Notochordless Xenopus embryos were produced by u.v. irradiation of the uncleaved fertilized egg. The spinal cords were examined using intermediate filament staining for glial cells, retrograde HRP staining for neuronal morphology and an anti-glycinergic antibody to reveal commissural cells and axons. The floorplate cells of the normal cord appear to be absent and their position along the ventral midline of the cord is occupied by motor neurones, Kolmer-Agduhr cells, radial glial cells and a ventrally placed marginal zone containing the longitudinal axons. Motor neurone number is reduced to 15 % of control values, and the sensory extramedullary cell number is increased twentyfold. Commissural axons are still able to cross the ventral cord but do so at abnormal angles and some commissural axons continue to grow circumferentially up the contralateral side of the cord rather than turning to grow longitudinally. Extracellular electrophysiological recordings from motor axons reveal that the normal alternation of locomotor activity on the left and right side of the embryo is lost in notochordless animals. These results suggest that the notochord and/or the normal floor plate structure are important for the development of the laterality of spinal cord connections and may influence motor neurone proliferation or differentiation.

The vertebrate neuroectoderm forms as a result of one or more inductive interactions between mesoderm and ectoderm around the time of gastrulation. The exact nature of these interactions is poorly understood, although it appears that some mesodermal components are not only able to induce a broad ‘neural’ identity in the adjacent ectoderm cells, but are also able to specify which neural structures will develop. That is, mesoderm may impose positional values on the neuroectoderm (see reviews by Holtfreter and Hamburger, 1955 and more recently by Saxen, 1989).

One of the possible interactions between mesoderm and neuroectoderm is that the notochord may induce the development of a specialised group of ventral neuroepithelial cells known as the floor plate. In the early stages of neural tube development, this ventral midline structure is composed of simple cuboidal or columnar cells that span the thin floor beneath the neural canal. The basal surfaces of these cells are in close contact with the notochord. The evidence for an inductive interaction is twofold. First, at appropriate developmental stages, ectopically grafted notochords are capable of inducing secondary, ectopic floor plate structures in the lateral walls of the neural tube of chick embryos (van Straaten et al. 1988; Placzek et al. 1990b, Yamada et al. 1991)). Second, surgical removal of the chick notochord results in floorplateless neural tubes (Placzek et al. 1990b, Yamada et al. 1991) and, in amphibians, neural tubes that develop in the absence of notochord appear not to develop floor plate structures, rather the ventral midline of the tube is composed of a thick basal mass of cells and axons (Holtfreter and Hamburger, 1955; Youn and Malacinski, 1981).

In the present paper, we examine in detail the neuroanatomical and functional consequences of neural tube development in notochordless Xenopus embryos. Several recent reports suggest that the floor plate may play one or more important roles in neuronal development; therefore, if the floor place is absent from spinal cords in notochordless Xenopus embryos, this may have profound effects upon neural tube organisation. We have paid particular attention to three related aspects of neural tube development:

(1) The floor plate

We have used morphological criteria, ultrastructure and intermediate filament staining to determine whether the floor plate is missing from notochordless embryos and how the other neuroepithelial cells (radial glia) are reorganised.

(2) Sensory and motor neurone development

It has been suggested (Jessell et al. 1988; Wagner et al. 1990; Yamada et al. 1991; Hirano et al. 1991) that the floor plate, perhaps together with the underlying notochord, may play a role in the patterning of the dorsoventral axis of the neural tube. That is, cells that differentiate close to the floor plate will develop ventral cell characteristics, and cells that differentiate distant from the floor plate will develop dorsal cell characteristics. Floor plateless neural tubes may thus be more ‘dorsalised’ than normal. We have examined this proposal by analysing the number, position and morphology of dorsal sensory neurones (Rohon-Beard cells and extramedullary cells, see Clarke et al. 1984) and ventral neurones (motor neurones).

(3) The development of commissural axons

Commissural neurones have a ventrally directed axon, which crosses over to the contralateral marginal zone by growing through the basal regions of the floor plate cells. In vitro experiments suggest that floor plate cells can exert a major influence over the growth of commissural axons, since in rats the floor plate has been shown to release a chemotropic factor capable of redirecting commissural axons within a collagen gel (Tessier-Levigne et al. 1988) and within cultured neuroepithelium (Placzek et al. 1990a). Also, in situ, the change in direction of commissural axon growth, from circumferential growth round the ipsilateral cord and across the floor plate to longitudinal growth within the contralateral marginal zone, occurs at or close to the lateral border of the floor plate (Bovolenta and Dodd, 1990) and may result from a change in the expression of specific cell surface glycoproteins (and hence cell affinities) on commissural axons (Dodd el al. 1988). The floor plate may be the source of the signal that initiates the change in glycoprotein expression, and may thus act as an intermediate target in directing commissural axon growth. Thus if floor plate development is disturbed by inhibiting notochord development this may have important consequences for the growth of commissural axons. We have used an antibody that is specific for glycinergic commissural neurones (Roberts et al. 1988) to assess the growth of their axons in notochordless embryos.

Notochordless embryos can be produced by ultraviolet irradiation of the vegetal pole of fertilized Xenopus laevis eggs before their first cleavage. This procedure produces mesodermal deficits in a dosedependent manner (see Fig. 1), and can result in a variety of embryonic types, ranging from a completely vegetalised embryo containing no axial mesoderm, to a nearly normal individual whose mesodermal deficits are limited to partial or complete deletion of the notochord and/or prechordal plate. (Malacinski et al. 1977; Youn and Malacinski, 1981). Notochord and prechordal plate development are thus most susceptible to the effects of u.v. irradiation, which is consistent with the view that irradiation disrupts the distribution of mesoderminducing factors within the vegetal pole of the egg, and that the development of the notochord and prechordal plate requires the highest concentration of these factors (Scharf and Gehart, 1983).

Fig. 1.

Xenopus embryos at stage 35/36 (A) and stage 29/30 (B) or their u.v.-treated equivalent. From top to bottom: normal embryo, IAD Type 1/2, IAD Type 3, IAD Type .4. Lines on background grid are 1mm apart.

Fig. 1.

Xenopus embryos at stage 35/36 (A) and stage 29/30 (B) or their u.v.-treated equivalent. From top to bottom: normal embryo, IAD Type 1/2, IAD Type 3, IAD Type .4. Lines on background grid are 1mm apart.

Our results confirm that the structure of the spinal cord in notochordless embryos is abnormal and, according to anatomical criteria and intermediate filament expression, the normal floor plate cells are absent. Retrograde horseradish peroxidase staining shows that all the normal neuronal classes are present in experimental cords, but for some their position, morphology and number are disturbed. Unlike the normal cord, which has two lateral marginal zones, all longitudinal axons in the experimental cord run in a single marginal zone that covers the ventral and ventrolateral surfaces of the cord. Within this zone, however, relative axon position is normal. Glycinergic commissural axons are still able to cross the ventral surface of the cord, but often do so at unusually shallow angles and, in some cases, do not turn to run longitudinally along with the other contralateral axons, but continue to grow circumferentially up to the dorsal surface of the cord. Electrophysiological evidence shows that, although the experimental cords are able to generate a rhythmic motor output, they are unable to generate a normal alternating rhythm on the left and right sides, rather the whole cord acts like a single-sided preparation where all motor neurones fire nearly synchronously. These results may all be consequent upon removing the normal interaction between notochord and neuroepithelium, although the disruption of other neural induction mechanisms cannot be ruled out.

Egg production and u.v. irradiation

Eggs were obtained by injecting female Xenopus laevis with 500i.u. hCG (Sigma) and then stripping twelve hours later. The eggs were fertilised by gentle rubbing with macerated testes and were then covered with 1/10 Barth X solution (Gurdon, 1977). Control eggs were left to develop at room temperature. Experimental eggs were left to rotate until the pigmented animal pole was clearly uppermost and then transfered to a quartz glass dish positioned 7 cm above the ultraviolet lamp (a mercury pen ray lamp – Ultra Violet Products Ltd, Cambridge, UK). Eggs were irradiated for 3 min at 254 nm and then transferred to an aerated container of 1/10 Barth X at room temperature.

Wax histology

Embryos were fixed in Perfix (Fisher Diagnostics) for 2h, dehydrated through Lang’s alcohols (Lang, 1937), cleared and embedded in wax. Serial 8 μm transverse sections were stained with H and E.

Electron microscopy

Embryos were fixed in Karnovsky’s fixative for 2h, dehydrated, osmicated and embedded in Epon resin. 1 μm sections were stained with methylene blue and ultrathin sections were stained with lead citrate and uranyl acetate.

HRP

Embryos were anaesthetised in MS222 (Sigma) and a small area of skin removed to allow HRP (Boehringer grade 1) application. HRP was recrystallised onto the tips of electrolytically sharpened pins, and these were then pushed into appropriate regions of the CNS or into the myotome blocks. Embryos were allowed to recover in 1/10 Barth X for 3 to 6 h, before being reanaesthetised and fixed in 2.5% glutaraldehyde in 0.1M phosphate buffer, pH 7.4 (PB) for 1.5 h. The skin and sometimes the myotomes were removed from the CNS and the remaining tissue washed several times in PB, before being soaked in 0.04% diaminobenzidene (DAB, Sigma) for 15 min. H2O2 was added at 0.03% for a further 15min. The tissue was washed in PB, the brain and cord dissected free, dehydrated, cleared and mounted in DPX between two coverslips. Some specimens were sectioned in wax rather than prepared as whole mounts.

