THE JOURNAL OF COMPARATIVE NEUROLOGY 303316-328 (1991)

GmAergic Circuitryin the R~stralVentral Medulla of the Rat and Its Relationship to DescendingAntinociceptive Controls HEE JUNG CHO AND ALLAN I. BASBAUM Departments of Anatomy and Physiology, University of California, San Francisco, California 94143

ABSTRACT This study used postembedding immunocytochemistry to examine the organization of GABA-immunoreactive synapses in the rostral ventral medulla (RVM) of the rat. To determine whether the outflow neurons of the RVM are under GABAergic control, we examined the distribution of GABA-immunoreactive synapses upon bulbospinal projection neurons that were labelled by retrograde transport of wheatgerm agglutinin-HRP from the cervical spinal cord. To study the possible convergence of GABAergic and periaqueductal gray (PAG) synaptic inputs to RVM neurons, we also made lesions in the PAG and examined the relationship between degenerating PAG axons and GABA-immunoreactive terminals. Approximately 45%of all synapses in the RVM,which includes the midline nucleus raphe magnus and the nucleus reticularis paragigantocellularis lateralis, were GABA-immunoreactive. The vast majority of GABA-immunoreactive terminals contained round, clear, and pleomorphic vesicles and made symmetrical axodendritic synapses; axoaxonic synapses were not found. Almost 50% of the retrogradely labeled dendrites in the NRM were postsynaptic to GABA-immunoreactive terminals. Several examples of convergence of degenerating PAG terminals and GABAergic terminals onto the same unlabelled dendrite were also found. These data indicate that the projection neurons of the RVM are under profound GABAergicinhibitory control. The results are discussed with regard to the hypothesis that the analgesic action of narcotics and electrical stimulation of the midbrain PAG involves the regulation of tonic GABAergic inhibitory controls that are exerted upon spinally-projecting neurons of the nucleus raphe magnus. Key words: postembeddingimmunocytochemistry, pain control, nucleus raphe magnus, morphine

The rostral ventral medulla (RVM), including the 5HTrich midline nucleus raphe magnus (NRM), constitutes an important link in the circuits through which opiates and electrical stimulation of the midbrain periaqueductal gray (PAG) produce analgesia (Basbaum and Fields, '78, '84). Anatomical studies have established a direct connection from the PAG to spinally-projecting neurons of the NRM that are an integral part of this circuitry (Abols and Basbaum, '81; Fardin et al., '84; Gallagher and Pert, '78; Lakos and Basbaum, '88). The latter project to and inhibit the firing of nociresponsive neurons of the spinal dorsal horn (Basbaum and Fields, '79; Basbaum et al., '78, '86; Fields et al., '78; Willis et al., '77). Recent studies of these analgesia circuits have focussed on the excitatory and inhibitory controls that are exerted on RVM neurons. In addition to the well-described excitatory connection from the PAG (Aimone and Gebhart, '86; Pomeroy and Behbehani, '791, there is pharmacological and electrophysiological O

1991 WILEY-LISS, INC.

evidence for a tonic GABAergic inhibitory control of raphe neurons (Fields and Heinricher, '85). Drower and Hammond ('88) reported that injection of the GABA-A antagonist, bicuculline, into the NRM produced a mild increase in nociceptive threshold. They hypothesized that this hypoalgesia resulted from the loss of a tonic GABAergic inhibition of "antinociceptive" output neurons of the NRM. In electrophysiological studies, Fields and colleagues ('83a) demonstrated that "off' cells in the RVM, which are characterized by a pause in their spontaneous activity just prior to the occurrence of a noxious heat stimulus-evoked tail-flick reflex, are inhibited by iontophoreAccepted September 5,1990. Address reprint requests to Allan I. Basbaum, Department of Anatomy, University of California, San Francisco, CA 94143. H.J. Cho is now at the Department of Anatomy, Kyungpook National University, School of Medicine, 2-101, Dongin Dong, Taegu, Korea.

