Ž .Brain Research 820 1999 20–30

Research report

Ultrastructure of glutamate and GABA immunoreactive axon terminals of therat nucleus tractus solitarius, with a note on infralimbic cortex afferents

Fernando Torrealba ) , Celia Muller¨Departamento de Ciencias Fisiologicas, Facultad de Ciencias Biologicas, Pontificia UniÕersidad Catolica de Chile, Alameda 340, Casilla 114-D,´ ´ ´

Santiago, Chile

Accepted 8 December 1998

Abstract

The principal fast neurotransmitters in the CNS are glutamate and GABA. Our aim was to provide a baseline account on theŽ .ultrastructure of the axon terminals immunoreactive to glutamate or GABA present in the nucleus tractus solitarius NTS of the rat. In

addition, we wanted to complete our study of cortico-solitary afferents at the electron microscopic level, by analyzing the inputs from theinfralimbic cortex. Using post-embedding immunogold, we found that nearly 61% of the axon terminals were glutamatergic, and 36%were GABAergic in the rat visceral NTS. In general, axons making asymmetric synaptic contacts were enriched in glutamate, comparedto axons involved in symmetric synapses. In contrast, the vast majority of the GABAergic axon terminals made symmetric synapticcontacts. We could discern five types of glutamatergic and two types of GABAergic axon terminals that differed in their fine structure.Afferents from the infralimbic cortex were small, with clear synaptic vesicles and no dense core vesicles; they made asymmetric contactswith fine dendrites, and were glutamatergic. We conclude that most axon terminals in the NTS use glutamate or GABA as fasttransmitters, in addition to being a heterogeneous population of morphological types. q 1999 Elsevier Science B.V. All rights reserved.

Keywords: Glutamate; GABA; Immunogold; Solitarius; Infralimbic; Ultrastructure

1. Introduction

The primary visceral receptors from the vagalw x w x3,6,11,12,14,26,27 , and the glossopharyngeal 4 nervesmake synaptic connections on neurons within the caudal,

Ž .and larger, region of the nucleus tractus solitarius NTS .Because these primary afferents provide an input of clearphysiological relevance, it is customary to view the vis-ceral NTS as a relay station that distributes the differentvisceral inputs from the periphery to other centers so as toproduce appropriate responses to the eventual changes in

w xvisceral activity 2 . In addition to primary visceral affer-ents, NTS neurons receive inputs from a large variety ofstructures in the forebrain, midbrain and spinal cordw x16,22,30 . Only a few projections to the NTS have beenstudied at the ultrastructural level. The axon terminalsfrom the vagus and the carotid sinus nerve have beenidentified as among the largest in this nucleus, and Sykes

w xet al. 27 have shown that most vagal afferents are gluta-w xmatergic. Also, afferents from the amygdala 20 and from

) Corresponding author. Fax: q 56-2-222-5515; E-mail:ftorreal@genes.bio.puc.cl

w xthe insular cortex 29 have been analyzed at the electronmicroscope. The axons from the amygdala form symmetricsynaptic contacts with small dendrites. The insular cortexafferents are among the smallest in the NTS, and theymake single, asymmetric synaptic contacts with thin, distaldendrites. These cortical axons are immunoreactive to

w xglutamate 29 . There are no systematic studies in the rataimed to describe the variety of axon terminals that areimmunoreactive to the principal fast neurotransmitters,

w x w xglutamate and GABA 1 . One recent study 27 mentionedthat glutamatergic axons in the rat NTS are a homoge-

w xneous population. Our previous observation 29 suggestedto us that this is not the case, and that the existence of avariety of axon terminals that differ in relevant cytologicalfeatures better characterizes the rat NTS.

The aim of the present study was to characterize gluta-matergic and GABAergic axon terminals in two regions ofthe visceral NTS: the commissural subnucleus caudal tothe area postrema level, and the medial and commissuralsubnuclei at the area postrema level. These amino acids arethe principal fast neurotransmitters in the CNS and in the

w xNTS 1 . To complement our previous study on the insularw xcortex afferents to the NTS 29 , we analysed the ultra-

0006-8993r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved.Ž .PII: S0006-8993 98 01326-2

( )F. Torrealba, C. MullerrBrain Research 820 1999 20–30¨ 21

structure and immunoreactivity to glutamate of the secondmajor cortical input to the NTS, that arising from theinfralimbic cortex.

2. Experimental procedures

2.1. Surgery

Eight male Sprague–Dawley rats weighting 200–250 gŽwere anesthetized with thiopental 50 mgrkg, intraperi-

.toneal , and supplemented with 10 mgrkg when needed.The animals were placed on a stereotaxic apparatus, with

w xthe skull in a horizontal position 18 . After a parasagittalincision of the skin, a craniotomy allowed access to themedial frontal cortex. Under stereotaxic guidance, 1%

Ž .WGA-HRP Sigma in sterile saline was injected into theinfralimbic cortex, at 3.0 mm rostral to bregma, 0.6 mmlateral to midline. and at 4.0 mm below the cortical surfacew x18 . The tracer was pressure injected through a calibratedglass micropipette with a ca. 30 mm tip diameter. Thevolume of each injection varied from 20 to 80 nl deliveredthrough a 20 min period.

