I have studied superior cervical ganglion cells in mouse, hamster, rat, guinea pig, and rabbit by electron microscopy to determine how the distribution of synapses on these neurons is affected by the systematic differences in dendritic morphology and preganglionic convergence that are evident in the superior cervical ganglia of these species (Purves, D., and J.

This study of spinal cord injury in bullfrog (Rana catesbeiana) tadpoles using the neuroanatomical tracer horseradish peroxidase (HRP) was undertaken to determine (1) whether the same anatomical regions that normally give rise to ascending or descending spinal tracts do so following complete spinal cord transection and (2) whether the course of behavioral recovery could be related to the anatomical results. The results of this study show that (1) spinal cord continuity is readily restored in tadpoles subjected to spinal cord transection, but nerve fibers crossing the site of injury end within 1 to 2 mm of the lesion site; (2) tadpoles with spinal cord transections held through metamorphosis show, as juvenile frogs, restoration of lumbar projections from all brainstem regions that normally project to the lumbar spinal cord; (3) neither long ascending projections from dorsal root ganglion cells nor those from spinal neurons caudal to the transection traverse the transection site, even after metamorphosis; and (4) consistent with the anatomical results, tadpoles show only minimal behavioral recovery, but these same animals as juvenile frogs show recovery of behaviors that are dependent upon connections to supraspinal regions. In other experiments, [3H]thymidine or [3H]apo-HRP was combined with HRP histochemistry to determine if new brainstem neurons projecting to the spinal cord are born in the metamorphic period and if, in normal animals, brainstem projections to the lumbar spinal cord persist through metamorphosis. We found no evidence that neurons with lumbar spinal cord projections are born during metamorphosis; however, evidence was found that most brainstem neurons that project to the lumbar spinal cord before metamorphosis retain this projection in the juvenile frog.

In the rabbit, ciliary ganglion neurons with dendrites maintain inputs from several different axons during the period of synaptic rearrangement that occurs in early postnatal life. Neurons without dendrites, on the other hand, lose the majority of their initial inputs and are innervated in maturity by the terminals of only one or two axons (Purves, D., and R.I. Hume (1981) J. Neurosci. 1: 441-452; Hume, R.I., and D. Purves (1981) Nature 293: 469-471). We have explored the basis of this phenomenon by individually marking preganglionic axons and the neurons they innervate with horseradish peroxidase. In general, the innervation of geometrically complex (multiply innervated) neurons by individual preganglionic axons is regional. That is, the synaptic contacts made by an axon on these neurons are limited to a portion of the postsynaptic surface that includes some, but not all, of the dendrites. This regional innervation of target neurons is consistent with the view that dendrites allow multiple innervation to persist by providing relatively separate postsynaptic domains for individual preganglionic axons. Such regional innervation may mitigate competitive interactions between the several axons which initially innervate the same neuron.

The number of preganglionic inputs that innervate rabbit ciliary ganglion cells is directly correlated with the number of dendrites arising from each ganglion cell (Purves and Hume, 1981). In general, the innervation of multiply innervated ciliary neurons by individual preganglionic axons is regionally restricted to a portion of the postsynaptic surface that usually includes the cell body and some, but not all, of the dendrites (Forehand and Purves, 1984). These observations suggest that dendrites modulate convergence to each cell by providing relatively separate postsynaptic domains for individual inputs. To examine this possibility further, I have assessed the distribution of synaptic boutons from individually labeled preganglionic axons on ciliary ganglion cells at the ultrastructural level. The results show that at least a third of the dendrites of these neurons are contacted exclusively by synaptic boutons from a single preganglionic axon. However, at least half of the dendrites (and nearly all of the cell bodies) of multiply innervated ganglion cells are innervated by at least 2 different preganglionic axons. Moreover, synapses from 2 different inputs often coexist in close proximity on the postsynaptic surface. Thus, individual preganglionic axons do not require exclusive dominion over a particular part of a postsynaptic cell in order to maintain their connection with the cell. These results suggest that competitive interactions between the inputs to these cells occur between the sets of boutons arising from different inputs, rather than at the level of individual boutons.(ABSTRACT TRUNCATED AT 250 WORDS)