Cell-type-specific antibodies have been used to follow the appearance of neurones and glia in the developing nervous system of the amphibian embryo. Differentiated neurones were recognized with antibodies against neurofilament protein while glial cells were identified with antibodies against glial fibrillary acidic protein (GFAP). The appearance of neurones containing the neurotransmitters 5-hydroxytryptamine and dopamine has been charted also.
In Xenopus, neurofilament protein in developing neurones was observed occasionally at NF stage 21 and was present reliably in the neural tube and in caudal regions of the brain at stage 23. Antibodies to the low molecular weight fragment of the neurofilament triplet recognized early neurones most reliably. Radial glial cells, identified with GFAP antibody, were identified from stage 23 onwards in the neural tube and caudal regions of the brain. In the developing spinal cord, GFAP staining was apparent throughout the cytoplasm of each radial glial cell. In the brain, the peripheral region only of each glial cell contained GFAP. By stage 36, immunohistochemically recognizable neurones and glia were present throughout the nervous system.
In the axolotl, by stage 36 the pattern of neural and glial staining was identical to that observed in Xenopus. GFAP staining of glial cells was obvious at stage 23, although neuronal staining was clearly absent. This implies that glial cells differentiate before neurones.
5-HT-containing cell bodies were first observed in caudal regions of the developing brain on either side of the midline at stage 26. An extensive network of 5-HT neurones appeared gradually, with a substantial subset crossing to the opposite side of the brain through the developing optic chiasma. 5,7-dihydroxytryptamine prevented the appearance of 5-HT. Depletion of 5-HT had little effect on development or swimming behaviour. Dopamine-containing neurones in the brain first differentiated at stage 35-36 and gradually increased in number up to stage 45-47, the latest stage examined. The functional role of 5-HT- or dopamine-containing neurones remains to be elucidated.
We conclude that cell-type-specific antibodies can be used to identify neurones and glial cells at early times during neural development and may be useful tools in circumstances where functional identification is difficult.
The mechanisms that control the differentiation of neurones and glial cells within the developing nervous system are very poorly understood. One way of approaching this problem is to use agents that may interfere with primary differentiation and to examine the consequences for neural development. Thus inhibiting the sodium pump during the midneural fold stages of development in the amphibian embryo prevents the subsequent differentiation of central nervous system neurones (e.g. Messenger & Warner, 1979; Breckenridge & Warner, 1982). Treatment of cleavage-stage embryos with lithium has the opposite effect, potentiating rather than inhibiting the differentiation of nerve cells (Breckenridge et al. 1987). In such experiments, it can be important to have markers that allow the identification of neural and glial cells as soon as possible after differentiation. Studies that examine the capacity of different embryonic cells to differentiate in novel environments (e.g. Heasman et al. 1984) also rely heavily on histological identification of different cell types.
Functional neurones appear early during the development of Xenopus. Endplate potentials driven by innervating motor neurones can be recorded in the myotomes shortly after the neural tube has closed, at developmental stage 21-24, although functional innervation can occur as early as stage 19, while the neural tube is closing (Blackshaw & Warner, 1976a,b). Flexions of the body initiated by touching the skin also first appear at about stage 21–24 (Muntz, 1975), implying that Rohon-Beard cells differentiate equivalently early. We have used antibodies to chart the immunohisto chemical recognition of neurones within the developing nervous system of Xenopus in order to establish whether such antibodies can provide an early marker of neuronal differentiation. We have compared the timetable for Xenopus with that for a urodelean amphibian, the axolotl. Antibodies to glial fibrillary acidic protein have been used to provide time tables for the immunohistochemical recognition of differentiated glial cells within the developing brain and spinal cord.
We include also studies on the appearance of the neurotransmitter molecules 5-hydroxytryptamine and dopamine within the developing nervous system of Xenopus.
