- Email: [email protected]
Non-Purkinje cell GABAergic innervation of the deep cerebellar nuclei: A quantitative immunocytochemical study in C57BL and in Purkinje cell degeneration mutant mice
Brain Resea'~ch, 399 (1986) 125-135 Ei;evier
125
BRE 12223
Non-Purkinje Cell GABAergic Innervation of the Deep Cerebellar Nuclei: A Quantitative Immunocytochemical Study in C57BL and in Purkinje Cell Degeneration Mutant Mice* MARION WASSEF l, JACQUES SIMONS 1, MARCEL L. TAPPAZ2 and CONSTANTINO SOTELO 1 1Laboratoire de Neuromorphologie (I. N. S. E. R.M. U-106), HOpital de la SalpOtri~re, Paris (France) and 21.N.S.E.R.M. U-171, Saint Genis Laval (France)
(Accepted 20 May 1986) Key words: Cerebellum- Mutant m o u s e - y-Aminobutyric acid (GABA) --Purkinje c e l l Deep cerebellar nucleus --Immunocytochemistry
Purkinje cell degeneration (pcd) mutant mice, 3-4 months old, were used to identify and quantify the non-Purkinjecell GABAergic innervationof deep cerebellar nuclei. Glutamic acid decarboxylase (GAD) immunoreactive structures appeared as dark dots throughout the 4 nuclei. Ultrastructural examination confirmed that each dot corresponded to an axon terminal. GAD-labeled boutons were large, contained tightly packed flattened vesicles and established Gray type II synapses with all nuclear neuronal populations. Thus, cytological criteria did not distinguish between Purkinje cell and non-Purkinje cell GAD-positive nerve terminals, since they shared many common features. The number of GAD-immunoreactive axon terminals in the deep nuclei of pcd cerebella was compared to that of normal C57BL mice. Despite an almost complete disappearance of Purkinje cells in the pcd mouse (less than 0.05% of these neurons remained in the mutants), the surface density of GAD-positive nerve terminals in the deep nuclear region was 37% of control value. Taking into account a volumetric decrease of 58% for the deep nuclei of the mutant cerebellum, we estimated the percentage of GAD-positive boutons innervating these nuclei to be 15% of normal values. This important residual innervation of the deep nuclei might arise from local GABAergic neurons, which were identified in the normal and mutant cerebella by immunostaining with an antiGABA antibody.
INTRODUCTION
compared to the contralateral intact side 5 and a 73% decrease in the n u m b e r of G A D - i m m u n o r e a c t i v e
Most Purkinje cell (PC) axons terminate in the cerebellar nuclei and in the dorsal part of the lateral vestibular nucleus (Deiters' nucleus). PC axons have been shown to form inhibitory synapses 1°, using ~,aminobutyric acid ( G A B A ) as a neurotransmitter 16'17. In confirmation of this finding, a decrease in G A B A content 2° and in the activity5'2° or i m m u n o r e activity9 of the G A B A - s y n t h e s i z i n g enzyme, glutamate decarboxylase ( G A D ) , has been observed in the deep cerebellar and Deiters' nuclei after partial or subtotal elimination of PC input, by cortical lesioning in cat and rat cerebella 5'2°. After appropriate cortical lesioning, there was a 70% decrease in G A D activity in the interposed nucleus on the lesioned, as
terminals in the dorsal part of the vestibular nucleus of a lesioned rat as compared to a control animal9. At least part of the remaining G A D activity and immuZ noreactivity may be due to the small G A D - i m m u n o reactive neurons which have been demonstrated to exist in the deep nuclei9'15'18. The pcd m u t a n t mouse seemed appropriate for the evaluation of n o n - P C G A B A e r g i c innervation of the deep nuclei, since in this m u t a n t PCs degenerate progressively starting at 21 days after birth. A t 2 months, less than 1% of the n o r m a l PC population remains, mostly in the nodulus 14. According to Roffier-Tarlov et at. 2°, the weight of the deep nuclei in pcd mice diminishes by 35% at 45 days. At 28 days, a 50% fall in
* Part of this work has been presented in abstract form25. Correspondence: M. Wassef, I.N.S.E.R.M. U-106, H6pital de la Salp6tri6re, 75651 Paris, Cedex 13, France. 0006-8993/86/$03.50 © 1986 Elsevier Science Publishers B.V. (Biomedical Division)
126 G A B A concentration in the deep nuclei occurs in pcd mutants as compared to normal littermates. No further changes in these two parameters have been observed up to 80 days after birth. We used pcd and C57BL control mice between 3 and 4 months old in order to avoid the period of decrease in deep nuclear volume and to obtain an almost total absence of PCs. Different parameters were measured. (1) The volumes of the deep nuclei in pcd and C57BL mice were calculated from their surface area on sections of paraffin-embedded cerebella. (2) The number of remaining PCs were counted in 3 pcd cerebella after staining with an immune serum (anti-cyclic GMP-dependent protein kinase (cGK)) which specifically labels PCs. (3) The surface density of GAD-immunoreactive boutons was obtained from the deep nuclei of pcd and C57BL mice by postembedding techniques on 1-/~m-thick sections of plastic-embedded cerebella. Two additional elements were also studied: (a) the comparative distribution of G A B A immunoreactive neurons in the deep nuclei of both pcd and C57BL; and (b) the ultrastructural features and the synaptology of GAD-positive boutons in the pcd deep nuclei, in order to determine whether GABAergic nerve terminals of PC origin could be differentiated from those of non-PC origin. MATERIALS AND METHODS
Animals C57BL mice were obtained from I F F A - C R E D O (Lyon), and pcd mutants were obtained from Dr. J.L. Guenet (Institut Pasteur, Paris). All animals were between 3 and 6 months old.
Histological techniques Fixation. Mice were anesthetized with ether and perfused through the ascending aorta with a fixative solution containing 4% formaldehyde in 0.12 M sodium Ehosphate buffer, pH 7.2-7.4. In some cases 0.01 M periodic acid and 0.075 M lysine were added according to the method of McLean and Nakane 13. Two additional pcd mice were fixed for electron microscopy (4% paraformaldehyde and 0.1% glutaraldehyde in the same buffer solution). The brains were dissected out and a block containing the whole cerebellum was immersed overnight in the formaldehyde
fixative at 4 °C. For the detection of G A B A immunoreactivity, a stronger fixative was used, containing 5% glutaraldehyde in the same phosphate buffer (one pcd and one C57BL). Embedding. After a rapid rinse in phosphate-buffered saline (PBS), 3 C57BL and 3 pcd cerebella were embedded in paraffin and cut at 10/~m thickness. The paraffin sections stained with Cresyl violet were used to evaluate the volume of the deep nuclei. The two pcd cerebella for electron microscopy were cut with a Vibratome into 50-~m-thick sections, immunostained with the anti-GAD antiserum, osmicated and flat embedded in Araldite. From two C57BL and two pcd cerebella, 300-400~m-thick sections were cut with a Vibratome and flat embedded in Araldite. From these blocks, 1-j~mthick sections of the cerebellum, separated by at least 3 j~m, were collected on gelatin-coated slides. Araldite was removed with sodium methanolate, according to the method of Lane and Europa 11. A 5-min pretreatment with 3% H202 was used to eliminate endogenous peroxidase. The slides were then treated for 15 rain at 37 °C with a solution containing 1/~g/ml of pronase (Sigma). After immunostaining, the semithin sections were used to count labeled boutons or to observe GABA-immunoreactive neurons. Frozen sections. After fixation, the cerebella were infiltrated for 2 days at 4 °C in 3 changes of 30% sucrose in sodium phosphate buffer. Twenty-five-j~mthick sections were cut on a freezing microtome. Frozen sections were processed for G A D or cGK immunostaining. In preliminary experiments some of the GAD-immunoreacted sections were flat embedded in Araldite; 1-~m-thick sections were cut from these blocks and the immunoreactive terminals were counted. Usually, the immunoreacted sections were mounted on gelatin-coated slides and dried overnight. They were then immersed for 30 s in osmic acid, 0.01%, in distilled water, dehydrated and mounted with coverslips. Immunocytochemistry. After removal of Araldite and pronase pretreatment, mounted semithin sections were processed for G A D and G A B A immunocytochemistry according to the method of Sternberger et al. 23. Freely floating frozen sections were similarly processed for cGK, except that 0.25% Triton X-
127 100 was added to the antisera. Antisera. The anti-GAD antiserum raised in sheep has been extensively characterized t8,19. Anti-GABA was purchased from Immunotech (Marseille, France). It was raised in rabbit against reduced G A B A linked to bovine serum albumin with glutaraldehyde 21. Anti-cGK 4'12'24 (gift of Pr. P. Greengard) and antiG A B A 21 are rabbit antisera and have been previously characterized. These antisera were applied overnight at room temperature at a dilution of 1/2000 for anti-GAD, 1/3000 for anti-GABA, and 1/1000 for anti-cGK. Anti-rabbit immunoglobulin (IgG) produced in goat and anti-goat IgG produced in rabbit were purchased from Cappel and used at 1/100 dilution. Rabbit peroxidase-antiperoxidase (PAP) (UCB, France) and goat PAP (Sternberger-Meyer) were used at 1/200 dilution.