After 2 days’ survival, the subjects were deeply anes-thetized with thiopental and perfused through the leftventricle with a brief saline rinse containing 2000 i.u. ofheparin, followed by 600 ml of 1% paraformaldehyde, and

Ž .2.5% glutaraldehyde in 0.1 M phosphate buffer PB , pH7.4. The interval from the thorax opening to the beginningof the fixation was kept below 90 s. Two hundred millilitersof the fixative were delivered in 5 min, and the rest duringa 40-min period. The excess of fixative was removed bywashing with 600 ml of cold 0.1 M PB perfused during 30to 40 min. The skull was opened and left in PB for 1 h; thebrain removed and prepared into two blocks, correspond-

ing to the injected telencephalon and to the brainstem.Each block was sectioned at 50 mm in the coronal planewith a vibratome, while kept in cold PB.

2.2. Histochemical procedures

The frontal cortex and the brainstem sections wereprocessed with the Gu et al. protocol for the demonstration

w xof peroxidase at the electron microscopic level 5 . Thesections were soaked for 20 min in a 1% sodium tungstateŽ .Merck solution with 0.01 N HCl, 0.007% tetramethyl-

Ž .benzidine Sigma; dissolved in absolute ethanol , in 0.1 MPB, pH 5.2. The incubation, made under dim light, startedby adding 70 ml every 10-min of 0.3% hydrogen peroxideto each 10 ml of the pre incubation solution, during onehour. The sections were rinsed in cold 0.1 M PB, pH 6.0,followed by a rinse in 0.1 M TrisrHCl buffer, pH 7.4 for2 min. The tissue was soaked in a 0.5% cobalt chloridesolution in 0.1 M TrisrHCl buffer, pH 7.4 for 10 min;rinsed in TrisrHCl buffer; and the reaction product wasstabilized and intensified for 10 min at 378C in a TrisrHCl

Ž .buffer 0.1 M solution pH 7.4 containing 0.05% di-Ž .aminobenzidine Sigma , 0.01% hydrogen peroxide and 10

mM imidazole. The sections were rinsed in 0.2 M PB, pH7.4, and the region of the NTS dissected under a micro-scope with a scalpel. The pieces were treated with 1%OsO in 0.1 M PB, pH 7.4 at room temperature for 504

min; rinsed in PB; dehydrated in ethanol; and flat embed-Ž .ded between plastic slides Polyscience in Durcupan ACM

Ž .Fluka .The flat plastic sections containing the NTS were exam-

Žined under the light microscope to select the medial at. Žarea postrema level and commissural caudal to the area

. Ž .postrema NTS see Fig. 1 subnuclei with WGA-HRPlabeled axons, on either side of the brain. With the guid-

Ž .Fig. 1. Drawings of transverse sections through the rat medulla. The asterisks indicate the center of the regions analyzed in the intermediate left andcaudal regions of the nucleus tractus solitarius. AP, area postrema; Cu, nucleus cuneatus; Gr, nucleus gracilis; sol, tractus solitarius; 10, dorsal motornucleus of the vagus; 12, hypoglossal nucleus.

( )F. Torrealba, C. MullerrBrain Research 820 1999 20–30¨22

ance of landmarks like blood vessels, area postrema ormyelinated regions, a 1 mm2 piece was cut with a razorblade and glued with cyanoacrylate to a blank resin block.Ultrathin sections were mounted on 200 mesh thin barnickel grids, and processed with postembedding immuno-gold procedures to reveal either GABA or glutamate im-munoreactivity.

2.3. Immunocytochemistry

We used a rabbit polyclonal antibody against glutamateŽ . w x1:5000 dilution , a kind gift from Dr. P. Petrusz 7 . The

Ž .GABA antibody Incstar , raised in rabbits using GABA

conjugated to bovine serum albumin, was used at a 1:3000or 1:6000 dilution. The specificity of the antibodies was

w xtested in previous studies 7,29 . We followed the opti-mized immunogold procedures described by Phend et al.w x19 . After incubation with the primary antibody, the sec-tions were treated with goat anti-rabbit IgG-coated col-

Ž .loidal gold 10 or 20 nm, Sigma, 1:25 dilution , rinsed andstained with uranyl acetate and lead citrate.

Four rats were processed for postembedding immuno-gold, in the same way as the rats injected with the tracer,except for the peroxidase histochemical procedures.