Materials and methods
Adult Xenopus laevis (Xenopus Ltd) were induced to mate and lay eggs by injection of chorionic gonadotrophin (Pregnyl, Organon or Chorulon, Intervet, females: 600 i.u., males 400 i.u.). Embryos were reared in tap water at 16-22°C until they reached the required stage and staged according to the Normal Life Table (Nieuwkoop & Faber, 1956).
Axolotl embryos were obtained from spontaneous matings of mature adults. We are indebted to Dr N. Holder for donating eggs from his own supplies.
Glial cells were identified with a rabbit polyclonal antibody against glial fibrillary acidic protein (GFAP) (Pruss et al. 1981; Jessen & Mirsky, 1980; gift of M. Raff) or with a rabbit polyclonal raised against fish optic nerve (anti-Band 3, Maggs and Scholes, 1986; gift of J. Scholes) that recognizes fish vimentin and stains astrocytes in fish optic nerve. An antibody to vimentin (Hynes and Destree, 1974; gift of D. Lawson) was used to compare the staining patterns of Band 3, GFAP and vimentin antibodies.
Neural cells were recognized with a rabbit polyclonal antibody against the Mr 70K fragment of neurofilament protein (gift of D. Bray) and compared with the staining produced by RT97 (Anderton et al. 1982; gift of B. Anderton) and a rabbit polyclonal against fish optic nerve that recognizes a low molecular weight form of neurofilament protein (antiBand 2; Maggs and Scholes, 1986; gift of J. Scholes).
The distribution of 5-hydroxytryptamine was determined with a rat monoclonal anti-5-HT, supplied as a purified 1gG preparation (Sera-Lab) or with a rabbit polyclonal anti-5HT (Immunodiagnostics Ltd). Staining patterns were identical with both antibodies.
A mouse monoclonal antibody raised against a dopamine-glutaraldehyde-BSA conjugate (Geffard et al. 1984; gift of M. Geffard) was used to recognize dopamine.
Unfixed embryos were used for the study of the appearance of intermediate filaments. For recognition of 5-HT and dopamine, embryos were fixed in either 4% paraformaldehyde, 0·1 M-phosphate buffer, pH7·5 (5-HT) or 5 % glutaraldehyde, 0·lM-sodium cacodylate, pH 7·5 containing 10 mg ml− sodium metabisulphite (dopamine) for 2–3 h at room temperature. Fixed embryos were washed for 2–3 h in 5mM-Tris-HCl and incubated in 5% sucrose in 0·1 M-phosphate buffer with 0·1 % sodium azide at pH 7·5. 10 mg ml−1 sodium metabisulphite was included in this and all subsequent steps when the dopamine distribution was to be examined.
Embryos were embedded in O.C.T. (Miles Scientific), frozen in isopentane at —70°C and sectioned at 8-10μm on a Bright Cryostat. The sections were collected on gelatinized slides and stored in an airtight box at -20°C until they were stained.
For immunofluorescent staining the sections were incubated as follows: (i) 30 min in blocking serum, (ii) three 10 min washes in buffer, pH 7·5. (iii) Overnight at room temperature in primary antibody appropriately diluted in buffer with 0·1 % Triton X-100, 2% fetal calf serum (Sera Lab), 0·1 % sodium azide, pH 7·5. (iv) 2–3 h at room temperature with appropriate biotinylated second layer antibody (Vector) or FITC-labelled goat anti-rat IgG (Nordic Immunology Ltd) followed by 3×10 min washes in buffer, (v) biotinylated sections incubated in FITC-Streptavidin (Vector) diluted 1:100 in buffer with 10 % foetal calf serum. After a final 30 min wash in buffer, sections were mounted either in Citiflour (City University, London) or 90% glycerol containing 10 mg ml−1p-phenylene-diamine to reduce fading.