Measures. (1) Volume of the deep nuclei in pcd and C57BL. The surface of the deep nuclear area in one or both hemicerebella was measured on every fifth 10-/~mthick paraffin section with a camera lucida and a Hewlett-Packard microcomputer interfaced with a graphics tablet. The volume of the deep nuclei was then calculated in 3 pcd and 3 C57BL mice. (2) Density of GAD-immunoreactive terminals. GAD-immunoreactive terminals were counted on the 555/~m z squares of a grid with a x 100 oil-immersion objective on semithin sections of the cerebellar nuclei of two C57BL and two pcd mice. In each animal, sections separated by at least 3 ktm were obtained from 3 different levels of the deep nuclei. (3) Number of cGK-immunoreactive PCs in pcd. The cerebella of 3 pcd mice were cut at 25/~m on a freezing microtome and reacted freely floating for cGK immunocytochemistry. The sections were mounted on gelatin-coated slides and the immunoreactive cell bodies were counted. RESULTS
Technical considerations In a preliminary attempt to count labeled terminals in the deep nuclei, 25-/~m-thick frozen sections were immunoreacted for G A D , freely floating, and flat embedded in Araldite. The counts were made on
semithin sections cut from these blocks. However, we always observed differences in staining intensity between the superficial and internal parts of the blocks, even if the frozen sections were incubated overnight with 0.25% Triton X-100 added to the primary antiserum. The later use of desaraldited semithin sections gave an even immunoreactivity throughout the whole extent of the deep cerebellar region in cerebella cut on the coronal plane and permitted a good sampling of the nuclear region (Figs. 2A, B and 3A, B). Boutons in the deep nuclei of C57BL are very tightly packed, and were probably underestimated as compared to pcd. Labeling of the cellular bodies of GABAergic interneurons was more intense with anti-GABA than with anti-GAD (Figs. 2 and 3). However, the G A B A immunoreactivity of PC axons in the white matter and around and inside the deep nuclear region prevented the counting of individual terminals. On the contrary, G A D was highly concentrated in axon terminals but presynaptic axonal segments were almost devoid of labeling, thus allowing the enumeration of individual boutons.
Cerebellar cortex Number and localization of remaining PCs in pcd. In C57BL mice, cyclic GMP-dependent protein kinase labeled all PCs, and exclusively this cell population, in the cerebellum. In two 4-month-old pcd mutants, the number of labeled cellular bodies was 103 and 112, respectively. As previously described, most of the remaining PCs were located in the nodulus (Fig. 1C). In one 2-month-old pcd, the number of PCs was 1053. Comparison of GABAergic cortical innervation in pcd and C57BL. The thickness of the cortex was reduced in pcd mice as compared to C57BL due to the elimination of PCs and to the progressive death of granule cells 7. G A D immunostaining of the terminals in the cortex was always more intense in pcd than in C57BL mice cerebella processed in parallel (Fig. 1A, B). The short alignments of immunoreactive boutons observed in the molecular layer of normal mice (Fig. 1B) probably represent stellate cell axon terminals or ascending collaterals of basket cell axon terminals in synaptic contact with the shaft of PC dendrites. In the molecular layer of pcd, GAD-immunoreactive termi-
128
Fig. 1. A and B: semithin sections through the cerebella ofpcd mutant (A) and control C57BL (B) mice. In C57BL mice (B), Purkinje cells are slightly G A D immunoreactive (asterisk). The organization of the cortical layers (molecular layer, m; Purkinje cell layer, PC; granule cell layer, gr) is preserved in pcd mutants. In spite of the absence of Purkinje cells in these animals, the PC layer remains as a row of empty pinceau formations. A, x420; B, x420. C: general view of the cerebellum of a 4-month-old pcd mouse immunostained with anti-cGK to show PCs (indicated by arrowheads). Two PCs are shown at higher magnification in the inset, x45; inset, x 150.
129
Fig. 2. Semithin sections through the deep nuclear region of C57BL (A) and pcd (B) mice. A: in C57BL, the somata and thick dendrites are underlined with an almost continuous row of GAD-immunoreactive terminals, x600. B: at the same magnification, the GABAergic innervation of pcd deep nuclei, although reduced when compared to C57BL (A), is still appreciable. This is due in part to the almost 60% decrease in deep nuclear volume in the mutant, x600. C: higher magnification of the GAD-immunoreactive terminals in the pcd mutant, x 1800.
130 nals are more uniformly distributed (Fig. 1A). Lightly stained Purkinje cell bodies were surrounded by an almost continuous ring of terminals in C57BL, and the basket fibers of the pinceau formation were always well stained. In pcd mice (Fig. 1A), there was an apparent increase in the number of pinceau formations which in some regions seemed even longer than their normal counterparts, as is the case in the hemispheres of the nervous cerebellum 22. The density of isolated labeled terminals was lower in the PC layer than in the adjacent molecular and granular layers. Golgi cell terminals were reduced in number in the granular layer of pcd but were intensely stained and typical rosettes were frequent. With anti-GAD serum, GABAergic perikarya were at most lightly stained (Figs. 1A, B and 2A, B). On the contrary, with anti-GABA serum, GABAergic deep nuclear neurons (arrows in Fig. 3A, B and arrowheads in Fig. 3C, D) and Golgi cells in the cortex (arrowheads in Fig. 3A, B) were intensely labeled. Interestingly, PC bodies were almost unstained with anti-GABA although their axons were intensely labeled as soon as they emerge from the cell bodies. This suggests a different compartmentation of G A B A in PCs.
