/. Embryol. exp. Morph. 90, 233-250 (1985) 233Printed in Great Britain © The Company of Biologists Limited 1985

Factors guiding regenerating retinotectal fibres in the

frog Xenopus laevis

J. W. FAWCETTDevelopmental Neurobiology Laboratory, The Salk Institute for Biological Studiesand the Clayton Foundation for Research, California Division, PO Box 85800,San Diego, California 92138, U.S.A.

SUMMARYI have examined the pathways of retinotectal fibres regenerating back to the contralateral

tectum, and also to innervated and 'virgin' ipsilateral tecta in postmetamorphic Xenopus. Thefibres were visualized by HRP labelling of either the whole optic nerve, or a selected quadrant ofthe retina.

Most fibres grow into either the ipsilateral or contralateral optic tract, although a smallproportion go down the outside of the contralateral optic nerve. In the tracts, many fibres growsuperficially, close beneath the pia, but a variable proportion runs more deeply. Axonal growthis not, therefore, restricted absolutely to the subpial region in the postmetamorphic Xenopusbrain.

Fibres growing onto the contralateral, or a 'virgin' tectum mostly grow straight onto therostral margin of the tectal lobe, without growing around its margin in the form of a medial orlateral brachium. Most of these fibres grow through the deeper part of the tectal layer whichnormally contains optic neuropil, but a proportion of them grow immediately deep to the pia.Fibres regenerating to an innervated ipsilateral tectum mostly enter either the medial or lateralbrachium of the optic tract, and only leave this close to their site of termination. In the brachiathe fibres run superficially under the pia, but when they leave the brachia they mostly runthrough the deeper retinorecipient layers. These observations provide further evidence thatingrowing optic fibres have their pathways influenced by the axons which have preceded them.

INTRODUCTION

The retinotectal projection of fish and frogs differs from that of birds andmammals in that both the retina and the tectum continue to grow almostcontinuously throughout the life of individual animals, new neurons being addedto the periphery of the retina, and to the caudal margin of the tectum (Gaze,Keating, Ostberg & Chung, 1979; Beach & Jacobson, 1979; Johns, 1977). Anotherimportant and possibly related difference, is that the axons of the retinal ganglioncells in fish and frogs are able to regenerate back to their original target within thetectum, a capacity that is completely lacking in amniotes. This ability to regenerateis not only very interesting in itself, but has also proven to be a valuable tool for thestudy of the factors that guide optic fibres to their targets.

Key words: retina, optic tectum, frog, regeneration, axonal guidance, Xenopus laevis.

234 J. W. FAWCETT

When the optic nerve of the frog Xenopus laevis is cut near the chiasm, opticfibres regenerate back not just to the contralateral tectal lobe, which theyoriginally innervated, but also to the ipsilateral tectum, which is innervated by theother eye. In tadpoles these regenerating fibres grow immediately beneath the pia(Gaze & Grant, 1978), as do newly growing fibres in the course of normaldevelopment (Fawcett, Taylor, Gaze, Grant & Hirst, 1984; Reh, Pitts &Constantine-Paton, 1983; Steedman, Stirling & Gaze, 1979). These observationshave led to the idea that this subpial layer may possess special properties whichmake it uniquely hospitable to growing axons.

The ability of regenerating fibres to innervate the ipsilateral tectum has beenused in a number of recent studies as a means of examining the interactions thatoccur between regenerating fibres, and in situ fibres from the opposite eye, whoseoptic nerve has not been cut (Gaze & Straznicky, 1980; Straznicky & Tay, 1981;Straznicky, Tay & Glastonbury, 1980). In the same way, it has been possible tostudy the pattern of innervation by regenerating optic fibres of a so-called 'virgin'tectum; that is a tectum which has been deprived of optic input by removal of itscontralateral eye before this sends fibres to the brain (Feldman, Gaze & Keating,1971; Straznicky & Gaze, 1982; Gaze & Fawcett, 1983). Collectively, theseexperiments have provided information not only about the behaviour ofregenerating fibres, but also provided clues as to the behaviour of fibres duringnormal development. One can draw inferences about normal development fromsuch experiments, because the frog brain is a continuously developing structure:consequently the fibres regenerating to, for instance, an innervated ipsilateraltectum, grow through much the same environment as those newly-formed fibreswhich are growing out from the eye with the intact optic nerve.

In one such experiment Gaze & Fawcett (1983) examined the pathways taken byfibres regenerating from both normal and 'compound' eyes in Xenopus. Both thepresent, and our previous study are in agreement with many other reports(Fawcett & Gaze, 1981; Gaze & Grant, 1978; Cook, 1983; Stuermer & Easter,1984; Udin, 1978; Meyer, 1980), in showing that fibres regenerating back to theiroriginal target (the contralateral tectum) follow courses that are very differentfrom those which they took during their original development. The regeneratingfibres tend to grow directly over the tectum, rather than around its peripheralmargin, and cross over one another frequently rather than growing parallel. Fibresgrowing to a 'virgin' tectum showed a similar, but more pronounced behaviour.Perhaps the most interesting finding to come from this study, however, was thatmany of the fibres regenerating to an innervated ipsilateral tectum seemed tomimic the behaviour of the most recently arrived fibres from the normallydeveloping contralateral eye: that is to say they grew around the periphery of thetectum in either the medial or lateral brachium, and only entered onto the tectumat a point on its periphery close to their eventual termination. This result hasrecently been confirmed in a study by Taylor & Gaze (personal communication),in which the behaviour of a selected portion of the fibres regenerating from anormal eye was examined. Both these studies were done using whole-mount

