The optic tract of the goldfish splits into two brachia just before it reaches the tectum, normal optic axons being distributed systematically between the two according to their retinal origins. The orderliness of this division, like that of the retinotectal projection itself, is conventionally attributed to a system of specific axonal guidance cues. However, the brachial distribution of regenerated axons is much less orderly; and, since there is evidence that these axons have many collateral branches in the nerve and tract, the gross order that remains after regeneration could potentially arise secondarily, in parallel with refinement of the retinotectal map, by a preferential loss of collaterals from the inappropriate brachium.

The brachial paths of normal axons, and axons regenerated after optic nerve cut for periods ranging from 19 days to 5 years, were therefore studied by anterograde labelling with horseradish peroxidase from discrete retinal lesions or retrograde labelling of ganglion cells from a cut brachium.

From 19 to 28 days, regenerating axons showed little or no preference for their normal brachium. During this period (which includes the first week of tectal synaptogenesis) an average of 46·3 % of cells retrogradely labelled from a cut medial brachium were in dorsal retina, compared with only 1· 45% in normal fish. Some preference for the normal brachium was evident at 35 days and significant order had returned by 42–70 days, when the average proportion of labelled cells in dorsal retina had fallen to 25·4 % though the average number in the whole retina was unchanged. Thus a brachial refinement had occurred in parallel with refinement of the retinotectal map.

These results support the idea of a selective loss of axon collaterals from the inappropriate brachium, though they do not exclude the possibility of some concurrent gain in the appropriate one. We suggest that refinement may depend on a process we term ‘sibling rivalry’: competition between different collaterals of the same axon to form a critical number of stable tectal synapses, in which the most-normally-routed branches have the best chance of succeeding and surviving. Developing normal axons might also make use of collateral formation and ‘sibling rivalry’ to generate and refine the complex interwoven patterns of the normal optic tract.

The retinotectal system of lower vertebrates has been the basis of a broad range of influential studies of axonal growth and synaptic specificity over four decades (Cowan & Hunt, 1985). In this paper, we aim to shed some new light on the old problem of axonal guidance in this system by focusing on just one region of the goldfish optic pathway, that where the optic tract splits into two brachia just before it reaches the optic tectum. In normal development, the axons of retinal ganglion cells seem to be very specific in their choice of brachium: most axons from dorsal retina take the lateral brachium and most from ventral retina the medial (Attardi & Sperry, 1963; Horder, 1974; Bunt, 1982; Springer & Mednick, 1985, 1986a). A similar division occurs in Xenopus. Straznicky, Gaze & Horder (1979) showed that optic axons from both halves of a double ventral ‘compound’ eye, formed by the fusion of two embryologically ventral half eye rudiments, tended to select their embryologically correct brachium, the medial. This has been taken as evidence for active guidance in the region of the bifurcation, contributing to the formation of the retinotopic projection to the tectum (Straznicky et al. 1979; Fawcett & Gaze, 1982).

However, brachial selection is not as precise in animals where the optic nerve has been cut and allowed to regenerate (fish: Horder, 1974; Cook, Pilgrim & Horder, 1983; frog: Straznicky et al. 1979; Gaze & Fawcett, 1983; Taylor & Gaze, 1985). Stuermer & Easter (1984) labelled regenerating goldfish optic axons with horseradish peroxidase (HRP) from the retina and traced their routes through the optic tract. They found brachial selection still in evidence, but estimated that 20% of regenerating axons entered the inappropriate brachium in contrast to only about 1 % of normal axons.

In the optic tectum, too, many regenerated axons are misrouted (Meyer, 1980; Cook, 1983a; Cook et al. 1983; Stuermer & Easter, 1984). The retinotopic map that is formed initially is crude and an activitydependent process is involved in its subsequent refinement to a degree of retinotopy that approximates the normal (Meyer, 1983; Schmidt & Edwards, 1983; Schmidt & Eisele, 1985; Rankin & Cook, 1986; Cook & Rankin, 1986). When refinement is complete, even the most grossly misrouted axons can still terminate retinotopically (Horder, 1974; Cook et al. 1983). These observations are inconsistent with the proposal of Attardi & Sperry (1963) that axons (both normal and regenerating) are guided direct to their retinotopic termination sites; and make the significance of the precise brachial selection in normal fish less clear.

Murray (1982) and Murray & Edwards (1982) showed, by axonal counts in a light and electron microscopic study, that damaged retinal ganglion cell axons branch profusely in the optic nerve, tract and tectum, though many of the branches are subsequently lost. Though this work is often cited, one of its most interesting implications seems to have gone unrecognized, at least in print: an axon possessing such collateral branches could send them into both brachia in search of its tectal target. Chance selection of the normal brachium by some collaterals would bring them closer to their retinotopic tectal sites: this might enable them to form stable synapses before other collaterals of the same axons, whose subsequent selective loss could underlie the crude order seen in the fully regenerated pathway.

The present experiments were designed to investigate the possibility that branching in regeneration initially produces a disorderly, non-retinotopic pathway that is subsequently refined. Two approaches were taken. First, the paths of normal and regenerating axons within the optic tract were studied directly,by labelling a small number of them with HRP from discrete retinal lesions and tracing them through serial sections. Second, the axons in a single brachium, either medial or lateral, were cut and filled with HRP to allow us to analyse numerically the retinal distribution of their cell bodies. Some of the results have been published in abstract form (Becker & Cook, 1986).

Animals

Goldfish (Carassius auratus; 60–63 mm long snout to tail base) were obtained locally. All surgery was performed under anaesthesia, initiated by immersion in a 1 % aqueous solution of MS222 (ethyl 3-aminobenzoate, methanesulphonic acid salt; Aldrich) and maintained by passage of a 0·1 % solution over the gills. Right optic nerves were cut cleanly in mid orbit with fine iridectomy scissors. Care was taken not to damage accompanying blood vessels and fish that showed signs of haemorrhage were eliminated at this point. The remainder were then maintained under diurnal lighting at 20±0·5°C for between 19 and 84 days. Four larger fish (75–90mm long), whose right optic nerves had been cut in the same way 1251–1823 days (3–5 years) earlier, had been kept at room temperature (16–24°C).

HRP application and specimen preparation

Anterograde tracing

The cornea of the right eye was cut three quarters of the way around its margin and folded back. The lens was gently removed without retinal damage and the aqueous humour was aspirated. Groups of axon fascicles in dorsal or ventral retina were cut with a fine tungsten needle, without damage to major blood vessels. A pledget of gelfoam (Sterispon No. 3; Allen & Hanbury) soaked in 30 % HRP (Sigma type VI in distilled water) was placed over the damaged area, the lens replaced and the corneal flap returned. The fish were then maintained at room temperature for a further 24 h to allow anterograde transport to take place, and the brains were removed and fixed in 5 % glutaraldehyde in 0·2 M-phosphate buffer for 2h. Coronal sections were cut at 150μm on a Vibratome (Oxford), reacted to reveal transported HRP (Hanker, Yates, Metz & Rustioni, 1977) and mounted on gelatin-subbed slides.