Deep nuclei Volume of the deep nuclei. The deep nuclear region was more easily delineated in pcd than in C57BL mice. For this reason, the nuclear volume was calculated in both hemicerebella in 3 C57BL and in only one in 3 pcd (Table I). The nuclear volume in pcd decreased to 42% of control volume. GAD and GABA immunoreactivity in the deep nuclei. As expected, the density of GAD-immunoreactive terminals was greatly decreased in the deep nuclei of pcd. In C57BL, neurons and thick dendrites were delineated by an almost continuous row of immunoreactive boutons. In pcd, GAD-positive dots were scattered randomly. In both mouse types, antiG A B A intensely stained many neurons which were not easily seen with anti-GAD without a colchicine pretreatment. In C57BL, Purkinje cell axons in the white matter and inside the nuclear region were also G A B A immunoreactive. The white matter in pcd mice was devoid of G A B A immunoreactive fibers except for some labeled axons near the midline. Ultrastructure of GAD-positive axon terminals in pcd deep nuclei. In the pcd mutant, the perikaryal
surface of the large neurons was partially covered with a small number of axosomatic synapses, which appeared either grouped into small clusters of 2-3 boutons or isolated. Some of these synapses always exhibited G A D immunoreactivity. The labeled axon terminals were large and elongated, with a largest diameter of about 5/~m and a smallest diameter between 1 and 2/zm (Fig. 4 A - D ) . They contained a pleomorphic population of tightly packed synaptic vesicles, the vast majority of which were flattened. At the synaptic interface, pre- and postsynaptic membranes established several patches of synaptic junctions with poorly marked postsynaptic differentiations, as in type II synapses (Fig. 4 A - D ) . The small neurons also received GAD-positive synapses on their perikarya. Generally, these axon terminals were smaller than those synapsing on the large cells. They appeared as bulbous ends filled with pleomorphic vesicles and established type II synaptic contacts (Fig. 4B). Large (3-4 btm in diameter) and smaller (1-2/~m in diameter) GAD-positive axon terminals were frequently encountered in synaptic contact with dendrites of either large or small neurons. Although the main dendritic stems were privileged sites for axodendritic GAD-positive inputs, labeled boutons also established synaptic contacts on thin dendritic elements (Fig. 4C) and on sessile spinous protrusions (Fig. 4D). These axon terminals shared their intrinsic features with those at somatic locations, and, like them, established mainly type II synapses. Only occasionally did they establish synapses of an intermediate type, characterized by a somewhat more marked postsynaptic differentiation (Fig. 4C). These labeled boutons were present in all 4 deep nuclei, especially in the lateral nucleus.
Number of GABAergic boutons in pcd and C57BL deep nuclei. The surface density of GAD-immunoreactive boutons in the deep nuclei of pcd was about one-third of that observed in C57BL (Table II). Although direct observation of the sections gave the impression that GABAergic innervation was more uniform in pcd than in C57BL, mean variation was about 25% in both cases. Taking into account the decrease in deep nuclear volume in pcd, we calculated that the total number of GABAergic boutons in pcd deep nuclei was 15% of that in C57BL.
131
Fig. 3. Semithin sections of C57BL (A, C) and pcd (B, D) deep nuclear regions stained with anti-GABA serum. A and B: general view. Immunoreactive neurons (arrows) are scattered throughout the deep nuclei of C57BL (A) and pcd (B). In the cortex, Golgi cells are intensely stained (arrowheads in A and B). In the white matter of C57BL mice (wm in A), PC axons are G A B A immunoreactive. A and B, x 115. C and D: detail of GABA-immunoreactive structures. GABAergic neurons (arrowheads) are completely stained, with the exception of the nucleolus. Their small size can be appreciated by comparing them to unstained neurons (N). The higher background in C as compared to D is probably due to leakage of G A B A (highly concentrated in normal mice) during fixation. C and D, × 420.
132 TABLE I Mean volume of deep nuclei in C57BL and PCD mice Animal
Mean volume + S.E.M. (108~m 3)
C57BL (n = 6) 4.5 + 0.84 PCD(n=3) 1.9+0.17
PCD volume/C57BL volume
0.42
DISCUSSION The existence of a non-Purkinje cell GABAergic innervation in the cerebellar deep nuclei was hypothesized as an explanation of the rather high G A D activity remaining in the deep nuclei after deafferentation from Purkinje cells 5'6. After large cortical lesions or in pcd mutants, G A D activity is never lowered to less than 30% of control values. In contrast, transection of the striatonigral tract at the premammillary level causes a 90% reduction of G A D activity in the pars reticulata of the substantia nigra 2. However, it is difficult to appreciate the importance of non-PC GABAergic innervation of the deep nuclei in animals with large cortical lesions because the PC projection to the deep nuclei is somewhat diffuse, although ordered. According to Fonnum et al. 5, it is impossible to destroy all the PCs which send their axons to a particular region of the deep nuclei. For this reason, pcd mice seem more appropriate for this evaluation. The number of PCs in C57BL was estimated by Herrup and Mullen s to be 196,000. Using this value, we can calculate that only 0.5% of the normal PC population remains in 2-month-old pcd mutants; this drops to 0.05% in 4-month-old animals. In pcd mice, the deep nuclear volume decreases to 42% of control value (Table I), more than would be expected. According to Roffier-Tarlov et al. 2°, the weight of the pcd cerebellum decreases to 65% of control value at 79-85 days. A simple subtraction of the volume occupied by PC axons and boutons does not entirely explain this reduction. In the cat, for example, the entire population of axons and boutons occupies less than 60% of the total deep nuclear volume 6. The massive deafferentation of the deep nuclei might induce a shrinkage or a moderate loss of deep nuclear neurons. This decrease in volume is probably responsible for the high concentration of G A B A (55% of con-
trol) found by Roffier-Tarlov et al. 2° in pcd deep nuclei and for the relatively high density of GAD-immunoreactive terminals (37% of control) found in the present study. The discrepancy between the two values could be due to the fact that in the study of Roffler-Tarlov et al., deep nuclear samples contained up to 30% of cerebellar cortex in which there was no important decrease in G A B A concentration 2°. Many factors could have infuenced our estimation of non-PC GABAergic innervation. Contribution o f remaining PCs
The contribution of the 0.5-0.05% of PCs remaining in pcd mice to the GABAergic innervation of the deep nuclei is probably negligible because: (1) most of them are located in the nodulus and thus merely project to the vestibular nuclei, with the exception of Deiters' nucleusl; (2) at 4 months, remaining PCs seem altered when observed with anti-cGK (Fig. 1C); and (3) PC axon terminals in the deep nuclei degenerate earlier than the corresponding PC somata and dendrites in the cerebellar cortex 2°. Therefore, PC projections to the deep nuclei are probably inexistent in 4-month-old pcd mice. Sprouting o f n o n - P C G A B A e r g i c axons
There is some evidence of axon sprouting in the deep nuclei of pcd mice following PC axon loss. According to Roffier-Tarlov et al. 2°, the somatic circumference of the large neurons of the lateral cerebellar nucleus does not decrease in pcd mice. The proportion of the circumference occupied by synaptic terminals (40% of control C57BL mice) increases from 7% in 23-day-old pcd mice to 12% in 60-day-old animals. We have observed (unpublished data) a very important increase in the density of serotonin immunoreactive fibers in the deep nuclei of pcd as compared to C57BL mice. However, without a precise calculation we cannot rule out the possibility that the higher density of 5-HT immunoreactive fibers in pcd mice is due simply to the decrease in deep nuclear volume. At present, it is not possible to determine the importance of an eventual sprouting of non-PC G A B A ergic axons.
133
Fig. 4. Electron micrographs of GAD-positive axon terminals in the deep nuclei of the pcd cerebellum. A: lateral nucleus. A large, elongated, labeled axon terminal is in synaptic contact with the perikaryon of a large neuron (LN). The arrows point to the small patches of type II synaptic junctions, x25,000. B: medial nucleus. The immunolabeled axon terminal is directly apposed to the perikaryal surface of a small neuron (SN). x22,000. C: medial nucleus. The slightly labeled axon terminal establishes an intermediate type of synaptic junction (arrows) with a small dendritic profile (SD). x50,000. D: lateral nucleus. The GAD-positive axon terminal is in synaptic contact with the perikaryon of a large neuron (LN), as well as with a small spinous protrusion (S). The arrows mark the type II synaptic complexes, x 31,000.
