THE JOURNAL OF COMPARATIVE NEUROLOGY 313:349-367 (1991)

GABA-Synthesizing Neurons in the Medulla: Their Relationship to Serotonin-Containing and Spinally Projecting Neurons in the Rat BARBARA E. JONES, COLIN J. HOLMES, ELISIA RODRIGUEZ-VEIGA, AND LYNDA MAINVILLE Department of Neurology and Neurosurgery, Montreal Neurological Institute, McGill University, Montreal, Quebec, Canada H3A 2B4 (B.E.J., C.J.H.), and Departamento de Anatomia Y Anatomia Pathologica Comparadas, Facultad de Veterinaria, Universidad Complutense, Madrid, Spain 28040 (E.R.-V.)

ABSTRACT GAl3A-synthesizing neurons were identified in the medulla of the rat by peroxidaseantiperoxidase (PAP) immunohistochemistry for glutamic acid decarboxylase (GAD). Using diaminobenzidine (DAB) either alone or intensified with silver, a relatively large number of GAD-immunoreactive neurons were evident within the reticular formation, raphe nuclei and vestibular nuclei. In all these areas, profuse GAD-immunoreactive varicosities appeared to contact the soma and dendrites of both non-GABA and GABA neurons. These observations suggest that GABA neurons may act as interneurons or local projection neurons within the medulla and accordingly exert a potent inhibitory and/or disinhibitory control on bulbar projection neurons. Within the ventral reticular formation (pars alpha and ventralis of the gigantocellular reticular field) and raphe magnus, large numbers of prominent GAD-immunoreactive neurons resembled in size and morphology and overlapped in distribution the serotonin-immunoreactive neurons of the same regions. However, by sequential double immunostaining utilizing DAB as a chromogen for serotonin (5-HT) and benzidine dihydrochloride (BDHC) for GAD, it was found that GAD-containing neurons were distinct from 5-HT-containing neurons. Following injections of wheat germ agglutinin conjugated to horseradish peroxidase (WGA-HRP) into the upper cervical spinal cord and combined processing for WGA-HRP (using tetramethylbenzidine [TMB] with cobalt) and immunohistochemistry(with DAB), a contingent of spinally projecting neurons were found to contain GAD. The GAD-immunoreactive reticuloand raphe-spinal neurons were most frequent within the pars alpha and ventralis of the gigantocellular reticular fields and the raphe magnus, where they were approximately equal in number to the coexistent, but distinct 5-HT spinally projecting neurons. GABA neurons of the medulla may thus contribute directly to the bulbar inhibitory influence upon spinal sensory and motor systems. Key words: glutamic acid decarboxylase, reticular formation,raphe nuclei, vestibular nuclei, spinal cord

Gamma-aminobutyric acid (GABA) has long been known to be an important inhibitory neurotransmitter within the central nervous system (Krnjevic and Schwartz, '661, although its presence and importance within the brainstem and spinal cord, and particularly within bulbospinal systems, has been less well recognized (Krnjevic et al., '77). With the more recent immunohistochemical revelation of glutamic acid decarboxylase (GAD), the synthetic enzyme for GABA, it has become evident that GABA-synthesizing o 1991 WILEY-LISS, INC.

neurons are present within the brainstem, and that GABAsynthesizing nerve terminals are distributed through the brainstem and the spinal cord (Mugnaini and Oertel, '85). A prominent group of GAD-immunoreactive neurons has Accepted July31,1991. Address reprint requests to Dr B.E. Jones, Neuroanatomy Laboratory, Montreal Neurological Institute, 3801 University St., Montreal, Quebec, Canada, H3A 2B4.

B.E. JONES ET AL.

350

been identified within the ventral medullary reticular formation (Mugnaini and Oertel, '85; Ruggiero et al., '85), overlapping in position with the region that Magoun and Rhines ('46) originally identified as an area from which inhibition of motor activity and spinal reflexes could be elicited by electrical stimulation. This inhibition was shown to be due in part to a direct, postsynaptic inhibition of spinal motor neurons through a ventral reticulospinal system (Llinas and Terzuolo, '64; Jankowska et al., '68). Inhibition of neck motor neurons produced by stimulation of the gigantocellular reticular field was subsequently found to involve a monosynaptic projection (Peterson et al., '78). Inhibition of sensory transmission was also found to occur through direct, postsynaptic inhibition of interneurons and sensory relay neurons, in addition to presynaptic inhibition of primary afferents, via a dorsal reticulospinal system (Engberg et al., '68). The inhibition of the spinoreticular and spinothalamic neurons by stimulation of the gigantocellular reticular field and raphe may also involve a monosynaptic projection (Fields et al., '77; Gerhart et al., '81; Giesler et al., '81; Light et al., '86). The presence of GABA neurons within the region of the medullary reticular formation and raphe from which such inhibitory effects were evoked, raises the possibility that GABA reticulospinal neurons could participate in such inhibitory processes. This study was undertaken, first, to examine in detail the identity and distribution of GABA neurons within the medulla by immunohistochemical staining for GAD and, second, to determine if medullary GABA neurons project directly to the spinal cord and could thus contribute to bulbospinal inhibitory mechanisms. In the process of this investigation, it also became obvious that the GAD-immunoreactive neurons overlapped extensively with serotonin (5-HT)-containing neurons of the medullary raphe and reticular formation, and indeed it was brought forward by Millhorn et al. ('87) that GABA might be co-localized with serotonin in raphe neurons, including raphespinal projection neurons. The putative inhibitory role in sensory transmission attributed to serotonin raphespinal projections into the dorsal horn (Fields and Basbaum, '78) could be due to such co-localization of GABA with serotonin in the same neurons. This study also

Abbreviations

I 12

A DP, DPGi Gi GiA GiV

I0 IRt LLPGi LP, LPGi LVe MVe PCR PM, PMn PrH RM RO RPO Sol SP5 SpVe SuVe

facial nucleus hypoglossal nucleus ambiguus nucleus dorsal paragigantocellular field gigantocellular reticular field gigantocellular reticular field, pars alpha gigantocellular reticular field, pars ventralis inferior olivary nuclei intermediate reticular field lateral, lateral paragigantocellular field lateral paragigantocellular field lateral vestibular nucleus medial vestibular nucleus parvocellular reticular field paramedian reticular field prepositus hypoglossal nucleus raphe magnus nucleus raphe obscurus nucleus raphe pallidus and obscurus nuclei solitary tract and nucleus spinal trigeminal nucleus spinal vestibular nucleus superior vestibular nucleus

examined in detail the relationship of GAD-immunoreactive neurons to 5-HT-immunoreactive neurons within the medulla by sequential double immunohistochemical staining and investigated the relationship of the GABA cells to serotonin raphespinal neurons. Preliminary results of this study have been presented (Rodriguez-Veiga et al., '90).

MATERIALS AND METHODS Animals, surgery, and perfusion Male Wistar (Charles River) rats, weighing approximately 250 g, were anaesthetized with sodium pentobarbital (Somnotol, 60 mg/kg) for surgery and for sacrifice. In one series of animals, in which the brains were processed for simple immunohistochemistry, several animals received local injections of colchicine (10-75 kg in 0.1-0.75 p1) into the medulla, several received injections of ) the lateral ventricle and colchicine (60 kg in 30 ~ 1 into others received no colchicine pretreatment. The colchicinepretreated animals were sacrificed within 24 hours after injection by perfusion-fixation. In another series of animals, in which retrograde transport was studied, the rats received injections of 25, 50 or 100 k1 of wheat germ agglutinin conjugated to horseradish peroxidase (WGA-HRP; in a 5% solution) via a capillary tube (cemented to a Hamilton microsyringe) into the upper cervical spinal cord (Cl). These animals were sacrificed approximately 48 hours later by perfusion-fixation. Several of these animals received colchicine pretreatment 24 hours prior to perfusion. For fixation of the brain, animals were perfused transcardially with three consecutive solutions: (1)< 100 ml normal saline in 0.1M phosphate buffer (PB), (2) = l o 0 0 ml 3% paraformaldehyde in PB (with 0.1% glutaraldehyde and/or 0.2% picric acid in some cases), and ( 3 ) 250 ml of 10% sucrose in PB. The brains were subsequently removed, placed in 30%sucrose in PB for approximately 24 hours at 4°C. frozen to -50°C and stored at -80°C.