Alternatively, either the PAP (peroxidase-antiperoxidase) or the ABC Vectastain peroxidase kit was used to recognize the serotonin antibody. For the PAP method, the second layer antibody was either sheep anti-rabbit 1gG (Nordic, 1:40) or goat anti-rat 1gG (Nordic, 1:100) in 0·lM-buffer at pH7·5, followed by 2 h incubation at room temperature with peroxidase-antiperoxidase complex (Dakopatts, Copenhagen) diluted 1:50 in buffer. Sections were incubated for 10 min in 0·05M-Tris, pH 7·5, containing 0·5 mg ml−1 diaminobenzidine (DAB, Sigma), transferred to this solution with 0·006% hydrogen peroxide (Analar, BDH) for 10 min. After two 10 min washes in phosphate buffer the sections were mounted in glycerol.
For the ABC method, the sections were incubated with 0·3% hydrogen peroxide in methanol for 30 min, then incubated with primary antibody overnight. After washing, biotinylated antibody was applied for 1–2 h at room temperature, sections incubated for 1 h at room temperature in Vectastain ABC reagent and the stain developed for 2–6 min in 0·1 M-Tris, pH 7·2 with 0·02% hydrogen peroxide and 0-1% DAB.
Sections were examined on a Zeiss microscope equipped with incidence fluorescence. Photographs were taken with Tri-X Pan 400 (fluorescence) or Plus-X Pan 125. The results are based on examination of at least 4 embryos of the appropriate stage.
The dorsal parts of 40–80 embryos at stage 35/36 were manually homogenized on ice and spun for 3 min at 11600g in an Eppendorf centrifuge to precipitate intact yolk platelets. Intermediate filament protein was precipitated from the supernatant with 1% Triton in 0·1 M-phosphate buffer (pH 7·5) for 10 min at 4°C. The Triton-insoluble material was spun down in an Eppendorf for 10 min, resuspended in loading buffer (0·0625 M-Tris-HCl pH 7·6, 2% SDS, 10% glycerol, 5% β-mercaptoethanol, 0·001% bromophenol blue), boiled for 5–10 min and spun to remove any solid material. Samples were loaded onto a 10% polyacrylamide gel. The protein was transferred onto nitrocellulose, blocked with 3 % BSA and stained with the appropriate antibody at 1:25. The distribution of antibodies was recognized with a Protein A-125I complex. Intermediate filament preparations from fish optic nerve were prepared as in Maggs & Scholes (1986) and stained with the same antibodies for comparison. We are indebted to C. R. Green, A. Maggs and J. Scholes for advice.
The antibody against the 70K fragment of neurofilament protein recognized a single band at Mr 70K on immunoblots of intermediate filaments extracted from the dorsal parts of Xenopus embryos (Fig. 1A). The antibody against fish optic nerve neurofilament protein (Maggs & Scholes, 1982), which identifies a single band at Mr 61K in the fish (Fig. 1C), recognized a band of similar size in Xenopus, but also stained higher Mr bands of unknown identity (Fig. IB).
The pattern of staining in the neural tube at stage 36 with these two antibodies was identical (Fig. 2C,D). Furthermore it matched precisely the staining observed with RT97, a mouse monoclonal against the 200K neurofilament protein (Anderton et al. 1982; see also Breckenridge et al. 1987; Godsave et al. 1986). Thus the additional bands recognized by anti-Band 2 in Xenopus did not produce staining of cells in the spinal cord other than neurones. Fig. 3 shows staining with anti-70K at stage 36, where it recognizes neurites in the white matter of ventrolateral regions of the developing brain.