Immunohistochemistry

Glycine

Spinal cords were stained for glycinergic neurones as described previously (Dale et al. 1986: Roberts et al. 1988) with the exception that the ABC Vector Elite avidin/biotin system was used rather than the PAP method. The specificity of the staining was checked using test filters containing spots of various amino acid-tissue-glutaraldehyde conjugates (Dale et al. 1986). All immunoreactivity was removed by omitting the primary antibody step and by adding glycine-glutaralde-hyde conjugates to the primary antiserum.

Vimentin

Embryos were fixed for 4 h in a 2 % solution of trichloroacetic acid, washed in PBS, dehydrated and embedded in polyethyleneglycol 400 distearate wax (BDH). Transverse sections were cut at 10μm, dried onto glass slides and dewaxed through a series of graded acetones to PBS. Non-specific binding was blocked with 4% bovine serum albumin, 2% newborn calf serum in PBS for 15 min The anti-vimentin antibody, Z9 (supplied by Dr Chris Wylie), was diluted 1:100 and applied to sections overnight at 4°C. After washing in PBS the primary antibody was detected with a fluorescein-tagged anti-rabbit secondary antibody for 2h at room temperature. Sections were washed in PBS and mounted in DABCO (Sigma).

Electrophysiology

Detailed accounts of the methods used here have been published elsewhere (Kahn and Roberts, 1982). Briefly, the embryos were immobilised with curare (70–100μM d-tubocurarine) and then pinned to a Sylgard dish and superfused with Xenopus saline (Na, 115; K, 2.5; Ca, 2.0; Mg, 1.0; Cl, 121; HCO3, 2.5; Hepes, 10; all in mM, pH7.6). Suction electrodes applied to the intermyotomal clefts recorded ventral root activity and a stimulating suction electrode was used to activate sensory endings in the skin. Ventral root activity was recorded on tape and then played back through a thermal arraycorder (Graphtek) for analysis and presentation.

A grading system to describe the abnormalities characteristic of the u.v.-irradiated embryos has been proposed by Malacinski (Malacinski et al. 1974) and slightly modified by Scharf and Gerhart (1983). The grades or indices of axial deficiency (IAD) are:

0 – normal in all externally visible respects

1– reduced forehead; eyes smaller than normal

and sometimes joined

2– eyes fused or cyclopic; but at least some

retinal pigment visible

3– no visible retinal pigment; otic vesicles or

single vesicle still present

4– no otic vesicle present; somites present in

trunk or portion thereof

5– no somites present; trace of tail mesenchyme

occasionally seen

Nearly all u.v.-irradiated embryos used in our study were grade 1 or 2 (Fig. 1). Individuals with total or partial notochord deletions could be selected as this deficit results in a constriction of the body width just ventral to the level of the neural tube. The assignment of embryos into these grades is straightforward after 2 days of development, but less easy earlier before the development of clear head structures. It is possible, then, that some of the embryos analysed before 2 days may have fallen into the grade 3 category. This possibility is discussed in more detail later.

Normal cord structure

At 2 and 3 days of development the spinal cord is a gently tapering tube of approximately oval crosssection. The neurocoel is roughly triangular in crosssection and eccentrically placed towards the ventral surface of the cord (Fig. 2A). Most ependymal cells lining the neurocoel are radial glia and have branching processes that span the cord out to its surface. Along the lateral surface of the cord, these branching processes expand to form end-feet and interdigitate to form the glia limitans. Vimentin staining labels the peripheral branches of these processes (Fig. 2E). In contrast, the ventral surface of the neurocoel is lined by 3, 4 or 5 roughly cuboidal cells, which do not have branching processes, do not stain for vimentin, and form the floor plate.

Fig. 2.

(A) Photomicrograph of an unstained transverse section of a normal stage 35/36 spinal cord. Note lateral marginal zones and ventrally placed central canal. (B) Toluidine-blue-stained transverse section of a normal stage 35/36 IAD Type 1 spinal cord. Note central canal is now placed dorsally and the pale-staining marginal zone (*) lines the ventral half of the cord. (C,D) Electron micrographs of the ventral midline of Type 1 spinal cord shown in B. Note the characteristic radial glia-like endfeet, normally only seen in the lateral marginal zones, the large extracellular spaces and, in D, several axonal profiles lying immediately adjacent to the basal lamina (arrows). (E,F) Transverse sections of normal (E) and Type 1 spinal cord (F) stained with the anti-vimentin antibody Z9. Vimentin-positive radial processes and endfeet are present only in the lateral marginal zones of normal cords, but are present also across the ventral midline of the uv-treated cord. My, myotomes; not, notochord; fp, floor plate; c, central canal; mz, marginal zone; ef, endfoot. Scale bar in A is 50μm and applies to A, B, E and F, and is 1, μm in C and D.

Fig. 2.

(A) Photomicrograph of an unstained transverse section of a normal stage 35/36 spinal cord. Note lateral marginal zones and ventrally placed central canal. (B) Toluidine-blue-stained transverse section of a normal stage 35/36 IAD Type 1 spinal cord. Note central canal is now placed dorsally and the pale-staining marginal zone (*) lines the ventral half of the cord. (C,D) Electron micrographs of the ventral midline of Type 1 spinal cord shown in B. Note the characteristic radial glia-like endfeet, normally only seen in the lateral marginal zones, the large extracellular spaces and, in D, several axonal profiles lying immediately adjacent to the basal lamina (arrows). (E,F) Transverse sections of normal (E) and Type 1 spinal cord (F) stained with the anti-vimentin antibody Z9. Vimentin-positive radial processes and endfeet are present only in the lateral marginal zones of normal cords, but are present also across the ventral midline of the uv-treated cord. My, myotomes; not, notochord; fp, floor plate; c, central canal; mz, marginal zone; ef, endfoot. Scale bar in A is 50μm and applies to A, B, E and F, and is 1, μm in C and D.

All longitudinal running axons he within the two lateral marginal zones. Occasionally, the dorsal part of the marginal zones is separated from the rest by a superficial neuronal cell body (Roberts and Clarke, 1982). The majority of neuronal cells bodies He in the mantle zone between the ependymal layer and the marginal zones. Rohon-Beard cells are the major exception to this, and they form a cell column along the dorsal midline of the cord. Their superficial cell surfaces have a less extensive covering of neuroepithelial endfeet than other cord surfaces (Taylor and Roberts, 1983).

Notochordless embryo spinal cord

The structure of spinal cords that develop in the absence of a notochord differ from control cords in three obvious ways:

  1. The neurocoel lies in a more dorsal position, often with only a thin cellular bridge (less than 5 μm) separating it from the dorsal cord surface (Fig. 2B). Occasionally the neurocoel is much reduced in size and can be difficult to see in wax-sectioned material (Fig. 4D,F).
    Fig. 3.

    Camera-lucida drawings of motoneurones. (A) Side-view whole-mounted spinal cord from a normal embryo. Neurons filled from a myotome HRP application. Large black cell is a primary motor neurone and small black cell is a secondary motor neurone. Positions and sizes of other motor neurone cell bodies are shown as unfilled outlines. Rohon-Beard cells are shown stippled along the dorsal cord. Rostral is to the left. (B) Transverse sections from a normal embryo showing two large primary motor neurones and three smaller secondary motor neurones. Note dendrites are restricted to one side only. No motor neurones are found ventral to the neurocoel. (C,D,E) Ventral views of whole-mounted spinal cords from u.v.-treated embryos. Motor neurone.s and their caudally running axons lie on or close to the ventral midline. In C both cells have dendrites that clearly extend in both left and right halves of the marginal zone. Arrows indicate positions where axons exit the cord. The most rostral cell in C is unusual in that the motor dendrites overlap with the Rohon-Beard axons (shown dashed). One Rohon-Beard axon is shown crossing the cord ventrally; this was only seen once. (F,G) Transverse sections from u.v.-treated embryos showing motor neurones with clear bilateral dendrites. Rohon-Beard axons are shown as short wavy lines dorsolateral to the motor cells. Rohon-Beard and extramedullary cells are stippled. (H,I) The position of motor neurones (ventrally) and Rohon-Beard and extramedullary cells (dorsally) plotted onto standardised section outlines. H is from a normal embryo and I from a u.v.-treated embryo. Note the large number of cells outside the spinal cord in I. My, myotome; not, notochord. Scale bar is 50 μm for B, C, H, I and 30 μm for A, D, E, F, G.

    Fig. 3.

