THE JOURNAL OF C O M P A R A m NEUROLOGY 3W162-182 (1990)

Li&t Microscopic andultrastrudurall Analysis - ofGABA-Immunoreactiveh f d e s in the Monkey Spinal Cord SUSAN M. CARLTON AND ELIZABETH S. HAYES Marine Biomedical Institute and Department of Anatomy and Neurosciences, University of Texas Medical Branch, Galveston, Texas 77550

ABSTRACT

It is hypothesized that terminals containing y-aminobutyric acid (GABA) participate in presynaptic inhibition of primary afferents. To date, few convincing GABA-immunoreactive (GABA-IR)axo-axonic synapses have been demonstrated in support of this theory. The goal of this study is to document the relationship between GABA-IR profiles and central terminals in glomerular complexes in lumbar cord of the monkey (Macaca fuscicularis). In addition, the relationship between GABA-IR profiles and other neural elements are analyzed in order to better understand the processing of sensory input in the spinal cord. GABA-IR cell bodies were present in Lissauer's tract (LT) and in all laminae in the spinal gray matter except lamina IX. GABA-IR fibers and terminals were heavily concentrated in LT; laminae I, 11,and 111;and present in moderate concentration in the deeper laminae of the dorsal horn, ventral horn (especially in association with presumed motor neurons), and lamina X. Electron microscopic analysis confined to LT and laminae I, 11, and I11 demonstrated GAJ3A-IR cell bodies, dendrites, and myelinated and unmyelinated fibers. GABA-IR cell bodies received sparse synaptic input, some of which was immunoreactive for GABA. The majority of the synaptic input to GABA-IR neurons occurred at the dendritic level. Furthermore, the presence of numerous vesicle-containing GABA-IR dendrites making synaptic interactions indicated that GABA-IR dendrites also provided a major site of output. Two consistent arrangements were observed in laminae 1-111 concerning vesicle-containing GABA-IR dendrites: 1) they were often postsynaptic to central terminals and 2) they participated in reciprocal synapses. The majority of GABA-IR axon terminals observed contained round clear vesicles and varying numbers of dense core vesicles. Only on rare occasions were GABA-IR terminals with flattened vesicles observed. GABA-IR terminals were not observed as presynaptic elements in axo-axonic synapses; however, on some occasions, GABA-IR profiles presumed to be axon terminals were observed postsynaptic to large glomerular type terminals. Our findings suggest that a frequent synaptic arrangement exists in which primary afferent terminals relay sensory information into a GABAergic system for further processing. Furthermore, GABA-IR dendrites appear to be the major source of input and output for this inhibitory system. The implications of this GABAergic neurocircuitry are discussed in relation to the processing of sensory input in the superficial dorsal horn and in terms of mechanisms of primary afferent depolarization (PAD). Key words: primate, immunohistochemistry,electron microscopy, primary afferent depolarization

Physiological and pharmacological data indicate a role for y-aminobutyric acid (GABA) in inhibitory mechanisms (Roberts, '76; Levy, '77). Two types of inhibition are usually discussed, presynaptic and postsynaptic. GABA is particularly implicated in presynaptic inhibition of primary afferents and it has been postulated that GABA terminals form axo-axonic synapses on primary afferent terminals (Eccles et al., '62a, '63a). Although there are physiological and O

1990 WILEY-LISS, INC.

pharmacological data to support this hypothesis (Eccles et al., '63a-c; Barker and Nicoll, '72; Davidoff, '72; Miyata and Otsuka, '72, '75; Otsuka and Konishi, '76; Levy, '77), few convincing axo-axonic GABA synapses have been anatomically demonstrated in which the polarity of the synapse is beyond doubt (McLaughlin et al., '75; Barber et al., '78; Accepted July 10,1990.

GABA-IMMUNOREACTIVITYIN THE MONKEY SPINAL CORD Ribeiro-Da-Silva and Coimbra, '80; Westman et al., '84; Magoul et al., '87; Matthews et al., '88; however, see Basbaum et al., '86). Localization of GABAergic profiles has previously been accomplished by staining immunohistochemically for glutamic acid decarboxylase (GAD),the biosynthetic enzyme of GABA. In those studies, however, GABAergic cell bodies and dendrites were not visible unless the animal was pretreated with colchicine, an inhibitor of axonal transport (McLaughlin et al., '75; Wood et al., '76; Barber et al., '78, '82; Westman et al., '84). By contrast, using anti-GABA antibodies, GABAergic neurons are visible, and pretreatment with colchicine is unnecessary (Magoul et al.,'87; Meinecke and Peters, '87; Broman and Westman, '88; Carlton, '88; Broman and Blomqvist, '89; Hayes and Carlton, '89; Todd and McKenzie, '891, making ultrastructural studies of GABA-containing neurons and dendrites possible. In the present study, a detailed analysis of serial sections at the ultrastructural level reveals unanticipated synaptic arrangements involving GABA-containing profiles. The synaptology revealed here suggests alternative mechanisms through which this transmitter may exert its effects on sensory processing in the dorsal horn.

MATERIALSANDMETHODS

163

was observed were measured directly on the microscope slides.

Electronmicroscopic(EM) analysis For EM analysis, three animals were anesthetized as described above and perfused with a solution containing 3% paraformaldehyde, 3% glutaraldehyde, and 0.1% picric acid in 0.1 M PB. The entire spinal cord was removed, cut into small blocks, and postfixed for 6 hours. The tissue was placed in PB overnight at 4"C, and L 3 4 segments were sectioned on avibratome (Lancer Series 1000) at 25 pm and collected in 0.1 M PB. Tissue was incubated in 1% sodium borohydride for 30 minutes and rinsed in 0.1 M PB. Prior to incubation in the primary antibody, the tissue was exposed to a graded series of ethyl alcohol (lo%, 25%, 40%, 25%, and 10%) in 0.1 M PB for 5 minutes at each concentration to increase antibody penetration. The tissue was immunostained as described above, but Triton X-100 was not used. Following immunostaining, the tissue was placed in 1.0% osmic acid in cacodylate buffer for 1hour, stained with 1% uranyl acetate en block, dehydrated in ethanols, and flat embedded in a mixture of Epon-Araldite. Ultrathin serial sections were collected on bar grids or Formvar-coated slot grids.

Controls

GABA was conjugated to either hemocyanin (HC) or bovine serum albumin (BSA) according to the protocol of Hepler et al. ('88). The conjugates were then incubated with their proper antibody (GABA-HC GABA antibody from Chemicon; GABA-BSA + GABA antibody from Incstar). A complete absence of specific GABA immunostaining was observed in monkey spinal cord tissue when GABA antiserum (1:1000) was preadsorbed with the appropriate conjugated GABA antigen at concentrations of 100 kg/ml. Unconjugated forms of GABA, glycine, L-glutamate, beta-alanine, Light microscopicanalysis L-alanine, L-aspartic acid, and taurine processed with GABA antiserum at 1500 resulted in no reaction product For light microscopy, two deeply anesthetized animals were perfused through the aorta with warm heparinized using the immunoblot technique (Larsson, '81). There was saline, followed by a solution containing 4% paraformalde- also no specific immunostaining in sections incubated in hyde and 0.5% glutaraldehyde in phosphate buffer (PB). solutions lacking primary antiserum. The entire spinal cord was then removed, cut into small blocks and postfixed for 6-8 hours in fresh fixative. The RESULTS tissue was stored overnight at 4°C in PB containing 30% Light microscopic (LM)observations sucrose. Twenty-five micron sections from the L3-4 segGABA immunoreactive (GABA-IR) profiles were obments were cut on a freezing microtome. The sections were incubated in 1% sodium borohydride for 30 minutes and served at all levels of the primate spinal cord. In this study rinsed in 0.1 M PB. The tissue was immunostained with a we concentrated on the distribution of GABA-IR cell bodies modified version of the Sternberger ('79) peroxidase anti- and the synaptology of GABA-IR terminals in laminae I, 11, peroxidase (PAP) method. Antibodies to GABA obtained and I11 in lumbar cord. The concentration of GABA-IR cells from Chemicon and Incstar were used at dilutions of 1 5 0 0 was greatest in laminae 1-111, with a slight decrease in to 1:3,000. All immunoreagents contained Triton X-100 to laminae IV-VI (Figs. 1, 2A). Some GABA-IR cells were increase antibody penetration. Sections were incubated in usually visible in Lissauer's tract. The medial and lateral 3% normal goat serum (NGS) in phosphate buffered saline regions of dorsal lamina VII had small populations of (PBS) for 0.5 hour and placed in primary antibody for 24 GABA-IR cells, and some were seen in the dorsal part of hours at room temperature or 48 hours at 4°C. The tissue lamina VIII and in lamina X around the central canal. No was then washed in 1% and 3% NGS for 0.5 hour each and GABA-IR cells were observed in the ventral regions of placed in goat anti-rabbit gamma immunoglobulin (IgG, laminae VII and VIII or in any part of IX. In laminae 1-111, the GABA-IR cells were small and had a 1 5 0 ) for 1 hour. Following this incubation, the tissue was washed with 1% and 3% NGS for 0.5 hour each and placed round or oval profile. The nuclei were also immunostained in a solution of PAP (1:lOO) for 1 hour. The tissue was (Fig. 2B). Diameters of GABA-IR cells which displayed a rinsed in PB and reacted with DAB (0.05%) and H,O, for nucleus in these laminae ranged from 8.1 to 21.7 pm2,with 6-10 minutes. Following several rinses in PB, the sections an average of 12.5 ? .37 pm2 (X f S.E., N = 50). Very few were mounted on subbed slides, cleared, and coverslipped. dendrites could be seen radiating from these cells in the Diameters of GABA-IR positive cells in which the nucleus transverse plane.