GABAERGIC CIRCUITRY IN THE ROSTRAL VENTRAL MEDULLA sis of GABA into the RVM (Heinricher et al., '87); iontophoresis of bicuculline blocked the "off' cell's pause. Since PAG injection of morphine increased the firing of "off' cells and blocked the pause that precedes the reflex, these authors proposed that PAG injection of morphine activates the NRM output neuron, via inhibition of the GABAergic inhibitory interneurons which tonically regulate the NRM projection neurons (Fields et al., '83b). In addition to the antinociceptive controls originating in the midline NRM, there is an important contribution of more lateral regions, in particular the nucleus reticularis paragigantocellularis lateralis (Rpgl). For example, Gray and Dostrovsky ('85)reported that electrical stimulation of the Rpgl inhibits the firing of spinal cord dorsal horn neurons and Sandkiihler and Gebhart ('84) demonstrated that the analgesic action of PAG stimulation is eliminated only when the outputs of both the midline raphe and the lateral medulla are blocked with local anesthetic. Other studies have implicated the lateral medulla in tonic descending inhibitory controls that are exerted on dorsal horn nociresponsive neurons and have established a contribution of GABAergic mechanisms in the Rpgl to the antinociceptive controls generated by electrical stimulation in the PAG (Foong and Duggan, '86; Lovick, '87). Since changes in cardiovascular tone are produced by noxious stimulation, it is of particular interest that the region from which Foong and Duggan ('86) characterized the GABAergic regulation overlaps with the area that has been implicated in bulbospinal cardiovascular control (Lovick, '85, '86; Siddal and Dampney, '89; Sun and Guyenet, '86). In fact, the Rpgl neurons that project to and excite cardioacceleratory preganglionic sympathetic neurons of the intermediolateral cell column of the thoracic cord appear to be regulated by GABAergic mechanisms in a manner very similar to that seen for neurons at the origin of descending antinociceptive controls (Keeler et al., '84; Lovick, '87; Ruggiero et al., '85; Willette et al., '83; Yamada et al., '82, '84). Despite these many pharmacological and physiological studies demonstrating similarities in the GABAergic control of ventral medullary nociceptive and cardiovascular control systems, electron microscopic studies of GABAergic circuitry in the RVM have been limited to the cardiovascular regulatory neurons in the lateral medulla (Milner et al., '87). In fact, there are only a few light microscopic studies of GABA neurons in the region of the NRM and Rpgl (Bowker et al., '88; Meeley et al., '85; Millhorn et al., '87, '88; Mugnaini and Oertel, '85; Reichling and Basbaum, '90). TO extend the ultrastructural analysis to the GABAergic circuitry that regulates antinociceptive control systems of the rostroventral medulla, we have used a postembedding immunocytochemical approach to analyze the synaptic organization of GABA-immunoreactive synapses in the medial and lateral regions of the RVM. To specifically address whether NRM projection neurons are directly influenced by GABA terminals, we examined the input to NRM neurons that were retrogradely labelled from the spinal cord. Finally, to study the possible convergence of GABAergic and PAG synaptic inputs to NRM neurons, we

Abbreviations

GABA NRM PAC WGA-HRP

y-aminobutyric acid nucleus raphe magnus periaqueductal gray wheat germ agglutinin coupled to horseradish peroxidase

317

made lesions in the PAG and examined the relationship between degenerating PAG axons and GAl3Aergic terminals. A preliminary report of this work has been published (Reichling et al., '88).

METHODS These studies were performed in 250-300 g male rats (Bantin and Kingman, Fremont, CA). Although some rats were untreated prior to perfusion fixation, most underwent one or both of the following procedures. To make spinal cord injections of the retrograde tracer, the rats were anesthetized with pentobarbital (50 mg/kg) and a laminectomy performed over the cervical enlargement. The dura was incised and four 0.5 Km injections of a 5% solution of wheatgerm agglutinin conjugated to horseradish peroxidase (WGA-HRP) were made bilaterally into the spinal cord. The overlying muscle and skin were sutured and the rats returned to their cage. Two of these rats also had a lesion made in the midbrain PAG. The rats were anesthetized and placed in a stereotaxic instrument, the skull was exposed, and a hole drilled over appropriate stereotaxic coordinates, taken from the atlas of Paxinos and Watson ('86), to target the ventrolateral part of the midbrain PAG. The lesion was made with a Radionics (Burlington, MA) thermal lesion maker. Two days after the lesion was made in the PAG, and up to 1 week after the spinal injection of WGA-HRP, rats were deeply anesthetized and perfused intracardially with the following solutions: 0.1 M phosphate-buffered saline (PBS) followed by a fixative solution containing 2% paraformaldehyde and 2% glutaraldehyde in a phosphate buffer. The brain and spinal cord were removed and placed in the same fixative for an additional 6 hours. Blocks containing the RVM were then dissected out, washed in PBS, and sectioned at 100 pm on a vibratome. The retrogradely transported WGA-HRP was first visualized with diaminobenzidine in the presence of H,O, and then sections were osmicated in 0.5% osmium tetroxide for 1hour, dehydrated in graded alcohols and propylene oxide, and then embedded in a relatively soft embedding matrix that was readily etched and thus compatible with postembeddingimmunocytochemistry. The embedding resin was made up of equal parts of (1) dodecenylsuccinic anhydride (Ted Pella Inc.; Reading, CAI-25 g and Poly/Bed 812 (Polysciences; Warrington, PA)-23.5 g; (2) nadic methyl anhydride (Ted Pella Inc.; Reading, CA)-24.1 g and Polyhed 812-37 g. To this mixture DMP-30 (Polysciences; Warrington, PA) was added. Two antisera were used in these studies, a rat anti-GABA antiserum that was kindly provided by Dr. Andrew Towle of Cornell University Medical School, and a rabbit anti-GABA antiserum purchased from IncStar (Minneapolis, MN). These antisera are directed against a glutaraldehyde conjugate of GABA and bovine serum albumin (BSA). The postembedding protocol (Basbaum, '89) is adapted from that described by De Biasi and Rustioni ('88) and results in central nervous system tissue quality superior to that found with hydrophilic resins. The sections were first etched with sodium metaperiodate for 30 to 45 minutes, rinsed in distilled water, and then incubated in sodium borohydride. After several washes the sections were incu-

