Transneuronal changes of the inhibitory circuitry in the macaque somatosensory thalamus following lesions of the dorsal column nuclei.

The inhibitory circuitry of the ventroposterolateral nucleus (VPL) of the macaque somatosensory thalamus was analyzed in normal animals and in those surviving for a few days or several weeks following a unilateral lesion of the cuneate nucleus, the source of medial lemniscal (ML) axons carrying information from the contralateral upper extremity. Inhibitory synaptic terminals in the VPL were defined as those that contain flattened or pleomorphic synaptic vesicles and that can be shown to be immunoreactive for gamma-aminobutyric acid (GABA). There are two types of these profiles: F axon terminals that arise from neurons of the thalamic reticular nucleus, and perhaps from VPL local circuit neurons (LCNs); and the dendritic appendages of LCNs that form presynaptic dendrites (PSDs). ML terminals normally have extensive synaptic interactions with PSDs but not with F axon terminals. Electron microscopic analyses revealed that cuneatus lesions resulted in a rapid loss of ML terminals and a statistically significant reduction in both F and PSD synaptic profiles. Confocal scanning microscopy also demonstrated a profound loss of GABA immunoreactivity in the deafferented VPL. These changes persisted for more than 20 weeks, without any evidence of reactive synaptogenesis of surviving sensory afferents or of inhibitory synapses. The changes in GABA circuitry are transneuronal, and the possible mechanisms that may underlie them are discussed. It is suggested that the altered GABAergic circuitry of the VPL in the monkey may serve as a model for understanding changes in somatic sensation in the human following peripheral or central deafferentation.

Electron microscopic imaging of multiple markers in glutaraldehyde fixed CNS tissue of Macaca fascicularis: maximizing information from a single experimental animal.

Sensory and motor pathways in the central nervous system (CNS) of macaque monkeys were visualized by anterograde or retrograde axonal transport of wheatgerm agglutinin-horseradish peroxidase (WGA-HRP) reacted with the chromagen tetramethylbenzidine (TMB), or by the use of anterograde degeneration after specific ablation lesions. To maximize information from each animal we combined the results of the anterograde and retrograde axonal transport with several pre- and post-embedding markers at both the light and electron microscopic levels while maintaining good preservation of tissue. Pre-embedding techniques included those for cytochrome oxidase activity and the calcium-binding proteins calbindin D-28k and parvalbumin. Post-embedding techniques included immunocytochemistry for gamma-aminobutyric acid (GABA) or other amino acid neurotransmitters. We believe that the methods described here provide superior tissue preservation, thus permitting a more detailed analysis of tissue prepared after experiments concerned with neural circuitry.

Corticorubral synaptic organization in Macaca fascicularis: a study utilizing degeneration, anterograde transport of WGA-HRP, and combined immuno-GABA-gold technique and computer-assisted reconstruction.

The macaque red nucleus receives afferents from two major sources, the cerebral cortex and the deep cerebellar nuclei. Approximately 90% of the corticorubral afferent axons project to pars parvicellularis of the red nucleus, the neurons of which transmit information to the cerebellum by way of the inferior olivary nucleus. The remaining 10% project to pars magnocellularis of the red nucleus, the major projection of which is to the spinal cord. In this study, corticorubral terminations labeled following lesions or injections of wheatgerm agglutinin conjugated to horseradish-peroxidase into the topographically defined hand area of the primary motor cortex were quantitatively studied via electron microscopy. Cortical afferent terminals within pars parvicellularis and pars magnocellularis synapse upon all regions of the dendritic arbors of rubral projection neurons. However, the majority of these labeled afferents synapse upon thin-diameter shafts or presumed spinous processes of rubral distal dendrites as well as upon vesicle-containing profiles of presynaptic dendrites of local circuit interneurons that are gamma-aminobutyric acid-immunoreactive, as identified by postembedding immunohistochemistry. Synaptic contacts formed by the labeled cortical terminal were large in width and extended through several serial sections. Synaptic contacts formed by the presynaptic dendritic profiles, on the other hand, were more punctate and could be seen in only one or two serial sections. These latter synaptic interactions probably provide a modification of the effects of cortical input to rubral projection neurons as suggested by previous physiological studies that indicated the dominance of cortical input onto distal dendrites as well as involvement with inhibitory circuits. An example of the complexities of these synaptic interactions is further demonstrated by a three-dimensional computer reconstruction. This quantitative study of corticorubral afferents in the macaque monkey provides insight into the interactions of cerebral cortical afferents with rubral projection neurons and their relationship with local circuit inhibitory interneurons to elucidate the role played by the cortex in the activation of rubral neurons.

