THE JOURNAL OF COMPARATIVE NEUROLOGY 295~370-384( 1990)

Choline Acetyltransferase-Immunoreactive Profiles Are Presynaptic to Primary Sensory Fibers in the Rat Superficial Dorsal Horn A. RIBEIRO-DA-SILVA AND A. CLAUD10 CUELLO Department of Pharmacology and Therapeutics, McGill University, Montreal, Quebec, Canada H3G 1Y6

ABSTRACT The specific aim of this study was to search for morphological counterparts to the known antinociceptive effects of cholinomimetic drugs a t the spinal cord level. For this, the light microscopic and ultrastructural distribution of choline acetyltransferase immunoreactivity was studied in laminae 1-111 of the rat cervical spinal cord. Immunoreactivity was present in cell bodies in lamina 111, and in dendrites and axons of all three laminae. Immunoreactive axonal varicosities were often presynaptic to the central varicosities of type I1 synaptic glomeruli in lamina I1 and lamina 111, less often presynaptic to the central elements of type I glomeruli in lamina 11, and often presynaptic to dendrites in both type I and type I1 glomeruli. In addition, immunoreactive dendrites were often postsynaptic to the central varicosities of glomeruli of all morphological types. These results indicate that 1) primary sensory fibers excite cholinergic interneurons; 2) the acetylcholine released by the axon terminals of these interneurons modulates both nociceptive and non-nociceptive sensory information a t the spinal cord level through both pre- and postsynaptic mechanisms. Furthermore, our results reinforce current ideas on reciprocal sensory interaction between thick and fine afferent fibers. K e y words: synaptic glomeruli, acetylcholine, immunocytochemistry, ultrastructure, substantia gelatinosa

For several years, cholinergic mechanisms have been known to be associated with antinociception as, for example, cholinomimetic drugs potentiate opiate analgesia (for review, see Green and Kitchen, '86). In addition, when applied intrathecally to the spinal cord surface, cholinergic agonists have antinociceptive effects (Taylor et al., '82; Yaksh et al., '85). Such effects are antagonized by atropine, but not by nicotinic antagonists, indicating that they are muscarinic in nature (Taylor et al., '82; Yaksh et al., '85). In agreement with this, high concentrations of muscarinic receptors have been detected in the superficial laminae of the spinal cord dorsal horn in several mammalian species, including the human (Wamsley et al., '81; Seybold, '85; Villiger and Faull, '85; Gillberg et al., '88; Quirion et al., '89). Although as yet very little is known of the cholinergic circuitry in this area of the central nervous system (CNS), these findings favor a contribution of acetylcholine to sensory processing. There is a wealth of information on the distribution of choline acetyltransferase (ChAT)-immunoreactive (IR) neurons and fibers in the CNS (Kimura et al., '81; Houser et al., o 1990 WILEY-LISS, INC.

'83; Cuello and Sofroniew, '84; Butcher and Woolf, '86; Kasa, '86). However, rather little is known of the contribution of cholinergic neurons to sensory processing systems. The dorsal horn of the rat spinal cord was reported to contain a prominent plexus of ChAT-IR axons and varicosities that was particularly dense in lamina I11 (Barber et al., '84; Phelps et al., '84; Borges and Iversen, '86). The axonal plexus was intermingled with dendrites from ChAT-IR neurons whose cell bodies were considered to be mainly located in laminae 111-V (Barber et al., '84). Since very little is known of the synaptic relationships of these axonal boutons and dendrites (Barber e t al., '84),we have undertaken a study of such interactions, placing emphasis on the relationship of ChAT-IR profiles to the synaptic glomeruli. Glomeruli are the most striking ultrastructural feature in Accepted December 21,1989. Address reprint requests to A. Claudio Cuello, Department of Pharmacology and Therapeutics, McGill University, Montreal, Quebec, Canada H3G 1Y6.

371

CHAT IN GLOMERULI OF R A T GELATINOSA the superficial laminae of rat dorsal horn and represent termination sites of many primary sensory neurons (Coimbra et al., '70, '74, '84; Knyihar and Gerebtzoff, '73; Ribeiroda-Silva and Coimbra, '82); therefore, we believe attention to glomerular neurochemistry is justified. Clarification of the ultrastructural neurochemistry could bring novel information on cholinergic participation in the modulation of primary sensory information, thereby providing insight into pharmacological manipulation of nociceptive input at the spinal cord level.

MATERIALS AND METHODS Eight adult male Wistar rats, 240-260 g in weight, were used in this study. They were anesthetized with Equithesin (4 ml/kg i.p.1 and perfused through the left ventricle with 15-20 ml of perfusion buffer (for composition see Connaughton et al., '86), followed by either of two fixative mixtures at room temperature: a) 500 ml of a mixture of 4% paraformaldehyde, O.lC1c glutaraldehyde, and 0.2%picric acid in 0.1 M phosphate buffer (pH 7.4), followed by 500 ml of the same fixative without glutaraldehyde; b) 1,000 ml of a mixture of 1% paraformaldehyde and 1% glutaraldehyde in 0.1 M phosphate buffer (pH 7.2-7.4) for 30 minutes. After perfusion, segments C,-C, of the cervical spinal cord were fixed by immersion in a) the 4 % paraformaldehyde/0.2 7c picric acid ' paraformaldemixture without glutaraldehyde or b) the 1% hyde/l% glutaraldehyde mixture, for 1.5-2 hours at 4°C. Subsequently, the tissue was infiltrated overnight at 4°C in 30% sucrose-0.1 M phosphate buffer, quickly frozen by immersion in liquid nitrogen, and thawed in phosphate buffer at 25OC. Transverse and parasagittal 50 pm-thick slices of the dorsal horns were obtained with a Vibratome. Prior to incubation in the antibody, the tissue sections were treated with 1 Yo sodium borohydride in phosphate buffered saline (PBS) for 30 minutes and extensively washed in PBS until all bubbling disappeared (Kosaka et al., '86).