Immunohistochemistry Sections were cut at 25 krn thickness on a freezing microtome. Adjacent series were collected every 400 km for immunohistochemistry. The peroxidase-antiperoxidase (PAP) technique was employed for all antibodies (Sternberger, '79). A sheep GAD antiserum that was previously developed and carefully characterized by Oertel, Mugnaini and colleagues (Oertel et al., '81; Oertel et al., '82) was kindly supplied by E. Mugnaini for this study. It was employed at a dilution of 1:2000 and in association with sheep preimmune serum employed on control sections at the same dilution. Anti-5-HT antiserum from rabbit (INCSTAR) was diluted at 1:7500 and in association with normal rabbit serum employed on control sections at the same dilution. The sections were incubated overnight at room temperature with the primary antisera, for 30 minutes with secondary antisera (ICN rabbit anti-sheep or goat anti-rabbit),and for 30 minutes with PAP (Dimension goat or rabbit), and were rinsed in Tris-saline between incubations. A double bridge procedure (Vacca et al., '75) was employed for GAD immunohistochemistry. PAP was revealed with diaminobenzidine (DAB) in single label experiments. In some series, the DAB was intensified with silver according to a previously adapted procedure (Gallyas et al., '82; van den Pol and Gorcs, '86; Jones, '90a).

GABA NEURONS WITHIN THE MEDULLA

Fig. 1. Photomicrographs of sections through the rostral (A, P2.3 approximately) and caudal (B, P3.1 approximately) medulla that have been processed by PAP immunohistochemistq with DAB for GAD.

351

GAD-immunoreactive cells are evident through the reticular formation, raphe nuclei and vestibular nuclei upon a variable blanket of GAD-immunoreactive varicosities. Magnification bar = 200 pm.

B.E. JONES ET AL.

352

Figure 2A

In series from several animals, sections were doubly immunostained by a sequential procedure (Levey et al., '86) for serotonin and GAD, with DAB as a chromogen for serotonin processed in the first position, and benzidine &hydrochloride (BDHC) as a chromogen for GAD processed in the second position. In this procedure, the dilution

of the serotonin antiserum was 1:10,000 and of the GAD antiserum 1:2000;secondary antibodies were from donkey (Jackson donkey anti-sheep and donkey anti-rabbit). Using sheep preimmune serum and normal rabbit serum, three control series were run for the double labelling procedure by substituting the normal antisera for one or both of the

GAJ3A NEURONS WITHIN THE MEDULLA

353

1

Fig. 2. Atlas figures of four levels (P1.5, P2.3, P3.1, and P3.9) through the medulla in which outlines of structures were drawn from Nissl stained sections with the aid of a computer image analysis system. Stereotaxic levels and neuroanatomical terms correspond approldmately to those established by Paxinos and Watson ('86). Open circles

indicate GAD+ cells and closed squares indicate 5-HT+ cells that were mapped at high magnification from adjacent series of sections which were superimposed upon the computer template of the corresponding section from the same animal (SC32).

primary antisera (against GAD or 5-HT). The double labelling procedure was perfected so that no staining or coloration (by DAB or BDHC) was present in the absence of a primary antibody. The same procedure has been used for other chemicals and neurotransmitters and, as confirmed in our laboratory, has been found to provide double staining of cell soma containing two chemical markers (Sugimoto et al., '85).

Combined WGA-HRP histochemistry and immunohistochemistry WGA-HRP was revealed by using DAB in one series and tetramethylbenzidine (TMB) in another series of sections (1 every 800 km) with cobalt intensification in both cases according to previously employed (Jones, 'goal, published procedures (Mesulam, ' 7 8 Mesulam et al., '86; Rye et al.,

B.E. JONES ET AL.

354

neurons is proportional to the cell density in different regions, being highest in the periventricular zone (medial vestibular nucleus and prepositus hypoglossi nucleus) and lowest through the giant cell field where cells lie scattered within the large fascicles of longitudinal fiber bundles of the reticular core. The GAD+ cells are very frequent within the ventromedial reticular formation, and particularly within the pars alpha and ventralis of the gigantocellular field, the Analysis lateral paragigantocellular field, and the raphe magnus. GAD-immunoreactive varicosities are relatively dense Sections were viewed with a Leitz Orthoplan microscope equipped with an image analysis system (Biocom, Paris). through the medulla, ostensibly providing an innervation Outlines of sections and structures were drawn at low to neurons through the vestibular nuclei, the reticular core, magnification from Nissl stained sections at intervals of and the raphe nuclei (in addition to the spinal trigeminal 800 pm in one rat (SC32). These graphic templates were and solitary tract nuclei; Figs. 1 and 3). Extremely fine superimposed over the video image of corresponding sec- GAD+ fibers with prominent, bulbous varicosities run in tions from all rats, allowing mapping of labelled elements at between the large fiber fascicles of the reticular core in close high magnification. Accordingly, cells that were GAD-, proximity to the dendritic processes and cell bodies of 5-HT-, WGA-HRP-positiveor double labelled, in which the neurons, including GAD-immunoreactiveneurons (Fig. 3). nucleus was evident, were mapped in adjacent series from This filigree of GAD+ fibers reaches cells throughout the the three animals which received injections of 100 p1 reticular formation and raphe, including most prominently WGA-HRP into the spinal cord. Crude counts of the the raphe magnus, but also the raphe pallidus and obscumapped cells were obtained automatically in each structure rus. As strikingly evident in silver intensified material, the for each series. Due to the wide range in cell sizes, estimates GAD+ varicosities envelop the large reticular neurons, of maximal numbers of cells were made by simple multipli- surrounding their soma and proximal and distal dendrites cation of the crude cell counts, obtained in 25 pm thick (Fig. 4). A similar dense innervation of large and giant sections (every 800 pm), by the number of sections (32) in neurons is evident within the lateral vestibular nucleus and, to a lesser extent, within the spinal vestibular nucleus. the interval between samples. Through the periventricular zones (including the medial vestibular nucleus and prepositus hypoglossi nucleus) and RESULTS the parvicellular reticular field, GAD+ varicosities are also relatively dense, although their specific association with the Technical considerations of GAD processes of the smaller neurons was not clearly evident. immunohistochemistry Nonetheless, in the immunohistochemical material proIn preliminary studies of GAD-immunoreactivity,colchi- cessed with DAB or silver intensified DAB, it appeared that cine pretreatment was employed to enhance the levels of the vast majority of neuronal cell bodies through the GAD within the cell soma. It was subsequently found, reticular formation, raphe and vestibular nuclei, could be however, that by employing perfusion-fixation of the brain contacted by GABA varicosities. and by utilizing 3%paraformaldehyde alone or in combination with picric acid (but in the absence of glutaraldehyde) Relationship of GAD+ neurons to that the pretreatment was not necessary for visualization of the GAD-immunoreactive neurons within the brainstem. 5-HT neurons Furthermore, local injections of colchicine virtually elimiGAD-immunoreactiveneurons are of prominent size and nated retrogradely transported WGA-HRP, and intraven- number within the ventromedial medulla in a position that tricular injections greatly decreased the same in medullary overlaps with that of the serotonin raphe neurons (Fig. 5 neurons. For this reason, all results reported for retro- and Fig. 2). In fact, the GAD+ neurons are morphologically gradely labelled cells, and all quantitative results concern- indistinguishable from the 5-HT+ neurons in this region, ing numbers of immunoreactive and retrogradely labelled both being bi- or multi-polar medium sized cells (average cells, are based entirely upon nonpretreated animals that long axis of GAD+ neurons = 20.6 pm as compared to that were perfusion-fixed with 3% paraformaldehyde. of 5-HT+ neurons = 20.0 pm in 3 animals). The overlap is particularly striking within the more rostral raphe, here GAD-immunoreactiveneurons and processes called for simplicity the raphe magnus (at levels P1.5 and GAD-immunoreactive neurons are distributed through P2.3) where the numbers of GAD+ neurons are similar to the medulla in the vestibular nuclei, reticular formation, the numbers of 5-HT+ neurons (Table 2), in contrast to the and raphe nuclei, as well as in other sensory and relay more caudal raphe, here called the raphe pallidus-obscurus nuclei (Figs. 1and 2). Across these structures, the size and (at levels P3.1 and P3.9), where the numbers of GAD+ shape of GAD+ neurons varies according to region. In most neurons are inferior to those of 5HT+ neurons (Table 2). structures, such as the gigantocellular field (Fig. 3A), they The overlap of GAD+ and 5-HT+ neurons extends from are small to medium in size (measuring 12 to 18 pm across the raphe laterally into the reticular formation where, the long axis) and oval or fusiform. The largest GAD+ within the pars alpha and ventralis of the gigantocellular neurons are found within the ventral reticular formation field, similar numbers of GAD+ and 5-HT+ neurons and raphe (in the pars alpha and ventralis of the gigantocel- coexist. Given the overlap and similarity of the GAD+ and M a r field and the lateral paragigantocellular field, and in the raphe magnus), where they measure 17 to 26 pm 5-HT+ cells, the possible co-localization of serotonin with GABA in the same neurons was investigated by sequential (across the long axis) and are often multipolar (Fig. 3B). GAD-immunoreactive neurons are relatively numerous double immunohistochemical staining for serotonin and through the vestibular nuclei, reticular fields, and raphe GAD with DAB as a chromogen for 5-HT, revealed first, and nuclei (Figs. 1, 2 and Table 1). The density of GAD+ BDHC as a chromogen for GAD, revealed second (Fig. 6). In