134 TABLE II Density of GAD-immunoreactive terminals in the deep nuclei of C57BL and pcd mice Animal
555 ktm2 squares counted
Mean number of S.E.M. terminals per square
Numerical density (terminals/lO0 ktm2)
Numerical density pcd/ numericaldensity C57BL
83164 (C57BL) 83165 (C57BL) 83168 (pcd) 83169 (pcd)
154 148 65 145
72 78 27 27
13 14 5 5
0.37
Origin o f non-Purkinje cell G A B A e r g i c innervation o f the deep nuclei The G A B A e r g i c neurons of the d e e p nuclei are the most likely source of non-PC G A B A e r g i c innervation. These neurons are distributed in all 4 subdivi-
sions of the d e e p nuclei in pcd as well as in C57BL mice (Fig. 3). Recently, these neurons were r e p o r t e d to project to the inferior olive, at least in part Is. It is not known whether there are two classes of G A B A e r g i c n e u r o n in the deep nuclei or w h e t h e r the projecting neurons have axon collaterals and are also responsible for local G A B A e r g i c innervation in the d e e p nuclei. Although less p r o b a b l e , the existence of other G A B A ergic afferents cannot be excluded. The present ultrastructural analysis of G A D - p o s i tive axon terminals of the lateral nucleus in pcd mice was a i m e d at characterizing the intrinsic features of this axonal p o p u l a t i o n and at helping to d e t e r m i n e its origin. Chan-Palay 3 has characterized 6 axon terminal types in the rat lateral nucleus and has ascribed a different origin to each. In comparison, our results are rather disappointing. The cytological features of the G A D - p o s i t i v e b o u t o n s are very much like those of class A . F o r C h a n - P a l a y 3, these boutons belong to PC axons, but, as discussed above, this axonal pop-
REFERENCES 1 Angaut, P. and Brodal, A., The projection of the 'vestibulocerebellum' onto the vestibular nuclei in the cat, Arch. Ital. Biol., 105 (1967) 441-479. 2 Brownstein, M.J., Mroz, E.A., Tappaz, M.L. and Leeman, S.E., On the origin of substance P and glutamic acid decarboxylase (GAD) in the substantia nigra, Brain Research, 135 (1977) 315-323. 3 Chan-Palay, V., Cerebellar Dentate Nucleus. Organization, Cytology and Transmitters, Springer, Berlin, 1977, pp. 212-238.
+18 _+19 _+7 _+6
ulation is missing in the m u t a n t cerebellum. M o r e over, the vast majority of l a b e l e d axons might belong to the small G A B A e r g i c neurons present within the deep nuclei, although they are very different from those of Chan-Palay's class E 3, considered as originating from local neurons. Clearly, cytological criteria alone cannot establish the origin of the non-Purkinje cell G A B A e r g i c innervation of the deep nuclei. These results are, however, i m p o r t a n t since they not only prove that the i m m u n o s t a i n e d dots counted in the light micrographs c o r r e s p o n d to axon terminals, but also d e m o n s t r a t e that the G A B A e r g i c innervation forms conventional synapses with all possible postsynaptic targets present in the deep nuclei. In conclusion, the 15% contribution of local neurons to the total G A B A e r g i c innervation of the deep nuclei as calculated in the p r e s e n t study is p r o b a b l y an overestimation. It can, however, be p r o p o s e d as a reasonable approximation. ACKNOWLEDGEMENTS The authors wish to thank Prof. P. G r e e n g a r d for the gift of a n t i - c G K antiserum, D. Le Cren for photographic w o r k and B. A l v a r a d o for improving the English and typing the manuscript.
4 De Camilli, P., Miller, P., Levitt, P., Walter, U. and Greengard, P., Anatomy of cerebellar Purkinje cells in the rat determined by a specific immunohistochemical marker, Neuroscience, 11 (1984) 761-817. 5 Fonnum, F., Storm-Mathisen, J. and Walberg, F., Glutamate decarboxylase in inhibitory neurons. A study of the enzyme in Purkinje cell axons and boutons in the cat, Brain Research, 20 (1970) 259-279. 6 Fonnum, F. and Walberg, F., An estimation of the concentration of v-aminobutyric acid and glutamate decarboxylase in the inhibitory Purkinje axon terminals in the cat, Brain Research, 54 (1973) 115-127.
135 7 Ghetti, B., Alyea, C.J. and Muller, J., Studies on Purkinje cell degeneration (pcd) mutant: primary pathology and transneuronal changes, J. Neuropathol. Exp. Neurol., 37 (1978) 617. 8 Herrup, K. and Mullen, R.J., Regional variation and absence of large neurons in the cerebellum of the staggerer mouse, Brain Research, 172 (1979) 1-12. 9 Houser, R.C., Barber, R. and Vaughn, J.E., Immunocytochemical localisation of glutamic acid decarboxylase in the dorsal lateral vestibular nucleus: evidence for an intrinsic and extrinsic GABAergic innervation, Neurosci. Lett., 47 (1984) 213-220. 10 Ito, M. and Yoshida, M., The origin of cerebellar-induced inhibition of Deiter's neurons. I. Monosynaptic initiation of the inhibitory postsynaptic potentials, Exp. Brain Res., 2 (1966) 330-349. 11 Lane, B.P. and Europa, D.L., Differential staining of ultrathin sections of Epon-embedded tissues for light microscopy, J. Histochem. Cytochem., 3 (1965) 579-582. 12 Lohman, S.M., Walter, U., Miller, P.E., Greengard, P. and De Camilli, P., Immunohistochemical localization of cyclic GMP dependent protein kinase in mammalian brain, Proc. Natl. Acad. Sci. U.S.A., 78 (1981) 653-657. 13 McLean, I.W. and Nakane, P.K., Periodate lysine paraformaldehyde fixative. A new fixative for immunoelectron microscopy, J. Histochem. Cytochem., 22 (1974) 1077-1083. 14 Mullen, R.J., Eicher, E.M. and Sidman, R.L., Purkinje cell degeneration, a new neurological mutation in the mouse, Proc. Natl. Acad. Sci. U.S.A., 73 (1976) 208-212. 15 Nelson, B., Barmack, N.H. and Mugnaini, E., A GABAergic cerebello-olivary projection in the rat, Soc. Neurosci. Abstr., 10 (1984) 539. 16 Obata, K., G A B A in Purkinje cells and motoneurons, Experientia, 25 (1969) 1285. 17 Obata, K. and Takeda, K., Release of y-aminobutyric acid into the fourth ventricle induced by stimulation of the cat's cerebellum, J. Neurochem., 16 (1969) 1043-1047.
18 Oertel, W.H., Mugnaini, E., Schmechel, D.E., Tappaz, M.L. and Kopin, I.J., The immunocytochemical demonstration of gamma-aminobutyric acid-ergic neurons - methods and application. In S.L. Palay and V. Chan-Palay (Eds.), Cytochemical Methods in Neuroanatomy, Alan Liss, New York, 1982, pp. 297-329. 19 Oertel, W.H., Schmechel, D.E., Mugnaini, E., Tappaz, M.L. and Kopin, I.J., Production of a specific antiserum to rat brain glutamic acid decarboxylase by injection of an antigen-antibody complex, Neuroscience, 6 (1981) 2715-2735. 20 Roffler-Tarlov, S., Beart, P.M., O'Gorman, S. and Sidman, R.L., Neurochemical and morphological consequences of axon terminal degeneration in cerebellar deep nuclei of mice with inherited Purkinje cell degeneration, Brain Research, 168 (1979) 75-95. 21 Seguela, P., Geffard, M., Buijs, R.M. and Le Moal, M., Antibodies against 7-aminobutyric acid: specificity and immunochemical results, Proc. Natl. Acad. Sci. U.S.A., 81 (1984) 3888-3892. 22 Sotelo, C. and Triller, A., Fate of presynaptic afferents to Purkinje cells in the adult nervous mutant mouse: a model to study presynaptic stabilization, Brain Research, 175 (1979) 11-36. 23 Sternberger, L.A., Hardy, P.H., Jr., Cuculis, J.J. and Meyer, H.G., The unlabelled antibody method of immunohistochemistry. Preparation and properties of soluble antigen complex and its use in identification of spirochetes, J. Histochem. Cytochem., 18 (1970) 315-333. 24 Walter, U., Miller, P., Wilson, F., Menkes, D. and Greengard, P., Immunological distinction between guanosine 3',5'-monophosphate dependent protein kinases, J. Biol. Chem., 255 (1980) 3757-3762. 25 Wassef, M., Tappaz, M.L., Oertel, W.E. and Sotelo, C., GAD immunocytochemistry in cerebellum of mutant mice with Purkinje cell defect, Neurosci. Lett., Suppl. 10 (1982) 513.