Immunocytochemistry The slices were incubated overnight in a rat anti-ChAT monoclonal antibody (Boehringer Mannheim), which displays the same immunological characteristics of the antibody reported by Eckenstein and Thoenen ('82). The antibody was demonstrated to recognize purified ChAT in Western blot and by immunoprecipitation and gel electrophoresis studies. The anti-ChAT antibody was used diluted 1:lOO in PBS, a t 4OC. (Unless stated otherwise, all subsequent incubations were carried out a t room temperature, and PBS was used for the washes and to dilute the antibodies.) After two washes, the tissue sections were incubated for 1 hour in a rabbit anti-rat IgG antibody (raised in our laboratory), diluted 1 5 0 and given two further washes. The slices were then incubated for 2 hours in rat monoclonal anti-horseradish peroxidase (HRP) antibody (Medicorp, Canada; Cuello et al., '84). After one wash, the tissue was incubated for 2 hours in PBS containing 5 pg/ml of HRP (Sigma type VI). After three washes, the sections were treated for peroxidase with 3,3-diaminobenzidinetetrahydrochloride (DAB, Sigma) dissolved in 0.1 M phosphate buffer, pH 7.4, containing cobalt chloride and nickel ammonium sulfate (Adams, '81). After the DAB reaction, the sections were rinsed three times in phosphate buffer, and osmicated for 90 minutes in osmium tetroxide in 0.1 M phosphate buffer at 4°C. Some sections were block stained for 45

minutes in a 1.5% aqueous solution of uranyl acetate. Afterwards, all sections were dehydrated in ascending alcohols and flat-embedded in Epon. After selecting the appropriate fields, the sections were trimmed and reembedded. Ultrathin sections were obtained with a Reichert Ultracut microtome, collected on Formvar-coated one-slot grids, and observed on a Philips 410 electron microscope. Some of the grids had been counterstained with uranyl acetate and lead citrate. To assess the relative frequency of each type of ChAT-IR profile, counts were made on the EM screen a t a magnification of x 12,600. This involved examining 12 ultrathin sections: one from each of three blocks from four animals perfused with the higher glutaraldehyde concentration. These sections were obtained from blocks trimmed to include the 500 pm segment of the mediolateral extent of laminae I, I1 and 111, which was equidistant from both medial and lateral edges of the dorsal horn. The laminae were identified according to the criteria established by Ribeiro-da-Silva and Coimbra ('82). All immunostained dendrites and axonal boutons were counted and their synaptic relationships identified. The results are expressed as percentage per type (dendrite or bouton) and per lamina (Table 1). In control sections which were incubated with normal rat serum instead of the primary antibody, no immunostaining was ohserved.

RESULTS Light microscopy The light microscopic distribution of ChAT-IR profiles was best observed in the flat-embedded material from rats fixed with the paraformaldehyde/glutaraldehyde/picricacid mixture. A prominent network of very fine fibers with small varicosities could he seen in the dorsal half of lamina 111and ventralmost third of lamina I1 (Figs. la,b, 2). The density of the fiber network was lower in the ventral half of lamina 111, in the outer two-thirds of lamina TI and in lamina I (Fig. la). Immunoreactive cell bodies could be seen primarily in laminae I11 and IV (Fig. 1) but also in lamina V. The morphology of the cells in laminae I11 and IV could best be appreciated in parasagittal slices (Fig. 2). They were multipolar neurons, with average cell body dimensions in parasagittal sections of 25 x 15 pm. Frequently, the main dendrites TABLE 1. Number of CUT-IR Axonal Boutons and Dendrites in Laminae I-111 of the Rat CervicalDorsal Horn' LI ~~~~

~

~

LIIA

LIIB

LIIl

69 69 (100)

275 214 (77.8)

514 374 (72.7)

0

61 (22.2)

140 (27.2)

0

21 (7.6)

0

27 (9.8)

~

Axonal boutom itnial) Nonglomerular boutnns vz profile3 in glomeruli (total) v2profilesin type I glomeruli v2profiles in subtype IIaglomed V2profilea in subtype IIb glomeruli Dendrites (total] Nonglomerular dendrites Glomeruh dendrites (total) Dendritic profiles in type I glomeruli Dendritic profiles in subtype Ilaglomeruli Dendritic profiles in subtype IIb glomeruli

0

39 (7.6) 101 (19.6)

0 3 3 (100)

0

13 (4.6)

9 9 (100)

77 39 (50.6)

69

0

0

38 (49.4)

25 (36.2)

0

27 (35.1)

0

8 (10.4)

15 (21.7)

0

3 (3.9)

10 (14.5)

44 (63.8)

0

'Values in parenthesas represent percentages of the total number of dendrites or boutom per lamina Total number of profilescounted = 1,070.

A. RIBEIRO-DA-SILVA AND A.C. CUELLO

372

Fig. 1. Light microscopic distribution of ChAT immunoreactivity in the superficial laminae of the rat cervical dorsal horn as observed in 50 pm-thick Epon flat-embedded coronal slices. a: Fibers with varicosities (small arrows) are seen in all laminae, but are mostly concentrated in the deeper part of ventral lamina I1 (LIIB) and in the dorsal half of lamina I11 (LIII); an immunoreactive neuron (large arrow) is present at the

lamina 111-IV transition. b: At higher magnification, a presumptive axon with varicosities (small arrows) is seen to originate from a dendritic trunk of a ChAT-immunoreactive (ChAT-IR) neuron of lamina 111. Large arrows point to ChAT-IR cell bodies. LT = Lissauer’s tract. Scale bars = 20 rm.

were oriented towards lamina 11, and it was not uncommon to find a dendrite coursing longitudinally for a considerable distance along lamina I1 (Fig. 2). Three to four ChAT-IR cell bodies could be observed in laminae 111-IV in a single 50 Fm-thick parasagittal slice.

