Proc. NatL Acad. Sci. USA Vol. 79, pp. 5093-5096, August 1982 Neurobiology

Presynaptic and perikarya in deafferented cerebellar cortex (reactive synaptogenesis/dendrodendritic /Golgi cells/granule cells/morphological plasticity) J6ZSEF HAMORI AND J6ZSEF SOMOGYI First Department ofAnatomy, Semmelweis University Medical School, Tuzolt6 u. 58, H-1450, Budapest, Hungary Communicated by J. Szentdgothai, April 29, 1982

ABSTRACT Two to 30 days after complete isolation of the This paperprovides experimental evidence that the observed cerebellar cortex of adult rats, small Golgi II type neurons in the potential of cerebellar granule and Golgi neurons to develop granular layer develop presynaptic sites on their dendrites and presynaptic dendrites and somata can be induced with surgical perikarya. The newly developed presynaptic dendrites estab- deafferentation of the adult cerebellar cortex-i.e., without al- lished dendrodendritic synaptic contacts with dendritic digits of tering significantly the number of granule cells and other in- granule cells. In addition, some granule cells were observed to terneurons in the cortex. It will also be shown that the observed develop presynaptic sites on their somata and dendritic digits, re- plastic potential of the two nerve cell types is not restricted to sulting in the formation of(sometimes reciprocal) dendrodendritic the synapses between granule cells or between granule cells and Gol- developing (10). gi neurons. The development ofthese unusual synaptic formations indicates a persisting synaptic plasticity ofthese two cerebellar cell MATERIALS AND METHODS types. It is suggested that this unorthodox formation de- Deafferentation of the cerebellar cortex was achieved two velops as part of a reactive compensatory process to synaptic de- by saturation brought about different surgical approaches. (i) In six adult rats the upper cer- by the deafferentation of the cerebellum. ebellar vermis was undercut with a surgical knife, resulting in the destruction ofall afferent fibers. In Neurons with dendrites and soma exhibiting both pre- and post- (ii) three adult rats the synaptic sites cerebellar peduncles were destroyed, resulting in the destruc- (presynaptic dendrites and perikarya) have been tion ofall climbing fibers, catecholamine afferents, and the ex- found in many regions of the central nervous system (for ref- tracerebellar mossy terminals, but leaving intact the nucleo- erences, see refs. 1 and 2). It was soon recognized that in most cortical mossy endings (11) and the majority of the efferent cases (e.g., retina, thalamic nuclei, tectal nuclei, dorsal horn Purkinje . ofthe spinal cord) neurons with presynaptic dendrites are short- The interneurons. It was also established that the interneurons animals were perfused 2, 15, and 30 days after the op- having presynaptic dendrites were eration with an aldehyde fixative. Small blocks from the vermal localized almost exclusively cortex were excised, treated with osmic acid, dehydrated in in subcortical nuclei or regions, whereas in the cerebral cortex, and embedded in apart from the hitherto unconfirmed observation of Sloper (3), ethanol, Durcupan. Serial ultrathin sections all neurons are were cut with an LKB Ultrotome, collected on Formvar-coated "classical," with only postsynaptic dendrites and single-hole grids, and stained by uranyl acetate and lead citrate. soma. Similarly, the billions ofnerve cells in the cerebellar cor- Sections were viewed and tex are, under normal circumstances, all classical neurons. Ac- photographed in a JEM 100 B elec- cordingly, synaptic arrangements between the five neuronal tron microscope. types of the cerebellar cortex and the three main afferents are identified as axodendritic or axosomatic synapses (4), except for RESULTS a few axoaxonal synapses on the initial segment ofPurkinje cells An unexpected finding after complete deafferentation was the (5, 6). Under abnormal conditions, as in organotypic cerebellar presence in the granular layer of the undercut vermis of den- cultures (7) or in cerebellar mutant mice (8, 9), however, gran- dritic processes exhibitingbothpre- andpostsynaptic sites. Two ule cells in the cerebellar cortex were shown to have accumu- structurally distinct presynaptic types were observed lations of synaptic vesicles in their thin soma or in their den- 15-30 days postoperatively within the slightly atrophic cere- drites and to establish somatodendritic or dendrodendritic bellar glomeruli. synaptic contacts with Purkinje neurons. Sotelo (2) has also de- (i) Large presynaptic dendrites occupying a considerable por- scribed Golgi neurons with presynaptic dendrites and perikarya tion of the glomerulus (Fig. 1A). By using a serial section re- in the rat cerebellar cortex after neonatal x-irradiation. Because construction technique, these dendrites were identified as the in the mouse mutants weaver (8) and reeler (9) as well as in tissue processes ofsmall (6-11 um in diameter) Golgi neurons, resid- culture (7) the number ofgranule cells was much lower than in ing mostly in the deeper halfof the granular layer. The Golgi- the normal cerebellum, it was proposed (2) that the appearance presynaptic dendrite was found to establish contact with 15-25 ofpresynaptic sites on soma and dendrites ofthe surviving gran- dendritic digits ofsurrounding granule cells. The same dendrite ule cells is a consequence ofthe loss ofgranule cells. In the case was also postsynaptic to axon terminals or of Golgi cells a causal relationship was proposed between the dendritic digits. The parent cell bodies of the Golgi cell pre- disappearance ofbasketand stellate neurons inx-irradiatedcere- synaptic dendrites were also seen to develop presynaptic sites bella and the concomitant development of presynaptic den- (Fig. 1B). As revealed by the reconstruction ofseven small Golgi drites on the surviving inhibitory neuron, the Golgi cell. cells, the somatic surface of all cell bodies established several somatodendritic synaptic contacts with the dendritic digits of The publication costs ofthis article were defrayed in part by page charge the granule cells. The synaptic vesicles in the perikaryon and payment. This article must therefore be hereby marked "advertise- in the dendrites ofthe small Golgi neurons were pleomorphic. ment" in accordance with 18 U. S. C. §1734 solely to indicate this fact. (ii) The second type ofpresynaptic dendrite, also within the 5093 Downloaded by guest on October 1, 2021 5094 Neurobiology: Hdmori and Somogyi Proc. Nad Acad. Sci. USA 79 (1982)