'84). Adjacent series of sections were processed for GAD or 5-HT immunohistochemistry with DAB as a chromogen. For each antibody, a series of controls (processed for DAB-Co" or TMB-Co" histochemistry) was processed using normal sera in lieu of the primary antisera. A series of TMB-Co" processed sections was stained for Nissl with thionin.

+

GABA NEURONS WITHIN THE MEDULLA

Fig. 3. Photomicrographs of GAD-immunoreactivecells and processes revealed by immunohistochemistry with DAB in the gigantocellular (Gi) field (A) and in the raphe magnus (RM) and adjacent pars alpha of the Gi field (GW (B). Magnification bar = 20 km.

355

B.E. JONES ET AL.

356

TABLE 1. Number of GAD+ Neurons in Vestibular. Reticular and Raohe Nuclei of the Medulla in Three Rats'

sc 33

sc 32 Structure

Left

Right

SpVe LVe Mve PrH DPGi PCR IRt LLPGi

94 0 215 130 82 94 96 37 2 124 18 67 59 109 81 1208 38656

101 0 247 115 78 71 86 34 2 105 17 65 45 95 98 1159 37088

PMn Gi RPO Giv

RM GiA2 LPGi Actual Total Estimated Total

Left 94 3 322 175 86 112 153 25 2 133 16

54 52 119 130 1476 47232

sc 34

Mean est. total

Mean

Right

Left

Right

Left

107 3 368 188 57 111 151 47 9 154 15 52 56 88 87 1493 47776

123 37 418 190 67 78 151 36 7 77 17 54 63 58 119 1495 47840

99 27 351 193 70 89 125 72 2 97 12 25 38 66 94 1360 43520

104 13 318 165 78 95 133 33 4 111 17 58 58 95 110 1393 44576

Right 102 10 322 165 68 90 121 51 4

119 15 47 46 83 93 1337 42795

Left

Right

3317 427 10187 5280 2507 3030 4267 1045 117 3563 544 1867 1856 3051 3520 44576

3275 320 10304 5291 2187 2891 3861 1632 139 3797 469 1515 1483 2656 2976 42795

'For each nucleus, cells were counted in (1-4) 25 pm thick sections a t 800 pm intervals (P3900, P3100, P2300, PI5001 (depending upon the length of the nucleus). Actual numbers of cells counted are presented for each nucleus of each animal and estimated total number of cells is calculated (as 32 times the actual total) for the mean number per nucleus and total number per animal. 'Cells counted in GiV and GiA include those located within the medial lemniscus and inferior olivary nuclei located ventral to the reticular fields.

It was apparent that, although the two immunostained cell groups did overlap in certain regions, they were nonetheless differentially distributed (Fig. 5 and Fig. 2). Thus the highest concentration of GAD+ neurons was found dorsal and lateral to that of the 5-HT+ neurons within the pars alpha and ventralis of the gigantocellular fields and within the lateral paragigantocellular field, where few serotonin neurons were found (Table 2).

Relationship of GAD+ neurons to spinally projecting neurons

Fig. 4. Photomicrograph of GAD-immunoreactive varicosities revealed by immunohistochemistry employing silver intensification of the DAB reaction product. Varicosities surround the cell bodies and dendrites of three large reticular neurons within the gigantocellular field. Magnification bar = 20 km.

the slides, a distinct differentiation between the 5-HT+ neurons, which were stained a homogeneous brown by DAB, and the GAD+ neurons, which were stained a granular blue by BDHC, was apparent. No evidence for co-localization was found in the double labelled material in any of the regions of overlap, including the raphe magnus and pars alpha of the gigantocellular field, where the overlap was the most striking (Fig. 6). When counts of GAD+ and 5-HT+ cells in the double labelled sections were compared with the counts of these in single labelled sections, it was found that the respective totals of the two cell types corresponded approximately in the different series, further militating against the co-localization of GAD and 5-HT in one population of cells.

As within the medulla, GAD-immunoreactive cells and processes are present within the gray of the spinal cord (Fig. 7A). GAD-immunoreactive varicosities are particularly dense within the superficial laminae of the dorsal horn, but are also distributed through the intermediate zone and ventral horn, as well as the cornu commissuralis. In the latter regions, GAD-immunoreactive terminals surround the soma and proximal dendrites of large neurons, including most prominently, the motor neurons. The possibility that (at least a portion of) the GABA innervation in the spinal cord originates from GAD+ neurons within the medulla was investigated by retrograde transport of WGA-HRP injected into the upper cervical spinal cord (Fig. 7B). In initial experiments, 25, then 50, and finally 100 pl of a 5% solution of WGA-HRP was injected into C1 in an effort to retrogradely label the maximum number of GAD+ cells in the medulla. Although the number of retrogradely labelled GAD+ cells appeared greater following 50 pl relative to 25 pl, it did not appear to be greater following 100 p1 relative to 50 pl, reaching a saturation point for the uptake and retrograde transport from the region. To insure estimation of the maximal retrograde labelling, cell counts were performed in the three animals having received injections of 100 p1 WGAHRP (Tables 1-5). The injection sites were centered within the intermediate zone and extended into the ventral horn and over the dorsal horn, as well as into the cornu commissuralis. With 100 pl WGA-HRP, the label extended bilaterally through the cornu commissuralis in the three animals and in one animal (SC32 as shown), into the contralateral medial intermediate zone.

GABA NEURONS WITHIN THE MEDULLA

Fig. 5. Photomicrographs of sections through the rostral medulla (somewhat caudal to P2.3, Fig. 2) processed by immunohistochemistry with DAB for GAD (A) and 5-HT (B). GAD-immunoreactive cells are distributed through the raphe magnus (RM) and adjacent pars alpha of

357

Gi (GiA) and lateral paragigantocellular (LPGi) reticular fields (A) in a position overlapping with, yet different from that of the 5-HTimmunoreactive neurons (B). Magnification bar = 200 km.

B.E. JONES ET AL. TABLE 2. Number of 5-HT+ Neurons in Vestibular, Reticular and Raphe Nuclei of the Medulla in Three Rats'

sc 32

sc 33

sc 34

Structure

Left

Right

Left

Right

Left

Right

SpVe

0 0 0 0 0 0 0 0 1 0 76 41 28 43 5 194 6208

0 0 0 0 0 0 0 0 0 0 65 58 18 40 6 187 5984

0 0 0 0 0 0 0 1 0 3 66 26 74 88 5 263 8416

0 0 0 0 0 0 0 0 0 5 59 23 38 75 9 209 6688

0 0 0 0 0 0 0 0 3 3 62 29 54 68 8 227 7264

0 0 0 0 0 0 0 0

LVe Mve PrH DPGi PCR IRt LLPGi PMn Gi RPO Giv RM GiA2 LPGi Actual Total Estimated Total

Mean est. total

Mean

4

2 51 20 48 63 20 208 6656

Left

Right

Left

Right

0 0 0 0 0 0 0 0 1 2 68 32 52 66 6 228 7296

0 0 0 0 0 0 0 0 1 2 58 34 35 59 12 201 6443

0 0 0 0 0 0 0 11 43 64 2176 1024 1664 2123 192 7296

0 0 0 0 0 0 0 0 43 75 1867 1077 1109 1899 373 6443

'.'See Table 1.

