- Email: [email protected]
Plasticity of the parallel Fiber-Purkinje cell synapse by spine takeover and new synapse formation in the adult rat
Brain Research, 240 (1982) 205-220 Elsevier Biomedical Press
205
Research Reports
Plasticity of the Parallel Fiber-Purkinje Cell Synapse by Spine Takeover and New Synapse Formation in the Adult Rat SUZANNE CHEN and DEAN E. HILLMAN Department of Physiology and Biophysics, New York University Medical Center, 550 First Avenue, New York, N Y 10016 (U.S.A.)
(Accepted October 22nd, 1981) Key words: deafferentation - - reactive synaptogenesis - - Purkinje cell spines - - parallel fibers - - plasticity - -
giant spines - - cerebellum - - synapse
Alteration in synaptic connectivity between Purkinje cell spines and parallel fibers of the cerebellum were studied following partial deafferentation of Purkinje cells in the adult rat. Transection of parallel fibers by two lesions placed at a I mm interval on the folial crest were used to produce degeneration of these afferents. Ultrastructural analysis of synapses on Purkinje cell spines revealed degeneration with vacating of postsynaptic sites within 6 h. Reactive synaptogenesis as takeover of Purkinje cell spines by formation of new synapses from remaining parallel fibers occurred even before degenerating parallel fibers had vacated postsynaptic sites. This was accompanied by a marked increase in the number of dual innervations by reactive parallel fibers within one day. Some vacated postsynaptic sites were lost as indicated by a reduction in the number of synapses and others may have been taken over by newly formed synapses on spines. In addition, new synapses formed between the shafts of Purkinje cell branchlets and parallel fibers. Sprouting of parallel fibers occurred as small extensions without tubules while Purkinje cell spines reacted by forming elongated and multiple heads which contacted different parallel fibers. After 5 days degenerating boutons were rarely found. Enlarged spine heads were each capped by a proportionally enlarged parallel fiber bouton and joined by an elongated synaptic junction to parallel fibers. Some parallel fiber boutons were greatly enlarged and capped numerous profiles of spines. This study shows that formation of new pre- and postsynaptic sites takes precedence over reoccupation of original contacts and that multiple synapses on individual spines are being eliminated to give rise to single contacts with boutons. This elimination resulted in enlargement of synaptic contact areas between Purkinje cell spines and parallel fibers by taking over postsynaptic sites from some vacated and eliminated boutons. INTRODUCTION In the last decade reactive synaptogenesis t h r o u g h sprouting has been demonstrated in a n u m b e r o f regions o f the central nervous system such as the h i p p o c a m p u s s,3°,34,85, septum la,45, optic system 2s, 29,41, olfactory bulblV,Ss and m a n y other regions having highly organized geometrical arrangements. This growing b o d y of evidence has shown that after partial deafferentation, certain spared afferents can establish synaptic connections on denervated neurons 9. The magnitude o f the response and the onset depended on the animal's agea,9,15 and the region o f the brain involvedZ,S. The cerebellum is a particularly well-suited region tbr study o f plasticity since the parallel fiber projection and Purkinje cell arborization are regularly arranged and parallel fibers monopolize the afferent
input to spines. In the present study we examined synaptic plasticity o f parallel fibers and Purkinje cells after partial deafferentation o f this major input to Purkinje cells. Since parallel fibers usually do not branch, we wanted to determine whether the response to deafferentation in the adult was primarily a function o f Purkinje cell spines or o f parallel fibers and to investigate specifically if remaining parallel fibers t o o k over relinquished postsynaptic sites on spines or formed new connections. MATERIALS AND METHODS The molecular layer in regions adjacent to small lesions placed in the apex o f cerebellar folia was analyzed for ultrastructural changes. Sixty adult female rats f r o m 60 to 90 days o f age were anesthetized with a 0.25 m g injection o f pentobarbital
206 and i n h a l a t i o n of ether. The caudal pole of the
the molecular layer and then moved along the
c r a n i u m was opened for exposure of the cerebellar
sagittal plane just penetrating the depth of tile
vermis. A pair of small transfolial lesions separated
molecular layer. The w o u n d was closed and the
by approximately 1 mm, were directed across the tip
animals were allowed to survive for periods of 2, 4,
of a vermal folium. A microknife, fashioned from a
6, 9, 12, 15, 18, 21 h or 2, 3, 5, 10, 15, 30, 60 days.
30-gauge needle was inserted through the dura, into
Anesthetized animals were perfused through the
Fig. 1. Electron micrograph of parallel fibers m various stages of degeneration. At earlier stages, mitochondria show changes, then the axoplasm of the parallel fibers becomes dense. Note a small dense profile attached to Purkinje cell spine and a long dense shaft of parallel fibers which contains many recognizable organelles. More advanced degeneration consists of phagocytosis by glial process (stippled area) and finally degenerated remnants appeared as dense bodies in the glial processes. (4 h post-lesion; bar equals 0.5/~m.) Fig. 2. Electron micrograph of normal Purkinje cell spine (S) each shrouded by a degenerating parallel fiber. The spine profile in the upper left has continuity with the Purkinje cell dendrite. Note that the parallel fiber caps the spine and extends onto the neck (arrow). (15 h post-lesion; bar equals 0.5 #m.) Fig. 3. Detailed view of capped Purkinje cell spine. A 4-layered membrane complex consists of membrane from a degenerating parallel fiber that is sandwiched between the plasma membranes of a spine and a glial process. Note that one portion of the degenerating parallel fiber remains enlarged over the postsynaptic site. (22 h post-lesion; bar equals 0.5/ml.) Fig. 4. Electron micrograph of vacated Purkinje cell spines. Vacated postsynaptic site (arrow) opposed to glia. Two spine synapses with one parallel fiber bouton are seen in the lower left hand corner. (28 h post-lesion; bar equals 0.5/~m.)
207 aorta with 1 ~o glutaraldehyde and 1 ~ paraformaldehyde buffered with 0.09 M phosphate. Thin sagittal slices of the folium were taken from between and just outside the dual lesions. Control samples were excised from adjacent vermal folia and distant folia on the hemispheres of the same animals and also the same vermal region in normal animals. The aldehyde perfused tissues were postfixed in osmium tetroxide and potassium ferrocyanide for 2 h and then immersed in 2 ~o aqueous uranyl acetate overnight. These tissues were dehydrated through a graded series of alcohol solutions, infiltrated with resin in propylene oxide and embedded in DER epon. Semithin sections (1 /~m) were taken from each block for examination with a light microscope in order to identify the primary region of interest. Thin sections ranging from 60 to 80 nm were mounted on formvar coated, wide, oval slot grids and stained with uranyl acetate and lead citrate. All preparations were examined and photographed on a JEOL 100C transmission electron microscope. RESULTS
Light microscopy Less than one day alter transecting parallel fibers, examination of semithin plastic sections with light microscopy revealed vacuoles as well as dense bodies along the course of the affected parallel fiber beams. The thickness of the molecular layer was still normal at 6-9 h but by 15 h some thinning of the layer was apparent as a small depression in the surface of the folial crest. A marked thinning of the molecular layer was completed by 5-10 days.
Degeneration of parallel fibers Ultrastructural observations at 2 h did not reveal any clearly definable changes in severed parallel fibers. By 4 h the parallel fibers had early signs of degeneration with the most obvious changes in mitochondria. Some shafts of parallel fibers already began to fragment and were being phagocytized by glia. At 6 h, parallel fiber remnants were prominent as phagosomes in glia while dense boutons remained on spines (Fig. 1). These boutons had markedly altered mitochondria and clumped synaptic vesicles in a dense cytoplasmic mass. At 6-24 h, about 3
spines per 1000/~m 2 were capped by a collapsed ghost of a parallel fiber that formed a very narrow cytoplasmic matrix with one small expanded zone containing a dense cytoplasm (Figs. 2 and 3). Essentially Bergmann glial processes surrounded the spine head and a set of sandwiched parallel fiber membranes so that together they formed a 4-component membrane ensemble surrounding the Purkinje cell spine cytoplasm (Fig. 3). During the period from 10-24 h numerous phagosomes were found in the processes of Bergmann astrocytes. By 36-48 h the number of degenerating remnants decreased and were rarely detected after 3 days. Quantitation of parallel fibers at 7 days after surgery revealed 30-60 ~ decreases in the number of parallel fibers per unit area of molecular layer (from 2.4/#m 2 to 4.27/#m2). In addition the thickness of the molecular layer was reduced on the average of 15 ~, thus producing 45-75 ~ reduction in the total number of parallel fibers. Vacated postsynaptic membrane thickenings were observed on Purkinje cell spines as early as 6 hours and increased markedly until 24-36 h (0.3-0.8/1000 #m2), after which they decreased and were rarely observed. These spines appeared normal and had a small postsynaptic thickening (Fig. 4).
Dual innervation of spines: degenerating and normal boutons Purkinje cell spines are rarely contacted by more than one bouton in the normal adult cerebellum. Ten hours following the transections a small percentage of spines had one degenerating bouton opposed to a postsynaptic junction and a normal parallel fiber bouton opposed to a second postsynaptic site (Figs. 5-9). In each case, the degenerating bouton was on the tip of the spine while the bouton with normal features, characteristic of parallel fibers, was positioned on the side of the spine heads. The degenerating bouton had a presynaptic site that was nearly obscured by the dense cytoplasm and vesicles. On occasion a spine was found which had a normal bouton in addition to a vacated synaptic site that faced glial processes containing dense phagosomes (see Fig. 10). Some of these vacated sites appeared to be partially reinnervated by expansion of a reactive bouton onto the vacated area (Fig. 10).
208
Figs. 5-9. Electron micrographs of dual innervation on Purkinje cell spines. One bouton is degenerating and the other bouton is normal. In Figs. 5, 6 and 8, the spines (S) emerge from dendrites (D) and the degenerating boutons are on the tip of the spines while the newly-formed boutons are on the side of the spines. In Fig. 5, the degenerating bouton (B) is being separated from the postsynaptic site by glia. The mitochondrion of the degenerating bouton in Fig. 5 is distinctly altered while in Figs. 6 and 8 the
209 A phagosome was seen in the glial cytoplasm indicating that a degenerating parallel fiber bouton had been removed.
Dual innervation of spines: two normal boutons Accompanying the occurrence of dual innervated spines, that were contacted by a degenerating and a normal bouton, there was also an increase in dual innervations of spines where both presynaptic eleraents appeared normal (Figs. 11 and 12). These synaptic complexes became prominent (approximately 0.2-0.8/1000 #m 2) as compared to their nearly complete lack in control regions. Dual boutons were on opposite sides of the spine head as compared to original synapses which were on the tip of the spine. Usually both boutons were large though on occasion one was small and had only a few synaptic vesicles (Fig. 12). While this figure may represent dual sectioning of the same profile, reconstruction of numerous other terminals showed that dual innervation was present. These normal boutons on the dual innervated spine contained round uniformly sized vesicles distributed similar to those found in parallel fibers, but on occasion one of these dual synapses on spines was a climbing fiber bouton (Fig. 13). At later stages, the dual innervation of spines was not detected.