Retrograde tracing

A cranial flap was opened to give access to the brain. Excess fluid was sucked away and the left lobe of the forebrain was gently displaced to reveal the brachia of the left optic tract. One brachium was severed with a fine tungsten needle and a 5 % solution of HRP in 0·2 M-KCI, buffered to pH 7·9 with 0·05M-Tris, was applied to the cut on a piece of gelfoam. The cranial flap was then replaced, being held by its bevelled edges. Every batch of fish thus labelled contained at least two normal fish to confirm the consistency of the operative procedure. All fish were maintained for a further 8 days for retrograde transport to take place. Fish were dark adapted for at least 45 min before excision of the right eye. Prior to dissection, a needle was used to pierce through the sclera, choroid and retina to mark the positions of the irideal darts (Springer & Mednick, 1986a), and a cut was made through all these layers along the line of the choroid fissure. Retinae were then dissected free, in 0·9% NaCl, from the outer coats and pigment epithelium. The cut marking the fissure was extended almost to the optic disc and three shorter cuts were made to allow flattening: one dorsally, in line with the first, and two through the needle holes in the retina marking the position of the irideal darts, each extending half way to the optic disc. Retinae were flattened, fixed in 1 % glutaraldehyde in 0·2M-phosphate buffer for 30min, rinsed in saline for 10min and processed to reveal HRP by a slightly modified Hanker-Yates method (Cook, 1983a). They were then dehydrated, cleared in methyl salicylate and flatmounted in DePeX, with a piece of film bearing a grid of 0·25 mm squares sandwiched between the retina and the slide (Cook, 1987) to make it possible to define any point on the retina and return to it reliably.

Analysis

Camera-lucida drawings of anterogradely labelled axons in the optic tract were made from serial coronal Vibratome sections. Each drawing was matched to its neighbours, with the aid of landmarks such as blood vessels, to create a composite view of the whole brachial region showing the pattern of innervation. Reliable estimates of axonal number could not be made, however, because of the difficulty of counting axons that travelled together in fascicles.

To study the distribution of retrogradely labelled retinal ganglion cells after brachium fills, retinae were notionally divided into quadrants, two dorsal and two ventral. Two different methods of doing this have been proposed but neither is well established, so a combination of both was employed. The first method subdivides each flatmount by area, starting at the choroid fissure (Cook, 1983b). A camera-lucida drawing of the retina was made on paper with squares matching those of the grid on which the retina was mounted and its area was measured on a graphics tablet. Straight lines were then drawn from the centre of the optic disc to the edge of the retina, starting ventrally at the cut marking the choroid fissure and moving round in quarters of the total retinal area. The second method relies on the irideal darts as natural boundary markers (Springer & Mednick, 1986a). For this, additional straight lines were drawn from the centre of the optic disc to the cuts marking the dart positions. All these lines are shown in Fig. 3. A combination of both methods was effective in predicting the important dorsoventral boundary in normal fish (Fig. 2): squares between the two estimates of this boundary, and squares within 250μm of either estimate, were discounted so that only those falling entirely within a quadrant as defined by both methods of estimation would be used in the analysis.

Ganglion cells labelled from the medial brachium were then counted in every eighth remaining complete square, selected by randomly placing a point grid of appropriate spacing over the retinal drawing. Cells intersecting the edges of a sampled square were only counted if they lay on its top or left side. The number of labelled cells in each quadrant was estimated by dividing the total count for the sampled squares in that quadrant by the total area of these squares to obtain an estimate of the mean cell density and then multiplying this by the quadrant area. The nasal and temporal quadrants of each half-retina (dorsal or ventral), showed no consistent differences and were later pooled. Thus the number of labelled cells in the dorsal half-retina, (which does not normally project through the medial brachium) could be expressed as a percentage of the total number labelled, a percentage which directly reflects the proportion of ganglion cells having at least one axon in the inappropriate brachium. This proportion was relatively uniform from fish to fish, though it varied with time from nerve cut. The actual number of labelled cells was much more variable, especially at the earliest stages of regeneration, though it did not vary systematically with time. We discarded one retina in which fewer than 30 labelled cells were found within the sampled squares.

Normal axons

Most of the axons in the medial brachium of the normal optic tract are known to originate from the ventral region of the retina, and most of those in the lateral brachium from the dorsal region (see Introduction). Retrograde transport of HRP from a cut brachium in each of 28 normal fish (the medial in 15 and the lateral in 13) labelled ganglion cells that were largely confined to the expected retinal region, as illustrated in Fig. 1A. Representative fields of normal dorsal and ventral retina containing ganglion cells labelled from the medial brachium are shown in Fig. 1B,C.

Fig. 1.

(A) Part of a flat-mounted retina from a normal goldfish, reacted by the Hanker–Yates method to reveal HRP. Ganglion cells and their axons labelled from the cut medial brachium of the optic tract are almost entirely confined to ventral retina (lower half). Dorsal is up, nasal to the right and the optic disc at the left. Bar, 250μm. (B) Typical field from the dorsal region of such a retina. A ring surrounds a single labelled ganglion cell. Erythrocytes in blood vessels also stain by this method, and the outlines of unlabelled ganglion cells can be seen, lying in rows between fascicles of optic axons. Bar, 100 μm. (C) Typical field from the ventral region of the same retina, to the same scale. Labelled ganglion cells are very numerous.

Fig. 1.

(A) Part of a flat-mounted retina from a normal goldfish, reacted by the Hanker–Yates method to reveal HRP. Ganglion cells and their axons labelled from the cut medial brachium of the optic tract are almost entirely confined to ventral retina (lower half). Dorsal is up, nasal to the right and the optic disc at the left. Bar, 250μm. (B) Typical field from the dorsal region of such a retina. A ring surrounds a single labelled ganglion cell. Erythrocytes in blood vessels also stain by this method, and the outlines of unlabelled ganglion cells can be seen, lying in rows between fascicles of optic axons. Bar, 100 μm. (C) Typical field from the ventral region of the same retina, to the same scale. Labelled ganglion cells are very numerous.

Since the brachial distribution becomes blurred in regeneration, it was important, for sampling purposes, to have an independent method of identifying these dorsal and ventral regions and the boundary between them. Two independent estimates of its position were obtained for each retina: one based on pigmented marks, termed irideal darts (Springer & Mednick, 1986a), which are associated with blood vessels in the iris; the other on area measurements and the ventral choroid fissure (Cook, 1983b). An idea of the precision of these estimates can be gained from Fig. 2, which shows a series of plots of retrogradely labelled cells across the boundary zone, with the estimated boundary lines superimposed, for four normal fish. There were no consistent differences in precision between nasal and temporal retina or between fish with medial and lateral brachial labelling. As can be seen from the plots, the boundary was never discrete: the density of labelled cells tailed off over some 200–400μm. This transition region was therefore avoided in sampling (see Materials and methods): all counts, in both normal and regenerated retinae, were made in regions that were unequivocally dorsal or ventral in terms of their normal projection pattern.