125
BRE 12223
Non-Purkinje Cell GABAergic Innervation of the Deep Cerebellar Nuclei: A Quantitative Immunocytochemical Study in C57BL and in Purkinje Cell Degeneration Mutant Mice* MARION WASSEF l, JACQUES SIMONS 1, MARCEL L. TAPPAZ2 and CONSTANTINO SOTELO 1 1Laboratoire de Neuromorphologie (I. N. S. E. R.M. U-106), HOpital de la SalpOtri~re, Paris (France) and 21.N.S.E.R.M. U-171, Saint Genis Laval (France)
(Accepted 20 May 1986) Key words: Cerebellum- Mutant m o u s e - y-Aminobutyric acid (GABA) --Purkinje c e l l Deep cerebellar nucleus --Immunocytochemistry
Purkinje cell degeneration (pcd) mutant mice, 3-4 months old, were used to identify and quantify the non-Purkinjecell GABAergic innervationof deep cerebellar nuclei. Glutamic acid decarboxylase (GAD) immunoreactive structures appeared as dark dots throughout the 4 nuclei. Ultrastructural examination confirmed that each dot corresponded to an axon terminal. GAD-labeled boutons were large, contained tightly packed flattened vesicles and established Gray type II synapses with all nuclear neuronal populations. Thus, cytological criteria did not distinguish between Purkinje cell and non-Purkinje cell GAD-positive nerve terminals, since they shared many common features. The number of GAD-immunoreactive axon terminals in the deep nuclei of pcd cerebella was compared to that of normal C57BL mice. Despite an almost complete disappearance of Purkinje cells in the pcd mouse (less than 0.05% of these neurons remained in the mutants), the surface density of GAD-positive nerve terminals in the deep nuclear region was 37% of control value. Taking into account a volumetric decrease of 58% for the deep nuclei of the mutant cerebellum, we estimated the percentage of GAD-positive boutons innervating these nuclei to be 15% of normal values. This important residual innervation of the deep nuclei might arise from local GABAergic neurons, which were identified in the normal and mutant cerebella by immunostaining with an antiGABA antibody.
INTRODUCTION
compared to the contralateral intact side 5 and a 73% decrease in the n u m b e r of G A D - i m m u n o r e a c t i v e
Most Purkinje cell (PC) axons terminate in the cerebellar nuclei and in the dorsal part of the lateral vestibular nucleus (Deiters' nucleus). PC axons have been shown to form inhibitory synapses 1°, using ~,aminobutyric acid ( G A B A ) as a neurotransmitter 16'17. In confirmation of this finding, a decrease in G A B A content 2° and in the activity5'2° or i m m u n o r e activity9 of the G A B A - s y n t h e s i z i n g enzyme, glutamate decarboxylase ( G A D ) , has been observed in the deep cerebellar and Deiters' nuclei after partial or subtotal elimination of PC input, by cortical lesioning in cat and rat cerebella 5'2°. After appropriate cortical lesioning, there was a 70% decrease in G A D activity in the interposed nucleus on the lesioned, as
terminals in the dorsal part of the vestibular nucleus of a lesioned rat as compared to a control animal9. At least part of the remaining G A D activity and immuZ noreactivity may be due to the small G A D - i m m u n o reactive neurons which have been demonstrated to exist in the deep nuclei9'15'18. The pcd m u t a n t mouse seemed appropriate for the evaluation of n o n - P C G A B A e r g i c innervation of the deep nuclei, since in this m u t a n t PCs degenerate progressively starting at 21 days after birth. A t 2 months, less than 1% of the n o r m a l PC population remains, mostly in the nodulus 14. According to Roffier-Tarlov et at. 2°, the weight of the deep nuclei in pcd mice diminishes by 35% at 45 days. At 28 days, a 50% fall in
* Part of this work has been presented in abstract form25. Correspondence: M. Wassef, I.N.S.E.R.M. U-106, H6pital de la Salp6tri6re, 75651 Paris, Cedex 13, France. 0006-8993/86/$03.50 © 1986 Elsevier Science Publishers B.V. (Biomedical Division)
126 G A B A concentration in the deep nuclei occurs in pcd mutants as compared to normal littermates. No further changes in these two parameters have been observed up to 80 days after birth. We used pcd and C57BL control mice between 3 and 4 months old in order to avoid the period of decrease in deep nuclear volume and to obtain an almost total absence of PCs. Different parameters were measured. (1) The volumes of the deep nuclei in pcd and C57BL mice were calculated from their surface area on sections of paraffin-embedded cerebella. (2) The number of remaining PCs were counted in 3 pcd cerebella after staining with an immune serum (anti-cyclic GMP-dependent protein kinase (cGK)) which specifically labels PCs. (3) The surface density of GAD-immunoreactive boutons was obtained from the deep nuclei of pcd and C57BL mice by postembedding techniques on 1-/~m-thick sections of plastic-embedded cerebella. Two additional elements were also studied: (a) the comparative distribution of G A B A immunoreactive neurons in the deep nuclei of both pcd and C57BL; and (b) the ultrastructural features and the synaptology of GAD-positive boutons in the pcd deep nuclei, in order to determine whether GABAergic nerve terminals of PC origin could be differentiated from those of non-PC origin. MATERIALS AND METHODS
Animals C57BL mice were obtained from I F F A - C R E D O (Lyon), and pcd mutants were obtained from Dr. J.L. Guenet (Institut Pasteur, Paris). All animals were between 3 and 6 months old.
Histological techniques Fixation. Mice were anesthetized with ether and perfused through the ascending aorta with a fixative solution containing 4% formaldehyde in 0.12 M sodium Ehosphate buffer, pH 7.2-7.4. In some cases 0.01 M periodic acid and 0.075 M lysine were added according to the method of McLean and Nakane 13. Two additional pcd mice were fixed for electron microscopy (4% paraformaldehyde and 0.1% glutaraldehyde in the same buffer solution). The brains were dissected out and a block containing the whole cerebellum was immersed overnight in the formaldehyde
fixative at 4 °C. For the detection of G A B A immunoreactivity, a stronger fixative was used, containing 5% glutaraldehyde in the same phosphate buffer (one pcd and one C57BL). Embedding. After a rapid rinse in phosphate-buffered saline (PBS), 3 C57BL and 3 pcd cerebella were embedded in paraffin and cut at 10/~m thickness. The paraffin sections stained with Cresyl violet were used to evaluate the volume of the deep nuclei. The two pcd cerebella for electron microscopy were cut with a Vibratome into 50-~m-thick sections, immunostained with the anti-GAD antiserum, osmicated and flat embedded in Araldite. From two C57BL and two pcd cerebella, 300-400~m-thick sections were cut with a Vibratome and flat embedded in Araldite. From these blocks, 1-j~mthick sections of the cerebellum, separated by at least 3 j~m, were collected on gelatin-coated slides. Araldite was removed with sodium methanolate, according to the method of Lane and Europa 11. A 5-min pretreatment with 3% H202 was used to eliminate endogenous peroxidase. The slides were then treated for 15 rain at 37 °C with a solution containing 1/~g/ml of pronase (Sigma). After immunostaining, the semithin sections were used to count labeled boutons or to observe GABA-immunoreactive neurons. Frozen sections. After fixation, the cerebella were infiltrated for 2 days at 4 °C in 3 changes of 30% sucrose in sodium phosphate buffer. Twenty-five-j~mthick sections were cut on a freezing microtome. Frozen sections were processed for G A D or cGK immunostaining. In preliminary experiments some of the GAD-immunoreacted sections were flat embedded in Araldite; 1-~m-thick sections were cut from these blocks and the immunoreactive terminals were counted. Usually, the immunoreacted sections were mounted on gelatin-coated slides and dried overnight. They were then immersed for 30 s in osmic acid, 0.01%, in distilled water, dehydrated and mounted with coverslips. Immunocytochemistry. After removal of Araldite and pronase pretreatment, mounted semithin sections were processed for G A D and G A B A immunocytochemistry according to the method of Sternberger et al. 23. Freely floating frozen sections were similarly processed for cGK, except that 0.25% Triton X-
127 100 was added to the antisera. Antisera. The anti-GAD antiserum raised in sheep has been extensively characterized t8,19. Anti-GABA was purchased from Immunotech (Marseille, France). It was raised in rabbit against reduced G A B A linked to bovine serum albumin with glutaraldehyde 21. Anti-cGK 4'12'24 (gift of Pr. P. Greengard) and antiG A B A 21 are rabbit antisera and have been previously characterized. These antisera were applied overnight at room temperature at a dilution of 1/2000 for anti-GAD, 1/3000 for anti-GABA, and 1/1000 for anti-cGK. Anti-rabbit immunoglobulin (IgG) produced in goat and anti-goat IgG produced in rabbit were purchased from Cappel and used at 1/100 dilution. Rabbit peroxidase-antiperoxidase (PAP) (UCB, France) and goat PAP (Sternberger-Meyer) were used at 1/200 dilution.