Spinal cord tissue from five monkeys was used in the present study. The animals were initially anesthetized with a mixture of nitrous oxide, oxygen, and halothane gases and then injected with sodium pentobarbital (40 mg/kg, IV). Electrophysiological recordings in L6-7 segments were done in these animals for data collection in an unrelated study before perfusion. For this reason, L3-4 segments were taken for our anatomical studies.

+

164

S.M. CARLTON AND E.S. HAYES

n

Fig. 1. A map showing the laminar distribution of all GABA-IR cellular profiles in a 25 km section from the lumbar enlargement, generated using a camera lucida and focusing through all planes of the section.

Laminae IV and V appeared to have fewer GABA-IR neurons, but the sizes and shapes were similar to those observed in the superficial laminae. Many of the GABA-IR cells observed medial in laminae VI and VII were larger, with a mean diameter of 16.15 -+ .35 pm2 (N = 50), and at least one large dendrite was often seen in the transverse plane (Fig. 2C). In addition, along the gray/white matter border in medial laminae VI and VII, small, darkly staining GABA-IR cells were observed. This cell type was also found in lamina X around the central canal in no specific pattern. A small number of GABA-IR ependymal cells were observed with processes extending into the central canal. GABA-IR dendrites, axons and axon terminals. The intense amber color visible in all PAP immunostained lumbar sections was presumably due to the presence of GABA containing dendrites, axons and axon terminals distributed in the neuropil; this labeling was heaviest in

laminae I, 11, and I11 with small dark profiles randomly scattered throughout these layers (Fig. 2A). Presumed immunoreactive terminals could be observed in close apposition to unlabeled as well as to GABA-IR dorsal horn neurons. The size and appearance of GABA-IR profiles in Lissauer's tract and the dorsal horn indicated that both GABA-containing myelinated and unmyelinated fibers were present. A large population of presumed myelinated GABA-IR fibers was observed throughout the ventral horn; GABA-IR terminals were observed encrusting unlabeled presumed motoneurons (Fig. 2D). In the region of the central canal, GABA-IR fibers radiated toward the ependyma1 layer, and into the subependymal region.

Electron micrasmpic (EM) observations Variability of antibody penetration probably accounted for the differences in electron density of the immunoreac-

GABA-IMMUNOREACTJYI'IY IN THE MONKEY SPINAL CORD

165

Fig. 2. A A photomicrograph of the medial 7 3 of the dorsal horn demonstrating dense GABA immunoreactivity in the neuropil in the form of punctate dots (several indicated by open arrows), fiber-like processes (arrowheads), and GABA-IR neurons (arrows). EM analysis demonstrated that the immunoreactive punctate dots were GABA-IR dendrites, unmyelinated axons, and axon terminals; the fibers were myelinated axons. B: Higher magnification of round and oval GABA-IR

cells in lamina I (arrows). A nucleus (N) which is also GABA-IR is seen in one cell. C: An example of a GABA-IR neuron in lamina VI. These cells often demonstrated robust dendrites (arrowheads). The immunoreactive nucleus (N) is visible in this neuron. Another GABA-IR soma lies adjacent to this cell. D GABA-IR terminals (several identified by arrowheads) encrust two presumed motor neurons in the ventral horn. (Nomarski optics used in B-D). Bar = 100 pm for A,B; 10 pm for C,D.

tion product at the EM level. Some profiles were very heavily labeled (with the ultrastructure obscured) and others were lightly labeled, appearing only slightly denser than background. No positive immunoreactivity was observed in glial profiles. All thin sections were analyzed without lead citrate staining. Lissauer 's tract (LT). Numerous GABA-IR profiles were observed in the LT. These profiles included somata, small to medium sized dendrites, myelinated and unmyelinated axons, and axon terminals (Fig. 3A). The cell bodies in LT were similar in morphology and input to cells in the dorsal horn in that they were surrounded mainly by glial processes, receiving few axosomatic contacts (on the average, 4-5 contacts in one plane of section). Furthermore, two

or three of these contacts were usually GABA-IR (Fig. 3A). The GABA-IR cells typically had immunoreaction product associated with the rough endoplasmic reticulum in the cytoplasm and the outer mitochondria1 membranes (Fig. 3B). The cell illustrated in Figure 3 was contacted by a GABA-IR vesicle-containingdendrite.

DorsalHorn-LaminaeI,II,andIII Myelinated and unmyelinated GABA-IRaxons. In the dorsal and ventral horns, numerous axons were labeled for GABA. They ranged in size from small unmyelinated fibers, usually found in axonal bundles (Fig. 4A), to small and medium diameter myelinated axons which were heavily labeled with immunoreaction product (Fig. 4B).The immu-

166

S.M. CARLTON AND E.S. HAYES

Figure 3

GABA-IMMUNOREACTIVITYIN THE MONKEY SPINAL CORD

167

Fig. 4. Numerous unmyelinated (A) and myelinated (B)GABA-IR axons (arrows) are observed throughout the dorsal horn. The immunoreaction product is mainly associated with the microtubules but is also seen diffusely labeling the axoplasm. Bars = 0.5 km.

noreaction product was associated with the microtubules and also seen as diffuse labelingin the axoplasm. GABA-IRsomata. At the ultrastructural level, immunostained cell bodies were round or oval with an intensely immunoreactive nucleus and a thin rim of cytoplasm. GABA-IR cell bodies in the dorsal horn were similar to those in LT, being surrounded mainly by glial processes and unmyelinated axons, with relatively few somatic contacts.

Fig. 3. A An electron micrograph demonstrating the various GABA-IR profiles found in Lissauer’s tract, including a GABA-IR soma (nucleus is labeled with “N”), dendrites (D),myelinated farrow) and unmyelinated (arrowheads) axons and terminals (open arrows). This GABA-IR soma is typical of those found throughout the dorsal horn, being mainly surrounded by glial processes and a few unmyelinated axons. Of the five synaptic profiles contacting the soma, three are GABA-JR. Bar = 1.5 &m. B: Higher magnification of the GABA-IR dendrite shown in A, makinga dendrosomatic synapse. A small group of vesicles can be seen near a widened synaptic cleft (arrows). Immunoreaction product can be seen associated with the mitochondrial membranes and the rough endoplasmic reticulum in the cell cytoplasm (arrowheads). An unlabeled terminal (a) is also observed synapsing on the cell. Bar = 0.5 pm.

GABA-IR dendrites and their synaptic interactions. The most common immunoreactive profiles in the dorsal horn were GABA-IR dendrites. These were oriented either perpendicular or parallel to the long axis of the spinal cord. Heavy reaction product was associated with the microtubules and the outer mitochondrial membranes. Figure 5 illustrates a GABA-IR dendrite cut in longitudinal section showing both of these immunostaining characteristics. This figure also illustrates the fact that GABA-IR dendrites receive a great deal of synaptic input along their length, in contrast to the limited synaptic input to GABA-IR cell bodies. A consistent feature of GABA-IR dendrites was the presence of round clear vesicles (Figs. 3B, 6-10). These vesicles were usually located in discrete pocket^" within the dendrite. In any one plane of section, not all GABA-IR dendrites demonstrated vesicles; but serial sections often revealed their presence. Furthermore, serial sections also showed that in the majority of cases, these vesiclecontaining profiles were not varicose enlargements separated by a narrow process. Rather, the diameter of the dendritic profiles remained relatively constant, and only gradually increased or decreased over the distance ana-

S.M. CARLTON AND E.S. HAYES

168

Fig. 5. A longitudinal section through a GABA-IR dendrite in the dorsal horn. Immunoreaction product is seen associated with the microtubules (several identified by arrowheads) and the outer mitochondria1 membranes. Several unlabeled terminals are observed malung synaptic contact (arrows) along the length of the dendrite. Bar

= 0.5 wm.

lyzed. Examples were observed, however, where there appeared to be a narrowing of the dendrite before the terminal region. Figure 6 demonstrates a GABA-IR neuron and proximal dendrite with numerous microtubules and rough endoplasmic reticulum. Note that the dendrite does appear to taper, prior to giving rise to a vesicle-containing profile. Vesicle-containing GABA-IR dendrites were observed making dendro-somatic, dendro-dendritic, and dendroaxonic synaptic contacts. In the case of dendro-somatic and dendro-dendritic contacts, the postsynaptic elements were often also GABA-IR (Figs. 3B, 7A). The dendro-axonic contacts included GABA-IR vesicle-containing dendrites presynaptic to small unlabeled terminals containing mainly round clear vesicles (Fig. 7C) or large central terminals (Fig. 10).Vesicle-containingGABA-IR dendrites were found postsynaptic to unlabeled dendrites (not shown); small axon terminals (Fig. 7B); and large, unlabeled glomerular type terminals (Figs. 7A, 8-10).