Development of new spines Besides takeover of spines, there was also formation of new synaptic contacts between parallel fibers and Purkinje cell dendrites. This reinnervation occurred on the shafts of spiny branchlets in regions where dendritic shafts were not shrouded by glial sheaths and parallel fibers were in direct contact with a Purkinje cell dendrite (Fig. 15). This sequence of reactive synaptogenesis began at about 10-14 h with the development of primitive contacts (Figs. 15 and 16). First there was an increase in pinocytosis by
Purkinje cell spiny branchlets giving rise to numerous coated vesicles (Figs. 17-19). Purkinje cell dendrites appeared to sample the parallel fiber plasmalemma by phagocytizing a portion of the membrane through coated vesicle formation (Fig. 20). Then, small membrane densities appeared on the membranes opposed to the axon and dendrite (Fig. 15). During this early phase, synaptic vesicles had not accumulated at the newly-formed synaptic site. In the next phase, it appeared that synaptic vesicles accumulated and the parallel fiber bouton began to enlarge. During the same period, stubby spines emerged from the dendritic shaft with a parallel fiber attached to its tip. By 24 h newly-formed spines could not be distinguished from normal spine profiles.
Giant spines and enlarged synaptic contacts Enlargement of spine heads in the area where parallel fibers were degenerating occurred already by 15 h and reached a peak at about 4 days (Figs. 16 and 22). Each enlarged spine was accompanied by a marked elongation of synaptic profiles. The presynaptic bouton was enlarged and the total number of synaptic vesicles was increased. These modified synaptic contacts were still present at 2 months after the lesion was made. Each of these enlarged spines was characterized by a proliferation of endoplasmic reticulum in the head of the spinel Enlarged spines were most prominent in areas where parallel fiber shafts were greatly reduced in number and Purkinje cell dendrites were closely aligned in the neuropile.
Sprouting of Purkinje cell dendritic processes Active searching for parallel fibers by Purkinje cell spines was suggested by the appearance of elongated spine profiles and complex spine formations (Figs. 14 and 17). This was most frequently
degenerating bouton is dense. Note that the glia appear to be separating the degeneratingboutons (Fig. 8) from the postsynaptic site. The endoplasmicreticulum in the dual innervated spine is very elaborate. In Fig. 7 note the closeapposition of the two boutons indicating that the normal bouton may be extendingonto the postsynapticjunction to take over the postsynapticsite being vacated by the degenerating parallel fiber. (22 h post-lesion; bar equals 0.5/~m.) Fig. 10. Electron micrograph showingone slightlyenlarged spine with part of its postsynaptic membrane specializationopposed by a normal parallel fiber bouton and part of its postsynaptic membrane unoccupied (large arrow). Note one vacated spine (small arrow) and two phagosomesin a glial process. The apposition of glia with phagosome to a spine suggests that the degenerated bouton was recently removed from the spine and that the parallel fiber is taking over the postsynaptic site. (24 h post-lesion; bar equals 0.5 ,urn.)
Figs. 11 and 12. Electron micrograph of dual innervation of spines by two reactive boutons. In Fig. 11, note that the spine (S) contacts a parallel fiber bouton on the left and one on the lower right. In Fig. 12, note that one parallel fiber bouton is very small and the other is comparatively larger. (22 h post-lesion; bar equals 0.5/*m.) Fig. 13. Electron micrograpb showing dual innervation of a Purkinje cell spine by a climbing fiber and a parallel fiber. The climbing fiber (CF) is characterized by its apposition to the longitudinal profile of the dendrite (D) and by large and small vesicles. The apposed parallel fiber on the lower edge of the spine profile has equal size vesicles. (15 h post-lesion; bar equals 0.5/,~m.)
211 observed in regions where parallel fibers had been reduced to about 25 ~ of their original number. Elongated multi-branch spines extended between glial processes and even other parallel fibers to reach available presynaptic elements.
Sprouting of parallel fibers In longitudinal sections of folia, small protuberances of parallel fibers that contacted Purkinje cell spines increased in frequency (Fig. 21). These side branches were 1 or 2 # m in length and never contained tubule cores. Parallel fiber sprouting appeared less often than elongation of Purkinje cell spines. At later stages it appeared that the cut ends of parallel fibers grew out and attempted to circumvent the incision site. This was seen as a dense packing of parallel fibers near the incision where Purkinje cells were lacking as the result of the trauma. Unsectioned parallel fibers may have grown from their tips into the region of parallel fiber degeneration but this could not be determined by these methods. In addition, remaining parallel fibers presumably increased the number of boutons along their course to compensate the loss (unpublished results).
Aberrant synapses Synapses were frequently observed between Purkinje cell spines and stellate or basket cell axons within one day;these synapses are infrequently seen in the normal cerebellum14, 3s. The postsynaptic junction was opposed to basket axons, but there was a lack of vesicle accumulation or presynaptic thickenings (Figs. 23 and 24).
Later changes Five to 30 days after surgery, aberrant synapses
and vacated synaptic sites were rarely seen. However, giant spines with enlarged parallel fiber varicosities prevailed throughout the neuropile in which the parallel fibers were greatly reduced in number (Fig. 22). These enlarged boutons capped the Purkinje spine in a characteristic dome shape so that the entire head of the spine was surrounded by the bouton. The synaptic contact sometimes appeared in multiple profiles on a single spine; however, reconstruction of these boutons revealed that the synapse was very irregular in shape. Some enlarged boutons were contacted by numbers of spine heads. The heads of these spines usually depressed the surface of the large bouton giving the appearance of a scalped rosette (Fig. 25) similar to that described by Sotelo 53 in the cerebellum of the X-irradiated rabbit. Characteristically, these boutons were bounded by other parallel fibers. Other bouton profiles that were contacted by numerous heads of spines were completely shrouded by glia (Fig. 26). These boutons had dense core vesicles and were classified as climbing fibers. DISCUSSION Rapid degeneration of axons was previously described at 10-12 h following lesions of parallel fibers 4 and of afferent fibers to the interpeduncle nucleus ~9. Climbing fibers in the cerebellum also degenerated rapidly after transection of axons or chemical destruction of somas11,54, 55. Our findings revealed not only a rapid onset for degeneration of parallel fibers, but the complete removal of some of these afferents resulting in vacated postsynaptic sites on spines within 6 h of the lesion. Nevertheless, the number of degenerating boutons remained at about 3 ~ during this 2-3-day
Fig. 14. Electron micrograph of complex spine (CS) and a primitive synapse on a normal spine (S). A dumbbell-shaped spine (CS) is in synaptic contact with two parallel fibers. Another spine (S) is forming a new synap'ic contact (arrow) with a parallel fiber. Note that the parallel fiber bouton is not yet enlarged and vesicles have not significantly accumulated. Also note that the junctional complex is very small and is on the side of the spine head. (15 h post-lesion; bar equals 0.5/*m.) Fig. 15. Electron micrograph showing new synaptic contacts on the spine and on a Purkinje cell dendrite. Arrows point to small attachment sites which represent early synapse formation. Note that synaptic vesicles are beginning to accumulate. Two of these contacts are apparently on one existing stubby spine and a third is made directly with a dendrite from which an additional spine will presumably emerge. (20 h post-lesion; bar equals 0.50/~m.) Fig. 16. Electron micrograph showing formation of new synapse on a developing spine and dual synapses on an enlarged Purkinje cell spine. Note that the short stubby spine protrudes from dendrite without a distinct neck and is contacted by a parallel fiber on a very small postsynaptic site. Only a very few vesicles have accumulated. In the upper righthand corner, an enlarged spine (S) with elongated synaptic contact is continuous between the two boutons of reactive parallel fibers. (22 h post-lesion; bar equals 0.5 !tm.)
212
Fig. 17. Low power electron micrograph of an elaborated Purkinje cell dendrite (D) showing two elongated spines (S) and coated vesicles in the cytoplasm. Many coated vesicles (single arrows) are either attached to the plasma membrane or floating free in the cytoplasm. Note a double membrane-coated vesicle (double arrow) which arose by inclusion of a piece of membrane from another process. (23 h post-lesion; bar equals 0.5/~m.)
213 period. This may be due to assorted lengths of severed segments arising from their T-shaped interdigitating arrangement. For example, degeneration may occur very early in short segments and phagocytosis of these may be completed even before longer segments were beginning to show degenerative changes. In the cat, parallel fibers are much longer and can be seen to degenerate as late as 5 days following granule cell lesions 4. The distal portion of longer segments may survive for an additional length of time, being supported by continued axonal transport of substances needed to maintain the integrity of the axonal segment and ultimately synaptic contact 16. This suggests that the length of the distal segment may be a factor in the rate of degeneration in addition to caliber.
Phagocytosis vs vacating of postsynaptic sites Degeneration of parallel fibers following lesions was thought to result in removal of degenerated boutons together with opposed spines from dendrites TM. A number of recent studies from various regions of the brain have shown that postsynaptic sites remained on the dendrite and may even be reutilized2,18,ag,46,sL We did not observe phagocytosis of Purkinje cell spines and their postsynaptic junctions together with parallel fiber boutons. Postsynaptic sites remained on dendritic spines following removal of the degenerating boutons by glial processes.
Reinnervation of Purkinje cell spines: reoccupation of synaptic sites vs formation of new synapses Vacating of postsynaptic sites following degeneration of afferents has been described in a number of regions in the central nervous system2,9,36,39,40,45, 51, 58. However, whether or not these postsynaptic sites
are re-utilized as original units, has not been dearly defined. Raisman and his co-workers45, 46 suggested that postsynaptic sites are permanent and are reoccupied by sprouting afferents. Investigations on mutant mice 25,33,44,52 and induced agranular cerebellalS,~4, 27 suggested that synaptic sites are permanent since vacated postsynaptic sites on spines remained throughout life. This apparent permanence of postsynaptic sites suggests that afferents in the area could easily re-occupy these specializations. One of the first demonstrations of re-occupation of postsynaptic membrane was provided by Westrum 5s. He found evidence that boutons having fiat vesicles (symmetrical synaptic sites) could position themselves over sites previously occupied by boutons with spherical vesicles (asymmetrical type of junction). Evidence is lacking that these contacts are more than transient appositions or have a functional role. In studies of the superior cervical ganglion, vacated sites remained for long periods of time after degeneration of afferents and were believed to be reinnervated by afferents which were allowed to regrow into the cell mass 4°. This study revealed a reciprocal relationship between the number of vacated sites and number of intact synapses throughout the reinnervation period. More recently, Smollen 51 found that the number of vacated sites remained constant throughout development following early denervation of the ganglion, although approximately 50 ~ of the normal number of afferents were attained by adulthood. He 51 concluded that contacts were made on entirely new synaptic sites. Our studies indicate that some vacated synaptic sites may be re-occupied by original or other types of afferents (e.g. basket cell axons), at least for a
Figs. 18-20. Detailed views of coated vesicles. In Fig. 18, 4 coated vesicles are opposed to the dendritic membrane; note that each vesicle marks a site overlying a parallel fiber. In Fig. 19, note coated vesicle is attached to a membrane of a spine and is adjacent to a postsynaptic site. In Fig 20, note the coated vesicle is encapsulating a portion of parallel fiber plasma membrane. Double membrane-coated vesicle can be seen suspended in the cytoplasm of Purkinje cell dendrite. (20-23 h post-lesion; bar equals 0.25/~m). Fig. 21. Electron micrograph showing sprouting of a parallel fiber giving rise to a small process. The tubules of the parallel fiber continue to the left while the sprouting process (on the right) extends into a large bouton synapse on a Purkinje cell spine. (9 h post-lesion; bar equals 0.5/am.) Fig. 22. Electron micrograph showing a giant spine(s) with an elongated synaptic contact. Note that there is a marked increase in the endoplasmic reticulum within the spine head and that no tubules or mitochondria are present. The presynaptic terminal is also proportionally enlarged and filled with abundant round synaptic vesicles characteristic of parallel fibers. (5 days post-lesion; bar equals 0.5 ,urn.)