Analysis of gamma-aminobutyric acidergic synaptic contacts in the thalamic reticular nucleus of the monkey.

Gamma-aminobutyric acidergic (GABAergic) neurons in the thalamic reticular nucleus (TRN) spontaneously generate a synchronous bursting rhythm during slow-wave sleep in most mammals. A previous study at the electron microscopic level in cat anterior TRN has suggested that synchronous bursting activity could result from the large number of presumably GABAergic dendrodendritic synaptic contacts. However, little is known about the synaptology of the monkey thalamic reticular nucleus and whether it contains dendrodendritic contacts. To address this issue, we examined tissue obtained from Macaca fascicularis that was prepared for electron microscopy using postembedding techniques to demonstrate GABA immunoreactivity. Examination of the anterior (motor) and posterior (somatosensory) portions of the TRN disclosed the following: The majority of synaptic contacts (87.5% of 958) were formed by axon terminals showing no GABA immunoreactivity and making asymmetric synaptic contacts on dendrites or cell bodies. A further 6.4% of synaptic contacts was composed of GABA-immunoreactive presynaptic terminals making symmetric contacts with the dendrites of TRN neurons. The majority resembled the pleomorphic vesicle containing F-terminals seen in the dorsal thalamus and known to originate from axons of TRN. A subset or possible second class did not resemble any previously described class of GABA-immunoreactive terminals in the TRN. Both classes of these terminals making symmetric contacts may originate wholly or partially within the nucleus. There was one dendrodendritic synaptic contact and only a small number (3.2%) of axodendritic contacts with synaptic vesicles visible both pre- and postsynaptically. We conclude that dendrodendritic contacts are probably not responsible for the synchronized bursting neuronal activity seen in the slow-wave sleep of monkeys, and that, if TRN neurons are coupled synaptically, the most likely mechanism is through the synapses formed by recurrent axon collaterals of TRN neurons onto TRN dendrites.

Medial lemniscal and spinal projections to the macaque thalamus: an electron microscopic study of differing GABAergic circuitry serving thalamic somatosensory mechanisms.

The synaptic relationships formed by medial lemniscal (ML) or spinothalamic tract (STT) axon terminals with neurons of the somatosensory ventroposterolateral thalamic nucleus of the macaque monkey have been examined quantitatively by electron microscopy. ML and STT axons were labeled by the anterograde axon transport of WGA-HRP following injection of the tracer into the contralateral dorsal column nuclei, or the dorsal horn of the spinal cord, respectively. Thalamic tissue was histochemically reacted for the presence of HRP. Serial thin sections were stained with a gold-labeled antibody to GABA, to determine which neuronal elements exhibited GABA immunoreactivity (GABA-ir). Serially sectioned thalamic structures were recorded in electron micrographs and reconstructed in three dimensions by computer. Individual ML axon terminals form multiple synaptic contacts with segments of the proximal dendritic trees of thalamocortical relay neurons and also synapse upon the dendritic appendages of GABA-ir interneurons (local circuit neurons). These GABA-ir dendritic appendages contain synaptic vesicles and are presynaptic (presynaptic dendrites) to the same segments of relay neuron dendrites that receive ML contacts. When analyzed in serial sections and reconstructed by computer, the ML terminals form triadic relationships (ML, GABA appendage, and relay neuron dendrite) or more complex glomerular arrangements involving multiple appendages, all of which then contact the relay neuron dendritic segment. In contrast, multiple STT terminals make synaptic contacts along segments of projection neuron dendrites and are usually the only type of profile to contact that segment of dendrite. More than 85% of the spinal afferents form simple axodendritic synapses with relay cells and do not contact GABA-ir appendages. The thalamic synaptic relationships of ML terminals are fundamentally different from those formed by the STT. Because STT neurons predominatly transmit information about noxious stimuli, the simple axodendritic circuitry of the majority of these spinal afferents suggests that the transmission of noxious information is probably not subject to GABAergic modulation by thalamic interneurons, in contrast to the GABAergic processing of non-noxious information carried by the ML afferents. The differences in the GABAergic circuits of the thalamus that mediate ML and STT afferent information are believed to underlie differential somatosensory processing in the forebrain. We suggest that changes in thalamic GABAergic dendritic appendages and GABA receptors following CNS injury may play a role in the genesis of some central pain states.

