THE JOURNAL OF COMPARATIVE NEUROLOGY 31554-84 (1992)

Cerebellar Nuclei and the Nucleocortical Projections in the Rat: Retrograde Tracing Coupled to GABA and Glutamate Immunohistochemistry CESIRA BATINI, CLAUDE COMPOINT, CATHERINE BUISSERET-DELMAS, HERVE DANIEL, AND MARYVONNE GUEGAN Laboratoire de physiologie de la motricite, CNRS URA 385, Universite Pierre et Marie Curie, CHU Pitie-Salpetriere, 75013, Paris, France

ABSTRACT The amino acids GABA and glutamate (Glu) are thought to be the principal substances in the central nervous system responsible for neuronal inhibition and excitation. Their distributions among the different neurons in a defined pathway may thus be indicative of the contributions of the cells to pathway function. Examples of such neurons are those of the cerebellar nuclei which, while regulating output from the Purkinje cells of the cerebellar cortex, are also found to project back to the cerebellar cortex. Immunohistochemical experiments were done to identify GABA and glutamate (Glu) containing cells in the adult rat cerebellar nuclei. Consecutive semithin and serial vibratome sections were incubated with antisera raised in rabbit against GABA and Glu. In semithin sections, only small neurons were intensely GABA immunoreactive (GABA-IR) (31.7%), and the majority (80.5%) were Glu immunoreactive (Glu-IR) of different sizes. Consistent with Glu being a metabolic precursor for GABA, 75.4% of the GABA-IR population colocalized Glu. In vibratome sections GABA-IR neurons showed some local differences in number, whereas the Glu-IR were uniformly distributed in the three nuclei studied. Measured mean diameters for these neurons showed a distinct size difference for the GABA- and Glu-IR with little overlap. Cerebellar nuclei neurons projecting to the cortex (nucleocortical neurons, NCN) were identified by locally preinjecting the retrograde transported WGA-apoHRP-colloidal gold complex in the cerebellar cortex. Vibratome sections of these cerebella were silver intensified for the retrograde tracer and double labeled for GABA and Glu. Of the total number of identified NCN, 8.7% were GABA-IR (10 animals) and 47.7% Glu-IR (5animals). Many retrograde labeled NCN in the core of the thick sections were immunonegative for both amino acids due to poor antibody penetration, thus underestimating the proportions of cells containing GABA and Glu. The size distributions for the GABA-IR and Glu-IR NCN were similar to those measured in non-retrograde labeled nuclei in thick sections. The conclusions reached are that GABA-IR neurons of the cerebellar nuclei, including the NCN, use GABA as the presumed inhibitory neurotransmitter and that Glu-IR neurons may use Glu or another excitatory neurotransmitter. Key words: immunohistochemistry, cerebellar nuclei, GABA, glutamate

Purkinje cells of the cerebellar cortex send axons to the cerebellar nuclei with a mediolateral organization conventionally divided into sagittal zones A to D as well as sub-zones (see Brodal and Kawamura, '80; Fig. 6). Some of the nuclei cells, receiving input from the Purkinje cells, project back to the same cortical zones (the nucleocortical neurons, NCN) and have been intensively studied (Gould and Graybiel, '76; Tolbert et al., '76, '77; Batini et al., '78; Haines and Pearson, '79; Tolbert and Bantli, '79; Dietrichs and Walberg, '79; '80; Dietrichs, '81; '83; Hess, '82; Payne, '83; Trott et al., '90; Umetani, '90). O

1992 WILEY-LISS, INC.

In the rat, the majority of the NCN project to the ipsilateral cerebellar cortex (Buisseret-Delmas and Angaut, '89). These neurons are known as the reciprocal NCN and make closed feedback loops with the Purkinje cells within a cerebellar zone. In addition to the reciprocal projections,

Accepted August 21, 1991. Address reprint requests to Dr. C. Batini, Laboratoire de physiologie de la motricit6, CNRS, VRA 385, Universite Pierre et Marie Curie, 91 Bod de l'H6pital,75013 Paris, France.

