number in the ganglion cell layer was calculated by proportionality. To draw density profiles, cells/(100 μlm)² were also counted at intervals across the nasotemporal and dorsoventral axes centering on the optic nerve head; in the adult the nasotemporal axis transected the area centralis. Axons were counted from sample electron micrographs and the total estimated by proportionality having estimated nerve area from a scanmode montage (Beazley & Dunlop, 1983).

In untreated animals (group i), cells at all stages of mitosis were seen along the sclerad edge of the ciliary margin although their frequency was low (Fig. 1,i).

After exposure to colchicine (group ii) arrested metaphase figures accumulated in this region; they had deeply stained clumped chromatin and a thin rim of clear cytoplasm (Fig. 1,ii). The ciliary margin was always longer after colchicine treatment (Fig. 1,ii) although the numbers of arrested metaphase figures did not correlate with dose or exposure time. Mitotic figures were not observed away from the ciliary margin.

In all animals injected with [³H]thymidine and killed within 24h (group iii), label was almost exclusively limited to the ciliary margin and was found throughout its depth (Fig. 1,iii). Most labelled cells were spindle-shaped but a few labelled mitotic figures were seen adjacent to the pigment epithelium. In addition, small numbers of labelled cells were observed apparently randomly distributed throughout the retina, their appearance and position suggesting they were haematogenous.

Animals injected with [³H]thymidine and sacrificed at a subsequent stage (group iv) presented a consistent picture in that all labelled cells occupied a distinct band across the width of the nuclear layers at a distance from the ciliary margin (Fig. 1,iv; A & B). The most heavily labelled cells tended to be nearer the optic nerve head. Labelled cells extended slightly more centrally in the inner nuclear layer presumably reflecting the time at which cells left the mitotic cycle (Morris, Wylie & Miles, 1976).

All tadpoles injected with colchicine (group ii) had significantly (P<·05) more mitotic figures in nasal and temporal retina compared to dorsal and ventral. In untreated tadpoles (group i) this difference was apparent but not significant. Untreated and colchicine-treated animals at metamorphic climax and juveniles had more mitoses in nasal and temporal compared to dorsal and ventral retina. This difference was significant (P < 0·05) for all except one animal at metamorphic climax. At all stages lowest counts were usually in dorsal retina. In general numbers of mitotic figures were highest at metamorphic climax and lowest in juveniles. Our mitotic figure counts were reflected in the size of the ciliary margin (Fig. 2).

Animals injected with [³H]thymidine and killed within 24 h (group iii) had consistently more label in nasal and temporal than dorsal and ventral poles. Density of label and packing of cells precluded counts of labelled cells in some animals, particularly those at metamorphic climax. However two tadpoles and a juvenile were suitable and in these there were significantly (P<0·05) more labelled cells nasally and temporally than dorsally and ventrally. All animals injected with [³H]thymidine and killed at a subsequent stage (group iv) showed significantly (P < 0·05) more labelled cells in the ganglion cell layer nasotemporally than dorso ventrally. This result indicated that histogenetic events at the ciliary margin analysed in groups (i–iii) were reflected in the number of cells entering the ganglion cell layer (group iv).

The majority of cells in the ganglion cell layer had a similar appearance at any stage with rounded dark nuclei and sparse cytoplasm (Fig. 3A). From tadpole to juvenile, cell number and retinal area (Fig. 4) increased to similar extents (×2·7 and ×2·4 respectively) although between juvenile and adult, cell addition (×l·8) was less than areal enlargement (×3·4). In tadpoles cell densities were lowest centrally and dorsally. At metamorphosis and in juveniles high-density patches extended from nasal and temporal peripheries toward the optic nerve head while lower densities were observed dorsally and ventrally. In adults a visual streak enclosed the area centralis in temporal retina; densities were lower dorsally than ventrally. Changing cell distributions are shown as isodensity contours and density profiles (Figs 5,6). Axon (Fig. 3B) counts were 29 % (tadpole), 28 % (juvenile) and 35 % (adult) below their respective total cell estimates.

Here we have examined the site and extent of cell division in the retina of Limnodynastes dorsalis. This species resembled other Anura studied (Dunlop & Beazley, 1981,1984) by developing from metamorphosis onwards a visual streak within the ganglion cell layer.