With all three antibodies some variable staining was frequently observed in the epidermal layers, although not in any other region of the embryo. For anti-Band 2, this suggests that the other proteins apparent on immunoblots are of epidermal origin. But the labelling with the anti-70K fragment, which identified a single 70K band only, and RT97 (see Breckenridge et al. 1987) implies that some ectoderm cells possess intermediate filaments that share an epitope with neurofilament protein. This epitope is not recognized by the antibody 2f7.c7 (Jones, 1985), which specifically labels epidermal cells, and can be used together with RT97 or anti-70K to identify any epidermal cells expressing the epitope recognized by neurofilament antibodies, as in Breckenridge et al. (1987). Otherwise the neuronal staining with anti-70K and anti-Band 2 was specific and identical and either antibody could be used to follow the appearance of neurofilament protein within developing neurones. All the results reported here give patterns of expression of neurofilament protein determined with both antibodies. RT97 proved of limited use at early stages (see below).
At stage 36, glial cells in the spinal cord could be identified by: anti-Band 3 (raised against fish optic nerve, Maggs & Scholes, 1986), anti-GFAP (Pruss et al. 1981) and anti-vimentin (Hynes & Destree, 1978; gift of D. Lawson). Fig. 3B-F shows the pattern of staining with anti-GFAP in the spinal cord (B), brainstem (C), and forebrain (E,F). In the neural tube and brain stem, the antibody reveals extensive radial glial cells, with GFAP present throughout the cytoplasm. The marginal zone contains a wide band of overlapping end-feet, which must provide a dense glial substrate for elongating axons. A different staining pattern was apparent in the brain. Fig. 3E,F show anti-GFAP staining on a complete transverse section through the forebrain (E), while F illustrates a higher power view of the ventrolateral margin of the forebrain in the adjacent section. Dense staining is apparent in the marginal zone and the filamentous nature of brain glia is apparent in the mantle zone. However no GFAP is apparent in the deeper layers of the developing brain, or in the ependymal layer. Immunoblots showed that anti-Band 3 and anti-GFAP recognized bands of the appropriate size in Xenopus intermediate filament preparations. Anti-GFAP staining was restricted to the neural tube and specific for glial cells. We have therefore used GFAP antibodies to identify the distribution of glial cells.
The initial appearance of immunohistochemically recognizable neurones and glia
In Xenopus, staining of neurones in the neural tube was occasionally found as early as stage 21, and was routinely present by stage 23. Examples of the staining pattern seen with the anti-70K and Band 2 antibodies in the neural tube and brain at stage 23 are shown in Fig. 2E,F. Fig. 2E illustrates weak neurofilament staining, recognized with the anti-70K antibody, at the ventrolateral margin of the developing brain. Serial sections showed that at stage 23 neurofilament staining was restricted to the caudal end of the brain. Fig. 2F shows a horizontal section through the neural tube at stage 23 stained with anti-Band 2; anti-70K produced identical staining. Filamentous axons are visible in the developing white matter. Stained axons were found along the length of the segmented somites, but did not extend back into the segmenting region.
By stage 28, neurofilament staining was obvious at all levels of the rostrocaudal axis and extended to latero-dorsal regions of the developing brain (see Breckenridge et al. 1987). As the embryo progressed through to stage 36, the general pattern of neurofilament staining did not change, but became more extensive in both brain and spinal cord (see Figs 2 and 3).
An interesting difference emerged between the ability of.the various antibodies to stain neurofilaments at different stages of development. Both 70K and the Band 2 neurofilament antibodies stained from the early 20s onward. However, prior to stage 28, RT97 always failed to recognize neurofilaments. At stages 28-33, by which time both 70K and Band 2 antibodies stained axons extensively and reliably, staining with RT97 was apparent (e.g. Breckenridge et al. 1987), but capricious. Reliable staining of neurofilament protein with RT97 was only obtained at relatively late stages, long after functional neuronal differentiation. This probably accounts for the failure of Godsave et al. (1986), who used RT97, to identify neuronal staining at or before stage 33/34, despite the functional evidence for differentiated neurones.