either round or pleomorphic synaptic vesicles (Fig. 4b) and were very frequently components of synaptic glomeruli (see below). Some ChAT-IR varicosities were presynaptic to a cell body, although such contacts were rare. The immunostained axonal profiles in lamina I and outer two-thirds of lamina 11, where the density of immunostaining was lower than in the inner third of lamina I1 and outer half of lamina 111, were very similar to those described above (Fig. 5a,b; see also Fig. 8). However, the ChAT-IR profiles established axosomatic synapses more often in lamina I (Figs. 5a,b) and outer lamina I1 (lamina IIA) than in deeper laminae. The size of the immunostained varicosities measured from 0.6 x 0.4 Fm to 1.0 x 0.6 pm. Ultrastructural analysis of the ChAT-IR cell bodies in laminae 111-IV (Fig. 6a,b) revealed an ovoid nucleus with infolding of the nuclear envelope, abundant free ribosomes, sparse cisterns of rough endoplasmic reticulum, and several dictyosomes of the Golgi apparatus. Although ChAT reaction product was diffusely distributed in the cytoplasm, it was more concentrated in areas rich in free ribosomes (Fig. 6b). The Golgi apparatus seemed devoid of immunostaining (Fig. 6a,b). ChAT immunoreactivity occurred in numerous

Electron microscopy Non-glomerular profiles. The examination of ultrathin serial sections confirmed that the immunoreactive fiber plexus in deep lamina I1 and lamina I11 consisted mainly of axonal varicosities connected by unmyelinated axons (Figs. 3, 4a-c). The ChAT-IR axonal boutons contained numerous pleomorphic agranular synaptic vesicles (Fig. 4) and, occasionally, one or two dense-core vesicles (Figs. 3, 4a). The shape of the varicosities varied; some were dome-shaped (Fig. 3 ) and others had a more irregular contour (Fig. 4a). The immunoreactive boutons established symmetric or marginally asymmetric synaptic contacts with a single dendrite (Figs. 3,4a,b) and less often with two (Fig. 412). Clearly asymmetric synaptic contacts were also detected but rarely (Fig. 4d). ChAT-IR boutons were often apposed to nonimmunoreactive axonal profiles containing

373

CHAT IN GLOMERULI OF RAT GELATINOSA

Fig. 2. Photomontage and camera lucida drawing of a lamina 111ChAT-TR neuron, depicting the caudal course of a very long dendrite along lamina I1 (arrowheads).Parasagittal slice. Scale har = 20 p m .

dendritic profiles, particularly those of inner lamina I1 (lamina IIB) and lamina 111. The immunoprecipitate in dendrites was concentrated over microtubules and neurofilthe inner surface Of the plasma as as membrane (Fig. 6c). Profiles in synaptic glomeruli. The synaptic glomeruli could be recognized readily in the immunostained prepsrations. described previous~yby one of the authors (Ribeiro-da-Silva and Coimbra, '82), such glomeruli were found to be of two populations: type I glomeruli, which prevailed in the middle third of lamina 11, had central

varicosities (C,) of sinuous contour which were poor in mitochondria and had dense axoplasm filled with densely packed synaptic vesicles p i g . 7a); type 11 glomeruli ( ~ i ~ ~ . 7b, 8a,b) which had larger, rounder and lighter central varicosities (C1J richer in mitochondria, prevailed in the Part Of lamina 'I and in lamina 'I1* Two varieties of type I1 glomeruli could be identified, depending on the ~ I S W K X (subtype IIa-FigS. 7 h 8a) or Presence (subtype IIh-Fig. 8b) of neurofilament bundles in the centralvaricosity.

374

A. RIBEIRO-DA-SILVA AND A.C. CUELLO

Fig. 3. Low magnification electron micrograph of five ChAT-IR axonal varicosities (B) in deep lamina I1 that are apposed to dendrites (D). Two of the ChAT-IR boutons are presynaptic to the dendrites (D), forming symmetric contacts (arrowheads). Immunostaining is also present in unmyelinated axons (arrows). Scale bar = 1 pm.

Immunoreactive dendrites were observed to be postsynaptic to central varicosities of both type I (Fig. 7a) and type I1 (Fig. 7b) synaptic glomeruli (Table 1). Immunoreactive axonal varicosities containing pleomorphic synaptic vesicles which filled the entire terminal were apposed to C,, profiles frequently (Figs. 8a,b, 9, 10a) and to C, profiles less fre-

quently (Fig. lob; Table 1).They were often presynaptic to the central varicosity and to glomerular dendrites (D) (Figs. 8b, 9, 10a,b). The synaptic contacts were of the symmetric type. Furthermore, analysis of serial sections revealed the frequent occurrence of a synaptic triadic arrangement in which a single immunolabeled profile was presynaptic to

Fig. 4. Features of the ChAT-IR axonal profiles in lamina 111. a: Two ChAT-IR varicosities (B) are connected by an unmyelinated axon (arrow). b A dome-shaped ChAT-IR varicosity (B) with agranular synaptic vesicles estahlishes a symmetric synaptic contact (arrowhead) with a dendrite (D) which also receives synapses from a nonimmunostained varicosity containing round synaptic vesicles (asterisk).

c: A ChAT-IR axonal bouton (B) is presynaptic t o two dendrites (D), the

synapse with the dendrite on the left appearing slightly asymmetric. d Example of an immunoreactive axonal varicosity (B) establishing an asymmetric contact (arrowhead) with a dendrite (D). Arrowheads point to synapses. Scale bars = 0.5 pm.