Measures. (1) Volume of the deep nuclei in pcd and C57BL. The surface of the deep nuclear area in one or both hemicerebella was measured on every fifth 10-/~mthick paraffin section with a camera lucida and a Hewlett-Packard microcomputer interfaced with a graphics tablet. The volume of the deep nuclei was then calculated in 3 pcd and 3 C57BL mice. (2) Density of GAD-immunoreactive terminals. GAD-immunoreactive terminals were counted on the 555/~m z squares of a grid with a x 100 oil-immersion objective on semithin sections of the cerebellar nuclei of two C57BL and two pcd mice. In each animal, sections separated by at least 3 ktm were obtained from 3 different levels of the deep nuclei. (3) Number of cGK-immunoreactive PCs in pcd. The cerebella of 3 pcd mice were cut at 25/~m on a freezing microtome and reacted freely floating for cGK immunocytochemistry. The sections were mounted on gelatin-coated slides and the immunoreactive cell bodies were counted. RESULTS
Technical considerations In a preliminary attempt to count labeled terminals in the deep nuclei, 25-/~m-thick frozen sections were immunoreacted for G A D , freely floating, and flat embedded in Araldite. The counts were made on
semithin sections cut from these blocks. However, we always observed differences in staining intensity between the superficial and internal parts of the blocks, even if the frozen sections were incubated overnight with 0.25% Triton X-100 added to the primary antiserum. The later use of desaraldited semithin sections gave an even immunoreactivity throughout the whole extent of the deep cerebellar region in cerebella cut on the coronal plane and permitted a good sampling of the nuclear region (Figs. 2A, B and 3A, B). Boutons in the deep nuclei of C57BL are very tightly packed, and were probably underestimated as compared to pcd. Labeling of the cellular bodies of GABAergic interneurons was more intense with anti-GABA than with anti-GAD (Figs. 2 and 3). However, the G A B A immunoreactivity of PC axons in the white matter and around and inside the deep nuclear region prevented the counting of individual terminals. On the contrary, G A D was highly concentrated in axon terminals but presynaptic axonal segments were almost devoid of labeling, thus allowing the enumeration of individual boutons.
Cerebellar cortex Number and localization of remaining PCs in pcd. In C57BL mice, cyclic GMP-dependent protein kinase labeled all PCs, and exclusively this cell population, in the cerebellum. In two 4-month-old pcd mutants, the number of labeled cellular bodies was 103 and 112, respectively. As previously described, most of the remaining PCs were located in the nodulus (Fig. 1C). In one 2-month-old pcd, the number of PCs was 1053. Comparison of GABAergic cortical innervation in pcd and C57BL. The thickness of the cortex was reduced in pcd mice as compared to C57BL due to the elimination of PCs and to the progressive death of granule cells 7. G A D immunostaining of the terminals in the cortex was always more intense in pcd than in C57BL mice cerebella processed in parallel (Fig. 1A, B). The short alignments of immunoreactive boutons observed in the molecular layer of normal mice (Fig. 1B) probably represent stellate cell axon terminals or ascending collaterals of basket cell axon terminals in synaptic contact with the shaft of PC dendrites. In the molecular layer of pcd, GAD-immunoreactive termi-
128
Fig. 1. A and B: semithin sections through the cerebella ofpcd mutant (A) and control C57BL (B) mice. In C57BL mice (B), Purkinje cells are slightly G A D immunoreactive (asterisk). The organization of the cortical layers (molecular layer, m; Purkinje cell layer, PC; granule cell layer, gr) is preserved in pcd mutants. In spite of the absence of Purkinje cells in these animals, the PC layer remains as a row of empty pinceau formations. A, x420; B, x420. C: general view of the cerebellum of a 4-month-old pcd mouse immunostained with anti-cGK to show PCs (indicated by arrowheads). Two PCs are shown at higher magnification in the inset, x45; inset, x 150.
129
Fig. 2. Semithin sections through the deep nuclear region of C57BL (A) and pcd (B) mice. A: in C57BL, the somata and thick dendrites are underlined with an almost continuous row of GAD-immunoreactive terminals, x600. B: at the same magnification, the GABAergic innervation of pcd deep nuclei, although reduced when compared to C57BL (A), is still appreciable. This is due in part to the almost 60% decrease in deep nuclear volume in the mutant, x600. C: higher magnification of the GAD-immunoreactive terminals in the pcd mutant, x 1800.
130 nals are more uniformly distributed (Fig. 1A). Lightly stained Purkinje cell bodies were surrounded by an almost continuous ring of terminals in C57BL, and the basket fibers of the pinceau formation were always well stained. In pcd mice (Fig. 1A), there was an apparent increase in the number of pinceau formations which in some regions seemed even longer than their normal counterparts, as is the case in the hemispheres of the nervous cerebellum 22. The density of isolated labeled terminals was lower in the PC layer than in the adjacent molecular and granular layers. Golgi cell terminals were reduced in number in the granular layer of pcd but were intensely stained and typical rosettes were frequent. With anti-GAD serum, GABAergic perikarya were at most lightly stained (Figs. 1A, B and 2A, B). On the contrary, with anti-GABA serum, GABAergic deep nuclear neurons (arrows in Fig. 3A, B and arrowheads in Fig. 3C, D) and Golgi cells in the cortex (arrowheads in Fig. 3A, B) were intensely labeled. Interestingly, PC bodies were almost unstained with anti-GABA although their axons were intensely labeled as soon as they emerge from the cell bodies. This suggests a different compartmentation of G A B A in PCs.
Deep nuclei Volume of the deep nuclei. The deep nuclear region was more easily delineated in pcd than in C57BL mice. For this reason, the nuclear volume was calculated in both hemicerebella in 3 C57BL and in only one in 3 pcd (Table I). The nuclear volume in pcd decreased to 42% of control volume. GAD and GABA immunoreactivity in the deep nuclei. As expected, the density of GAD-immunoreactive terminals was greatly decreased in the deep nuclei of pcd. In C57BL, neurons and thick dendrites were delineated by an almost continuous row of immunoreactive boutons. In pcd, GAD-positive dots were scattered randomly. In both mouse types, antiG A B A intensely stained many neurons which were not easily seen with anti-GAD without a colchicine pretreatment. In C57BL, Purkinje cell axons in the white matter and inside the nuclear region were also G A B A immunoreactive. The white matter in pcd mice was devoid of G A B A immunoreactive fibers except for some labeled axons near the midline. Ultrastructure of GAD-positive axon terminals in pcd deep nuclei. In the pcd mutant, the perikaryal
surface of the large neurons was partially covered with a small number of axosomatic synapses, which appeared either grouped into small clusters of 2-3 boutons or isolated. Some of these synapses always exhibited G A D immunoreactivity. The labeled axon terminals were large and elongated, with a largest diameter of about 5/~m and a smallest diameter between 1 and 2/zm (Fig. 4 A - D ) . They contained a pleomorphic population of tightly packed synaptic vesicles, the vast majority of which were flattened. At the synaptic interface, pre- and postsynaptic membranes established several patches of synaptic junctions with poorly marked postsynaptic differentiations, as in type II synapses (Fig. 4 A - D ) . The small neurons also received GAD-positive synapses on their perikarya. Generally, these axon terminals were smaller than those synapsing on the large cells. They appeared as bulbous ends filled with pleomorphic vesicles and established type II synaptic contacts (Fig. 4B). Large (3-4 btm in diameter) and smaller (1-2/~m in diameter) GAD-positive axon terminals were frequently encountered in synaptic contact with dendrites of either large or small neurons. Although the main dendritic stems were privileged sites for axodendritic GAD-positive inputs, labeled boutons also established synaptic contacts on thin dendritic elements (Fig. 4C) and on sessile spinous protrusions (Fig. 4D). These axon terminals shared their intrinsic features with those at somatic locations, and, like them, established mainly type II synapses. Only occasionally did they establish synapses of an intermediate type, characterized by a somewhat more marked postsynaptic differentiation (Fig. 4C). These labeled boutons were present in all 4 deep nuclei, especially in the lateral nucleus.