Fig. 4 compares staining in the neural tube at stage 23 with anti-Band 3 (B), anti-GFAP (transverse section, D; horizontal section, E) and anti-vimentin (F). In all cases staining is apparent in the marginal zone. None of the antibodies reliably revealed staining elsewhere in the neural tube, although occasionally some cytoplasmic staining across the tube was apparent (e.g. F) and at the marginal zone (e.g. C). This variability decreased as the embryos matured and by stage 25 the whole of each radial glial cell could be recognized. Prior to stage 23, staining with anti-GFAP antibodies was sparse and variable, although both the anti-Band 3 and anti-vimentin antibodies generated a staining pattern very similar to that illustrated in Fig. 4. This fits with the likelihood that the Band 3 antibody recognized a form of vimentin (see Maggs & Scholes, 1986), rather than GFAP and suggests that vimentin is expressed at an earlier stage than the glial-specific intermediate filament. In keeping with this suggestion, anti-vimentin antibodies stained neural plate cells at stage 17, although no reproducible pattern of staining was apparent that might allow distinction between neuronal and glial precursors. At later stages the pattern of anti-vimentin staining always paralleled GFAP staining, rather than neurofilament staining. If vimentin is produced in differentiating neurones, as suggested by Godsave et al. (1986), the antibodies used here did not recognize the appropriate epitope(s).
GFAP staining was also present in the ventrolateral marginal zone in caudal regions of the brain at stage 23.
Anti-vimentin and anti-Band 3 antibodies showed scattered and irreproducible staining of other structures in the embryo, particularly the epidermis (both) and somites and notochord (vimentin only). By contrast, anti-GFAP stained glial cells only. We conclude that differentiated glial cells can first be recognized histo-chemically at stage 23 with anti-GFAP, closely similar to the time when neurofilament staining was first apparent.
Comparison with the axolotl
The pattern of neurofilament and GFAP staining was determined also in the axolotl, which does not display spontaneous movements until late tailbud stages and functional innervation of the myotomes is apparently delayed until this stage (Keeter et al. 1975).
Fig. 5 shows adjacent transverse sections through the developing brainstem taken through an axolotl embryo at stage 23. The location of the two illustrated regions relative to the whole section is given in A, which reproduces diagrammatically the pattern of GFAP staining (B), which was apparent throughout the ventrolateral margins of the brainstem and in the neural tube. By contrast, anti-Band 2 or anti-70K staining was sparse and patchy (e.g. C) and, if present at all, was located at the ependymal margins of the cells only. By stage 36, anti-GFAP stained glial cells in the brain strongly, with a virtually identical pattern of staining to that seen in Xenopus (compare Fig. 5D with Fig. 4C,D). The pattern of axonal staining, revealed with anti-70K or Band 2 antibodies, was closely similar to that observed in Xenopus (compare Fig. 5E with Fig. 4E). Thus neurofilament staining appeared later in the axolotl than in Xenopus, radial glial cells clearly differentiated first.
The development of 5-HT-containing neurones
Nerve cell bodies containing 5-HT were first apparent in the ventral midline in rostral regions of the brain at stage 26. By stage 28, the number of stained cell bodies had increased and axonal processes were visible also. Fig. 6A shows a diagrammatic side view of the brain, giving the location of the 10-12 5-HT-containing cell bodies and their axons at stage 28, constructed on the basis of serial sections through a number of embryos. B shows a transverse section, taken at the end of the hindbrain, showing the lateroventral location of the cell bodies, while C illustrates a horizontal section, taken through a string of cell bodies. At this stage, there is little axonal outgrowth, although axons beginning to extend in a dorsorostral direction could be recognized (D).
The number of neurones containing 5-HT increased with age and by stage 35/36 a string of lateroventral neurones had differentiated, extending forward as far as the optic cups. An extensive set of 5-HT-containing axons was apparent running both rostrally into the brain and caudally into the spinal cord (Fig. 7A). B, D shows horizontal (B) and transverse (D) sections taken through the neuronal cell bodies. The growth cones of axons extending through the marginal zone towards the midbrain were clearly recognizable (C) as were axons running in the lateral tracts of the developing spinal cord (E). F shows a section from an embryo that had been exposed to 5,7-dihydroxytryptamine (5,7-DHT, Baumgarten et al. 1973) by making a small hole in the belly wall at stage 25, to allow drug penetration into the intercellular spaces. The drug prevents the synthesis of 5-HT and, as expected, 5-HT-containing axons were no longer visible. 6 other embryos treated similarly showed complete depletion of 5-HT-containing neurones.