376

A. RIBEIRO-DA-SILVA AND A.C. CUELLO

Fig. 5. Axosomatic contacts of two ChAT-IR varicosities (B1and B2) on a lamina I nerve cell body (S). b, which is an adjacent section to the one shown in a,demonstrates a t higher magnification t h at both B1 and B2 establish symmetric synapses on the cell body (arrows), although B2 is also presynaptic (arrowhead) to a dendrite (D). Scale bars = 1Fm.

both the central varicosity and a dendrite postsynaptic to the central bouton (Fig. 9). However, such an arrangement could be detected even in isolated sections (Fig. 10a). These immunolabeled varicosities were identical in size, density of synaptic vesicles, and synaptic connections to other varicos-

ities in glomeruli which were not immunolabeled. In respect of their ultrastructural features and synaptic connections, both the immunostained and the non-immunostained glomerular peripheral axonal boutons belonged to the V, type of Coimbra and collaborators ('74). Counts of V, profiles in

CHAT IN GLOMERULI OF RAT GELATINOSA

Fig. 6. ChAT immunoreactivity in cell bodies and dendrites. a: General features of a ChAT-IR perikaryon in lamina 111include a deeply infolded nuclear envelope (arrows), sparse rough endoplasmic reticulum, and well-developed Golgi apparatus (G). b Enlargement of a portion of Figure 6a reveals that the DAB reaction product is particu-

377

larly concentrated in localized areas of the cytoplasm (arrowheads)that are rich in free ribosomes. c: An immunoreactive dendrite (arrow) is postsynaptic to a nonimmunoreactive varicosity (asterisk).Scale bars = 0.5pm.

A. RIBEIRO-DA-SILVA A N D A.C. CUELLC

378

Fig. 7. ChAT immunoreactivity in dendrites of synaptic glomeruli. a: In the middle third of lamina 11, an immunoreactive dendrite (arrow)

is postsynaptic to the central varicosity of a type I synaptic glomerulus

glomeruli revealed that approximately one-half of their total number were immunostained for ChAT (423 immunostained and non-immunostained V? profiles counted in type 1, IIa, and IIb glomeruli in the same ultrathin sections used to obtain Table 1). Furthermore, the immunostained V, represented approximately one quarter of the immunolabeled varicosities in ventral lamina I1 and lamina I11 (Table 1).

DISCUSSION The light microscopic findings of this study concur with previous observations made in the same species (Barber et al., '84; Borges and Iversen, '86). The new observations emerging from this study are those obtained at the ultrastructural level, viz., the detection of ChAT-IR dendrites in both type I and type I1 synaptic glomeruli and ChAT-IR presynaptic peripheral axons (V, profiles) in synaptic glomeruli. Furthermore, such V, axons were found to be part of triadic type synaptic arrangements.

Origin of ChAT-IR profiles The use of a double metal intensification of the peroxidase reaction product according to Adams ('81) allowed good visualization of cell bodies and processes in osmicated, flat-embedded 50 /*m-thick tissue slices. The origin of the dendritic processes in laminae I1 and 111from cells located in laminae 111-IV could sometimes be observed. Therefore, it can be speculated that the immunostained dendritic

((2,). b A similarly immunoreactive dendritic profile (arrow) is postsynaptic to the central bouton of a subtype IIa glomerulus (C,,J. Scale bars = 0.5 pm.

branches in laminae 1-111 all have a similar origin. However, the source of the fine terminal arborization of axons in deep lamina I1 and lamina I11 was not clear. A primary sensory origin for these ChAT-positive axons can, nonetheless, be excluded since ChAT has never been observed in the dorsal root ganglia (Barber e t al., '84). It also seems unlikely that the stained axons originate from descending fibers because spinal transections do not affect ChAT activity caudal to the lesion (Kanazawa et al., '79). Based on these findings, and the observation of ChAT-IR fibers in the dorsolateral funiculus, as also reported by Barber and colleagues ('84), most, if not all, the immunoreactive axonal varicosities probably come from intrinsic spinal cord neurons located either in the same or neighboring segments.

C U T - I R V,-profiles in glomeruli Although most ChAT-IR profiles in our electron microscopic preparations were non-glomerular, a considerable number of them were part of synaptic glomeruli in ventral lamina I1 and lamina 111 (Table 1).The significance of this finding is related to the fact that glomeruli represent important prototypes for the study of synaptic interactions in the dorsal horn. The central axonal varicosity of glomeruli is most likely of exclusive primary sensory origin as it has been demonstrated that normal synaptic glomeruli disappear from the substantia gelatinosa following dorsal rhizotomy (Duncan and Morales, '78; Coimbra et al., '84; Murray and Goldberger, '86). The morphological classification of

Fig. 8. Electron micrographs of elements in lamina IIB. a: In the middle third of lamina 11, an immunoreactive axon (V,) is apposed to the central varicosity (CIIJ of a subtype IIa synaptic glomerulus (recognizable by the light matrix of the C,,, and occurrence of three V, profiles). The immunoreactive V, is continuous with a nonglomerular varicosity (B) via a short connecting segment. On the left side of the figure, a second nonglomerular axonal varicosity (B) also displays ChAT immunoreactivity. b: In the ventralmost part of lamina IT, a ChAT-IR

glomerular V, is observed to be presynaptic (small arrows) to both the central varicosity of a subtype IIb glomerulus (CIIb)and to a glomerular dendrite (D). Note t h e neurofilament bundle (nf) in the C,,, profile, which distinguishes it from the non-neurofilamentous C,,,. An immunoreactive dendrite (arrow) is postsynaptic to the central varicosity of a type I glomerulus (C,), and is also apposed by an immunoreactive varicosity (B). Asterisks indicate nonimmunoreactive V, profiles. Scale bars = 1pm.