GABA-IRprofilespostsynaptic to glomerular terminals. One striking feature was the observation that presumed GAI3A-IR terminals and more frequently GABA-IR vesicle-containing dendrites were positioned postsynaptic to central terminals of glomeruli. These glomerular terminals were often scalloped, indented by several postsynaptic profiles, and were very similar to those types ckescribed by Knyihar et al. i'82a). One type contained many clear round

vesicles of varying sizes, no dense core vesicles but a dense axoplasm and was called a dense sinusoid axon (DSA, Figs. 7A, 9B); another contained clear round vesicles of a small uniform size and was called a regular synaptic vesicle terminal iRSV, Figs. 8, 10A); and a third had numerous dense cores, as well as clear round vesicles and was called a large dense core vesicle terminal (LDCV; Figs. 9A, 12). As illustrated in the drawing in Figure 13, vesicle-containing GABA-IR dendrites were observed postsynaptic to all three types of primary afferent terminals. Reciprocal synapses involving GABA profiles. On several occasions, analysis of serial sections demonstrated the presence of reciprocal synapses between GABA-IR vesicle-containing dendrites and large unlabeled glomerular terminals. This arrangement was seen most often in lamina 111. Figure 10 demonstrates an asymmetrical and a symmetrical synapse, polarized in opposite directions. This particular reciprocal synapse occurred on a small intervaricose chain, in close proximity to an enlarged varicosity. In all cases where reciprocal synapses were followed in serial sections, the GABA-IR profile was a vesicle-containing dendrite that interacted with a large scalloped, central terminal. GABA-IR synaptic profiles: presumed axon terminals. GABA-IR profiles that were completely filled with vesicles were considered to be representative of axon terminals. GABA-IR axon terminals observed in the dorsal horn

GABA-IMMUNOREACTIVITYIN THE MONKEY SPINAL CORD

Fig. 6. A: In this micrograph, a GABA-IR soma with an immunostained nucleus (N)gives off a proximal dendrite containing rough endoplasmic reticulum (arrows). The dendrite in turn gives rise to a swelling that contains mitochondria, clear round vesicles, and a few dense core vesicles. Bar = 1.0 km. B: A higher magnification of the

169

vesicle-containing swelling shown in A. This profile makes an asymmetrical synapse on an unlabeled dendrite (D), demonstrating a widened cleft and a postsynaptic density (arrowheads). C : A serial section demonstrating an aggregation of clear round vesicles, widened cleft, and postsynaptic density. Bar = 0.34 km for B,C.

170

Fig. 7. Dendrodendritic synapses. A: Large glomerular type axon terminal (A), with morphology typical of a dense sinusoid axon (DSA), is making synaptic contact (arrows) with a GABA-IR vesicle-containing dendrite of medium diameter. This dendrite in turn makes synaptic contact (arrowheads) with a smaller GABA-IR dendrite. €3: A small cluster of vesicles is observed in this GABA-IR dendrite (arrows), where

S.M. CARLTON AND E.S. HAYES

it synapses on an unlabeled dendrite. This GABA-IR dendrite receives synaptic contacts (arrowheads) from two unlabeled axon terminals (A). C: Dendroaxonic synapse: a vesicle-containing GABA-IR dendrite is synapsing (arrows) on a small unlabeled terminal. Another unlabeled axon terminal (A) is seen making synaptic contact (arrowhead) with this same profile. Bars = 0.5 pm.

GABA-IMMUNOREACTIVITYIN THE MONKEY SPINAL CORD

Fig 8. This glomerular primary afferent terminal (A) has morphology typical of a regular synaptic vesicle terminal (RSV).Three GABA-IR dendrites (D) are observed in contact with it, two demonstrate asymmetric synaptic interactions (arrows). Taxi bodies are evident at the

171

postsynaptic sites. One dendrite has a small collection of clear round vesicles (arrowhead). An unlabeled axon terminal (a) is seen making synaptic contact with the glomerular terminal (arrowhead). Bar = 0.5 pm.

S.M. CARLTON AND E.S. HAYES

172

Figure 9

GABA-IMMUNOREACTIVITY IN THE MONKEY SPINAL CORD were of three morphological types. The most common had round clear vesicles which were loosely arranged and contained some dense core vesicles (Fig. 11A). A second terminal type contained round clear vesicles which were very densely packed and some dense core vesicles (Fig. 11B). These two types were observed more or less randomly distributed throughout the dorsal horn. A third type, rarely observed, contained flattened or oblong clear vesicles and no dense core vesicles (Fig. 11C). GABA-IR terminals made synaptic contact with labeled and unlabeled dendrites and cell bodies. They were not observed as presynaptic elements in axo-axonic synapses; however, on some occasions, GABA-IRprofiles presumed to be axon terminals were observed postsynaptic to large glomerular type terminals (Fig. 12). On several occasions, presumed GABA-IR axon terminals were observed in contact with glomerular type terminals. Analysis of serial sections, however, failed to show a synaptic junction between the two profiles. In the ventral horn, medium to large diameter GABA-IR axon terminals packed with round clear vesicles and numerous mitochondria were observed (Fig. 11D). These synapsed on cell bodies, dendrites of all diameters, and dendritic spines. G A B A - Z R synaptic complexes in laminae I, II, and ZZZ. There were many examples of GABA-IR profiles participating in complex synaptic arrangements. The most common involved a GABA-IR vesicle-containing dendrite postsynaptic to an axon terminal (either glomerular or simple dome-shaped terminal). The GABA-IR vesiclecontaining dendrite was in turn presynaptic to a small diameter dendrite or dendritic spine (which could be labeled or unlabeled). In Figures 7A and 9B, the first presynaptic element in each complex was a glomerular type axon terminal. In Figure 7B, two small, dome-shaped axon terminals synapse on a GABA-IR vesicle-containing dendrite, which in turn synapses on an unlabeled dendritic spine. These GABAergic interactions are summarized in Figure 13.

DISCUSSION To our knowledge, the present study is the first concerning GABA-IR labeling in the lumbar dorsal horn of the monkey. Previous studies of the GABAergic system and its neurocircuitry in rat and cat spinal cord and medullary dorsal horn were hampered by the fact that glutamic acid decarboxylase (GAD)-containing cell bodies and dendrites were not visible unless colchicine was used to block axonal transport of the amino acid out of the cell body (McLaughlin et al., '75; Wood et al., '76; Barber et al., '78, '82; Westman et al., '84). With anti-GABA, cell bodies as well as dendrites and axon terminals immunostain without colchicine pretreatment, allowing excellent localization of GABA-

Fig. 9. A Demonstration of a large dense core vesicle terminal (LDCV) glomerular type axon terminal (A) presynaptic to two GABA-IR profiles (arrows indicate the presence and polarity of the specializations). The postsynaptic profiles are a dendrite (D) and a smaller profile which appears to be a terminal containing round clear and dense core vesicles. Both synapses appear to be asymmetrical. B Demonstration of an RSV glomerular type axon terminal (A) presynaptic (arrows) to a GABA-IR vesicle-containing dendrite. The dendrite in turn synapses on an unlabeled dendritic spine (arrows). Bar = 0.5 pm.

173

containing profiles and GABAergic synaptic interactions at the LM and EM levels.

GABA-JRprofiles in Lissauer's tract Our observation of numerous GABA-IR profiles in LT, including cell bodies, dendrites, myelinated, and unmyelinated axons, was surprising since this region has not been previously highlighted as a "GABA-enriched" area. Magoul et al. ('87) described only GABA labeled myelinated fibers in LT of the rat and Todd and McKenzie ('89) observed what they termed fine immunoreactive profiles. Both propriospinal and primary afferent fibers, as well as a small population of neurons, are present in the LT (Rexed, '52; Earle, '52; LaMotte, '77; Chung and Coggeshall, '82, '83), with propriospinal fibers composing 20% of the fiber population in the primate LT (Coggeshall et al., '81). To date, GABA has not been localized within dorsal root ganglion cells, nor is it reduced in the spinal cord following dorsal rhizotomy (Roberts and Keen, '74; Takahashi and Otsuka, '75). Thus it can be presumed that the majority of the GABA-IR axons in LT are of propriospinal origin. In support of this hypothesis, lesions of LT in the monkey produces degeneration almost entirely within laminae 1-111 (LaMotte, '77). This is consistent with our report of a large population of GABA-IR cells and fibers in these laminae. Integration of dorsal horn function is achieved in part by way of LT (see Kerr, '75 for refs.; Scheibel and Scheibel, '68), and the presence of numerous GABA-IR cell bodies, fibers, and axon terminals in this region suggests that GABA plays an important role in the coordination of intersegmental activities.