A set of 5-HT-containing axons crossed the developing optic chiasma. The course of these axons at stage 37 is traced with a set of serial transverse sections in Fig. 8.
A shows a composite diagram illustrating the full distribution of 5-HT-containing neurones. B illustrates diagrammatically the location of the axons seen in transverse section (C,D) as they dive down towards the optic chiasma. The dense array of 5-HT-containing axons as they cross over in the chiasma is shown in E,F.
The very early appearance of these 5-HT-containing neurones suggests that they may reflect a class of nerve cells involved in the generation of swimming movements. A number of embryos reared in 5,7-DHT, and therefore depleted of 5-HT, were examined for any obvious defect in swimming. The only apparent difference between these tadpoles and their controls at stage 36/37 was the length of the burst of swimming initiated by touching the epidermis. In the 5-HT-depleted tadpoles, swimming was initiated more readily and lasted for longer than in controls.
Dopamine was first apparent in axons at stage 35/36, and, despite careful serial sectioning of a number of tadpoles, we were unable to locate any positive cell bodies. Fig. 9A gives a reconstruction of the location of these axons, based on serial sections through a number of embryos. The staining was sparse, but clearly visible in both transverse (Fig. 9C) and horizontal sections (Fig. 9E).
Dopamine-containing cell bodies were found in ventrolateral regions of the brain at the level of the optic cups at stage 37. Fig, 9B gives the location of these cell bodies and their axons, while D shows a few of the cell bodies in a transverse section at this level. Their axons ran back towards the spinal cord in the lateroventral tracks of the hindbrain (Fig. 9F).
By stage 39, the number of dopamine-containing neurones had increased, and because of forward growth of the brain, were now located anterior to the developing eyes (Fig. 10A). Two transverse sections show some of the brightly staining cell bodies (B) and their axons coursing back towards the spinal cord (C). D shows a horizontal section through the axon tracts in the hindbrain. The axons possess marked varicosities, visible also in those that had grown back into the spinal cord (E). At stage 45-47, the number of dopamine-containing neurones had increased substantially. Fig. 11A gives a composite side view of the brain to show the rostrocaudal location of the cell bodies while B shows that there were two groups of neurones, on either side of the midline at the same level in the midbrain. The rest of the figure illustrates sections taken from a complete serial series, beginning at the forebrain (C), anterior to the cell bodies, and showing dopamine-containing fibres crossing the midline in the marginal zone. D, E, F and G are transverse sections taken through the cell bodies to show the relative positions of the dopamine-containing neurones. H and I give transverse sections through the axon tracts running in the hindbrain towards the spinal cord. A small number of axons extend well back into the spinal cord (J).
Differentiated neurones can be recognized immunohis-tochemically within the developing nervous system of Xenopus close to the time when functional studies first show that physiologically differentiated neurones are present. Radial glial cells, expressing GFAP, can be recognized immunohistochemically as early as (Xenopus’), or before (axolotl), neurones. These results imply that antibodies to neurofilament protein and GFAP provide early markers of the neuronal and glial phenotype that can be used to identify neurones and glia in circumstances where physiological identification may be difficult, or impossible.