GABA, GLU IN CEREBELLAR AND NUCLEOCORTICAL NEURONS

75

some of the NCN in each zone project to contralateral nucleus lateralis (NL). Zone B was also occasionally inhomologous zones or to other ipsilateral zones. These are, jected, but was not studied as it does not have nucleocortical respectively, known as the symmetrically and non-recipro- projections. cally projecting NCN. They are considered an open feedback Perfusion loop, since their target Purkinje cells are not directly afferent to them. While the anatomy of the nucleocortical Under deep sodium pentobarbital anesthesia (40 mg/kg) pathways has been extensively studied, their function is not the rats were intracardially perfused, rapidly a t first with established. Gamma aminobutyric acid (GABA) is found in about 200 ml of Ringer's solution and then with 1,000ml of some NCN and is considered their putative neurotransmit- the fixative. The fixative solution contained paraformaldeter, giving them an inhibitory function (Chan-Palay et al., hyde (PF) (4%)in phosphate buffer (PB) (0.1 M, pH 7.4) '79; Tolbert and Bantli, '80; Hamori and Takacs, '89). In alone or with glutaraldehyde (GA) (0.1-0.3%). Two rats, the present study, the GABA and glutamate (Glu) content used for semithin sections, were intracardially perfused of these neurons has been examined immunocytochemi- directly with 250 ml of PF (4%)and GA (0.2%)in PB (0.1 M; cally for the possibility that this pathway may have two pH 7.2) and subsequently with 500 ml of PA (4%) in PB (0.1 populations of NCN, one having an inhibitory function and M; pH 7.2). The cerebella were then removed whole and kept from 3 hours to overnight in the same fixative the other excitatory. Baseline experiments were first done in semithin and solution. thick sections of the cerebellar nuclei in order to develop Immunohistochemistry criteria for identifying GABA and Glu containing neurons Antibodies raised in rabbit against GABA conjugated to reported in the literature (Oertel et al., '81, '82; Mugnaini and Oertel, '85; Monaghan et al., '86; Kumoi et al., '88; bovine serum albumin were bought from Immunonuclear Batini et al., '89, Kaneko et al., '89; Nelson and Mugnani, (see Maley and Newton, '85) and used diluted 1/5,000 or '89). The semithin sections had two advantages: both from Sigma (see Hodgson et al., '85) and used diluted GABA and Glu antibodies completely penetrated the sec- 1/8,000-10,000; antibodies raised in rabbit against Glu tions, thus binding to all the available antigen; hemisec- conjugated to bovine serum albumine were bought from tioned neurons could be identified in consecutive sections. Immunotech and used diluted 1/ 100-200. Serial vibratome sections 40 or 50 pm thick were first Treating such sections with different antibodies allowed incubated, free-floating, a t room temperature for 30 minunambiguous identification of intracellular colocalizations. utes in normal goat serum (2% in phosphate buffer salt, By using the established criteria, GABA and Glu immuno- PBS), then incubated in one of the primary antibodies reactive neurons of the cerebellar nuclei were quantita(diluted in PBS) at room temperature overnight or at 4°C tively analyzed in thick sections to study the NCN particifor 24 to 48 hours, then treated for 1 hour with antirabbit pating in the nucleocortical pathway. Part of this work has IgG serum diluted 11400 in PBS. They were finally probeen published in a preliminary report (Batini et al., '89). cessed by the indirect peroxidase-antiperoxidase (PAP, Sternberger, '86), avidin biotin (Hsu et al., '81) or extravidin methods. DAB (diaminobenzidine chloride) or AEC (3 MATERIALS AND METHODS amino- 9 ethylcarbazole) were used as chromagens. In each The experiments were done with 22 rats; of these 12 had experiment one control section was processed without the a retrograde tracer injected in the cerebellar cortex before primary antibody to confirm that there was no immunopreparation for histological experiments. Histological prep- staining in these sections. arations of all 22 animals were treated immunohistochemiIn three protein gold complex injected animals, alternate cally with anti-GABA and anti-Glu antibodies. thick sections of the cerebellum were treated for GABA and Glu immunostaining. Each pair of adjacent sections was Tracing the NCN mounted in such a way that the two surfaces, resulting The complex of wheat germ agglutinin-apohorseradish from the same vibratome knife cut, were facing the coverperoxidase coupled to colloidal gold is only retrograde slip. Thus any neurons cut in two by the knife displayed one transported (Basbaum and Menetrey, '87). This tracer was part immunostained for GABA and the other for Glu as pressure injected through micropipettes with tip diameters mirror images. The cerebella of the two animals prepared for semithin of 30-50 pm at a depth of about 1 mm in the cerebellar cortex of pentobarbital (40 mg/kg) anesthetized rats. In sections were first cut in the frontal plane at 50 pm with the each animal, 0.2-1 p1 of tracer was injected in 2-5 nearby vibratome. These thick sections were then osmicated, dehypoints of the cerebellar cortex. One to three day survival drated, and embedded in epon. The areas of interest were times were allowed for retrograde transport of the tracer trimmed out, attached to epon blocks, and serial semithin before intracardiac perfusion and vibratome sectioning as sections cut at 1 km were mounted on gelatinized slides. described below. The thick sections were first processed The sections were etched for 10 min in 200 ml methanol, with a silver intensification method in a darkroom by 100 ml propylenoxide, and 40 g potassium ethoxide, rinsed following the recipe extensively described by Basbaum and in methanol and again in a methanol PBS solution, deosmiMenetrey ('87) to reveal the protein gold complex. They cated in 1% NaIO,, and finally immersed rapidly in NaBH, were subsequently treated immunohistochemically to re- (0.1M) solutions. Alternate serial sections were then treated for GABA veal the GABA or Glu content as double labeled cells. The injection sites, determined histologically in each (diluted 1/2,500) and Glu (diluted 1/ 100) immunostaining. experiment, covered more than one zone. In different After rinsing in PBS, they were incubated for 20 min in animals the following zones were retained for analysis: zone normal goat serum (10% in PBS), then overnight in the A connected to the nucleus medialis (NM), zone C (includ- primary antibody, then for 30 min in antirabbit IgG serum. ing its subzones) connected to the nucleus interpositus They were then stained with the PAP method. The DAB (NI), and zone D (including its subzones) connected to the reaction product was intensified by immersing the sections

C. BATINI ET AL.

76

Fig. 1. Examples of t h e immunostaining obtained with glutamate (A) and GABA (B) in non-consecutive semithin sections of the nucleus interpositus of t h e cerebellum.

in OsO, (0.1%) for 5 min or in a gold intensification solution. The slices were finally dehydrated and coverslipped with Eukitt.

Cell counting and measuring Cell counting was based on criteria which tried to minimize subjective error. The protein-gold complexes were unambiguously clear black points in the microscope images. All cells containing numerous particles, confined to the cytoplasm, were classified as being retrograde labeled and counted throughout the thick sections. Judging the GABA and Glu immunoreactivity was more subjective. A general quality rule was adopted that only cells showing uniformly dark or distinctively reticulated cytoplasmic staining were to be counted for each antibody (see for examples Figs. 1-3). Cells with general light staining were considered as only background stained and were not counted. While no differences were noted for the antibodies from different sources, variability in coloration from animal to animal was encountered. Therefore, the above quality criteria were always applied, allowing only for the overall degree of staining in each case. To avoid double counting of cells, alternate vibratome thick sections or semithin sections separated by at least 100 p,m were used. Since cell body shapes varied, their sizes were determined by measuring their long (X) and short (Y)diameters, then calculating the mean diameter as (X + Y)/2. Corrections for shrinkage were not attempted.