'A- synaptic glomerulus, was the small dendritic digit of granule cells (Fig. 2 A and B). The dendritic digits are the terminal bul- bous portions ofthe tiny granule celldendrites and are the main postsynaptic partners ofmossy terminals. The hundreds ofdig- its in a single glomerulus are mechanically coupled to each other by junctional complexes, the so-called "symmetric attachment 9.:S f W: plaques"-a characteristic landmark in the identification ofthe in t.. t. l:.f digits. Two days after deafferentation, but more so after 15 and 30 days, spheroid synaptic vesicles were accumulated in some den- :. gas A. dritic digits, resulting in asymmetric dendrodendritic synapses. '- a-.! The postsynaptic dendrite was either another granule cell den- ., .. drite (Fig. 2 B and C) or a Golgi dendrite (Fig. 1A). Although i+,; :' only a small percentage (about 4%) ofall dendritic digits became

C' presynaptic, in serial section such neoformations were found to .. occur in alldeafferented synaptic glomeruli investigated. More- over, several granule cell bodies were also observed to contain .g :' ;. accumulations of synaptic vesicles and to develop presynaptic ... .tS= -.5 < .. ..jij;^.. : -2a :.: sites, resulting in somatodendritic or, rarely, somatosomatic (Fig. 2D) synaptic contacts, with other granule or Golgi neurons m --. A. as postsynaptic elements. 4 !. So The two othernerve cell types ofthe granularlayer, the large ._ -E [9- to 18-,um] Golgi cells and the rare Lugaro cells, as well as r- Purkinje cells and the interneurons ofthe molecularlayer, were