Fig. 2.

Tracings of labelled ganglion cell distributions across the boundary between dorsal and ventral retina in four normal fish, with independent estimates of the boundary position. (A–C) Medial brachium fills. (D) Lateral brachium fill. Each column represents a strip 500 μm by 2000μm, midway between the optic disc and the retinal margin. Dorsal is up. Lines marked d, with long dashes, run from the disc towards each irideal dart. Lines marked a, with short dashes, divide each retina into four quadrants of equal area (see Fig. 3 and Materials and methods). Each dot represents one labelled ganglion cell. N and T identify strips from temporal and nasal retina: paired strips are from the same specimen, and the optic disc lay between them. The actual boundary of the heavily labelled region usually falls between the two estimates (A) but may match either the dart line (B,C) or the area line (D). The dart line is commonly the more dorsal of the two in nasal retina and the more ventral in temporal retina (A,B), as implied by Springer & Mednick (1986a); but the relationship can also be reversed (C,D).

Fig. 2.

Tracings of labelled ganglion cell distributions across the boundary between dorsal and ventral retina in four normal fish, with independent estimates of the boundary position. (A–C) Medial brachium fills. (D) Lateral brachium fill. Each column represents a strip 500 μm by 2000μm, midway between the optic disc and the retinal margin. Dorsal is up. Lines marked d, with long dashes, run from the disc towards each irideal dart. Lines marked a, with short dashes, divide each retina into four quadrants of equal area (see Fig. 3 and Materials and methods). Each dot represents one labelled ganglion cell. N and T identify strips from temporal and nasal retina: paired strips are from the same specimen, and the optic disc lay between them. The actual boundary of the heavily labelled region usually falls between the two estimates (A) but may match either the dart line (B,C) or the area line (D). The dart line is commonly the more dorsal of the two in nasal retina and the more ventral in temporal retina (A,B), as implied by Springer & Mednick (1986a); but the relationship can also be reversed (C,D).

A typical set of cell counts from a normal retina labelled from the medial brachium is shown as Fig. 3. In this instance, only 1·34 % of the labelled cells were estimated to be in the dorsal half-retina, as defined by area measurements. For the 15 normal fish labelled in this way, the mean was 1·45%. Despite the strong numerical bias towards ventral retina, the range of intensity of the labelling among individual labelled cells was very similar in the two regions.

Fig. 3.

Tracing of a flat-mounted normal retina labelled from a cut medial brachium. Lines of long dashes run from the optic disc towards the irideal darts. Lines of short dashes divide the flat mount into four quadrants of equal area, based on the position of the ventral choroid fissure along which the first cut (V) was made. N, T and D indicate the directions of the nasal, temporal and dorsal poles. Labelled ganglion ceils were counted in each of the 0·25 mm squares shown, avoiding quadrant boundaries, and the total number in each quadrant was estimated. In this example, only 1· 34 % of the labelled cells were in the dorsal half of the retina. The mean for 15 such retinae was 1·45 %. Typical fields from a similar retina are shown in Fig. 1. Bar, 1 mm.

Fig. 3.

Tracing of a flat-mounted normal retina labelled from a cut medial brachium. Lines of long dashes run from the optic disc towards the irideal darts. Lines of short dashes divide the flat mount into four quadrants of equal area, based on the position of the ventral choroid fissure along which the first cut (V) was made. N, T and D indicate the directions of the nasal, temporal and dorsal poles. Labelled ganglion ceils were counted in each of the 0·25 mm squares shown, avoiding quadrant boundaries, and the total number in each quadrant was estimated. In this example, only 1· 34 % of the labelled cells were in the dorsal half of the retina. The mean for 15 such retinae was 1·45 %. Typical fields from a similar retina are shown in Fig. 1. Bar, 1 mm.

Anterograde labelling of normal axons, from lesions restricted to regions well within either dorsal or ventral retina (as defined at the time of surgery by the irideal darts), showed their brachial distribution directly (Fig. 4). Although the vast majority of axons from one such retinal region travelled together through one brachium, a few well-labelled individuals could usually be seen taking a deviant route through the opposite one, confirming the retrograde results. The discrete distribution of anterograde label also confirmed that label uptake was restricted to the retinal lesion site under these conditions.

Fig. 4.

Coronal sections through the brachial region of the left optic tract in two normal fish. MBr, LBr mark the medial and lateral brachia; NR the nucleus rotundus that separates them. Dorsal is up, towards the tectum. Bar, 100μm. (A) Normal axons in the medial brachium, labelled by anterograde transport of HRP from a small lesion confined to ventral retina. (B) Normal axons in the lateral brachium, similarly labelled from a larger lesion confined to dorsal retina. A few labelled axons (bracket) can also be seen in the medial brachium.

Fig. 4.

Coronal sections through the brachial region of the left optic tract in two normal fish. MBr, LBr mark the medial and lateral brachia; NR the nucleus rotundus that separates them. Dorsal is up, towards the tectum. Bar, 100μm. (A) Normal axons in the medial brachium, labelled by anterograde transport of HRP from a small lesion confined to ventral retina. (B) Normal axons in the lateral brachium, similarly labelled from a larger lesion confined to dorsal retina. A few labelled axons (bracket) can also be seen in the medial brachium.

Regenerating axons

The quality of anterograde labelling was generally much poorer in regenerates than in normal fish (see Discussion). Both the number of visible axons and the intensity of their labelling decreased. Of 40 fish operated upon at all stages of regeneration, 24 were adequately labelled. Details of these are given in Table 1. Poor labelling, the shallow depth of field associated with high-resolution objectives and the tortuous paths taken by many regenerating axons made them difficult to photograph, so camera-lucida drawings of serial sections through the tract were used to illustrate their paths.

Table 1.

Brachial distribution of HRP transported anterogradely from a small retinal lesion at different times after optic nerve cut

Brachial distribution of HRP transported anterogradely from a small retinal lesion at different times after optic nerve cut
Brachial distribution of HRP transported anterogradely from a small retinal lesion at different times after optic nerve cut

Early stages: anterograde labelling

In the early stages of regeneration, between 20 and 29 days after optic nerve cut, axons labelled from either dorsal or ventral retina were usually seen in approximately equal proportions in both brachia: indeed, more were sometimes visible in the inappropriate brachium than in the appropriate one (Table 1). Fig. 5 shows four representative distributions, two after ventral and two after dorsal lesions: the lesion sites cannot be deduced from the axon patterns. Individual axons often entered the nearest brachium, even when it was retinotopically inappropriate. A photomontage showing some of the less tortuous and better-labelled axons in the inappropriate brachium of Fig. 5B, followed through 20 focal planes in two consecutive sections, is reproduced as Fig. 6A.

Fig. 5.