RESULTS GABA and glutamate immunostained neurons in the cerebellar nuclei Semithin sections of the cerebellar nuclei treated with anti-GABA antibodies showed densely packed fibers and

intensely marked bouton-like profiles. Intensely GABA immunoreactive (GABA-IR)somata, small, round or ovoid, with large nuclei surrounded by little cytoplasm were also present (Figs. 1,2). The nucleus was usually more intensely stained than the cytoplasm. In the sections treated for Glu, most of the somata as well as fibers and bouton-like material were very immunoreactive (Glu-IR) (Figs. 1, 2); the somata were of all the neuron types described previously in the cerebellar nuclei (see Chan-Palay, '77). A total of 205 neurons, recognized in consecutive GABA and Glu immunostained sections, were counted in the three cerebellar nuclei. Of these neurons, 56.6%were Glu-IR only and 7.8% GABA-IR only, 11.7%were negative for both GABA and Glu. Some neurons (23.9%)had GABA and Glu colocalization and were in the small mean diameter range (Fig. 2). As GABA containing cells, they would represent 75.4%of the GABA-only population. This assignment of the colocalizing cells may be justified since Glu, as the precursor, is expected to be present in cells containing GABA. Table 1 summarizes these results. TABLE 1. Neurons Immunostained With GABA and/or Glu in Semithin Sections' No. of GABA + No. of Glu+ No. of GluTotal

49 i23.9%) 16 (7.8%) 65 (31.7%)

NO.of GABA116 (56.68) 24 (11.76) 140 (68.3%)

Total 165 (80.5%) 40 (19.5%) 205 (100%)

'The percentages (given in parentheses) are of the total of 205 immunoreactive neurons identified in t h e semithin sections. GABA+ and Glu+ are the immunoreactive and GABA- and Glu- are the non-immunoreactive neurons

In thick sections treated with anti-GABA antibodies, many immunoreactive fibres and terminals and some cell bodies were observed. Again, the nuclei were usually more

GABA, GLU IN CEREBELLAR AND NUCLEOCORTICAL NEURONS

Fig. 2. Demonstration of colocalized and non-colocalizedGABA and Glu in the same neuron. Two consecutive semithin sections of the nucleus interpositus of the cerebellum, one immunostained for GABA (A) and the other for Glu (B). Both sections contained the same two

77

neurons, a small one (black arrows) and a large one (white arrows). Note that the GABA-IR neuron on the right in A is also Glu-IR in B. The Glu-IR soma in B is surrounded by densely marked bouton-like terminals for GABA in A; these could be from Purkinje cell axons.

Fig. 3. Examples of neurons in 40 em thickvibratome sections, small GABA-IR (A) and larger Glu-IR (B)

intensely stained than the cytoplasm (see also Ottersen and Storm-Mathisen, '84). In each animal, the well-known GABA immunoreactivity in the overlying cerebellar cortex (Gabbott et al., '86) was taken as indicating successful immunostaining. Only those experiments where label was

present in the soma of stellate, basket, and Golgi interneurons, but not the soma of the Purkinje cells, were retained for further studies. In all these experiments, regardless of the sera from two sources and the histochemical technique used, only small neurons, round or oval, similar to those

78

C. BATINI ET AL.

observed in the semithin sections, were intensely GABA-IR (Fig. 3). Glu immunoreactivity in thick sections of the cerebellar nuclei also was found in large numbers of fibers and terminals as well as cell bodies. The specificity of the Glu antibodies was also judged by the known immunostaining of the overlying cerebellar cortex (Ottersen and StormMathisen, '84). Glu immunoreactivity and neuron type (Fig. 3) was clearly reproducible between experimental animals. While somewhat variable in intensity, they could be classified as Glu-IR or negative. Positional relations of GABA-IR and Glu-IR neurons within the nuclei were examined in alternate serial sections treated with antiGABA and antiGlu sera. Small, heavily labeled GABA-IR neurons were scattered in the cerebellar nuclei but in their ventral parts and in the NL were more numerous and in clusters of 3 to 7 cells. Glu-IR neurons of all sizes were more evenly distributed throughout the cerebellar nuclei. Figure 4 illustrates these differences in two consecutive sections at low power as mirror images and also show that the two types of cells are intermingled. Clear evidence for GABA versus Glu immunostaining was provided by the hemisectioned neurons which could be identified in adjacant alternately immunostained semithin sections. In such neurons, the presence or absence of GABA and Glu immunoreactivity could be confirmed in the same neuron. In thick sections, neurons unstained for GABA were Glu-IR (see Fig. 8) and were in the medium to larger sizes. Hemisectioned GABA-IR neurons were rarely found in adjacent vibratome slices treated with antiGlu sera, perhaps because of their small size. The above results in semithin and thick sections indicated that only small neurons were GABA-IR, whereas neurons of all sizes were Glu-IR. To better quantify their size relations, the mean diameters of identified GABA-IR and Glu-IR neurons of the NM, NI, and NL were measured in serial thick sections from three experiments. The

GABA-IR cell diameters were very similar in the three nuclei, ranging from 5 to 22.5 p.m with a peak near 10 km (Fig. 5A). These results showed that the GABA-IR neurons of the cerebellar nuclei are a rather homogeneous population of small cells. The Glu-IR neurons of the three nuclei included sizes from 10 to 35 (*mwith apeak at about 20 pm. This peak was slightly broader than for the GABA-IR and could be considered, morphologically, to be more heterogeneous (Fig. 5B). The similarities of GABA- and Glu-IR cells in the three nuclei as well as the differences between the two populations were clearly revealed when the measurements for each were averaged over the three cerebellar nuclei and plotted together as shown in Figure 5C. The size range of the overlap between the two cumulated spectra strongly indicated that these were the GABA-IR neurons colocalizing Glu.

Marking the nucleocortical neurons The protein gold complex injected in the cerebellar cortex retrogradely labeled cell bodies in the cerebellar nuclei. Because of the topographical arrangement of the nucleocortical pathways, the location and size of the injection sites determined whether labeled neurons were found in only one of the cerebellar nuclei or in more than one. In the present experiments, large injection sites were used to assure a sufficient number of retrogradely labeled neurons. As a consequence, fine nucleocortical subzonal relationships could not be individually discriminated, but when included in the injection site, zonal subdivisions were always confirmed (Table 2). While reciprocal and symmetrical projections were easily distinguished, non-reciprocal projections were observed only exceptionally. This was because the large injection sites included successive cortical zones precluding the separation of reciprocal and nonreciprocal projections. Table 2 lists the injection sites used and the reciprocal and symmetrical projections obtained in

Fig. 4. Typical distribution of immunoreactive neurons in the nucleus lateralis of the cerebellum. Consecutive sections immunostained for glutamate (A) and GABA (B) are shown as mirror images (see text). Note the difference in size of the stained neurons in the facing surfaces.