380

A. RIBEIRO-DA-SILVA AND A.C. CUELLO

Fig. 9. Immunostained V, profiles from glomeruli in lamina I11 as revealed in serial sections. a: One immunoreactive V, profile (V,,,,) is apposed to both C,,,and C,,,but ispresynaptic (small arrows) only to the neurofilament (nf)-containing C,,, and to a dendrite (U). b and c demonstrate that D is also postsynaptic to C,,, (small arrows), thereby establishing a triadic arrangement (see text). b and c also show that V2(L,

and V,,,, are in continuity via a short connecting segment (large arrow) and that V2,2,is presynaptic to C,,, at a long synaptic contact (small arrow in 9c). In d, which reveals a level somewhat removed from the series, C, is recognizable but all irnmunoreactive V, have disappeared from the plane of section. However, the axon from which V,!,, and V,(,, originate can be identified (arrow).Scale bar for all micrographs = 1pm.

synaptic glomeruli into types according to their ultrastructural features has been carried out in rat (Ribeiro-da-Silva and Coimbra, '82; Ribeiro-da-Silva et al., '85) and monkey (Knyihar-Csillik e t al., '82a,b). In the rat, small, sinuous, and dense central varicosities of type I glomeruli (C,) have been shown to be capsaicin-sensitive (Ribeiro-da-Silva and Coimbra, '84),and are therefore probably of unmyelinated

origin. Furthermore, C, boutons have been shown to cnnstitute a heterogeneous population, in terms of both morphology and neurochemistry. Most were poor in large granular vesicles (LGV) and contained fluoride resistant acid phosphatase activity although up to 20% were rich in LGV, had a simpler synaptic architecture and were immunoreactive for certain peptides (Ribeiro-da-Silvaet al., '86, '89; Riheiro-da-

CHAT IN GLOMERULI OF RAT GELATINOSA

381

Fig. 10. Synaptic contacts of immunoreactive V, profiles in glomeruli. a: In a small type IIb glomerulus of lamina 111,a ChAT-IR V, profile is seen t o establish a triadic arrangement with the C, and a dendrite. Small arrows indicate synapses. b: In a type I glomerulus of the middle

third of lamina 11, an immunoreactive V, profile is shown to be presynaptic (small arrow) to the central varicosity (CJ. Scale bars = 0.5 um.

Silva and Cuello, '87). In contrast, the larger, less sinuous central boutons (CJ of type TI glomeruli, which were richer in mitochondria, were interpreted to be of myelinated sensory origin (Ribeiro-da-Silva et al., '85). Like C, profiles, they were shown to comprise a heterogeneous population, although only morphological elements were considered for this subdivision. On this basis, C,, boutons have been divided into two subtypes: C,,, (without neurofilaments) and CIlb (with neurofilament bundles) (Coimbra et al., '84; Ribeiro-da-Silva et al., '85). Analysis of our material revealed that ChAT-IR v, profiles were considerably more frequent around C,, than C, profiles. This is not surprising since the numbers of V, detected around C,, in conventional, non-immunostained morphological preparations were also found to be greater (Ribeiro-da-Silva and Coimbra, '82; Ribeiro-da-Silva et al., '85). Within the two varieties of C,l, the neurofilamentous Cllh had synaptic triads involving immunoreactive V, profiles more frequently than did the CIIa.Again, this could be predicted because this synaptic arrangement has been shown to prevail in the CIIhvariety (Ribeiro-da-Silva et al., '85). The modalities of sensations subserved by type I1 glomeruli are as yet unknown because studies combining intracellular recording and labeling of primary sensory fibers have not been undertaken in the rat. However, since type I1 glomeruli predominate in the ventralmost part of lamina I1 and in lamina 111, which are areas of termination of non-nociceptive afferents (for review, see Brown, '82f, these glomeruli may participate in the modulation of non-nociceptive sensory information. Furthermore, the morphology of terminals of non-nociceptive afferent fibers of the D-hair A delta type (Rethelyi et al., '82) and of thicker afferents (Ralston et al., '84) in the cat dorsal horn was very similar to that of type I1 glomeruli in the rat. This

has been substantiated by a preliminary rat study in which the H R P labeling of sensory fibers demonstrated that some afferent fibers exhibiting the typical flame-shaped arborization of hair follicle afferents terminated as C,,, glomerular profiles (Cruz et al., '87a,b). The ChAT-IR V, profiles observed were identical morphologically to the axonal profiles which were shown to contain GAD immunoreactivity in the rat dorsal horn (McLaughlin et al., '75; Barber et al., '78) and in the substantia gelatinosa of the cat trigeminal nucleus (Basbaum et al., '86) as well as to take up radiolabeled GABA in the rat dorsal horn (Ribeiro-da-Silvaand Coimbra, '80). A coexistence of GABA and ChAT has recently been demonstrated in cells of several areas of the CNS including the dorsal horn of the spinal cord (Kosaka et al., '88). Since only one-half of the V, profiles were immunoreactive for ChAT and no information is available on the number of GAD positive V, boutons, we cannot predict whether a coexistence of both substances occurs. Although we can anticipate their coexistence in some of these terminals, only a double labeling study can clarify this point.

The significance of ChAT-IR profiles in the dorsal horn Almost without exception, the ChAT-IR profiles described in this study contained pleomorphic synaptic vesicles and established symmetric synapses. Rapisardi and Lipsenthal ('84) reported variations in the thickness of the postsynaptic density after serial section examination of synapses in the dorsal lateral geniculate nucleus of the cat and monkey. These variations indicated that synapses can be erroneously classified after examination of single sec-