H.J. CHO AND A.I. BASBAUM

318

Figure 1

GABAERGIC CIRCUITRY IN THE ROSTRAL VENTRAL MEDULLA

Fig. 1. These electron micrographs illustrate several examples of 10 nm immunogold-labelled GABA-immunoreactive axon terminals that make synaptic contact in the NRM. GABA-immunoreactive axon terminals ( G ) make symmetrical synaptic contacts (arrowheads) with

319

the cell body (cb in A),proximal dendrite (pd in B) and distal dendrites (d in C and D); the distal dendrites are commonly also postsynaptic to unlabelled axon terminals (a1 and a2 in C and a in D). Calibration bars = 0.5 km.

320

Fig. 2. A few of the GABA-immunoreactive axon terminals (G in A and G2 in B) make asymmetrical synaptic contacts (arrowheads) with dendrites (d) in the RVM.Another GABA-immunoreactive axon termi-

H.J. CHO AND A.I. BASBAUM

nal (G1 in B) makes symmetrical synaptic contact with the same dendrite. The dendrite in B is also postsynaptic to an unlabelled axon (a). Calibration bars = 1.0 pm.

GABAERGIC CIRCUITRY IN THE ROSTRAL VENTRAL MEDULLA

321

Fig. 3. Many GABA-immunoreactive axon terminals (G in A, G1, G2 and G3 in B) contain dense core vesicles (arrows). Some of these terminals make synaptic contacts (arrowheads) with unlabelled den-

drites (dl, d2 in B). The labelled terminal varicosity in A appears to be continuous with another labelled varicosity (arrow). Calibration bars = 0.5 km.

bated in normal goat serum in Tris-buffered saline followed by a 1:1,000 dilution of the primary antiserum for 1 hour. After multiple washes, the sections were incubated in a 1:25 dilution of 10 nm colloidal-goldlabelled goat antirabbit IgG or goat antirat IgG (Janssen Life Sciences/Ted Pella, Inc.; Reading, CA) for 1 hour. After further washes and then drying, the sections were grid stained with uranyl acetate and lead citrate. Control sections were stained with the primary antiserum preabsorbed for 24 hours with 10 K r n GABA-BSA conjugate.

To provide a quantitative estimate of the labelling in the different regions of the RVM, we counted 200 terminals in each of four different regions of the RVM. These were the midline NRM, the nucleus reticularis paragigantocellularis (Rpg), located just lateral to the NRM, the nucleus reticulark gigantocellularis, pars alpha (Rpgcx), located just ventral to the Rpg and the nucleus reticularis paragigantocellularis lateralis (Rpgl), located lateral to the Rpg. All terminals were counted, whether or not a synaptic contact was recognizable.

322

Fig. 4. The electron micrograph in A illustrates a GABA-immunoreactive axon terminal that is presynaptic to a cell body (Cb), that is retrogradely labelled with WGA-HRP. Arrows identify dense bodies that contain the DAB reaction product. The arrowheads point to the synaptic density between the GABA terminal and the cell body. Inset B

H.J. CHO AND A.I. BASBAUM

is a higher magnification of the GABA-immunoreactive terminal outlined in A. A symmetrical synaptic contact between the GABAimmunoreactive axon terminal (GI and the retrogradely labelled cell body can be clearly identified. Calibration bars = 1.0 pm in A and 0.5 pm in B.

GABAERGIC CIRCUITRY IN THE ROSTRAL VENTRAL MEDULLA

323

Fig. 5 . The electronmicrograph in A illustrates a dendrite (d) that is retrogradely labelled with WGA-HRP; arrow points to labelled dense body. This dendrite is postsynaptic to GABA-immunoreactive and unlabelled axon terminals. Arrowheads point to the synaptic densities.

The enlargement in B better illustrates the area outlined in A. Convergence of a GABA-immunoreactive axon terminal (G) and unlabelled axon terminal (a) onto retrogradely labelled dendrites was common. Calibration bars = 1.0 pm in A and 0.5 km in B.