Regenerating frog retinotectal fibres 235

preparations of the Xenopus brain, and although these allow a very clear overallview of the fibre pathway, one cannot see which layers of the tectum or tract theregenerating fibres are running in. Consequently, one can say little about theenvironment which the optic fibres encounter as they regenerate. In theexperiments reported in this paper, I have essentially repeated these earlierexperiments, but have used sectioned material. This work has provided additionalevidence that regenerating optic fibres actively follow optic fibres which havealready reached the tectum, and that regenerating fibres are able to grow throughthe substance of the frog brain, not just in the layer immediately beneath the pia.

METHODSXenopus laevis frogs, obtained from NASCO, Fort Atkinson, Wisconsin, were used for all the

experiments. They were kept in 10 % Holtfreters solution, and fed on Tubifex worms.

Embryonic eye removalsPairs of adult Xenopus were induced to breed by chorionic gonadotrophin injections. Fertile

eggs were collected into 10 % Niu Twitty solution and demembranated. When the embryosreached Nieuwkoop & Faber (1956) stage 28, they were hatched, and then transferred to fullstrength Niu Twitty, containing 0-005% MS222 (Tricaine methane sulphonate, Sigma). Theembryos were laid in trenches in wax, and one eye was removed using electrolytically sharpenedtungsten needles. After half an hour the animals were transferred to 20 % Niu Twitty, and then,after 24 h, to 10 % Holtfreters solution, in which rearing was continued. The tadpoles were fedon Nasco frog brittle.

Optic nerve sectionAnimals which had metamorphosed about 2 months previously were anaesthetized with

MS222 (Tricaine methane sulphonate, Sigma). They were pinned down, belly up, and theirlower jaws were opened wide. A cut was made with fine spring scissors down the midline of thecranium, which then was prised open to reveal the ventral surface of the diencephalon, and theoptic nerves. The left optic nerve was cut between cranium and chiasm. The cut edges of thecranium were then pushed together, and the animal put in oxygenated 50 % Holtfreters solutionto recover from the anaesthetic.

HRPfillsWhole optic nerve fillsHRP (Sigma type VI) was dissolved in a small drop of water (approx. 1 [A) until the solution

became syrupy. The water was then allowed to evaporate, leaving a hard deposit of HRP. Thiswas broken up into small blocks, about 0-5 to 1 mm across.

The frogs were anaesthetized in MS222, then the eye on the side to be labelled was removedwith fine scissors, after cutting through the optic nerve. When bleeding had ceased one or twoblocks of HRP were put on the optic nerve stump. The animal was then wrapped in tissue, andlaid in 50% Holtfreters solution, with the labelled orbit, out of the water. After 20min theanimals were fully immersed, and bubbled with oxygen.

Partial eye fills with HRPHRP (Sigma type VI) was dissolved in a small drop of water until it went brown. As the water

evaporated, a sticky deposit of HRP formed at the edge of the drop, and some of this was used tocoat the tip of a sharpened tungsten needle.

236 J. W. FAWCETT

The animal was anaesthetized with MS222, and the skin was parted over the quadrant of theeye to be labelled. A hole was made in the eye with a needle at the point to be labelled, then theneedle was rotated round so as to contact the retina, and the retina was lesioned by pressingagainst the needle with a pair of forceps from outside the eye. The HRP-coated needle was nextinserted into the eye, and left adjacent to the retinal lesion until the HRP had dissolved. Theanimals were allowed to recover in oxygenated 50 % Holtfreters solution.

HRP histochemistry48 h after HRP labelling, the animals were anaesthetized, and killed by perfusion through the

heart with saline, followed by 2-5 % glutaraldehyde in 0-1 M-phosphate buffer. The brains wereremoved, fixed for a further hour in 2-5 % glutaraldehyde in buffer, rinsed in 10 % sucrose in0-1 M-phosphate buffer, then put into gelatin albumen overnight (90 g Fisher Albumen no.A388+60 g Sigma albumen no. A5253 in 200 ml 01 % PO4 buffer, and 1-5 g gelatin in 100 ml PO4buffer, mixed together, then 60 g sucrose added). The brains were then embedded by setting thegelatin albumen by adding 1 part in 10 of 25 % glutaraldehyde. The blocks were sectioned on afreezing microtome, and the sections mounted on subbed slides, which were reacted usingAdam's cobalt chloride-diamino benzidine technique (Adams, 1977), and counterstained withneutral red.

RESULTS

Control non-regenerates

In order to make the observations on the regenerates which follow compre-hensible, and to report a few new observations in normal animals, I will brieflydescribe the appearance of the control animals in which either the whole opticnerve or part of the retina was labelled.

