The results of the various observations and experiments described in this
dissertation have been discussed already in their own sections. Therefore
instead of rehashing the same arguments I propose here to consider these
results in the light of previous experience with using C. elegans as
a model developmental animal, and to speculate in which type of direction
future work, particularly on the genetics of neural specification, might
The studies described in both parts of this dissertation have relied on the fact that the C. elegans nervous system is both extremely simple and highly reproducible, so that information can be gained from a comparatively small amount of data. However there is also a possible penalty to be paid in studying an organism with a very small number of cells, all of which are reproducible from individual to individual. These properties potentially permit structures to be put together piecemeal by some form of internal program specific to each part, rather than by general mechanisms.
The initial reason for attempting a computer database analysis of the synapse and connectivity data was to attempt to find internal logical patterns in the connectivity data which might allow rules to be proposed for specifying which cells connected to which, for instance by placing the neurons in possibly overlapping "super-classes" that might have common recognition properties, so that if two cells were in compatible classes and also in contact then they would form a connection. There are examples of pairs or groups of cells that are in different places and make mostly different connections, but which make similar connections to cells that they both contact, and which share other properties in common (White et al., 1983). However an overall search for such grouping reveals nothing that is statistically significant. One possible problem that may be important is that regional specialisation of neurons, as discussed in Chapter 7, would create complications in any search for classes of neurons with equivalent synaptic potential. This does not mean that label receptor matching systems for determining synaptic connectivity do not exist, but merely that there are too few cells and there is too much variation to deduce them from the final connectivity data.
A similar observation was made when the complete cell lineage was determined, which is more reproducible than the nervous system. Although there are a few suggestive repeated motifs, the overall arrangement of which precursors produce which cells is essentially haphazard and mosaic, correlating as much with position as with pattern in the lineage (Sulston, 1983). This could be taken to indicate that external interactions with extracellular environment were important in determining cell fate, but abalation experiments largely revealed no effect on adjacent cells (Sulston and White, 1980, Sulston et al., 1983). Overall this suggests intrinsic programming, but it has an advantage for the study of intercellular determination, which is that those instances where specific cell interaction is important, of which there are a number of clear examples (Sulston and White, 1980), may be comparatively isolated. A number of the cell lineage mutants that have been obtained affect situations where induction or regulation takes place (Sternberg and Horvitz, 1984), and these may provide an excellent tool to study specific determinitive cell interactions during development in vivo. One particular gene of this type has recently been cloned and sequenced, and its protein sequence has homology to a family of extacellular proteins including growth factors and their receptors (Greenwald, 1985). Indeed there is an argument that clean developmental switch genes, which cause the change of cell fate from one type to another, will often be associated with inductive or regulative situations: a defect in a single component of an extracellular signalling pathway, such as the signal or the receptor, would cause an effective loss of signal, while internal choice determination may be a complex activity requiring many components simultaneously at each stage, and with no clear default behaviour. Having obtained one of the components for an interactive mechanism via a mutant, one then has a genetic handle on the subsequent parts of the mechanism.
The relative positioning of neuronal processes is much more complex than that of most other types of cells, and it must be expected that a large amount of intracellular interaction is required for process positioning and synapse formation. However much of this may be non-specific. As with the lineage ablation studies, the ablation experiments described in Chapter 4 in general had remarkably little effect on other cells. The DD3/DD5, DVC and PVPL removal experiments showed no immediate effect on guidance of other neurons at all. As discussed in Chapter 5 there are already mutants affecting process guidance in various ways. There are also mutants known that affect synaptic connectivity in the ventral and dorsal nerve cords in a way that can be interpreted as switching the specificity of certain cells from one type to another (J. White, L. Nawrocki, personal communication). It is possible that some of these mutants may also affect comparatively isolated determinative intercellular interactions, which may provide models for similar interactions in more complex animals. Even if not they may still reveal interesting mechanisms involved in specific guidance and synaptic connectivity. However, by itself, genetics can be problematical because it may be hard to determine what one is studying. It is ultimately in combining genetics with the detailed and specific anatomical observations and experiments that are possible in such a simple organism that I believe
C. elegans has most to offer development neuroscience. If I were to continue working with C. elegans I would investigate the early anatomical development of some of the guidance mutants and follow up the molecular and genetic opportunities they generate.