Synaptic connectivity of local circuit neurons in laminae III and IV of hamster spinal cord.

The present study was undertaken to examine the morphological bases of local synaptic interactions between dorsal horn interneurons. Seven interneurons responding to innocuous mechanical stimuli were intracellularly recorded in lamina III/IV of an isolated preparation of hamster spinal cord with partially intact innervation from an excised patch of hairy skin. Axonal arborizations were stained with horseradish peroxidase (HRP) and examined with an electron microscope. Five cells had extensive synaptic terminations (375-1,785 boutons/axon) with localized distributions (rostrocaudal distance, 425-1,251 microns) overlapping the dendritic trees. Two cells gave rise to deep stem axons that bifurcated into rostrocaudal daughter branches with collaterals ventral to the parent cell bodies (79-661 boutons/axon). Axons of local interneurons were thinly myelinated and formed terminal and en passant enlargements (mean [+/- S.D.] diameter = 0.88 +/- 0.24 microns, n = 157) containing clear, round vesicles 20-60 nm in diameter. Collateral branches of deep axon cells produced round, vesicle-containing boutons comparable in diameter (0.93 +/- 0.22 microns, n = 31) to local axon cells. Both types of interneurons formed asymmetric synaptic contacts with dendritic profiles, but not with cell bodies or axon terminals. Postsynaptic profiles contained sparse ribosomes and had a mean diameter of 1.0 +/- 0.5 microns (n = 49), significantly smaller than a population of identified proximal dendrites (2.3 +/- 0.9 microns, n = 47). HRP-labeled boutons were rarely (5/45 or 11%) in synaptic contact with more than one profile. We conclude that lamina III/IV interneurons make axodendritic synapses predominantly with distal dendrites. Thus, terminations of deep dorsal horn interneurons appear to have a postsynaptic distribution overlapping with axodendritic contacts formed by several functional classes of cutaneous sensory fibers signaling innocuous mechanical stimuli. Such overlap suggests that local spinal networks selectively and strongly influence afferent signals at initial stages of somatosensory integration.