Number of GABAergic boutons in pcd and C57BL deep nuclei. The surface density of GAD-immunoreactive boutons in the deep nuclei of pcd was about one-third of that observed in C57BL (Table II). Although direct observation of the sections gave the impression that GABAergic innervation was more uniform in pcd than in C57BL, mean variation was about 25% in both cases. Taking into account the decrease in deep nuclear volume in pcd, we calculated that the total number of GABAergic boutons in pcd deep nuclei was 15% of that in C57BL.
131
Fig. 3. Semithin sections of C57BL (A, C) and pcd (B, D) deep nuclear regions stained with anti-GABA serum. A and B: general view. Immunoreactive neurons (arrows) are scattered throughout the deep nuclei of C57BL (A) and pcd (B). In the cortex, Golgi cells are intensely stained (arrowheads in A and B). In the white matter of C57BL mice (wm in A), PC axons are G A B A immunoreactive. A and B, x 115. C and D: detail of GABA-immunoreactive structures. GABAergic neurons (arrowheads) are completely stained, with the exception of the nucleolus. Their small size can be appreciated by comparing them to unstained neurons (N). The higher background in C as compared to D is probably due to leakage of G A B A (highly concentrated in normal mice) during fixation. C and D, × 420.
132 TABLE I Mean volume of deep nuclei in C57BL and PCD mice Animal
Mean volume + S.E.M. (108~m 3)
C57BL (n = 6) 4.5 + 0.84 PCD(n=3) 1.9+0.17
PCD volume/C57BL volume
0.42
DISCUSSION The existence of a non-Purkinje cell GABAergic innervation in the cerebellar deep nuclei was hypothesized as an explanation of the rather high G A D activity remaining in the deep nuclei after deafferentation from Purkinje cells 5'6. After large cortical lesions or in pcd mutants, G A D activity is never lowered to less than 30% of control values. In contrast, transection of the striatonigral tract at the premammillary level causes a 90% reduction of G A D activity in the pars reticulata of the substantia nigra 2. However, it is difficult to appreciate the importance of non-PC GABAergic innervation of the deep nuclei in animals with large cortical lesions because the PC projection to the deep nuclei is somewhat diffuse, although ordered. According to Fonnum et al. 5, it is impossible to destroy all the PCs which send their axons to a particular region of the deep nuclei. For this reason, pcd mice seem more appropriate for this evaluation. The number of PCs in C57BL was estimated by Herrup and Mullen s to be 196,000. Using this value, we can calculate that only 0.5% of the normal PC population remains in 2-month-old pcd mutants; this drops to 0.05% in 4-month-old animals. In pcd mice, the deep nuclear volume decreases to 42% of control value (Table I), more than would be expected. According to Roffier-Tarlov et al. 2°, the weight of the pcd cerebellum decreases to 65% of control value at 79-85 days. A simple subtraction of the volume occupied by PC axons and boutons does not entirely explain this reduction. In the cat, for example, the entire population of axons and boutons occupies less than 60% of the total deep nuclear volume 6. The massive deafferentation of the deep nuclei might induce a shrinkage or a moderate loss of deep nuclear neurons. This decrease in volume is probably responsible for the high concentration of G A B A (55% of con-
trol) found by Roffier-Tarlov et al. 2° in pcd deep nuclei and for the relatively high density of GAD-immunoreactive terminals (37% of control) found in the present study. The discrepancy between the two values could be due to the fact that in the study of Roffler-Tarlov et al., deep nuclear samples contained up to 30% of cerebellar cortex in which there was no important decrease in G A B A concentration 2°. Many factors could have infuenced our estimation of non-PC GABAergic innervation. Contribution o f remaining PCs
The contribution of the 0.5-0.05% of PCs remaining in pcd mice to the GABAergic innervation of the deep nuclei is probably negligible because: (1) most of them are located in the nodulus and thus merely project to the vestibular nuclei, with the exception of Deiters' nucleusl; (2) at 4 months, remaining PCs seem altered when observed with anti-cGK (Fig. 1C); and (3) PC axon terminals in the deep nuclei degenerate earlier than the corresponding PC somata and dendrites in the cerebellar cortex 2°. Therefore, PC projections to the deep nuclei are probably inexistent in 4-month-old pcd mice. Sprouting o f n o n - P C G A B A e r g i c axons
There is some evidence of axon sprouting in the deep nuclei of pcd mice following PC axon loss. According to Roffier-Tarlov et al. 2°, the somatic circumference of the large neurons of the lateral cerebellar nucleus does not decrease in pcd mice. The proportion of the circumference occupied by synaptic terminals (40% of control C57BL mice) increases from 7% in 23-day-old pcd mice to 12% in 60-day-old animals. We have observed (unpublished data) a very important increase in the density of serotonin immunoreactive fibers in the deep nuclei of pcd as compared to C57BL mice. However, without a precise calculation we cannot rule out the possibility that the higher density of 5-HT immunoreactive fibers in pcd mice is due simply to the decrease in deep nuclear volume. At present, it is not possible to determine the importance of an eventual sprouting of non-PC G A B A ergic axons.
133
Fig. 4. Electron micrographs of GAD-positive axon terminals in the deep nuclei of the pcd cerebellum. A: lateral nucleus. A large, elongated, labeled axon terminal is in synaptic contact with the perikaryon of a large neuron (LN). The arrows point to the small patches of type II synaptic junctions, x25,000. B: medial nucleus. The immunolabeled axon terminal is directly apposed to the perikaryal surface of a small neuron (SN). x22,000. C: medial nucleus. The slightly labeled axon terminal establishes an intermediate type of synaptic junction (arrows) with a small dendritic profile (SD). x50,000. D: lateral nucleus. The GAD-positive axon terminal is in synaptic contact with the perikaryon of a large neuron (LN), as well as with a small spinous protrusion (S). The arrows mark the type II synaptic complexes, x 31,000.