Optic tract

As described by previous authors (Levine, 1980; Scalia & Coleman, 1974;Steedman et al. 1979; Scalia & Fite, 1974) the optic tract just after the chiasma is a

Table 1. Numbers and types of operations

Regeneration time Retinal source of filled axons Number of cases

Animals with left optic nerve cuts after metamorphosis10 days Whole eye 314 days Whole eye 521 days Whole eye 51 month Whole eye 121£ months Whole eye 54 months Whole eye 717 days Ventral 41 month Nasal 42 months Ventral 44 months Nasal 5

Animals with embryonic eye removal, and postmetamorphic optic nerve cuts5 weeks Ventral 35 weeks Dorsal 35 weeks Nasal 3

Regenerating frog retinotectal fibres 237

very deep structure, with some fibres, probably the oldest (Steedman et al. 1979';Fawcett et al. 1984) turning dorsally as they leave the optic nerve, to run quite closeto the underside of the ventricle (see Fig. 1). As the fibres approach the tectum,the tract widens in the anteroposterior dimension, and becomes shallower again.

Optic tectum

The optic neuropil of the Xenopus tectum is not layered in the precise andobvious fashion seen in Rana (Scalia & Coleman, 1974; Potter, 1969; Scalia, 1973;Szekely & Lazar, 1976), in fact, using [3H]proline as an anterograde tracer, Levine(1980) reported no layering at all. However, in my HRP-labelled material, anexample of which is illustrated in Fig. 2, and also in the cobalt-filled preparationsdescribed by Steedman et al. (1979), a rather imprecise layering is seen in theneuropil. Immediately under the pia there is a layer which contains no HRP-reaction product, and, below this is a thick layer of neuropil. This layer of neuropilis divided in half by a thin layer of less-dense label. The deeper part of the deeperof the two sublayers contains many thick fibres which run tortuous courses.

Most of the fibres, enter the tectum through well-defined medial or lateralbrachia, which bound the tectal periphery as it then is, as described previously.This is most clearly seen in the brains of animals in which peripheral nasal retinawas filled with HRP. In these cases the fibres travel in either the medial or lateralbrachium until they reach the caudal part of the tectum, where they branch off,and grow across the tectum to their termination site. The fibres in the brachia atthe tectal periphery run close to the pia, but as fibres leave the brachia, they pass

Fig. 1. A transverse section through the optic chiasma of a normal frog. Fibres fromthe left eye (to the right of the picture) have been filled with HRP. Dorsal is upwards,ventral is downwards. The optic tract at this point is very wide in the dorsoventraldimension. The most dorsal labelled fibres are probably the oldest axons which arrivefrom the centre of the retina. Bar = 250 jum.

238 J. W. FAWCETT

into the layer of coarse fibres described above, in the deeper part of the neuropil,and remain in this layer until they terminate (see Fig. 3).

For the purposes of this study, it is important only to distinguish the neuropillayers, and this layer of fibres of passage. Rather, therefore, than speculating onexactly which tectal layers in Xenopus correspond to the various schemes oflayering nomenclature in other amphibia, I will just refer to the neuropil layers,and to the deep fibre layer, which lies in the deepest of the neuropil layers.

RegeneratesOptic nerves

In animals in which the whole optic nerve was filled with HRP, reaction productwas seen throughout the nerve, both central and peripheral to the cut. However,there was generally a decrease in the density of the reaction product towards thecentre of the nerve, and there were also some quite large unlabelled areas, whichprobably represent glial cell bodies. In animals which received partial andperipheral fills of the retina, the filled fibres were more clearly restricted to theperiphery of the nerve.

A variable proportion of the regenerating fibres did not grow into the optictracts, but instead grew down the opposite optic nerve, as previously reported inRana by Bohn & Stelzner (1981). These axons were grouped in a narrow band allaround the very periphery of the nerve, and seldom penetrated into its centre (seeFig. 4A,B). In animals with partial retinal HRP fills, fibres were still seen only inthe periphery of the contralateral nerve, and were not restricted to those parts ofthe nerve that are occupied by fibres from the corresponding part of thecontralateral retina (Fawcett, 1981).

Optic tract

The regenerating fibres grew in both optic tracts, in most cases apparently inequal numbers (see Gaze & Straznicky, 1980; Gaze & Grant, 1978). In about halfthe cases, all the regenerated fibres were confined to the layer immediatelyunderlying the pia, as reported in Xenopus tadpoles by Gaze & Grant. However,in the remaining cases, a proportion of the regenerated fibres were found moredeeply in the optic tract (see Fig. 5). This was seen both in animals filled with HRPand killed shortly after optic nerve cut (as little as 14 days) and in animals killed upto 4-5 months after the nerve was cut. Regenerating fibres from all parts of theretina are apparently able to grow deep within the optic tract, I saw no particulartendency for regenerating fibres in the 11 cases with ventral retinal HRP fills to runmore superficially than fibres from cases with whole eye fills. Equally out of thesmall number of dorsally filled cases, two had both deep and superficial fibres, andone had only superficial ones. The pattern seen in normal animals, in which fibresfrom dorsal retina generally run deeper than those from ventral retina is not,therefore, repeated in regeneration (Fawcett etal. 1984). There also seemed to beno difference in the distribution within the optic tract of fibres regenerating to thecontralateral tectum, or to an innervated or virgin ipsilateral tectum. However,

Regenerating frog retinotectal fibres 239

the presence or absence of deep regenerating fibres appeared to correlate to someextent with the way in which the cut end of the optic nerve had reconnected to thebrain. In cases where the optic nerve had reconnected to its stump, the passage ofoptic fibres into the brain was fairly smooth, and few deep axons were found,whereas where the optic nerve had reconnected directly to the diencephalon, therewas frequently a neuroma where fibres entered the brain, and fibres exiting fromthis neuroma frequently ran deep. The normal optic tract is deep near the chiasm,

Fig. 2. (A) A parasagittal section of the optic tectum from a normal postmetamorphicfrog to show the rather indistinct layering of the optic neuropil. The optic fibres havebeen filled with HRP. Note that there are no labelled fibres immediately beneath thepia. In, and just below the deeper layer of neuropil, there are many thick, tortuousfibres. Bar = 50/urn. (B) This layer of tortuously running fibres is more clearly seen inthis transverse section, in which the neuropil has been stained much less densely.