A large number of bulbar neurons were retrogradely labelled from the cervical spinal cord (Fig. 8 and Table 3). Without correction for cell size, it was estimated that up to 25,000 and 15,000 bulbar neurons on the ipsilateral and contralateral sides, respectively, project to the upper cervical cord. These neurons were most highly concentrated within the vestibular nuclei, the medial reticular formation, and the raphe nuclei. Medium and large neurons within these regions were densely packed with retrogradely labelled granules of WGA-HRP, as evident in both DAB-Co" and TMB-Co" processed sections. A small number of GAD+ neurons were retrogradely labelled from the upper cervical spinal cord (Fig. 8 and Table 4).Without correction for cell size, it was estimated that up to 2100 and 1850 GAD+ bulbar neurons on the ipsilateral and contralateral sides, respectively, project to the upper cervical spinal cord. These neurons were most frequent within the (spinal) vestibular nuclei, medial reticular formation, and the rostral raphe. Although such GAD+ /WGA-HRP+ cells could be visualized occasionally in sections processed with DAB-Cot+ for WGA-HRP, they were only evident in significant numbers in sections processed with TMB-Co". Even in this material, the amount of retrogradely transported WGA-HRP evident in the soma of the GAD+ neurons, was usually only light to moderate (Fig. 9). All the bulbospinal neurons retrogradely labelled with WGA-HRP appeared to be contacted or surrounded by GAD-immunoreactive varicosities within the vestibular nuclei, reticular formation and raphe nuclei. The GAD+/ WGA-HRP+ cells also appeared to have GAD+ varicosities in close proximity to their soma in these regions. Of all the bulbo- (vestibulo-, reticulo-, raphe-) spinal neurons projecting to the upper cervical spinal cord, the GAD + neurons represented approximately 10% (Table 6). In areas such as the spinal vestibular nucleus and the ventromedial reticular formation and raphe, they represented up to one third of the bulbospinal neurons. Of all the GAD+ cells in the medulla, only approximately 5% were retrogradely labelled from the upper cervical spinal cord (Table 7). In regions where significant numbers of GAD+ cells were labelled, such as in the ventromedial reticular formation, this proportion reached 15%.

Relationship of GAD+ to 5-HT+ spinally projecting neurons There was an overlap in the distributions of retrogradely labelled GAD+ and 5-HT+ neurons within the ventromedial medulla (Fig. 8). Morphologically the two groups of bulbospinal neurons are also similar, both being largemedium cells (the retrogradely labelled cells being the same size as the nonretrogradely labelled cells for both GAD+ and 5-HT+ neurons). In general, however, it appeared that the 5-HT+ neurons were more densely retrogradely labelled than the GAD+ neurons, the dark brown 5-HT+ cell bodies being filled with black granules (so as to preclude the photomicrographic illustration in black and white of the double labelling of these cells). Despite the overlap and similarity of GAD+ and 5-HT+ bulbospinal neurons, the distributions and frequencies of the two cell types were found to be somewhat different (Fig. 8 and Table 5). The 5-HT+ bulbospinal neurons were most numerous within the caudal raphe, here called pallidusobscurus, whereas the GAD bulbospinal neurons were most numerous within the rostral raphe, here called raphe magnus. Through the adjacent ventralis and alpha gigantocellular fields of the reticular formation, the 5-HT+ and GAD neurons are similar in frequency. Of all the neurons projecting to the upper cervical cord from the medulla, the 5-HT+ neurons represented approximately 5% (Table 6). These cells are all located within the raphe and ventromedial reticular formation where they represented approximately one fourth of the cervical cord projecting neurons. In the caudal raphe pallidus-obscurus, they represented up to 80% of the total. In the same regions, the GAD + neurons represented approximately 15%of the projecting neurons, and in the raphe pallidusobscurus less than 10%. The proportion of bulbospinal neurons represented by GAD+ neurons was similar to that represented by 5-HT+ neurons in the ventromedial reticular formation and raphe magnus. Together the two comprised one third of the total population of bulbospinal neurons in the latter regions. Of all the serotonin neurons in the medulla, less than one fourth were retrogradely labelled from the upper cervical spinal cord (Table 7). Of all the regions, the largest propor-

+

+

GABA NEURONS WITHIN THE MEDULLA

359

Fig. 6. Photomicrographs taken from one section through the raphe magnus (A) and the lateral paragigantocellular field (LPGD (B) and processed for double sequential immunohistochemical staining first for 5-HT, which is revealed by the floccular, brown DAB product and

second for GAD, which is revealed by the granular, blue BDHC product (in an animal pretreated with an intraventricular injection of colchicine). GAD is contained in different neurons (arrowheads) than 5-HT. Magnification bar = 20 km.

tion (up to one third) was retrogradely labelled in the caudal raphe (pallidus-obscurus)and adjacent reticular formation (pars ventralis). In the more rostra1 regions, the proportion of 5-HT neurons that were retrogradely labelled was similar to that of the GAD+ cells.

DISCUSSION

+

The results of this investigation have shown that a large population of GABA-synthesizing neurons are present within the medulla, and particularly within the vestibular

360

B.E. JONES ET AL.

GABA neurons and processes in the medulla

Fig. 7. Photomicrographs of sections through the cervical spinal cord processed by PAP-immunohistochemistry with DAB for GAD (A) and by histochemistry using DAB-Co" for WGA-HRP to reveal the injection site (of SC32) (B). GAD-immunoreactive varicosities are particularly dense within the superficial laminae of the dorsal horn but are also distributed through the intermediate zone and ventral horn as well as the corm commissuralis. The injection of 100 ~1 of WGA-HRP into the spinal gray resulted in a dense labelling of the ipsilateral intermediate zone and ventral horn and moderate labelling of the ipsilateral dorsal horn, as well as moderate labelling of cornu commissuralis (and contralateral medial intermediate zone in this animal). Magnification bar = 200 pm.

nuclei, reticular formation, and raphe nuclei, where neurons also appear t o receive an innervation by GABA axon terminals. The GABA neurons could act primarily as interneurons within the medulla and influence the prominent bulbospinal projection neurons. In addition, a certain percentage of GAD-immunoreactiveneurons were found to project to the cervical spinal cord. Many of these were located in the medial reticular formation and raphe, where they were in part intermingled with, yet distinct from, serotonin raphespinal neurons.

The results of this study confirm the presence of GABAsynthesizing neurons within the medulla (Mugnaini and Oertel, '85), and document their relatively high frequencies through all structures, notably the vestibular nuclei, reticular formation and raphe nuclei. Similar observations were recently reported for the rabbit (Blessing, '90). Through the medulla of the rat, up to 45,000 GABA neurons were estimated t o be present on each side. The highest density of the GABA neurons was found within the vestibular nuclei (and prepositus hypoglossi nucleus), where a generally high density of cell packing exists. Through the reticular fields, GABA neurons are more sparsely distributed, as are the indigenous large reticular neurons with which they lie among the fascicles of fibers intrinsic to this structure. The ratio of GABA neurons to other neurons within the reticular formation is not necessarily much lower than that within the vestibular nuclei, given the difference in general cell densities between the two structures. Through the reticular core as well as other bulbar nuclei, neurons appear to be richly innervated by GABA nerve terminals. The dendrites and soma of the large reticular neurons within the central gigantocellular field are surrounded by GAD-immunoreactivevaricosities. As observed in the same material, the density of this peri-somatic and peri-dendritic innervation is equal to that surrounding Deiter's neurons within the lateral vestibular nucleus (Mugnaini and Oertel, '85). In addition, through the reticular formation and the raphe nuclei, almost all cells, including GABA neurons, appeared at the light microscopic level t o be contacted by GAD-immunoreactivevaricosities. The ubiquitous presence of GABA neurons and GABA axon terminals through the medial, as well as the lateral, reticular formation suggests that these cells may function as interneurons within this structure. Yet from the early Golgi studies of Cajal ('09) and the later studies of the Scheibels ('58) and Valverde ('611, it would not appear that short-axoned Golgi type I1 neurons exist in the reticular core. It is possible that such apparent absence was due to the technical limitations of the Golgi technique, being selective for certain types of cells. Indeed most cells described in these Golgi studies within the medial reticular formation concerned large or giant neurons, whereas the GAD-immunoreactive cells identified in this study within the central gigantocellular field were small to medium in size. On the other hand, the GABA neurons could be local projection neurons extending their axons through the reticular core and emitting short local collaterals, as reticular neurons have been visualized to do in Golgi material (Scheibel and Scheibel, '58). The largest GAD-immunoreactive neurons were found within the ventral reticular formation and raphe nuclei, where they were also the most concentrated within the reticular formation. These GABA neurons had previously been identified by Mugnaini and Oertel('85) and by Ruggiero et al. ('85). The GABA neurons through this region may number up to 10,000 per side. According to their prominent size and multipolar architecture, they resemble projection neurons, the majority of which are known to give rise to descending projections to the spinal cord from this region (Jones andYang, '85).