79

GABA, GLU IN CEREBELLAR AND NUCLEOCORTICAL NEURONS TABLE 2. Cortical Zone and GABA-IR and Glu-IR Neurons in the

Cerebellar Nuclei'

B Glu Rat GlOl N 107

N 108 N 117 N 118 N 124 N 130 N 139 N 142 N 145 N 146 N 147

Cortical zones injected C-D A-GD A-C-D A-C A-C-D C-D A-C-D A A-C A-C-D A-C-D C-D

GABA-IR Recipr C-D A-C-D A-C-D A-C A-CD G D A-CD A A-C

Symm

Glu-IR Recipr

Symm

A A A-GD A-C-D C-D

A

A-C A A-C A-C-D C

C-D

A-C-D A-C-D

'Sagittal zones included in the injection sites in each of the 12 rats (left column) and reciprocally (recipr)and symmetrically (symm) connected neurons found to be GABAand Glu-IR in the nuclei. Zone B, projecting to nucleus vestibularis lateralis, is not included because it does not retrogradely label the vestibular neurons.

hemi-neurons could not always be identified unambiguously in the Glu stained sections.

GABA and glutamate immunoreactive nucleocortical neurons

I,NL 10

20

Fig. 5. Immunoreactive cell diameter distributions in the nucleus medialis (NM), the nucleus interpositus (NU and the nucleus lateralis (NL) of the cerebellum. The populations of GABA-IR (A) and Glu-IR (B) neurons were clearly peaked and very similar in size in all three nuclei. In C the spectra from the three nuclei are averaged for both populations and plotted together. The size range of the GABA and Glu overlap is the same as for cells positivelyidentified as colocalizing GABA and Glu. Abscissae: diameter of the neurons in km (class interval 2.5 Fm). Ordinates: percentage of neurons in each diameter class.

individual experiments, and Figure 6 shows the extent of each of the injection sites. Sections of retrograde labeled cerebella contained neurons with the same morphological properties as the GABA-IR and Glu-IR neurons of the nonretrogradely marked cerebellar nuclei (Fig. 7). Double labeled neurons were present in the surfaces of the thick sections and became progressively less numerous in the depths of the sections. This observation was consistent with a previous report that the protein gold complex histochemical reaction is effective throughout a 40-50 km section but neither of the immunohistochemical procedures penetrates more than about 10 Fm into each surface of the slice (Batini et al., '89). In three retrograde labelling experiments, serial thick sections were treated alternately with GABA and Glu antisera. In facing sections, retrogradely labeled hemiGlu-IR neurons were readily seen to be negative for GABA (Fig. 8). However, the equivalent in strongly GABA-IR

Since the reciprocal and symmetrical projecting NCN could be identified, their proportions immunoreactive for GABA and Glu were evaluated in individual cerebellar nuclei. In each of the three nuclei examined, only a small proportion of the total number of retrogradely labeled NCN was found to be GABA-IR. Furthermore, the proportions of NCN containing GABA were very similar whether the nuclei gave reciprocal or symmetrical projections. In spite of the subjective counting method and the use of different immunostaining methods, the percentage of GABA-IR NCN was very constant across ten experiments, as shown in Figure 9A. When grouped, the numbers of GABA-IR NCN in the three nuclei from ten experiments gave an overall average for the retrogradely labeled GABA-IR NCN of 8.7 -C 2.2%. The percentages of Glu-IR NCN were also roughly the same in the nuclei reciprocally or symmetrically connected to the cortical injection sites in five animals (Fig. 9B). The overall averaged proportion of Glu-IR NCN was much larger than for the GABA-IR at 47.7 -C 3.5%, with very small variability from one animal to another. The disproportionality between GABA-IR and Glu-IR neuron populations was clearly evident in alternate GABA and Glu immunostained sections from each of three animals (Fig. 9C). In two animals, the sizes of the GABA-IR and Glu-IR NCN were measured in alternate serial sections. The general distribution of diameters for all the retrogradely labeled neurons in all three nuclei were very similar as shown in Figure 10. The GABA-IR NCN had small sizes, and the Glu-IR NCN appeared to be of all sizes, as was found for the non-retrograde labeled neurons above (compare Fig. 5 with Fig. 10). These results demonstrate that few NCN projecting neurons contain the inhibitory neurotransmitter GABA, while most contain the presumed excitatory neurotransmitter Glu. They also indicate that morphologically, the amino acid containing NCN in the cerebellar nuclei are not different from the non-NCN.

DISCUSSION The new aspect of this work is the quantitative evaluation of the GABA and Glu containing neurons of the cerebellar nuclei and of those participating as NCN in the

80

C. BATINI ET AL.

Ni

Fig. 6. The areas of the twelve retrograde tracer injection sites relative to the sagittal zones A-D. The numbered injection areas (left) and their relations to the sagittal zones A-D (right) are identified in

Table 2. The right-hand figure is a schematic representation of the sagittal zones A to D, including the subzones (for C and D) that were not considered separately in this study.

reciprocal and symmetrical pathways. The results support the following conclusions: 1) GABA- and Glu-IR neurons may be distinguished by their sizes. 2) The majority of the neurons of the cerebellar nuclei are Glu-IR only and many of the GABA-IR neurons colocalize Glu. 3) The proportions of GABA-IR and Glu-IR neurons in the three cerebellar nuclei are very similar. For the NCN the following conclusions can be reached: 1) The mean sizes of GABA-IR and Glu-IR neurons are also distinctly different. 2) The GABA-IR population is much smaller than the Glu-IR but is clearly not negligible. 3) The proportions of GABA-IR and Glu-IR NCN are similar in the three cerebellar nuclei. 4)The reciprocal and the symmetrical projections have the same proportions of GABA-IR and Glu-IR neurons. 5) For a given cortical injection, the number of symmetrical NCN is smaller than the reciprocal NCN.