240 J. W. FAWCETT

but as fibres actually approach the tectum they run in a superficial sheet. This isalso the case in regenerates. The deep running fibres do not enter the tectum as aseparate group from the superficially running ones, but rather they all mergetogether, and approach the tectum as a wide sheet of fibres which run close to thesurface of the brain.

Optic tecta

Contralateral or virgin ipsilateral tecta. Fibres regenerating to either of thesetwo types of tectum behaved similarly. As reported previously, fibres regeneratingto a non-innervated tectum do not grow in well-defined medial or lateral brachia,but rather tend to grow directly onto the rostral margin of the tectum. However,the pattern of ingrowth is not random, and there is some evidence for specific fibreguidance. Regenerating fibres of ventral retinal origin tend to grow over themedial part of the tectum, giving the appearance of a rather ill-defined medialbrachium, whereas fibres of nasal or dorsal origin are also found both medially andlaterally. A similar pattern of abnormal pathways of regeneration has beenreported in the retinotectal projections of other species of frog (Udin, 1978) andgoldfish (Meyer, 1980; Cook, 1983; Stuermer & Easter, 1984).

Regenerating fibres grow in the tectum in one of two layers. Many of the fibresgrow through the layers in which optic fibres are found in non-regenerates, theneuropil and deep fibre layers (see page 237). Here they tend to be concentrated in

Fig. 3. Optic fibres, when they leave the periphery of the tectum to run to their sites oftermination, are found largely in the layer of fibres in and under the deeper layer of thetectal neuropil. This transverse section is from a normal animal whose nasoventralretina was labelled with HRP. The section is taken just rostral to the area of tectumwhich contains labelled neuropil. Most of the labelled fibres have left the margin of thetectum, and are spreading out through the deeper retinorecipient layers. Bar = 200 jian.

Regenerating frog retinotectal fibres 241

"*

4A

B

Fig. 4. Sections through the optic nerves (A) ipsilateral and (B) contralateral to thesite of HRP filling in an animal whose optic nerve had been cut on the labelled side 21days previously. These photographs are from a transverse section of the brain, whichaccounts for the oblique angle of section of the optic nerves. The labelled optic nerve(A) is filled throughout with reaction product, but the fibres visualized in the othernerve are confined to a ring around its periphery. Regenerating fibres may grow downthe contralateral optic nerve instead of into the optic tract, but they seldom penetrateinto it. Bars = 100 jum.

the deep fibre layer, and the deeper parts of the neuropil layer, but some are alsofound in its more superficial part. There are also many fibres which grow in a layerimmediately beneath the pia (see Fig. 6). This is a layer in which one never sees

242 J. W. FAWCETT

optic fibres in the normal brain. These superficial regenerating fibres are so closeto the surface that it appears as if they have grown just under the surface of thetectum. The fibres almost certainly grew in these layers, since the same pattern isseen regardless of whether the period of regeneration is 14 days or 4 months.Moreover, when regenerates had only their nasal retina filled, so as to selectivelylabel the fibres terminating in caudal tectum, the labelled fibres in rostral tectum,which must all be fibres of passage, also ran largely in these same two layers,

-•t

B

Fig. 5. (A) A transverse section through the brain of an animal whose optic nerve hadbeen sectioned 2 months previously. The regenerating fibres have grown into bothoptic tracts, but instead of occupying their entire cross section (see Fig. 1), they arelimited to the region immediately under the pia. (B) In this animal which was killed1 month after the optic nerve had been cut, many fibres can be seen running deeply inthe tracts, on their way to the ipsilateral (right) and contralateral (left) tecta. In thiscase more fibres had regenerated to the ipsilateral side. On both sides, near the top ofthe picture, label is seen in the neuropil of Bellonci. Bars = 250 jum.

Regenerating frog retinotectal fibres 243

directly under the pia, and in the deeper layer. The areas in which terminallabelling was seen in animals with partial retinal fills were in the correct generalposition, but rather widespread, particularly on 'virgin' tecta.

Innervated ipsilateral tecta. As previously reported, optic fibres regenerating to atectum innervated by non-regenerating fibres from the other eye behave verydifferently from those regenerating to a non-innervated tectum. Instead ofgrowing straight onto the rostral pole of the tectum, they mostly grow around themargin of the tectum in either the medial or lateral brachium. In the presentexperiments, as in those reported previously, the majority of the regeneratingfibres grew in the medial brachium, and a minority in a very clearly defined lateralbrachium. There was no evidence that fibres from different parts of the retina werechoosing the 'correct' brachium (see Fig. 7). There was also a variable, but usuallysmall, number of fibres which failed to get into either brachium, and grew directlyonto the tectum.

