Flujograma

along the rostrocaudal tectal axis (Figure 8). Axons migrate further from the rostral (older) fascicles than from the caudal (younger) fascicles, where in fact the fibres can be seen to turn rostrally, a point which can be inferred from the graph also.

5:4: Assessment of the map during regeneration under standard laboratory conditions using retrograde transport of WGA-HRP

An anatomical assay of the map refinement which is known to occur after optic nerve regeneration has been presented. Retrograde transport of iontophoretically injected wheatgerm agglutinin conjugated to HRP (WGA- HRP) from two separate tectal sites resulted in patterns of labelled retinal ganglion cells (RGC) which became more compact and retinotopic as regeneration progressed. Initially the cells were widely distributed with little retinotopic bias. They were later found in the “correct” retinal quadrant and later a cluster became discernible. Finally, the cluster was compact and the number of scattered cells was few. However, the clusters of cells were not as compact as in normal animals. Occasionally, later in regeneration, after single injections of WGA-HRP, double clusters were seen in the retinotopic quadrant.

The wide distribution of very few labelled cells seen in some examples early in regeneration could have been due to an increase in the diffusion of the WGA-HRP from the injection site, at a time when there were few axon terminals available to take up the tracer. Staining of the injection sites at 22 days of regeneration in a separate group of 3 fish compared to 3 normal controls showed no evidence of a change in the injection site to support this explanation for the cell distributions seen early in regeneration. WGA-HRP, unlike unbound HRP, does not diffuse widely from injection sites, binding strongly to membranes near the pipette tip (Mesulam, 1982). In those cases where there were many labelled cells, despite the small injection size, it is possible that there was exuberant proliferation of early axon terminal branches with much overlapping, therefore many terminal structures were available to take up the tracer. The reason for the great variation in the number of labelled cells early in regeneration is not known.

It is known that in regeneration the first axons to reach the tectum do so at about day 14 after optic nerve cut, at 250C (Stuermer and Easter, 1984b). They enter the tectum at the rostral pole between the two brachia. Axons which are destined to synapse caudally will also pass through this

zone. Therefore the wide scatter of cells could be a reflection of the fact that cells from all retinal quadrants pass through the area early in regeneration and their terminals are available to take up the tracer.

The ventral bias which was seen in some retinae early in regeneration could have reflected a bias in the route of regenerating axons, such that most axons from the ventral retina pass through the medial brachium and are therefore more likely to be labelled from the rostromedial injection site. However, evidence from the area indices at these stages of

regeneration does not support this hypothesis. The area indices from

caudal injection sites at 24 days (6 days later) and at 28 days were as great a those from rostral sites, indicating that no such bias existed.

It is known that the paths of regenerating RGC axons remain disordered after the map is refined in the goldfish (Horder, 1974; Cook, 1983; Stuermer and Easter, 1984b). Even if there is no early bias in routes of axons, as shown in later experiments (Becker and Cook, 1987; 1988), the early diffuse WGA-HRP labelling in some retinae and the wide variety of numbers of axons could reflect a period of ‘trial and error’ synaptogenesis. These ‘abnormal’ routes would not be seen at later stages because this tracer detects terminals only. The period of trial and error may form part of the self-ordering that is thought to occur among the terminals of regenerating axons (Willshaw and von der Malsburg, 1976). The self-ordering would allow the map to become more refined, and this would be reflected in a fall in the value of the area index, the clusters becoming more compact. The regenerating system may rely on self-ordering more than the developing system does: the choices of sites for synapsis for the newly-arriving axons in the latter are in theory more limited. In the regenerating system, the axons may branch exuberantly thereby increasing the chances of finding a near neighbour on the tectum. Axon counts in the optic nerve, tract and tectum of animals after optic nerve cut compared to normal animals would suggest that there is an increase in axon branch number, as do direct observations of HRP-filled axons (Murray, 1976; Murray, 1982; Murray and Edwards, 1982; Fujisawa, Tani, Watanabe and Ibata, 1982).

It Is possible to over-interpret the findings from the WGA-HRP studies, especially from the earlier stages of regeneration. The nature of the terminal structures at these stages has not been well-characterised, nor has their ability or inability to take up WGA-HRP. Therefore the possibility of a trial and error phase of synaptogenesis during early regeneration must remain a supposition.