    Camera-lucida drawings of motoneurones. (A) Side-view whole-mounted spinal cord from a normal embryo. Neurons filled from a myotome HRP application. Large black cell is a primary motor neurone and small black cell is a secondary motor neurone. Positions and sizes of other motor neurone cell bodies are shown as unfilled outlines. Rohon-Beard cells are shown stippled along the dorsal cord. Rostral is to the left. (B) Transverse sections from a normal embryo showing two large primary motor neurones and three smaller secondary motor neurones. Note dendrites are restricted to one side only. No motor neurones are found ventral to the neurocoel. (C,D,E) Ventral views of whole-mounted spinal cords from u.v.-treated embryos. Motor neurone.s and their caudally running axons lie on or close to the ventral midline. In C both cells have dendrites that clearly extend in both left and right halves of the marginal zone. Arrows indicate positions where axons exit the cord. The most rostral cell in C is unusual in that the motor dendrites overlap with the Rohon-Beard axons (shown dashed). One Rohon-Beard axon is shown crossing the cord ventrally; this was only seen once. (F,G) Transverse sections from u.v.-treated embryos showing motor neurones with clear bilateral dendrites. Rohon-Beard axons are shown as short wavy lines dorsolateral to the motor cells. Rohon-Beard and extramedullary cells are stippled. (H,I) The position of motor neurones (ventrally) and Rohon-Beard and extramedullary cells (dorsally) plotted onto standardised section outlines. H is from a normal embryo and I from a u.v.-treated embryo. Note the large number of cells outside the spinal cord in I. My, myotome; not, notochord. Scale bar is 50 μm for B, C, H, I and 30 μm for A, D, E, F, G.

    Fig. 4.

    (A.B) Transverse sections of normal cords containing HRP-filied motor neurones. A shows two large, heavily filled primary motor neurones and one small, pale secondary (arrowed). B shows a large primary and its dendrites confined to the ipsilateral marginal zone. (C,D) Transverse sections of u.v.-treated cords containing large HRP-filled motor neurones along the ventral midline. In C a labelled Rohon-Beard cell is present in the dorsolateral cord. (E,F) Transverse sections of u.v.-treated cords containing HRP-labelled sensory and motor axons and several extramedullary cells (curved arrows) attached in lines or bunches to the dorsal cord surface. In E note that the sensory axons (filled straight arrows) occupy the dorsolateralmost region of the ventral marginal zone and the motor axon (open arrow) lies midventrally. My, myotomes; not, notochord; asterisk, melanocyte. Scale bar is 50 μm.

    Fig. 4.

    (A.B) Transverse sections of normal cords containing HRP-filied motor neurones. A shows two large, heavily filled primary motor neurones and one small, pale secondary (arrowed). B shows a large primary and its dendrites confined to the ipsilateral marginal zone. (C,D) Transverse sections of u.v.-treated cords containing large HRP-filled motor neurones along the ventral midline. In C a labelled Rohon-Beard cell is present in the dorsolateral cord. (E,F) Transverse sections of u.v.-treated cords containing HRP-labelled sensory and motor axons and several extramedullary cells (curved arrows) attached in lines or bunches to the dorsal cord surface. In E note that the sensory axons (filled straight arrows) occupy the dorsolateralmost region of the ventral marginal zone and the motor axon (open arrow) lies midventrally. My, myotomes; not, notochord; asterisk, melanocyte. Scale bar is 50 μm.

  2. Ventral to the neurocoel, there is no morphological evidence for any floor plate cells, i.e. no cuboidal cells span the distance to the ventral cord surface. This area is now filled with neurone cell bodies, radial glial cells with branching processes and Kolmer-Agduhr cells (see later).

  3. The ventral and ventrolateral surfaces of the cord are now fined with a single cresent-shaped marginal zone that contains the longitudinal axons. There is no obvious boundary between left and right marginal zone (Fig. 2B and 4E,F).

Vimentin staining and electron microscopy (Fig. 2C,D,F) indicate that the ventral marginal zone, including the midline, contains the branching processes of radial glia cells. The glia limitans, formed by the radial glial endfeet, appears to be less complete than in normal cords, so that occasionally neuronal processes are found contacting the basal lamina that lines the cord (Fig. 2D). The packing of neuronal processes, within the marginal zone is less dense than that of control specimens.

In serial wax sections, the dorsal surface of the experimental cords is less uniformly smooth than ventral and lateral surfaces. This is because cells are frequently found to lie outside but attached to the regular cord surface. Some of these dorsal cells are shown to be ectopic Rohon-Beard cells (see later).

Neurones revealed by HRP labelling

Normal cord (n=20)

The stage 37/38 (Nieuwkop and Faber, 1956) Xenopus embryo spinal cord has previously been extensively analysed by retrograde HRP staining (Roberts and Clarke, 1982). We have repeated this analysis on the slightly earlier stages 33 through to 36 to coincide more closely with the developmental stages of the u.v.-treated animals. We find that the neuronal types present at stages 33/34 and 35/36 closely resemble those in the stage 37/38 cord, and will thus describe these only briefly.

Three types of HRP application were used and each filled characteristic cell classes.

(1) Myotome HRP applications. This labels three classes of neurone; motor neurones, Rohon-Beard cells and extramedullary cells. Motor neurones (Figs 3A,B and 4A,B) lie in a ventrolateral or lateral position within the spinal cord, their dendrites extend into the ipsilateral marginal zone but do not enter the ventral floor plate or contralateral marginal zone, and their axons usually run caudally for a variable distance in the ventrolateral marginal zone before exiting in a ventral root. Later in development, axial motor neurones can be subdivided into two groups – primary and secondary. Primary motor neurones (Fig. 3A,B) are born first, have large cell bodies, extensive dendritic arbors and a synaptic relationship with Mauthner’s axon. Secondary motor neurones (Fig. 3A,B) are smaller, bom later, have less extensive dendritic arbors and are not contacted by Mauthner’s axon (Hughes, 1959; Blight, 1978; Forehand and Farel, 1982; van Mier et al. 1985; Nordlander, 1986). At early stages of development only the criterion of cell size is easily analysed and Fig. 7 demonstrates a skewed distribution of cell sizes that is not clearly separated into two groups. However, in the tail, at slightly later stages of development (stage 40), all primary motor neurons are larger than all secondary motor neurones (Nordlander, 1986); therefore, it is likely that the more numerous small cells of Fig. 7 are secondaries and the larger cells are primaries.

Rohon-Beard cells (Fig. 3A) lie on or close to the dorsal midline of the cord. They usually have large smooth cell bodies with both an ascending and a descending axon in the dosolateral part of the marginal zone. They also have a peripheral axon, which exits the cord dorsally and then runs over or through the myotomes to the skin.

Extramedullary cells are present in only small numbers (see Hughes 1957 and Roberts and Clarke, 1982). Their morphology is essentially the same as for Rohon-Beard cells except that their cell bodies lie outside the spinal cord, often suspended within the connective tissue dorsal to the cord or adjacent to the myotomes.

(2) Rostral cord/caudal hindbrain HRP applications. These label five neuronal classes; Rohon-Beard cells, ascending interneurones, dorsolateral commissural intemeurones, commissural interneurones and Kolmer-Agduhr cells.

Ascending intemeurones have an ascending ipsilateral axon and can have either unipolar or multipolar cell bodies with dendrites in the ipsilateral marginal zone. They were originally divided into two classes according to their dorsal or ventral position (Roberts and Clarke, 1982) but have since been classed together on the basis of their GABA immunoreactivity (Roberts et al. 1987).

Commissural interneurones (Fig. 5A) have unipolar cell bodies, a thick primary process that curves ventrally round the cord just inside the marginal zone, dendrites within the ipsilateral marginal zone and an axon that crosses the ventral floor plate to the contralateral marginal zone where it either ascends, descends or branches and does both. These neurones can be selectively stained with an anti-glycine antibody (Fig. 7 and Roberts et al. 1988). A few commissural interneurones also have an axon within the ipsilateral marginal zone (Roberts et al. 1988).

Fig. 5.

Camera-lucida drawings of neurones with commissural axons. (A) Side view, and (B) ventral view of whole-mounted spinal cords from normal embryos. Cells labelled from HRP applied to the contralateral rostral spinal cord. A shows the two morphological classes seen. Dorsolateral commissural intemeurones have multipolar cell bodies and a thin ventrally directed axon, while commissural intemeurones (stippled) have unipolar cell bodies and dendrites that arise from a thicker ventrally directed process. B shows that the axons from these cells cross the ventral surface at angles close to 90° to the long axis. (C) Side view and (D) ventral view of whole-mounted spinal cords from u.v.-treated embryos. Cells labelled from HRP applied to the contralateral rostral cord. C shows that both morphological classes of commissural neurone are present (compare with A). D shows that the angle at which their axon crosses the ventral midline (shown dotted) is more variable than in control spinal cords. Scale bar is 50μm.

Fig. 5.

Camera-lucida drawings of neurones with commissural axons. (A) Side view, and (B) ventral view of whole-mounted spinal cords from normal embryos. Cells labelled from HRP applied to the contralateral rostral spinal cord. A shows the two morphological classes seen. Dorsolateral commissural intemeurones have multipolar cell bodies and a thin ventrally directed axon, while commissural intemeurones (stippled) have unipolar cell bodies and dendrites that arise from a thicker ventrally directed process. B shows that the axons from these cells cross the ventral surface at angles close to 90° to the long axis. (C) Side view and (D) ventral view of whole-mounted spinal cords from u.v.-treated embryos. Cells labelled from HRP applied to the contralateral rostral cord. C shows that both morphological classes of commissural neurone are present (compare with A). D shows that the angle at which their axon crosses the ventral midline (shown dotted) is more variable than in control spinal cords. Scale bar is 50μm.

Dorsolateral commissural interneurones (Fig. 5A) have multipolar cell bodies adjacent to the Rohon-Beard cells with dendrites in the dorsal half of the marginal zone. A thin ventral axon runs ventrally just inside the marginal zone to cross the ventral floor plate and then ascend, descend or branch in the contralateral marginal zone.