Fibres regenerating to an innervated tectum almost always grew in the deeperretinorecipient layers after leaving the tectal periphery. Some were seen directlyunder the pia, but their number was much smaller than on either the 'virgin' or thecontralateral tecta (see Fig. 8). Some animals received fills of only their nasalretina. In these cases the vast majority of the labelled regenerating fibres grew ineither the medial or lateral brachium, but a few grew directly onto the rostral poleof the tectum. About one third of these 'wanderers' ran directly under the pia; amuch larger proportion than amongst the fibres which entered a brachium before

Fig. 6. Fibres regenerating back to the contralateral tectum, or to a virgin tectum,frequently grow just beneath the pia in a zone in which optic fibres are not normallyfound (see Fig. 2). This is a transverse section from the tectum of an animal killed l imonths after nerve section. Coarse fibres can be seen just under the pia (arrowed), aswell as in the deeper retinorecipient layers (bracketed). Bar = 100 jum.

244 J. W. FAWCETT

Fig. 7. A transverse section through both tecta of an animal whose nasal retina wasfilled with HRP 1 month after nerve section. On the contralateral side (left) there arefibres spread all across the tectum, whereas on the ipsilateral side (which is alsoinnervated by the eye whose optic nerve was intact), the fibres of passage are mostlyfound in either the medial brachium, or in the lateral brachium (arrowed). Some'wanderers' are seen in the centre of the tectum, and also some fibres which havestarted to fan out from the two brachia. This preparation was drawn, because many ofthe fibres were too fine for adequate photographic reproduction. Bar = 250 jum.

growing onto the tectum. In the animals in which a selected part of the retina waslabelled with HRP, terminal labelling in the tectum was restricted to theappropriate area. The area occupied by these labelled terminals was generallysmaller on the innervated ipsilateral tectum than on the same animal's contra-lateral tectum, particularly after the shorter regeneration times. The layering ofterminals in the tectum was also more noticeable on the innervated ipsilateral side

Fig. 8. A transverse section of the tectum ipsilateral to the cut optic nerve in an animalthat had been allowed to regenerate for 1£ months. This tectum has an intact projectionfrom the other eye on it, as well as the newly arrived regenerated fibres. On this sidethere are only a few fibres running under the pia, as compared to the contralateraltectum from the same animal, which is shown in Fig. 6. The layering of the neuropil isalso rather more marked on the ipsilateral side. Bar = 100 jum.

Regenerating frog retinotectal fibres 245

than on the contralateral tectum of animals with short regeneration times, but thedensity of labelled neuropil was usually lower on the ipsilateral than contralateraltectum, as reported previously by Straznicky etal. (1980).

DISCUSSION

The regeneration of axons in the frog retinotectal projection has long been ofinterest both as a model for the regeneration of axons in the CNS, a capabilitythat mammals lack, and also as a tool for doing experiments relating to thedevelopment of the retinotectal projection. Such experiments are meaningful,when carefully interpreted, because the frog retinotectal projection is essentially adeveloping system, since both retina and tectum continue to grow throughout thelife of an animal. Regenerating fibres therefore grow through an environmentwhich is able to interact with normally developing optic fibres. This is especiallytrue in those cases where fibres regenerate to an innervated ipsilateral tectum;here not only are the neurons and glia of the brain present, but also the axons fromthe contralateral eye, which have not been damaged by the optic nerve cut. Thismeans that fibres regenerating to the ipsilateral side might be expected to showsome of the interactions of newly growing fibres with already established fibresthat may help to guide axons to their targets.

In a recent experiment, Gaze & Straznicky (1980) looked at the termination ofregenerated fibres from compound eyes on the ipsilateral and contralateral tecta ofXenopus, and showed that there must be interactions between the incumbent andregenerating fibres, such that regenerating fibres only terminate near to non-regenerated fibres which carry the same retinal positional label. Gaze & Fawcett(1983), and Taylor & Gaze (personal communication) took this approach further,and looked at the pathway taken by fibres as they regenerated to both innervatedand non-innervated tecta. These studies showed that the presence of intact opticfibres made a great difference to the pathways taken by the regenerating axons;fibres regenerating to a non-innervated tectum tended to grow directly onto therostral pole of the tectum, and then across it to their termination, whereas fibresregenerating to an innervated tectum mostly grew around the very periphery ofthe tectum mainly in the medial brachium, and grew onto the tectum close to theireventual termination sites. This observation is particularly interesting, since fibresarriving at the tectum during normal development also travel around the tectalperiphery in this way (Fawcett & Gaze, 1982; Reh etal. 1983; Stuermer & Easter,1984; Rusoff, 1984). The study reported in this paper is a more detailed re-examination of this same experimental paradigm, using sectioned material ratherthan whole mount preparations.