To quantify the glutamate immunoreactivity, we calcu-lated the immunogold particle density for different NTSprofiles by dividing the number of particles over a profile

Ž .Fig. 2. Glutamate ir axon terminals in synaptic contact with different postsynaptic structures, postembedding immunogold, 10 nm gold particles. A TwoŽ . Ž .small axon terminals forming asymmetric contacts arrows with medium caliber dendrites in the commissural NTS. B One medium size GLU-ir axon

Ž . Ž . Ž .terminal forming three asymmetric contacts arrows with the same proximal dendrite pd . An axon not GLU-ir forms a symmetric contact arrowheadŽ . Ž .with the same proximal dendrite. C A large GLU-ir axon, making an synaptic contact of the asymmetric type large arrow with a vesicle containing

Ž .profile; the same axon is bound to a small dendrite through a puncta adherens small arrow . Note the deep fold of the plasma membrane, between the twoŽ . Ž . Ž .mitochondria. D An axon terminal similar to that in B in asymmetric synaptic contact with two spines sp ; note the partial segregation of the synaptic

vesicles. Bars, 0.5 mm.

( )F. Torrealba, C. MullerrBrain Research 820 1999 20–30¨ 23

by its area. The particles were counted manually under amagnifying glass, on electron micrographs printed at afinal magnification of =30,000 to 43,000. The areas weremeasured with the help of a graphics tablet and the Sig-maScan measuring system. To compare the gold particledensity of different cell profiles, we applied nonparametrictests for independent samples.

3. Results

3.1. Axon terminals immunoreactiÕe to glutamate

Immunogold particles coding for glutamate were pre-sent, with different densities, over most of the profiles

Ž .within a section see Figs. 2–5 . The labeling over glialŽprofiles was low in all sections 5.01 " 3.12

2 .particlesrmm , ns180 , and the labeling over resin wasŽ 2 .still lower -0.5 particlesrmm . The distribution of

immunogold particle density within the population of axonterminals did not show a clear break that might havehelped to set a threshold value to separate glutamatergicfrom non-glutamatergic terminals. We adopted as a crite-rion to obtain that threshold, the average gold particledensity of the tissue q2.576 S.D., that is the 99% confi-dence limit. We examined 569 axon terminals, with and

Žwithout synaptic specialization, and found that 61.6% n.s351 were glutamatergic and 38.3% were non-glutama-Ž .tergic ns218 .

The size of the glutamatergic terminals was similar inŽ 2the caudal mean"S.D.s0.811"0.645 mm ; medians

2 . Ž0.66 mm and the intermediate 0.867"0.642; medians2 .0.67 mm NTS. We observed no cytological differences

Ž .between GLU-ir axon terminals in the commissural caudalŽ .and the medial intermediate NTS, so the data were

pooled for the analysis. In these NTS regions, 68% of theŽ 2 .largest terminals -1.5 mm and, 55% of the terminals

smaller than 1.5 mm2 were glutamatergic. In contrast tosize, a good correlation was observed between the type of

Ž .synaptic contact symmetric vs. asymmetric and the levelof GLU-ir. Terminals making asymmetric synaptic con-

Ž . Žtacts ns243 had higher levels of GLU-ir 4.022=.GLU-ir over dendrites than axons making symmetric

Žsynapses ns74; GLU-irs1.823; p-0.001, Mann–.Whitney test .

Several types of GLU-ir axon terminals could be dis-cerned, based on their fine structure.

3.1.1. Large, folded, dark terminalsA typical example of the dark type is the cluster of

terminals shown in Fig. 3. They were irregular in shape

Fig. 3. A very large, lobulated GLU-ir axon terminal, probably a primary afferent, is shown occupying the center of the electron micrograph. In the planeŽ .of the section this terminal forms three asymmetric synaptic contacts arrows with three different large spines. This large terminal showed a

characteristically dark cytoplasm, due in part to the high density of small synaptic vesicles. Also typical is the presence of many swollen and palemitochondria and of deep folds where the terminal abuts itself. Note the presence of a few dense core vesicles of a medium size. The terminal and adjacent

Ž .dendritic profiles were completely surrounded by glial processes asterisks . Bars, 0.5 mm.

( )F. Torrealba, C. MullerrBrain Research 820 1999 20–30¨24

Fig. 4. Three examples of large GLU ir terminals with pale cytoplasm. Note that the clear synaptic vesicles are evenly distributed over the cytoplasm,except for the active zone, where they clustered; these terminals also contained medium size dense core vesicles and mitochondria. These pale, large

Ž . Ž . Ž .terminals formed short and simple asymmetric synaptic contacts thick arrows with medium size dendrites d or with spines sp . Note the punctaŽ .adherens between two of these terminals thin arrow in B . Bars, 0.5 mm.

and larger than 1.5 mm2, full with small, clear synapticvesicles, which made them look darker than other axons;they had a few dense core vesicles, and pale, centralmitochondria; the plasma membrane showed several tight

folds. In a single plane of section, these terminals mademultiple asymmetric synaptic contacts with spines, with

Ždendrites of different caliber, mainly medium 0.5 to 1.0.mm in diameter to large, and in occasions with somas.