Tracings compiled from several adjacent Vibratome sections of the brachial region in each of four fish, early in optic nerve regeneration. Orientation and abbreviations as for Fig. 4. Bar, 200μm. (A) Axons labelled from a discrete retinal lesion of ventral retina, 24 days after optic nerve cut. Many are present in both brachia. (B) Axons labelled from a similar ventral lesion after 29 days. Most, in this case, are in the inappropriate lateral brachium. A photomontage of the boxed region is shown as Fig. 6A. (C) Axons labelled from a similar lesion in dorsal retina after 20 days. They are evenly divided between the brachia. (D) Axons labelled from a similar dorsal lesion after 24 days. In this case, as in B, there are more axons in the lateral brachium, though this time it happens to be the appropriate one. Axons in this brachium may be slightly easier to identify in coronal sections (see Table 1).

Fig. 5.

Tracings compiled from several adjacent Vibratome sections of the brachial region in each of four fish, early in optic nerve regeneration. Orientation and abbreviations as for Fig. 4. Bar, 200μm. (A) Axons labelled from a discrete retinal lesion of ventral retina, 24 days after optic nerve cut. Many are present in both brachia. (B) Axons labelled from a similar ventral lesion after 29 days. Most, in this case, are in the inappropriate lateral brachium. A photomontage of the boxed region is shown as Fig. 6A. (C) Axons labelled from a similar lesion in dorsal retina after 20 days. They are evenly divided between the brachia. (D) Axons labelled from a similar dorsal lesion after 24 days. In this case, as in B, there are more axons in the lateral brachium, though this time it happens to be the appropriate one. Axons in this brachium may be slightly easier to identify in coronal sections (see Table 1).

Fig. 6.

(A) Photomontage of the boxed region of the lateral brachium in Fig. 5B, reconstructed from 20 focal planes in two 150 μm Vibratome sections. The regenerating axons or fascicles shown were labelled with HRP, 29 days after optic nerve cut, from a discrete lesion of ventral retina. Thus they are in the inappropriate brachium Dorsal is up, medial to the left. Bar, 100μm. (B) Photomontage of axons and possible collateral branches in the inappropriate medial brachium, labelled from dorsal retina 28 days after optic nerve cut and reconstructed from 10 focal planes. The main structure may be an axon fascicle: asterisks mark loops that appear to be axons briefly leaving and rejoining it. Arrowheads mark putative branch points, the first of which, at least, may be only a point of defasciculation. Dorsal is up, medial to the left. Bar, 100μm.

Fig. 6.

(A) Photomontage of the boxed region of the lateral brachium in Fig. 5B, reconstructed from 20 focal planes in two 150 μm Vibratome sections. The regenerating axons or fascicles shown were labelled with HRP, 29 days after optic nerve cut, from a discrete lesion of ventral retina. Thus they are in the inappropriate brachium Dorsal is up, medial to the left. Bar, 100μm. (B) Photomontage of axons and possible collateral branches in the inappropriate medial brachium, labelled from dorsal retina 28 days after optic nerve cut and reconstructed from 10 focal planes. The main structure may be an axon fascicle: asterisks mark loops that appear to be axons briefly leaving and rejoining it. Arrowheads mark putative branch points, the first of which, at least, may be only a point of defasciculation. Dorsal is up, medial to the left. Bar, 100μm.

To confirm that the broad distribution of labelled regenerated axons was not an artefact, resulting perhaps from a generalized uptake of HRP by chromatolytic ganglion cells during the period when the neuronal response to injury is at its peak (Murray & Grafstein, 1969), we placed HRP in the usual way on unlesioned retinae, in ten fish between 19 and 29 days after optic nerve cut. None of these fish showed any labelling in the tract. Nine similar fish were given the usual retinal lesions at the same time and six of these showed axonal labelling. The link between lesioning and labelling was statistically highly significant (P = 0·003; Fisher exact probability test), so we conclude that the labelling, even at this early stage, was restricted to axons crossing (or arising within) the lesion site.

The limited resolution of the light microscope makes it impossible to see the finest regenerating axons, to tell how many are in a fascicle, or to distinguish with confidence between branch points of a single axon and points of divergence of individual fasciculated axons (Tosney & Landmesser, 1985). One such axon fascicle is shown both as a photomontage and as a camera-lucida drawing in Fig. 6B, and may well show both divergence and branching. Because of this, it would have been inappropriate to quantify the anterograde results in more detail than that given in Table 1 and even the key given there can only be a rough guide to the minimum numbers.

Early stages: retrograde labelling

To make a quantitative assessment of brachial selection by regenerating optic axons, we labelled those in the medial brachium at different times in regeneration and analysed the pattern of their retinal origins. Between 19 and 28 days after nerve cut, 20 fish were successfully labelled in this way. The distribution of their labelled ganglion cells was almost random. Representative dorsal and ventral fields from a retina early in regeneration are shown in Fig. 7A,B. Approximately equal numbers of labelled retinal ganglion cells are present in both. A typical set of cell counts from a retina of this type, labelled from the medial brachium only 24 days after nerve cut, is shown as Fig. 8. In this instance, 43·1% of the labelled cells were in dorsal retina. For all 20 fish, the mean was 46·3 %.

Fig. 7.

Typical fields from a retina labelled from the medial brachium early in regeneration, 24 days after optic nerve cut. Bar, 100 μm. (A) Dorsal region. (B) Ventral region. In extreme contrast to the normal pattern (Fig. 1), the two fields contain very similar numbers of well-labelled ganglion cells. The rows of unlabelled cells are more clearly visible than normal as a result of the cellular enlargement that follows axotomy and reaches its peak at about this stage.

Fig. 7.

Typical fields from a retina labelled from the medial brachium early in regeneration, 24 days after optic nerve cut. Bar, 100 μm. (A) Dorsal region. (B) Ventral region. In extreme contrast to the normal pattern (Fig. 1), the two fields contain very similar numbers of well-labelled ganglion cells. The rows of unlabelled cells are more clearly visible than normal as a result of the cellular enlargement that follows axotomy and reaches its peak at about this stage.

Fig. 8.

Tracing of a flat-mounted retina labelled, 24 days after optic nerve cut, from the medial brachium, showing counts of labelled ganglion cells. Procedure and conventions as for Fig. 3. In this example, almost half the labelled ganglion cells (43·1 %) were in dorsal retina. The mean for 20 such retinae, labelled between 19 and 28 days, was 46· 3 %. Bar, 1 mm.

Fig. 8.

Tracing of a flat-mounted retina labelled, 24 days after optic nerve cut, from the medial brachium, showing counts of labelled ganglion cells. Procedure and conventions as for Fig. 3. In this example, almost half the labelled ganglion cells (43·1 %) were in dorsal retina. The mean for 20 such retinae, labelled between 19 and 28 days, was 46· 3 %. Bar, 1 mm.

To confirm that this widespread retrograde labelling arose from axons in the cut medial brachium alone, we applied HRP to the regenerating medial brachium without cutting it in seven fish, 19 days after nerve cut. None of these retinae showed any labelling.