GABA and Glu immunoreactivity in the cerebellar nuclei While GABA and Glu immunoreactivity was found in the fibres and terminals as well as in the somata of the neurons in the cerebellar nuclei, only the latter will be considered for discussion as pertinent to this report. With the criteria used to identify immunoreactivity, a homogeneous population of small cells could be identified as GABA-IR in the three cerebellar nuclei. Previous authors, using antibodies raised against GAD (glutamic acid decarboxylase, the enzyme synthetising GABA) or GABA, have also reported small immunostained neurons in the cerebellar nuclei (Oertel et d.,'81;'82; Mu*aini and Oerte1> '85; Iiumoiet '88; Batini et d.,'89; Buisseret-Delmas et al., '89; Nelson and Mugnaini, '89). We consider that the population of small

Fig. 7, Typical appearance of retrograde labeling in vibratome sections. The neurons shown, containing black grains of the tracer, were in the nucleus interpositus and were also immunostained with glutamate (A) and with GABA (B).

81

L1

10.

0.

Fig. 8. Retrograde labeled, Glu-IR (A) and GABA negative (B) nucleocortical neurons in vibratome sections mounted a s mirror images.

%IA

El GABA El Glu

20

0

GABA ,

1

B n

NLNL

9 'O 0I

Fig. 9. Percentages of the retrograde labeled nucleocortical neurons identified as GABA-IR in ten animals (A) or as Glu-IR in five animals (B). For three animals, alternate sections were immunoreacted for GABA and Glu to determine populations in the same animal (C). Abscissae: individual experiments. Ordinates: percentage of immunoreactive retrogradely labeled nucleocortical neurons.

Fig. 10. Diameter distributions of immunostained, retrograde double labeled, nucleocortical neurons. Two animals with different injection sites (cameralucida drawings, a and b) are shown in A and B. Note that in A the injection was bilateral. Cells were measured and counted in serial sections alternately immunoreacted for GABA (a',b') and Glu (a",b).GABA-IR and Glu-IR neuron distributions are shaded, while the empty parts show the immunonegative (simply retrogradely labeled) nucleocortical neurons. Note that the histograms include only the reciprocal nucleocortical neurons. NM = nucleus medialis; NI = nucleus interpositus; NL = nucleus lateralis. Abscissae: diameter of the nucleocortical neurons in pm (class interval 2.5 pm). Ordinates: numbers of nucleocortical neurons per class.

GABA-IR neurons, colocalizing Glu or not, are inhibitory using GABA as the agonist. Since Glu is a precursor for GABA, its presence in GABA-IR neurons would be expected and was confirmed in this study. Such colocalizations are also found in other brain structures (see Ottersen and Storm-Mathisen, '84). However, some of the GAI3A-IR neurons (24.6%of the G B A - I R neurons sampled in the semithin sections) were shown not to colocalize Glu. This has also been noted in other nervous structures (see Ottersen and Storm-Mathisen, '84;Walberg et al., '90). Glu-IR neurons in the three cerebellar nuclei have a broad distribution of sizes and were not particularly grouped within the nucleus. Glutamate is generally recognized as an excitatory neurotransmitter, but it is also thought to have other metabolic neuronal functions (Fonnum, '84). Thus, the presence of Glu immunoreactivity in a neuron cannot be interpreted to mean that it uses Glu only as a neurotransmitter. Furthermore, the level of immunostaining of this amino acid in somata containing only a metabolic pool of Glu is not necessarily low (Ottersen and Storm-Mathisen, '84). In the semithin sections of this study, a larger

proportion of the cerebellar neurons (56.6%)were found to be only Glu-IR, while a few cerebellar neurons were negative for both GABA and Glu. We postulate that these two groups of neurons of the cerebellar nuclei are excitatory, using Glu or another, as yet unidentified, neurotransmitter. The possibility that another presumed inhibitory neurotransmitter is colocalized with Glu seems small, since taurine is found only exceptionally in the neurons of the cerebellar nuclei (Ottersen, '88; Ottersen et al., '88) and glycine is found only in small neurons of the cerebellar nuclei (Fort et al., '90). From extensive electrophysiological evidence, many neurons from each of the cerebellar nuclei projecting to brainstem structures are known to be excitatory on their target cells (see Ito, '84). Immunohistochemical evidence to identify the excitatory agonist(s)has, so far, not been definitive. In favor of glutamate, Monaghan et al. ('86)report cerebellar nuclei neurons immunoreactive for glutaminase, the enzyme converting glutamine to Glu, and suggest a glutamatergic function. Recently, Kaneko et al. ('89) described scattered immunostained neurons in all the cerebellar

82

C. BATINI ET AL.

nuclei using a monoclonal antibody to the phosphate activated glutaminase thought to be found only in neurons using Glu as neurotransmitter (Kaneko and Mizuno, '88). Colocalization of Glu and aspartate may occur in the cerebellar nuclei as in other structures (Zhang et al., '90) since aspartate containing neurons have been described in the cerebellar nuclei (Aoki et al., '87; Mad1 et al., '87; Kumoi et al., '88). These authors suggest that such neurons use aspartate as the excitatory agonist. Glutamate containing cholinergic neurons have been described in other structures (Ottersen and Storm-Mathisen, '84) and a cholinergic transmission has been suggested for the excitatory interposito-rubral pathway (Nieoullon and Dusticier, '81a,b; Bernays et al., '88).If so, many of the Glu-IR neurons of the NI must also colocalize this neurotransmitter or its precursor.

