The data reviewed above have implicated the floor plate in directing axonal growth towards the midline, in directing the behavior of axons at the midline, and finally, in directing the longitudinal growth of axons alongside the midline. In the case of growth to the midline, there is clear evidence for the existence of two types of cues that collaborate to direct growth - short-range cues that can direct axons along the edge of the spinal cord and a long-range chemoattractant secreted by the floor plate cells whose main function may be to direct later-extending commissural axons that must migrate through the complex environment of the developing motor column. Determining the precise contribution of these cues will require identifying them and perturbing them in vivo. The cues that direct growth along the edge are unknown; their identification in the spinal cord would be of quite general significance, since the growth of axons parallel to (but not in contact with) the pial surface is a widespread feature of early axon growth at all axial levels of the neural tube. A strong candidate for the long-range chemoattractat is netrin-1, a homologue of the UNC-6 protein of C. elegans and a distant relative of laminin, which is expressed by floor plate cells and which can both promote and orient commissural axon outgrowth. Netrin-1 may also influence growth of other populations of neurons that exhibit stereotyped behaviors near the ventral midline. Much less is known about the exact role of the floor plate in directing axon growth at the midline, though it is clearly required for accurate guidance. In the absence of the floor plate, a range of errors has been found, the most prominent of which are aberrant midline crossings and errors in longitudinal growth near the ventral midline. The severity of these errors varies with species, which could result from either the variable importance of the floor plate in the different species or the fact that, so far, quite different manipulations of the ventral midline region have been performed in different species. The most specific perturbation of the ventral midline occurs in the zebrafish cyc-1 mutant, where the selective loss of the floor plate leads to stereotyped misrouting events. Perhaps surprisingly, virtually all axons that grow to the midline turn longitudinally (although sometimes in the wrong direction). The observed defects in the hindbrain can be interpreted to result primarily from the loss of a barrier to recrossing provided by the floor plate so that axons do not become confined to the appropriate side of the embryo after turning and wander across the midline at a certain frequency. We have suggested that perhaps this can account for most of the errors observed in the spinal cord too. However, other cues from the floor plate directing navigation at the midline may certainly be present, but the exact form of these cues (e.g. crossing, turning, or overriding signals, as we have discussed) remains to be resolved. If these other types of signals do exist, it seems unlikely that the floor plate is their sole source (at least in zebrafish), and there is in fact evidence that the motochord contributes at least some of these signals. There is at present no information on the molecular nature of any of the cues that direct guidance events at the midline. Studies of the role of the floor plate in axon guidance have highlighted the difficulty of analyzing such intricate guidance events and elucidating their molecular basis. Sharp turn made by axons at particular landmarks like the floor plate are made in many regions of the nervous system that are less amenable to experimental analysis. Lessons that are learned from studying axons guidance at the floor plate are therefore likely to provide insights into similar types of decisons made in other regions of the nervous system.