Changes with time: anterograde labelling

The paths taken by optic axons later in regeneration (70–84days) were as tortuous as before, but the proportion following grossly inappropriate routes had fallen (Table 1). A tracing from the brachial region of a fish labelled from a dorsal retinal lesion 70 days after nerve cut is shown as Fig. 9A. The bias towards the appropriate brachium seen in all five fish at this stage, though admittedly assessed from only a small number of labelled axons in each fish, consistently reflected the site of the retinal lesion. It thus serves not only to confirm earlier findings based on similar methods (Stuermer & Easter, 1984) but also to show that the lack of such a bias in the larger number of fish studied within the first 29 days was not due to any gross deficit in our method.

Fig. 9.

(A) Tracing compiled from several adjacent Vibratome sections of the brachial region late in regeneration, showing axons labelled from a dorsal retinal lesion 70 days after optic nerve cut. By this time, most axons were in the appropriate lateral brachium. Orientation and abbreviations as for Fig. 4. Bar, 200μm. (B) Tracing of a flat-mounted retina labelled, 70 days after optic nerve cut, from the medial brachium, showing counts of labelled ganglion cells. Procedure and conventions as for Fig. 3. The proportion in dorsal retina had fallen to 25·4 %, which was also the mean for the 15 retinae labelled between 42 and 70 days. Similar distributions were found in fish labelled up to 5 years after nerve cut. Bar, 1mm.

Fig. 9.

(A) Tracing compiled from several adjacent Vibratome sections of the brachial region late in regeneration, showing axons labelled from a dorsal retinal lesion 70 days after optic nerve cut. By this time, most axons were in the appropriate lateral brachium. Orientation and abbreviations as for Fig. 4. Bar, 200μm. (B) Tracing of a flat-mounted retina labelled, 70 days after optic nerve cut, from the medial brachium, showing counts of labelled ganglion cells. Procedure and conventions as for Fig. 3. The proportion in dorsal retina had fallen to 25·4 %, which was also the mean for the 15 retinae labelled between 42 and 70 days. Similar distributions were found in fish labelled up to 5 years after nerve cut. Bar, 1mm.

Changes with time: retrograde labelling

As regeneration progressed, the proportion of retrogradely labelled ganglion cells in the dorsal half of retinae labelled from the medial brachium also fell. Representative dorsal and ventral fields from a lateregenerated retina, labelled in this way 70 days after nerve cut, are shown in Fig. 10A,B. Fig. 9B shows a typical set of cell counts from another 70-day retina labelled in the same way. In this instance, only 25·4 % of the labelled cells were in dorsal retina. For the 15 fish labelled in this way between 42 and 70 days after nerve cut, the mean was also 25·4 %. The difference in distribution between early-regenerated (19–28 day) and late-regenerated (42–70 day) retinae was highly significant (P< 0·0001; one-tailed Mann-Whitney U test), though a highly significant difference (P< 0·0001) also remained between the late-regenerated retinae and the normal ones.

Fig. 10.

Typical fields from a retina labelled from the medial brachium late in regeneration, 70 days after optic nerve cut. Bar, 100, μm. (A) Dorsal region. (B) Ventral region. The normal pattern (Fig. 1) has not been restored after regeneration: the dorsal field still contains many labelled ganglion cells. Nevertheless, it now contains only about one third as many as the ventral field, implying that many dorsal retinal axons or collateral branches have been lost from the inappropriate medial brachium.

Fig. 10.

Typical fields from a retina labelled from the medial brachium late in regeneration, 70 days after optic nerve cut. Bar, 100, μm. (A) Dorsal region. (B) Ventral region. The normal pattern (Fig. 1) has not been restored after regeneration: the dorsal field still contains many labelled ganglion cells. Nevertheless, it now contains only about one third as many as the ventral field, implying that many dorsal retinal axons or collateral branches have been lost from the inappropriate medial brachium.

The course of brachial refinement during regeneration is presented graphically in Fig. 11A, which includes the results of all successful medial brachium fills in normal fish and in regenerated fish up to 70 days, including four fish at 35 days that were excluded from the U tests because their results were obviously intermediate between those of earlier and later groups. The data from fills made up to 42 days after nerve cut are adequately described by a regression line with a coefficient of –1·045 %/day and a constant of 70·37 %. According to the regression equation, ganglion cells with at least one axon in the medial brachium were, on average, completely evenly distributed across dorsal and ventral retina at 19·5 days, when their axons were just about to form synapses in the tectum (Schmidt, Edwards & Stuermer, 1983; Stuermer & Easter, 1984). By 42 days, however, almost three quarters of them were in ventral retina and only about a quarter (26·48 %) in dorsal retina. By 70 days, there had been little or no further change. Very similar results were also obtained from a group of four fish that had had their optic nerves cut between 3 and 5 years previously. In these, the proportion of labelled cells in dorsal retina ranged from 22·0 % to 37·2 % (mean 28·8 %). Although they were excluded from the U tests described above (and from Fig. 11) for the sake of statistical rigour, being older and larger than the other fish and perhaps having undergone some additional neurogenesis at the retinal margin (though this would have had little or no influence on the results because of our sampling procedure, which tended to exclude the margin), they demonstrate that the brachial distribution of regenerated axons never returned to normal. The equivalent proportion in four normal retinae of similar age and size ranged from 0· 8% to 3·8% (mean 1·8%), the difference being statistically significant (P = 0·014; one-tailed U test).

Fig. 11.

(A) Graph showing the percentages of ganglion cells, labelled from the medial brachium, in the inappropriate dorsal quadrants of all 15 normal retinae, and all 39 regenerated retinae studied between 19 and 70 days after optic nerve cut. Short lines for the normal and 70 day groups indicate mean values. The regression line is limited to the period between 19 and 42 days since little change was seen thereafter (note breaks in time axis). This line begins near 50 % and ends near 25 %. Its parameters are given in the text. (B) Graph showing the average densities of labelled ganglion cells in the same 39 fish between 19 and 70 days after optic nerve cut. There was no systematic change in labelled cell density with time: the slope of the regression line is nonsignificant and very close to zero. Note that the density varied far more, at all stages of regeneration, than did the pattern of distribution between dorsal and ventral within the same set of retinae. Ringed points in both graphs represent the retinae of Figs 8 and 9B.

Fig. 11.

(A) Graph showing the percentages of ganglion cells, labelled from the medial brachium, in the inappropriate dorsal quadrants of all 15 normal retinae, and all 39 regenerated retinae studied between 19 and 70 days after optic nerve cut. Short lines for the normal and 70 day groups indicate mean values. The regression line is limited to the period between 19 and 42 days since little change was seen thereafter (note breaks in time axis). This line begins near 50 % and ends near 25 %. Its parameters are given in the text. (B) Graph showing the average densities of labelled ganglion cells in the same 39 fish between 19 and 70 days after optic nerve cut. There was no systematic change in labelled cell density with time: the slope of the regression line is nonsignificant and very close to zero. Note that the density varied far more, at all stages of regeneration, than did the pattern of distribution between dorsal and ventral within the same set of retinae. Ringed points in both graphs represent the retinae of Figs 8 and 9B.