RESULTS Identificationof GABA-jmmunoreactive

reactive peptides are evaluated with postembedding methods (De Biasi and Rustioni, ’88). Although the quantitative analysis was not performed stereologically, the results from counting synapses were remarkably uniform throughout the different regions of the RVM and established several important features of the pattern of GABA-immunoreactive terminal labelling in the RVM. First, and perhaps most importantly, we found that over 40% of all terminals in the RVM were GABAimmunoreactive. Second the levels were comparable in all regions; the percentages of GABA terminals were 48.0, 43.5, 45.5, and 46.5 in the NRM, Rpg, Rgca, and Rpgl, respectively. When the analysis was made from only those terminals that made synaptic contacts, the numbers were comparable; 48.8,42.2,45.3, and 44.9%, respectively. Axodendritic synapses were by far the most common GABA-immunoreactive synaptic interaction observed (Figs. 1B-D, 2,3B; see also Figs. 5,6). They comprised 91.5, 93.0, 94.3, and 93.1% of the total number of synapses in the NRM, Rpg, Rgccu, and Rpgl, respectively. We found no examples of GABA-immunoreactive axoaxonic synapses in any of the four regions examined. The remaining GABAimmunoreactive terminals contacted cell bodies (Figs. lA, 4). Convergence of labelled and unlabelled terminals onto dendrites was very common (Figs. 1C,D, 2B, 5 ) . Almost all of the GABAergic terminals formed symmetrical synapses; however, a few examples of asymmetrical junctions were observed in each of the four areas studied (Fig. 2); the percentages were 3.4, 2.3, 1.9, and 3.4 in the NRM, Rpg, Rgw, and Rpgl, respectively. Figure 2B illustrates an unlabelled dendrite that is postsynaptic to two GABA-

termjnals Colloidal gold labelled GABA-immunoreactive terminals were readily distinguished from background labelling. Counts of gold particles over synaptic profiles demonstrated that “labelled” terminals typically had from three to four times the number of gold particles compared to unlabelled profiles in the same grid square. When the primary antibody was absorbed with the GABA-BSA conjugate, the immunoreactivity was abolished and only a few scattered gold particles were found. In the present study we found that most of the labelling was located over synaptic profiles; a few myelinated and unmyelinated axons were also labeled with equivalent density. In contrast, we could not detect dendritic or somatic labelling above background. This was true even when colchicine was administered intracisternally. Despite the fact that there was very dense labelling of large numbers of terminals, gold particles did not exclusively overly vesicles. Several terminals also had very concentrated gold label over mitochondria (Figs. lA,C,D, 2B). In fact, a small percentage of labelled terminals were unusual in containing very large numbers of mitochondria. GABA-immunoreactive terminals contained clear round and pleomorphic vesicles (Figs. 1, 2B, 3B); only rarely did we observe terminals that predominantly contained flat vesicles (Fig. 2A). Although a few labelled terminals also contained dense cored vesicles (Fig. 3), the gold labelling was not located over the latter, as is typical when immuno-

Fig. 6. This electron micrograph montage illustrates a dendrite in the NRM that is both postsynaptic to a degenerating axon terminal from the PAG (rectangle A) and to several GABA-immunoreactive axon terminals (arrowheads). Several unlabelled axon terminals are also presynaptic to this dendrite (arrows). Calibration bars equal 2.0 pm. See Figure 7A,B for higher magnification of the degenerating PAG terminal (rectangleA) and the GABA-immunoreactive terminal (rectangle B).

+

Figure 7 (see overleaf for legend)