GABA NEURONS WITHIN THE MEDULLA

361

TABLE 3. Number of WGA-HRP+ Neurons Retrogradely Labelled From the Spinal Cord in Vestibular, Reticular and Raphe Nuclei of the Medulla in Three Rats'

Structure

Left

SpVe LVe Mve

Right

42 12 64 4 55 61 60 7

PrH DPGi PCR IRt LLPGi PMn Gi RPO

48 24 43 4 30 121 89 5 7 231 32 135 33 68 20 890 28480

16

GiV RM GiA'

LPGi Actual Total Estimated Total

183 29 84 23 40 12 692 22144

sc 34

sc 33

SC 32

Left 18

8 30 2 34 23 36 1 10

126 8 9 14 12 9 340 10880

Right

Left

22 35 16 0 21 77 70 2 4 170 13 6 20 32 6 494 15808

9 53 4 32 62 66 9 4 110 13 11 21 33 10 453 14496

Right 32 34 34 5 24 180 147 6 3 266 16 31 46 79 23 926 29632

16

Mean est. total

Mean Left

Right

25 10 49 3 40 49 54 6 10 140 17 35 19 28

34 31 31 3 25 126 102 4 5 222 20

811 309 1568 107 1291

1088 992 992 96 800 4032 3264 139 149 7115 651 1835 1056 1909 523 24640

1728 181 320 4469 533 1109 619 907 331 15841

57

10

Right

1557

33 60 16 770 24640

495 15841

Left

~~

','See Table 1.

TABLE 4. Number of GAD+/WGA-HRP+ Neurons in Vestibular, Reticular and Raphe Nuclei of the Medulla in Three Rats' ~

Structure SpVe LVe Mve PrH DPGi PCR

IRt LLPGi

PMn Gi RPO

Left

Right

8 0 8 1 3 5 9 2 0 20

11 0 3 0 5

5 4 2 2 14 1 13 9

1

GiV

RM GiA2 LPGi Actual Total Estimated Total

20 8 4 3 92 2944

sc 34

sc 33

SC 32

5 3 77 2464

Left

~

_ _ _ _ _ _ _ ~

Mean est. total

Mean

Right

Left

Right

Left

Right

Left

Right

3 0 3 0

8 0 2

1 0

0 1 1 1 1 8 1 3 3 6 3 39 1248

7 1 3 0 3 1

6 0 3 1 3 3 5

7 0 3 0 3 2 4 2

192 0 107 21 85 85 149 43 21 448 32 256 181 160 85 1866

224 11 96 0 96 64 117 53 32 555 43 224 192 288 149 2144

2 0 0 0 5 2 4 1 1 14

2 1 0

20 2 3

1 1

5

6 5 2 44 1408

9 4 53 1696

1

5

2 1 18 1

5 4 13 7 71 2272

1

1 14 1 8 6 5 3 58 1866

1

17 1

7 6 9

5 67 2144

','See Table 1.

TABLE 5. Number of 5-HT+ IWGA-HRP+ Neurons in Vestibular, Reticular and Raphe Nuclei of the Medulla in Three Rats' ~

SpVe LVe Mve PrH DPGi PCR

Left

Right

0 0

0 0 0 0 0 0 0 0 0 0 25 28 0 10 0 63 2016

0 0 0 0 0 0 0

IRt LLPGi PMn GI RPO GlV

0 21 18

RM CIA'

0 4

LPGi

1

Actual Total Estimated Total

44 1408

sc 34

sc 33

SC 32 Structure

Left

Right

0 0 0 0

0 0 0 0 0 1 7 0

5 2 1

16 512

0 0 0 0 0 0 0 0 1 0 10 0 4 4 0 19 608

Left 0 0 0 0 0 0 0 0 0

1 8 0 2 5 1

17 544

~

~

~~

Mean est. total

Mean Right 0 0 0 0 0 0 0 0

3 1 14 3 6 20 5 52 1664

Left 0 0 0 0 0 0 0 0 0 1

12 6 2 4 1 26 822

Right 0 0

0 0 0 0 0 0 1 0 16 10 3 11 2 45 1430

Left

Right

0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 43 11 523 331 107 363 53 1430

21 384 192 75 117 32 822

','See Table 1

Relationship of GABA neurons to serotonin neurons Within the raphe and ventromedial reticular formation, GAD-immunoreactive neurons were found to be similar in

shape and size to serotonin-immunoreactive neurons and to overlap the latter in distribution. However, in contrast to previous studies (Millhorn et al., '87; Millhorn et al.,'88), no evidence for co-localization of GABA and serotonin in the same cells was found in this investigation. The reason

B.E. JONES ET AL.

362

Figure 8A

for this discrepancy may be technical, since a sequential immunostaining procedure using PAP was employed in this study and a simultaneous immunostaining process using immunofluorescencewas employed by Millhorn et al. ('87, '88). The dilution of the primary antisera was 1:ZOOO and

1:10,000 for anti-GAD and anti-5-HT, respectively, in this study, whereas it was 1:400for anti-GAD and anti-5-HT (or 1:ZOO for anti-GABA)in the latter study. It is possible that our procedure, which was perfected to eliminate any nonspecific staining or coloration, lacked the sensitivity to reveal a

GABA NEURONS WITHIN THE MEDULLA

363

I

Fig. 8. Atlas sections on which retrogradely labelled neurons were plotted (in SC32, as in Fig. 2) at high magnification from TMB-Co" treated adjacent series of sections. Plus signs indicate WGA-HRP + neurons, open circles indicate GAD+ IWGA-HRP+ neurons, and filled squares indicate 5-HT+ IWGA-HRP+ neurons that were retrogradely labelled from the upper cervical spinal cord.

subset of double labelled cells. On the other hand, the distribution of the GAD+ and 5-HT+ neurons was also found to be different on adjacent sections, and the numbers of the individually stained neurons on adjacent sections corresponded approximately to the numbers of neurons stained in the double labelled sections. We conclude from our results that the GABA-synthesizing neurons are distinct from the serotonin neurons with which they partially overlap as a group in the ventral reticular formation and raphe.

Relationship of GABA neurons to bulbospinal neurons UP to 40,000 I - I ~ ~ Owere ~ S estimated to be retrogradely labelled from the upper cervical spinal cord. The distribution of these neurons within the medulla is similar to previously observed distributions of spinally projecting neurons (Jones and Yang, '85). It appeared that, in the sections employing DAB-Co" for WGA-HRP and DAB for GAD, all of the retrogradely labelled neurons in the reticu-

B.E. JONES ET AL.

364

Fig. 9. Photomicrograph showing GAL-immunoreactive neurons (revealed with DAB) th at ar e retrogradely labelled with WGA-HRP (revealed with TMB-Go") which was injected into the cervical spinal

cord (of rat SC34). Th e retrogradely labelled cells (arrows) are located in t h e raphe magnus (RM). Magnification bar = 20 km.