Kohner-Agduhr cells (Fig. 6A) are ciliated cerebrospinal-fluid-contacting neurones, whose cell bodies lie within the ventral ependymal lining of the neurocoel close to the ventral floor plate cells. They have ipsilateral ascending axons and no dendrites.

Fig. 6.

Camera-lucida drawings of Kolmer-Agduhr cells. (A) Transverse section containing a Kolmer-Agduhr cell filled caudal to a rostral cord HRP application in a normal embryo. This cell has a typically simple morphology, contacts the cerebrospinal fluid and occupies a position adjacent or close to the floor plate cells. (B) Transverse section containing a Kolmer-Agduhr cell filled caudal to a rostral cord HRP application in a u.v.-treated embryo. This cerebrospinal fluid contacting cell has characteristic simple morphology but now occupies a position close to the midline beneath the neurocoel. (C) The positions of seven Kolmer-Agduhr cells from a u.v.-treated embryo have been superimposed on a typical transverse section. All but one occupy a position close to the midline. (D) Side view of whole-mounted spinal cord. In this u.v.-treated case, four Kolmer-Agduhr cells have been filled rostral to a caudal HRP application. This is never seen in normal embryos. The outline of the neurocoel is indicated, caudal is to the left and ventral at the bottom. Scale bar is 50 μm.

Fig. 6.

Camera-lucida drawings of Kolmer-Agduhr cells. (A) Transverse section containing a Kolmer-Agduhr cell filled caudal to a rostral cord HRP application in a normal embryo. This cell has a typically simple morphology, contacts the cerebrospinal fluid and occupies a position adjacent or close to the floor plate cells. (B) Transverse section containing a Kolmer-Agduhr cell filled caudal to a rostral cord HRP application in a u.v.-treated embryo. This cerebrospinal fluid contacting cell has characteristic simple morphology but now occupies a position close to the midline beneath the neurocoel. (C) The positions of seven Kolmer-Agduhr cells from a u.v.-treated embryo have been superimposed on a typical transverse section. All but one occupy a position close to the midline. (D) Side view of whole-mounted spinal cord. In this u.v.-treated case, four Kolmer-Agduhr cells have been filled rostral to a caudal HRP application. This is never seen in normal embryos. The outline of the neurocoel is indicated, caudal is to the left and ventral at the bottom. Scale bar is 50 μm.

Fig. 7.

Histograms to show the variability in motor neurone size in control and u.v.-treated embryos. The large numbers of small cells (less than 100 μm2) are absent from u.v.-treated embryos. Data collected from 5 control animals and 4 experimental animals. The cross-sectional area of cell bodies was measured from enlarged camera-lucida drawings of HRP stained motor neurones.

Fig. 7.

Histograms to show the variability in motor neurone size in control and u.v.-treated embryos. The large numbers of small cells (less than 100 μm2) are absent from u.v.-treated embryos. Data collected from 5 control animals and 4 experimental animals. The cross-sectional area of cell bodies was measured from enlarged camera-lucida drawings of HRP stained motor neurones.

(3) Caudal spinal cord HRP applications. These label five neuronal classes; Rohon-Beard cells, commissural and dorsolateral commissural intemeurones, descending intemeurones and motor neurones.

Descending interneurones are the only cell class to be specifically labelled by caudal HRP application; the other classes have been described above. Descending intemeurones have multipolar cell bodies, dendrites that extend throughout the ipsilateral marginal zone and ipsilateral descending axons.

Spinal cords from notochordless embryos (n=51) These will be described by dealing with each neuronal type in turn.

The number of Rohon-Beard cells within the cord filled from a standardised HRP application to the myotomes is the same in control and experimental embryos (Table 1). There is, however, a dramatic increase (approximately twentyfold) in the number of labelled extramedullary cells. These cells were often found together in bunches or streams of cells attached to the cord surface, but could also be found singly as far away from the cord as the dorsolateral edge of the myotomes (Fig. 31). In some cases, these streams or bunches of cells were continuous with Rohon-Beard cells within the cord, suggesting that they may be misplaced Rohon-Beard cells rather than a separate cell type. The irregular nature of parts of the dorsal cord surface and the sometimes obscured neurocoel, suggest that the mechanisms of neural tube closure have been incomplete, and this may have allowed for the abnormal migration of the Rohon-Beard cells.

Table 1.

A comparison of cell numbers in control and u.v.-treated embryos

A comparison of cell numbers in control and u.v.-treated embryos
A comparison of cell numbers in control and u.v.-treated embryos

The central axons of Rohon-Beard cells run in the most dorsolateral part of the marginal zone (Fig. 4E).

The number of labelled motor neurones in experimental cords is reduced to approximately 15 % of the numbers found in control cords (Table 1), as judged from standardised HRP applications to the myotomes. An analysis of size (Fig. 7) reveals that this reduction in number is due largely to a loss of the smallest cells. This is strongly suggestive that the motor neurones present in notochordless embryos are all primary motor neurones and that the smaller, later born secondary motor neurones are not present. Motor neurone position is also altered, with their cell bodies now placed midventrally and the dendrites from single motor neurones often extend into the marginal zone on both sides of the midfine (Fig. 3C,F,G and 4C,D). As in normals, the dendrites do not extend into the most dorsolateral part of the marginal zones and thus are not directly contactable by Rohon-Beard axons. Motor neurone axons run caudally along the ventral midline of the marginal zone before exiting the cord (Fig. 3C,D,E). Ventral roots form midventrally.

Retrograde HRP reveals that both commissural and dorsolateral commissural interneurones are present in notochordless embryos (Figure 5C,D). Their respective dendritic and proximal axon morphologies are normal. However, the angle at which their axons cross the ventral midline is often, but not always, abnormal. In normal cords most commissural axons cross the ventral surface in a very direct manner, i.e. at close to 90 degrees to the long axis of the cord, and thus close to the level of their cell bodies (Fig. 5B). This angle of crossing is much more variable in notochordless embryos, and although some axons still cross in a direct manner, many cross at more oblique angles and some at very shallow angles indeed (Fig. 5D). In the notochordless embryos, the commissural axons remain on the inside edge of the ventrally placed marginal zone, in contrast to those in normals that he close to the basal (superficial) surface of the ventral floor plate cells. The presence of some cells with extremely shallow crossing angles raises the possibility that some ‘would-be-commissural’ axons do not actually cross the ventral midline and these would thus not be recognised as commissurals on morphological grounds (they would be classified as ascending or descending interneurones). Any such ‘non-crossing commissural’ axons will, however, be specifically recognised by their glycinergic immunoreactivity (Roberts et al. 1988); see later section.

Most Kolmer-Agduhr cells have rostrally directed axons and are filled, as in normals, from rostral HRP applications. They are identified by having a ciliated CSF-contacting surface and a single process. In notochordless embryos, their cell bodies lie along the midfine, ventral to the neurocoel or within one cell diameter of the midline (Fig. 6B,C). Since the neurocoel is more dorsal than in normals, the Kolmer-Agduhr cells are correspondingly more dorsal. In some notochordless embryos rostral Kolmer-Agduhr cells are filled from caudal HRP applications (Fig. 6D); these cells therefore have caudally directed axons. This is never seen in control embryos.

Ascending interneurones and descending interneurones, both with ipsilateral axons, are revealed by HRP applications to the rostral cord or caudal cord, respectively (Fig. 8A,B,C). The dendritic organisation of these cells has not been analysed in detail, but most have a clear ventrally directed major process..

Fig. 8.

Interneurones. (A,B) Camera-lucida drawings of transverse sections containing neurones filled from rostral cord (in B) and caudal cord (in A) in u.v.-treated embryos. Notice how many cells have a clear dorsal to ventral polarity of their main process. Stippled cells are Rohon-Beard cells and extramedullary cells. Unfilled cell profiles are from cells with no clear processes and which are thus probably oriented in a rostrocaudal direction. (C) Photomicrograph of the side view of a whole-mounted u.v.-treated cord. HRP was applied to the rostral cord (to the left). Note the predominant dorsal-to-ventral orientation of most of the labelled cells. Scale bar is 50 μm for A and B, and 100 μm for C.

Fig. 8.

Interneurones. (A,B) Camera-lucida drawings of transverse sections containing neurones filled from rostral cord (in B) and caudal cord (in A) in u.v.-treated embryos. Notice how many cells have a clear dorsal to ventral polarity of their main process. Stippled cells are Rohon-Beard cells and extramedullary cells. Unfilled cell profiles are from cells with no clear processes and which are thus probably oriented in a rostrocaudal direction. (C) Photomicrograph of the side view of a whole-mounted u.v.-treated cord. HRP was applied to the rostral cord (to the left). Note the predominant dorsal-to-ventral orientation of most of the labelled cells. Scale bar is 50 μm for A and B, and 100 μm for C.

Glycinergic immunoreactivity

Normal cords (n=32)

The development of glycinergic neurones in normal Xenopus embryos has been described previously (Roberts et al. 1988). Our own observations on normal cords confirm this earlier description and will thus be summarised only briefly.

All glycinergic cells have a commissural axon. They have unipolar cell bodies and dendrites within the ipsilateral marginal zone. The axon is directed ventrally along the medial edge of the marginal zone, until it reaches the lateral edge of the floor plate where it then lies close to the superficial (basal) surface of the floor plate cells. For cells differentiating at the caudal end of the developing cell column, the initial growth of the axon within the contralateral marginal zone is almost always in a rostral direction. For cells that develop late within areas already containing glycinergic cells, some axons may initially turn caudally or branch to form a rostral and a caudally directed axon.