Ž .Fig. 5. GLU-ir axon terminals characterized by the presence of numerous large, some larger than 100 nm in diameter dense core vesicles, in addition toŽ . Ž . Ž .small clear synaptic vesicles. They formed asymmetric synaptic contacts arrows with small A, B or medium size C dendrites. Bars, 0.5 mm.

( )F. Torrealba, C. MullerrBrain Research 820 1999 20–30¨ 25

Some of these large axon terminals and their postsynapticdendrites where surrounded by a glial sheath.

3.1.2. Large, smooth and pale terminalsAlso larger than 1.5 mm2, they had oval shape with a

smooth, unfolded plasma membrane. These large terminalsŽ .Fig. 4 had a lower density of clear synaptic vesicles,which made the cytoplasm look pale; these vesicles tendedto cluster at the active zone of the synapses. They also haddense core vesicles evenly distributed within the cyto-plasm, and dark mitochondria. They made few or singleasymmetric synaptic with spines or dendrites. On occa-

sions, two of these terminals were joined by a punctumadherens.

3.1.3. Small, round boutonsThe most frequent type of axon terminal was a small,

Ž .round bouton, filled with small clear vesicles Fig. 2A .Some of them also had a few dense core vesicles, andothers, like the cortico-solitary axons, lacked dense corevesicles. These small axonal boutons tended to make sin-gle asymmetric contacts with dendrites, spines and somas.The active zone and the postsynaptic density occupied alarge portion of the membranes in contact.

Ž .Fig. 6. Photomicrographs of the most common type of GABA immunoreactive profiles asterisks in the NTS, postembedding immunogold method, usingŽ .10 nm gold particles. Note the high signal to noise ratio of the immunolabeling for GABA. A Low power photomicrograph showing one GABA-ir axon

Ž . Ž .terminal making a synaptic contact with a proximal dendrite d , and three unlabeled axon terminals at . Terminal at makes an asymmetric synapse with3Ž . Ž . Ž .the same basal dendrite. B Large GABA-ir terminal that makes a short and simple symmetric synapse arrowhead with a large dendrite. C A small

Ž . Ž .GABA-ir axon terminal in a symmetric synapse with a small dendrite. D Two symmetric synaptic contacts arrowheads with small dendrites. Note thatŽ . Ž .one of the axons is a GABA-ir terminal, while the other at is unlabeled. Also is shown an unlabeled bouton at forming an asymmetric synapse4 5

Ž .arrow with a dendrite. Bars, 0.5 mm.

( )F. Torrealba, C. MullerrBrain Research 820 1999 20–30¨26

3.1.4. Terminals with large dense core ÕesiclesAn infrequently observed type of axon terminal

Ž .24r569 , present in caudal and intermediate NTS, wasŽ 2 .characterized by the presence of many 12–15 per mm

Žunusually large dense core vesicles around 100 nm in. Ž .diameter within a single plane of section Fig. 5 . The

2 Ž .terminals had a mean area of 0.61 mm range, 0.2–1.56and the cytoplasm contained many round, small and clearsynaptic vesicles, and a few mitochondria. Nearly 80% of

Ž .these axon terminals 19r24 were immunoreactive toglutamate, with a mean particle count of 4.04=mean

Ž .tissue range, 0.47–6.81 . They made asymmetric synapticcontacts with dendrites of all calibers and dendritic spines.

3.1.5. Medium-sized, rectangular terminalsA less prevalent type of GLU-ir axon terminal was

characterized by an accumulation of small and clear synap-tic vesicles near the active zones; the terminals were of

Ž .Fig. 7. Examples of GABAergic axon terminals asterisks forming split symmetric synapses in the NTS. Arrowheads indicate the separate synaptic zones.Ž .A A medium size terminal from the intermediate NTS makes a contact with a medium size dendrite. Note the presence of a few dense core vesicles

Ž . Ž .within the terminal. B In the caudal NTS, a medium size terminal is making a split symmetric synapse with a small dendrite. A spine sp apparentlyŽ .forms two symmetric synapses with different GABA-ir terminals. C High magnification of a split symmetric synapse between a GABA-ir large axon

Ž .terminal and a large dendrite left side . Note the thinning of both pre- and postsynaptic densities at the central cleft, and the narrowing of the extracellularŽ .space. D A large GABA-ir terminal from the intermediate NTS forms two synapses with medium caliber dendrites. The contact with the dendrite to the

Ž . Ž . Ž . Ž .left has a cleft, which cannot be resolved in the contact with the bottom dendrite. Bars, 0.5 mm for A , B and D ; 0.1 mm for C .