GABA-IRsomata Most of the studies in rat and cat show a higher concentration of GABA cells in laminae I, 11, and I11 compared to the deeper laminae (Ribeiro-Da-Silva and Coimbra, '80; Hunt et al., '81; Magoul et al., '87; Todd and McKenzie, '89; however, see Fuji et al., '85). Although the largest population of immunostained cells is found in the superficial dorsal horn in monkey, our data demonstrate that the deeper laminae also contain many GABA-IR cells (lamina IX excluded). This discrepancy between species may be due to a limited penetration into the spinal cord gray matter of exogeneously applied colchicine. This would account for the large concentration of cells observed in the superficial laminae, while few to none are visible in deeper laminae in other studies. As with previous observations in other species (RibeiroDa-Silva and Coimbra, '80; Takasu et al., '87; Magoul et al., '87; Broman and Blomqvist, '89; however, see deBiasi et al., '86), GABA-IR cell bodies in the monkey receive few somatic contacts, being enclosed by glial processes and unmyelinated axons. Of the few somatic contacts observed in any one plane of section, approximately half are GABAIR. Interactions between GABA-IR profiles suggests a feedback mechanism through which GABA input and output are regulated.

GABA-JRfibers The presence of myelinated and unmyelinated GABA-IR and GAD-IR fibers in the dorsal horn has also been reported in the rat and cat (Hunt et al., '81; Magoul et al., '87; Matthews et al., '88), and thinly myelinated GABA-IR axons have been described in the monkey lateral cervical nucleus (Broman and Blomqvist, '89). The presumed source

174

Fig. 10. A Demonstration of a reciprocal synapse (arrowheads) between a GABA-IR vesicle-containing dendrite (D) and a large RSV glomerular type axon terminal (A). Bar = 0.5 pm. B: Higher magnification of the reciprocal complex shown in A A small collection of clear vesicles and an associated postsynaptic density is present in the

S.M. CARLTON AND E.S. HAYES

axodendritic component of the reciprocal. The dendroaxonic component also demonstrates a collection of clear round vesicles, however, there is no apparent postsynaptic density. C : Although the clear vesicles are no longer evident, the postsynaptic density is still apparent in this serial section through the axodendritic synapse. Bar = 0.25 p,m for B and C.

GABA-IMMUNOREACTIVITY IN THE MONKEY SPINAL CORD

175

Fig. 11. The morphology of presumed GABA-IR terminal profiles observed in the dorsal horn included A) profiles loosely packed with round clear vesicles and many dense core vesicles (arrowheads); B) profiles very densely packed with round clear vesicles and a few dense core vesicles (arrowheads); C) profiles containing flattened or oblong clear vesicles and no dense cores. Due to the dense immunoreaction product, it was difficult to determine the nature of these synaptic

interactions, however, many appeared to be asymmetric (arrows in A and B). The GABA-IR terminals synapsed on dendritic shafts of small and medium diameter. D: In the ventral horn, the synaptic profiles had a very different character, being much larger in diameter with many mitochondria, numerous round clear vesicles and few dense core vesicles. This synaptic profile is in contact with a dendrite and dendritic spine (s)of a presumed motor neuron. Bars = 0.5 bm.

of these fibers are the numerous GABA-containing neurons in the spinal cord gray matter, but the actual segmental origin of any one fiber is unknown since it is now clear that GABA cells send fibers to other segmental levels via LT. Furthermore, Millhorn et al. ('87) have demonstrated a population of GABA-containing cells in the caudal medullary raphe which project a t least to the thoracic cord. Thus, the presence of GABA-IR fibers in the cord originating from a supraspinal source must also be considered.

GABA-IRv&cleeontainingprofdes axon terminalsor dendrites? It is a difficult task trying to classify GABA-IR vesiclecontaining profiles as axon terminals or dendrites. Due to the fact that the immunoreaction product obscures the presence of ribosomes in the cytoplasm, this distinguishing feature used in previous studies for recognition of vesiclecontaining dendrites (Ralston, '71; Gobel, '76; Gobel et al.,

S.M. CARLTON AND E.S. HAYES

176

Fig. 12. Electron micrograph demonstrating a GABA-IR terminal (a) postsynaptic (arrowheads) to a large central terminal (A). A thin process, presumably an axon cylinder, leads into the terminal swelling (arrows).Bar = 0.5 Km.

'80; Knyihar-Csillik et al., '82a,b; Ribeiro-Da-Silva et al., '85) could not be used in the present study. Furthermore, the demonstration of a vesicle-filled spine branching from a GABA-IR dendrite (not shown) indicates the impossibility of predicting the origin of any vesicle-filled profile, based on morphology, from a single section. Figure 6 is another case in point, demonstrating a vesicle-filled profile extending from a proximal dendrite. Cut in a different plane of section, the vesicle-filled profile might have been classified as a terminal bouton arising from an axon. These particular examples force one to be extremely cautious when interpreting the origin of a vesicle-containing profile. In the present study, the characteristics of GABA-IR vesicle-containing profiles were first established through analysis of serial sections. In some cases, immunostained profiles were followed in serial sections back toward the cell body of origin, to the point where organelles such as rough endoplasmic reticulum were visible, indicating that the profile was dendritic in nature. Other vesicle-containing profiles were followed through serial sections to ascertain if they were varicosities along a beaded chain (axons en passant) or axon terminals (axon terrnznaun). If the former were true, the varicosities were replaced by intervaricose axons or thin axon cylinders (shown in Fig. 12) in subsequent serial sections. If the latter were true, the diameters

of the profiles changed in the sections nearing either end of the terminal (as in slicing through a sphere), and then the profiles would disappear. Following several GABA-IR profiles which we presumed to be vesicle-containing dendrites, the diameters of profiles

__ Fig. 13. Drawing summarizing the synaptic interactions of a GABAergic cell (G) and its associated dendrites. With the exception of the primary afferent terminals labeled DSA, LDCV, and RSV, all profiles are GABA-IR. In our analysis, we observed three morphological types of primary afferent terminals in synaptic contact with GABA dendrites. Sometimes these interactions were in the form of reciprocal synapses (dendrite 1 and 2) or as a simple postsynaptic element (dendrite 3). As illustrated, triadic arrangements were often observed in which a primary afferent made synaptic contact with a vesiclecontaining GABA-IR dendrite, which in turn synapsed on another GABA-IR dendrite. Presumed GABAergic axon terminals (a) have been observed postsynaptic to LDCV primary afferents, however, the origin of the GABAergic axan terminals (intrinsic or descending source) is unknown. GABAergic axon terminals have been observed in relation to RSV primary afferents, however, no synaptic specializations have been found. The GABAergic cell body has limited synaptic input, often being contacted by GABAergic axon terminals and vesicle-containing dendrites. The simple axodendritic synapses observed between GABA-IR terminals and unlabeled dendrites are not pictured here.