Cerebellar terminations in the red nucleus of Macaca fascicularis: an electron-microscopic study utilizing the anterograde transport of WGA:HRP.

The red nucleus (RN) of the macaque monkey is divided into a rostral two-thirds, the parvicellularis (RNp), which projects to the cerebellum by way of the inferior olivary nucleus, and a caudal third, the magnocellularis (RNm), which projects to the spinal cord via the rubrospinal tract. The RNp and RNm receive afferents from two principal sources: the cerebral motor cortices and the deep cerebellar nuclei. The terminations of these two afferent projections tend to be spatially segregated on rubral neurons, in that most corticorubral afferents terminate on more distal dendrites, and those from the deep cerebellar nuclei terminate more proximally. The present electron-microscopic analysis of the cerebellar terminations in the macaque RN provides anatomical evidence for the presence of labeled afferents in both divisions of this motor nucleus, following injection of wheatgerm agglutinin conjugated to horseradish peroxidase (WGA:HRP) into the deep cerebellar nuclei and the anterograde transport of the tracer to the RN. The cerebellar terminal afferents are large; contain numerous mitochondria and primarily rounded synaptic vesicles; and form asymmetric synaptic contacts with rubral neurons. Unlike other terminals in the nucleus, they possess an electron-lucent cytoplasmic matrix and less densely packed synaptic vesicles. They are termed "large, round, pale" (LRP) terminals because of the morphological characteristics that distinguish them from other afferent terminal types found in RN. Labeled cerebellar afferents in RNp and RNm contact primarily neuronal somata, proximal dendrites emerging from the cell body, large-diameter dendrites, and the spines of rubral neurons that arise from somata and proximal dendrites.(ABSTRACT TRUNCATED AT 250 WORDS)

Inhibitory synaptic input to identified rubrospinal neurons in Macaca fascicularis: an electron microscopic study using a combined immuno-GABA-gold technique and the retrograde transport of WGA-HRP.

Rubrospinal neurons of the magnocellular division of the red nucleus of Macaca fascicularis were retrogradely labeled following spinal cord microinjections of wheat germ agglutinin-horseradish peroxidase, as demonstrated by the chromagen tetramethylbenzidine, identifying the mesencephalic cells of origin of this descending motor pathway. The tissue was processed for electron microscopy and subsequently tested on the electron microscope grid for immunoreactivity of gamma aminobutyric acid (GABA) in presumed local circuit neuronal somata, in dendrites, and in axonal terminals. Results demonstrate the presence of retrogradely labeled rubrospinal neurons of medium and large diameters (30-90 microns) and immunoreactive neurons of small size (less than 20 microns in diameter) within the nucleus. In addition, there are substantial numbers of GABAergic, presumably inhibitory, synaptic structures contacting somata and primary, medium, and small sized dendrites, as well as spineheads of rubrospinal neurons. The immunoreactive presynaptic profiles exhibit two different morphological appearances: one axonal and the other dendritic. Axonal terminals contain densely packed pleomorphic to flattened vesicles and form primarily symmetrical synapses with somata and all regions of the dendritic arbor. GABAergic profiles resembling presynaptic dendrites (PSDs) are also present. These profiles possess scattered flattened to pleomorphic synaptic vesicles in a translucent cytoplasm and are often postsynaptic to axonal terminals of unknown origin, or to GABAergic profiles. GABAergic local circuit neurons (LCNs), the neurites of which remain within the confines of the nucleus, appear to be contacted primarily by cortical and cerebellar afferents. These LCNs may or may not possess axons and thus may represent both the source of the GABAergic axonal terminals as well as that of the PSDs. Inhibitory afferents from other sources, such as the mesencephalic reticular formation, may also account for GABAergic terminals involved in this inhibition. We propose that the level of excitability of rubrospinal neurons and their subsequent activation of spinal motor neurons and interneurons is significantly regulated by the local circuit GABAergic inhibitory interneuronal population of the nucleus proper and probably by axons entering the nucleus from an extranuclear source.

Red nucleus of Macaca fascicularis: an electron microscopic study of its synaptic organization.