X. ..s ., it: seen to develop neither presynaptic dendrites nor perikarya. FE, I. ," I :'; The only change in the molecular layer was the increase ofva- f cant dendritic spines (12), caused most probably by climbing fiber deafferentation (13). Presynaptic dendrites and perikarya were observed both in undercut cortex and in totally isolated cerebellum, although the frequency of these neoformations was somewhat higher in the A, ;' Is A* t~~~~ first case. DISCUSSION A. F : We have provided evidence for the development ofpresynaptic dendrites and perikarya in the adult cerebellar cortex by deaf- ferentation. The potential of Golgi cells (2) and granule cells (8, 9) to develop presynaptic dendrites was described earlier in mutant mice and in x-irradiated agranular cerebellum. A ten- tative hypothesis was put forward (2) to explain the appearance ofpresynaptic sites on the dendritic surfaces. According to this wo;A-! hypothesis, the newly developed presynaptic sites and peri- karya would be formed by the two nerve cell types as a result ofthe numerical deficiency (8, 9) or absence (2) ofinhibitory and ~~~ ~ ~ ~ ~ ~ ~ ~ I 0 excitatory interneurons and their axons. Because deafferenta- 'g-S @ -t f tion alone did not appear to change nerve cell numbers, at least 'e s,£'t 'e not during the relatively short postoperative survival time, the B.,: h"' AES,& +, b unusual presynaptic formations found in this study must be at- tributed to the removal ofthe afferent fibers. It is the absence of the mossy fibers that appears to induce the development of presynaptic dendrites. Climbing fiber deafferentation is fol- lowed by the appearance of an excess of free dendritic spines h~~~~~~~~~~~~~~~~~~~~~~~~i (12, 13) and by the partial myelinization ofinterneuronal peri- karya and of Purkinje cell dendrites (14) but does not lead to the formation ofpresynaptic dendrites. The removal of mossy afferents, on the other hand, is equivalent to a massive synaptic *f --a zfiZ S deprivation ofthe whole granularlayer. In contrast to the septal FIG. 1. Electron micrographs of the soma and dendritic process of nuclei (15) and hippocampus (16), where the loss of synaptic a small Golgi neuron in the cerebellar cortex of adult rat 15 days after equilibrium through partial deafferentation was followed by the complete deafferentation. (A) The Golgi dendrite within the cerebellar excess sprouting of new axonal endings from surviving intact glomerulus contains several aggregates of synaptic vesicles. The den- afferents occupying the vacant postsynaptic sites, no such "reaf- drite is postsynaptic (ringed arrows) to an axonal profile (a) and to a ferentation" could be expected in a totally deafferented cere- dendritic digit (d), and presynaptic (arrows) to other dendritic digits; bellar cortex. In trial to m, remnant of a degenerating dark mossy terminal. (x32,400.) (B) fact, any restore the disturbed synaptic Small Golgi cell soma establishing somatodendritic synapse (arrow) equilibrium ofthe granular layer should arise from local nerve with a dendritic digit of the atrophic glomerulus. (x 16,200.) (Inset) cells. However, Purkinje cells after complete deafferentation Details of the synaptic contact. (x 108,000.) were observed to develop only a few hypertrophic axon collat- Downloaded by guest on October 1, 2021 Neurobiology: Halmori and Sornogyi Proc. Natl. Acad. Sci. USA 79 (1982) 5095 erals (17). Likewise, a slight hypertrophy ofthe Golgi cell axons A was observed after undercutting of the cortex (18), although in the present electron microscopic study the number of Golgi axon terminals around the atrophic cerebellar glomeruli did not C .: O-:d. o~~ft :: appear to increase significantly. It appears that, in the absence of real axonal sprouting of the local neurons, the dendritic pro- cesses and perikaryaofthe granule cells and small Golgi neurons developed presynaptic sites. s * By destroying mossy terminals the only excitatory input (4) to the granular layer was abolished. It could be expected, there- fore, that the newly d~eveloped presynaptic sites would be of excitatory character. However, not only the excitatory granule cells but also inhibitory Golgi neurons developed presynaptic .:::Ol- dendrites, indicating that, in the case of restoration of synaptic

..- iL :. equilibrium of individual cells or the whole system, specificity .'s. im - of the synaptic connections not be essential. ;F may Large Golgi neurons (10-18 Am) were not observed to de- velop presynaptic dendrites or somata, although their cell bod- ies are located in the granular layer. There are two possible "T.!" explanations: (i) Small and large Golgi neurons differ genetically

.0 in their synaptic plasticity. (ii) The dendritic tree ofsmall Golgi -A cells arborizes within the granular layer and its main synaptic _5 input-similar to that of granule cells-is, under normal cir- cumstances, the mossy fiber system. Large Golgi cells, on the other hand, send their dendrites up to the molecular layer, where the main input is the parallel fiber system (19). Although possible genetic differences regulating intrinsic potentialities of the two Golgi neuronal types may also control the formation of presynaptic dendritic sites, our assumption is that the ob- served difference in synaptogenetic plasticity is more likely to mIl, .s be related to the differences in synaptic inputs to the two cell -p.1r X types. :: 2 Is An unexpected feature ofthis reactive synaptogenesis is that e:C. to -a An. many dendrodendritic synaptic contacts develop between gran- ir ibit > s *4 !.j _,S' - :P4 g -v-, 0 ule cells. This observation indicates that the synaptic depriva- lust Fib .. ,j,_ .. , J tion ofthe granular layer leads to a reactive synaptogenesis ne- e t£. ._. glecting the original specificity of synaptic connections in the