This conclusion differs from that previously drawn by Godsave et al. (1986), who did not obtain GFAP staining before the swimming tadpole stage and do not report staining with neurofilament antibodies until NF stage 48, approximately 3 days later than found here. A number of factors seem likely to have contributed to these differing results. The use of unfixed, rather than fixed, cryosections may be important, since even mild fixation can reduce antigenicity. The increased sensitivity afforded by using biotinylated reagents to recognize bound antibodies may have been an additional factor. However, the major cause of the discrepancy almost certainly lies in the antigens recognized by different neurofilament and GFAP antibodies. We found that antibodies against the larger neurofilament proteins, or those recognizing a phosphorylated form of neurofilament protein, such as RT97 (Anderton et al. 1982), stained the early amphibian nervous system capriciously and did not give reliable neuronal staining until the swimming tadpole stage, by which time the nervous system is relatively well-developed functionally. Antibodies recognizing the 70K neurofilament protein identified neurones much earlier than RT97 and the antibody raised against fish optic nerve neurofilament protein (Band 2, Maggs & Scholes, 1986) was equivalently reliable. The finding that RT97 only stains neurones reliably at relatively late stages implies that neurofilament protein is modified during development. Since RT97 recognizes a phosphorylated form of the 200K neurofilament protein, the simplest explanation is that initially neurofilament proteins are not phosphorylated, but become so as development proceeds.
Developing glial cells expressed vimentin, an intermediate filament protein very similar to that found in fish optic nerve glial cells (possibly fish vimentin, Maggs & Scholes, 1986) and glial fibrillary acidic protein. GFAP stained at later stages than either anti-vimentin or Band 3 antibodies, suggesting that the pattern seen with GFAP was most likely to represent differentiated glial cells. This conclusion is reinforced by the observation that the anti-vimentin antibody stained cells in the neural plate at stage 17, when commitment to particular classes of cell in the nervous system is probably still in progress (Messenger & Warner, 1979; Breckenridge & Warner, 1982). Since we found reliable staining of glial cells with GFAP antibodies from very early stages of neural development, it is surprising that Godsave et al. (1986), using an antibody to rat brain GFAP, saw no staining until stage 33/34. Presumably this reflects antibodies against different epitopes in the polyclonal antibodies used. It could suggest that GFAP becomes modified as the nervous system gradually matures.
The most interesting finding was the early recognition of differentiated glial cells, both in the brain and spinal cord. The rapidity of development in Xenopus made it difficult to separate the time of initial appearance of differentiated neurones and glia, but in the axolotl, which develops more slowly, cells staining for GFAP were clearly present before neurones. This implies that commitment to a glial, rather than a neuronal, fate occurs at about the same time, as or possibly before, commitment to neurones. The evidence is that commitment to differentiate into the neuronal phenotype takes place before the neural tube has closed.
The pattern of GFAP staining in the developing spinal cord is very different from that seen in the adult amphibian. In the adult spinal cord, GFAP is expressed only in that part of the glial cell that lies in the marginal zone (Miller & Luzzi, 1986). Presumably the pattern of expression of GFAP changes at some relatively late stage of development.
The time course of appearance of 5-HT-containing neurones is very similar to that charted by Van Mier et al. (1986). However, in addition to the cells in the raphe nucleus, we found a group of neurones that sent axons to the opposite side of the brain through the optic chiasma. The functional role of these 5-HT-containing neurones remains to be defined, although their very early appearance, similar to that for neurones containing the neurotransmitter glycine (Dale et al. 1986), suggests that they may be involved in controlling the swimming movement of the young tadpole. However, any such role must be restricted to subtle modulation of swimming, since depletion of 5-HT had a rather small effect on the ability of the tadpoles to swim. Dopamine-containing neurones appear relatively late and are not present in substantial numbers until the tadpole is approaching metamorphosis. The functional role of such dopamine-containing neurones remains to be elucidated.
This work was supported by the Medical Research Council and the Royal Society. We are indebted to B. Anderton, D. Bray, M. Geffard, M. Raff and J. Scholes for generous gifts of antibody. C. R. Green, A. Maggs and J. Scholes gave us much advice and assistance with the immunoblotting.