A. RIBEIRO-DA-SILVA A N D A.C. CUELLO

382

I

I I I

I I I I I I

I

I I

I II

u

c

-

J

I I

I I I

I 5 ; iI

CHAT IN GLOMERULI OF RAT GELATINOSA tions. However, in our material, only a few examples of clearly asymmetric contacts were found, even after serial section examination. This confirmed that the great majority of ChAT-IR profiles we studied really established symmetric contacts and had pleomorphic vesicles. As reported in the classical studies of Gray ('61) and Uchizono f'65), such morphological features suggest an inhibitory action. Most of the cholinergic synapses we detected may indeed be inhibitory; for example, intracellular studies on spinal neurons by Zieglgansberger and Reiter ('74) demonstrated a potentiation of evoked inhibitory postsynaptic potentials by acetylcholine. In addition, a participation of acetylcholine in antinociception has been reported, based on physiological and pharmacological data (for review see Green and Kitchen, '86). Furthermore, muscarinic acetylcholine receptors, some of which seem to be associated with inhibition (see, e.g., Kelly and Rogawsky, '85; Bonner, '89), have been detected in high amounts in the substantia gelatinosa (see Introduction). The hypothesis that such inhibitory effects are exerted a t the spinal level is supported by the demonstration of an anti-nociceptive effect of cholinomimetic drugs when applied intrathecally to the spinal cord surface (Taylor et al., '82; Yaksh et al., '85). Consequently, our morphological findings are compatible with a role for cholinergic neurons in antinociceptive circuits a t the spinal level. The origin of the dendritic profiles contacted by ChAT-IR varicosities was not identified. Unfortunately, there is no suitable morphological marker for these dendrites, contrasting with the central varicosities of glomeruli which are of primary sensory origin (see above). It is, however, probable that most of these dendritic profiles belong to interneurons or projecting neurons of the dorsal horn. Determining the origin of non-glomerular axons with round synaptic vesicles which were postsynaptic to an immunostained varicosity (a finding limited to inner lamina I1 and lamina 111) is also problematic. However, serial section analysis of these postsynaptic axons demonstrated that they often represented a tangential section of the central bouton of a synaptic glomerulus. In reality, the number of varicosities corresponding to glomerular V, profiles may be higher than was detected since many of them cannot be identified as such in single sections. The finding that primary sensory fibers in glomeruli were synaptic targets for ChAT-IR profiles is compatible with the decrease in muscarinic receptor binding observed in the superficial dorsal horn after dorsal rhizotomy (Gillberg and Wiksten, '86) and suggests a cholinergic presynaptic modulation of primary sensory neurons.

Fig. 11. Diagram of the hypothesized cholinergic synaptic circuits in the superficial dorsal horn (see Discussion). Three distinct types of primary sensory fibers (1, C; 2, A-delta; and 3, A-beta) are shown to terminate in central varicosities of synaptic glomeruli, types I (C,), IIa (C,,,), and IIb (C,,,), respectively. Each central glomerular varicosity is presynaptic a t an asymmetric contact to dendrites of a projection neuron (b) or t o an excitatory neuron ( c ) contacting a projection neuron (a) and to a C h h T + interneuron (in heavy outline) situated in lamina 111. T h e axons of the ChAT+ cells are presynaptic both to the central elements of glomeruli (CI, C,,,, and C,,,) and to t h e dendrites of cells linked to the ascending sensory pathway forming a triadic arrangement as described in the text. For the sake of simplicity, only one ChATimmunoreactive (ChAT +) neuron is shown, and only one of t h e primary sensory fibers (C fiber) is represented terminating in nonglomerular varicosities (although the latter are known to represent t h e main form of termination of primary sensory fibers).

383 Our observation that half of the immunostained dendritic profiles were part of synaptic glomeruli in ventral lamina I1 and in lamina I11 (Table 1) allows us t o postulate that the identified cholinergic interneurons in laminae 111-IV, whose dendrites extend to lamina 11, receive sensory information there from either type I or type I1 central varicosities (Fig. 11). The axons of these ChAT-IR neurons would, in turn, inhibit the transmission of primary sensory impulses in glomeruli of types I, IIa, and IIb through both pre- and postsynaptic mechanisms (Fig. 11). Such an arrangement would concur with the physiologically based pain theories (Melzack and Wall, '65; Cervero and Iggo, '78). However, the higher number of V, profiles in type I1 glomeruli (presumed to be non-nociceptive) compared to type 1glomeruli (probably nociceptive), favors a greater role for cholinergic mechanisms in the modulation of innocuous impulses than of nociceptive ones. This correlates well with the preponderance of ChAT immunoreactivity in areas not associated with nociception (ventralmost lamina I1 and lamina 111). However, the fact that most of the ChAT-IR varicosities described in this study are probably unrelated to nociception should not detract from the finding that some are. A cholinergic inhibition of nociception through pre- and postsynaptic mechanisms in type I glomeruli is worthy of consideration. This may explain the anti-nociceptive actions of muscarinic agonists a t the spinal level and may provide a basis for the pharmacological manipulation of nociception by the use of cholinomimetic agents.

ACKNOWLEDGMENTS This research study has been supported by grants from the NIH (USA), grant NS26415; Office of the Dean, Faculty of Medicine (McGill University); and Medicorp (Canada). Dr. A. Ribeiro-da-Silva acknowledges a grant from the Gulbenkian Foundation, Lisbon, Portugal. We thank Drs. Paul B.S. Clarke and Paul A. Lapchak for helpful suggestions. We are particularly grateful to Dr. Erik P. Pioro and to Dr. Margaret R. Matthews for revising the manuscript and to Ms. Diane Leggett and Ms. Rosa Maria Greco for editorial and secretarial assistance. We thank Ms. Kathy Hewitt, Mr. Sylvain CBtB, and Mr. Alan Forster for technical and photographic assistance.