( )F. Torrealba, C. MullerrBrain Research 820 1999 20–30¨ 27

medium size and of a rectangular shape; they made 2 or 3Ž .asymmetric contacts Fig. 2B,D with dendrites or with

spines.Other GLU-ir axon terminals, that were less frequent

than the above mentioned, were left unclassified.The antibody against glutamate allowed us to quantify

the proportion of glutamategic dendrites in the NTS. In theŽ .commissural NTS, 90% of the dendrites 244r271 were

glutamatergic, according to the criteria we used in thisŽ .study see first paragraph of Section 3 . In contrast, only

57.6% of the dendrites in the medial NTS subnucleusŽ .91r158 were glutamatergic.

3.2. Terminals immunoreactiÕe to GABA

The identification of axon terminals as GABA-ir neededno quantification of the gold particle density, since the

Žsignal was much higher than the background Figs. 6 and.7 . This strong labeling was present over very thin unmye-

linated axons, over some axon terminals and, in occasions,over myelinated axons. Labeling over cell bodies or den-drites was less frequent and, when present, it was muchweaker. A simple inspection of the photographs revealedthat the GABA-ir axon terminals had mitochondria withhigh levels of gold particles. Also, mitochondria fromnon-GABAergic terminals, dendrites or glia had negligiblelevels of immunoreactivity to GABA. Of 202 terminalsexamined, 36% were GABA-ir.

The GABA-ir terminals in the caudal commissural sub-nucleus were much smaller than terminals in the intermedi-ate levels of the NTS. The average cross-sectional area ofGABA-ir axon terminals in the caudal commissural NTS

2 Ž . 2was 0.37 mm "0.24 S.D. , with a median of 0.29 mm .They were filled with small, clear, pleiomorphic vesicles,and had few and sometimes no mitochondria. One third ofthe GABA-ir profiles, irrespective of size, had one to threedense core vesicles in the plane of section. The vastmajority of the GABA terminals made symmetric synapticcontacts, most frequently, 38% and 42%, respectively,

Ž . Žwith medium 0.5 to 1.0 mm in diameter or small less.than 0.5 mm dendritic shafts. Synaptic contacts with

dendrites thicker than 1 mm, or with perikarya were 16%and 3% of the total, respectively. Thirteen percent of thesynaptic contacts were of the asymmetric type, judged by athick postsynaptic density.

Fig. 8. Cortico-solitary axon terminals labeled by the anterograde trans-port of WGA-HRP after injections into the contralateral infralimbic

Ž .cortex. Caudal commissural NTS. A Axon terminal labeled by a largeŽ .crystal of reaction product. Note the asymmetric synaptic contact arrow

Ž .with a small dendrite. B A GLU-ir cortical terminal that makes anŽ . Ž .asymmetric synaptic contact arrow with a small dendrite. C A cortical

terminal double labeled with WGA-HRP and gold particles coding forŽ . Ž .GLU-ir arrowheads . Twenty-nanometer gold particles for B and 10Ž .nm for C . Bars, 0.5 mm.

( )F. Torrealba, C. MullerrBrain Research 820 1999 20–30¨28

In the intermediate NTS, 36% of the axon terminalsŽ . Ž .ns62 were immunoreactive to GABA totals172 .Most terminals made symmetric synaptic contacts, mainly

Ž . Ž .with medium size 57% or large dendrites 24% , and lessŽfrequently with thin dendrites or cell bodies 11% and 8%,

.respectively . Only 6% of the terminals formed asymmet-ric synapses. Dense core vesicles were present in 37% ofthe terminals in a single plane of section. The distributionof sizes of the GABA-ir axon terminals in the intermediateNTS suggested a mixed population of different types ofterminals. The majority of GABA-ir terminals in the inter-mediate NTS showed no cytological differences with thoseof the caudal NTS. One exception was a population oflarge terminals with a split synaptic zone.

We found in the intermediate NTS a type of GABA-irŽaxon terminal not previously described in the NTS Fig.

.7 . They were among the largest GABA terminals, usuallylarger than 1 mm2, with several mitochondria, and contain-ing many small, clear vesicles and a few dense corevesicles. Their most unusual characteristic was a symmet-ric synaptic contact with thin pre- and postsynaptic densi-ties having a central discontinuity. They resembled, in thisrespect, the perforated synapses described in the cerebral

w xcortex 2 . The intercellular space was narrower at thediscontinuity, relative to the surrounding synaptic space.The postsynaptic structures were medium size dendrites,and sometimes two dendrites were contacted in a singleplane of section.

The GABA antibody we used did not allow us torecognize GABAergic dendrites, since the density of im-munogold particles of dendrites was not statistically differ-ent from that of glial profiles.