.S ... \ cerebellar cortex, which normally does not contain synapses A. ::: :. .; .,

.f between granule cells. , S.-- 'C: In the absence of correlative electrophysiological studies , A. _ nothing can be said about the functional potentialities of the ; deafferented and partially reorganized granular layer of the . if-= ! cerebellar cortex. The morphological observations, however, S % clearly demonstrate that the morphogenetic potential ofgranule cells and small Golgi neurons to establish new synaptic contacts is a persisting property ofthese two cell types, present in both .S*-G _x':*\RS developing (10) and adult cerebella. 1. Hdmori, J. & Mezey, E. (1977) Exp. Brain Res. 30, 259-273. 2. Sotelo, C. (1977) Neuroscience 2, 275-283. 3. Sloper, J. J. (1971) Brain Res. 34, 186-192. 4. Eccles, J., Ito, M. & Szentdgothai, J. (1967) The Cerebellum as a Neuronal Machine (Springer, Berlin). FIG. 2. Electron of somata and dendrites: s 5 -of . ! (1965) Acta Acad. Sci. micrographs granule 5. Hamori, J. & Szentagothai, J. Biol Hung. cells in the cerebellar cortex of adult rat 15(A) and 30 (B-D) days after 15, 465-479. complete deafferentation. (A) Part of a cerebellar glomerulus showing 6. Somogyi, P. & Hamori, J. (1976) Neuroscience 1, 361-365. a degenerating remnant of the centrally located mossy terminal (i) 7. Kim, S. U. (1974) Exp. Neurol. 45, 659-662. that is surrounded by several dendritic digits of granule cells. The 8. Sotelo, C. (1975) Brain Res. 94, 19-44. digits are connected to each other by symmetrical attachment plaques 9. Mariani, J., Crepel, F., Mikoshiba, H., Changeux, J.-P. & So- (arrowheads). The arrow points at a dendrodendritic synapse. Note the telo, C. (1977) Philos. Trans. R. Soc. London Ser. B. 281, 1-28. presence of postsynaptic densities (p) facing degenerating mossy ter- 10. Hdmori, J. & Lakos, I. (1979) Neurosci. Lett. 3, 363. minals. (x42,200.) (B and C) Dendrodendritic synaptic contact (arrow) 11. Hamori, J., Mezey, E. & Szentagothai, J. (1981) Exp. Brain Res. between two granule cell dendritic digits, as revealed by serial sec- 44, 97-100. tions. Synaptic attachment plaques characteristic for dendritic digits 12. Halmori, J. (1980) in Advances in Physiological Sciences, eds. are indicated by arrowheads. (x62,100.) (D) Somatosomatic synaptic Szentagothai, J., HAmori, J. & Palkovits, M., (Pergamon, contact between a granule cell (g) and a small Golgi neuron (G). Akademiai Kiado, Budapest, Hungary), Vol. 2, pp. 117-131. (x 73,600.) 13. Sotelo, C. (1978) Prog. Brain Res. 48, 149-170. Downloaded by guest on October 1, 2021 5096 Neurobiology: Himori and Somogyi Proc. Nati Acad. Sci. USA 79 (1982)

14. Hgmori, J., Lakos, I. & Mezey, 1. (1980) J. Hirnforsch. 21, 18. Ram6n y Cajal, S. (1959) Degeneration and Regeneration of the 391-407. Nervous System, translated by May, R. M. (Hafner, New York), 15. Raisman, G. & Field, P. (1973) Brain Res. 50, 241-264. p. 629. 16. Cotman, C. W. & Nadler, J. V. (1978) in Neuronal Plasticity, ed. 19. Palay, S. L. & Chan-Palay, V. (1974) in Cerebellar Cortex, Cy- Cotman, C. W. (Raven, New York), pp. 227-271. tology and Organization (Springer, Berlin), p. 132. 17. Hgnori, J. & Lakos, I. (1980) Cell Tissue Res. 212, 415-427. Downloaded by guest on October 1, 2021