H.J. CHO AND A.I. BASBAUM

326

likely that most, if not all, of the GABAergic terminals that we observed derive from cell bodies located in the RVM. Light microscopic studies have, in fact, demonstrated large numbers of GABA immunoreactive, or glutamic acid decarboxylase positive neurons, in the medial and lateral RVM (Bowker et al., '88; Meeley et al., '85; Millhorn et al., '87, '88; Mugnaini and Oertel, '85; Reichling and Basbaum, '90). The objective of the present study was to evaluate the anatomical basis for the hypothesis that the analgesia produced by opiates or electrical stimulation of the midbrain PAG results from blockade of a tonic GABAergic inhibitory control that is exerted upon raphe-spinal neuGABA inputs to spinal projection neurons rons (Fields and Heinricher, '85). The simplest circuitry When diaminobenzidine is used to detect cells retro- that could account for the proposed disinhibitory mechagradely labelled with WGA-HRP, one is usually restricted to nism would include the following: (1)GABA-immunoreacthe analysis of soma and proximal dendrites of retrogradely tive terminals in the NRM should be found presynaptic to labelled neurons. In spite of this, we found numerous examples of retrogradely labelled cells that were postsynap- raphe-spinal neurons, (2) the morphology of the GABAergic tic to GABA-immunoreactive terminals (Figs. 4, 5). Most of terminals should be of the inhibitory type, i.e., symmetrical, the GABA-immunoreactive inputs contacted retrogradely and (3) PAG terminals should lie presynaptic to, and labelled dendrites. In fact, almost 50% of the dendrites in inhibit, GABAergic neurons in the NRM. Consistent with the proposal that the raphe-spinalprojecthe NRM that unequivocally contained retrogradely labelled dense bodies, were postsynaptic to GABA-immunore- tion neurons are under GABAergic control, we found that active terminals. When the analysis was repeated in the two many projection neurons of the NRM receive a somatic or animals that also received a PAG lesion, we found several dendritic GABAergic input. In fact, almost 50% of the examples of convergence of degenerating terminals and retrogradely labelled cells in the NRM were postsynaptic to GABAergic terminals onto the same, unlabelled dendrite GABAergic terminals. Since retrograde label was rarely (Figs. 6, 7). In general, these inputs were widely dispersed, found in distal dendrites, this number probably underestiwhich made it difficult to demonstrate in a single photo- mates the number of projection neurons that receive graph. Thus we have used a very low magnification image of GABAergic control. Our study did not address whether the a large dendrite in the NRM (Fig. 6), with magnified insets 5HT neurons of the NRM, which have been specifically (Fig. 7A,B), to illustrate examples of convergence onto the implicated in control of spinal nociresponsive neurons by same cell. Finally, although we detected some degenerating PAG stimulation (Basbaum and Fields, '78, '84; Roberts, terminals on retrogradely labelled cell bodies, we did not '841, receive a GABAergic input. The fact that a high find convergence of GABA and PAG terminals onto projec- percentage of the spinally-projecting NRM cells can be tion neurons. double-labelled with 5HT antisera (Bowker et al., '881, however, is compatible with the suggestion that a significant proportion of the NRM projection neurons that receive DISCUSSION GABA-immunoreactive synaptic contacts contain 5HT. The most striking observation in the present study is the The morphology of the GABAergic terminal is also very large and relatively constant percentage of GABAconsistent with the suggestion that there is a tonic inhibiimmunoreactive synapses in the medial and lateral subregions of the rostra1 ventral medulla (approximately 45%). tory GABAergic control of the raphe-spinal neuron. Thus Since our quantitative analysis was not stereological and almost all of the GABA-immunoreactivc terminals formed did not take into account differences in synapse size, we symmetrical synaptic contacts, with either unlabelled cell cannot comment on the relative volume density of GABAer- bodies or dendrites. Since we found no evidence for GABAgic and non-GABAergic synapses. Nevertheless, the rather immunoreactive axoaxonic synapses, it is likely that the uniform and high percentage indicates that the outflow inhibitory GABAergic effects in the RVM are generated neurons of the RVM, including those contributing to anti- exclusively via postsynaptic mechanisms. Interestingly, benociceptive as well as cardiovascular regulatory mecha- tween 30 and 40%of the GABA-immunoreactive terminals nisms, are subject to profound inhibitory control. Although in the RVM contained dense-cored vesicles. This raises the we did not label GABAergic cell bodies or dendrites in this possibility that some of the GABAergic neurons in the RVM study, probably due to the protocol used for immunocy- co-store and release a peptide neurotransmitter, an observatochemistry (see De Biasi et al., '88 for Discussion), it is tion consistent with the presence of many peptide contain-

immunoreactive terminals; one of the contacts was symmetrical, the other asymmetrical. All GABA-immunoreactive terminals that made symmetrical contacts contained round or slightly oval, clear vesicles; somewhat surprisingly, the only flattened vesicles were found in those few labelled terminals that made asymmetrical contacts (Fig. 2A). In a separate analysis we specifically evaluated whether any of the labelled terminals contained dense cored vesicles (Fig. 3). Fifty terminals in each of the four areas were counted; at least one dense core vesicle was detected in 38, 34, 32, and 40% of the terminals in the NRM, Rpg, Rgca and Rpgl, respectively.

~

Fig. 7. A. Higher magnification electronmicrograph of rectangle A from Figure 7. This degenerating axon terminal (asterisk) from PAG makes an asymmetrical synaptic contact with the postsynaptic dendrite (d).An unlabelled axon terminal (a) is also presynaptic to this dendrite. B. Higher magnification electronmicrograph of rectangle B outlined from Figure 7 illustrates that the GABA-immunoreactive axon terminal (GI makes a symmetrical synaptic contact with the same dendrite (d). C.This electronmicrograph illustrates the convergence of a GABA-

immunoreactive axon terminal (GI, a degenerating axon terminal (asterisk) from PAG and an unlabelled axon terminal (a) onto the same cell body (Cb). D. This inset is a higher magnification electronmicrograph of a serial section through the degenerating axon terminal described in C and demonstrates that there is a synaptic contact (arrowheads) between this degenerating axon terminal and the cell body. Calibration bars = 1.0 pm in A, B, and C, and 0.5 km in D.