214
Fig. 23. Electron micrograph ofa stellate or basket cell axon (B) making contact with a dendritic shaft (D) and with a long spine(s). Note the pleomorphic vesicles at the junction with the dendritic shaft (D) and a lack of synaptic vesicles near the spine contact (S) (arrow) which has a distinct postsynaptic thickening. (I 5 h post-lesion; bar equals 0.5 !~m.)
215 transient period. However, this reinnervation is preceded by new synapse formation. These newlyformed synapses developed on original spines as well as by formation of new spines. Dual innervation of spines in the normal adult cerebellum is uncommon while partial denervation of Purkinje cell spines resulted in 0.8-1.2 complexes for every 1000/~m 2 of section (i.e. 7-27 ~ of the degenerated bouton observed). During early periods, degenerating boutons were present on the same spine that had a normal bouton (Fig. 27). Later, there was a shift toward single boutons on spines as revealed by reconstruction of spines which contacted more than one bouton profile. Some of these elongated boutons had incomplete capping of spines resulting in the occasional appearance of multiple bouton profiles on spines at 5 and 7 days. Reconstruction of serial sections revealed that these multiple structures were single parallel fiber contacts with irregular and elongated synaptic contact area. This dual innervation of spines found here may have similarities to multiple innervations found during development in climbing fiber 1°, muscle 47 or submandibular ganglion 26 where a one-to-one innervation emerges following multiple innervations. A selective5 or competitive elimination 43 process for stabilization of synapses was suggested in the transformation of multiple synapses on individual spines to single contacts. This may also be reflected in our study since a single synapse on a spine is the stable synaptic relationship in the cerebellum. Four possible processes of reactive re-organization that give rise this stabilization following reduction of afferents by degeneration are illustrated in Fig. 27. An enlarged single contact on a spine was a prominent occurrence after 5 days following lesions. This was achieved through bouton elimination and the extension of postsynaptic sites so that the vacated postsynaptic site merged with the newly formed site (Fig. 27Ca ana 4). Also postsynaptic sites and their
spines must be lost since the total number of synapses is reduced while the number of Purkinje cells remained constant. Vacated sites disappear after 5-7 days. Thus molecules of postsynaptic sites are reabsorbed and either decomposed or possibly reutilized in the enlargement ofpostsynaptic sites (Fig. 27C1 and 2)-
Giant spines and elongation of synaptic profiles Enlargement of spines with multiple innervations was previously described in the reeler mutants 33. Giant spines appear following malnutrition in the developing rat 6. This study shows that spines as well as synapses enlarge in the adult cerebellum when afferents are reduced. Remodeling of synapses in the molecular layer following lesions also resulted in enlargement of spines and an associated increase in the size of contact areas. The amount of this enlargement was related to the reduction in number of parallel fibers. Even though remaining parallel fibers compensated by forming new synapses on spines and presumably were able to take over some vacated synaptic sites, the compensation was limited by the maximum capability of the granule cell axons to respond to a net deficit in the number of synapses on Purkinje cells Thus numerous spines and their synapses were lost. Accompanying the loss of contact areas there was an increase in the length of the synaptic profiles. We have demonstrated in a malnutrition model, that average synaptic contact areas increased inversely with a deficit in the number of synapses. This increase in individual contact size was shown to be due to a constancy of total contact area for spine synapses on each Purkinje cell2L Reductions in the number of parallel fibers by lesioning the adult cerebellum also supports the finding of constancy in total contact area. Preliminary measurements show that the area of synaptic contact attains values which are expected for reductions in the number of synapses 23.
Fig. 24. Electron micrograph of a basket axon contacting a spine (S) having a postsynaptic membrane specialization. There are no vesicles opposed to this junction (arrow). (28 h post-lesion; bar equals 0.5 ,urn.) Fig. 25. Electron micrograph of a reactive parallel fiber terminal synapsing with 6 Purkinje cell spines (S). Note, one spine is completely encircled within the terminal which contains uniform synaptic vesicles. (28 days post-lesion; bar equals 0.5 ~m.) Fig. 26. Electron micrograph of a reactive climbing fiber varicosity forming multiple synaptic contacts with 5 moderately enlarged Purkinje cell spines (S). The bouton has densely packed large and small vesiclescharacterized by a climbing fiber bouton. (24 h postlesion; bar equals 0.5/zm.)
REACTIVE-REORGANIZATION OF SYNAPSES
A NORMAL
C SYNAPTIREORGANI C ZATION D STABILIZATION ABSORP|TO I NI~ :~'~ 1 I ~
BDEGENERATION 3
~rENst,.,.
Fig. 27. Diagram showing rapid synaptic reorganization after partial deafferentation of Purkinje cell spine synapses. Sectioning of parallel fibers in adult cerebella results in rapid degeneration in distal segments. The first change is seen in mitoehondria. This is followed by an active phagocytosis of dense debris by glia processes. Parallel fibers which remain in the area reinnervate some of these spines by forming new contact sites and by reoccupying vacated postsynaptic sites while other spines are vacated and lost. A: normal bouton-spine relationship in the neuropile showing synapse on the tip of a spine which are nearly completely capped by a glial sheath (stippled). B: a degenerating bouton on spine and two degenerating parallel fibers (dotted). Glia react to assuming a phagocytotic role retract from the neck of the spine and allow parallel fibers to come in direct opposition to the spine. C: synaptic reorganization is a transient phase where trial contacts may occur (e.g. aberrant synapses and dual innervations). This reorganization gives rise to stable connections through selective or competitive elimination. Shown are 4 processes of synaptic reorganization which are supported by our morphological findings. In the first type, a presynaptic bouton was removed by glial phagocytosis and a vacated postsynaptic site came in contact with glia. This site could either be reinnervated by available parallel fibers or else be absorbed with retraction of the spine. A second type represents reinnervation of spines by a new synapse and occurs even before the degenerating bouton vacates the postsynaptic site. This indicates that the onset of reactive synaptogenesis ensues as soon as degeneration starts or possibly functional synaptic activity ceases. This synaptic site could also be absorbed and the new synapse would remain. In the third type, the new synapse takes over the vacated postsynaptic site by extension of the bouton and its contact site so that new and old postsynaptic sites merge. In type 4, dual innervation by 2 reactive parallel fibers results in formation of one new synaptic site and the takeover of vacated synaptic site by another fiber. Also new sites can be formed by two new boutons. D: stabilization phase represents the loss of some uninnervated spines and postsynaptic sites. A competitive elimination of dual synapses on spines results in one-to-one bouton-spine synapses and allows compensation for the loss of synapses by increasing synaptic sizes. These giant spines with single enlarged synapses seem to be the resulting stable and possibly more efficient synapses.
217 The mechanism for the enlargement of synapses, that is suggested in this study, is the takeover of vacated sites by newly-formed synapses as well as synapse elimination. The newly-formed boutons, presumably can expand and merge with vacated sites. Likewise, dual-innervated spines that develop during the reaction period relinquish one bouton. Such vacated sites could also be incorporated into a single giant site. This suggests that the macromolecules making up the postsynaptic membrane thickenings are mobile and can move from vacated sites to new synapes or else there is a continual turnover a9 of macromolecules possibly form a pool that allows synapses to develop where they are demanded. Mobility of attachment and receptor molecules and their shifting to adjacent sites would have to occur if the components of the membrane densities are actually permanent rather than their location. The question of permanency of these synaptic attachment molecules versus their continual turnover with ability to be inserted at new active sites remains unanswered. Nevertheless, the ability of new synapses to form in a very short time period, even before synaptic sites are vacated, strongly supports a turnover mechanism. The apparent permanence of postsynaptic sites may actually have an underlying continuous rapid turnover of macromolecules related to postsynaptic membrane thickenings. This turnover may allow deposition to occur at more favorable sites. Alternatively, there is a reservoir of macromolecules that responds to the deafferentation. Coated vesicles were first observed in degeneating dendrites 57 and were thought to transport and insert macromolecules at synaptic attachment sitesL Attachment molecules and receptor molecules may likely be separable entities.
Formation of new synapses and new spines Developmental synaptogenesis follows a progressive course where primitive contacts form and then synaptic vesicles accumulate resulting in bouton and spine development 37. We observed a similar process in the adult cerebellum as reactive synaptogenesis following the removal of parallel fiber afferents. The loss of afferents to Purkinje cell spines appeared to produce an immediate increase in the insertion of attachment molecules at new sites. A very interesting observation was that Purkinje
cell dendrites actually appeared to pinch off a portion of the parallel fiber membrane (Fig. 20). This could indicate a 'tasting' mechanism whereby one process recognized the other. Coated vesicles have been attributed to membrane recycling12,20 through a basket retrieval mechanism 19. Others suggested that coated vesicles served a role in insertion of postsynaptic membrane attachment sitesl,4s, 56. Privat a2 proposed that coated vesicles were part of a mechanism for removing the attachment site. Recent studies showed a marked association of these coated vesicles with axonal growth during development 12 as well as in reactive sproutingZ~,32,sL A major question remains as to the direction that these specialized vesicles move or whether they represent two simultaneous membrane displacing processes; one inserting molecules and the other shifting membrane as well as 'sampling' the environment. Glia appeared to be a major factor in determining where new synapses were formed on dendrites. New synaptic contacts were found only in those regions where parallel fibers were in direct contact with spiny branchlets. Glial involvement could explain why there was an irregular shift of parallel fiber-synaptic contacts down main dendritic trees of Purkinje cells following climbing fiber deafferentation54,55 (Hillman, unpublished observations). Bergmann astrocytic processes relinquished their position on main dendritic trees following climbing fiber deafferentation, may be due to their active role in phagocytosis of degenerating remnants. This concept is supported by the occurrence of spines on the larger dendritic shafts where glial ensheathment of dendrites is lacking, such as in reptiles 21.
Sprouting of parallel fibers and dendrites Parallel fibers, on occasion, make small side projections. These were observed more frequently following deafferentation, thus indicating an early form of collateral sprouting. However, it was clear that these side processes did not contain tubules and thus apparently did not extend for more than a few microns. Axons that were in apposition to dendrites and spines appeared to be the major contributor to reactive synaptogenesis rather than growth of axon through sprouting. Some parallel fiber boutons were greatly enlarged and capped by multiple spine heads; others were
218 enlarged but smaller and capped one spine. These enlarged boutons may have been on proximal segments of parallel fibers that remained after a significant portion of their axon had been severed. Sprouting or regrowth from the ends of parallel fibers that were not severed may have occurred, but was not detectable by methods used here. However, the cut ends of parallel fibers grew and had a very compact arrangement as they attempted to circumvent the scar. Sprouting of spines was also evident as shown by multiple spine heads which extended for 3-4 ttm into the neuropile. Some synapses on these complex spines were on small protuberances near the neck of spines while others were individual spine-like processes that were based on a 'spine trunk'. Similar complex spines were described in the hippocampusa4,aL Reactive synaptogenesis in the cerebellar molecular layer: presynaptic vs postsynaptic control Reactive sprouting of axons has revealed a potential for reconnectivity in the adult peripheral and central nervous system that is not possible by regenerative growth. The message for instituting these outgrowths in axons is not understood. In the peripheral nervous system, evidence suggests that substances directing sprouting appears confined to nerve paths (Schwann cells and perineural sheaths) which connect denervated muscles to sheaths containing normal fibers 5°. Parallel fibers have a form of side sprouting in that small processes can emerge from fibers along their course. These parallel fiber sprouts, however, are greatly limited, presumably due to the fact that the T-shaped parallel fiber has reached a limit in its bifurcation potential and does not have sufficient numbers of tubules to support additional branching. In this study reactive synaptogenesis occurred independent of vacating of sites with the initiation of synapse formation following severing of afferent
REFERENCES 1 Altman, J., Coated vesicles and synaptogenesis. A developmental study in the cerebellar cortex of the rat, Brain Research, 30 (1971) 311-312. 2 Bernstein, J. J. and Bernstein, M. E., Neuronal alteration
fibers to Purkinje cells. The rapid occurrence of synaptogenesis indicates that the postsynaptic element, the Purkinje cell itself, contains the potential to form synapses with afferents which are in direct contact. Furthermore, postsynapticjunctional molecules were available to form new synaptic sites. This reactive synaptogenesis revealed a potential of these neurons to form multip!e synaptic contacts on individual spines. Our findings further support the contention that postsynaptic structures can reorganize to shift position and size of synaptic connections, possibly by returning to an immature-like state. The potential of Purkinje cells to make new connections appears to be related to the immediate availability of postsynaptic membrane structures which can be inserted at appositional zones between the dendrite and presynaptic elements. The signal appears to be related to a communication mechanism which is activated in the postsynaptic cell and allows receptor molecules and attachment molecules to concentrate at potential synaptic sites. A second stage of postsynaptic control over reactive synaptogenesis was the elimination of multiple synapses on spines and the ultimate incorporation of postsynaptic membranes into enlarged synaptic sites. The principle underlying compensatory enlargement through elimination, may be dictated by the postsynaptic neuron in its response to maintain a constant level of synaptic contact area2L In this hypothesis, the response of both 'appositional' and 'sprouting' reactive synaptogenesis are mediated through the availability of postsynaptic macromolecules to be inserted into active zones. These could be activated through cessation of transmitter release following severing of axons. ACKNOWLEDGEMENTS Research was supported by U.S. Public Health Service grants HD-10934 and NS-13742 from N I C H D and N I N C D S .