GABA and Glu immunoreactivity of the nucleocortical neurons The NCN were identified by a cortically injected tracer which was transported only in the retrograde direction and whose histochemically revealing reaction did not affect the subsequent immunohistochemistry. The immunoreactivity of the NCN to GABA (from two sources) and Glu antibodies was evaluated in thick sections using the criteria defined above. Two distinct populations of identified NCN were present, one was the homogeneous group of small GABA-IR cells, and the other was the more heterogeneous sized group of Glu-IR cells. While the mean size of each population was distinct, some overlapping of the sizes was present (cfr Fig. 10). This overlapping can be accounted for by the GABA and Glu colocalizing cells confirmed in the thin sections of non-retrograde labeled cerebella. In the thick sections, GABA-IR hemisectioned cells could not be clearly identified in the facing Glu reacted sections. Unfortunately, penetration of the antibodies was limited to about 10 bm of the two surfaces of the 40 bm thick sections leaving about 50% of the section with only retrogradely labeled neurons. Since all the retrograde labeled cells were counted, the proportions of the NCN counted as GABA-IR or Glu-IR were underestimated. In thick sections of the cerebellar nuclei, only 8.7% of the total identified NCN were GABA-IR, whereas in the semithin sections, 31.7% of the identified neurons (see Table 1) were found to be GABA-IR. Even with a correction for 50% underestimation, the GABA-IR NCN are still only about half of that found in the semithin sections. The difference is likely to be accounted for by GABAergic neurons projecting to the inferior olive (Angaut and Sotelo, '89; BuisseretDelmas et al., 1989; Nelson and Mugnaini, '89) as well as by those reported to be intrinsic neurons (Wassef et al., '86). All these neurons contain sufficient GABA to be considered GABAergic and therefore inhibitory, whether or not they colocalize Glu. While the same conclusion was reached earlier using retrograde transport of a GAD antigenantibody complex (Chan-Palay et al., '79; see Chan-Palay, '821, the present results using specific antibodies adds new quantitative data. The proportion of Glu-IR NCN was five times (see Fig. 8C) that of the GABA-IR, being 47.7% of the total population of identified NCN. For the neurons sampled in the semithin sections (non-retrogradely marked in Table 11, the proportion of Glu-IR neurons was 80.5%.Their size distributions were almost identical (compare the histograms of

Figs. 5 and 9). If the Glu-IR NCN were underestimated by about 50%in the thick sections, their real proportion will be close to that found for the semithin sections. The present results do not allow the interpretation that this group of neurons use Glu as neurotransmitter, but their excitatory function is indirectly suggested since 1) many did not contain GABA, the major inhibitory neurotransmitter (present results) but they may colocalise another excitatory neurotransmitter (see above); 2) from other published work (see above) they probably do not colocalize some other inhibitory neurotransmitter; and 3) some of them send branches to extracerebellar structures (Tolbert et al., '76; McCrea et al., '781, where they have an excitatory function (see Ito, '84). We therefore postulate that the reciprocal and the symmetrical projections of the nucleocortical pathway consist of two contingents with opposite function, one inhibitory and the other excitatory. Other data support this conclusion: the NCN terminate as mossy fibres (Chan-Palay, '77; Dietrichs and Walberg, '79; Tolbert et al., '80; Hamori et al., '81) which are generally considered as excitatory (see Ito, '84); most of the mossy fibres have been shown to be not only strongly Glu immunoreactive (Somogyi et al., '86; Clemens et al., '87) but also glutaminase immunoreactive (Kaneko et al., '89) and therefore presumably excitatory. GABAergic mossy terminals of the NCN are also present (Chan-Palay, '82; Hamori and Takacs, '89) and therefore may be presumed inhibitory.

Functional implications Stimulation of the cerebellar cortex is known to produce antidromic invasion of the NCN (Ito et al., '64; Tolbert et al., '76). The excitatory or inhibitory function of the NCN is not electrophysiologicallyestablished since specific orthodromic stimulation of the NCN is made difficult by the presence of fibres of extracerebellar origin in the nuclei also terminating as mossy fibers. If, as discussed above, the terminals of the GABA-IR and Glu-IR NCN are mossy fibers, they are both synaptic inputs to the granule cells (see Hamori and Takacs, '89). The granule cells will receive inhibitory and excitatory inputs from the reciprocal nucleocortical pathway. Their outputs, being excitatory through the parallel fibers on the Purkinje cells, will be respectively disfacilitatory and facilitatory on the Purkinje cells. If it is assumed that all nuclear neurons, including the GABA-IR, receive innervation from ipsilatera1 inhibitory Purkinje cells (see Ito, '841, the net effect within the reciprocal feedback loop will be dependent on the activity of the Purkinje cell itself. This reciprocal feedback loop, representing a short-path self-regulatory system, can respond very rapidly, thus contributing to fine control of the Purkinje cell activity. The circuit will therefore function as a local microintegrator for the longer cerebellofugal pathways. The fact that the proportions of GABA- and Glu-IR neurons were nearly the same in the symmetric and reciprocal projecting nuclei could also be interpreted to mean that the symmetric projections are collaterals of reciprocal neurons. In terms of feedback loops, symmetric projections arising from collaterals of the reciprocal neurons would give the open and closed feedback loops a common origin. This possibility has important implications for the function of these pathways. The question of the origins of the symmetric and reciprocal connections is presently under investigation.

GABA,GLUINCEREBELLARANDNUCLEOCORTICALNEURONS

ACKNOWLEDGMENTS We wish to thank Dr R.T. Kado for critical reading and English correction of the manuscript.