Optic tract

Gaze & Grant (1978) examined the pathway of regenerating optic fibres in theoptic tract of Xenopus tadpoles, and found that they all grew immediately underthe pia. They suggested that normally developing fibres might behave in the same

246 J. W. FAWCETT

way, and that this would lead to an age-related layering of fibres in the optic tract,the oldest ones being found deepest, and the youngest fibres most superficially.That this is indeed so has recently been confirmed in Xenopus (Fawcett et al. 1984).In the present series of experiments, in which I have looked at the optic tract ofrather older, postmetamorphic Xenopus, I often saw a distinct layer of re-generating fibres immediately beneath the pia, but in many cases there were alsofibres running deep in the optic tract. These deep running fibres were seen in thecases with the shortest as well as the longest regeneration times, so it is probablethat the fibres had actually grown through the deep part of the tract. There is,therefore, no absolute restriction of new fibre growth to the surface of the brain.The presence or absence of established fibres for the regenerating fibres to followseemed to make no difference to whether the regenerating fibres ran deep or not;the pathway of fibres in the optic tract was similar regardless of whether they weregoing to an innervated or non-innervated tectum. The subpial layer is probably thepath of least resistance both for regenerating and normally developing fibres.Immediately under the pia of the optic tract is a layer of basal lamina, which isnormally covered by glial end feet (Fawcett, unpublished observations). It seemsprobable that the growth cones can part this glial layer, and grow on the layer ofbasal lamina, as reported by Easter, Bratton & Scherer (1984) in the goldfishretina. However, axons are also able to grow deep in the Xenopus brain if theymiss the subpial layer. The Xenopus brain does not, therefore, seem to becomeinhospitable to axonal growth as it matures in the way that the mammalian brainappears to.

Optic tectum

As far as the gross organization of the regenerating fibres in the tectum isconcerned, my results largely confirm previous studies; fibres regenerating to anuninnervated tectum, whether contralateral or 'virgin' show only a slight tendencyto grow in brachia, whereas those regenerating to an innervated ipsilateral tectumshow a clearer brachial organization.

Fibres of passage in the normal non-regenerated Xenopus retinotectalprojection are found largely deep to the layers of neuropil. Regenerating fibresalso tend to run in this layer, although this arrangement is not as clearly defined inregenerates as in non-regenerates. There may, however, be some specific guidanceof axons into this particular tectal layer. Many regenerating fibres, though, growover the tectum in a layer immediately under the pia, which is a layer in whichoptic fibres are never seen in non-regenerates. The highest proportion ofsuperficially running fibres was seen amongst fibres projecting to 'virgin' tecta, thenext highest amongst those regenerating to uninnervated contralateral tecta, andthe lowest proportion on innervated ipsilateral tecta. The presence of fibres fromthe contralateral eye, therefore, seems to make it easier for regenerating fibres toenter the deeper tectal layers. However, while incumbent fibres seem to exert apositive influence on the growth of the regenerating fibres, they appear, asStraznicky et al. (1980) have previously noted, to make it harder for them to form

Regenerating frog retinotectal fibres 247

arbors, since in animals whose regenerating fibres were labelled only a short timeafter the optic nerve crush, the terminal labelling on the ipsilateral tectum waslighter than on the contralateral side. It seems, then, that the rules for theinteraction between regenerating and incumbent fibres are different in thedifferent parts of the optic pathway; in the optic tract the presence or absence ofintact fibres makes no difference to the proportion of regenerating fibres whichgrow deep in the brain. In the region of the tectum, the presence of incumbentfibres makes it easier for regenerating axons to penetrate the deeper retino-recipient layers and also tends to guide regenerating fibres into one of the brachia,indicating that ingrowing fibres probably find the incumbent fibres an attractivegrowth surface here. It seems possible that the glial environment of the brainchanges between the optic tract and tectum, such that fibres prefer to grow onfibres in the region of the tectum, but on other surfaces in the tract. A similarchange in the type of glia, and their interactions with optic axons has recently beenreported by Scholes (personal communication) in the cichlid fish.

Given that these interactions occur between regenerating and in situ fibres, is itpossible to infer anything useful about the guidance of fibres arriving during thecourse of normal development?

During development newly arriving fibres grow around the periphery of thetectum, leading, because the tectum is constantly growing, to age-related 'rings' offibres on its surface (Cook, Rankin & Stevens, 1983; Stuermer & Easter, 1984;Rusoff, 1984; Reh etal. 1983). Regenerating fibres may also grow around the tectalperiphery, but only when there are incumbent fibres to follow does a substantialproportion of them seem to do so. It seems reasonable to extrapolate from thisbehaviour of regenerating fibres, and hypothesize that in development, newlyarriving fibres grow around the tectal margin because they are following the mostrecently arrived fibres. Why only the most recently arrived ones? One possibleexplanation for this is that the older fibres have become 'masked' by beingovergrown by the tectum, and are now deep in the neuropil, whereas the recentlyarrived fibres are still superficial.

If, then, there are interactions between incoming and incumbent fibres, arethese in any way specific; do fibres from, say, the ventral retina only interact withincumbent fibres from ventral retina? Neither in this present study, nor in previousstudies of Xenopus regeneration has it been possible to convincingly demonstratethis sort of specific fibre following by the regenerating fibres. During normaldevelopment, fibres from the ventral part of the retina all grow in the medialbrachium, while fibres from dorsal retina all grow in the lateral brachium.However, when regenerating fibres grow to an innervated tectum, their brachialselection seems to be hardly influenced by their position of retinal origin. But,having entered one or other brachium, regenerating fibres only enter onto thetectum in the 'correct' place, which could be due to fibre-fibre interactions.