LITERATURE CITED Adams, J.C. (1981) Heavy metal intensification of DAB-based HRP reaction product. J. Histochem. Cytochem. 29:775. Barber, R.P., P.E. Phelps, C.R. Houser, G.D. Crawford, P.M. Salvaterra,and J.E. Vaughn (1984) The morphology and distribution of neurons containing choline acetyltransferase in the adult rat spinal cord: An immunocytochemical study. J. Comp. Neurol. 229:329-346. Barber, R.P., J.E. Vaughn, K. Saito, B.J. McLaughlin, and E. Roberts (1978) GABAergic terminals are presynaptic to primary afferent terminals in the substantia gelatinosa of the rat spinal cord. Brain Res. 141:35-55. Basbaum, A.I., E.J. Glazer, and W. Oertel (1986) Immunoreactive glutamic acid decarboxylase in the trigeminal nucleus caudalis of the cat: A lightand electron-microscopicanalysis. Somatosens. Kes. 4%-94. Bonner, T.I. (1989) The molecular basis of muscarinic receptor diversity. Trends Neurosci. 12:148-151. Borges,L.F., and S.D. Iversen (1986) Topography of choline acetyltransferase immunoreactive neurons and fibers in the rat spinal cord. Brain Res. 362,140-148. Brown, A.G. (1982) The dorsal horn of the spinal cord. Q. J. Exp. Physiol. 67:193-212. Butcher, L.L., and N.J. Woolf (1986) Central cholinergicsystems: Synopsis of anatomy and overview of physiology and pathology. In A.B. Scheibel and

384 A.F. Wechsler (eds): The Biological Substrates of Alzheimer's Disease. Orlando, Florida: Academic Press, pp. 73-86. Cervero, F., and A. Iggo (1978) Reciprocal sensory interaction in the spinal cord. J. Physiol. (Lond.) 284:84P-85P. Coimbra, A., M.M. Magalhaes, and B.P. SodrB-Borges (1970) Ultrastructural localization of acid phosphatase in synaptic terminals of the substantia gelatinosa Rolandi. Brain Res. 22142-146. Coimbra, A., B.P. SodrB-Borges, and M.M. Magalhles (1974) The substantia gelatinosa Rolandi of the rat. Fine structure, cytochemistry (acid phosphatase) and changes after dorsal root section. J. Neurocytol. 3:199-217. Coimbra, A,, A. Ribeiro-da-Silva, and D. Pignatelli (1984) Effects of dorsal rhizotomy on the several types of primary afferent terminals in laminae 1-111 of the rat spinal cord. Anat. Embryol. (Berl.) 170:279-287. Connaughton, M., J.V. Priestley, M.V. Sofroniew, F. Eckenstein, and A.C. Cuello (1986) Inputs to motoneurones in the hypoglossal nucleus of the ratr Light and electron microscopic immunocytochemistry for choline acetyltransferase,substance P and enkephalins using monoclonal antibodies. Neuroscience 17:205-224. Cuello, A.C., C. Milstein, B. Wright, S. Bramwell, J.V. Priestley, and J. Jarvis (1984) Development and application of a monoclonal rat peroxidaseantiperoxidase (PAP) immunocytochemical reagent. Histochemistry 80: 257-261. Cuello, A.C., and M.V. Sofroniew (1984) The anatomy of CNS cholinergic neurons. Trends Neurosci. 734-78. Cruz, F., D. Lima, and A. Coimbra (1987a) Combined light and electron microscopic demonstration of synaptic endings in the substantia gelatinosa after HRP labeling. Neuroscience 22:S715. Cruz, F., D. Lima, and A. Coimbra (1987b) Several morphological types of terminal arborizations of primary d e r e n t s in laminae 1-11 of the rat spinal cord, as shown after HRP labeling and Golgi impregnation. J. Comp. Neurol. 261~221-236. Duncan, D., and R. Morales (1978) Relative numbers of several types of synaptic connections in the substantia gelatinosa of the cat spinal cord. J. Comp. Neurol. 182:601-610. Eckenstein, F., and H. Thoenen (1982) Production of specific antisera and monoclonal antibodies to choline acetyltransferase: Characterization and use for identification of cholinergic neurons. EMBO J. 1:363-368. Gillberg, P.-G., R. d'Argy, and S:M. Aquilonius (1988) Autoradiographic distribution of [3H]acetylcholine binding sites in the cervical spinal cord of man and some other species. Neurosci. Lett. 90:197-202. Gillberg, P.-G., and B. Wiksten (1986) Effects of spinal cord lesions and rhizotomies on cholinergic and opiate receptor binding sites in rat spinal cord. Acta Physiol. Scand. 126:575-582. Gray, E.G. (1961) The granule cells, mossy synapses and Purkinje spine synapses of the cerebellum: Light and electron microscope observations. J. Anat. (Lond.) 95345-356. Green, P.G., and I. Kitchen (1986) Antinociception opioids and the cholinergic system. Prog. Neurobiol. 261119-146. Houser, C.R., G.D. Crawford, R.P. Barber, P.M. Salvaterra, and J.E. Vaughn (1983) Organization and morphological characteristics of cholinergic neurons: An immunocytochemical study with a monoclonal antibody to choline acetyltransferase. Brain Res. 266~97-119. Kanazawa, I., D. Sutoo, I. Oshima, and S. Saito (1979) Effect of transection on choline acetyltransferase, thyrotropin releasing hormone and substance P in the cat cervical spinal cord. Neurosci. Lett. 13:325-330. Kasa, P. (1986) The cholinergic systems in brain and spinal cord. Prog. Neurohiol. 26:211-272. Kelly, J.S., and M.A. Rogawsky (1985) Acetylcholine. In M.A. Rogawsky and J.L. Barker (eds): Neurotransmitter Actions in the Vertebrate Nervous System. New York: Plenum Press, pp. 143-197. Kimura, H., P.L. McGeer, J.H. Peng, and E.G. McGeer (1981) The central cholinergic system studied by choline acetyltransferase immunohistochemistry in the cat. d. Comp. Neurol. 200:151-201. Knyihar, E., and M.A. Gerebtzoff (1973) Extra-lysosomal localization of acid phosphatase in the spinal cord of the rat. Exp. Brain Res. 18:383-395. Knyihar-Csillik, E., B. Csillik, and P. Rakic (1982a) Ultrastructure of normal and degenerating glomerular terminals of dorsal root axons in the substantia gelatinosa of the rhesus monkey. J. Comp. Neurol. 210:357375. Knyihar-Csillik, E., B. Csillik, and P. Rakic (1982b) Periterminal synaptology of dorsal root glomerular terminals in the substantia gelatinosa of the spinal cord of the rhesus monkey. J. Comp. Neurol. 210:376-399. Kosaka, T, I. Nagatsu, J.-Y. Wu, and K. Hama (1986) Use of high concentrations of glutaraldehyde for immunocytochemistry of transmitter-