TABLE 6. Percent of WGA-HRP+ Neurons Which Were GAD+ or 5-HT+',' Based on the Mean Values of Three Animals (Tables 3,4,5)

TABLE 7. Percent of GAD+ or 5-HT+ Neurons Which Were WGAHRP+'.' Based on the Mean Values of Three Animals (Tables 1 ,2,4,5)

~~

(WGA-HRP + i

(GAD+IWGA-HRP+)

Structure

Left

Right

Left

Right

SpVe LVe Mve PrH DPGi PCR IRt LLPGi PMn Gi

24% 0%

21% 1% 10% 0% 1290 2% 4% 39% 21% 8% 7% 12% 18% 15% 29% 9%

0% 0%

0% 0% 0% 0% 0% 0%

RPO GiV RM GiA2 LPGi Total

7% 20%

7% 5% 9% 23% 7% 10% 6% 23% 29% 18% 26% 12%

(GAD+) + (GAD + IWGA-HRP+)

(WGA-HRP+)

+ (5-HT + IWGA-HRP+I

0% 0% 0% 0% 0%

0%

0%

0%

0% 0% 72% 17% 12% 13%

28% 0% 80% 18% 10% 19% 10% 6%

10%

5%

Structure SpVe LVe Mve PrH DPGi PCR IRt LLPGi

(5-HT+) + (5-HT+IWGA-HRP+)

Left

Right

Left

Right

6% 0% 1%

7% 3%

0% 0% 0% 0% 0% 0% 0% 0% 0% 34%

0% 0% 0% 0% 0% 0% 0%

0% 3%

3% 4% 490

1% 0% 4% 2% 3% 390 23%

PMn

18%

Gi RPO GiV

13% 6% 14% 10% 5%

15% 9% 15% 13% 11%

18% 19% 5% 6%

2%

5% 5%

17% 11%

RM GiA2 LPGi Total

4%

0%

100% 14% 2890 3 I% 10% 19% 14% 22%

'.'See Table 1.

','See Table 1

lar formation and raphe were surrounded by GADimmunoreactive varicosities. Indeed, a high percentage of all synaptic contacts upon spinally projecting neurons located in the rostral ventral medulla and raphe magnus was recently documented by electron microscopic study to be represented by GABA-immunoreactive terminals (Cho and Basbaum, '90). These results would indicate that GABA neurons within the medulla may exert a potent control over bulbospinal neurons, including importantly, the reticulo- and raphe-spinal neurons. Of the bulbospinal neurons retrogradely labelled in this study, approximately 10%were GAD-immunoreactiveneu-

rons. Spinal projections from GABA-synthesizing neurons in the medulla have also recently been reported in the rabbit brain (Blessing et al., '87; Blessing, '90). In the present study, a large number of the retrogradely labelled GAD+ cells were found within the vestibular nuclei, notably the spinal and medial nuclei, where approximately 25% and 10% of the retrogradely labelled cells were GAD+. Through the medial reticular formation, notably the gigantocellular field, GAD-immunoreactive cells represented up to 10% of the reticulospinal neurons. These results would indicate that, from this field of large and giant cells long known to have a strong excitatory influence upon spinal

GABA NEURONS WITHIN THE MEDULLA neurons, a direct inhibitory influence may also originate. These results also show that GABA neurons within the reticular formation are indeed, at least in part, projection neurons. The largest number of retrogradely labelled GAD+ neurons was located within the more ventral reticular formation and raphe, where up to one third of the reticulospinal and raphespinal neurons may be GABA-synthesizing cells.

Relationship of GABA to serotonin bulbospinal neurons GAD-irnmunoreactive spinally projecting neurons were found through the caudal and rostral regions of the raphe and adjacent reticular formation. Spinal projections of GABA neurons were also recently identified in the raphe magnus of the rat by others (Reichlingand Basbaum, '90). In this study, the distribution of serotonin spinally projecting neurons was also examined on adjacent sections. The GAD-immunoreactive neurons were found to be predominantly distributed lateral and dorsal to the 5-HT+ spinally projecting neurons of these areas. In the caudal raphe, here called the pallidus-obscurus, the great majority of the raphespinal neurons were 5-HT+, in conformity with the original reports by Bowker et al. ('83). Only a small percentage were found to be GAD+ here. On the other hand, through the pars ventralis and pars alpha of the gigantocellular field, and through the raphe magnus, only a minor portion of the spinally projecting neurons were 5-HT+, in contrast to previous reports of a major proportion by Bowker et al. ('83; Bowker and Abbott, '901, yet in closer agreement with the report by Skagerberg and Bjorklund ('85).In these areas, the GAD+ neurons represented as important a percentage of the spinally projecting neurons as did the serotonin neurons. Together, GABA and 5-HT neurons comprised less than half of the projection neurons found in these regions. These results suggest, in direct contradiction to Bowker and colleagues, that the raphespinal system cannot be considered as a homogeneous pathway comprised of serotonin neurons. It is in fact a heterogeneous pathway made up of multiple cell types, among which distinct GABA neurons comprise perhaps as important a contingent as the serotonin neurons. Although retrogradely labelled GAD+ neurons may be as frequent as retrogradely labelled 5-HT+ neurons, they are rarely as densely retrogradely labelled as the latter or as any of the other reticular neurons in the region. The relatively small amounts of retrogradely transported WGAHRP in GAD+ neurons from the first cervical spinal gray may reflect a more limited axonal arborization by these neurons in this region than that by serotonin or other reticular neurons. The GAD-immunoreactiveneurons also appear to have very fine axons, which unlike those of serotonin neurons, are difficult to visualize (in coronal, sagittal, or horizontal sections) in immunohistochemical material at the light microscopic level.

Functional significance of GABA neurons in the medulla Given the frequency of GABA neurons within the medullary reticular formation and raphe, as well as within vestibular and other nuclei, and of GABA axonal terminals within the same regions, it may be assumed that these cells must play an important role in bulbar circuits. The GABA neurons may act in part as interneurons or local projection neurons innervating other projection neurons, including

365 the reticulospinal and raphespinal neurons. As in other areas of the central nervous system, GABA has been shown to produce an inhibitory response when applied to reticular neurons (Hosli and Tebecis, '70; Greene and Carpenter, '85). Accordingly, the GABA innervation within the medulla may serve as a potent inhibitory control of reticuloand raphe-spinal, as well as vestibulospinal, excitatory systems. Such disfacilitation may represent an important component of supraspinal inhibition (Llinas, '64).GABA neurons may also be innervated by GABA terminals and thus, as has been found to be the case in many areas of the central nervous system (Mugnaini and Oertel, '851, be involved in mechanisms of disinhibition of bulbar or bulbospinal circuits. The GABA neurons also appear to contribute to the heterogeneous output from the reticular formation, raphe and vestibular nuclei to the spinal cord. The intermingling of GABA neurons with other neurons would suggest that the output from any one field of the reticular formation is heterogeneous and may have excitatory and inhibitory influences on cells within the cord. The relatively high concentration of GAD-immunoreactive spinal projection neurons within the medial vestibular nucleus would perhaps explain the demonstration of a monosynaptic inhibitory influence on neck motoneurons originating from this nucleus (Wilson and Yoshida, '69). Despite the heterogeneity within the reticular formation, higher concentrations of GABA neurons are present within the ventral reticular formation and raphe, compared to the central gigantocellular field. It is from this ventral zone that inhibition of spinal reflexes was originally evoked with stimulation of the reticular formation by Magoun and Rhines ('46), and from where monosynaptic inhibition of cervical motor neurons was elicited by Peterson et al. ('78). This coincidence leads to the suggestion that the GABA spinally projecting neurons may represent at least one substrate of this descending inhibitory system to the cervical spinal cord. Within the ventral reticular formation and raphe, different zones have been shown to project preferentially to the ventral versus the dorsal horn (Basbaum et al., '78; Basbaum and Fields, '79). Thus the caudal raphe (here referred to as the raphe pallidus-obscurus) and the pars ventralis of the gigantocellular reticular field project most heavily to the ventral horn; whereas the rostral raphe (here referred to as the raphe magnus) and the pars alpha of the gigantocellular reticular field, project most heavily to the dorsal horn, including the superficial laminae. In this study, in which both the ventral and dorsal horns, and the intermediate zone, were encompassed by the WGA-HRP injection, it appeared that spinally projecting GAD+ neurons were present within both of these zones. These results would indicate that GABA reticulo- and raphe-spinal neurons may influence neurons in both the ventral and dorsal horns as well as in the intermediate zone. As observed in this study and as reported by others previously (McLaughlin et al., '75; Magoul et al., '87), GABA nerve terminals appear to provide an innervation to neurons throughout the spinal gray, including rnotoneurons, which could originate in part from supraspinal bulbar neurons in addition to spinal GABA neurons. The GABA neurons concentrated within the pars ventralis of the gigantocellular field could participate in the direct inhibition of neck motoneurons and interneurons as part of the ventral reticulospinal system (Llinas and Terzuolo, '64;Jankowska et al., '68; Peterson et al., '78). Those GABA neurons concentrated within the pars

B.E. JONES ET AL.