GABAERGIC CIRCUITRY IN THE ROSTRAL VENTRAL MEDULLA ing neurons in the RVM (Bowker et al., '88; Johansson et al., '81; Men6trey and Basbaum, '87). Since we were unable to immunostain GABA cell bodies or dendrites and since we never found GABA-immunoreactive terminals postsynaptic in an axoaxonic synaptic relationship, we have no information on the synaptic input of PAG neurons to the GABAergic neurons in the RVM. We, however, confirmed our previous observation that some PAC: axons terminate within and make synaptic contact with spinally-projecting neurons of the NRM (Lakos and Basbaum, '88). Importantly, all degenerating terminals from the PAG contained round vesicles and appeared to make asymmetrical synaptic contacts (the precise morphology of the contact was difficult to determine in the degenerating terminal). This synaptic morphology suggests that the PAG input to the R V M is excitatory, i.e., it is unlikely that the PAG disinhibits the NRM projection neuron, at least not via a direct inhibitory input to local GABAergic neurons in the RVM. Rather, in addition to exciting the RVM projection neuron, the PAG may evoke a feedforward inhibitory control of the output cell, via excitation of GABAergic inhibitory interneurons. Thus although our data neither confirm nor disprove the hypothesis concerning the disinhibitory regulation of the NRM output neuron, they indicate that the analgesic action of electrical brain stimulation of the PAG or of microinjection of morphine into the midbrain PAG results, at least in part, from direct excitation of the NRM output neuron. Finally, although we found convergence of GABA and PAG inputs to unlabelled dendrites in the RVM, we could not establish that GABAergic and PAG terminals converge onto the same retrogradely labelled cells. The numbers of degenerating PAG axon terminals, however, is very small, and thus the likelihood of finding convergence onto retrogradely labelled cells is even smaller, particularly if the convergence is predominantly on distal dendrites (which would not contain retrograde label). To get around this problem, we are examining the GABAergic inputs to physiologically identified cells that have been intracellularly labelled with HRP (Skinner et al., '90).This approach not only permits one to examine the inputs to distal dendrites, but also to establish, electrophysiologically, that the neurons receive a PAG input and project to the spinal cord. Our results on the GABAergic circuitry in the RVM are similar to those of others who addressed the GABAergic synaptic relationships in the cardiovascular control areas of the lateral medulla. Thus Milner et al. ('87) reported that the vast majority of GABAergic terminals in the lateral RVM, identified with antisera directed against glutamic acid decarboxylase (GAD), made symmetrical axodendritic synapses. In contrast to the present report, they found very few examples in which GAD positive terminals contained dense core vesicles. Furthermore, they found a few examples in which presumptive GABA terminals were located presynaptically, in an axoaxonic relationship. The fact that we did not find axoaxonic synapses may reflect the fact that the Milner et al. ('87) study was concentrated on the GABAergic synapses upon the PNMT positive neurons of the C1 epinephrine-containing cell group, which is located somewhat caudal to the focus of our study. Subtle differences in the synaptic organization in different regions of the medulla may reflect differences in the GABAergic interactions with functionally distinct populations of cells. In fact, Lovick ('87) demonstrated that nociceptive and cardiovascular control neurons of the Rpgl

327

could be distinguished. She reported that although bilateral microinjection of bicuculline into the Rpgl increased the latency of both the heat-evoked tail-flick reflex and increased heart rate and blood pressure, the time course of the analgesia and pressor response was different. Furthermore, injection of physostigmine into the Rpgl evoked a pressor response but did not produce analgesia. These data indicate the importance of studying the inputs to physiologically identified cells in order to establish the anatomical basis for the GABAergic regulation of nociceptive and cardiovascular control neurons of the ventral medulla. In conclusion, using postembedding immunocytochemistry, for GABA, in the rostral ventral medulla, we found that GABA-immunoreactive synapses constitute close to half of the total synaptic population. Our results indicate that the GABAergic inhibitory control is mediated via postsynaptic regulation of RVM neurons. In double-labelling studies, we established that at least 50% of the NRM projection neurons receive a GABAergic input. We also confirmed our observation that some of the NRM projection cells receive an input from the midbrain PAG. Finally, we demonstrated that undefined dendrites of the NRM receive a convergent input, both from GABAergicterminals and from degenerating axon terminals of the PAG. These results indicate that spinal projection neurons of the rostral medulla are under GABAergic control, an observation consistent with the growing body of physiological and pharmacological evidence implicating a tonic GABAergicinhibition in the antinociceptive controls exerted by opiates and midbrain electrical brain stimulation. The anatomical substrate for the hypothesized disinhibition of the NRM output neuron, however, remains to be determined.

ACKNOWLEDGMENTS We thank Dr. Jon Levine for many helpful discussions during the preparation of the manuscript. We also thank Ms. Bonnie Lord and Ms. Simona Ikeda for excellent help with the photographic plates. This work was supported by grants from NIH: NS14627 and 21445.