The parvicellular and magnocellular divisions of the red nucleus of the old world monkey, Macaca fascicularis, were analyzed at an electron microscopic level to examine the morphology of the synaptic profiles terminating on rubral neurons and to categorize them by their individual characteristics. The parvicellular division, or anterior two-thirds of the nucleus, is composed of small (10-15 microns) and medium-size (20-30 microns) cells, which are uniformly distributed with high packing density throughout this portion of the nucleus. These cells have invaginated nuclei and are often indented by blood vessels and glial cell somata (satellite cells) that lie in close proximity. The magnocellular portion, occupying the caudal one-third of the nucleus, is composed of an additional population of large cells, ranging from 50-90 microns in diameter, which often contain prominent lipofuscin granules and are frequently indented by blood vessels. Satellite glial cells are not a prominent feature in the magnocellularis portion of the nucleus. The large cells are separated one from the other by fields of myelinated axons either coursing through the nucleus or projecting to and from the nucleus itself. Although the divisions of the nucleus in the Macaca fascicularis are spatially distinct, each possesses a morphological similarity in regard to the categories of synaptic profiles seen at the electron microscopic level. These synaptic profiles are classified as follows: large terminals containing numerous, predominantly rounded vesicles (LR), which can often be seen to form the central profile in a synaptic glomerular arrangement; terminals of similar size with predominantly rounded vesicles but with a pale axoplasmic matrix (LRP); small profiles with rounded vesicles (SR); profiles containing granular dense-cored vesicles (DCV); profiles with numerous flattened vesicles (F); profiles containing pleomorphic vesicles (PL), some of which can be interpreted as presynaptic dendrites (PSD) because they are seen to be postsynaptic and contain ribosomes; and profiles with rounded synaptic vesicles, which are associated with subsynaptic Taxi bodies (T). Most of the various synaptic profile types were found to have similar distributions on the dendritic arbors of rubral neurons in both divisions of the nucleus. However, the LRP-type terminal predominates on the cell bodies and proximal dendrites of the large neurons in magnocellularis. Unlike other regions in the nervous system, F type terminals are rarely seen to contact neuronal somata. This study provides a basis for future experimental studies of afferents to the nucleus in this species.(ABSTRACT TRUNCATED AT 400 WORDS)

Ultrastructural evidence for direct monosynaptic rubrospinal connections to motoneurons in Macaca mulatta.

The magnocellularis division of the red nucleus of the Macaca mulatta, a midbrain structure involved in processing motor information, is known by light microscopic analysis to project, via the rubrospinal tract, to the contralateral intermediate horn of the spinal cord. Physiological studies, however, provide additional evidence for direct monosynaptic connections to motoneurons subserving distal musculature. This electron microscopic study demonstrates, by analyzing the anterograde transport of 5% wheatgerm agglutinin-horseradish peroxidase injected into the red nucleus, the presence of labeled terminals synapsing upon somata and proximal dendrites of motoneurons in the lateral portion of the ventral horn of the cervical enlargement of the spinal cord. We conclude that this anatomical evidence confirms the presence of direct monosynaptic connections to spinal motoneurons in the primate.

Anatomical evidence for red nucleus projections to motoneuronal cell groups in the spinal cord of the monkey.

In 4 rhesus monkeys wheat germ agglutinin-horseradish peroxidase (WGA-HRP) injections were made in the mesencephalic tegmentum. In 3 cases with injections involving the red nucleus (RN), rubrospinal fibers descended mainly contralaterally to terminate in laminae V, VI and dorsal VII of the spinal cord and in the lateral motoneuronal cell groups at the level of the cervical and lumbosacral enlargements. In all 4 cases the area of the interstitial nucleus of Cajal (INC) was injected, which resulted in labeled interstitiospinal fibers in the medial part of the ipsilateral ventral funiculus of the spinal cord. The results indicate that there is no major qualitative difference between the mesencephalic (RN and INC) and motor cortical projections to the spinal cord.

The use of specimen tilt in transmission electron microscopy of the central nervous system.

Thin sections of nervous tissue were viewed at different tilt angles using a transmission electron microscope equipped with a eucentric goniometer stage. In a comparison study of various degrees of tilt, one can observe additional morphological features within synaptic profiles, define subsynaptic structures such as Taxi-bodies, and clearly see the crystalline formation of cytochemical tracers. This study demonstrates the value of tilting thin-sections in the analysis of synapses and other biological material at the ultrastructural level.

Non-myelinated axons are rare in the medullary pyramids of the macaque monkey.