366 alpha of the gigantocellular fields and raphe magnus could participate in the inhibition of sensory relay neurons, and interneurons in the dorsal horn, as part of the dorsal reticulospinal system (Engberg et al., '68; Fields et al., '77; Gerhart et al., '81; Giesler et al., '81; Light et al., '86). These same GABA neurons could play an important role in sensory-motor inhibitory processes that occur during sleep, and that have been shown to depend upon neurons located in the ventral medullary reticular formation and raphe (Pompeiano, '76; Sakai et al., '81; Chase and Morales, '89; Siegel, '89; Jones, 9Ob).

GABA and glycine as inhibitory neurotransmitters of bulbospinal neurons From the results of early pharmacological studies, it appeared that glycine played a more important role than GABA in the direct inhibition of spinal motor neurons (Curtis, '69; Llinas, '64), but that GABA nonetheless had an inhibitory action upon motor neurons and interneurons, in addition to potentially acting upon primary afferent terminals, within the spinal cord (Curtis et al., '71). Recently, glycine has been identified by immunocytochemistry within neurons and/or processes in the spinal cord (van den Pol and Gorcs, '88) and in the medulla, as well as within neurons in the vestibular nuclei (Walberg et al., '90) and reticular formation (Fort et al., '90). However, glycine was also found to be co-localized with GABA within these neurons in the spinal cord and vestibular nuclei (Walberg et al., '90; Todd and Sullivan, '90) and could also be colocalized with GABA within the neurons in the ventral medullary reticular formation and raphe. Thus, from pharmacological and histochemical results, it would appear that GABA may play a role together with glycine in descending bulbospinal inhibition either by co-release from the same nerve terminals or by independent release from separate parallel processes.

SUMMARY AND CONCLUSIONS GABA-synthesizing neurons are present in significant numbers throughout the medulla, including the reticular formation, where they may serve as the source of a rich GABA innervation of reticular and raphe neurons. As potential interneurons or local projection neurons, they may exert a powerful inhibitory control over the large reticular and raphe projection neurons. As ostensibly contacted by GABA varicosities themselves, GABA neurons may also be important in processes of disinhibition. A small contingent of GABA neurons in the medulla also gives rise t o long descending projections to the cervical spinal cord, in parallel with other vestibulo-, reticulo- and raphe-spinal neurons. Accordingly, they may provide a direct inhibitory influence from these structures in what is a heterogeneous output from each. Within the reticular formation and raphe, the highest concentration of GABAspinally-projecting neurons was found in the ventral medulla in the region originally identified by Magoun and Rhines as that from which inhibition of spinal reflexes could be elicited. The GABA neurons represent one contingent of neurons within this region, the serotonin raphespinal projection neurons, with which they overlap in distribution, being another distinct population. According to their distribution and projections, these GABA projection neurons may participate in processes of reticulo- and raphe-spinal sensory and motor inhibition.

ACKNOWLEDGMENTS We would like to thank Dr. Enrico Mugnaini for kindly supplying the antibody against GAD and advice concerning its use. We express our appreciation to Charles Hodge for photographic work, and Beverley Lindsay for secretarial assistance. Elisia Rodriguez-Veiga participated in this research while on leave from Complutense University, Madrid. The work was supported by the Medical Research Council of Canada.

LITERATURE CITED 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., and H.L. Fields (1979) The origin of descending pathways in the dorsolateral funiculus of the spinal cord of the cat and r a t Further studies on the anatomy of pain modulation. J. Comp. Neural. 187t513531. Blessing, W.W. (1990) Distribution of glutamate decarboxylase-containing neurons in rabbit medulla oblongata with attention t o intramedullary and spinal projections. Neurosci. 37.171-185. Blessing, W.W., S.C. Hedger, and W.H. Oertel (1987) Vestibulospinal pathway in rabbit includes GABA-synthesizing neurons. Neurosci. Lett. 80;158-162. Bowker, R.M., and L.C. Abbott (1990) Quantitative re-evaluationof descending serotonergic and non-serotonergic projections from the medulla of the rodent: Evidence for extensive co-existence of serotonin and peptides in the same spinally projecting neurons, but not from the nucleus raphe magnus. Brain Res. 51215-25. Bowker, R.M., K.N. Westlund, M.C. Sullivan, J.F. Wilber, and J.D. Coulter (1983) Descending serotonergic, peptidergic and cholinergic pathways from the raphe nuclei: A multiple transmitter complex. Brain Res. 288:3348. Cajal, S.R. (1909) Histologie du Systeme Nerveux de 1'Homme et des Vertebrks, Val. I and 11. Paris: Maloine. Chase, M.H., and F.R. Morales (1989) The control of motoneurons during sleep. In M.H. Kryger, and T. Roth (eds): Principles and Practice of Sleep Medicine. Philadelphia: Saunders, pp. 74-85. Cho, H.J., and A.I. Basbaum (1990) GABAergic circuitry in the rostral ventral medulla of the rat and its relationship to descending antinociceptive controls. J. Comp. Neural. 303.2-14. Curtis, D.R. (1969) The pharmacology of spinal postsynaptic inhibition. In K. Akert, and P.G. Waser (eds): Progress in Brain Research, Val. 31: Mechanisms of Synaptic Transmission. Amsterdam: Elsevier, pp. 171189. Curtis, D.R., A.W. Duggan, D. Felix, and G.A.R. Johnston 11971) Bicuculline, an antagonist of GABA and synaptic inhibition in the spinal cord of the cat. Brain Res. 3269-96. Engberg, I., A. Lundberg, and R.W. Ryall(1968) Reticulospinal inhibition of transmission in reflex pathways. J. Physiol. 194;201-223. Fields, H.L., and A.I. Basbaum (1978) Brain stem control of spinal pain transmission neurons. Ann. Rev. Physiol. 4Ut193-221. 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. Fort, P., P.-H. Luppi, R. Wenthold, and M. Jouvet (1990) Neurones immunoreactifs a la glycine dans le bulbe rachidien du chat. C.R. Acad. Sci. Paris. 311:205-212. Gallyas, F., T. Gorcs, and I. Merchenthaler (1982) High-grade intensification of the end-product of the diaminobenzidine reaction for peroxidase histochemistry. J. Histochem. Cytochem. 30.183-184. Gerhart, K.D., T.K. Wilcox, J.M. Chung, and W.D. Willis (1981) Inhibition of nociceptive and nonnociceptive responses of primate spinothalamic cells by stimulation in medial brain stem. J. Neurophysiol. 45:121-136. Giesler, G.J., Jr., K.D. Gerhart, R.P. Yezierski, T.K. Wilcox, and W.D. Willis (1981) Postsynaptic inhibition of primate spinothalamic neurons by stimulation in nucleus raphe magnus. Brain Res. 204:184-188. Greene, R.W., and D.O. Carpenter (1985) Actions of neurotransmitters on pontine medial reticular formation neurons of the cat. J. Neurophysiol. 54:520-531. Hosli, L., and A.K. Tebecis (1970) Actions of amino acids and convulsants on bulbar reticular neurons. Exp. Brain Res. 1ltlll-127.