Abols, LA., and A.I. Basbaum (1981) Afferent connections of the rostral medulla of the cat: a neural substrate for midbrain-medullary interactions in the modulation of pain. J. Comp. Neurol. 201:285-297. Aimone, L.D., and G.F. Gebhart (1986) Stimulation-produced spinal inhibition from the midbrain in the rat is mediated by an excitatory amino acid neurotransmitter in the medial medulla. J. Neurosci. 6r1803-1813. Basbaum, A.I. (1989) A rapid and simple silver enhancement procedure for ultrastructural localization of the retrograde tracer WGAapoHRP-Au and its use in double-label studies with post-embedding immunocytochemistry. J. Histochem. Cytochem. 37:1811-1815. Basbaum, A.I., and H.L. Fields (1978) Endogenous pain control systems: Review and hypothesis. Annals of Neurology 4r451-462. Basbaum, A.I., and H.L. Fields (1979) The origin of descending pathways in the dorsolateral funiculus of the cord of the cat and rat: Further studies on the anatomy of pain modulation. J. Comp. Neurol. 187:513-532. Basbaum, A.I., and H.L. Fields (1984) Endogenous pain control systems: Brainstem spinal pathways and endorphin circuitry. Ann. Rev. Neurosci. 7:309-338. Basbaum, A.I., C.H. Clanton, and H.L. Fields (1978) Three bulbospinal pathways from the rostral medulla of the cat. An autoradiographic study of pain modulating systems. J. Comp. Neurol. 178.209-224. Basbaum, A.I., D.D. Ralston, and H.J. Ralston (1986) Bulbospinal projections in the primate: A light and electron microscopic analysis of a pain modulatory system. J. Comp. Neurol. 250:311-325. Bowker, R.M., L.C. Abbott, and R.P. Dilts (1988)Peptidergic neurons in the

328 nucleus raphe magnus and the nucleus gigantocellularis: Their distributions, interrelationships and projections to the spinal cord. Prog. Brain Res. 77:95-127. De Biasi, S., and A. Rustioni (1988) Glutamate and substance P coexist in primary afferent terminals in the superficial laminae of the spinal cord. Proc. Natl. Acad. Sci. (USA) 857820-7824. De Biasi, S., C. Frasoni, and R. Spreafico (1988)The intrinsic organization of the ventroposterolateral nucleus and related reticular thalamic nucleus of the rat: A double-labeling ultrastructural investigation with y-aminobutyric acid immunogold staining and lectin-conjugated horseradish peroxidase. Somatosens. Res. 5:187-203. Drower, E.J., and D.L. Hammond (1988) GABAergic modulation ofnociceptive threshold effects of THIP and bicuculline microinjected in the ventral medulla of the rat. Brain Res. 450t316-324. Fardin, V., J.L. Oliveras, and J.M. Besson (1984) Projections from the periaqueductal gray matter to the B3 cellular area (nucleus raphe magnus and nucleus reticularis paragigantocellularis) as revealed by the retrograde transport of horseradish peroxidase in the rat. J. Comp. Neurol. 223:483-500. Fields, H.L., and M.M. Heinricher (1985) Anatomy and physiology of a nociceptive modulatory system. Phil. Trans. R. SOC.Lond. B 308t361374. Fields, H.L., A.I. Basbaum, C.H. Clanton, and S.D. Anderson (1977) Nucleus raphe magnus inhibition of spinal cord dorsal horn neurons. Brain Res. 126t441-453. Fields, H.L., J. Bry, I. Hentall, and G. Zorman (1983a) The activity of neurons in the rostral medulla of the rat during withdrawal from noxious heat. J. Neurosci. 32545-2552. Fields, H.L., H. Vanegas, I.D. Hentall, and G. Zorman (1983b) Evidence that disinhibition of brain stem neurones contributes to morphine analgesia. Nature 306(59441:684-686. Foong, F.W., and A.W. Duggan (1986) Brain-stem areas tonically inhibiting dorsal horn neurones: studies with microinjection of the GABA analogue piperidine-4-sulphonic acid. Pain 27:361-371. Gallager, D.W., and A. Pert (1978)Afferents to brain stem nuclei (brain stem raphe, nucleus reticularis pontis caudalis and nucleus gigantocellularis) in the rat as demonstrated by microiontophoretically applied horseradish peroxidase. Brain Res. 144,257-275. Gray, B.G., and J.O. Dostrovsky (1985)Inhibition of feline spinal cord dorsal horn neurons following electrical stimulation of nucleus paragigantocellularis lateralis. A comparison with nucleus raphe magnus. Brain Res. 348:261-273. Heinricher, M.M., C.M. Haws, and H.L. Fields (1987) Iontophoresis of the GABA antagonist bicuculline blocks off-cell pause: Evidence for GABAmediated control of putative pain modulating neurons in the rostral ventromedial medulla. Neurosci. Abst. Z3:299. Johansson, O., T. Hokfelt, B. Pernow, S.L. Jeffcoate, N. White, H.W.M. Steinbusch, A.A.J. Verhofstad, P.C. Emson, and E. Spindel (1981) Immunohistochemical support for three putative transmitters in one neuron: Coexistence of 5-hydroxytryptamine, substance P- and thyrotropin-releasing hormone-like immunoreactivity in medullary neurons projecting to the spinal cord. Neurosci. 6: 1857-1881. Keeler, J.R., C.W. Shults, T.N. Chase, and C.J. Helke (1984) Ventral surface of the medulla in the r a t pharmacologic and autoradiographic localization of GABA-induced cardiovascular effects. Brain Res. 297t217-224. Lakos, S., and A.I. Basbaum (1988) An ultrastructural study of the projections from the midbrain periaqueductal gray to spinally projecting, serotonin-immunoreactive neurons of the medullary nucleus raphe magnus in the rat. Brain Res. 443:383-388. Lovick, T.A. (1985) Ventrolateral medullary lesions block the antinociceptive and cardiovascular responses elicited by stimulating the dorsal periaqueductal grey matter in rats. Pain 21t241-252. Lovick, T.A. (1986) Analgesia and the cardiovascular changes evoked by stimulating neurones in the ventrolateral medulla in rats. Pain 25:259268. Lovick, T.A. (1987) Tonic GABAergic and cholinergic influences on pain