LITERATURE CITED Angaut, P., and C. Sotelo (1989) Synaptology of the cerebello-olivary pathway. Double labeling with anterograde axonal tracing and GABA immunocytochemistry in the rat. Brain Res. 479:361-365. Aoki, E., R. Semba, K. Kato, and S. Kashiwamata (1987) Purification of specific antibody against aspartate and immunocytochemical localization of aspartatergic neurons in the rat brain. Neuroscience 21:755-765. Basbaum, A.I., and D. Menetrey (1987) Wheat germ agglutinin-apoHRP gold: A new retrograde tracer for light and electron-microscopic singleand double-label studies. J. Comp. Neurol. 261r306-318. Batini, C., C. Buisseret-Delmas, C. Compoint, and H. Daniel (1989) The GABAergic neurons of the cerebellar nuclei in the rat: projections to the cerebellar cortex. Neurosci. Lett. 99.351-256. Batini, C., C. Buisseret-Delmas, J. Corvisier, 0. Hardy, and D. JassikGerschenfeld (1978) Brainstem nuclei giving fibers to lolules VI and VII of the cerebellar vermis. Brain Res. 153241-261. Brodal, A., and K. Kawamura (1980) Olivocerebellar projections: a review. Adv. Anat. Embryol. Cell Biol. 64.1-140. Bernays, R.L., L. Heeb, M. Cuenod, and P. Streit (1988) Afferents to the rat red nucleus studied by means of D-(3H)aspartate, (3H)choline and non-selective tracers. Neuroscience 26:601-6 19. Buisseret-Delmas, C., and P. Angaut (1989) Anatomical mapping of the cerebellar nucleocortical projections in the rat: A retrograde labeling study. J. Comp. Neurol. 288r297-310. Buisseret-Delmas, C., C. Batini, C. Compoint, H. Daniel, and D. Menetrey (1989) The GABAergic neurons of the cerebellar nuclei: projections to the inferior olive and to the bulbar reticular formation. Exp. Brain Res. Z7:106-110. Chan-Palay, V. (1977) Cerebellar Dentate Nucleus. Berlin: Springer-Verlag, 548 pp. Chan-Palay, V. (1982) Neurotransmitters and receptors in the cerebellum: immunocytochemical localization of glutamic acid decarboxylase, GABAtransaminase, and cyclic GMP and autoradiography with 3H-muscimol. Exp. Brain Res. ISuppl.1552-586. Chan-Palay, V., J.Y. Wu, and S.L. Palay (1979) Immunocytochemical localization of -aminobutyric acid transaminase a t cellular and ultrastruct u r d levels. Proc. Natl. Acad. Sci. U.S.A. 4.2067-2071. Clemens, J.R., P.L. Monaghan, and A.J. Beitz (1987) An ultrastructural description of glutamate-like immunoreactivity in the rat cerebellar cortex. Brain Res. 421.343-348. Dietrichs, E. (1981) The cerebellar corticonuclear and nucleocortical projections in the cat as studied with anterograde and retrograde transport of horseradish peroxidase. IV.The paraflocculus. Exp. Brain Res. 44.235242. Dietrichs, E. (1983) The cerebellar corticonuclear and nucleocortical projections in the cat as studied with anterograde and retrograde transport of horseradish peroxidase. V. The posterior lobe vermis and the flocculonodular lobe. Anat. Embryol. (Berl.) 167:449-462. Dietrichs, E., and F. Walberg (1979) The cerebellar corticonuclear and nucleocortical projections in the cat as studied with anterograde and retrograde transport of horseradish peroxidase. I. The paramedian lobule. Anat. Embryol. (Bed.) 158:13-39. Dietrichs, E., and F. Walberg (1980) The cerebellar corticonuclear and nucleocortical projections in the cat as studied with antemgade and retrograde transport of horseradish peroxidase. 11. Lobulus simplex, Crus I and 11. Anat. Embryol. (Berl.) 161:83-103. Fonnum, F. (1984) Glutamate: a neurotransmitter in mammalian brain. J. Neurochem. 52: 1-1 1. 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. Gabbott, P.L.A., J. Somogyi, M.G. Stewart, and J. Hamori (1986) GABAimmunoreactive neurons in the rat cerebellum: A light and electron microscope study. J. Comp. Neurol. 251.474-490. Gould, B.B., and A.M. Graybiel (1976) M e r e n t s to the cerebellar cortex in the cat: evidence for an intrinsic pathway leading from the deep nuclei to the cortex. Brain Res. 110:601-611. Hamori, J., E. Mezey, and J. Szentagoyhai (1981) Electron microscopic