GABA-IMMUNOREACTIVITY IN THE MONKEY SPINAL CORD

Figure 13

L 77

178

S.M. CARLTON AND E.S. HAYES

gradually increased or decreased through the serial sec- postulated that GABAergic terminals synapse on primary tions, as might be expected when moving proximally or afferents (axo-axonic synapses) to produce primary afferent distally through a dendrite. Ralston ('71) described profiles, depolarization (PAD) (see Schmidt, '71; Levy, '77 for which he presumed to be vesicle-containing dendrites, as review). In pain pathways, Melzack and Wall ('65) proposed having vesicles that occupied less than one half of the total a gate control mechanism in which incoming afferent area of the dendritic profile. Conversely, profiles thought to impulses are modulated by substantia gelatinosa inhibitory be axon terminals were nearly filled with synaptic vesicles interneurons (possibly GABAergic neurons). Data corroboand usually had no visible neurofilaments. Applying this rating these theories has been generated from a study in morphological criterion to our data would lead us to the which dorsal rhizotomies were performed and GABAergic same conclusions concerning the origin of profiles (e.g., axons were observed contacting degenerating profiles (Barcompare density and arrangement of vesicles in the axon ber et al., '78). The three classical features that define a terminals in Fig. 11with those in dendrites in Fig. 7). There synapse (clustering of vesicles near the synaptic membrane were several examples in which vesicle-containing den- on the presynaptic side, the presence of densities at the drites received synaptic contacts along the shaft (see Figs. synaptic zone, and a widened cleft (Peters et al., '76)), are 7-10). It would be unusual to find synaptic contacts on an extremely difficult to discern if one element is degenerating. axonal shaft (however, see Duncan and Morales, '78). We Thus it is difficult to determine the presence and polarity of did not serially section through all of the GABA-IR vesicle- synapses between GABAergic profiles and degenerating containing profiles described here, and, therefore, we have primary afferents. Adding to the confusion is the presence based conclusions as to the nature of certain profiles on of vesicle-containing dendrites in the dorsal horn, which data accumulated from serial sections and from previous makes classification of a presumed presynaptic profile from studies concerning vesicle-containing dendrites (Lund, '69; one ultrathin section tenuous. In another EM study, MaxRalston, '71; Lieberman, '73; Gobel, '76; Ralston, '79; well and Noble ('87) iontophoretically labeled hair follicle Gobel et al., '80; Knyihar-Csillik et al., '82a,b). In all cases, primary afferent terminals with horseradish peroxidase every attempt was made to avoid "over-interpretation'' of (HRP) and subsequently immunostained this tissue with the data and only conservative conclusions are reported. anti-GAD. Due to the difficulties imposed by the combined techniques, it was impossible to identify any postsynaptic GABA-IRaxon terminals densities when GAD-labeled boutons were intimately assoThe distribution of GABA-IR terminals in the monkey is ciated with HRP-labeled primary afferents. Consequently, similar to that described in the rat (McLaughlin et al., '75; to date few convincing axo-axonic synapses have been Hunt et al., '81; Magoul et al., '87), with a slightly heavier demonstrated in which a GABAergic terminal is presynapconcentration in the upper laminae and somewhat less in tic to a primary afferent. the remaining layers of the gray matter. In the present study, axo-axonic synapses have been There are three possible sources of GABA input to any observed in which unlabeled terminals are presynaptic to spinal cord segment: 1)intrinsic GABA neurons giving rise glomerular terminals (see Fig. 8).We also observed GAEiA-IR to fibers which remain within the segment, 2) GABA fibers terminals in apposition to glomerular terminals. Following originating from other segments, coursing via LT or other exhaustive analysis of serial sections, however, the classical spinal cord funiculi to adjacent andlor distant segments, 3) characteristics of a synapse have not been observed in the GABA fibers arising from supraspinal sites (Millhorn et al., latter case. Thus, no unequivocal examples of GABA-IR '87). Although only speculative at this time, the morphol- terminals presynuptic to large glomerular terminals have ogy of a GABA-IR terminal may be indicative of its origin been demonstrated. Since there are reports in other species since in this study, three morphologically different GABA-IR of GABA-IR axon terminals presynaptic to primary afferent terminals are present in the neuropil. endings, our findings in the monkey could represent a Gray type I1 synapses, which display symmetrical densi- species difference. We believe this is unlikely, especially in ties at the synaptic specializations and contain flattened or light of the fact that review of the micrographs in the pleomorphic vesicles have been considered representative previous studies leave room for doubt as to the polarities of of axon terminals with inhibitory functions (Uchizono, '65). many of the interactions. Since GABAergic terminals in the GABA is presumed to have an inhibitory function in the presynaptic position were an "expected" anatomical arcentral nervous system; however, most GABA-IR terminals rangement, forecasted from the physiological and pharmaobserved in the present study do not fit this classical cological data, the anatomical interactions observed in description. The vast majority form asymmetrical synapses earlier studies may have been overinterpreted. Another and contain round clear vesicles with varying numbers of possibility is that the previous studies demonstrated one dense core vesicles. Only on rare occasion are pleomorphic component of the reciprocal synapse reported here (i.e., the vesicles observed in GABA-IR terminals. This is in agree- GABAergic dendrite synapsing on the glomerular t,ermiment with the findings of Broman and Blomqvist ('89) in nal). This is a distinct possibility, since analysis of serial the monkey, Matthews et al. ('88) and Basbaum et al. ('86) sections was often necessary to document the presence of in the cat, and Magoul et al. ('87) in the rat. However, this the two synaptic specializations in reciprocal synapses. It observation disagrees with the findings of Takasu et al. may also be possible that the presence of a presynaptic ('87) in the monkey and McLaughlin et al. ('75) in the rat. density and an aggregation of synaptic vesicles may have Since a predominance of round vesicles in GABA-IR termi- been obscured by the dense PAP reaction product in the nals is found in most studies, a difference in fixation GABA-IR terminals. Future studies in which GABA termimethods may account for the disparity. nals are labeled with post-embedding immunogold may Presynaptic inhibition has been suggested as one mecha- clarify whether the observed apposition is really a synaptic nism through which inhibitory actions of GABA are ex- interaction. Finally, it may be that GABAergic terminals pressed in the dorsal horn. For example, it has been exert their affects "nonsynaptically" through diffusion,

GABA-IMMUN'OREACTIVITYIN THE MONKEY SPINAL CORD making synaptic specializations unnecessary, as is the case in some peripheral autonomic interactions (Richardson, '64; Jonakait et al., '79; Burnstock, '79). There were several observations of presumed GABA-IR terminals postsynaptic to central terminals. An axo-axonic arrangement with a primary afferent as the presynaptic element and an axon terminal as the postsynaptic element has been described previously in the dorsal horn (Ralston, '68a; Knyihar et al., '74; Egger et al., '81). Furthermore, in a study of the cat trigeminal nucleus caudalis, Basbaum et al. ('86) demonstrated a similar arrangement where central terminals were presynaptic to GAD immunoreactive terminals. In further support of these findings, we have identified primary afferents containing calcitonin gene-related peptide presynaptic to GABA-IR axon terminals (Hayes and Carlton, '89).

GABA-IRdendrites Most of the interactions in which GABA-IR cells are involved occurred at the dendritic level. Large and medium sized GABA-IR dendrites cut in cross section are literally covered by axon terminals making synaptic contacts. The presence of numerous GABA-IRdendrites containing aggregations of clear round vesicles a t synaptic specializations is additional evidence that dendrites of GABA cells are a major site for processing of output as well as input. It should be noted that GABAergic vesicle-containing dendrites have also been reported in other regions of the neuroaxis (Ribak et al., '77; Streit et al., '78) and GABAergic vesicle-containing dendrites have been electrophysiologically demonstrated to be the source of inhibition through local dendritic interactions (Jahr and Nicoll, '80). Ralston ('71) addressed the possible functions of vesiclecontaining dendrites present in the cat ventrobasal thalamus, suggesting that some action potentials may invade the cell body and cause release of the synaptic transmitter from vesicles located only in proximal dendrites, leaving the distal dendrites more or less unaffected. In addition, local membrane changes occurring in discrete regions of a GABAergic dendrite produced by terminals synapsing on it could result in a local potential change and local release of GABA transmitter. Thus, each branch of a GABA dendritic tree might be capable of responding separately and discretely to incoming input, regardless of the neural activity occuring in other dendritic branches or in the parent cell. In a system where each dendritic branch or even multiple sites along the same dendritic branch can independently influence postsynaptic elements, many possibilities arise as to the final output(s) of one particular GABAergic neuron. The normal rules governing spatial and/or temporal summation which result in a single output do not apply. Furthermore, based upon the mathematical neuron models presented by Rall('62, '641, when a GABAergiccell is activated, depolarization might spread from the cell body into the dendritic branches, resulting in release of inhibitory transmitter at synaptic specializations in the GABA-IR dendrites. A scheme in which inhibition is mediated by nonpropagated depolarization of dendritic trees rather than by conduction through axons is suggested by Rall et al. ('66) in the olfactory bulb. GABAprofiles are postsynaptic to glomerular terminals. Knyihar-Csillik et al. ('82a) demonstrated the presence of three different types of primary afferent glomerular terminals in the monkey dorsal horn: dense sinusoid axons

179

(DSA), large dense core vesicle (LDCV) terminals, and regular synaptic vesicle (RSV) terminals. Their classification is based mainly on the content and distribution of synaptic vesicles. In the present study, we demonstrate vesicle-containing GABA-IR dendrites postsynaptic to each type of glomerular primary afferent terminal described by Kniyhar-Csillik et al. ('82a). For example, the glomerular terminals in Figures 7A and 9B fit the description of a DSA terminal, the terminal in Figure 9A fits the description of an LDCV terminal, and those in Figures 8 and 10 fit the description of an RSV terminal. Thus, all three of these primary afferent terminal types are presynaptic to GABA-IR profiles (summarized in Fig. 13). Each terminal type is believed to arise from a different class of axon. The DSA and LDCV terminals are regarded as arising from A delta and C fibers (Light and Perl, '79; Knyihar-Csillik et al., '82a; Carlton et al., '87; McNeill et al., '881, and RSV terminals probably arise from A beta and A delta fibers (LaMotte, '77; Light and Perl, '77; Knyihar-Csillik et al.,'82a). Thus, our findings suggest that different types (i.e., modalities) of primary afferent information are subjected to similar types of GABA processing. At the first synapse in the dorsal horn, the three glomerular types transmit their information to a GABAergic system. Presumably this information is modulated by GABAergic local circuit neurons before transmission to higher centers. This puts GABAergic neurons in a critical position in the dorsal horn, and any disruption of the GABAergic system could lead to widespread changes in the processing of sensory input. GABA synaptology allows for disinhibition. The complex synaptic interactions occuring between GABAergic profiles (summarized in Fig. 13)allows for the phenomenon of disinhibition described by Roberts ('76, '84). Numerous examples of GABA-IR dendrites and GABA-IR terminals synapsing on GABA-IR dendrites and somata confirm the fact that inhibition of inhibitory neurons or "disinhibition" could be a prevalent mechanism in laminae 1-111 of the spinal cord. The demonstration in this study of primary afferent terminals presynaptic to GABA-IR profiles which are in turn presynaptic to other GABA-IR profiles strongly suggests that the excitatory input relayed through primary afferents can result in disinhibition via GABAergic pathways. Inhibition and disinhibition, as well as direct excitation, are considered important synaptic mechanisms in higher centers in the mammalian nervous system (Roberts, '76, '84). This appears also to be true for the spinal cord.

