Xenopus temporal retinal neurites collapse on contact with glial cells from caudal tectum in vitro

During the development of the visual system of Xenopus laevis retinal fibres grow along the optic nerve, enter the brain and pass contralaterally up the optic tract and terminate on the optic tectum. The terminals are distributed in a topographically ordered fashion over the tectum, such that temporal retinal axons grow to the most rostral parts of the tectum, nasal fibres grow to caudal tectum, and ventral and dorsal fibres project to medial and lateral tectum respectively. The result of this ordered growth is that a map of the retina, and thus of the visual field is projected onto the tectum.

Several mechanisms have been proposed to account for the development of neural maps, and all require the interaction of growth cones with molecular cues present in the target area (for a review of molecular mechanisms see Dodd and Jessel, 1988). Recent experiments using chick and goldfish tissues in vitro have supported the idea that a graded series of cues on the tectum is recognised by retinal growth cones (Bonhoeffer and Huf, 1982; Walter et al. 1987a;Vielmetter and Stuermer, 1989). These workers have all used elegant in vitro assays to examine the behaviour of retinal neurites growing on cells or plasma membranes isolated from different parts of the optic tectum. The results showed that temporal retinal axons recognise and respond to a property of the tectal cells that is graded rostrocaudally. When presented with a choice of growing on either rostral or caudal tectal cells or membranes, temporal fibres always grew on the more rostral substratum; nasal fibres made no distinctions and grew equally well on both types of substratum. The experiments of Walter et al. (1987b) suggest that the graded property to which temporal fibres are responding is an inhibitory factor present on the surfaces of caudal tectal cells, and therefore temporal fibres choose to grow on rostral cells because they are inhibited from doing so on caudal cells.

In previous papers (Gooday, 1990; Jack et al. 1991), we have examined the growth of Xenopus neurites from nasal and temporal retina on pure substrata of glial cells isolated from the diencephalon and from the rostral and caudal thirds of the optic tectum. On glial cells from the diencephalon, there were no differences in the growth of nasal and temporal fibres; both types of fibre grew as long thin fascicles. On tectal glial cells, however, temporal fibres were restricted to growing in short fat fascicles. This property of tectal glial cells is more concentrated in glia from the caudal third of the tectum; nasal and temporal fibres show different growth patterns on rostral tectal glia, although this difference is less marked. This restriction of the growth of temporal fibres by caudal tectal glial cells is in accordance with the inhibitory theory proposed by Walter et al. (1987b).

The phenomenon of neurite inhibition has been recognised for many years. The importance of growth cone inhibition in the formation of neuronal connections during normal development and in regeneration, however, has only recently been realised (for a review see Patterson, 1988). The in vitro experiments of Bray et al. (1980) and of Kapfhammer et al. (1986) showed that embryonic chick retinal and sympathetic neurites did not mix in culture; they suggested that active avoidance of heterotypic neurites could serve as a guidance mechanism during development. Kapfhammer and Raper (1987) showed that this avoidance was caused by the collapse of growth cones and retraction of neurites on contact with heterotypic neurites. This work suggests that growth cone collapse is an integral part of neurite inhibition in vitro and may then serve as a useful indicator in examining other inhibitory phenomena.

In this paper, we have analysed encounters between retinal growth cones and individual glial cells in vitro to examine whether retinal neurites are inhibited by the glial cells that they may encounter as they grow to their targets in the optic tectum.

Sparse cultures of glial cells were prepared from the tecta and diencephalons of Xenopus laevis tadpoles at stage 54 (Niewkoop and Faber 1967), the stage at which there is a maximal growth of retinal fibres (Jacobson 1976; Beach and Jacobson 1979). Explants of retina from the most nasal and temporal parts of the eye were cultured with the glial cells and encounters between growth cones and glial cells were recorded by time-lapse videomicroscopy.

Plastic tissue culture Petri dishes (Nunc, 35 mm) were coated with laminin by applying a solution of laminin (20 μgml^-1 in Niu Twitty saline, Gibco/BRL). The dishes were left for 2h at room temperature, and were rinsed in Niu Twitty saline before use.

The tecta and diencephalons from four or five stage 54 tadpoles were dissected aseptically and placed in 500 μl of calcium- and magnesium-free Niu Twitty saline. The tissues were flushed three times through a 1 ml syringe fitted with a 21G needle, three times through a 23G needle and finally twice through a 26G needle. Each passage was performed slowly to avoid entrapment of air bubbles. The dissociated tissues were centrifuged at 2000 revs min^-1 for 2 min and the pellet was resuspended in 1 ml of culture medium. Aliquots of cell suspension (100 μl) were placed in microculture chambers made by sealing a 13 mm diameter silicone rubber ring (Sylgard 184 Silicone rubber elastomer, Dow Coming) to the base of a laminin-coated Petri dish. The suspensions were left for 3 –4 days at 20°C in a refrigerated incubator; the rings were removed and the dishes were filled with 2ml of medium. Before use, the neuronal cells were removed by flushing with medium expelled from a Pasteur pipette. The cultures were allowed to grow until the cell density was around 6 cells mm^-2.