Later in regeneration, clusters of labelled cells appeared in retinotopic sites and there was a gradual elimination of the scattered cells

surrounding the clusters. This confirms the findings of other mapping

methods, electrophysiology (Humphrey and Beazley, 1982; Northmore and Masino, 1984) and autoradiography (Meyer, 1980), that the map refinement seen as regeneration progresses is a consequence of the elimination of non-retinotopic terminal structures from a tectal site. Such elimination of non-retinotopic terminal structures and the maturation of retinotopic terminal arborisations would result in the labelling of more compact clusters of RGC after iontophoretic injection of WGA-HRP, which is in fact what is observed.

There was no further refinement of the map after 70 days according to this assay. However, when the clusters are so compact, a few scattered cells will increase the area index considerably. In order to investigate the possibility of any further refinement, much larger numbers of animals would have to be studied. However, for practical purposes, regeneration beyond 70 days in these standard laboratory conditions did not lead to an appreciable increase in map refinement.

Another feature of late regeneration was the appearance of compact but double clusters in 8 of the retinae, 5 after caudal and 3 after rostral injections. They were in the retinotopic quadrant, and in those following caudal injections they had a radial orientation while those after rostral injections all had different orientations. It is possible that these distributions arose because of completely overlapping terminal structures, therefore one injection resulted in the labelling of both sets. If so, it is likely that such overlap would have been extensive; it is unlikely that one injection of tracer detected the only area of overlap in a projection. Was this period of possible

overlap a stage in the process of continuing map refinement? The fact that they were seen in the mid- to late stages of regeneration, including at 524 days, argues against this idea. In addition, the clusters, though double, were relatively compact with only a few scattered cells, again suggesting that much of the retraction of non-retinotopic terminals had taken place. An alternative possibility is that the terminal arbors from the two clusters were segregated into adjoining fragments of a piece-wise continuous map, so that one injection of WGA-HRP was taken up by both ‘fragments’ of the map. It is interesting that these double clusters have only been seen in the retinae of regenerating systems. It is possible that when the mechanisms by which axons achieve retinotopy depend more heavily on self-ordering, as postulated above for regeneration, the incidence of these double clusters increases as does the piece-wise nature of the map. Why were these double clusters not seen more often? If they were due to an overlapping of terminal structures, it suggests that this was a relatively rare event. As such a projection would be less efficient, because the tectal cells would be doubly innervated over all, a mechanism to avoid overlapping would be advantageous. If the double clusters were due to formation of a piece-wise continuous map, their rarity in these studies might be due to greater areas of separation between the adjacent fragments in some animals, or to the small size or immaturity of terminal structures, which may also be small in number, in an adjacent fragment. The latter could occur if there were an imbalance in the sizes of the fragments. Indeed it is possible that the sizes of the fragments were continuously changing as the self-ordering mechanism continued to act.

The piece-wise continuous map in the regenerated retinotectal projection in the goldfish was concurrently described by Meyer, Sakurai and

Schauwecker (1985). Using anterograde W G A -H R P tracing, they

discovered that the regenerated map was usually in this form. The

regenerated map resolved into patches of reaction product, and one injection yielded more than one patch. The patches were of variable size. In contrast, in normal goldfish, one injection yielded one patch. The time course of the appearance of the double clusters of retrogradely labelled RGC at 56-524 days of regeneration corresponds well with the appearance

of the patches at 59-158 days in their anterograde tracer study. Although these authors noted that the sum of the areas of the individual clumps was not different from the area of a normal patch, it is not clear how much the initial branching of axons which occurs after optic nerve section contributed to the separate patches. It appears that the resolution of this retrograde tracer study using WGA-HRP is not great enough to detect the fragments in most cases, possibly because the distances between the patches are usually too great.

In four cases, two rostral and two caudal, the two clusters were approximately equal in size and in packing density of labelled cells but in the remaining four cases one cluster was distinctly larger or denser than the other. As is the case for the basis of the double clusters, the basis for their different sizes in some instances is not known. It is possible that this finding represents a transient situation in a continuously changing map, with the process of self-ordering altering the sizes of map fragments all the time.

The 5 cluster pairs which followed caudal injections had a radial