Notochordless embryos (n=38)

Glycinergic neurones are present in notochordless embryos in approximately normal numbers (Table 1). In almost all cases, the glycinergic cells had axons that crossed the ventral side of the cord to the contralateral marginal zone and, as in normals, they had unipolar somas with a thick proximal axonal process. These cells differed from those in controls in three ways:

(1) The angle of crossing was much more variable in experimental cords compared to normals (Figs 9 and 10). In normal cords, 66% (49/74) of measured axons crossed at angles that deviated less than 10° away from 90° to the long axis of the cord, and only 7 % (5/74) crossed at greater than 25° away from 90°. In experimental cords, these figures decreased to 20% (15/73) and increased to 52% (38/73) respectively.

Fig. 9.

Photomicrographs of the ventral surface of whole-mounted cords stained for glycinergic axons. A and B are from control embryos, A from the rostral cord/caudal hindbrain area and B from midtrunk cord. Note the stained axons cross the ventral cord in a fairly direct manner. C and D are from equivalent regions of a u.v. type 1 embryo. Note now how the angle of crossing is highly variable. Scale bar is 50 μm and applies to all.

Fig. 9.

Photomicrographs of the ventral surface of whole-mounted cords stained for glycinergic axons. A and B are from control embryos, A from the rostral cord/caudal hindbrain area and B from midtrunk cord. Note the stained axons cross the ventral cord in a fairly direct manner. C and D are from equivalent regions of a u.v. type 1 embryo. Note now how the angle of crossing is highly variable. Scale bar is 50 μm and applies to all.

Fig. 10.

Histograms to show the variability of angle of crossing of glycinergic commissural axons in control and u.v.-treated embryos. Angles were measured relative to the ventral midline of the spinal cord, those directed rostrally have values less than 90° and those directed caudally have values greater than 90°. Angles were allocated into bin widths of 10°. Note the increased variability in the experimental cords and the increased proportion that are directed caudally (hatched) rather than rostrally.

Fig. 10.

Histograms to show the variability of angle of crossing of glycinergic commissural axons in control and u.v.-treated embryos. Angles were measured relative to the ventral midline of the spinal cord, those directed rostrally have values less than 90° and those directed caudally have values greater than 90°. Angles were allocated into bin widths of 10°. Note the increased variability in the experimental cords and the increased proportion that are directed caudally (hatched) rather than rostrally.

(2) A larger proportion of the measured axons were directed caudally (Figs 9, 10) in experimental cords (34% or 25/73) compared to normals (13 % or 10/74). This was evident for axons of the most caudal cells, which in normals are always seen to grow rostrally, as well as for axons from more rostral cells.

(3) In some preparations, more often in those analysed at approximately stage 30 than in older embryos, it was clear that some axons (up to 21 in a single preparation) did not turn to grow longitudinally, rather they continued to grow circumferentially, past the other longitudinal axons and up to the dorsal surface of the cord or hindbrain (Fig. 11A,B,C,D). Some of these stained dorsal axons ended abruptly suggesting that they may have left the cord at this point, while others turned to grow ventrally back down on the same side to the majority of stained axons (Fig. 11C). Three axons were found that circumnavigated the cord.

Fig. 11.

Camera-lucida drawings of glycinergic cells and axons. (A,B) Ventral views of whole-mounted spinal cords from u.v.-treated embryos. The positions of all glycinergic cell bodies within the drawn segments are indicated, but only some of the axons are included for clarity. Most of the undrawn axons run longitudinally along the ventral surface of the cord and would thus obscure other details. The arrowed axons have all grown dorsally and probably all arise from contralateral cell bodies. This is clearly seen in three of the filled cells in A and in all four of the filled cells in B. The remaining two filled cells in A have caudally directed axons which remain ipsilateral for unusually long distances before eventually crossing to the other side of the cord. The growth cones of these cells were clear. Dotted lines represent axons growing on dorsal surface of cord. (C,D) Side views of whole-mounted spinal cords from u.v.-treated embryos. Arrows indicate trajectories of dorsally positioned glycinergic axons. In some cases (C) axons turn to grow back down ventrally on the same side of the cord, while in others they continue circumferentially and grow over the dorsal surface and down ventrally on the contralateral side of the cord (dotted in D). Stipple in C indicates position of large numbers of stained longitudinal axons. Scale bar is 50μm.

Fig. 11.

Camera-lucida drawings of glycinergic cells and axons. (A,B) Ventral views of whole-mounted spinal cords from u.v.-treated embryos. The positions of all glycinergic cell bodies within the drawn segments are indicated, but only some of the axons are included for clarity. Most of the undrawn axons run longitudinally along the ventral surface of the cord and would thus obscure other details. The arrowed axons have all grown dorsally and probably all arise from contralateral cell bodies. This is clearly seen in three of the filled cells in A and in all four of the filled cells in B. The remaining two filled cells in A have caudally directed axons which remain ipsilateral for unusually long distances before eventually crossing to the other side of the cord. The growth cones of these cells were clear. Dotted lines represent axons growing on dorsal surface of cord. (C,D) Side views of whole-mounted spinal cords from u.v.-treated embryos. Arrows indicate trajectories of dorsally positioned glycinergic axons. In some cases (C) axons turn to grow back down ventrally on the same side of the cord, while in others they continue circumferentially and grow over the dorsal surface and down ventrally on the contralateral side of the cord (dotted in D). Stipple in C indicates position of large numbers of stained longitudinal axons. Scale bar is 50μm.

Electrophysiology

Normal embryos

Xenopus embryos paralysed by curare will respond to normal sensory stimulation by generating fictive locomotory activity within their spinal cords, which can be monitored by placing extracellular suction electrodes close to the motor axons within the intermyotomal clefts. This activity underlies the alternating myotome contractions that drive normal swimming movements in freely moving embryos (Kahn and Roberts, 1982).

During fictive swimming in curarised normal embryos bursts of motor axon activity alternate on the two sides of the body (Fig. 12A). Within each burst individual motor neurones fire only once (Roberts et al. 1981). During a single episode of fictive swimming, which can last for up to several minutes, the cycle period (time between motor axon bursts) gradually increases from approximately 45 ms to 80ms (Figure). Occasionally during an episode of fictive swimming a few cycles of motor activity occur, in which discharge is synchronous on the two sides (Fig. 12A). The cycle period during this so-called ‘synchrony’ ranges between 15 and 25 ms, and is approximately half that of adjacent swimming cycles.

Fig. 12.

Motor neurone activity in paralysed embryos. (A) Control embryo. Extracellular action potentials recorded from one left (L) and one right (R) intermyotome cleft region. Activity was evoked by a single electrical stimulus to the skin (not shown). Note that the activity in left and right usually alternates. Occasionally short periods of synchronous activity occurs, usually towards the start of an episode. Phase relations are indicated by dotted Unes. (B) U.v.-treated embryo. Extracellular potentials recorded from left and right intermyotome regions now show no periods of alternating motor activity, rather the activity is always synchronous. Scale bar is 50ms. syn, synchrony.

Fig. 12.

Motor neurone activity in paralysed embryos. (A) Control embryo. Extracellular action potentials recorded from one left (L) and one right (R) intermyotome cleft region. Activity was evoked by a single electrical stimulus to the skin (not shown). Note that the activity in left and right usually alternates. Occasionally short periods of synchronous activity occurs, usually towards the start of an episode. Phase relations are indicated by dotted Unes. (B) U.v.-treated embryo. Extracellular potentials recorded from left and right intermyotome regions now show no periods of alternating motor activity, rather the activity is always synchronous. Scale bar is 50ms. syn, synchrony.

Notochordless embryos (n=4)

Notochordless embryos do not swim normally. In response to both gentle and more damaging stimulation of the trunk skin, the embryo is only able to generate bursts of shuddering movements during which the animal contracts along its body axis but does not make any alternating side-to-side movements. No propulsive force is generated.

In curarised preparations, the motor axon activity recorded from intermyotomal clefts is, as in normals, clearly rhythmic. However, in contrast to normals, activity does not alternate on left and right sides but is always synchronous on the two sides (Fig. 12B). As in normals, episodes of fictive motor activity can be evoked either by stroking the skin with a fine gerbil hair or by electrically stimulating the skin with a suction electrode. The cycle period between motor axon bursts ranged between 20 ms at the start of episodes to 45 ms at the end of episodes.

The effects of u. v. irradiation

Ultraviolet irradiation of the fertilized but uncleaved Xenopus egg subsequently produces dose-dependent deficits in the development of the axial mesoderm (Malacinski et al. 1977; Youn and Malacinski, 1981). At low doses of u.v. the deficit can be restricted to an absence of notochord with the remaining mesoderm developing intact (Youn and Malacinski, 1981). Under these conditions neural induction and neuralation proceed, but an abnormal neural tube is produced as described in the present paper. Motor neuron numbers are reduced, extramedullary cell numbers are increased, commissural axon trajectories are altered, the floor plate is apparently missing, and longitudinal axons line the ventral surface of the spinal cord. Are these abnormalities due solely to the disruption of an interaction between notochord and neural tube, or could the u.v. treatment be acting via another mechanism? The dorsal marginal zone (DMZ) of the early gastrula is responsible for organising the axial structures of the later embryo, i.e. somite, notochord and neural tube formation (Spemann, 1938; Smith and Slack, 1983), and if the DMZ is replaced in u.v. embryos, either by direct transplantation from a normal embryo (Chung and Malacinski, 1975) or by induction, following transplantation of the dorsal-most vegetal blastomeres from a normal blastula (Gimlich and Gerhart, 1984), then the development of normal axial structures can be rescued. These experiments demonstrate both that the DMZ is altered by the u.v. treatment and that any perturbations that consequently arise in the neuroectoderm, the large majority of which is not derived from the DMZ (Keller, 1975), do so in response to the DMZ lesion rather than via other routes.