The optic nerves of stage 54 tadpoles were cut bilaterally under MS222 anaesthesia to stimulate later outgrowth of fibres (Agranoff et al. 1976). The animals were allowed to recover for 10 days. They were then killed, the eyes removed and the retinas dissected free. Small parts of the retina, about 250 μm square, were cut from the most nasal, and temporal regions and were placed in the middle of the glial cell cultures. The explants were held in place by small squares of polyacrylamide gel weighted with fragments of glass slide. The gels and glass were removed 24 h later.

Tissue culture medium consisted of 60% L15 (Flow), 6 g I^-1 D-glucose (BDH), 0.3g^-1 L-glutamine (Flow), 16mgl^-1 putrescine (Sigma), 5mgl^-1 transferrin (Sigma), 5mgl^-1 insulin (Sigma) and 10% foetal calf serum (Sera Lab).

Tissue for anti-GFAP staining was fixed in methanol at – 20°C for 30 min and was then immersed in a cryoprotectant solution of 30 % sucrose in PBS and left overnight. The tissue was then snap-frozen in a mixture of acetone and dry ice, mounted on chucks with OCT compound (BDH), and 20 μm cryostat sections were cut and mounted on subbed slides. The sections were post-fixed in methanol at – 20°C for 6min, rinsed three times in PBS containing 0.1% Triton X-100 for 5 min each and then transferred to 10% horse serum in PBS for 30 min, drained of excess serum and incubated overnight at 4°C in anti-GFAP (1:50, clone G-A-5, Boehringer and Sigma); control slides were incubated in 0.1% BSA in PBS. After incubation in primary antiserum slides were rinsed three times 15 min each in PBS, incubated in biotinylated anti-mouse IgG 1:100 (Vector) for 30min, rinsed again three times and incubated with ABC complex (Vector) for 30 min, then rinsed three times in PBS. The HRP was visualised using DAB as the chromogen.

To confirm the glial nature of the cultured cells, some of the cultures were stained with anti-GFAP after photographing a growth cone collapse. The cells in question were identified for future location by scratching the plastic dish around the cells using a fine needle. The cultures were then fixed in methanol at – 20°C for 6 min and were then treated according to the above protocol for tissue sections, with the exception that Avidin-Texas Red 1:100 (Vector) was used to visualise the cells.

After addition of the retinal explants, the cultures were left for 24 h before being examined and recorded. Videorecordings were commenced when retinal growth cones were about to contact individual glial cells. To ensure that growth cones had not already contacted glial cells, situations were chosen in which there were no glial cells near the axon. In addition only those glial cells that were at least 200 – 300 μm from the explant were chosen to ensure that neurites were not encountering retinal glia. Recordings were made using a Leitz Fluovert microscope, a Panasonic WV-1500TV camera and Panasonic NV-8055 time-lapse video recorder with NV-F85 time date generator. Photographs were taken at selected intervals using Ilford FP4 35mm film and a Wild MPS 45 camera.

Time-lapse videorecordings were made of growth cone –glial cell encounters of each of six types: nasal and temporal growth cones were confronted with glia from the diencephalon, the rostral third of the optic tectum and the caudal third of the tectum. The number of recordings made and the number of growth cones to show collapsing behaviour and retraction are listed in Table 1.

Twenty encounters of this sort were recorded and, in all cases, the growth cone of the advancing neurite collapsed on contacting the glial cell. Growth cones collapsed and retracted from both flattened and elongated glial cells. Encounters of this type are shown in Figs 1 and 2. Typically temporal fibres advanced at a mean rate of 78 μm h^-1 . The growth cones of both nasal and temporal retinal neurites consisted of large irregular lamellipodia with a few filopodia, there tended to be more filopodia associated with the rear of the growth cones rather than at the leading edge. The lamellipodia exhibited extensive ruffling activity at the leading edge. On contact with caudal glial cells this ruffling activity did not immediately cease, and the growth cones continued to advance until extensive contact was made with the cell. Two or three minutes after initial contact, growth cone morphology began to change and forward advance ceased. The lamellipodia were withdrawn and the growth cone shrank in size and after 4 –5 min began to withdraw from the cell. After a mean time of 8 min in contact with the cells, the growth cones retracted completely from the glial cells and, in some cases, the neurites were seen to collapse and assume an irregular wavy profile as they withdrew from the cells. On withdrawing from the cells, fine stretched processes were often left attached to the glial cells as the neurites retracted up to 100 μm.