Previous electron microscopic studies of the medullary pyramids have concluded that non-myelinated axons constitute about 30-60% of all axons in the pyramid of the rat, and about 8-15% in the cat and monkey. Physiological studies of pyramidal tract axons have not found fibers conducting in the range predicted for non-myelinated axons, less than 1 m/s. This present study of the primate pyramid demonstrates that most of the profiles which could be interpreted as being non-myelinated axons when viewed in cross-section, are actually astroglial processes when examined in longitudinal section. We conclude that non-myelinated axons constitute less than 1% of the pyramidal tract axons in the old world adult primate.

Bulbospinal projections in the primate: a light and electron microscopic study of a pain modulating system.

The projections of the nucleus raphe magnus (NRM) and the immediately adjacent reticular formation were studied in the macaque monkey following injections of the rostroventral medulla with 3H-leucine and examination of the resultant labeled axons and terminals by light and electron microscopic autoradiography. Five monkeys had accurately placed injections, which resulted in fiber pathway labeling that coursed caudally, laterally, and dorsally to project to laminae I, II, and V of subnucleus caudalis of the trigeminal and then traveled in the dorsolateral funiculus of the cord and terminated in similar laminae of the spinal dorsal horn at cervical levels. The pathway was only lightly labeled caudal to the cervical enlargement and could not be readily discerned above background in the thoracic or lumbar cord. Electron microscopy revealed that axons and terminals serving this system constitute a heterogeneous population. Large-diameter myelinated axons (3-6-micron diameter), small myelinated axons (0.75-3-micron diameter), and clusters of nonmyelinated axons were labeled. Terminals in laminae I, II, and V contained mixtures of clear round and granular vesicles or clear pleomorphic and granular vesicles or formed the central element in synaptic glomeruli. The labeled profiles formed asymmetrical or symmetrical synapses on medium and small dendrites; labeled axosomatic synapses were not observed. In rare instances there were contacts between labeled profiles and vesicle-containing structures, which were probably dendritic, but whether the NRM axon was pre- or postsynaptic to such structures could not be determined. It was concluded that the NRM in the monkey is organized in a manner quite similar to that previously described in the cat. The wide variety of fiber types and synaptic terminals serving this system suggests that different classes of neurons participate in it, probably using several transmitter substances that result in varying postsynaptic effects on neurons located in the trigeminal complex and dorsal horn.

The terminations of corticospinal tract axons in the macaque monkey.

This study examined the corticospinal tract in monkey by utilizing the anterograde transport of wheat germ lectin conjugated to horseradish peroxidase (WGA HRP) at the light microscopic level and the axonal transport of 3H-proteins with both light and electron microscopic autoradiographic techniques. The animals survived 3-9 days after the injections of 3H-leucine or 3H-leucine/WGA HRP into either motor or sensory cortices. With the laminar schema of Rexed as a guide to the layers of the spinal gray matter, qualitative and quantitative analyses of labeled projections of the corticospinal tract (CST) were made. With the light microscope, axons from the sensory cortex labeled with WGA-HRP could be observed in the contralateral spinal gray from lamina I to the border of laminae VI/VII, the heaviest distribution being located in medial III-VI. There was a small ipsilateral projection to V and VI. With 3H label, laminae I and II revealed few overlying silver grains; many grains overlay laminae III-VI. Projections from the motor cortex labeled with either WGA-HRP or 3H extended from the contralateral laminae III/IV border into the motor nucleus (lamina IX) and were seen to be somewhat more dense in the lateral areas of the spinal gray. The motor cortex projected heavily to ipsilateral VIII, and in sparse amounts to ipsilateral V and VI. Electron microscopy of radioactive axons from the sensory cortex to dorsal horn revealed many radioactive myelinated fibers and some labeled non-myelinated axons. Labeled terminals contacted medium to small dendrites; there were a few labeled C-type profiles in glomeruli and occasional axo-axonal or dendro-axonal contacts, the labeled cortical axons being the postsynaptic structure. In ventral horn following motor cortex injections, the labeled axons were all myelinated. The synaptic contacts were found on small, medium, and large proximal dendrites as well as on cell bodies. Labeled terminals which formed the central element in glomeruli were also seen in this region. Most of the labeled corticospinal terminals in dorsal and ventral horn contained rounded vesicles, but a significant number revealed pleomorphic vesicles. The relationship of these morphological findings to physiological studies of the CST is presented.

Morphology and synaptic relationships of physiologically identified low-threshold dorsal root axons stained with intra-axonal horseradish peroxidase in the cat and monkey.