Keller has divided the marginal zone into two regions according to whether or not the cells involute during the movements of gastrulation (Keller et al. 1985). The notochord is derived from deep cells of the dorsal involuting marginal zone (IMZ), while the cells that normally give rise to the floor plate of the neural tube are derived from the dorsal region of the non-involuting marginal zone (NIMZ). How the properties and fates of those two areas of the DMZ are altered by the u.v. treatment is unknown. With the low doses of u.v. used in our study, it is possible that the cell movements are largely normal but that the dorsal IMZ is now fated to become somitic mesoderm, rather than notochord, and the dorsal NIMZ is still fated to become ventral neuroepithelium, but not to develop the specialisation of the floor plate.

Recent evidence suggests that the DMZ may be the source of a signal that spreads through the ectoderm and. contributes to neural induction (Dixon and Kintner, 1989; Ruiz i Altaba, 1990). Since, as discussed above, the DMZ is altered by u.v. treatment, this spreading signal may also be altered. However, with the low doses of u.v. used in our study, the dimensions of the neural tube are normal (Youn and Malacinski, 1981) and, with the exception of small motor neurones, all neuronal cell types are present within the embryonic spinal cord. Thus, it appears that, if the spreading signal is altered, it cannot be involved in these aspects of neural induction. The principal alterations in neural structure in u.v. embryos centre around the organisation of the ventral midline of the neural tube, and recent evidence suggests that, in the chick, interactions between the notochord and the neural tube are central to the patterning of this region (van Straaten et al. 1988, 1989, Wagner et al. 1990; Placzek et al. 1990b, Hirano et al. 1991; Yamada et al. 1991). Our present results are consistent with the view that similar interactions between notochord and neural tube are operating in the Xenopus embryo.

The notochord imposes laterality

In the Xenopus embryo, spinal cords that develop in the absence of a notochord show an impaired ability to maintain a strict left/right laterality relative to the cord’s ventral midline. This loss of laterality is manifest in motor neurones that occupy a midventral position beneath the neurocoel, with many single motor neurones having dendritic processes in the marginal zones on either side of the midline (Fig. 3). Normally motor neurone dendrites are restricted to the ipsilateral marginal zone. Also in normal cords, the left and right marginal zones are clearly separated from one another, but, in notochordless embryos, a continuous marginal zone stretches across the cord’s ventral surface. The loss of laterality is also revealed by the longitudinal axons that run close to the ventral midline in notochordless embryos. Some of these axons can be seen to wander across the midline, far from their cell bodies, in a manner never seen in normal cords. Consistent with these observations, glycinergic axons, which normally cross from side to side via a very direct route, often take less direct routes to the contralateral side, and a few may never reach the contralateral side. A lack of functional laterality is strikingly demonstrated by the inability of the notochordless cord to generate a normally alternating motor output (Fig. 12). Instead, they produce a pattern of motor output that resembles the activity generated by a surgically isolated single side of the cord, where all motor neurones in a single segment fire synchronously (Soffe, -1989).

The neural structure that may normally impose laterality on the organisation of the ventral cord is the floor plate, and it is possible that all the consequences of cord development without a notochord are in fact, more directly, the result of development without a floor plate. The evidence that the floor plate is missing in notochordless embryos is threefold:

(1) No cuboidal cells, characteristic of the floor plate in normal cords, are present ventral to the neurocoel. Instead the neuroepithelial cells present in this location are morphologically similar to the radial glial cells of the normal marginal zones (Fig. 2C,D).

(2) Normal floor plate cells do not express vimentin at 2 to 3 days of development (Fig. 2E), but in notochordless embryos vimentin-positive processes are present throughout the ventral midfine of the cord (Fig. 2F).

(3) The motor neurones, Kolmer-Agduhr cells and ventrolateral axons of the two sides of the cord are normally separated from each other only by the floor plate cells. Both these cell types and their axons lie directly ventral to the neurocoel in notochordless embryos and are not divided into left and right populations by any other cell type.

This evidence from the Xenopus embryo is further supported by recent experiments on chick embryos that provide strong evidence that the floor plate is induced by the notochord. Here surgical removal of the early notochord results in a floorplateless spinal cord, and a notochord grafted to the lateral wall of the neural tube can induce a secondary floor plate (van Straaten et al. 1988; Placzek et al. 1990b, Yamada et al. 1991). At present, there is no biochemical marker that is specific for the Xenopus embryo floor plate; thus, although cells with a normal floor plate morphology are absent from u.v. embryos, we cannot rule out the possibility that cells with other morphologies maintain the molecular characteristics of floor plate cells.

As a mechanism to impose the normal laterality of motor neurone dendrites and ventral, non-commissural axons, we suggest that the floor plate may present a less attractive growth substrate for these processes than the cell surfaces within the marginal zone.

The floor plate is not essential for the development of commissural axons, but probably does influence their growth

The floor plate has been suggested to play one or more important roles in the development of commissural axons. First, in rats it may secrete a diffusible factor that is able to redirect the growing commissural axons towards the floor plate (Tessier-Levigne et al. 1989). Furthermore, having reached the floor plate, rat commissural axons lose the expression of one cellsurface glycoprotein (TAG-1) and begin expression of another surface glycoprotein (LI, Dodd et al. 1988). The signal for this transformation may come from the floor plate, and may be required for the change in behaviour observed in commissural growth cones as they reach its contralateral side of the floor plate. Here they make an abrupt 90 degree turn to grow anteriorly in close association with the lateral wall of the floor plate cells (Bovolenta and Dodd, 1990). This change in direction may be guided by interactions between commissural axons and the cell surface properties of the floor plate, or by as yet unspecified interactions between commissural axons and longitudinal axons within the ventrolateral funiculus (Bovolenta and Dodd, 1990; Yaginuma et al. 1990). In zebrafish embryos, commissural axons bypass ventrolateral axons in the contralateral cord but may fasciculate with other more dorsal axons (Kuwada et al. 1990). In both zebrafish and rats, growth cone morphology is more complex at the floor plate compared to positions along their growth pathway, an observation which in other systems is coincident with the growth cone having reached so-called decision regions in their pathways (Bovolenta and Mason, 1987; Caudy and Bentley, 1968a,b; Tosney and Landmesser, 1985).

The evidence presented in this paper suggests that the floor plate is not essential either for the initial ventrally directed growth of axons or for the glycinergic axons to cross into the contralateral half of the marginal zone. In experimental cords analysed by retrograde HRP staining, the majority of spinal intemeurones have a clear dorsal-to-ventral polarity to their proximal axonal process (Fig. 8), and this is confirmed by the glycinergic staining (Figs 9 and 11) which, in normal cords, is specific for commissural intemeurones (Roberts et al. 1988). The overwhelming majority of glycinergic axons cross the ventral surface of the cord in notochordless embryos. An intrinsic dorsal-to-ventral growth of axons has also been reported in expiants of chick and rat spinal cord devoid of floor plate (Placzek et al. 1990a, Nomes et al. 1990).

The normal ventral structure of the spinal cord does, however, seem to be important for directing the angle of crossing of commissural axons. In the absence of the normal ventral structure, glycinergic axons often take very indirect routes across the ventral midline and one consequence of this is that some of the commissural axons are able to turn caudally into the contralateral marginal zone. In normal cords, the initial contralateral growth is almost always in the rostral direction (Roberts et al. 1988). Whether this means that positional cues that determine rostrally directed growth are lost, or whether the direction of growth is determined by the angle and direction of crossing is not certain. Evidence that some rostrocaudal cues are lost is suggested by the caudal growth of some Kolmer Agduhr axons (Fig. 6D). Normally these axons only grow rostrally (Dale et al. 1987a,b).

In some notochordless embryos, some glycinergic axons failed to turn into the contralateral marginal zone and continued to grow circumferentially up to the dorsal surface of the cord (Fig. 11). This axonal behaviour was most often seen in the hindbrain and rostral cord of the more affected (Grade 2 or 3) embryos. Since the u.v. treatment seems to affect rostral structures more than caudal ones, it is probable that this continued circumferential growth results from a more severe lesion to the cord. The factors that normally direct the glycinergic axons to turn and grow longitudinally may include the expression of cell adhesion molecules like LI (Dodd et al. 1988), which promote fasciculation between commissural axons and other longitudinal axons. We suggest therefore that the more severe lesions to the Xenopus cord may disrupt this fasciculation either by preventing the expression of fasciculation factors on commissural axons or other axons, or that the axons with which commissural axons may fasciculate are absent or reduced in number. In normal cords, the axons of Kolmer-Agduhr cells are in the appropriate position and are present at the right time in development (Dale et al. 1987a,b) to provide a scaffold onto which commissural axons may grow. We are therefore currently examining their distribution and growth in u.v.-treated embryos using GABA immunocytochemistry.