In contrast to the progressive change in distribution of labelled ganglion cells between dorsal and ventral retina, there was no consistent change in their total number: the mean cell density for all the sampled squares of each retina, dorsal and ventral together, showed no detectable change with time (Fig. 11B). Comparison of Fig. 11A and B shows that the dorsoventral distribution and overall density of the labelled cells were almost entirely independent of each other and this was confirmed by correlation analysis (Spearman’s rs = – 0·09).

Interpretation of anterograde results

In normal fish, the predominance of axonal label in one brachium after application of HRP to discrete dorsal or ventral retinal lesions showed clearly that uptake was limited to the region of the lesion site. Moreover, when we applied HRP in the same way to both lesioned and unlesioned retinae early in regeneration, labelling was confined to the fish with lesions. Thus we can be confident that the labelled axons seen entering the two brachia at early stages of regeneration were all from the same half-retina, and consequently that their even-handed distribution reflected an initial lack of brachial selection. In principle, the small numbers of axons labelled in each fish lay us open to the charge that their distributions might not be typical of the population as a whole; but in practice, the similarities between the anterograde and retrograde results, and between the results at later stages and those of Stuermer & Easter (1984), argue strongly against this. It is puzzling that regenerating axons should be so much harder to label from the retina than normal ones, especially when anterograde transport (at least of newly synthesized proteins) is known to be enhanced (Grafstein & Murray, 1969), but others have reported a similar problem (Murray & Edwards, 1982).

Interpretation of retrograde results

The complementary retrograde study addressed the question of brachial distribution quantitatively. The presence of label in a retinal ganglion cell was taken to mean that it had at least one axon in the cut brachium. We can be reasonably sure that the broad distribution of labelled cells early in regeneration was not due to HRP uptake by growth cones in the uncut brachium. The barrier to diffusion presented by the intact pia mater, and the small amount of HRP applied to the cut brachium, together made it seem unlikely that intense labelling of such uniformity across both dorsal and ventral retina could be achieved in this way; and a direct test for labelling in the absence of a brachial cut confirmed this. The anterograde results also argue against any such explanation: in this respect, the two tracing methods serve as extra controls for each other.

The apparently complete absence of a dorsoventral bias among the labelled ganglion cells early in regeneration might be thought to reflect a totally indiscriminate invasion of the two brachia by the regenerating axons. In fact, some caution is needed here. Although this inference would be justified if each ganglidn cell had only one axon, the situation in the presence of collaterals is more complex. Strictly, all we should infer is that ganglion cells in dorsal and ventral retina were equally likely to have at least one collateral in the medial brachium. If enough collaterals were formed, a small initial bias in their distribution might perhaps have passed undetected, though a substantial bias like that seen later would have been exposed by anterograde tracing.

The subsequent brachial refinement has, in principle, several potential explanations. For example, some axons might conceivably both regenerate later than others and show more selectivity. However, if that were so there should have been a substantial overall increase in the number of retrogradely labelled ganglion cells during the period of refinement. No such increase was found, neither was there any correlation among individual retinae between labelled cell number and dorsoventral distribution: thus brachial refinement could only be explained in such terms by assuming that ganglion cells became progressively harder to label (that is, more abnormal) as regeneration proceeded. At least three early retinae (two at 21 and one at 28 days: see Fig. 11B) already exhibited such high overall labelled cell densities that they could probably not have achieved the three-fold density increase in the ventral half needed for refinement by late axon outgrowth. Regenerated axons may, however, have formed additional collaterals during the refinement period (Murray, 1982).

Another potential explanation might be a massive selective loss of ganglion cells. Cell loss is certainly a feature of regeneration in some types of frog (Humphrey & Beazley, 1985; Scalia, Arango & Singman, 1985); but direct counts consistently show little change after regeneration in goldfish (Murray, Sharma & Edwards, 1982; Burmeister, Perry & Grafstein, 1983; Cook & Rankin, 1986).

New ganglion cells are also formed continuously, though at low rates under normal laboratory conditions (P. A. Raymond and E. C. C. Rankin, both cited in Stuermer, 1986). However, though new axons might conceivably show more brachial selectivity than regenerated ones and might not be distinguishable from them by anterograde tracing, their retrogradely labelled somata would be confined to the extreme retinal margin and thus easily distinguished. In the event, we saw no increase in selectivity at the margin of late-regenerated retinae and, in any case, our sampling protocol, in which only complete squares were counted, tended to exclude this region. New axon growth cannot, therefore, explain our results.

The explanation of brachial refinement that we favour is also the one most consistent with previous evidence to be discussed below. We assume that regenerating axons branch, many of them sending collaterals into both brachia, and that many branches later degenerate or are withdrawn, being lost preferentially from the inappropriate brachium. We hypothesize that the loss of redundant branches may be triggered by the prior establishment, by a sibling branch from the same ganglion cell, of a group of tectal synapses that have been stabilized by interaction with retinal neighbours. Collaterals entering the appropriate brachium, which leads directly towards retinotopic termination sites, would have a greater chance of establishing such synapses quickly, and therefore a greater chance of surviving, than misrouted collaterals; though a minority of the latter would still be expected to survive. Such a process might conveniently be termed ‘sibling rivalry’.

Relation to previous studies

Our results agree with many others in showing a high degree of retinotopic segregation between the brachia of normal animals, which is reduced after regeneration (see Introduction). Stuermer & Easter (1984) reported that the proportions of axons in the inappropriate brachia of normal and regenerated goldfish were in the region of 1 % and 20 % respectively. Our retrograde results confirm their estimate for normal fish. However, both our anterograde and retrograde results show appreciable changes in the brachial distribution during regeneration, whereas Stuermer & Easter seemed to imply that it was constant. Fig. 11A shows that an individual fish may, in fact, show a ‘late-regenerated’ pattern as early as 30 days after nerve cut. Thus, of the five regenerating fish studied by Stuermer & Easter at 18, 30, 60 (two fish) and 90 days, only the 18-day fish would necessarily be expected to have shown a clear ‘earlyregenerated’ pattern, even if it had been analysed by retrograde tracing. Moreover, their anterograde method was presumably subject to the same limitations as ours. The possibility of refinement by collateral loss was not discussed.

Murray (1982) made axon counts, at different times during regeneration, in the optic nerve and tract. She found that the number of axon profiles in the optic nerve increased fourfold early in regeneration, returning towards normal later, and concluded that the regenerating axons had branched to form multiple collaterals, many of which were subsequently lost. An even greater proliferation was seen in the tectum (Murray & Edwards, 1982). Widespread axonal branching has also apparently been seen directly, by light microscopy, during regeneration in the optic tectum of the newt (Fujisawa, Tani, Watanabe & Ibata, 1982) and fish (Schmidt, Buzzard & Turcotte, 1984) though the problem of distinguishing between true axonal branching and defasciculation, considered above in relation to our own anterograde results, applies to these as well.