134 TABLE II Density of GAD-immunoreactive terminals in the deep nuclei of C57BL and pcd mice Animal
555 ktm2 squares counted
Mean number of S.E.M. terminals per square
Numerical density (terminals/lO0 ktm2)
Numerical density pcd/ numericaldensity C57BL
83164 (C57BL) 83165 (C57BL) 83168 (pcd) 83169 (pcd)
154 148 65 145
72 78 27 27
13 14 5 5
0.37
Origin o f non-Purkinje cell G A B A e r g i c innervation o f the deep nuclei The G A B A e r g i c neurons of the d e e p nuclei are the most likely source of non-PC G A B A e r g i c innervation. These neurons are distributed in all 4 subdivi-
sions of the d e e p nuclei in pcd as well as in C57BL mice (Fig. 3). Recently, these neurons were r e p o r t e d to project to the inferior olive, at least in part Is. It is not known whether there are two classes of G A B A e r g i c n e u r o n in the deep nuclei or w h e t h e r the projecting neurons have axon collaterals and are also responsible for local G A B A e r g i c innervation in the d e e p nuclei. Although less p r o b a b l e , the existence of other G A B A ergic afferents cannot be excluded. The present ultrastructural analysis of G A D - p o s i tive axon terminals of the lateral nucleus in pcd mice was a i m e d at characterizing the intrinsic features of this axonal p o p u l a t i o n and at helping to d e t e r m i n e its origin. Chan-Palay 3 has characterized 6 axon terminal types in the rat lateral nucleus and has ascribed a different origin to each. In comparison, our results are rather disappointing. The cytological features of the G A D - p o s i t i v e b o u t o n s are very much like those of class A . F o r C h a n - P a l a y 3, these boutons belong to PC axons, but, as discussed above, this axonal pop-
REFERENCES 1 Angaut, P. and Brodal, A., The projection of the 'vestibulocerebellum' onto the vestibular nuclei in the cat, Arch. Ital. Biol., 105 (1967) 441-479. 2 Brownstein, M.J., Mroz, E.A., Tappaz, M.L. and Leeman, S.E., On the origin of substance P and glutamic acid decarboxylase (GAD) in the substantia nigra, Brain Research, 135 (1977) 315-323. 3 Chan-Palay, V., Cerebellar Dentate Nucleus. Organization, Cytology and Transmitters, Springer, Berlin, 1977, pp. 212-238.
+18 _+19 _+7 _+6
ulation is missing in the m u t a n t cerebellum. M o r e over, the vast majority of l a b e l e d axons might belong to the small G A B A e r g i c neurons present within the deep nuclei, although they are very different from those of Chan-Palay's class E 3, considered as originating from local neurons. Clearly, cytological criteria alone cannot establish the origin of the non-Purkinje cell G A B A e r g i c innervation of the deep nuclei. These results are, however, i m p o r t a n t since they not only prove that the i m m u n o s t a i n e d dots counted in the light micrographs c o r r e s p o n d to axon terminals, but also d e m o n s t r a t e that the G A B A e r g i c innervation forms conventional synapses with all possible postsynaptic targets present in the deep nuclei. In conclusion, the 15% contribution of local neurons to the total G A B A e r g i c innervation of the deep nuclei as calculated in the p r e s e n t study is p r o b a b l y an overestimation. It can, however, be p r o p o s e d as a reasonable approximation. ACKNOWLEDGEMENTS The authors wish to thank Prof. P. G r e e n g a r d for the gift of a n t i - c G K antiserum, D. Le Cren for photographic w o r k and B. A l v a r a d o for improving the English and typing the manuscript.
4 De Camilli, P., Miller, P., Levitt, P., Walter, U. and Greengard, P., Anatomy of cerebellar Purkinje cells in the rat determined by a specific immunohistochemical marker, Neuroscience, 11 (1984) 761-817. 5 Fonnum, F., Storm-Mathisen, J. and Walberg, F., Glutamate decarboxylase in inhibitory neurons. A study of the enzyme in Purkinje cell axons and boutons in the cat, Brain Research, 20 (1970) 259-279. 6 Fonnum, F. and Walberg, F., An estimation of the concentration of v-aminobutyric acid and glutamate decarboxylase in the inhibitory Purkinje axon terminals in the cat, Brain Research, 54 (1973) 115-127.
135 7 Ghetti, B., Alyea, C.J. and Muller, J., Studies on Purkinje cell degeneration (pcd) mutant: primary pathology and transneuronal changes, J. Neuropathol. Exp. Neurol., 37 (1978) 617. 8 Herrup, K. and Mullen, R.J., Regional variation and absence of large neurons in the cerebellum of the staggerer mouse, Brain Research, 172 (1979) 1-12. 9 Houser, R.C., Barber, R. and Vaughn, J.E., Immunocytochemical localisation of glutamic acid decarboxylase in the dorsal lateral vestibular nucleus: evidence for an intrinsic and extrinsic GABAergic innervation, Neurosci. Lett., 47 (1984) 213-220. 10 Ito, M. and Yoshida, M., The origin of cerebellar-induced inhibition of Deiter's neurons. I. Monosynaptic initiation of the inhibitory postsynaptic potentials, Exp. Brain Res., 2 (1966) 330-349. 11 Lane, B.P. and Europa, D.L., Differential staining of ultrathin sections of Epon-embedded tissues for light microscopy, J. Histochem. Cytochem., 3 (1965) 579-582. 12 Lohman, S.M., Walter, U., Miller, P.E., Greengard, P. and De Camilli, P., Immunohistochemical localization of cyclic GMP dependent protein kinase in mammalian brain, Proc. Natl. Acad. Sci. U.S.A., 78 (1981) 653-657. 13 McLean, I.W. and Nakane, P.K., Periodate lysine paraformaldehyde fixative. A new fixative for immunoelectron microscopy, J. Histochem. Cytochem., 22 (1974) 1077-1083. 14 Mullen, R.J., Eicher, E.M. and Sidman, R.L., Purkinje cell degeneration, a new neurological mutation in the mouse, Proc. Natl. Acad. Sci. U.S.A., 73 (1976) 208-212. 15 Nelson, B., Barmack, N.H. and Mugnaini, E., A GABAergic cerebello-olivary projection in the rat, Soc. Neurosci. Abstr., 10 (1984) 539. 16 Obata, K., G A B A in Purkinje cells and motoneurons, Experientia, 25 (1969) 1285. 17 Obata, K. and Takeda, K., Release of y-aminobutyric acid into the fourth ventricle induced by stimulation of the cat's cerebellum, J. Neurochem., 16 (1969) 1043-1047.
18 Oertel, W.H., Mugnaini, E., Schmechel, D.E., Tappaz, M.L. and Kopin, I.J., The immunocytochemical demonstration of gamma-aminobutyric acid-ergic neurons - methods and application. In S.L. Palay and V. Chan-Palay (Eds.), Cytochemical Methods in Neuroanatomy, Alan Liss, New York, 1982, pp. 297-329. 19 Oertel, W.H., Schmechel, D.E., Mugnaini, E., Tappaz, M.L. and Kopin, I.J., Production of a specific antiserum to rat brain glutamic acid decarboxylase by injection of an antigen-antibody complex, Neuroscience, 6 (1981) 2715-2735. 20 Roffler-Tarlov, S., Beart, P.M., O'Gorman, S. and Sidman, R.L., Neurochemical and morphological consequences of axon terminal degeneration in cerebellar deep nuclei of mice with inherited Purkinje cell degeneration, Brain Research, 168 (1979) 75-95. 21 Seguela, P., Geffard, M., Buijs, R.M. and Le Moal, M., Antibodies against 7-aminobutyric acid: specificity and immunochemical results, Proc. Natl. Acad. Sci. U.S.A., 81 (1984) 3888-3892. 22 Sotelo, C. and Triller, A., Fate of presynaptic afferents to Purkinje cells in the adult nervous mutant mouse: a model to study presynaptic stabilization, Brain Research, 175 (1979) 11-36. 23 Sternberger, L.A., Hardy, P.H., Jr., Cuculis, J.J. and Meyer, H.G., The unlabelled antibody method of immunohistochemistry. Preparation and properties of soluble antigen complex and its use in identification of spirochetes, J. Histochem. Cytochem., 18 (1970) 315-333. 24 Walter, U., Miller, P., Wilson, F., Menkes, D. and Greengard, P., Immunological distinction between guanosine 3',5'-monophosphate dependent protein kinases, J. Biol. Chem., 255 (1980) 3757-3762. 25 Wassef, M., Tappaz, M.L., Oertel, W.E. and Sotelo, C., GAD immunocytochemistry in cerebellum of mutant mice with Purkinje cell defect, Neurosci. Lett., Suppl. 10 (1982) 513.