and reinnervation following axonal regeneration and sprouting in the mammalian spinal cord, Brain Behav. EvoL, 8 (1973) 135-161. 3 Bjorklund, A. and Stenevi, L., Regeneration of monoaminergic and cholinergic neurons in the mammalian central nervous system, Physiol. Rev., 59 (1979) 62-100.
219 4 Brand, S., Dahl, A. L. and Mugnaini, E., The length of parallel fibers in the cat eerebellar cortex. An experimental light and electron microscopic study, Exp. Brain Res., 26 (1976) 39-58. 5 Changeux, J. and Danchin, A., Selective stabilisation of developing synapses as a mechanism for the specification of neuronal networks, Nature (Lond.), 264 (1976) 705-712. 6 Chen, S. and Hillman, D E., Giant spines and enlarged synapses induced in Purkinje cells by malnutrition, Brain Research, 187 (1980) 487-493. 7 Colonnier, M., Experimental degeneration in the cerebral cortex, J. Anat. (Lond.), 98 (1964) 47-53. 8 Cotman, C. W., Neuronal Plasticity, Raven Press, New York, 1978. 9 Cotman, C. W. and Nadler, J. V., Reactive synaptogenesis in the hippocampus. In C. W. Cotman (Ed.), Neuronal Plasticity, Raven Press, New York, 1978, pp. 229-260. 10 Crepel, F., Mariani, J. and Dehhaye-Bouchaud, N., Evidence for a multiple innervation of Purkinje cells by climbing fibers in the immature rat cerebellum, J. Neurobiol., 7 (1976) 567-578. 11 Desclin, J. C. and Escubi, J., An additional silver impregnation method for demonstration of degenerating nerve cells and processes in the central nervous system, Brain Research, 93 (1975) 25-39. 12 Eckenhoff, M. F. and Pysh, J. J., Double-walled coated vesicle formation: evidence for massive and transient conjugate internalization of plasma membranes during cerebellar development, J. NeurocytoL, 8 (1979) 623-638. 13 Field, P. M., Synapse formation after injury in adult brain: failure of fimbrial axons to reinnervate the bed nucleus of the stria terminars, Brain Research, 189 (1980) 91-101. 14 Fox, C. A., Hillman, D. E., Siegesmund, K. A. and Dutta, C. R., The primate cerebellar cortex: a Golgi and electron microscopic study. In C. A. Fox and R. S. Snider (Eds.), Progress in Brain Research, VoL 25, Elsevier Press, Amsterdam/London/New York, 1967, pp. 174-225. 15 Gall, C. and Lynch, G., Rapid axon sprouting in the neonatal rat hippocampus, Brain Research, 153 (1978) 357-362. 16 Goldowitz, D. and Cotman, C. W., Do neurotrophic interaction control synapse formation in the adult rat brain? Brain Research, 181 (1980) 325-344. 17 Graziadei, P. P. C. and Gradiadei, G. A. M., The olfactory system: a model for the study of neurogenesis and axon regeneration in mammals. In C. W. Cotman (Ed.), Neuronal Plasticity, Raven Press, New York, 1978, pp. 131-153. 18 Herndon, R. M., Tiophen induced granule cell necrosis in rat cerebellum. An electron microscopic study, Exp. Brain Res., 6 (1968) 49-68. 19 Heuser, J. E., Three-dimensional visualization of coated vesicle formation in fibroblasts, J. Cell Biol., 84 (1980) 560-583. 20 Heuser, J. E. and Reese, T. S., Evidence for recycling of synaptic vesicle membrane during transmitter release at the frog neuromuscular junction, J. Cell Biol., 57 (1973) 315-344. 21 Hillman, D. E., Neuronal organization of the eerebellar cortex in amphibia and reptilia. In R. Llinas (Ed.), Neurobiology of Cerebellar, Evolution and Development, Am. Med. Assoc., Chicago, 1969, pp. 279-325.
22 Hillman, D. E. and Chen, S., Vulnerability of cerebellar development in malnutrition. II. Intrinsic determination of total synaptic areas on Purkinje cell spines, Neuroscience, 6: 1263-1275, 1981. 23 Hillman, D. E. and Chen, S., Plasticity of synaptic size with constancy of total synaptic contact area on Purkinje cells in the cerebellum. In E. A. Vidrio and M. A. Galina (Eds.), Symposia Proceedings of the XI International Congress of Anatomy, Alan Liss, New York, pp. 229-245. 24 Hirano, A., Dembitzer, H. M. and Jones, M., An electron microscopic study of cyasin-induced cerebellar alteration, J. Neuropath. Exp. Neurol., 31 (1972) 113-125. 25 Landis, D. M. and Sidman, R., Electron microscopic analysis of postnatal histogenesis in the cerebellar cortex of staggerer mutant mice, J. comp. Neurol., 179: 831-864, 26 Lichtman, J. W., The reorganization of synaptic connexions in the rat submandibular ganglion during postnatal development, J. Physiol. (Lond.), 273 (1977) 155-177. 27 Llinas, R., Hillman, D. E. and Precht, W., Neuronal circuit reorganization in mammalian agranular eerebellar cortex, J. Neurobiol., 4 (1973) 69-94. 28 Lund, R. D., Development and Plasticity of the Brain, Oxford University Press, New York, 1978. 29 Lund, R. D. and Lund, J. S., Synaptic adjustment after deafferentation of the superior colliculus of the rat, Science, 171 (1971) 804-807. 30 Lynch, G., Rose, G., Gall, C. and Cotman, C. W., The response of the dentate gyrus to partial deafferentation. In M. Santini (Ed.), Golgi Centennial Symposium Proceedings Raven Press, New York, 1975, pp. 505-517. 31 McWilliams, R. and Lynch, G., Terminal proliferation and synaptogenesis following partial deafferentation: The reinnervation of the inner molecular layer of the dentate gyrus following removal of its commissural afferents, J. comp. Neurol., 180 (1978) 581-616. 32 McWilliams, J. R. and Lynch, G., Sprouting in the hippocampus is accompanied by an increase in coated vesicles, Brain Research, 211 (1981) 158-164. 33 Mariani, J., Crepel, F., Mikoshiba, K., Changeux, J. P. and Sotelo, C., Anatomical, physiological and biochemical studies of the cerebellum from reeler mutant mouse, Physiol. Trans. roy Soc. (Lond.), 281 (1977) 1-28. 34 Matthews, D. A., Cotman, C. W. and Lynch, G. S., A quantitative ultrastructural analysis of synaptic changes in the rat dentate gyrus following entorhinal lesions. I. Magnitude and time course of degeneration, Brain Research, 115 (1976) 1-21. 35 Matthews, D. A., Cotman, C. and Lynch, G., A quantitative ultrastructural analysis of synaptic changes in the rat dentate gyrus following entorhinal lesions. II. Reacquisition of morphologically normal synaptic contacts, Brain Research, 115 (1976) 23-41. 36 Mouren-Mathieu, A. M. and Colonnier, M., The molecular layer of the adult cat eerebellar cortex after lesions of the parallel fiber: an optic and electron microscopic study, Brain Research, 16 (1969) 307-323. 37 Mugnaini, E., Ultrastructural studies on the cerebellar histogenesis. II. Maturation of nerve cell populations and establishment of synaptic connections in the cerebellar cortex of the chick. In R. Llinas (Ed.), Neurobiology of Cerebellar Evolution and Development, Amer. Med. Assn., Chicago, 1969, pp. 749-782. 38 Mugnaini, E., The relation between cytogenesis and the formation of different types of synaptic contact, Brain
220
Research, 17 (I 970) 169-179. 39 Murray, M., Zimmer, J. and Raisman, G., Quantitative electron microscopic evidence for reinnervation in the adult rat interpeduncular nucleus after lesions of the fasciculus retroflexus, J. comp. Neurol., 187 (1979) 447-468. 40 Ostberg, A.-J. C., Raisman, G., Field, P. M., Iversen, L. L. and Zigmond, R. E., A quantitative comparison of the formation of synapses in the rat superior cervical sympathetic ganglion by its own and by foreign nerve fibers, Brain Research, 107 (1976) 445-470. 41 Pettigrew, J. D., The paradox of the critical period for striate cortex. In C. W. Cotman (Ed.), NeuronalPlasticity, Raven Press, New York, 1978, pp. 311-329. 42 Privat, A., A possible mechanism for the resorption of attachment plates in the growing rat brain, Brain Research, 69 (1974) 125-129. 43 Purves, D. and Lichtman, J. W., Elimination of synapses in the developing nervous system, Science, 210 (1980) 153-157. 44 Rakic, P., Synaptic specificity in the cerebellar cortex: study of anomalous circuits induced by single gene mutations in mice, CoM Spr. Harb. Symp. quant. BioL, 40 (1976) 333-346. 45 Raisman, G. and Field, P. M., A quantitative investigation of the development of collateral reinnervation after partial deafferentation of the septal nuclei, Brain Research, 50 (1973) 241-264. 46 Raisman, G., Field, P. M., Ostberg, A.-J. C., Iversen, L. L. and Zigmond, R. E., A quantitative ultrastructural and biochemical analysis of the process of reinnervation of the superior cervical ganglion in the adult rat, Brain Research, 71 (1974) 1-16. 47 Redfern, P. A., Neuromuscular transmission in newborn rats, J. Physiol. (Lond.), 209 (1970) 701-709. 48 Reese, R. P., Bunge, M. B. and Bunge, R. P., Morphological changes in the neuritic growth cone and target neuron during synaptic junction development in culture, J. cell Biol., 68 (1976) 240-263. 49 Ruffalo, R., Eisenbarth, G. S., Thompson, J. M. and
50
51
52
53
54
55
56
57
58
59
Nirenberg, M., Synapse turnover: a mechanism for acquiring synaptic specificity, Proc. nat. Acad. Sci. U.S.A., 75 (1978) 2281-2285. Slack, J. R., Hopkins, W. G., Williams, M. N., Nerve sheaths and motoneurone collateral sprouting, Natur6 (Lond.), 282 (1979) 506-507. Smolen, A. J., Postnatal development of ganglionic neurons in the absence of preganglionic input: morphological observations in synapse formation, Develop. Brain Res., 1 (1981) 49-58. Sotelo, C., Anatomical, physiological and biochemical studies of the cerebellum from mutant mice. II. Morphological study of cerebellar cortical neurons and circuits in the Weaver mouse, Brain Research, 94 (1975) 19-44. Sotelo, C., Synaptic remodeling in mutants and experimental animals. In F. Vital-Durand (Ed.), Aspects of neural plasticity, Vol. 43 pp. 167-190 and Jeannerod, M. INSERM 11-12, 1975. Sotelo, C. and Arsenio-Nunes, M. L., Development of Purkinje cell in absence of climbing fibers, Brain Research, 111 (1976) 389-395. Sotelo, C., Hillman, D. E., Zamora, A. J. and Llinas, R., Climbing fiber deafferentation : Its action on Purkinje cell dendritic spines, Brain Research, 98 (1975) 574-58l. Stelzner, D. J., Martin, A. H. and Scott, G. L., Early stages of synaptogenesis in the cervical spinal cord of the chick embryo, Z. Zellforseh., 138 (1973) 475-488. Walberg, F., Role of normal dendrite in removal of degenerating terminal boutons, Exp. NeuroL, 8 : 112-124, 1963. Westrum, L. E., Electron microscopy of degeneration in the lateral olfactory tract and plexiform layer of the prepyriform cortex of the rat, Z. Zellforsch., 98 (1969) 157 187. Westrum, L. E. and Black, R. G., Fine structural aspects of the synaptic organization of the spinal trigeminal nucleus (pars interpolaris) of the cat. Brain Research, 25 (1971) 265-287.