83

identification of cerebellar nucleo-cortical mossy terminals in the rat. Exp. Brain Res. 44~97-100. Hamori, J., and J. Takacs (1989) Two types of GABA-containing axon terminals in cerebellar glomeruli of cat: an immunogold-EM study. Exp. Brain Res. 74.471-479. Haines, D.E., and J.C. Pearson (1979) Cerebellar corticonuclear-nucleocortical topography: a study of the tree shrew (Tupaia) paraflocculus. J. Comp. Neurol. 187:745-758. Hess, D.T. (1982) Cerebellar nucleo-cortical neurons projecting to the vermis of lobule VII in the rat. Brain Res. 248:361-366. Hodgson, A.J., B. Penke, A. Erdei, I.W. Chubb, and P. Somogyi (1985) Antisera to g-aminobutyric acid. I Production and caracterisation using a new model system. J. Histochem. Cytochem. 33229-239. Hsu, S.M., L. Raine, and H. Fanger (1981) The use of avidin-biotinperoxidase complex (ABC) in immunoperoxidase techniques: a comparison between ABC and unlabeled antibody (PAP) procedure. J. Histochem. Cytochem. 29r577-580. Ito, M. (1984) The Cerebellum and Neural Control. New York: Raven press, 580 pp. Ito, M., M. Yoshida, and K. Obata (1964) Monosynaptic inhibition of the intracerebellar nuclei induced from the cerebellar cortex. Experientia 20r575-5 76. Kaneko, T., K. Itob, R. Shigemoto, and N. Mizuno (1989) Glutaminase-like immunoreactivity in the lower brain stem and cerebellum of the adult rat. Neuroscience 32~79-98. Kaneko, T., and N. Mizuno (1988) Immunohistochemical study of glutaminase-containing neurons in the cerebral cortex and thalamus of the rat J. Comp. Neurol. 267.590-602. Kumoi, K., N. Saito, K. Takayushi, and C. Tanaka (1988) Immunohistochemical localization of -aminobutyric acid and aspartate-contaming neurons in the rat deep cerebellar nuclei. Brain Res. 439:302-310. Madl, J.E., A.J. Beitz, R.L. Johnson, and A.A. Larson (1987) Monoclonal antibodies specific for fixative-modified aspartate: immunocytochemical localization in the rat CNS. J. Neurosci. 7.2639-2650. Maley, B., and B. Newton (1985) Immunohistochemistry of gaminobutyric acid in the cat nucleus tractus solitarius. Brain Res. 33Or364-368. McCrea, R.A., G.A. Bishop, and S.T. Kitai (1978) Morphological and electrophysiological characteristics of projection neurons in the nucleus interpositus of the cat cerebellum. J. Comp. Neurol. 181:397420. Monaghan, P.L., A.J. Beitz, A.A. Larson, R.A. Altschuler, J.E. Madl, and M.A. Mullett (1986) Immunocytochemical localization of glutamate-, glutaminase- and aminotransferase-like immunoreactivity in the rat deep cerebellar nuclei. Brain Res. 363:364-370. Mugnaini, E., and W.H. Oertel (1985) An atlas of the distribution of GABAergic neurons and terminals in the rat CNS as revealed by GAD immunohistochemistry. In A. Bjorklund and T. Hokfeld (eds): Handbook of Chemical Anatomy. GABA and Neuropeptides in the CNS, Vol. 4. Amsterdam: Elsevier, pp. 436-608. Nelson, B.J., and E. Mugnaini (1989) Origins of GABAergic inputs to the inferior olive. In P. Strata (ed): The Olivocerebellar System in Motor Control. Berlin: Experimental Brain Research, vol. 17, pp. 86-107. Nieoullon, A., and N. Dusticier (1981a) Decrease in choline acetyltransferase and in high affinity glutamate uptake in the red nucleus of the cat after cerebellar lesions. Neurosci. Lett. 24.267-271. Nieoullon, A., and N. Dusticier (1981b) Decrease in choline acetyltransferase activity in the red nucleus of the cat after cerebellar lesions. Neuroscience 6: 1633-1641. Oertel, W.H., D.E. Schmechel, E. Mugnaini, M.L. Tappaz, and I.J. Kopin (1981) Immunocytochemical localization of glutamate decarboxilase in rat cerebellum with a new antiserum. Neuroscience 62715-2735. Oertel, W.H., D.E. Schmechel, E. Mugnaini, M.L. Tappaz, and I.J. Kopin (1982) The immunocytochemical demonstration of gamma-aminobutyric acid-ergic neurons-methods and application. In Chan-Palay and S.L. Palay (eds): Cytochemical Methods in Neuroanatomy. New York Alan R. Liss, Inc., pp. 297-329. Ottersen, O.P. (1988) Quantitative assessment of taurine-like immunoreactivity in different cell types and processes in rat cerebellum: an electronmicroscopic study based on a postembedding immunogold labelling procedure. Anat. Embryol. (Berl.) 178.407-421. Ottersen, O.P., S. Madsen, J. Storm-Mathisen, P. Somogyi, L. Scopsi, and L.-I. Larsson (1988)Immunocytochemical evidence suggest that taurine is colocalized with GABA in the Purkinje cell terminals, but that the stellate cell terminals predominantly contain GABA a light- and electronmicroscopic study of the rat cerebellum. Exp. Brain Res. 72:407-416.

84 Ottersen, O.P., and J. Storm-Mathisen (1984) Glutamate- and GABAcontaining neurons in the mouse and rat brain, as demonstrated with a new immunocytochemical technique. J. Comp. Neurol. 229:374-392. Payne, J.N. (1983) The cerebellar nucleo-cortical projection in the rat studied by the retrograde fluorescent double-labeling method. Brain Res. 271:141-144. Somogyi, P., K. Halasi, J. Somogyl, J. Storm-Mathisen. and O.P. Ottersen (1986) Quantification of immunogold labeling reveals enrichment of glutamate in mossy and parallel fibres terminals in cat cerebellum. Neuroscience 19t1045-1050. Sternberger, L. (1986) The unlabeled antibody peroxidase-antiperoxidase (PAP) method. In Immunocytochemistry, 3rd Edn. New York: John Wiley, pp. 90-209. Tolhert, D.L., and H. Bantli (1979) An HRP and autoradiographic study of Cerebellar corticonuclear nucleocortical reciprocity in the Monkey. Exp. Brain Res. 36:563-571. Tolbert, D.L., and H. Bantli (1980) Uptake and transport of 3H-GABA (g-aminobutyric acid) injected into the cat dentate nucleus. Exp. Neurol. 70:525-538. Tolhert, D.L., H. Bantli, and J.R. Bloedel(1976) Anatomical and physiological evidence for a cerebellar nucleo-cortical projection in the cat. Neuroscience I ;205-2 17.

C. BATINI ET AL. Tolbert, D.L., H. Bantli, and J.R. Bloedel (1977) The intracerebellar nucleocortical projection in a primate. Exp. Brain Res. 30:425-434. Tolbert, D., K. Kultas-Ilinsky, and I. Ilinsky (1980) EM-autoradiography of cerebellar nucleocortical terminals in the cat. Anat. Embryol. (Berl.) 161:215-223. Trott, J.R., R. Apps, and D.M. Armstrong (1990) Topographical organization within the Cerebellar nucleocortical projection to the paravermal cortex of lobules Vb/c in the cat. Exp. Brain Res. 80;415-428. Umetani, T. (1990) Topographic organization of the cerebellar nucleocortical projection in the albino rat: an autoradiographic orthograde study. Brain Res. 5 0 7 2 1 6 2 2 4 . Walberg, F., O.P. Ottersen and E. Rinvik (1990) GABA, glycine, aspartate, glutamate and taurine in the vestibular nuclei: a n immunocytochemical investigation in the cat. Exp. Brain Res. 79:547-563. Wassef, M., J. Simons, M.L. Tappaz, and C. Sotelo (1986) Non-Purkinje cell GABAergic innervation of the deep cerebellar nuclei: a quantitative immunocytochemical study in C 57BL and in Purkinje cell degeneration mutant mice. Brain Res. 395125-135. Zhang, N., F. Walberg, J.H. Laake, B.S. Meldrum, and O.P. Ottersen (1990) Aspartate-like immunoreactivities in the inferior olive and climbing fibre system: a light microscopic and semiquantitative electron microscopic study in rat and baboon (Papio Anubis). Neuroscience 38:61-80.