The arborizations and synaptic relationships of intra-axonally stained horseradish peroxidase- (HRP) labeled primary afferent fibers to the dorsal horn of the cat and monkey spinal cord have been studied by light and electron microscopic methods. The light microscopic arborizations of the afferent fiber types (hair follicle afferents, pacinian corpuscle afferents, type I and type II slowly adapting afferents) are similar to those described by Brown and his colleagues (1) in the cat. The synaptic profiles formed by labeled afferents contain rounded synaptic vesicles. In serial thin sections, it was found that single dorsal root axons may make hundreds or thousands of synapses with neuronal structures of the dorsal horn. The vast majority of synaptic contacts are on the dendritic trees of dorsal horn neurons. The synapses made by these low-threshold afferent axons are almost all in the deeper laminae (III-VI) of the dorsal horn. The hair follicle afferent axons and the pacinian corpuscle afferents have numerous vesicle-containing structures that synapse on them to form either axoaxonal synapses or dendroaxonal synapses. The slowly adapting afferent axons are less often found to be postsynaptic to axons or dendrites. It is concluded that different physiological classes of primary afferent axons have different morphological characteristics, both at the light and electron microscopic level.

The distribution of dorsal root axons to laminae IV, V, and VI of the Macaque spinal cord: a quantitative electron microscopic study.

The projections of dorsal root axons to the deeper laminae (IV, V, and VI) of the Macaque spinal cord were examined by the use of experimentally induced degeneration following dorsal rhizotomy or by injection of dorsal root ganglia with tritiated amino acids followed by light and electron microscopic autoradiography. Following dorsal rhizotomy, neurofilamentous degeneration of synaptic profiles occurs in each of the three deep laminae, more commonly in laminae IV and V than in lamina VI. The neurofilamentous degeneration is seen both in central glomerular (C) profiles and in many of the round vesicle (R) profiles. Neurofilamentous degeneration occurs as early as 18 hours following rhizotomy and the degenerating terminals are most numerous at 3-4 days postrhizotomy. None are seen after 7 days survival. The neurofilamentous profiles form axodendritic and, occasionally, axosomatic synapses with neurons of the dorsal horn. They are also seen to be postsynaptic to flat vesicle (F) profiles in axoaxonal synapses. A second type of degeneration, electron-lucent degeneration, is seen in laminae V and VI, and only occasionally in lamina IV. The lucent degeneration occurs somewhat later after rhizotomy than does the neurofilamentous degeneration and reaches its peak at 5 days postrhizotomy. No lucent terminals are seen after 7 days survival. Electron-dense degeneration, so common in lamina II, is not seen in the deeper dorsal horn. Autoradiographic techniques show that both C and R terminals are labelled in the deeper dorsal horn. Both of these terminals form axodendritic synapses and a significant number are found to be postsynaptic in axoaxonal synapses. Most of the C terminals degenerate following rhizotomy or are labelled following injection of the parent dorsal root ganglia with tritiated amino acids. Approximately one-fifth of the R profiles are derived from dorsal roots. F profiles do not appear to be of dorsal root origin in any case. It is concluded that neurofilamentous alterations represent the degeneration of larger-diameter (A beta) axons which distribute to the deeper dorsal horn and that electron-lucent degeneration represents the termination of A delta fibers. Electron-dense degeneration thought to represent the termination of nonmyelinated axons (C fibers) in the superficial dorsal horn is not seen in the deeper dorsal and it is concluded that C fibers do not project to the deeper laminae.

Identification of dorsal root synaptic terminals on monkey ventral horn cells by electron microscopic autoradiography.

The projection of dorsal root fibres to the motor nucleus of the macaque monkey spinal cord has been examined utilizing light and electron microscopic autoradiography. Light microscopy demonstrates a very sparse labelling of primary afferent fibres in the ventral horn. Silver grains overlying radioactive sources are frequently clustered into small groups, often adjacent to dendritic profiles. Under the electron microscope, myelinated axons and a few large synaptic profiles containing rounded synaptic vesicles were overlain by numerous silver grains. These labelled profiles made synaptic contact with dendrites 1-3 micron in diameter. The labelled profiles did not contact cell bodies or large proximal dendrites of ventral horn neurons. Frequently, small synaptic profiles containing flattened vesicles were presynaptic to the large labelled terminals and it is suggested that these axoaxonal synapses may mediate presynaptic inhibition of the primary afferent fibres. The relationship of the present findings to previously published physiological and anatomical studies is discussed.