205
Research Reports
Plasticity of the Parallel Fiber-Purkinje Cell Synapse by Spine Takeover and New Synapse Formation in the Adult Rat SUZANNE CHEN and DEAN E. HILLMAN Department of Physiology and Biophysics, New York University Medical Center, 550 First Avenue, New York, N Y 10016 (U.S.A.)
(Accepted October 22nd, 1981) Key words: deafferentation - - reactive synaptogenesis - - Purkinje cell spines - - parallel fibers - - plasticity - -
giant spines - - cerebellum - - synapse
Alteration in synaptic connectivity between Purkinje cell spines and parallel fibers of the cerebellum were studied following partial deafferentation of Purkinje cells in the adult rat. Transection of parallel fibers by two lesions placed at a I mm interval on the folial crest were used to produce degeneration of these afferents. Ultrastructural analysis of synapses on Purkinje cell spines revealed degeneration with vacating of postsynaptic sites within 6 h. Reactive synaptogenesis as takeover of Purkinje cell spines by formation of new synapses from remaining parallel fibers occurred even before degenerating parallel fibers had vacated postsynaptic sites. This was accompanied by a marked increase in the number of dual innervations by reactive parallel fibers within one day. Some vacated postsynaptic sites were lost as indicated by a reduction in the number of synapses and others may have been taken over by newly formed synapses on spines. In addition, new synapses formed between the shafts of Purkinje cell branchlets and parallel fibers. Sprouting of parallel fibers occurred as small extensions without tubules while Purkinje cell spines reacted by forming elongated and multiple heads which contacted different parallel fibers. After 5 days degenerating boutons were rarely found. Enlarged spine heads were each capped by a proportionally enlarged parallel fiber bouton and joined by an elongated synaptic junction to parallel fibers. Some parallel fiber boutons were greatly enlarged and capped numerous profiles of spines. This study shows that formation of new pre- and postsynaptic sites takes precedence over reoccupation of original contacts and that multiple synapses on individual spines are being eliminated to give rise to single contacts with boutons. This elimination resulted in enlargement of synaptic contact areas between Purkinje cell spines and parallel fibers by taking over postsynaptic sites from some vacated and eliminated boutons. INTRODUCTION In the last decade reactive synaptogenesis t h r o u g h sprouting has been demonstrated in a n u m b e r o f regions o f the central nervous system such as the h i p p o c a m p u s s,3°,34,85, septum la,45, optic system 2s, 29,41, olfactory bulblV,Ss and m a n y other regions having highly organized geometrical arrangements. This growing b o d y of evidence has shown that after partial deafferentation, certain spared afferents can establish synaptic connections on denervated neurons 9. The magnitude o f the response and the onset depended on the animal's agea,9,15 and the region o f the brain involvedZ,S. The cerebellum is a particularly well-suited region tbr study o f plasticity since the parallel fiber projection and Purkinje cell arborization are regularly arranged and parallel fibers monopolize the afferent
input to spines. In the present study we examined synaptic plasticity o f parallel fibers and Purkinje cells after partial deafferentation o f this major input to Purkinje cells. Since parallel fibers usually do not branch, we wanted to determine whether the response to deafferentation in the adult was primarily a function o f Purkinje cell spines or o f parallel fibers and to investigate specifically if remaining parallel fibers t o o k over relinquished postsynaptic sites on spines or formed new connections. MATERIALS AND METHODS The molecular layer in regions adjacent to small lesions placed in the apex o f cerebellar folia was analyzed for ultrastructural changes. Sixty adult female rats f r o m 60 to 90 days o f age were anesthetized with a 0.25 m g injection o f pentobarbital
206 and i n h a l a t i o n of ether. The caudal pole of the
the molecular layer and then moved along the
c r a n i u m was opened for exposure of the cerebellar
sagittal plane just penetrating the depth of tile
vermis. A pair of small transfolial lesions separated
molecular layer. The w o u n d was closed and the
by approximately 1 mm, were directed across the tip
animals were allowed to survive for periods of 2, 4,
of a vermal folium. A microknife, fashioned from a
6, 9, 12, 15, 18, 21 h or 2, 3, 5, 10, 15, 30, 60 days.
30-gauge needle was inserted through the dura, into
Anesthetized animals were perfused through the
Fig. 1. Electron micrograph of parallel fibers m various stages of degeneration. At earlier stages, mitochondria show changes, then the axoplasm of the parallel fibers becomes dense. Note a small dense profile attached to Purkinje cell spine and a long dense shaft of parallel fibers which contains many recognizable organelles. More advanced degeneration consists of phagocytosis by glial process (stippled area) and finally degenerated remnants appeared as dense bodies in the glial processes. (4 h post-lesion; bar equals 0.5/~m.) Fig. 2. Electron micrograph of normal Purkinje cell spine (S) each shrouded by a degenerating parallel fiber. The spine profile in the upper left has continuity with the Purkinje cell dendrite. Note that the parallel fiber caps the spine and extends onto the neck (arrow). (15 h post-lesion; bar equals 0.5 #m.) Fig. 3. Detailed view of capped Purkinje cell spine. A 4-layered membrane complex consists of membrane from a degenerating parallel fiber that is sandwiched between the plasma membranes of a spine and a glial process. Note that one portion of the degenerating parallel fiber remains enlarged over the postsynaptic site. (22 h post-lesion; bar equals 0.5/ml.) Fig. 4. Electron micrograph of vacated Purkinje cell spines. Vacated postsynaptic site (arrow) opposed to glia. Two spine synapses with one parallel fiber bouton are seen in the lower left hand corner. (28 h post-lesion; bar equals 0.5/~m.)
207 aorta with 1 ~o glutaraldehyde and 1 ~ paraformaldehyde buffered with 0.09 M phosphate. Thin sagittal slices of the folium were taken from between and just outside the dual lesions. Control samples were excised from adjacent vermal folia and distant folia on the hemispheres of the same animals and also the same vermal region in normal animals. The aldehyde perfused tissues were postfixed in osmium tetroxide and potassium ferrocyanide for 2 h and then immersed in 2 ~o aqueous uranyl acetate overnight. These tissues were dehydrated through a graded series of alcohol solutions, infiltrated with resin in propylene oxide and embedded in DER epon. Semithin sections (1 /~m) were taken from each block for examination with a light microscope in order to identify the primary region of interest. Thin sections ranging from 60 to 80 nm were mounted on formvar coated, wide, oval slot grids and stained with uranyl acetate and lead citrate. All preparations were examined and photographed on a JEOL 100C transmission electron microscope. RESULTS
Light microscopy Less than one day alter transecting parallel fibers, examination of semithin plastic sections with light microscopy revealed vacuoles as well as dense bodies along the course of the affected parallel fiber beams. The thickness of the molecular layer was still normal at 6-9 h but by 15 h some thinning of the layer was apparent as a small depression in the surface of the folial crest. A marked thinning of the molecular layer was completed by 5-10 days.
Degeneration of parallel fibers Ultrastructural observations at 2 h did not reveal any clearly definable changes in severed parallel fibers. By 4 h the parallel fibers had early signs of degeneration with the most obvious changes in mitochondria. Some shafts of parallel fibers already began to fragment and were being phagocytized by glia. At 6 h, parallel fiber remnants were prominent as phagosomes in glia while dense boutons remained on spines (Fig. 1). These boutons had markedly altered mitochondria and clumped synaptic vesicles in a dense cytoplasmic mass. At 6-24 h, about 3
spines per 1000/~m 2 were capped by a collapsed ghost of a parallel fiber that formed a very narrow cytoplasmic matrix with one small expanded zone containing a dense cytoplasm (Figs. 2 and 3). Essentially Bergmann glial processes surrounded the spine head and a set of sandwiched parallel fiber membranes so that together they formed a 4-component membrane ensemble surrounding the Purkinje cell spine cytoplasm (Fig. 3). During the period from 10-24 h numerous phagosomes were found in the processes of Bergmann astrocytes. By 36-48 h the number of degenerating remnants decreased and were rarely detected after 3 days. Quantitation of parallel fibers at 7 days after surgery revealed 30-60 ~ decreases in the number of parallel fibers per unit area of molecular layer (from 2.4/#m 2 to 4.27/#m2). In addition the thickness of the molecular layer was reduced on the average of 15 ~, thus producing 45-75 ~ reduction in the total number of parallel fibers. Vacated postsynaptic membrane thickenings were observed on Purkinje cell spines as early as 6 hours and increased markedly until 24-36 h (0.3-0.8/1000 #m2), after which they decreased and were rarely observed. These spines appeared normal and had a small postsynaptic thickening (Fig. 4).
Dual innervation of spines: degenerating and normal boutons Purkinje cell spines are rarely contacted by more than one bouton in the normal adult cerebellum. Ten hours following the transections a small percentage of spines had one degenerating bouton opposed to a postsynaptic junction and a normal parallel fiber bouton opposed to a second postsynaptic site (Figs. 5-9). In each case, the degenerating bouton was on the tip of the spine while the bouton with normal features, characteristic of parallel fibers, was positioned on the side of the spine heads. The degenerating bouton had a presynaptic site that was nearly obscured by the dense cytoplasm and vesicles. On occasion a spine was found which had a normal bouton in addition to a vacated synaptic site that faced glial processes containing dense phagosomes (see Fig. 10). Some of these vacated sites appeared to be partially reinnervated by expansion of a reactive bouton onto the vacated area (Fig. 10).