Cortical terminals. The injection of WGA-HRP into theinfralimbic and adjacent cortices labeled a moderate amountof axons and terminals in the most caudal part of thecontralateral commissural NTS. Very few labeled termi-nals were seen in the intermediate NTS. Under the electron

Ž .microscope these cortical terminals Fig. 8 were small,with a mean cross-sectional area of 0.42 mm2 "0.056

Ž .S.E.M. ns30 . They contained a few if any mitochon-dria and only small, clear synaptic vesicles. Cortical termi-nals contacted small to medium size dendrites throughasymmetric synapses. They had in average 4.16="1.78Ž .S.D. the mean tissue glutamate-ir. In occasions we de-tected a transneuronal transfer of the reaction product thatlabeled dendrites and glial profiles adjacent to corticalterminals.

4. Discussion

Our results indicate that a large fraction, up to 61%, ofthe axon terminals in the intermediate and caudal NTS areimmunoreactive to glutamate, including the afferents origi-nating from the infralimbic and adjacent cortices. We alsofound that 36% of the terminals in the intermediate andcaudal NTS are immunoreactive to GABA.

The present study shows that glutamatergic axon termi-nals in the NTS form a heterogeneous population, withdistinct types of terminals according to cytological andconnectional features. This observation does not agree with

w xa previous study in the rat NTS 27 that described gluta-matergic terminals as a homogeneous population. It is

w xclear, for example, that the afferents from insular 29 andŽ .infralimbic cortices present study do belong to a popula-

tion of GLU-ir axon terminals of the smallest size, withsmall clear synaptic vesicles and no dense core vesicles.They make asymmetric synaptic contacts mainly with thin,probably distal dendrites. These cortical terminals sharplydiffer from the largest GLU-ir axon terminals describedhere, and from the presumably serotoninergic terminalscharacterized by the abundance of large dense core vesi-cles and small clear synaptic vesicles.

In the same line, the largest GLU-ir axon terminals inthe NTS include two distinct types. The clear differencesin synaptic vesicle density, presence of folds, and mito-chondrial features makes it unlikely that they belong toportions of the same terminal. Of course, apparently smallerterminals, may be part of one of the largest terminals, orthey may belong to a single fiber that has both large andsmall terminals.

The dark type of large terminal is morphologicallysimilar to some of the primary afferents from the carotid

Ž w x.sinus nerve see Figs. 5–10 in Ref. 4 and to gastric vagalŽ w x.afferents see Fig. 5b and d in Ref. 21 . The pale type of

axon terminal may belong to some vagal afferents, likew x w xthose described in rats 26,31 and cats 3,6,11,12,14 .

The axon terminals characterized by the presence ofmany unusually large dense core vesicles, in addition tosmall and clear synaptic vesicles, are remarkably similar to

w xthe serotoninegic terminals described in the NTS 13 .Since our study shows that most of these terminals areglutamatergic, then they should release both serotonin andglutamate. This conclusion agrees with studies by Johnsonw x10 , who has shown that single raphe neurons in cultureindeed act on postsynaptic neurons by releasing both trans-mitters. The origin of most of the small axon terminals,particularly those with dense core vesicles, and the sourceof the terminals with clustered synaptic vesicles and multi-ple active zones is unknown.

It was surprising to find a substantial difference in theproportion of dendrites immunoreactive to glutamate be-

Ž . Ž .tween the commissural 90% and the medial 57.6%NTS. This finding may imply that a higher fraction of theneurons in the commissural NTS are indeed glutamatergic.Unfortunately, the method we used to label GABAergicprofiles was of no use to identify dendrites immunoreac-tive to GABA. However, the lower fraction of glutamater-gic dendrites in the medial NTS may in part be explained

w xby a higher proportion of GABAergic neurons 15 , com-pared to the commissural NTS.

We found that about a third of the NTS axon terminalsare GABAergic, and that these terminals tend to contact

( )F. Torrealba, C. MullerrBrain Research 820 1999 20–30¨ 29

proximal dendrites and somas. These conclusions agreew xwith previous studies in rats and cats 8,9,14,23 . A novel

finding of the present study was the existence, in theintermediate but not the commissural NTS, of a populationof large GABAergic terminals with a split symmetric

w xsynapse, superficially similar to perforated synapses 2 . Acomparable ultrastructure has been recently described for

w xsome glycinergic terminals within the spinal cord 32 .A vast majority of the synaptic contacts were of the

symmetric type, as expected for GABAergic axon termi-nals. However, nearly 10% of the synapses were asymmet-ric. Since we did not make serial section reconstruction,we are not able to tell whether we are in the presence of asingle population of terminals with regional variation inthe thickness of the postsynaptic density or we have twopopulations of GABAergic terminals.

At present, the ultrastucture of axon terminals of knownorigin has been directly determined for cortical, vagal andglossopharyngeal afferents to the NTS, all of them gluta-matergic. In addition, the serotoninergic terminals withvery large dense core vesicles are likely to originate from

w xneurons in the medullary raphe nuclei 24,25,28 . How-ever, nearly 5% of the vagal afferents are serotoninergicw x17 , and it remains to be determined what ultrastructuraltype of terminals these axons have.