Cell division was confined to the ciliary margin of Limnodynastes in agreement with most other studies of both fish (Hollyfield, 1972; Meyer, 1978; Sharma & Ungar, 1980) and amphibia (Straznicky & Gaze, 1971; Beach & Jacobson, 1979; Gaze, Keating, Ostberg & Chung, 1979; Tay et al. 1982). To our knowledge there is only one report of mitosis away from the ciliary margin in amphibia, this being in the inner nuclear layer of larval and metamorphosing Xenopus (Hollyfield, 1971). In contrast mitosis has been reported in the outer nuclear layer of several fish (Lyall, 1957; Ahlbert, 1976; Sandy & Blaxter, 1980; Johns & Fernald, 1981; Johns, 1982). Asymmetric patterns of cell division around the ciliary margin such as we report for Limnodynastes were found in late larval Xenopus although in most studies only the dorsoventral axis was examined (Straznicky & Gaze, 1971; Hollyfield, 1971; Jacobson, 1976; Straznicky & Tay, 1977; Beach & Jacobson, 1979; Gaze et al. 1979). We have demonstrated also that patterns of mitosis at the margin are reflected in the number of cells entering the ganglion layer in accord with the finding of asymmetric addition of cells to this layer in metamorphosing Xenopus (Tay et al. 1982).

Variations in the extents of mitosis around the ciliary margin could be explained by differences in either the size of the precursor pool or cycle time. A correlation between ciliary margin length and amounts of cell division argues for the former. Such a conclusion would be in line with the finding of equal cycle times reported for dorsal and ventral Xenopus retina (Beach & Jacobson, 1979).

It would seem probable that our finding of increased cell division at nasal and temporal compared to dorsal and ventral poles (groups i–iii) directly accounts for the greater addition of cells to the retinal ganglion cell layer nasally and temporally (group iv). These results apparently provide a sufficient explanation for the changing distributions we have observed in the ganglion cell layer of wholemounts. It is not necessary therefore to envisage major roles for other cellular events such as death and migration in visual streak formation. Indeed thorough examination did not reveal dying cells in our untreated or [³H]-thymidine-injected material. We consider it unlikely that we have misidentified dying cells, since they were readily distinguished in the group treated with colchicine, known for its toxic effects (Hollyfield, 1968; Meller, 1981). Furthermore had there been a wave of cell death, cell numbers might have been expected to drop at some stage, whereas we have reported a continual increase.

Migration of cells during visual streak formation is a more difficult issue to resolve. Grafting of retinal segments between different marker mutants (Hunt & Ide, 1977) might prove useful to examine migration from dorsal and ventral retina into the streak or between nuclear layers. However we can exclude radial migration since in animals injected with pHJthymidine and sacrificed after several weeks’ labelled cells were always closely grouped throughout the retinal layers.

It is important to consider the part played by areal growth in changing cell distributions since retinal area increased several fold from tadpole to adult. Areal enlargement in the absence of asymmetric patterns of cell division could have produced the high density patches seen in juveniles only if the dorso ventral axis had grown more than the nasotemporal one. This did not occur. From our wholemounts it is apparent that axes were of approximately equal length until the juvenile stage. However between the juvenile and adult lesser extension of the temporal compared to other axes may explain the development of an area centralis in this region. Furthermore, the overall drop in cell density observed between these stages is presumably a result of areal enlargement outstripping cell addition.

This paper has considered developmental processes which might underlie changing distributions of the total cell population in the retinal ganglion cell layer. It would be of interest to extend these studies to analyse separately the various cell types comprising this layer. The shortfall of axon to cell counts suggests at least 30 % of cells are non-ganglion cells. In the frog Hyla moorei nonganglion cells represent a higher percentage of cells in dorsal and ventral peripheries than elsewhere (Humphrey & Beazley, in press). If Limnodynastes resembles Hyla in this respect, the implication is that cell types are added in different proportions around the ciliary margin. Thus we would predict a higher percentage of ganglion cells would be generated at the nasal and temporal compared to dorsal and ventral poles. A combined [³H]thymidine and retrograde horseradish peroxidase study would address this issue.

LDB is a Senior Research Fellow, National Health and Medical Research Council Australia (NH&MRC). These experiments were performed in partial fulfilment of an honours degree by L-A.C. This research was supported by NH&MRC grants 79/2087 and 82/0180 and the Muscular Dystrophy Research Association of Western Australia. We are grateful for the use of the Electron Microscopy Centre, University of Western Australia. We thank J. Durston, J. Darby, M. Stevens and H. Jurkiewicz for histological, art and photographic assistance. C. Pfaff, P. Basden and D. Anstey are thanked for typing the manuscript.