Dorsoventral patterning of the neural tube

The normal balance of dorsal and ventral neurones appears to be disrupted in notochordless embryos (Table 1). As judged from HRP labelling in the myotomes, the number of motor neurones is reduced to about 15 % of normal and the number of extramedullary cells is increase by twentyfold. Extramedullary cells are generated within the neural tube and subsequently migrate out of the dorsal cord (Hughes, 1957); their proliferation is thus likely to be controlled within the cord. It is possible that some or all of the supemumary extramedullary cells are in fact Rohon-Beard cells that have migrated outside the cord. This may be possible due to incomplete neural tube closure (as judged by the occasional apparent lack of a central canal) or the lack of an effective glia limitans, which may normally act to retain cells within the neural tube.

The increase in extramedullary cells and the decrease in motor neurones may at first appear consistent with the proposal that the notochord and/or floor plate can impose positional values on the dorsoventral axis of the neural tube (Jessell et al. 1988; Wagner et al. 1990; Yamada et al. 1991). A diffusible morphogen, perhaps a retinoid (Wagner et al. 1990), from the notochord or floor plate, could produce a gradient such that cells close to the floor plate, where the morphogen is most concentrated, become motor neurones, and those further away take on progressively more dorsal characteristics. In fact, if the u.v. treatment completely removes the notochord and the floor plate, then this hypothesis would predict all cells would acquire the most dorsal characteristics. This is clearly not the case. Perhaps a more likely explanation is that a particular interaction between the notochord and/or floorplate and the neuroepithelium normally induces the stem cells to produce motor neurones (van Straaten et al. 1988, 1989, Hirano et al. 1991; Yamada et al. 1991). In Xenopus, the absence of a notochord does not reduce the motor neurone population to zero and this may reflect the very early birth of some motor neurones during gastulation (Lamborghini, 1980), before any interaction with the notochord is possible. In fact, the only motor neurones to differentiate in notochordless embryos may be the so-called primary motor neurones, which in normal cords are large cells present in relatively small numbers compared to the more numerous, smaller secondary motor neurones (Figs 3, 4, 7 and Hughes, 1959; van Mier et al. 1985; Nordlander, 1986).

The increased number of extramedullary or Rohon-Beard cells may be explained if the proliferation of those cells were normally regulated by a mechanism that measured cell density within the neural tube. Thus, if Rohon-Beard cells were able to migrate out of the neural tube, more would be produced to maintain the normal cell density. In normal embryos, 80% of Rohon-Beard cells are bom before the end of gastrulation; however, some are bom as late as stage 24 (Lamborghini, 1980) at which time neural tube closure (stages 19 to 21, Nieuwkoop and Faber, 1956) is complete. Any increase in proliferation due to cell migration in notochordless embryos should occur around this time, rather than earlier in development.

A further possibility for the increased number of HRP-labelled extramedullary cells is that the total number of extramedullary and Rohon-Beard cells is unchanged, but their density has increased due to a shortening of the length of neural tube. While the notochordless embryos are shorter than normals by about 25 %, we feel this is unlikely to account for the observed threefold increase in combined Rohon-Beard and extramedullary cell numbers. Experiments to determine the total populations of these cells and motor neurones are however, in progress.

Functional consequences

The electrophysiological data show that the spinal cords of notochordless embryos are able to generate a rhythmic motor output but are unable to coordinate this rhythmicity in such a way as to give the alternating motor output on the two sides of the animal that characterise swimming (Fig. 12). Alternation is normally achieved by the reciprocal inhibitory connections made by the glycinergic commissural axons (Dale, 1985; Dale et al. 1986). The synchronous firing pattern recorded from left and right motor axons could result from a normally connected central arrangement of premotor and motor neurones if the motor axons from one side of the cord were able to innervate the contralateral as well as the ipsilateral myotomes. Thus, although the activity of left and right motor neurones would be alternating normally, this would be masked at the level of motor axons within the bilaterally innervated myotomes. This is unlikely to be the case here, as any underlying alternating motor axon activity would be expected to be revealed as a pattern of motor discharge whose characteristics (shape, amplitude, width) differed on alternate cycles. This was never seen. Also, intracellular recordings from ventral spinal neurones have revealed that each neurone fires on each cycle at the accelerated synchronous rate rather than the slower rate observed in normal cords (Soffe and Clarke unpublished observations).

More likely explanations for the synchronous left and right motor activity are either that the bilateral dendrites of motor neurones allow these cells to be rhythmically excited by two groups of premotor intemeurones, active in alternation on either side of the cord (thus motor neurones would fire at twice the frequency of premotor intemeurones) or that the whole spinal cord acts as a single-phase oscillator, like a single surgically isolated side of the cord, with all rhythmically active cells firing synchronously and at the same rate. The former proposal would require that the normal laterality of connections between premotor interneurones is maintained in the experimental cords. This seems unlikely in view of the tendency of HRP-labelled axons not to respect the ventral midline as a boundary; however, these alternatives need to be tested by a more detailed study of the dendritic and axonal projections of premotor intemeurones and the use of intracellular electrodes to determine synaptic connectivity and the firing patterns of premotor intemeurones.

Action Research and the SERC provided financial assistance. Thanks to Steve Franey for histology, Cynthia Civil and Sheila Collins for word processing and Leon Kelberman for photography. Tom Jessell, Malcolm Maden, Tahereh Kama-lati, Jim Hill and Emily Gale have all provided helpful discussion. Chris Wylie and Janet Heasman kindly provided the anti-vimentin antibody.