GABA NEURONS WITHIN THE MEDULLA Jankowska, E., S. Lund, A. Lundberg, and 0. Pompeiano (1968) Inhibitory effects evoked through ventral reticulospinal pathways. Arch. Ital. Biol. 106:124-140. Jones, B.E. (1990a)Immunohistochemical study of choline acetyl transferaseimmunoreactive processes and cells innervating the pontomedullary reticular formation. J. Comp. Neurol. 295t485-514. Jones, B.E. (199Ob) Paradoxical sleep and its chemicalistructural substrates in the brain. Neurosci. 4O:l-20. Jones, B.E., and T.-Z. Yang (1985) The efferent projections from the reticular formation and the locus coeruleus studied by anterograde and retrograde axonal transport in the rat. J. Comp. Neurol. 242:56-92. Krnjevic, K., and S. Schwartz (1966) Is gamma-aminobutyric acid an inhibitory transmitter? Nature. 21 1:1372-1374. Kmjevic, K., E. Puil, and R. Werman (1977) GABA and glycine actions on spinal motoneurons. Can. J. Physiol. Pharmacol. 55t658-669. Levey, A.I., J.P. Bolam, D.B. Rye, A.E. Hallanger, R.M. Demuth, M.-M. Mesulam, and B.H. Wainer (1986) A light and electron microscopic procedure for sequential double antigen localization using diaminobenzidine and benzidine dibydrocbloride. J. Histochem. Cytochem. 34:14491457. Light, A.R., E.J. Casale, and D.M. Menetrey (1986) The effects of focal stimulation in nucleus raphe magnus and periaqueductal gray on intracellularly recorded neurons in spinal laminae I and 11. J. Neurophysiol. 56555-571. Llinas, R. (1964) Mechanisms of supraspinal actions upon spinal cord activities. Pharmacological studies on reticular inhibition of alpha extensor motoneurons. J. Neurophysiol. 27: 1127-1137. Llinas, R., and C.A. Terzuolo (1964) Mechanisms of supraspinal actions upon spinal cord activities. Reticular inhibitory mechanisms on alphaextensor motoneurons. J. Neurophysiol. 27t579-591. Magoul, R., B. Onteniente, M. Geffard, 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. Neurosci. 20: 1001-1009. Magoun, H.W., and R. Rhines (1946) An inhibitory mechanism in the bulbar reticular formation. J. Neurophysiol. 9:165-171. McLaughlin, B.J., R. Berber, K. Saito, E. Roberts, and J.Y. Wu (1975) Immunocytochemical localization of glutamic decarboxylase in rat spinal cord. J. Comp. Neurol. 164:305-322. Mesulam, M.-M. (1978) Tetramethylbenzidine for horseradish peroxidase neurohistochemistry: A non-carcinogenic blue reaction-product with superior sensitivity for visualizing neural afTerents and efferents. J. Histochem. Cytochem. 26:106-117. Mesulam, M.-M., E.J. Mufson, and B.H. Wainer (1986) Three-dimensional representation and cortical projection topography of the nucleus basalis (Cb4) in the macaque: Concurrent demonstration of choline acetyltransferase and retrograde transport with a stabilized tetramethylbenzidine method for horseradish peroxidase. Brain Res. 367:301308. Millhorn, D.E., T. Hokfelt, K. Seroogy, 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.H. Oertel, A.A.J. Verhofstad, and J.-Y. Wu (1987) Immunohistochemical evidence for colocalization of gamma-aminobutyric acid and serotonin in neurons of the ventral medulla oblongata projecting to the spinal cord. Brain Res. 410:179-185. Mugnaini, E., and W.H. Oertel (1985) An atlas of the distribution of GABAergic neurons and terminals. In A. Bjorklund, and T. Hokfelt (eds): Handbook of Chemical Neuroanatomy. Vol. 4: GABA and Neuropeptides in the CNS, Part I. Amsterdam: Elsevier Science Publishers, pp. 436-608. Oertel, W.H., E. Mugnaini, D.E. Schmechel, M.L. Tappaz, and I.J. Kopin (1982) The immunocytochemical demonstration of gamma-aminobutyric acidergic neurons: Methods and application. In V. Chan-Palay, and S.L. Palay (eds): Cytochemical Methods in Neuroanatomy. New York: Alan R. Liss, pp. 297-329. Oertel, W.H., D.E. Schmechel, E. Mugnaini, M.L. Tappaz, and I.J. Kopin (1981) Immunocytochemical localization of glutamate decarboxylase in rat cerebellum with a new antiserum. Neurosci. 6:2715-2735.

367 Paxinos, G., and C. Watson (1986) The Rat Brain in Stereotaxic Coordinates. Sydney: Academic Press. Peterson, B.W., N.G. Pitts, K. Fukushima, and R. Mackel(1978)Reticulospinal excitation and inhibition of neck motoneurons. Exp. Brain Res. 32:471489. Pompeiano, 0. (1976) Mechanisms responsible for spinal inhibition during desynchronized sleep: experimental study. I n C. Guilleminault, and W.C. Dement (eds): Advances in Sleep Research, Vol. 3, Narcolepsy. New York: Spectrum, pp. 411449. Reichling, D.B., and A.I. Basbaum (1990) Contribution of brainstem GABAergic circuitry to descending antinociceptive controls. 1. GABAimmunoreactive projection neurons in the periaqueductal gray and nucleus raphe magnus. J. Comp. Neurol. 302:370-377. Rodriguez-Veiga, E., C. Holmes, and B.E. Jones (1990) Relationship of GABA neurons to reticulo- and raphe-spinal neurons in the medullary reticular formation. SOC.Neurosci. Abst. 16t851. 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 Research. 339t171-177. Rye, D.B., C.B. Saper, and B.H. Wainer (1984)Stabilization ofthe tetramethylbenzidine (TMB) reaction product application for retrograde and anterograde tracing, and combination with immunohistochemistry. J. Histochem. Cytochem. 32: 1145-1153. Sakai, K., J.-P. Sastre, N. Kanamori, and M. Jouvet (1981) State-specific neurons in the ponto-medullary reticular formation with special reference to the postural atonia during paradoxical sleep in the cat. I n 0. Pompeiano, and C. Ajmone-Marsan (eds): Brain Mechanisms and Perceptual Awareness. New York Raven Press, pp. 405429. Scheibel, M.E., and A.B. Scheibel (1958) Structural substrates for integrative patterns in the brain stem reticular core. In H.H. Jasper, and L.D. Proctor (eds): Reticular Formation of the Brain. Boston: Little, Brown & Co., pp. 31-68. Siege1 (1989) Brainstem mechanisms generating REM sleep. In M.H. Kryger, and T. Roth (eds): Principles and Practice of Sleep Medicine. Philadelphia: Saunders, pp. 104-120. Skagerberg, G., and A. Bjorklund (1985) Topographic principles in the spinal projections of serotonergic and non-serotonergic brainstem neurons in the rat. Neurosci. 15:445480. Sternberger, L.A. (1979) Immunocytochemistry. New York: John Wiley & Sons. Sugimoto, T., Itoh, K., Yukihiko, Y., Kaneko, T., and N. Mizuno (1985) Coexistence of neuropeptides in projection neurons of the thalamus in the cat. Brain Res. 347t381-384. Todd, A.J., and A.C. Sullivan (1990) Light microscope study of the coexistence of GABA-likeand glycine-likeimmunoreactivitiesin the spinal cord of the rat. J. Comp. Neurol. 296t496-505. Vacca, L.L., S.L. Rosario, E.A. Zimmerman, P.-Y.N. Tomashefsky, and K.C. Hsu (1975) Application of immunoperoxidase techniques to localize horseradish peroxidase peroxidase-tracer in the central nervous system. J. Histochem. Cytochem. 23:208-215. Valverde, F. (1961) Reticular formation of the pons and medulla oblongata.-A Golgi study. J. Comp. Neurol. I1 6:71-99. van den Pol, A.N., and T. Gorcs (1986) Synaptic relationships between neurons containing vasopressin, gastrin-releasing peptide, vasoactive intestinal polypeptide, and glutamate decarboxylase immunoreactivity in the suprachiasmatic nucleus: dual ultrastructural immunocytochemistry with gold-substituted silver peroxidase. J. Cornp. Neurol. 252507521. van den Pol, A.N., and T. Gorcs (1988) Glycine and glycine receptor immunoreactivity in brain and spinal cord. J. Neurosci. 8:47%492. Walberg, F., O.P. Ottersen, and E. Rinvik (1990) GABA, glycine, aspartate, glutamate, and taurine in the vestibular nuclei: an immunocytochemical investigation in the cat. Exp. Brain Res. 79.547-563. Wilson, V.J., and M. Yoshida (1969) Monosynaptic inhibition of neck motoneurons by the medial vestibular nucleus. Exp. Brain Res. 9:365380.