In some cases, we recorded retracted axons that organised new growth cones and advanced again to contact, collapse and retract from the same ceil. Fig. 3 shows such an example; fibre 1 has already retracted from cell a, and a fine stretched process can be seen linking the two. In Fig. 3B, fibre 1 has developed a new growth cone and it has advanced again to contact the cell. In Fig. 3C and D, it collapses and retracts from the cell again. Meanwhile fibre 2 contacts and withdraws from cell b.

In encounters between temporal retinal growth cones and glial cells from rostral tectum and diencephalon, and in encounters of nasal growth cones with glia from all three tissues, the majority of growth cones did not show any collapsing or retracting behaviour (Table 1). Examples of a temporal growth cone contacting a diencephalic glial cell, and a nasal growth cone encountering a rostral tectal cell are shown in Figs 4 and 5. In most cases, the rate of forward advance of the growth cones slowed on contact with glial cells, but ruffling activity of the lamellipodia continued. This gave the impression that the growth cones were ‘exploring’ the surfaces of the glial cells. The growth cones continued to advance over the surface of the cells and on crossing the cells continued growing on the plastic surface.

The cells in our cultures have been identified as glia in previous publications by staining with anti-GFAP (Gooday, 1990; Jack et al. 1991). In this paper, we have fixed and immunocytochemically characterised identified cells after they had induced the collapse of temporal retinal growth cones Figs 6 and 7; this was performed on six separate occasions and all the cells stained with anti-GFAP. In frozen sections of brain, the anti-GFAP stained only radial glial cells. These spanned the width of the tectum from the ventricle to the pial margin, where they terminated in enlarged end feet (Fig. 8).

The results of this study show that in vitro the growth cones of temporal retinal neurites collapse and retract from caudal tectal glial cells, but not from rostral cells, or from cells from the diencephalon. Nasal fibres rarely retract from any of the glia and continue to grow over the surface of the cells. The results suggest that temporal retinal fibres are inhibited from growing to an inappropriate target, the caudal optic tectum, due to the repulsive nature of the glia; nasal fibres do not find caudal glia repulsive. As yet we cannot say whether this repulsive effect is graded across the tectum. Due to the small size of the Xenopus brain it is feasible to dissect it into only three parts, the middle part was discarded to ensure no overlap of the cell types isolated. The stripe assays of Walter et al. (1987a) on chick and of Vielmetter and Stuermer (1989) on goldfish suggest that this effect does operate in a rostrocaudal gradient.

It is unclear at present how this collapsing activity relates to the in vivo guidance of retinal axons on the tectum, the culture methods in this paper detect collapsing activity only in the caudal third of the tectum, a part that is not encountered by temporal fibres in normal growth. Our earlier work (Gooday, 1990; Jack et al. 1991), which involved culturing retina on confluent monolayers of glia, showed that temporal fibres were highly growth-restricted on caudal tectal glia and, to a lesser extent on rostral tectal glia. However, the present study shows that rostral tectal glial cells do not cause the collapse of any temporal growth cones. The previous studies would have led us to expect that rostral cells would have caused a proportion of temporal growth cones to collapse. The lack of collapse induced by rostral glia in this present work could be an artefact of culture; the expression of an inhibitory molecule by the glia, or the threshold of its recognition by growth cones may be lower in vitro than in vivo. The discrepancy between the experiments could also be due to a quantitative effect due to the numbers of glial cells: the previous experiments were performed using confluent monolayers in contrast to the individual glial cells used in this study. The sections of brain stained with anti-GFAP (Fig. 8) show that there are many hundreds of radial glia in the tectum, so it is probable that growth cones in vivo are in contact with more than one glial cell at any time. In this respect the collapse assay is perhaps a more artificial environment than monolayer culture. Therefore, assuming that a gradient of inhibition exists over the tectum, a higher level of inhibitory activity per cell may be required to elicit a response in the collapse assay than in the monolayer assay or in vivo.