Axon collateralization is not, of course, restricted to regenerating systems. Braekevelt, Beazley, Dunlop & Darby (1986) found three times as many optic nerve axons as ganglion cells at one stage in the postnatal development of the small wallaby Setonix brachyurus. Embryonic and neonatal visual cortical and callosal systems in mammals also have initial excesses of axons, forming diffuse transient projections that are later removed or refined by selective collateral loss or neuronal death (Ivy, Akers & Killackey, 1979; Innocenti, 1981; O’Leary, Stanfield & Cowan, 1981 ; O’Leary & Stanfield, 1986). O’Leary & Stanfield have suggested that the formation and selective elimination of collaterals may be an important general strategy adopted by developing nervous systems to reduce the genetic information required for the generation of complex connection patterns.

Implications

Although axonal guidance is needed, at some level, to fix the polarity of the retinotectal projection (Willshaw & Malsburg, 1976), it would appear to be given to regenerating axons only after they have passed the brachial bifurcation. The evidence given here that these axons show no initial preference for their usual brachium suggests that brachial guidance may not, as originally proposed by Attardi & Sperry (1963) and seconded by Fawcett & Gaze (1982), contribute directly to the re-establishment of a retinotopic tectal map. Indeed, the converse relationship now seems as likely: retinotopic refinement of regenerated tectal terminal arbors may be a prerequisite for retinotopic refinement of their axons in the brachial tract. This would account just as well for the correlation between pathway patterning and the ordering of the retinotectal map noted by Gaze & Fawcett (1983).

The normal brachial distribution may even be refined in this way during development. Though this possibility seems never to have been discussed in print there is, to our knowledge, no convincing counter evidence. In late larval or adult fish and frogs, branched axons would normally be confined to the newly formed ganglion cells at the retinal margin (Johns, 1977) and would comprise only a very small (and unmyelinated) proportion of the total axon population, which might readily be overlooked (Scholes, 1979).

In fact, Maggs & Scholes (1986) have proposed, on the basis of evidence from a cichlid fish, that developing optic axons tend to branch at the point where, on entering the optic tract, they encounter a profound change in their glial environment. A developmental scenario involving (i) extensive collateral formation, (ii) a tendency for divergent sibling branches to fasciculate independently of each other with surviving branches of older axons, and (iii) a process of selective collateral loss on the basis of sibling rivalry for retinotopic termination sites, might economically account for the complex interweaving of axons with different retinal origins that has been found, in several species, at sites between the optic chiasm and brachial bifurcation (fish: Scholes, 1979; Bunt, 1982; Springer & Mednick, 1986b; amphibia: see Taylor, 1987). Though the details of such rearrangements tend to be species specific, the basic outcome is always the same: most mature axons enter the tectum through whichever brachium originally offered the shortest path around the tectal margin to their initial termination site (fish: Cook, Rankin & Stevens, 1983; frog: Taylor & Gaze, 1985; newt: Fujisawa, Watanabe, Tani & Ibata, 1981).

This work was supported by the Medical Research Council of Great Britain, through a Studentship to D.L.B. and a Project Grant to J.E.C.