H.J. CHO AND A.I. BASBAUM control and cardiovascular control neurons in nucleus paragigantocellularis lateralis in the rat. Pain 31t401-409. Meeley, M.P., D.A. Ruggiero, T. Ishituka, and D.J. Reis (1985) Intrinsic y-aminobutyric acid neurons in the nucleus of the solitary tract and the rostral ventrolateral medulla of the r a t a n immunocytochemical and cytochemical study. Neurosci. Lett. 58:83-89. Millhorn, D.E., T. Hokfelt, K. Seroogy, W. Oertel, and A.A.J. Verhofstad (1988) Extent of colocalization of serotonin and GABA in neurons of the ventral medulla oblongata in rat. Brain Res. 461:169-174. 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. Milner, T.A., V.M. Pickel, J. Chan, V.J. Massari, W.H. Oertel, D.H. Park, T.H. Joh, and D.J. Reis (1987) Phenylethanolamine N-methyltransferasecontaining neurons in the rostral ventrolateral medulla. 11. Synaptic relationships with GABAergic terminals. Brain Res. 421:4657. Mugnaini, E., and W.H. Oertel (1985) An atlas of the distribution of GABAergic neurons and terminals in the rat CNS as revealed by GAD immunohistochemistry. In A. Bjorklund and T. Hokfelt (eds): Handbook of Chemical Neuroanatomy, Vol. 4, GABA and Neuropeptides in the CNS. Part I. Amsterdam: Elsevier, pp 436-608. Paxinos, G., and C. Watson (1986) The Rat Brain in Stereotaxic Coordinates, 2nd ed. Orlando: Academic Press. Pomeroy, S.L., and M.M. Behbehani (1979) Physiologic evidence for a projection from the periaqueductal gray to nucleus raphe magnus in the rat. Brain Res. I76:143-147. Reichling, D.B., and A.I. Bashaum (1991) The contribution of hrainstem GABAergic circuitry to descending antinociceptive controls. I. GABAimmunoreactive projection neurons in the periaqueductal gray and nucleus raphe magnus. J. Comp. Neurol. (in press). Reichling, D.B., H.J. Cho, and A.I. Basbaum (1988)GABA-immunoreactive synaptic contacts onto projection neurons of the periaqueductal gray matter and the nucleus raphe magnus. SOC. Neurosci. Abs. 14:856. Roberts, M.H.T. (1984) 5-hydroxytryptamine and antinociception. Neuropharm. 23:1529-1536. Ruggiero, D.A., M.P. Meeley, M. Anwar, and D.J. Reis (1985) Newly identified GABAergic neurons in regions of the ventrolateral medulla which regulate blood pressure. Brain Res. 339t171-177. Sandkuhkler, J., and G.F. Gebhart (1984) Relative contributions of the nucleus raphe magnus and adjacent medullary reticular formation to the inhibition by stimulation in the periaqueductal gray of a spinal nociceptive reflex in the pentobarbitol-anesthetized rat. Brain Res. 305.77-87. Siddall, P.J., and R.A.L. Dampney (1989) Relationship between cardiovascular neurones and descending antinociceptive pathways in the rostral ventrolateral medulla of the cat. Pain 37:347-355. Skinner, K., A.I. Basbaum, P. Mason, and H.L. Fields (1990) Gabaimmunoreactive synapses onto physiologically identified neurons in the cat rostral ventromedial medulla. Pain Suppl. 5: S446. Sun, M.-K., and P.G. Guyenet (1986) Effect of clonidine and y-aminobutyric acid on the discharges of medullo-spinal sympathoexcitatory neurons in the rat. Brain Res. 368:l-17. Willette, R.N., A.J. Krieger, P.P. Barcas, and H.N. Sapru (1983) Medullary y-aminobutyric acid (GABA) receptors and the regulation of blood pressure in the rat. J. Pharmacol. Exper. Therap. 226:893-899. Willis, W.D., L.H. Haber, and R.G. Martin (1977)Inhibition of spinothalamic tract cells and interneurons by brainstem stimulation in the monkey. J. Neurophysiol. 40:968-981. Yamada, K.A., R.M. McAllen, and A.D. Loewy (1984) GABA antagonists applied to the ventral surface of the medulla oblongata block the baroreceptor reflex. Brain Res. 297t175-180. Yamada, K.A., W.P. Norman, P. Hamosh, and R.A. Gillis (1982) Medullary ventral surface GABA receptors affect respiratory and cardiovascular function. Brain Res. 248:71-78.