GABA and primary Serent depolarization (PAD) The phenomenon of PAD is attributed to axo-axonic synapses between GABA containing axons and primary afferent terminals (see Schmidt, '71; Levy, '77 for review; Eccles et al., '62a,b, '63a-c). The tremendous volume of data documenting the involvement of GABA in PAD led us to believe that numerous GABA-IR terminals would be observed presynaptic to central terminals. In fact, our analysis showed that 1) presumed GABA terminals are postsynaptic to central terminals; 2) vesicle-containing GABA-IR dendrites, are the most common GABA-IR element presynaptic to central terminals; 3)when this arrangement is present, it is frequently in the form of reciprocal synapses; and 4) the frequency of GABA-IR vesiclecontaining dendrites presynaptic to central terminals in laminae 11-111 is surprisingly low, given the fact that

180

this area is considered a major site of PAD in cutaneous afferents (Howland et al., '55; Wall, '62, '64). It must be emphasized that the central terminal in a glomerulus is not the only morphology for a primary afferent terminal as other types have been described in the primate following rhizotomy studies (Ralston and Ralston, '79), tooth pulp extirpation (Westrum and Canfield, '77) or labeling with CGRP (Carlton et al., '87, '88, '90). It may be that nonglomerular terminals, which could not be identified in our study as arising from primary afferents, are postsynaptic to GABAergic terminals. In light of our findings, however, several hypotheses might be made concerning the involvement of GABA in the production of PAD. First, PAD may be generated through GABA-IR vesicle-containing dendritzc interactions with primary afferents as well as axo-axonic interactions. The finding that GABAergic dendrites may also be responsible for PAD has minimal impact on the theory of the mechanism of PAD, however, it has a major impact on the neurocircuitry thought to be involved in the production of PAD. As summarized in Figure 13, incoming primary afferent input to one GABAergic dendrite could result in the inhibition of several primary afferent terminals through dendritic branches of the same GABA neuron. The simplest spinal pathway for PAD production, based on a central latency of at least 2.0 msec, has been postulated to include two synapses (Eccles et al., '62a; Andersen et al., '64). The neurocircuitry hypothesized in Figure 13 would satisfy this requirement. The fact that GABA-IR vesicle-containing dendrites participate in reciprocal synapses with primary afferents may mean that excitatory input carried by the primary afferent is responsible for a local potential change which depolarizes the GABAergic dendrite and then the same GABAergic dendrite inhibits or decreases further action by that particular primary afferent. Thus, spatially and temporally, a very short loop can exist between the primary afferent and the GABA-IR profile that will ultimately inhibit it. The fact that we observed GABA-IR vesicle-containing dendrites participating in reciprocal synapses with LDCV terminals in laminae I and I1 and RSV glomerular type terminals in lamina 111, is consistent with reports that these fiber types are subject to primary afferent depolarization (Eccles, '64; Calvillo, '78; Hentall and Fields, '79; Curtis and Lodge, '82; Calvillo et al., '82; Gmelin and Zimmermann, '83; Dickenson et al., '85; Carstens et al., '87). Reciprocal synapses. Rall et al. ('66) were the first to describe reciprocal synapses, observing them in the olfactory bulb, where mitral and granule cells demonstrate adjacent synaptic junctions with polarities in opposite directions. The same morphological arrangement is seen in the retina between amacrine cells and the central processes of bipolar cells (Dowling and Boycott, '66). Other locations of reciprocal synapses include the ventrobasal thalamus (Harding, '711, the lateral geniculate body (Famiglietti, '70; LeVay, '71; Lieberman and Webster, '72), the dorsal column nuclei (Ellis and Rustioni, '811, and the spinal cord dorsal horn (Gobel et al., '80). GABA appears to be a consistent component of reciprocal synapses, based on its proposed presence in olfactory and retinal reciprocal synapses. Thus, it is not surprising that a similar type of complex found in the spinal cord also involves GABA. However, in the present study, the profiles involved included a GABA-IR dendrite and a large glomeru-

S.M. CARLTON AND E.S. HAYES lar type terminal. As mentioned above, Gobel et al. ('80) demonstrate a reciprocal synapse between an islet cell dendrite and a primary afferent terminal. If we assume that we are analyzing the same system (i.e., some GABA cells are islet cells), we confirm Gobels' observations. Knyihar and Gerebtzoff ('73) describe what they believed to be axoaxonic synapses in the superficial dorsal horn onto intervaricose segments of primary afferent axons. In contrast, in Figure 10, we demonstrate a dendro-axonic synapse on an intervaricose segment. A reasonable hypothesis for the functional significance of this arrangement might be that the activation of a primary afferent terminal will lead to the immediate feedback inhibition of that afferent via the GABAergic interneuron (see Jahr and Nicoll, '80). As suggested by Shepherd and Greer ('86), the patterns of these synaptic connections provide the basis of microcircuits that carry out specific functional integration of incoming sensory information. Based on our data, as well as that of others, GABAergic dendrites appear to be an important element through which control of the excitability of central terminals is mediated. GABA is a candidate neurotransmitter in islet cells. It is hypothesized by Gobel et al. ('80) that lamina I1 islet cells function as inhibitory interneurons that contain GABA. The ultrastructural characteristics of GABA-IR cells in the present study strongly suggest that GABA may be contained in islet cells. Similarities include the fact that islet cells and some GABA-IR cells are found in the same region (lamina 11),both cell types have numerous vesiclecontaining dendrites, both demonstrate spines with vesicles (type I1 spine heads), both make dendro-dendritic and dendro-axonic connections, both can be found as postsynaptic elements at asymmetrical synapses with primary afferents and as presynaptic elements at symmetrical synapses with primary afferent terminals (i.e., both are involved in reciprocal synapses). Because of these similarities, we hypothesize that many islet cells contain GABA. Todd and McKenzie ('89) have confirmed this hypothesis in their study in the rat in which Golgi-stained islet cells in lamina I1 were subsequently immunostained for GABA. Processing of sensory input. Elucidating the synaptic interactions of GABAergic processes may predict the functional roles of this inhibitory transmitter. It becomes clear in this study that the processing of primary afferent input is intimately related to the structural and functional integrity of the GABA system. Disruptions of this system will undoubtedly result not only in widespread effects on processing of incoming sensory information, but also affect the ongoing activity of spinal cord projection neurons. Just as loss of inhibitory GABAergic terminals in cortical regions may be responsible for seizures (Roberts, '80, '811, loss of the same types of synapses in the spinal cord dorsal horn may result in transmission of aberrant sensory signals. The diversity and complexity of the spinal GABAergic system described here suggests a variety of ways in which disturbances in this system could result in pathological states.

ACKNOWIJ3DGMENTS The authors would like to thank Greg Hargett and Cindy Do-Shope for their excellent technical assistance and Deatra Clay for secretarial help. This work was supported by NS11255 and NS27910.