The origin of the GABAergic innervation of the NTShas not been directly studied. However, the NTS contains

w xmany GABAergic neurons 8,9,14,23 , and it is likely thata sizable fraction of the GABA-ir terminals are intrinsic. Inaddition, projections from the amygdala form symmetric

w xcontacts with NTS neurons 20 , making them a candidatefor extrinsic GABAergic innervation.

5. Conclusion

We conclude that the vast majority of NTS synapses areeither glutamatergic or GABAergic, which implies that thegreat number of other neurotransmitters described in NTSafferents colocalize with glutamate or GABA. In addition,the present study revealed that infralimbic and insularcortex afferents are similar in cytological features and inbeing glutamatergic. They differ only in their distribution,where insular cortex afferents terminate more rostrally inthe ipsilateral medial and commissural subnuclei, whileinfralimbic afferents synapse on neurons in the contralat-eral caudal commissural subnucleus. Further studies areneeded to elucidate the structure–function relationship foreach of the axon types described in the present research.

Acknowledgements

Financed by Fondecyt grant 194r0652. We thank Dr.P. Petrusz for supplying the glutamate antiserum, and Ms.Monica Belmar for technical help.´

References

w x1 M.C. Andresen, D.L. Kunze, Nucleus tractus solitarius—gateway toŽ .neural circulatory control, Ann. Rev. Physiol. 56 1994 93–116.

w x2 R.K.S. Calverley, D.G. Jones, Determination of the numerical den-sity of perforated synapses in rat neocortex, Cell Tissue Res. 248Ž .1987 399–407.

w x3 G. Chazal, A. Baude, A. Barbe, J.J. Puizillout, Ultrastructuralorganization of the interstitial subnucleus of the nucleus of thetractus solitarius in the cat; identification of vagal afferents, J.

Ž .Neurocytol. 20 1991 859–874.w x4 I.-L. Chen, J.T. Weber, R.W. Rieck, R.D. Yates, Synaptic connec-

tions of central carotid sinus afferents in the nucleus of the tractussolitarius of the rat: I. An electron microscopic study, J. Neurocytol.

Ž .21 1992 137–147.w x5 Y. Gu, Y. Chen, L. Ye, Electron microscopical demonstration of

horseradish peroxidase by use of tetramethylbenzidine as chromogenŽ .and sodium tungstate as stabilizer TMB-ST method : a tracing

method with high sensitivity and well preserved ultrastructural tis-Ž .sue, J. Neurosci. Methods 42 1992 1–10.

w x6 D.G. Gwyn, P.H. Wilkinson, R.A. Leslie, The ultrastructural identi-fication of vagal terminals in the solitary nucleus of the cat afteranterograde labelling with horseradish peroxidase, Neurosci. Lett. 28Ž .1982 139–143.

w x7 J.P. Hepler, C.S. Toomim, K.D. McCarthy, F. Conti, G. Battaglia,A. Rustioni, P. Petrusz, Characterization of antisera to glutamate and

Ž .aspartate, J. Histochem. Cytochem. 36 1988 13–22.w x8 P.N. Izzo, R.M. Sykes, K.M. Spyer, g-Aminobutyric acid im-

munoreactive structures in the nucleus tractus solitarius: a light andŽ .electron microscopic study, Brain Res. 591 1992 69–78.

w x9 H.G. Jia, B.R. Wang, Z.R. Rao, J.W. Shi, R. Shigemoto, T. Kaneko,N. Mizuno, GABAergic synapses upon neurons expressing sub-stance P receptors in the nucleus of the solitary tract: an immunocy-tochemical electron microscope study in the rat, Neurosci. Lett. 210Ž .1996 49–52.

w x10 M.D. Johnson, Synaptic glutamate release by postnatal rat serotoner-Ž .gic neurons in microculture, Neuron 12 1994 433–442.

w x11 M. Kalia, D. Richter, Morphology of physiologically identifiedslowly adapting lung stretch receptor afferents stained with intra-axonal horseradish peroxidase in the nucleus of the tractus solitariusof the cat: II. An ultrastructural analysis, J. Comp. Neurol. 241Ž .1985 521–535.

w x12 M. Kalia, D. Richter, Rapidly adapting pulmonary receptor affer-ents: II. Fine structure and synaptic organization of central terminalprocesses in the nucleus of the tractus solitarius, J. Comp. Neurol.