Recent work from several laboratories suggests that the collapsing activity is caused by a cell surface component. Raper and Kapfhammer (1990) have found collapsing activity in enriched, detergent-soluble extracts of chick brain plasma membranes of putative relative molecular mass 50 ×10³. Davies et al. (1990) have isolated a Peanut agglutinin-binding glycoprotein fraction from chick somites, which causes the collapse of dorsal root ganglion growth cones in vitro. They have proposed that this molecule restricts the growth of sensory and motor nerves to the anterior half of each sclerotome, and so brings about the segmental pattern of spinal nerves characteristic of vertebrates. Cox et al. (1990) prepared suspensions of plasma membranes from anterior and posterior chick tectum. Posterior membranes, and to a lesser extent anterior membranes, caused the growth cone collapse and retraction of temporal retinal fibres, whereas nasal fibres remained unaffected by either preparation.

All these extracts have so far been made from whole tissues, and the precise cellular locations of any inhibitory molecules are so far unknown. Our results using purified populations of glial cells firmly indicate that such molecules are present on the surfaces of glial cells from the caudal part of the optic tectum at least.

Our results also suggest that glial cells may provide positional information for nerve fibres in target areas in addition to their role as general mediators of axonal guidance. Further evidence for the provision of positional information comes from the recent work of Suzue et al. (1990) who raised monoclonal antibodies against rostrocaudally graded molecules in mammalian sympathetic ganglia. One antibody ROCA1 bound to a glial cell-surface component in rostral ganglia and the binding intensity decreased in caudal ganglia.

The role of glia in facilitating axon growth and in guiding neuronal trajectories during development is well documented (reviewed by Hatten, 1990), and recently the significance of inhibitory glial neuronal interactions has been realised. The failure of mammahan CNS neurons to regenerate may be due in part to two cell-surface components of 35 and 250 ×10³, found in CNS myelin and oligodendrocyte membranes, which repel neurites (Schwab and Caroni, 1988). Oligodendrocyte-type cells in the optic nerve of goldfish, however, support the growth of retinal neurones, and promote axonal regeneration in vitro. Goldfish retinal axons recognise and are inhibited by mammalian oligodendrocytes; however, goldfish oligo-dendrocytic cells also support the growth of chick retinal neurones in vitro (Bastmeyer et al. 1991). This suggests that a fundamental difference exists between the oligodendrocytes of birds and mammals and their equivalents in fish and amphibia, in the ability of these cells to support axon growth. The glial cells in our cultures are either elongated cells or fibroblast-like and are unlike cultured mammalian oligodendrocytes morphologically. They stain with an antibody to pig GFAP, and this same antibody only stains radial glial cells in frozen sections of Xenopus tectum. The developmental lineages of amphibian glial cells, and for that matter, goldfish glial cells, are not known and little antigenic characterization has been done, unlike the situation in the rat optic nerve. Therefore we are reluctant to assign a glial category to these cells. The cells in our cultures may be equivalent to radial glial cells and it is probable that growing retinal fibres contact radial glia in vivo, as the glial processes extend to the pial margin of the tectum, where they terminate in enlarged endfeet.

The role of glia in organising fibre pathways in the visual systems of several classes of vertebrates has already been suggested by others. Several lines of evidence suggest that at key positions in the optic pathway, where fibre reordering or rerouting takes place, there is a change in glial cell phenotype. In the optic system of cichlid fish, there is a change in intermediate filament expression in glial cells at the boundary between the nerve and the tract (Maggs and Scholes, 1986). A change in glial type between nerve and tract has also been found in the ferret (Guillery and Walsh, 1987).

Silver (1984) suggested that some glial structures could act as barriers to axon growth. He demonstrated the presence of two distinct glial structures in the optic chiasm of embryonic mice. Fibres from ventrotemporal and ventronasal retina occupy opposite sides of the tract and come into contact with different glial structures. Fibres from ventrotemporal retina encountered a maze-like system of glial processes and channels that seemed to direct growth ipsilaterally, whereas fibres from more nasal retina encountered a dense glial ‘knot’ which appeared to impede growth and deflect axons contralaterally. Other axon-deflecting glial structures have been identified: the roof plate of the developing spinal cord and brain may serve to maintain the side-specificity of developing axonal tracts, which do not normally cross the midfine. The roof plate glial cells secrete keratan sulphate glycosaminoglycan (Snow et al. 1990a), which has been shown to be inhibitory for neurite growth in vitro (Snow et al. 1990b). Interference with the roof plate of the mesencephalon in hamsters leads to the misrouting of axons across the midline (Poston et al. 1988; So, 1979; Wu et al. 1989).

With the current level of interest in the role of glial cells and inhibitory phenomena in the development of the nervous system, it seems probable that research into other aspects of neural development will result in other such phenomena being described.

We are grateful to Professor R.M. Gaze for help and advice and to James Jack, Maida Davidson and Christine Virtue for technical assistance. The project was funded by the Medical Research Council of Great Britain, and by a Summer Bursary from the Nuffield Foundation awarded to A.R.J.