During normal development brachial selection is certainly not random, so whatcould determine it? There seem to be three main possibilities: (a) growing fibresmight follow their neighbours all the way from retina to tectum. The first

248 J. W. FAWCETT

pioneering fibres search for and find appropriate positional markers on the tectum(Holt & Harris, 1983), and lay down a first pathway, which subsequent arrivalscould follow, leading, as the tectum grows, to the adult pattern. There is a certainamount of evidence against such simple fibre-following schemes, although none ofit is admittedly quite watertight (Fawcett, 1981; Reh et al. 1983; Scholes, 1979;Scalia & Arango, 1983). (b) There could be a 'signpost' which tells fibres as theyreach rostral tectum whether they should turn medially or laterally, themechanism presumably being swamped by the large numbers of fibres arrivingsimultaneously during regeneration. This means bringing yet another mechanisminto retinotectal development, but it certainly is a possibility, and there is noevidence either for or against it. (c) Fibres might search out and follow previousarrivals from the same retinal position, presumably at roughly the point wherefibres reach the tectum, and follow them to their target. This last scheme has themerit of providing the most economical explanation for pathway finding in theretinotectal system, despite our failure to find evidence for a specific guidance ofregenerating fibres in this way. However, this failure may be explicable on thegrounds that the large number of fibres which probably arrive simultaneously inregeneration may not find sufficient available surface area in the incumbent fibresfor them all to be able to interact with them. If fibres were guided in this way, theywould automatically grow to the correct region of the tectum, and would not haveto search all over its surface to find an appropriate target. In support of this generalidea I may cite the form of interaction between optic fibre terminals based onpositional labels, that is seen in experiments in which compound eyes regenerateonto innervated tecta (Gaze & Straznicky, 1980; Gaze & Fawcett, 1983) (thisshould be distinguished from electrically mediated terminal interactions, whichare responsible, for instance, for eye dominance stripes (Fawcett & Willshaw,1982; Meyer, 1982; Schmidt & Edwards, 1983)). It seems reasonable, therefore, toextrapolate from this, and postulate that as incoming optic fibres approach thetectum, they search out an appropriate pioneer to follow, and are guided by it tothe correct termination site. The first fibres to arrive, of course, have no pioneersto follow, but, since they arrive at a tectum that is not much larger than a singleterminal arbor, they can search much of its area, and use positional addresses onits surface to find their correct target area (Holt & Harris, 1983). Specificfibre-fibre interactions have been demonstrated in the chick retinotectal system invitro (Halfter, Claviez & Schwarz, 1981; Bonhoeffer & Huf, 1985) and in the frogin vivo (Fawcett, 1981; Fawcett & Gaze, 1982; Gaze & Fawcett, 1983), but so faronly between fibres of nasal and temporal origin.

To summarize, in this series of experiments I have demonstrated that opticfibres can regenerate through the substance of the brain, and not just under thepia. I have also further defined the interactions between ingrowing optic fibres andtheir environment. It seems likely that these interactions involve a specificinfluence of in situ fibres on ingrowing ones. It remains to be shown what thismeans in terms of the behaviour of individual growth cones.

Regenerating frog retinotectal fibres 249

This work was supported by grant EY-03653 from the National Eye Institute, and was carriedout while the author was in receipt of an Alexander Pigott Wernher travelling fellowship fromthe U.K. Medical Research Council. I should like to thank Mr James Rokos and Ms SarahBacon for technical assistance, Mr Kris Trulock for the photography, and Ms Pat Thomas fortyping the manuscript.

REFERENCESADAMS, J. C. (1977). Technical considerations on the use of HRP as a neuronal marker.

Neuroscience 2, 141-145.BEACH, P. M. & JACOBSON, M. (1979). Patterns of cell proliferation in the retina of the clawed

frog during development. /. comp. Neurol. 183, 603-611.BOHN, R. C. & STELZNER, D. J. (1981). The aberrant retino retinal projection during optic nerve

regeneration in the frog II. Anterograde labelling with Horseradish Peroxidase. /. comp.Neurol. 196, 621-632.

BONHOEFFER,F. & HUF, J. (1985). Position dependant properties ofaxons and their growth cones.Nature 315, 409-410.

COOK, J. R. (1983). Tectal paths of regenerated optic axons in the goldfish: evidence fromretrograde labelling with Horseradish Peroxidase. Expl Brain Res. 51, 433-442.

COOK, J. E., RANKIN, E. C. C. & STEVENS, H. P. (1983). A pattern of optic axons in the normalgoldfish tectum consistent with caudal migration of optic terminals during development. ExplBrain Res. 52, 147-151.

EASTER, S. S.,BRATTON,B. &SCHERER, S. S. (1984). Growth-related order of the retinal fiber layerin goldfish. /. Neurosci. 4, 2173-2190.

FAWCETT, J. W. (1981). How axons grow down the Xenopus optic nerve. /. Embryol. exp. Morph.65, 219-233.

FAWCETT, J. W. & GAZE, R. M. (1981). The organisation of regenerating axons in the Xenopusoptic nerve. Brain Res. 229, 487-490.

FAWCETT, J. W. & GAZE, R. M. (1982). The retinotectal fibre pathways from normal andcompound eyes in Xenopus. J. Embryol. exp. Morph. 72,19-37.

FAWCETT, J. W., TAYLOR, J. S. H., GAZE, R. M., GRANT, P. & HIRST, E. (1984). Fibre order in thenormal Xenopus optic tract near the chiasma. /. Embryol. exp. Morph. 83,1-14.

FAWCETT, J. W. & WILLSHAW, D. J. (1982). Compound eyes project stripes on the optic tectum inXenopus. Nature 296, 350-352.