208
Figs. 5-9. Electron micrographs of dual innervation on Purkinje cell spines. One bouton is degenerating and the other bouton is normal. In Figs. 5, 6 and 8, the spines (S) emerge from dendrites (D) and the degenerating boutons are on the tip of the spines while the newly-formed boutons are on the side of the spines. In Fig. 5, the degenerating bouton (B) is being separated from the postsynaptic site by glia. The mitochondrion of the degenerating bouton in Fig. 5 is distinctly altered while in Figs. 6 and 8 the
209 A phagosome was seen in the glial cytoplasm indicating that a degenerating parallel fiber bouton had been removed.
Dual innervation of spines: two normal boutons Accompanying the occurrence of dual innervated spines, that were contacted by a degenerating and a normal bouton, there was also an increase in dual innervations of spines where both presynaptic eleraents appeared normal (Figs. 11 and 12). These synaptic complexes became prominent (approximately 0.2-0.8/1000 #m 2) as compared to their nearly complete lack in control regions. Dual boutons were on opposite sides of the spine head as compared to original synapses which were on the tip of the spine. Usually both boutons were large though on occasion one was small and had only a few synaptic vesicles (Fig. 12). While this figure may represent dual sectioning of the same profile, reconstruction of numerous other terminals showed that dual innervation was present. These normal boutons on the dual innervated spine contained round uniformly sized vesicles distributed similar to those found in parallel fibers, but on occasion one of these dual synapses on spines was a climbing fiber bouton (Fig. 13). At later stages, the dual innervation of spines was not detected.
Development of new spines Besides takeover of spines, there was also formation of new synaptic contacts between parallel fibers and Purkinje cell dendrites. This reinnervation occurred on the shafts of spiny branchlets in regions where dendritic shafts were not shrouded by glial sheaths and parallel fibers were in direct contact with a Purkinje cell dendrite (Fig. 15). This sequence of reactive synaptogenesis began at about 10-14 h with the development of primitive contacts (Figs. 15 and 16). First there was an increase in pinocytosis by
Purkinje cell spiny branchlets giving rise to numerous coated vesicles (Figs. 17-19). Purkinje cell dendrites appeared to sample the parallel fiber plasmalemma by phagocytizing a portion of the membrane through coated vesicle formation (Fig. 20). Then, small membrane densities appeared on the membranes opposed to the axon and dendrite (Fig. 15). During this early phase, synaptic vesicles had not accumulated at the newly-formed synaptic site. In the next phase, it appeared that synaptic vesicles accumulated and the parallel fiber bouton began to enlarge. During the same period, stubby spines emerged from the dendritic shaft with a parallel fiber attached to its tip. By 24 h newly-formed spines could not be distinguished from normal spine profiles.
Giant spines and enlarged synaptic contacts Enlargement of spine heads in the area where parallel fibers were degenerating occurred already by 15 h and reached a peak at about 4 days (Figs. 16 and 22). Each enlarged spine was accompanied by a marked elongation of synaptic profiles. The presynaptic bouton was enlarged and the total number of synaptic vesicles was increased. These modified synaptic contacts were still present at 2 months after the lesion was made. Each of these enlarged spines was characterized by a proliferation of endoplasmic reticulum in the head of the spinel Enlarged spines were most prominent in areas where parallel fiber shafts were greatly reduced in number and Purkinje cell dendrites were closely aligned in the neuropile.
Sprouting of Purkinje cell dendritic processes Active searching for parallel fibers by Purkinje cell spines was suggested by the appearance of elongated spine profiles and complex spine formations (Figs. 14 and 17). This was most frequently
degenerating bouton is dense. Note that the glia appear to be separating the degeneratingboutons (Fig. 8) from the postsynaptic site. The endoplasmicreticulum in the dual innervated spine is very elaborate. In Fig. 7 note the closeapposition of the two boutons indicating that the normal bouton may be extendingonto the postsynapticjunction to take over the postsynapticsite being vacated by the degenerating parallel fiber. (22 h post-lesion; bar equals 0.5/~m.) Fig. 10. Electron micrograph showingone slightlyenlarged spine with part of its postsynaptic membrane specializationopposed by a normal parallel fiber bouton and part of its postsynaptic membrane unoccupied (large arrow). Note one vacated spine (small arrow) and two phagosomesin a glial process. The apposition of glia with phagosome to a spine suggests that the degenerated bouton was recently removed from the spine and that the parallel fiber is taking over the postsynaptic site. (24 h post-lesion; bar equals 0.5 ,urn.)
Figs. 11 and 12. Electron micrograph of dual innervation of spines by two reactive boutons. In Fig. 11, note that the spine (S) contacts a parallel fiber bouton on the left and one on the lower right. In Fig. 12, note that one parallel fiber bouton is very small and the other is comparatively larger. (22 h post-lesion; bar equals 0.5/*m.) Fig. 13. Electron micrograpb showing dual innervation of a Purkinje cell spine by a climbing fiber and a parallel fiber. The climbing fiber (CF) is characterized by its apposition to the longitudinal profile of the dendrite (D) and by large and small vesicles. The apposed parallel fiber on the lower edge of the spine profile has equal size vesicles. (15 h post-lesion; bar equals 0.5/,~m.)
211 observed in regions where parallel fibers had been reduced to about 25 ~ of their original number. Elongated multi-branch spines extended between glial processes and even other parallel fibers to reach available presynaptic elements.
Sprouting of parallel fibers In longitudinal sections of folia, small protuberances of parallel fibers that contacted Purkinje cell spines increased in frequency (Fig. 21). These side branches were 1 or 2 # m in length and never contained tubule cores. Parallel fiber sprouting appeared less often than elongation of Purkinje cell spines. At later stages it appeared that the cut ends of parallel fibers grew out and attempted to circumvent the incision site. This was seen as a dense packing of parallel fibers near the incision where Purkinje cells were lacking as the result of the trauma. Unsectioned parallel fibers may have grown from their tips into the region of parallel fiber degeneration but this could not be determined by these methods. In addition, remaining parallel fibers presumably increased the number of boutons along their course to compensate the loss (unpublished results).
Aberrant synapses Synapses were frequently observed between Purkinje cell spines and stellate or basket cell axons within one day;these synapses are infrequently seen in the normal cerebellum14, 3s. The postsynaptic junction was opposed to basket axons, but there was a lack of vesicle accumulation or presynaptic thickenings (Figs. 23 and 24).
Later changes Five to 30 days after surgery, aberrant synapses
and vacated synaptic sites were rarely seen. However, giant spines with enlarged parallel fiber varicosities prevailed throughout the neuropile in which the parallel fibers were greatly reduced in number (Fig. 22). These enlarged boutons capped the Purkinje spine in a characteristic dome shape so that the entire head of the spine was surrounded by the bouton. The synaptic contact sometimes appeared in multiple profiles on a single spine; however, reconstruction of these boutons revealed that the synapse was very irregular in shape. Some enlarged boutons were contacted by numbers of spine heads. The heads of these spines usually depressed the surface of the large bouton giving the appearance of a scalped rosette (Fig. 25) similar to that described by Sotelo 53 in the cerebellum of the X-irradiated rabbit. Characteristically, these boutons were bounded by other parallel fibers. Other bouton profiles that were contacted by numerous heads of spines were completely shrouded by glia (Fig. 26). These boutons had dense core vesicles and were classified as climbing fibers. DISCUSSION Rapid degeneration of axons was previously described at 10-12 h following lesions of parallel fibers 4 and of afferent fibers to the interpeduncle nucleus ~9. Climbing fibers in the cerebellum also degenerated rapidly after transection of axons or chemical destruction of somas11,54, 55. Our findings revealed not only a rapid onset for degeneration of parallel fibers, but the complete removal of some of these afferents resulting in vacated postsynaptic sites on spines within 6 h of the lesion. Nevertheless, the number of degenerating boutons remained at about 3 ~ during this 2-3-day
Fig. 14. Electron micrograph of complex spine (CS) and a primitive synapse on a normal spine (S). A dumbbell-shaped spine (CS) is in synaptic contact with two parallel fibers. Another spine (S) is forming a new synap'ic contact (arrow) with a parallel fiber. Note that the parallel fiber bouton is not yet enlarged and vesicles have not significantly accumulated. Also note that the junctional complex is very small and is on the side of the spine head. (15 h post-lesion; bar equals 0.5/*m.) Fig. 15. Electron micrograph showing new synaptic contacts on the spine and on a Purkinje cell dendrite. Arrows point to small attachment sites which represent early synapse formation. Note that synaptic vesicles are beginning to accumulate. Two of these contacts are apparently on one existing stubby spine and a third is made directly with a dendrite from which an additional spine will presumably emerge. (20 h post-lesion; bar equals 0.50/~m.) Fig. 16. Electron micrograph showing formation of new synapse on a developing spine and dual synapses on an enlarged Purkinje cell spine. Note that the short stubby spine protrudes from dendrite without a distinct neck and is contacted by a parallel fiber on a very small postsynaptic site. Only a very few vesicles have accumulated. In the upper righthand corner, an enlarged spine (S) with elongated synaptic contact is continuous between the two boutons of reactive parallel fibers. (22 h post-lesion; bar equals 0.5 !tm.)
212
Fig. 17. Low power electron micrograph of an elaborated Purkinje cell dendrite (D) showing two elongated spines (S) and coated vesicles in the cytoplasm. Many coated vesicles (single arrows) are either attached to the plasma membrane or floating free in the cytoplasm. Note a double membrane-coated vesicle (double arrow) which arose by inclusion of a piece of membrane from another process. (23 h post-lesion; bar equals 0.5/~m.)
213 period. This may be due to assorted lengths of severed segments arising from their T-shaped interdigitating arrangement. For example, degeneration may occur very early in short segments and phagocytosis of these may be completed even before longer segments were beginning to show degenerative changes. In the cat, parallel fibers are much longer and can be seen to degenerate as late as 5 days following granule cell lesions 4. The distal portion of longer segments may survive for an additional length of time, being supported by continued axonal transport of substances needed to maintain the integrity of the axonal segment and ultimately synaptic contact 16. This suggests that the length of the distal segment may be a factor in the rate of degeneration in addition to caliber.
Phagocytosis vs vacating of postsynaptic sites Degeneration of parallel fibers following lesions was thought to result in removal of degenerated boutons together with opposed spines from dendrites TM. A number of recent studies from various regions of the brain have shown that postsynaptic sites remained on the dendrite and may even be reutilized2,18,ag,46,sL We did not observe phagocytosis of Purkinje cell spines and their postsynaptic junctions together with parallel fiber boutons. Postsynaptic sites remained on dendritic spines following removal of the degenerating boutons by glial processes.