Attardi
,
D. G.
&
Sperry
,
R. W.
(
1963
).
Preferential selection of central pathways by regenerating optic fibers
.
Expl Neurol
.
7
,
46
64
.
Becker
,
D. L.
&
Cook
,
J. E.
(
1986
).
Refinement of brachial selection by regenerating axons in the goldfish optic tract may be due to collateral loss
.
J. Physiol
.
382
,
175P
.
Braekevelt
,
C. R.
,
Beazley
,
L. D.
,
Dunlop
,
S. A.
&
Darby
,
J. E.
(
1986
).
Numbers of axons in the optic nerve and of retinal ganglion cells during development in the marsupial Setonix brachyurus
.
Devi Brain Res
.
25
,
117
125
.
Bunt
,
S. M.
(
1982
).
Retinotopic and temporal organization of the optic nerve and tracts in the adult goldfish
.
J. comp. Neurol
.
206
,
209
226
.
Burmeister
,
D. W.
,
Perry
,
G. W.
&
Grafstein
,
B.
(
1983
).
Target regulation of the cell body reaction during regeneration of goldfish retinal ganglion cells
.
Soc. Neurosci. Abstr
.
9
,
694
.
Cook
,
J. E.
(
1983a
).
Tectal paths of regenerated optic axons in the goldfish: evidence from retrograde labelling with horseradish peroxidase
.
Expl Brain Res
.
51
,
433
442
.
Cook
,
J. E.
(
1983b
).
Growth patterns of goldfish optic axons reveal boundaries between retinal quadrants
.
J. Physiol
.
334
,
69
70P
.
Cook
,
J. E.
(
1987
).
A sharp retinal image increases the precision of the goldfish retinotectal projection during optic nerve regeneration in stroboscopic light
.
Expl Brain Res. (in press)
.
Cook
,
J. E.
,
Pilgrim
,
A. J.
&
Horder
,
T. J.
(
1983
).
Consequences of misrouting goldfish optic axons
.
Expl Neurol
.
79
,
830
844
.
Cook
,
J. E.
&
Rankin
,
E. C. C.
(
1986
).
Impaired refinement of the regenerated goldfish retinotectal projection in stroboscopic light: a quantitative WGA-HRP study
.
Expl Brain Res
.
63
,
421
430
.
Cook
,
J. E.
,
Rankin
,
E. C. C.
&
Stevens
,
H. P.
(
1983
).
A pattern of optic axons in the normal goldfish tectum consistent with the caudal migration of optic terminals during development
.
Expl Brain Res
.
52
,
147
151
.
Cowan
,
W. M.
&
Hunt
,
R. K.
(
1985
). The development of the retinotectal projection: an overview.
In Molecular Basis of Neural Development
(ed.
G. M.
Edelman
,
W. E.
Gall
&
W. M.
Cowan
), pp.
389
428
.
New York
:
Wiley
.
Fawcett
,
J. W.
&
Gaze
,
R. M.
(
1982
).
The retinotectal fibre pathways from normal and compound eyes in Xenopus
.
J. Embryol. exp. Morph
.
72
,
19
37
.
Fujisawa
,
H.
,
Tani
,
N.
,
Watanabe
,
K.
&
Ibata
,
Y.
(
1982
).
Branching of regenerating retinal axons and preferential selection of appropriate branches for specific neuronal connection in the newt
.
Devi Biol
.
90
,
43
57
.
Fujisawa
,
H.
,
Watanabe
,
K.
,
Tani
,
N.
&
Ibata
,
Y.
(
1981
).
Retinotopic analysis of fiber pathways in amphibians. I. The adult newt Cynops pyrrhogaster
.
Brain Res
.
206
,
9
20
.
Gaze
,
R. M.
&
Fawcett
,
J. W.
(
1983
).
Pathways of Xenopus optic fibres regenerating from normal and compound eyes under various conditions
.
J. Embryol. exp. Morph
.
73
,
17
38
.
Grafstein
,
B.
&
Murray
,
M.
(
1969
).
Transport of protein in goldfish optic nerve during regeneration
.
Expl Neurol
.
25
,
494
508
.
Hanker
,
J. S.
,
Yates
,
P. E.
,
Metz
,
C. B.
&
Rustioni
,
A.
(
1977
).
A new specific, sensitive and non-carcinogenic reagent for the demonstration of horseradish peroxidase
.
Histochem. J
.
9
,
789
792
.
Horder
,
T. J.
(
1974
).
Changes of fibre pathways in the goldfish optic tract following regeneration
.
Brain Res
.
72
,
41
52
.
Humphrey
,
M. F.
&
Beazley
,
L. D.
(
1985
).
Retinal ganglion cell death during optic nerve regeneration in the frog Hyla moorei
.
J. comp. Neurol
.
236
,
382
402
.
Innocenti
,
G. M.
(
1981
).
Growth and reshaping of axons in the establishment of visual callosal connections
.
Science
212
,
824
827
.
Ivy
,
G. O.
,
Akers
,
R. M.
&
Killackey
,
H. P.
(
1979
).
Differential distribution of callosal projection neurons in the neonatal and adult rat
.
Brain Res
.
173
,
532
537
.
Johns
,
P. R.
(
1977
).
Growth of the adult goldfish eye. III. Source of the new retinal cells
.
J. comp. Neurol
.
176
,
343
358
.
Maggs
,
A.
&
Scholes
,
J. H.
(
1986
).
Glial domains and nerve fiber patterns in the fish retinotectal pathway
.
J. Neurosci
.
6
,
424
438
.
Meyer
,
R. L.
(
1980
).
Mapping the normal and regenerating retinotectal projection of goldfish with autoradiographic methods
.
J. comp. Neurol
.
189
,
273
289
.
Meyer
,
R. L.
(
1983
).
Tetrodotoxin inhibits the formation of refined retinotopography in goldfish
.
Devi Brain Res
.
6
,
293
298
.
Murray
,
M.
(
1982
).
A quantitative study of regenerative sprouting by optic axons in goldfish
.
J. comp. Neurol
.
209
,
352
362
.
Murray
,
M.
&
Edwards
,
M. A.
(
1982
).
A quantitative study of the reinnervation of the goldfish optic tectum following optic nerve crush
.
J. comp. Neurol
.
209
,
363
373
.
Murray
,
M.
&
Grafstein
,
B.
(
1969
).
Changes in the morphology and amino acid incorporation of regenerating goldfish optic neurons
.
Expl Neurol
.
23
,
544
560
.
Murray
,
M.
,
Sharma
,
S. C.
&
Edwards
,
M. A.
(
1982
).
Target regulation of synaptic number in the compressed retinotectal projection of goldfish
.
J. comp. Neurol
.
209
,
374
385
.
O’leary
,
D. D. M.
&
Stanfield
,
B. B.
(
1986
).
A transient pyramidal tract projection from the visual cortex in the hamster and its removal by selective collateral elimination
.
Devi Brain Res
.
27
,
87
99
.
O’leary
,
D. D. M.
,
Stanfield
,
B. B.
&
Cowan
,
W. M.
(
1981
).
Evidence that the early postnatal restriction of the cells of origin of the callosal projection is due to the elimination of axonal collaterals rather than to the death of neurons
.
Devi Brain Res
.
1
,
607
617
.
Rankin
,
E. C. C.
&
Cook
,
J. E.
(
1986
).
Topographic refinement of the regenerating retinotectal projection of the goldfish in standard laboratory conditions: a quantitative WGA-HRP study
.
Expl Brain Res
.
63
,
409
420
.
Scalia
,
F.
,
Arango
,
V.
&
Singman
,
E. L.
(
1985
).
Loss and displacement of ganglion cells after optic nerve regeneration in adult Rana pipiens
.
Brain Res
.
344
,
267
280
.
Schmidt
,
J. T.
,
Buzzard
,
M. J.
&
Turcotte
,
J.
(
1984
).
Morphology of regenerated optic arbors in goldfish tectum
.
Soc. Neurosci. Abstr
.
10
,
667
.
Schmidt
,
J. T.
&
Edwards
,
D. L.
(
1983
).
Activity sharpens the map during the regeneration of the retinotectal projection in goldfish
.
Brain Res
.
269
,
29
39
.
Schmidt
,
J. T.
&
Eisele
,
L. E.
(
1985
).
Stroboscopic illumination and dark-rearing block the sharpening of the regenerated retinotectal map in goldfish
.
Neuroscience
14
,
535
546
.
Schmidt
,
J. T.
,
Edwards
,
D. L.
&
Stuermer
,
C.
(
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
.
269
,
15
27
.
Scholes
,
J. H.
(
1979
).
Nerve fibre topography in the retinal projection to the tectum
.
Nature, Lond
.
278
,
620
624
.
Springer
,
A. D.
&
Mednick
,
A. S.
(
1985
).
Topography of the goldfish optic tracts: implications for the chronological clustering model
.
J. comp. Neurol
.
239
,
108
116
.
Springer
,
A. D.
&
Mednick
,
A. S.
(
1986a
).
Relationship of ocular pigmentation to the boundaries of dorsal and ventral retina in a nonmammalian vertebrate
.
J. comp. Neurol
.
245
,
74
82
.
Springer
,
A. D.
&
Mednick
,
A. S.
(
1986b
).
Simple and complex retinal ganglion cell axonal rearrangements at the optic chiasm
.
J. comp. Neurol
.
247
,
233
245
.
Straznicky
,
C.
,
Gaze
,
R. M.
&
Horder
,
T. J.
(
1979
).
Selection of appropriate medial branch of the optic tract by fibres of ventral retinal origin during development and in regeneration: an autoradiographic study in Xenopus
.
J. Embryol. exp. Morph
.
50
,
253
267
.
Stuermer
,
C. A. O.
(
1986
).
Pathways of regenerated retinotectal axons in goldfish. I. Optic nerve, tract and tectal fascicle layer
.
J. Embryol. exp. Morph
.
93
,
1
28
.
Stuermer
,
C. A. O.
&
Easter
,
S. S.
Jr
(
1984
).
A comparison of the normal and regenerated retinotectal pathways of goldfish
.
J. comp. Neurol
.
223
,
57
76
.
Taylor
,
J. S. H.
(
1987
).
Fibre organization and reorganization in the retinotectal projection of Xenopus
.
Development
99
,
393
410
.
Taylor
,
J. S. H.
&
Gaze
,
R. M.
(
1985
).
The effects of the fibre environment on the paths taken by regenerating optic nerve fibres in Xenopus
.
J. Embryol. exp. Morph
.
89
,
383
401
.
Tosney
,
K. W.
&
Landmesser
,
L. T.
(
1985
).
Growth cone morphology and trajectory in the lumbosacral region of the chick embryo
.
J. Neurosci
.
5
,
2345
2358
.
Willshaw
,
D. J.
&
Malsburg
,
C. Von Der
(
1976
).
How patterned neural connections can be set up by selforganization
.
Proc. R. Soc. Lond. B
194
,
431
445
.