Using axonal tracers, we characterized the neurons projecting from the brain to the spinal cord as well as the terminal fields of ascending spinal projections in the brain of adult zebrafish with unlesioned or transected spinal cords. Twenty distinct brain nuclei were found to project to the spinal cord. These nuclei were similar to those found in the closely related goldfish, except that additionally the parvocellular preoptic nucleus, the medial octavolateralis nucleus, and the nucleus tangentialis, but not the facial lobe, projected to the spinal cord in zebrafish. Terminal fields of axons, visualized by anterograde tracing, were seen in the telencephalon, the diencephalon, the torus semicircularis, the optic tectum, the eminentia granularis, and throughout the ventral brainstem in unlesioned animals. Following spinal cord transection at a level approximately 3.5 mm caudal to the brainstem/spinal cord transition zone, neurons in most brain nuclei grew axons beyond the transection site into the distal spinal cord to the level of retrograde tracer application within 6 weeks. However, the individually identifiable Mauthner cells were never seen to do so up to 15 weeks after spinal cord transection. Nearly all neurons survived axotomy, and the vast majority of axons that had grown beyond the transection site belonged to previously axotomized neurons as shown by double tracing. Terminal fields were not re-established in the torus semicircularis and the eminentia granularis following spinal cord transection.

Previous studies indicated that the developing fish spinal cord was a simple system containing a small number of distinguishable neuronal cell types (Eisen et al., Nature 320:269-271, '86; Kuwada, Science, 233:740-746, '86). To verify this we have characterized the cellular anatomy of the spinal cord of developing zebrafish in order to determine the number, identities, and organization of the spinal neurons. Spinal neurons were labeled by intracellular dye injections, application of an axonal tracer dye to all or subsets of the axonal tracts, and application of antibodies which recognize embryonic neurons. We found that nine classes of neurons could be identified based on soma size and position, pattern of dendrites, axonal trajectory, and time of axonogenesis. These are two classes of axial motor neurons, which have been previously characterized (Myers, J. Comp. Neurol. 236:555-561, '85), one class of sensory neurons, and six classes of interneurons. One of the interneuron classes could be subclassified as primary and secondary based on criteria similar to those used to classify the axial motor neurons into primary and secondary classes. The early cord (18-20 hours) is an extremely simple system and contains approximately 18 lateral cell bodies per hemisegment, which presumably are post-mitotic cells. By this stage, five of the neuronal classes have begun axonogenesis including the primary motor neurons, sensory neurons, and three classes of interneurons. By concentrating on these early stages when the cord is at its simplest, pathfinding by growth cones of known identities can be described in detail. Then it should be possible to test many different mechanisms which may guide growth cones in the vertebrate central nervous system (CNS).

A specific antiserum was used to ascertain the distribution of microtubule-associated protein 2 (MAP2) in the rat brain at the light and electron microscope levels. Light microscopy showed MAP2 to be present only in neurons, and only in the dendrites and the perikaryon of each cell. This same polarized distribution pattern was found in the Purkinje, Golgi, basket, stellate, and granule cells of the cerebellum, and also in neurons of the hippocampus, the olfactory bulb, and the midbrain. While labelling of the dendritic arborization was extensive and intense, MAP2 density tended to decrease in the proximal dendritic trunk. Particularly in large neurons (e.g., Purkinje, Golgi, and pyramidal cells), staining was reproducibly weaker in the cell body than in the main dendrites. Dendritic contours generally appeared smooth, without any evidence of staining of dendritic spines. An electron microscope examination of the cerebellum confirmed the presence of MAP2 reactivity in neurons and its absence from axons and non-neuronal cells. MAP2 in dendrites was associated with microtubules, while MAP2 in neuronal perikarya was associated with polyribosomes. There was no evidence of specific staining in dendritic spines and in postsynaptic densities. MAP2 is a novel dendritic marker and labels part of a specific dendritic cytoskeleton, different from that in axons and non-neuronal cells.

The distributions of tubulin and high molecular weight microtubule-associated proteins (HMWPs) in brain were determined by immunoperoxidase histochemistry with specific antisera. Tubulin was found in microtubules of both neurons and glia and both axons and dendrites. HMWPs were found only in neurons where, in all cases examined, they were associated with dendritic microtubules but not those in axons. Both tubulin and * H HMWPs were also found in postsynaptic densities. These results indicate that brain microtubules vary in chemical composition. The preferential association of HMWPs with dendritic microtubules suggests that they may play a role in distinguishing between dendritic and axonal export routes from the cell body.

We analyzed changes in the expression of mRNAs for the axonal growth-promoting cell recognition molecules L1.1, L1.2, and neural cell adhesion molecule (NCAM) after a rostral (proximal) or caudal (distal) spinal cord transection in adult zebrafish. One class of cerebrospinal projection nuclei (represented by the nucleus of the medial longitudinal fascicle, the intermediate reticular formation, and the magnocellular octaval nucleus) showed a robust regenerative response after both types of lesions as determined by retrograde tracing and/or in situ hybridization for GAP-43. A second class (represented by the nucleus ruber, the nucleus of the lateral lemniscus, and the tangential nucleus) showed a regenerative response only after proximal lesion. After distal lesion, upregulation of L1.1 and L1.2 mRNAs, but not NCAM mRNA expression, was observed in the first class of nuclei. The second class of nuclei did not show any changes in their mRNA expression after distal lesion. After proximal lesion, both classes of brain nuclei upregulated L1.1 mRNA expression (L1.2 and NCAM were not tested after proximal lesion). In the glial environment distal to the spinal lesion, labeling for L1.2 mRNA but not L1.1 or NCAM mRNAs was increased. These results, combined with findings in the lesioned retinotectal system of zebrafish (Bernharnhardt et al., 1996), indicate that the neuron-intrinsic regulation of cell recognition molecules after axotomy depends on the cell type as well as on the proximity of the lesion to the neuronal soma. Glial reactions differ for different regions of the CNS.