Reinnervation of Purkinje cell spines: reoccupation of synaptic sites vs formation of new synapses Vacating of postsynaptic sites following degeneration of afferents has been described in a number of regions in the central nervous system2,9,36,39,40,45, 51, 58. However, whether or not these postsynaptic sites
are re-utilized as original units, has not been dearly defined. Raisman and his co-workers45, 46 suggested that postsynaptic sites are permanent and are reoccupied by sprouting afferents. Investigations on mutant mice 25,33,44,52 and induced agranular cerebellalS,~4, 27 suggested that synaptic sites are permanent since vacated postsynaptic sites on spines remained throughout life. This apparent permanence of postsynaptic sites suggests that afferents in the area could easily re-occupy these specializations. One of the first demonstrations of re-occupation of postsynaptic membrane was provided by Westrum 5s. He found evidence that boutons having fiat vesicles (symmetrical synaptic sites) could position themselves over sites previously occupied by boutons with spherical vesicles (asymmetrical type of junction). Evidence is lacking that these contacts are more than transient appositions or have a functional role. In studies of the superior cervical ganglion, vacated sites remained for long periods of time after degeneration of afferents and were believed to be reinnervated by afferents which were allowed to regrow into the cell mass 4°. This study revealed a reciprocal relationship between the number of vacated sites and number of intact synapses throughout the reinnervation period. More recently, Smollen 51 found that the number of vacated sites remained constant throughout development following early denervation of the ganglion, although approximately 50 ~ of the normal number of afferents were attained by adulthood. He 51 concluded that contacts were made on entirely new synaptic sites. Our studies indicate that some vacated synaptic sites may be re-occupied by original or other types of afferents (e.g. basket cell axons), at least for a
Figs. 18-20. Detailed views of coated vesicles. In Fig. 18, 4 coated vesicles are opposed to the dendritic membrane; note that each vesicle marks a site overlying a parallel fiber. In Fig. 19, note coated vesicle is attached to a membrane of a spine and is adjacent to a postsynaptic site. In Fig 20, note the coated vesicle is encapsulating a portion of parallel fiber plasma membrane. Double membrane-coated vesicle can be seen suspended in the cytoplasm of Purkinje cell dendrite. (20-23 h post-lesion; bar equals 0.25/~m). Fig. 21. Electron micrograph showing sprouting of a parallel fiber giving rise to a small process. The tubules of the parallel fiber continue to the left while the sprouting process (on the right) extends into a large bouton synapse on a Purkinje cell spine. (9 h post-lesion; bar equals 0.5/am.) Fig. 22. Electron micrograph showing a giant spine(s) with an elongated synaptic contact. Note that there is a marked increase in the endoplasmic reticulum within the spine head and that no tubules or mitochondria are present. The presynaptic terminal is also proportionally enlarged and filled with abundant round synaptic vesicles characteristic of parallel fibers. (5 days post-lesion; bar equals 0.5 ,urn.)
214
Fig. 23. Electron micrograph ofa stellate or basket cell axon (B) making contact with a dendritic shaft (D) and with a long spine(s). Note the pleomorphic vesicles at the junction with the dendritic shaft (D) and a lack of synaptic vesicles near the spine contact (S) (arrow) which has a distinct postsynaptic thickening. (I 5 h post-lesion; bar equals 0.5 !~m.)
215 transient period. However, this reinnervation is preceded by new synapse formation. These newlyformed synapses developed on original spines as well as by formation of new spines. Dual innervation of spines in the normal adult cerebellum is uncommon while partial denervation of Purkinje cell spines resulted in 0.8-1.2 complexes for every 1000/~m 2 of section (i.e. 7-27 ~ of the degenerated bouton observed). During early periods, degenerating boutons were present on the same spine that had a normal bouton (Fig. 27). Later, there was a shift toward single boutons on spines as revealed by reconstruction of spines which contacted more than one bouton profile. Some of these elongated boutons had incomplete capping of spines resulting in the occasional appearance of multiple bouton profiles on spines at 5 and 7 days. Reconstruction of serial sections revealed that these multiple structures were single parallel fiber contacts with irregular and elongated synaptic contact area. This dual innervation of spines found here may have similarities to multiple innervations found during development in climbing fiber 1°, muscle 47 or submandibular ganglion 26 where a one-to-one innervation emerges following multiple innervations. A selective5 or competitive elimination 43 process for stabilization of synapses was suggested in the transformation of multiple synapses on individual spines to single contacts. This may also be reflected in our study since a single synapse on a spine is the stable synaptic relationship in the cerebellum. Four possible processes of reactive re-organization that give rise this stabilization following reduction of afferents by degeneration are illustrated in Fig. 27. An enlarged single contact on a spine was a prominent occurrence after 5 days following lesions. This was achieved through bouton elimination and the extension of postsynaptic sites so that the vacated postsynaptic site merged with the newly formed site (Fig. 27Ca ana 4). Also postsynaptic sites and their
spines must be lost since the total number of synapses is reduced while the number of Purkinje cells remained constant. Vacated sites disappear after 5-7 days. Thus molecules of postsynaptic sites are reabsorbed and either decomposed or possibly reutilized in the enlargement ofpostsynaptic sites (Fig. 27C1 and 2)-
Giant spines and elongation of synaptic profiles Enlargement of spines with multiple innervations was previously described in the reeler mutants 33. Giant spines appear following malnutrition in the developing rat 6. This study shows that spines as well as synapses enlarge in the adult cerebellum when afferents are reduced. Remodeling of synapses in the molecular layer following lesions also resulted in enlargement of spines and an associated increase in the size of contact areas. The amount of this enlargement was related to the reduction in number of parallel fibers. Even though remaining parallel fibers compensated by forming new synapses on spines and presumably were able to take over some vacated synaptic sites, the compensation was limited by the maximum capability of the granule cell axons to respond to a net deficit in the number of synapses on Purkinje cells Thus numerous spines and their synapses were lost. Accompanying the loss of contact areas there was an increase in the length of the synaptic profiles. We have demonstrated in a malnutrition model, that average synaptic contact areas increased inversely with a deficit in the number of synapses. This increase in individual contact size was shown to be due to a constancy of total contact area for spine synapses on each Purkinje cell2L Reductions in the number of parallel fibers by lesioning the adult cerebellum also supports the finding of constancy in total contact area. Preliminary measurements show that the area of synaptic contact attains values which are expected for reductions in the number of synapses 23.
Fig. 24. Electron micrograph of a basket axon contacting a spine (S) having a postsynaptic membrane specialization. There are no vesicles opposed to this junction (arrow). (28 h post-lesion; bar equals 0.5 ,urn.) Fig. 25. Electron micrograph of a reactive parallel fiber terminal synapsing with 6 Purkinje cell spines (S). Note, one spine is completely encircled within the terminal which contains uniform synaptic vesicles. (28 days post-lesion; bar equals 0.5 ~m.) Fig. 26. Electron micrograph of a reactive climbing fiber varicosity forming multiple synaptic contacts with 5 moderately enlarged Purkinje cell spines (S). The bouton has densely packed large and small vesiclescharacterized by a climbing fiber bouton. (24 h postlesion; bar equals 0.5/zm.)
REACTIVE-REORGANIZATION OF SYNAPSES
A NORMAL
C SYNAPTIREORGANI C ZATION D STABILIZATION ABSORP|TO I NI~ :~'~ 1 I ~
BDEGENERATION 3
~rENst,.,.
Fig. 27. Diagram showing rapid synaptic reorganization after partial deafferentation of Purkinje cell spine synapses. Sectioning of parallel fibers in adult cerebella results in rapid degeneration in distal segments. The first change is seen in mitoehondria. This is followed by an active phagocytosis of dense debris by glia processes. Parallel fibers which remain in the area reinnervate some of these spines by forming new contact sites and by reoccupying vacated postsynaptic sites while other spines are vacated and lost. A: normal bouton-spine relationship in the neuropile showing synapse on the tip of a spine which are nearly completely capped by a glial sheath (stippled). B: a degenerating bouton on spine and two degenerating parallel fibers (dotted). Glia react to assuming a phagocytotic role retract from the neck of the spine and allow parallel fibers to come in direct opposition to the spine. C: synaptic reorganization is a transient phase where trial contacts may occur (e.g. aberrant synapses and dual innervations). This reorganization gives rise to stable connections through selective or competitive elimination. Shown are 4 processes of synaptic reorganization which are supported by our morphological findings. In the first type, a presynaptic bouton was removed by glial phagocytosis and a vacated postsynaptic site came in contact with glia. This site could either be reinnervated by available parallel fibers or else be absorbed with retraction of the spine. A second type represents reinnervation of spines by a new synapse and occurs even before the degenerating bouton vacates the postsynaptic site. This indicates that the onset of reactive synaptogenesis ensues as soon as degeneration starts or possibly functional synaptic activity ceases. This synaptic site could also be absorbed and the new synapse would remain. In the third type, the new synapse takes over the vacated postsynaptic site by extension of the bouton and its contact site so that new and old postsynaptic sites merge. In type 4, dual innervation by 2 reactive parallel fibers results in formation of one new synaptic site and the takeover of vacated synaptic site by another fiber. Also new sites can be formed by two new boutons. D: stabilization phase represents the loss of some uninnervated spines and postsynaptic sites. A competitive elimination of dual synapses on spines results in one-to-one bouton-spine synapses and allows compensation for the loss of synapses by increasing synaptic sizes. These giant spines with single enlarged synapses seem to be the resulting stable and possibly more efficient synapses.
217 The mechanism for the enlargement of synapses, that is suggested in this study, is the takeover of vacated sites by newly-formed synapses as well as synapse elimination. The newly-formed boutons, presumably can expand and merge with vacated sites. Likewise, dual-innervated spines that develop during the reaction period relinquish one bouton. Such vacated sites could also be incorporated into a single giant site. This suggests that the macromolecules making up the postsynaptic membrane thickenings are mobile and can move from vacated sites to new synapes or else there is a continual turnover a9 of macromolecules possibly form a pool that allows synapses to develop where they are demanded. Mobility of attachment and receptor molecules and their shifting to adjacent sites would have to occur if the components of the membrane densities are actually permanent rather than their location. The question of permanency of these synaptic attachment molecules versus their continual turnover with ability to be inserted at new active sites remains unanswered. Nevertheless, the ability of new synapses to form in a very short time period, even before synaptic sites are vacated, strongly supports a turnover mechanism. The apparent permanence of postsynaptic sites may actually have an underlying continuous rapid turnover of macromolecules related to postsynaptic membrane thickenings. This turnover may allow deposition to occur at more favorable sites. Alternatively, there is a reservoir of macromolecules that responds to the deafferentation. Coated vesicles were first observed in degeneating dendrites 57 and were thought to transport and insert macromolecules at synaptic attachment sitesL Attachment molecules and receptor molecules may likely be separable entities.
Formation of new synapses and new spines Developmental synaptogenesis follows a progressive course where primitive contacts form and then synaptic vesicles accumulate resulting in bouton and spine development 37. We observed a similar process in the adult cerebellum as reactive synaptogenesis following the removal of parallel fiber afferents. The loss of afferents to Purkinje cell spines appeared to produce an immediate increase in the insertion of attachment molecules at new sites. A very interesting observation was that Purkinje
cell dendrites actually appeared to pinch off a portion of the parallel fiber membrane (Fig. 20). This could indicate a 'tasting' mechanism whereby one process recognized the other. Coated vesicles have been attributed to membrane recycling12,20 through a basket retrieval mechanism 19. Others suggested that coated vesicles served a role in insertion of postsynaptic membrane attachment sitesl,4s, 56. Privat a2 proposed that coated vesicles were part of a mechanism for removing the attachment site. Recent studies showed a marked association of these coated vesicles with axonal growth during development 12 as well as in reactive sproutingZ~,32,sL A major question remains as to the direction that these specialized vesicles move or whether they represent two simultaneous membrane displacing processes; one inserting molecules and the other shifting membrane as well as 'sampling' the environment. Glia appeared to be a major factor in determining where new synapses were formed on dendrites. New synaptic contacts were found only in those regions where parallel fibers were in direct contact with spiny branchlets. Glial involvement could explain why there was an irregular shift of parallel fiber-synaptic contacts down main dendritic trees of Purkinje cells following climbing fiber deafferentation54,55 (Hillman, unpublished observations). Bergmann astrocytic processes relinquished their position on main dendritic trees following climbing fiber deafferentation, may be due to their active role in phagocytosis of degenerating remnants. This concept is supported by the occurrence of spines on the larger dendritic shafts where glial ensheathment of dendrites is lacking, such as in reptiles 21.