Ž .274 1988 574–594.w x13 B. Maley, R. Elde, The ultrastructural localization of serotonin

immunoreactivity within the nucleus of the solitary tract of the cat,Ž .J. Neurosci. 2 1982 1499–1506.

w x14 A. Maqbool, T.F.C. Baten, P.N. McWilliam, Ultrastructural relation-ships between GABAergic and cardiac vagal preganglionic mo-toneurons and vagal afferents in the cat: a combined HRP tracing

Ž .and immunogold labelling study, Eur. J. Neurosci. 3 1991 501–513.w x15 M.P. Meeley, D.A. Ruggiero, T. Ishitsuka, D.J. Reis, Intrinsic

g-aminobutyric acid neurons in the nucleus of the solitary tract andthe rostral ventrolateral medulla of the rat: an immunocytochemical

Ž .and biochemical study, Neurosci. Lett. 58 1985 83–89.w x16 E.J. Neafsey, K.M. Hurley-Gius, D. Arvanitis, The topographical

organization of neurons in the medial frontal, insular and olfactorycortex projecting to the solitary nucleus, olfactory bulb, periaqueduc-

Ž .tal gray and superior colliculus, Brain Res. 377 1986 261–270.w x17 A. Nosjean, C. Compoint, C. Buisseret-Delmas, H.S. Orer, N.

Merhi, J.J. Puizillout, R. Laguzzi, Serotoninergic projections fromthe nodose ganglion to the nucleus tractus solitarius: an immunocy-tochemical and double labelling study in the rat, Neurosci. Lett. 114Ž .1990 22–26.

( )F. Torrealba, C. MullerrBrain Research 820 1999 20–30¨30

w x18 G. Paxinos, C. Watson, The Rat brain in Stereotaxic Coordinates,2nd edn., Academic Press, Sydney, 1986.

w x19 K.D. Phend, R.J. Weinberg, A. Rustioni, Techniques to optimizepost-embedding single and double staining for amino acid neuro-

Ž .transmitters, J. Histochem. Cytochem. 40 1992 1011–1020.w x20 V.M. Pickel, E.J. van Bockstaele, J. Chan, D.M. Cestari, Amygdala

efferents form inhibitory-type synapses with a subpopulation ofcatecholaminergic neurons in the rat nucleus tractus solitarius, J.

Ž .Comp. Neurol. 362 1995 510–523.w x21 L. Rinaman, J.P. Card, J.S. Schwaber, R.R. Miselis, Ultrastructural

demonstration of a gastric monosynaptic vagal circuit in the nucleusŽ .of the solitary tract in rat, J. Neurosci. 9 1989 1985–1996.

w x22 C.A. Ross, D.A. Ruggiero, D.J. Reis, Afferent projections to cardio-vascular portions of the nucleus of the tractus solitarius in the rat,

Ž .Brain Res. 223 1981 402–408.w x23 S. Saha, T.F.C. Batten, P.N. McWilliam, Glutamate, g-aminobutyric

acid and tachynin-immunoreactive synapses in the cat nucleus trac-Ž .tus solitarii, J. Neurocytol. 24 1995 55–74.

w x24 N. Schaffar, J.P. Kessler, O. Bosler, A. Jean, Central serotoninergicprojections to the nucleus tractus solitarii: evidence from a double

Ž .labeling study in the rat, Neuroscience 26 1988 951–958.w x25 H.W.M. Steinbusch, Distribution of serotonin-immunoreactivity in

the central nervous system of the rat—cell bodies and terminals,Ž .Neuroscience 6 1981 557–618.

w x26 K.K. Sumal, W.W. Blessing, T.H. Joh, D.J. Reis, V.P. Pickel,Synaptic interactions of vagal afferents and catecholaminergic neu-

Ž .rons in the rat nucleus tractus solitarius, Brain Res. 277 198331–40.

w x27 R.M. Sykes, D. Jordan, K.M. Spyer, P.N. Izzo, Demonstration ofglutamate immunoreactivity in vagal sensory afferents in the nucleus

Ž .tractus solitarius of the rat, Brain Res. 762 1997 1–11.w x28 K.B. Thor, C.J. Helke, Serotonin and substance P-containing projec-

tions to the nucleus tractus solitarii of the rat, J. Comp. Neurol. 265Ž .1987 275–293.

w x29 F. Torrealba, C. Muller, Glutamate immunoreactivity of insular¨cortex afferents to the nucleus tractus solitarius in the rat: a quantita-

Ž .tive electron microscopic study, Neuroscience 71 1996 77–87.w x30 D. van der Kooy, L.Y. Koda, J.F. McGinty, C.R. Gerfen, F.E.

Bloom, The organization of projections from the cortex, amygdala,and hypothalamus to the nucleus of the solitary tract in rat, J. Comp.

Ž .Neurol. 224 1984 1–24.w x31 L. Velley, T.A. Milner, J. Chan, F.S. Morrison, V.M. Pickel,

Relationship of Met-enkephalin-like immunoreactivity to vagal af-ferents and motor dendrites in the nucleus of the solitary tract: alight and electron microscopic dual labeling study, Brain Res. 550Ž .1991 298–312.

w x32 B. Walmsley, F.J. Alvarez, R.E.W. Fyffe, Diversity of structure andŽ .function at mammalian central synapses, TINS 21 1998 81–88.