FELDMAN, J. D., GAZE, R. M. & KEATING, M. J. (1971). Delayed innervation of the optic tectumin Xenopus laevis. Expl Brain Res. 14, 16-23.

GAZE, R. M. & FAWCETT, J. W. (1983). Pathways of Xenopus optic fibres regenerating fromnormal and compound eyes under various conditions. /. Embryol. exp. Morph. 73,17-38.

GAZE, R. M. & GRANT, P. (1978). The diencephalic course of regenerating retinotectal fibres inXenopus tadpoles. /. Embryol. exp. Morph. 44, 201-216.

GAZE, R. M., KEATING, M. J., OSTBERG, A. & CHUNG, S.-H. (1979). The relationship betweenretinal and tectal growth in larval Xenopus: implications for the development of theretinotectal projection. /. Embryol. exp. Morph. 53, 103-143.

GAZE, R. M. & STRAZNICKY, C. (1980). Regeneration of optic nerve fibres from a compound eyeto both tecta in Xenopus: Evidence relating to the state of specification of the eye and thetectum. /. Embryol. exp. Morph. 60,125-140.

HALFTER, W., CLAVIEZ, N. & SCHWARZ, U. (1981). Preferential adhesion of tectal membranes toanterior embryonic chick retina neurites. Nature 292, 67-70.

HOLT, C. E. & HARRIS, W. A. (1983). Order in the initial retinotectal map in Xenopus: A newtechnique for labelling growing nerve fibres. Nature 301, 150-152.

JOHNS, P. R. (1977). Growth of the adult goldfish eye III: Source of new retinal cells. /. comp.Neurol. 176, 343-358.

LEVTNE, R. L. (1980). An autoradiographic study of the retinal projection in Xenopus laevis withcomparisons to Rana. J. comp. Neurol. 189,1-29.

MEYER, R. L. (1980). Mapping the normal and regenerating retinotectal projection of goldfishwith autoradiographic methods. /. comp. Neurol. 189, 273-289.

MEYER, R. L. (1982). Tetrodotoxin blocks the formation of ocular dominance columns ingoldfish. Science 218, 589-591.

250 J. W. FAWCETT

NIEUWKOOP, P. D. & FABER, J. (1956). Normal Table o/Xenopus laevis (Daudin). Amsterdam:North Holland.

POTTER, H. D. (1969). Structural characteristics of cell and fiber populations in the optic tectumof the frog. /. comp. Neurol. 136, 203-232.

REH, T. A., PITTS, E., CONSTANTINE-PATON, M. (1983). The organization of the fibers in the opticnerve of normal and tectumless Rana pipiens. J. comp. Neurol. 218, 282-296.

RUSOFF, A. C. (1984). Paths of axons in the visual system of Perciform fish and implications ofthese paths for rules governing axonal growth. /. Neurosci. 4, 1414-1429.

SCALIA, F. (1973). Autoradiographic demonstration of optic fibers in the stratum zonale of thefrog's tectum. Brain Res. 58, 484-488.

SCALIA, F. & ARANGO, V. (1983). The anti retinotopic organization of the frog's optic nerve. BrainRes. 266, 121-126.

SCALIA, F. & COLEMAN, D. R. (1974). Aspects of the central projection of the optic nerve in thefrog revealed by the anterograde migration of HRP. Brain Res. 79, 496-504.

SCALIA, F. & FITE, K. (1974). Retinotopic analysis of the central connections of the optic nerve inthe frog. /. comp. Neurol. 158, 455-478.

SCHMIDT, J. T. & EDWARDS, D. L. (1983). Activity sharpens the map during the regeneration ofthe retinotectal projection in goldfish. Brain Res. 268, 28-39.

SCHOLES, J. H. (1979). Nerve fibre topography in the retinal projection to the tectum. Nature 278,620-624.

STEEDMAN, J. G., STIRLING, R. V. & GAZE, R. M. (1979). The central pathways of optic fibres inXenopus tadpoles. /. Embryol. exp. Morph. 50, 199-315.

STRAZNICKY, C. & GAZE, R. M. (1982). The innervation of a virgin tectum by a double temporalor a double nasal eye in Xenopus. J. Embryol. exp. Morph. 68, 9-21.

STRAZNICKY, C. & TAY, D. (1981). Interactions between compound and normal eye projections indually innervated tectum: a study of optic nerve regeneration in Xenopus. J. Embryol. exp.Morph. 66, 159-174.

STRAZNICKY, C , TAY, D. & GLASTONBURY, J. (1980). Regeneration of an abnormal ipsilateralvisuotectal projection in Xenopus is delayed by the presence of optic fibers from the other eye./. Embryol. exp. Morph. 57, 129-141.

STUERMER, C. A. O. & EASTER, S. S. JR. (1984). Rules of order in the retinal fascicles of goldfish./. Neurosci. 4, 1052-1064.

STUERMER, C. A. O. & EASTER, S. S. JR. (1984). A comparison of the normal and regeneratedretinotectal pathways of goldfish. /. comp. Neurol. 223, 57-76.

SZEKELY, G. & LAZAR, G. (1976). Cellular and synaptic architecture of the optic tectum. In FrogNeurobiology (ed. R. Llinas & W. Precht). Berlin: Springer Verlag.

UDIN, S. (1978). Permanent disorganisation of the regenerating optic tract in the frog. ExplNeurol. 58, 455-470.

{Accepted 3 July 1985)