GABA-IMMUNOREACTIVITY IN THE MONKEY SPINAL CORD

Andersen, P., J.C. Eccles, R.F. Schmidt, and T. Yokota (1964) Slow potential waves produced in the cuneate nucleus by cutaneous volleys and by cortical stimulation. J. Neurophysiol. 27:7%9 1. Barber, R.P., J.E. Vaughn, and E. Roberts (1982) The cytoarchitecture of gabaergic neurons in rat spinal cord. Brain Res. 238r305-328. Barber, R.P., J.E. Vaughn, K. Saito, B.J. McLaughlin, and E. Roberts (1978) GABAergic terminals are presynaptic to primary afferent terminals in the substantia gelatinosa of the rat spinal cord. Brain Res. I41r35-55. Barker, J.L., and R.A. NicoH (1972) Gamma-aminobutyric acid: Role in primary afferent depolarization. Science 176t1043-1045. Basbaum, A.I., E.J. Glazer, and W. Oertel(1986) Immunoreactive glutamic acid decarboxylase in the trigeminal nucleus caudalis of the cat: A lightand electron-microscopic analysis. Somatosens. Res. 4:77-94. Broman, J., and A. Blomqvist (1989) GABA-Immunoreactive neurons and terminals in the lateral cervical nucleus of the cynomologus monkey. J. Comp. Neurol. 283t415-424. Broman, J., and J. Westman (1988) GABA-immunoreactive neurons and terminals in the lateral cervical nucleus of the cat. J. Comp. Neurol. 274r467482. Burnstock, G. (1979) Autonomic neuroeffector junction. Neurosciences Res. Prog. Bull. 17:388-391. Calvillo, 0. (1978) Primary afferent depolarization of C fibers in the spinal cord of the cat. Can. J. Physiol. Pharmacol. 56t154-157. Calvillo, O., J. Madrid, and P. Rudomin (1982) Presynaptic depolarization of unmyelinated primary afferent fibers in the spinal cord of the cat. Neuroscience 7r1389-1400. Carlton, S.M. (1988) A light and electron microscopic analysis of GABA immunostaining in monkey spinal cord. Soc. Neurosci. Abstr. 18t983. Carlton, S.M., D.L. McNeill, K. Chung, and R.E. Coggeshall (1987) A light and electron microscopic level analysis of calcitonin gene-related peptide (CGRP) in the spinal cord of the primate: An immunohistochemical study. Neurosci. Lett. 82.145-150. Carlton, S.M., D.L. McNeill, K. Chung, and R.E. Coggeshall(1988) Organization of calcitonin gene-related peptide-immunoreactive terminals in the primate dorsal horn. J. Comp. Neurol. 276527-536. Carlton, S.M., K.N. Westlund, D. Zhang, L.S. Sorkin, and W.D. Willis (1990) Calcitonin gene-related peptide containing primary afferent fibers synapse on primate spinothalamic tract cells. Neurosci. Lett. 10936-81. Carstens, E., H. Gilly, H. Schreiber, and M. Zimmermann (1987) Effects of midbrain stimulation and iontophoretic application of serotonin, noradrenaline, morphine and GABA on electrical thresholds of afferent Cand A-fibre terminals in cat spinal cord. Neuroscience 2Ir395-406. Chung, K., and R.E. Coggeshall(1982) Quantitation of propriospinal fibers in the tract of lissauer ofthe rat. J. Comp. Neurol. 211t418-426. Chung, K., and R.E. Coggeshall (1983) Propriospinal fibers in the rat. J. Comp. Neurol. 217:47-53. Coggeshall, R.E., K. Chung, J.M. Chung, and L.A. Langford (1981) Primary afferent axons in the tract of lissauer in the monkey. J. Comp. Neurol. 196r431-442. Curtis, D.R., and D. Lodge (1982) The depolarization of feline ventral horn group Ia spinal afferent terminations by GABA. Exp. Brain Res. 46.215233. Davidoff, R.A. (1972) Gamma-aminobutyric acid antagonism and presynap tic inhibition in the frog spinal cord. Science 175:331-333. deBiasi, S., C. Frassoni, and R. Spreafico (1986) GABA immunoreactivity in the thalamic reticular nucleus of the rat. A light and electron microscopical study. Brain Res. 399t143-147. Dickenson, A.H., C.M. Brewer, and N.A. Hayes (1985) Effects of topical baclofen on C fibre-evoked neuronal activity in the rat dorsal horn. Neuroscience 14r557-562. Dowling, J.E., and B.B. Boycott (1966) Organization of the primate retina: Lond. [Biol.] 166r80-111. Electron microscopy. Proc. R. SOC. Duncan, D., and R. Morales (1978) Relative numbers of several types of synaptic connections in the substantia gelatinosa of the cat spinal cord. J. Comp. Neurol. 182r601-610. Earle, K.M. (1952) The tract of lissauer and its possible relation to the pain pathway. J. Comp. Neurol. 96t93-111. Eccles, J.C. (1964) The Physiology of Synapses. New York: Academic Press. Eccles, J.C., P.G. Kostyuk, and R.F. Schmidt (1962a) Central pathways responsible for depolarization of primary afferent fibres. J. Physiol. (Lond.) 161237-257.

181

Eccles, J.C., F. Magni, and W.D. Willis (1962b) Depolarization of central terminals of group I afferent fibres from muscle. J. Physiol. (Lond.) 16Or62-93. Eccles, J.C., R.F. Schmidt, and W.D. Willis (1963a) Pharmacological studies on presynaptic inhibition. J. Physiol. (Land.) 168r500-530. Eccles, J.C., R.F. Schmidt, and W.D. Willis (1963b) The location and mode of action of the presynaptic inhibitory pathways on to group I afferent fibers from muscle. J. Neurophysiol. 26r506-522. Eccles, J.C., R.F. Schmidt, and W.D. Willis ( 1 9 6 3 ~ Depolarization ) of the central terminals of cutaneous afferent fibers. J. Neurophysiol. 26r646661. Egger, M.D., N.C.G. Freeman, S. Malamed, P. Masarachia, and E. Proshansky (1981) Electron microscopic observations of terminals of functionally identified afferent fibres in cat spinal cord. Brain Res. 207.157-162. Ellis, L.C., Jr., and A. Rustioni (1981) A correlative HRP, golgi, and EM study of the intrinsic organization of the feline dorsal column nuclei. J. Comp. Neurol. 197:341-367. Famiglietti, E.V. (1970) Dendro-dendritic synapses in the lateral geniculate nucleus of the cat. Brain Res. 2Ot181-191. Fuji, K., E. Senba, S. Fujii, I. Nomura, J.-Y. Wu, Y. Ueda, and M. Tohyama (1985) Distribution, ontogeny and projections of cholecystokinin-8, vasoactive intestinal polypeptide and y aminobutyrate-containing neuron systems in the rat spinal cord: An immunohistochemical analysis. Neuroscience 14t881-894. Gmelin, G., and M. Zimmermann (1983) Effects of y-aminobutyrate and bicuculline on primary afferent depolarization of cutaneous fibres in the cat spinal cord. Neuroscience 10:869-874. Gobel, S. (1976) Dendroaxonic synapses in the substantia gelatinosa glomeruli of the spinal trigeminal nucleus of the cat. J. Comp. Neurol. 167t165-176. Gobel, S., W.M. Falls, G.J. Bennett, M. Abdelmoumene, H. Hayashi, and E. Humphrey (1980) An EM analysis of the synaptic connections of horseradish peroxidase-filled stalked cells and islet cells in the substantia gelatinosa of adult cat spinal cord. J. Comp. Neurol. 194t781-807. Harding, B.N. (1971) Dendro-dendritic synapses including reciprocal synapses in the ventro-lateral nucleus of the monkey thalamus. Brain Res. 34t181-185. Hayes, E., and S.M. Carlton (1989) CGRP primary afferents are presynaptic to GABA terminals in primate dorsal horn: A post-embedding immunogold study. SOC. Neurosci. Abstr. 15:740. Hentall, I.D., and H.L. Fields (1979) Segmental and descending influences on intraspinal thresholds of single C-fibers. J. Neurophysiol. 42r15271537. Hepler, J.R., C.S. Toomim, K.D. McCarthy, F. Conti, G. Battaglia, A. Rustioni, and P. Petrusz (1988) Characterization of antisera to glutamate and aspartate. J. Histochem. Cytochem. 36.13-22. Howland, B., J.Y. Lettvin, W.S. McCulloch, W. Pitts, and P.D. Wall (1955) Reflex inhibition by dorsal root interaction. J. Neurophysiol. 18:l-17. Hunt, S.P., J.S. Kelly, P.C. Emson, J.R. Kimmel, R.J. Miller, and J.Y. Wu (1981) An immunohistochemical study of neuronal populations containing neuropeptides or y-aminobutyrate within the superficial layers of the rat dorsal horn. Neuroscience 6r1883-1898. Jahr, C.E., and R.A. Nicoll (1980) Dendrodendritic inhibition: demonstration with intracellular recording. Science 207:1473-1475. Jonakait, G.M., A.R. Gintzler, and M.D. Gershon (1979) Isolation of axonal varicosities (autonomic synaptosomes) from the enteric nervous system. J. Neurochem. 32t1387-1400. Kerr, F.W.L. (1975) Neuroanatomical substrates of nociception in the spinal cord. Pain 1t325-356. Knyihar, E., and M.A. Gerebtzoff (1973) Extralysosomal localization of acid phosphatase in the spinal cord of the rat. Exp. Brain Res. I8r383-395. Knyihar, E., I. Laszlo, and S. Tornyos (1974) Fine structure and fluoride resistant acid phosphatase activity of electron dense sinusoid terminals in the substantia gelatinosa Rolandi of the rat after dorsal root transection. Exp. Brain Res. 19t529-544. Knyihar-Csillik. E., B. Csillik, and P. Rakic (1982a) Ultrastructure of normal and degenerating glomerular terminals of dorsal root axons in the substantia gelatinosa of the rhesus monkey. J. Comp. Neurol. 210,357375. Knyihar-Csillik, E., B. Csillik, and P. Rakic (1982b) Periterminal synaptology of dorsal root glomerular terminals in the substantia gelatinosa of the spinal cord in the rhesus monkey. J. Comp. Neurol. 2101376-399. LaMotte, C. (1977) Distribution of the tract of lissauer and the dorsal root fibers in the primate spinal cord. J. Comp. Neurol. 172529-562.