A. RIBEIRO-DA-SILVA AND A.C. CUELLO synthesizing enzymes in the central nervous system. Neuroscience 18:975990. Kosaka, T., M. Tauchi, and J.L. Dahl (1988) Cholinergic neurons containing GABA-like and/or glutamic acid decarboxylase-like immunoreactivities in various brain regions of the rat. Exp. Brain Res. 70505417. McLaughlin, B.J., R. Barber, K. Saito, E. Roberts, and J.Y. Wu (1975) lmmunocytochemical localization of glutamate decarboxylase in rat spinal cord. J. Comp. Neurol. 164:305-322. Melzack, R., and P.D. Wall (1965) Pain mechanisms: A new theory. Science 150:971-979. Murray, M.. and M.E. Goldberger (1986) Replacement of synaptic terminals in lamina I1 and Clarke's nucleus after unilateral dorsal rhizotomy in adult cats. J. Neurosci. 6:3205-3217. Phelps, P.E., R.P. Barber, C.R. Houser, G.D. Crawford, P.M. Salvaterra, and J.E. Vaughn (1984) Postnatal development of neurons containing choline acetyltransferase in rat spinal cord An immunocytochemical study. J. Comp. Neurol. 229347-361. Quirion, R., D. Araujo, W. Regenold, and P. Boksa (1989) Characterization and quantitative autoradiographic distribution of [3H]acetylcholine muscarinic receptors in mammalian brain. Apparent labelling of an MJike receptor sub-type. Neuroscience 29:271-289. Ralston, H.J. 111, A.R. Light, D.D. Ralston, and E.R. Perl (1984) Morphology and synaptic relationships of physiologically identified low-threshold dorsal root axons stained with intra-axonal horseradish peroxidase in the cat and monkey. J. Neurophysiol. 51:777-792. Rapisardi, S.C., and L. Lipsenthal (1984) Asymmetric and symmetric synaptic junctions in the dorsal lateral geniculate nucleus of cat and monkey. J. Comp. Neurol. 224:415-424. RBthelyi, M., A.R. Light, and E.R. Perl (1982) Synaptic complexes formed by functionally defined primary afferent units with fine myelinated fibers. J. Comp. Neurol. 207:381-393. Ribeiro-da-Silva, A., and A. Coimbra (1980) Neuronal uptake of I3H]GABA and [3H]glycinein laminae 1-111 (substantia gelatinosa Rolandi) of the rat spinal cord. An autoradiographic study. Brain Res. 188:449464. Ribeiro-da-Silva, A., and A. Coimbra (1982) Two types of synaptic glomeruli and their distribution in laminae 1-111 of the rat spinal cord. J. Comp. Neurol. 209:176-186. Ribeiro-da-Silva, A,, and A. Coimbra (1984) Capsaicin causes selective damage to type I synaptic glomeruli in rat substantia gelatinosa. Brain Res. 290:380-383. Ribeiro-da-Silva, A,, D. Pignatelli, and A. Coimbra (1985) Synaptic architecture of glomeruli in superficial dorsal horn, as shown in serial reconstructions. J. Neurocytol. 14~203-220. Ribeiro-da-Silva. A,, J.M. Castro-Lopes, and A. Coimbra (1986) Distribution of glomeruli with FRAP-containing terminals in the substantia gelatinosa of the rat. Brain Res. 377~32S329. Ribeiro-da-Silva, A., and A.C. Cuello (1987) Substance P- and somatostatinlike immunoreactivities in synaptic glomeruli of the substantia gelatinosa, as revealed by bi-specific monoclonal antibodies. In J.L. Henry, R. Couture, A.C. Cuello. G. Pelletier, R. Quirion, and D. Regoli (eds): Substance P and Neurokinins. New York: Springer-Verlag, pp. 313-317. Ribeiro-da-Silva, A., P. 'l'agari. and A.C. Cuello (1989) Morphological characterization of substance P-like immunoreactive glomeruli in the superficial dorsal horn of the rat spinal cord and trigeminal subnucleus caudalis: A quantitative study. J. Comp. Neurol. 281:497-515. Seybold, V.S. (1985) Distribution of histaminergic, muscarinic and serotonergic binding sites in cat spinal cord with emphasis on the region surrounding the central canal. Brain Res. 342291-296. Taylor, J.E., T.L. Yaksh, and E. Richelson (1982) Agonist regulation of muscarinic acetylcholine receptors in rat spinal cord. J. Neurochem. 39:521-524. Uchizono, K. (1965) Characteristics of excitatory and inhibitory synapses in the central nervous system of the cat. Nature 207:642-643. Villiger, J.W., and R.L.M. Faull (1985) Muscarinic cholinergic receptors in the human spinal cord Differential localization of ["Hlpirenzepine and [lH]quinuclidinylbenzilate binding sites. Brain Res. 345:19&199. Wamsley, J.K., M.S. Lewis, W.S. Young 111,and M.J. Kuhar (1981) Autoradiographic localization of rnuscarinic cholinergic receptors in rat brainstem. J. Neurosci. 1:17&191. Yaksh, T.L., R. Dirksen, and G.J. Harty (1985) Antinociceptive effects of intrathecally injected cholinomimetic drugs in the rat and cat. Eur. J. Pharmacol. 117:81-88. Zieglglnsberger, W., and Ch. Reiter (1974) A cholinergic mechanism in the spinal cord of cats. Neuropharmacology 13519-527.