Blight
,
A.
(
1978
).
Golgi-staining of ‘primary’ and ‘secondary’ motoneurons in the developing spinal cord of an amphibian
.
J. comp. Neurol.
180
,
679
689
.
Bovolenta
,
P.
and
Dodd
,
J.
(
1990
).
Guidance of commissural growth cones at the floor plate in embryonic rat spinal cord
.
Development
109
,
435
447
.
Bovolenta
,
P.
and
Mason
,
C. A.
(
1987
).
Growth cone morphology varies with position in the developing mouse visual pathway from retina to first targets
.
J. Neurosci.
7
,
1447
1460
.
Caudy
,
M.
and
Bentley
,
D.
(
1986a
).
Pioneer growth cone morphologies reveal proximal increase in substrate affinity within leg segments of grasshopper embryos
.
J. Neurosci.
6
,
364
379
.
Caudy
,
M.
and
Bentley
,
D.
(
1986b
).
Pioneer growth cone steering along a series of neuronal and nonneuronal cues of different affinities
.
J. Neurosci.
6
,
1781
1795
.
Chung
,
H.
and
Malacinski
,
G. M.
(
1975
).
Repair of ultraviolet irradiation damage to a cytoplasmic component required for neural induction in the amphibian egg
.
Proc. natn. Acad. Sci. U.S.A.
72
,
1235
1239
.
Clarke
,
J. D. W.
,
Hayes
,
B. P.
,
Hunt
,
S. P.
and
Roberts
,
A.
(
1984
).
Sensory physiology, anatomy and immunohistochemistry of Rohon-Beard neurones in embryos of Xenopus laevis.
J. Physiol.
348
,
511
525
.
Dale
,
N.
(
1985
).
Reciprocal inhibitory interneurones in the Xenopus embryo spinal cord
.
J. Physiol. London
363
,
61
70
.
Dale
,
N.
,
Otterson
,
O. P.
,
Roberts
,
A.
and
Storm-Mathisen
,
J.
(
1986
).
Inhibitory neurones of a motor pattern generator in Xenopus revealed by antibodies to glycine
.
Nature
324
,
255
257
.
Dale
,
N.
,
Roberts
,
A.
,
Ottersen
,
O. P.
and
Storm-Mathisen
,
J.
(
1987a
).
The morphology and distribution of ‘Kolmer-Agduhr cells’, a class of cerebrospinal-fluid-contacting neurons revealed in the frog embryo spinal cord by GABA immunocytochemistry
.
Proc. R. Soc. London B
232
,
193
203
.
Dale
,
N.
,
Roberts
,
A.
,
Ottersen
,
O. P.
and
Storm-Mathisen
,
J.
(
1987b
).
The development of a population of spinal cord neurons and their axonal projections revealed by GABA immunocytochemistry in frog embryos
.
Proc. R. Soc. London B
232
,
205
215
.
Dixon
,
J. E.
and
Kintner
,
C. R.
(
1989
).
Cellular contact required for neural induction in Xenopus embryos: evidence for two signals
.
Development
106
,
749
757
.
Dodd
,
J.
,
Morton
,
S. B.
,
Karagogeos
,
D.
,
Yamamoto
,
M.
and
Jessell
,
T. M.
(
1988
).
Spatial regulation of axonal glycoprotein expression on subsets of embryonic spinal neurons
.
Neuron
1
,
105
116
.
Forehand
,
C. J.
and
Farel
,
P. B.
(
1982
).
Spinal cord development in anuran larvae. I. Primary and secondary neurons
.
J. comp. Neurol.
209
,
386
394
.
Glmlich
,
R. L.
and
Gerhart
,
J. C.
(
1984
).
Early cellular interactions promote embryonic axis formation in Xenopus laevis.
Devi Biol.
104
,
117
130
.
Gurdon
,
J. B.
(
1977
).
Methods for nuclear transplantation in Amphibia
.
Methods Cell Biol.
16
,
125
139
.
Hirano
,
S.
,
Fuse
,
S.
and
Sohal
,
G. S.
(
1991
).
The effect of the floor plate on pattern and polarity in the developing central nervous system
.
Science
251
,
310
313
.
Holtfreter
,
J.
and
Hamburger
,
V.
(
1955
).
Embryogenesis: Progressive differentiation
. In
Foundations of Experimental Embryology
(ed.
Willier
,
B. H.
et al.) pp.
230
296
.
Philadelphia
:
Saunders Press
.
Hughes
,
A.
(
1957
).
The development of the primary sensory system in Xenopus laevis (Daudin).
J. Anat.
91
,
323
338
.
Hughes
,
A.
(
1959
).
Studies in embryonic and larval development in amphibia. II. The spinal motor root
.
J. Embryol. exp. Morph.
1
,
128
145
.
Jessell
,
T. M.
,
Bovolenta
,
P.
,
Placzek
,
M.
,
Tessier-Lavigne
,
M.
and
Dodd
,
J.
(
1988
).
Polarity and patterning in the neural tube: the origin and function of the floor plate
. In
The Cellular Basis of Morphogenesis. CIBA Foundation Symposium
.
Kahn
,
J. A.
and
Roberts
,
A.
(
1982
).
The central nervous origin of the swimming motor pattern in embryos of Xenopus laevis.
J. exp. Biol.
99
,
185
196
.
Keller
,
R. E.
(
1975
).
Vital dye mapping of the gastrula and neurala of Xenopus laevis. I. Prospective areas and morphogenetic movements of the superficial layer
.
Devi Biol.
51
,
222
241
.
Keller
,
R. E.
,
Danilchik
,
M.
,
Gimlich
,
R.
and
Shih
,
J.
(
1985
).
The function and mechanism of convergent extension during gastrulation of Xenopus laevis.
J. Embryol. exp. Morph.
89
Suppl.
,
185
209
.
Kuwada
,
J. Y.
,
Bernhardt
,
R. R.
and
Chitnis
,
A. B.
(
1990
).
Pathfinding by identified growth cones in the spinal cord of zebrafish embryos
.
J. Neurosci.
10
,
1299
1308
.
Lamborghini
,
J. E.
(
1980
).
Rohon-Beard cells and other large neurons in Xenopus embryos originate during gastrulation
.
J. comp. Neurol.
189
,
323
333
.
Lang
,
A. G.
(
1937
).
The use of n-butyl alcohol in the paraffin method
.
Stain Technol.
12
,
113
117
.
Malacinski
,
G. M.
,
Alus
,
C. D.
and
Chung
,
H-M.
(
1974
).
Correction of developmental abnormalities resulting from localized ultra-violet irratiation of an amphibian egg (1)
.
J. exp. Zool.
189
,
249
254
.
Malacinski
,
G. M.
,
Brothers
,
A. J.
and
Chung
,
H-M.
(
1977
).
Destruction of components of the neural induction system of the amphibian egg with ultraviolet irradiation
.
Devi Biol.
56
,
24
39
.
Nieuwkoop
,
P. D.
and
Faber
,
J.
(
1956
).
Normal Tables of Xenopus laevis (Daudin). Amsterdam: North-Holland
.
Nordlander
,
R. H.
(
1986
).
Motoneurons of the tail of young Xenopus tadpoles
.
J. comp. Neurol.
253
,
403
413
.
Nornes
,
H. O.
,
Knapik
,
E.
and
Kroger
,
S.
(
1990
).
Polarised outgrowth of circumferential fibres from expiants of avian embryonic spinal cord
.
Soc. Neurosci. Abstr.
16
,
316
.
Placzek
,
M.
,
Tessier-Lavigne
,
M.
,
Jessell
,
T.
and
Dodd
,
J.
(
1990a
).
Orientation of commissural axons m vitro in response to a floor plate-derived chemoattractant
.
Development
110
,
19
30
.
Placzek
,
M.
,
Tessier-Lavigne
,
M.
,
Yamada
,
T.
,
Jessell
,
T.
and
Dodd
,
J.
(
1990ft
).
Mesodermal control of neural cell identity: floor plate induction by the notochord
.
Science
250
,
985
988
.
Roberts
,
A.
and
Clarke
,
J. D. W.
(
1982
).
The neuroanatomy of an amphibian embryo spinal cord
.
Phil. Trans. R. Soc. B
296
,
195
212
.
Roberts
,
A.
,
Dale
,
N.
,
Ottersen
,
O. P.
and
Storm-Mathisen
,
J.
(
1987
).
The early development of neurons with GABA immunoreactivity in the CNS of Xenopus laevis embryos
.
J. comp. Neurol.
261
,
435
449
.
Roberts
,
A.
,
Dale
,
N.
,
Ottersen
,
O. P.
and
Storm-Mathisen
,
J.
(
1988
).
Development and characterization of commissural intemeurones in the spinal cord of Xenopus laevis embryos revealed by antibodies to glycine
.
Development
103
,
447
461
.
Roberts
,
A.
,
Kahn
,
J. A.
,
Soffe
,
S. R.
and
Clarke
,
J. D. W.
(
1981
).
Neural control of swimming in a vertebrate
.
Science
213
,
1032
1034
.
Ruiz i
Altaba
, A
. (
1990
).
Neural expression of the Xenopus homeobox gene Xhox3: evidence for a patterning neural signal that spreads through trhe ectoderm
.
Development
108
,
595
604
.
Saxen
,
L.
(
1989
).
Neural induction
.
Int. J. Devi Biol.
33
,
21
48
.
Scharf
,
S. R.
and
Gerhart
,
J. C.
(
1983
).
Axis determination in eggs of Xenopus laevis: A critical period before first cleavage, identified by the common effects of cold, pressure and ultraviolet irradiation
.
Devi Biol.
99
,
75
87
.
Smith
,
J. C.
and
Slack
,
J. M. W.
(
1983
).
Dorsalisation and neural induction: properties of the organiser in Xenopus laevis.
J. Embryol. exp. Morph.
78
,
299
317
.
Soffe
,
S. R.
(
1989
).
Roles of glycinergic inhibition and N-methyl-D-aspartate receptor mediated excitation in the locomotor rhythmicity of one half of the Xenopus embryo central nervous system
.
Eur. J. Neurosci.
1
,
561
571
.
Spemann
,
H.
(
1938
).
Embryonic Development and Induction. New Haven: Yale University Press
.
Taylor
,
J. S. H.
and
Roberts
,
A.
(
1983
).
The early development of primary sensory neurites in an amphibian embryo: a scanning electron microscope study
.
J. Embryol. exp. Morph.
75
,
49
66
.
Tessier-Lavigne
,
M.
,
Placzek
,
M.
,
Lumsden
,
A. G. S.
,
Dodd
,
J.
and
Jessell
,
T. M.
(
1988
).
Chemotropic guidance of developing axons in the mammalian central nervous system
.
Nature
336
,
775
778
.
Tosney
,
K. W.
and
Landmesser
,
L.
(
1985
).
Growth cone morphology and trajectory in the lumbosacral region of the chick embryo
.
J. Neurosci.
5
,
2345
2358
.
Van Mier
,
R.
,
Van Rheden
,
R.
and
Ten Donkelaar
,
H. J.
(
1985
).
The development of the dendritic organisation of primary and secondary motoneurons in the spinal cord of Xenopus.
Anat. Embryol.
172
,
311
324
.
Van Straaten
,
H. W. M.
,
Hekking
,
J. W. M.
,
Beursgens
,
J. P. W. M.
,
Terwindt-Rouwenhorst
,
E.
and
Drukker
,
J.
(
1989
).
Effect of the notochord on proliferation and differentiation in the neural tube of the chick embryo
.
Development
107
,
793
804
.
Van Straaten
,
H. W. M.
,
Hekking
,
J. W. M.
,
Wiertz-Hoessels
,
E. J. L. M.
,
Thors
,
F.
and
Drukker
,
J.
(
1988
).
Effect of notochord on the differentiation of a floor plate area in the neural tube of the chick embryo
.
Anat. Embryol.
177
,
317
324
.
Wagner
,
M.
,
Thaller
,
C.
,
Jessell
,
T.
and
Eichele
,
G.
(
1990
).
Polarizing activity and retinoid synthesis in the floor plate of the neural tube
.
Nature
345
,
819
822
.
Yaginuma
,
H.
,
Shiga
,
T.
,
Homma
,
S.
,
Ishihara
,
R.
and
Oppenheim
,
R. W.
(
1990
).
Identification of early developing axon projections from spinal interneurons in the chick embryo with a neuron specific b-tubulin antibody: evidence for a new ‘pioneer’ pathway in the spinal cord
.
Development
108
,
705
716
.
Yamada
,
T.
,
Placzek
,
M.
,
Tanaka
,
H.
,
Dodd
,
J.
and
Jessell
,
T. M.
(
1991
).
Control of cell pattern in the developing nervous system: polarizing activity of the floor plate and notochord
.
Cell
64
,
635
647
.
Youn
,
B. W.
and
Malacinski
,
G. M.
(
1981
).
Axial structure development in ultraviolet-irradiated (notochord-defective) amphibian embryos
.
Devi Biol.
83
,
339
352
.