182 Larsson, L.-I. (1981) A novel immunocytochemical model system for specificity and sensitivity screening of antisera against multiple antigens. J. Histochem. Cytochem. 29r408-410. LeVay, S. (1971) On the neurons and synapses of the lateral geniculate nucleus of the monkey and the effects of eye enucleation. Z. Zellforsch. Mikrosk. Anat. 113:396-419. Levy, R.A. (1977)The role of GABA in primary afferent depolarization. Prog. Neurobiol. 9.211-267. Lieberman, A.R. (1973) Neurons with presynaptic p e r i k q a and presynaptic dendrites in the rat lateral geniculate nucleus. Brain Res. 59.35-59. Lieberman, A.R., and K.E. Webster (1972) Presynaptic dendrites and a distinctive class of synaptic vesicle in the rat dorsal lateral geniculate body. Brain Res. 42:196-200. Light, A.R., and E.R. Per1 (1977) Differential termination of large-diameter and small diameter primary afferent fibers in the spinal dorsal gray matter as indicated by labeling with horseradish peroxidase. Neurosci. Lett. 659-63. Light, A.R., and E.R. Per1 (1979) Spinal termination of functionally identified primary afferent neurons with slowly conducting myelinated fibers. J. Comp. Neurol. 186:133-150. Lund, R.D. (1969) Synaptic patterns of the superficial layers of the superior colliculus of the rat. J. Comp. Neurol. 135:179-208. Magoul, R., B. Onteniente, M. Gerrard, and A. Calas (1987) Anatomical distribution and ultrastructural organization of the gabaergic system in the rat spinal cord. An immunocytochemical study using anti-GABA antibodies. Neuroscience 20: 1001-1009. Matthews, M.A., G.K. McDonald, and T.V. Hernandez (1988) GABAdistribution in a pain-modulating zone of trigeminal subnucleus interpolaris. Somatosens. Res. 5.205-217. Maxwell, D.J., and R. Noble (1987) Relationship between hair-follicle afferent terminations and glutamic acid decarboxylase-containing boutons in the cat's spinal cord. Brain Res. 408t308-312. McLaughlin, B.J., R. Barber, K. Saito, E. Roberts, and J.Y. Wu (1975) Immunocytochemical localization of glutamate decarboxylase in rat spinal cord. J. Comp. Neurol. 164t305-322. McNeill, D.L., R.E. Coggeshall, and S.M. Carlton (1988) A light and electron microscopic study of calcitonin gene-related peptide in the spinal cord of the rat. Exp. Neurol. 99:699-708. Meinecke, D.L., and A. Peters (1987) GABA immunoreactive neurons in rat visual cortex. J. Comp. Neurol. 261:388-404. Melzack, R., and P.D. Wall (1965) Pain mechanisms: A new theory. Science I50:97 1-979. Millhorn, D.E., T. Hokfelt, K. Seroogy, W. Oertel, A.A.J. Verhofstad, and J.Y. Wu (1987) Immunohistochemical evidence for colocalization of y-aminobutyric acid and serotonin in neurons of the ventral medulla oblongata projecting to the spinal cord. Brain Res. 410:179-185. Miyata, Y., and M. Otsuka (1972) Distribution of y-aminobutyric acid in cat spinal cord and the alteration produced by local ischaemia. J. Neurochem. 29: 1833-1834. Miyata, Y., and M. Otsuka (1975) Quantitative histochemistry of yaminobutyric acid in cat spinal cord with special reference to presynaptic inhibition. J. Neurochem. 25.239-244. Otsuka, M., and S. Konishi (1976) GABA in the spinal cord. In E. Roberts, J.N. Chase, and D. Towers (eds): GABA in Nervous System Function. New York: Raven Press, pp, 197-202. Peters, A,, S.L. Palay, and H.D. Webster (1976) The Fine Structure of the Nervous System: The Neurons and Supporting Cells. Philadelphia: W.B. Saunders Company. Rall, W. (1962) Electrophysiology of a dendritic neuron model. Biophys. J. 2:145-167. Rall, W. (1964) Theoretical significance of dendritic trees for neuronal input-output relations. In R.F. Reiss (ed): Neural Theory and Modeling. Stanford, CA Stanford University Press, pp. 73-97. Rall, W., G.M. Shepherd, T.S. Reese, and W.M. Brightman (1966) Dendrodendritic synaptic pathway for inhibition in the olfactory bulb. Exp. Neurol. 14:44-56. Ralston, H.J. (1968) The fine structure of neurons in the dorsal horn of the cat spinal cord. J. Comp. Neurol. 132:275-302. Ralston, H.J. (1071) Evidence for presynaptic dendrites and a proposal for their mechanism of action. Nature 230:585-587. Ralston, H.J. (1979) The fine structure of laminae I, I1 and 111 of the macaque spinal cord. J. Comp. Neurol. 184:619-642. Ralston, H.J., and D.D. Ralston (1979) The distribution of dorsal root axons in laminae I, I1 and 111 of the macaque spinal cord: A quantitative electron microscope study. J. Comp. Neurol. 184:643-684.

S.M. CARLTON AND E.S. HAYES Rexed, B. (1952) The cytoarchitectonic organization of the spinal cord in the cat. J. Comp. Neurol. 96:415-495. Ribak, C.E., J.E. Vaughn, K. Saito, R. Barber, and E. Roberts (1977) Glutamate decarboxylase localization in neurons of the olfactory bulb. Brain Res. 126:l-18. Ribeiro-Da-Silva, A,, and A. Coimbra (1980) Neuronal uptake of ['HIGABA and [3H]glycinein laminae 1-111 (substantia gelatinosa Rolandi) of the rat spinal cord. An autoradiographic study. Brain Res. 188:449-464. Ribeiro-Da-Silva, A., D. Pignatelli, and A. Coimbra (1985) Synaptic architecture of glomeruli in superficial dorsal horn of rat spinal cord, as shown in serial reconstructions. J. Neurocytol. 14203-220. Richardson, K.C. (1964) The fine structure of the albino rabbit iris with special reference to the identification of adrenergic and cholinergic nerves and nerve endings in its intrinsic muscles. Am. J. Anat. 114:173205. Roberts, E. (1976) Disinhibition as an organizing principle in the nervous system-the role of the GABA system. Application to neurolobic and psychiatric disorders. In E. Roberts, T.N. Chase, and D.B. Tower (eds): GABA in Nervous System Function. New York: Raven Press, pp. 515-539. Roberts, E. (1980) Epilepsy and antiepileptic drugs: a speculative synthesis. In G.H. Glaser, J.K. Penry, and D.M. Woodbury (eds): Antiepileptic Drugs: Mechanisms of Action. New York: Raven Press, pp. 667-7 13. Roberts, E. (1981) A speculative consideration on the neurobiology and treatment of senile dementia. In T. Crook and S. Gerson (eds): Strategies for the Development of an Effective Treatment for Senile Dementia. New Canaan, CT: Mark Powley Assoc., pp. 247-320. Roberts, E. (1984) GABA neurons in the mammalian central nervous system: model for a minimal basic neural unit. Neurosci. Lett. 47:195200. Roberts, P.J., and P. Keen (1974) Effect of dorsal root lesion of amino-acids of rat spinal cord. Brain Res. 74:333-337. Scheibel, M.E., and A.B. Scheibel (1968) Terminals axonal patterns in the cat spinal cord. 111. The dorsal horn. Brain Res. 9:32-58. Schmidt, R.F. (1971) Presynaptic inhibition in the vertebrate central nervous system. Ergeb. Physiol. 6320-101. Shepherd, G.M., and C.A. Greer (1988) The dendritic spine: adaptations of structure and function for different types of synaptic integration. In R.J. Lasek and M.M. Black (eds): Intrinsic Determinants of Neuronal Form and Function. New York: Alan R. Liss, Inc., pp. 245-262. Sternberger, L.A. (1979) Immunocytochemistry. 2nd ed., New York: John Wiley & Sons. Streit, P., E. Knecht, J.-C. Reubi, S.P. Hunt, and M. Cuenod (1978) GABA-specific presynaptic dendrites in pigeon optic tectum: A high resolution autoradiographic study. Brain Res. 149:204-210. Takahashi, T., and M. Otsuka (1975) Regional distribution of substance P in the spinal cord and nerve roots of the cat and the effect of dorsal root section. Brain Res. 87:l-11. Takasu, N., T. Nakatani, T. Arikuni, and H. Kimura 11987) Immunocytochemical localization of y-aminobutyric acid in the hypoglossal nucleus of the macaque monkey, Macaca fuscata: A light and electron microscopic study. J. Comp. Neurol. 263:42-53. Todd, A.J., and J. McKenzie (1989) GABA-immunoreactive neurons in the dorsal horn of the rat spinal cord. Neuroscience 31:799-806. Uchizono, K. (1965) Characteristics of excitatory and inhibitory synapses in the central nervous system of the cat. Nature 207:642,643. Wall, P.D. (1962) The origin of a spinal cord slow potential. J. Physiol. (Lond.) 164:508-526. Wall, P.D. (1964) Presynaptic control of impulses a t the first central synapse in the cutaneous pathway. In J.C. Eccles and J.P. Schade (eds): Progress in Brain Research, Physiology of Spinal Neurons. Amsterdam: Elsevier, Vol. 12, pp. 92-118. Westman, J., A. Blomqvist, C. Kohler, and J.-Y. Wu (1984) Light and electron microscopic localization of glutamic acid decarboxylase and substance P in the dorsal column nuclei of the cat. Neurosci. Lett. 51:347-352. Westrum, L.E., and R.C. Canfield 11977) Light and electron microscopy of degeneration in the brainstem spinal trigeminal nucleus following tooth pulp removal in adult cats. In D.J. Anderson and B. Matthews (eds): Pain in the Trigeminal Region. Amsterdam: Elsevier, pp. 171-179. Wood, J.G., B.J. McLaughlin, and J.E. Vaughn (1976) Immunocytochemical localization of GAD in electron microscopic preparations of rodent CNS. In E. Roberts, T.N. Chase, and D.B. Tower (eds): GABA in Nervous System Function. New York: Raven Press, pp. 133-148.