Sprouting of parallel fibers and dendrites Parallel fibers, on occasion, make small side projections. These were observed more frequently following deafferentation, thus indicating an early form of collateral sprouting. However, it was clear that these side processes did not contain tubules and thus apparently did not extend for more than a few microns. Axons that were in apposition to dendrites and spines appeared to be the major contributor to reactive synaptogenesis rather than growth of axon through sprouting. Some parallel fiber boutons were greatly enlarged and capped by multiple spine heads; others were
218 enlarged but smaller and capped one spine. These enlarged boutons may have been on proximal segments of parallel fibers that remained after a significant portion of their axon had been severed. Sprouting or regrowth from the ends of parallel fibers that were not severed may have occurred, but was not detectable by methods used here. However, the cut ends of parallel fibers grew and had a very compact arrangement as they attempted to circumvent the scar. Sprouting of spines was also evident as shown by multiple spine heads which extended for 3-4 ttm into the neuropile. Some synapses on these complex spines were on small protuberances near the neck of spines while others were individual spine-like processes that were based on a 'spine trunk'. Similar complex spines were described in the hippocampusa4,aL Reactive synaptogenesis in the cerebellar molecular layer: presynaptic vs postsynaptic control Reactive sprouting of axons has revealed a potential for reconnectivity in the adult peripheral and central nervous system that is not possible by regenerative growth. The message for instituting these outgrowths in axons is not understood. In the peripheral nervous system, evidence suggests that substances directing sprouting appears confined to nerve paths (Schwann cells and perineural sheaths) which connect denervated muscles to sheaths containing normal fibers 5°. Parallel fibers have a form of side sprouting in that small processes can emerge from fibers along their course. These parallel fiber sprouts, however, are greatly limited, presumably due to the fact that the T-shaped parallel fiber has reached a limit in its bifurcation potential and does not have sufficient numbers of tubules to support additional branching. In this study reactive synaptogenesis occurred independent of vacating of sites with the initiation of synapse formation following severing of afferent
REFERENCES 1 Altman, J., Coated vesicles and synaptogenesis. A developmental study in the cerebellar cortex of the rat, Brain Research, 30 (1971) 311-312. 2 Bernstein, J. J. and Bernstein, M. E., Neuronal alteration
fibers to Purkinje cells. The rapid occurrence of synaptogenesis indicates that the postsynaptic element, the Purkinje cell itself, contains the potential to form synapses with afferents which are in direct contact. Furthermore, postsynapticjunctional molecules were available to form new synaptic sites. This reactive synaptogenesis revealed a potential of these neurons to form multip!e synaptic contacts on individual spines. Our findings further support the contention that postsynaptic structures can reorganize to shift position and size of synaptic connections, possibly by returning to an immature-like state. The potential of Purkinje cells to make new connections appears to be related to the immediate availability of postsynaptic membrane structures which can be inserted at appositional zones between the dendrite and presynaptic elements. The signal appears to be related to a communication mechanism which is activated in the postsynaptic cell and allows receptor molecules and attachment molecules to concentrate at potential synaptic sites. A second stage of postsynaptic control over reactive synaptogenesis was the elimination of multiple synapses on spines and the ultimate incorporation of postsynaptic membranes into enlarged synaptic sites. The principle underlying compensatory enlargement through elimination, may be dictated by the postsynaptic neuron in its response to maintain a constant level of synaptic contact area2L In this hypothesis, the response of both 'appositional' and 'sprouting' reactive synaptogenesis are mediated through the availability of postsynaptic macromolecules to be inserted into active zones. These could be activated through cessation of transmitter release following severing of axons. ACKNOWLEDGEMENTS Research was supported by U.S. Public Health Service grants HD-10934 and NS-13742 from N I C H D and N I N C D S .
and reinnervation following axonal regeneration and sprouting in the mammalian spinal cord, Brain Behav. EvoL, 8 (1973) 135-161. 3 Bjorklund, A. and Stenevi, L., Regeneration of monoaminergic and cholinergic neurons in the mammalian central nervous system, Physiol. Rev., 59 (1979) 62-100.
219 4 Brand, S., Dahl, A. L. and Mugnaini, E., The length of parallel fibers in the cat eerebellar cortex. An experimental light and electron microscopic study, Exp. Brain Res., 26 (1976) 39-58. 5 Changeux, J. and Danchin, A., Selective stabilisation of developing synapses as a mechanism for the specification of neuronal networks, Nature (Lond.), 264 (1976) 705-712. 6 Chen, S. and Hillman, D E., Giant spines and enlarged synapses induced in Purkinje cells by malnutrition, Brain Research, 187 (1980) 487-493. 7 Colonnier, M., Experimental degeneration in the cerebral cortex, J. Anat. (Lond.), 98 (1964) 47-53. 8 Cotman, C. W., Neuronal Plasticity, Raven Press, New York, 1978. 9 Cotman, C. W. and Nadler, J. V., Reactive synaptogenesis in the hippocampus. In C. W. Cotman (Ed.), Neuronal Plasticity, Raven Press, New York, 1978, pp. 229-260. 10 Crepel, F., Mariani, J. and Dehhaye-Bouchaud, N., Evidence for a multiple innervation of Purkinje cells by climbing fibers in the immature rat cerebellum, J. Neurobiol., 7 (1976) 567-578. 11 Desclin, J. C. and Escubi, J., An additional silver impregnation method for demonstration of degenerating nerve cells and processes in the central nervous system, Brain Research, 93 (1975) 25-39. 12 Eckenhoff, M. F. and Pysh, J. J., Double-walled coated vesicle formation: evidence for massive and transient conjugate internalization of plasma membranes during cerebellar development, J. NeurocytoL, 8 (1979) 623-638. 13 Field, P. M., Synapse formation after injury in adult brain: failure of fimbrial axons to reinnervate the bed nucleus of the stria terminars, Brain Research, 189 (1980) 91-101. 14 Fox, C. A., Hillman, D. E., Siegesmund, K. A. and Dutta, C. R., The primate cerebellar cortex: a Golgi and electron microscopic study. In C. A. Fox and R. S. Snider (Eds.), Progress in Brain Research, VoL 25, Elsevier Press, Amsterdam/London/New York, 1967, pp. 174-225. 15 Gall, C. and Lynch, G., Rapid axon sprouting in the neonatal rat hippocampus, Brain Research, 153 (1978) 357-362. 16 Goldowitz, D. and Cotman, C. W., Do neurotrophic interaction control synapse formation in the adult rat brain? Brain Research, 181 (1980) 325-344. 17 Graziadei, P. P. C. and Gradiadei, G. A. M., The olfactory system: a model for the study of neurogenesis and axon regeneration in mammals. In C. W. Cotman (Ed.), Neuronal Plasticity, Raven Press, New York, 1978, pp. 131-153. 18 Herndon, R. M., Tiophen induced granule cell necrosis in rat cerebellum. An electron microscopic study, Exp. Brain Res., 6 (1968) 49-68. 19 Heuser, J. E., Three-dimensional visualization of coated vesicle formation in fibroblasts, J. Cell Biol., 84 (1980) 560-583. 20 Heuser, J. E. and Reese, T. S., Evidence for recycling of synaptic vesicle membrane during transmitter release at the frog neuromuscular junction, J. Cell Biol., 57 (1973) 315-344. 21 Hillman, D. E., Neuronal organization of the eerebellar cortex in amphibia and reptilia. In R. Llinas (Ed.), Neurobiology of Cerebellar, Evolution and Development, Am. Med. Assoc., Chicago, 1969, pp. 279-325.
22 Hillman, D. E. and Chen, S., Vulnerability of cerebellar development in malnutrition. II. Intrinsic determination of total synaptic areas on Purkinje cell spines, Neuroscience, 6: 1263-1275, 1981. 23 Hillman, D. E. and Chen, S., Plasticity of synaptic size with constancy of total synaptic contact area on Purkinje cells in the cerebellum. In E. A. Vidrio and M. A. Galina (Eds.), Symposia Proceedings of the XI International Congress of Anatomy, Alan Liss, New York, pp. 229-245. 24 Hirano, A., Dembitzer, H. M. and Jones, M., An electron microscopic study of cyasin-induced cerebellar alteration, J. Neuropath. Exp. Neurol., 31 (1972) 113-125. 25 Landis, D. M. and Sidman, R., Electron microscopic analysis of postnatal histogenesis in the cerebellar cortex of staggerer mutant mice, J. comp. Neurol., 179: 831-864, 26 Lichtman, J. W., The reorganization of synaptic connexions in the rat submandibular ganglion during postnatal development, J. Physiol. (Lond.), 273 (1977) 155-177. 27 Llinas, R., Hillman, D. E. and Precht, W., Neuronal circuit reorganization in mammalian agranular eerebellar cortex, J. Neurobiol., 4 (1973) 69-94. 28 Lund, R. D., Development and Plasticity of the Brain, Oxford University Press, New York, 1978. 29 Lund, R. D. and Lund, J. S., Synaptic adjustment after deafferentation of the superior colliculus of the rat, Science, 171 (1971) 804-807. 30 Lynch, G., Rose, G., Gall, C. and Cotman, C. W., The response of the dentate gyrus to partial deafferentation. In M. Santini (Ed.), Golgi Centennial Symposium Proceedings Raven Press, New York, 1975, pp. 505-517. 31 McWilliams, R. and Lynch, G., Terminal proliferation and synaptogenesis following partial deafferentation: The reinnervation of the inner molecular layer of the dentate gyrus following removal of its commissural afferents, J. comp. Neurol., 180 (1978) 581-616. 32 McWilliams, J. R. and Lynch, G., Sprouting in the hippocampus is accompanied by an increase in coated vesicles, Brain Research, 211 (1981) 158-164. 33 Mariani, J., Crepel, F., Mikoshiba, K., Changeux, J. P. and Sotelo, C., Anatomical, physiological and biochemical studies of the cerebellum from reeler mutant mouse, Physiol. Trans. roy Soc. (Lond.), 281 (1977) 1-28. 34 Matthews, D. A., Cotman, C. W. and Lynch, G. S., A quantitative ultrastructural analysis of synaptic changes in the rat dentate gyrus following entorhinal lesions. I. Magnitude and time course of degeneration, Brain Research, 115 (1976) 1-21. 35 Matthews, D. A., Cotman, C. and Lynch, G., A quantitative ultrastructural analysis of synaptic changes in the rat dentate gyrus following entorhinal lesions. II. Reacquisition of morphologically normal synaptic contacts, Brain Research, 115 (1976) 23-41. 36 Mouren-Mathieu, A. M. and Colonnier, M., The molecular layer of the adult cat eerebellar cortex after lesions of the parallel fiber: an optic and electron microscopic study, Brain Research, 16 (1969) 307-323. 37 Mugnaini, E., Ultrastructural studies on the cerebellar histogenesis. II. Maturation of nerve cell populations and establishment of synaptic connections in the cerebellar cortex of the chick. In R. Llinas (Ed.), Neurobiology of Cerebellar Evolution and Development, Amer. Med. Assn., Chicago, 1969, pp. 749-782. 38 Mugnaini, E., The relation between cytogenesis and the formation of different types of synaptic contact, Brain
220
Research, 17 (I 970) 169-179. 39 Murray, M., Zimmer, J. and Raisman, G., Quantitative electron microscopic evidence for reinnervation in the adult rat interpeduncular nucleus after lesions of the fasciculus retroflexus, J. comp. Neurol., 187 (1979) 447-468. 40 Ostberg, A.-J. C., Raisman, G., Field, P. M., Iversen, L. L. and Zigmond, R. E., A quantitative comparison of the formation of synapses in the rat superior cervical sympathetic ganglion by its own and by foreign nerve fibers, Brain Research, 107 (1976) 445-470. 41 Pettigrew, J. D., The paradox of the critical period for striate cortex. In C. W. Cotman (Ed.), NeuronalPlasticity, Raven Press, New York, 1978, pp. 311-329. 42 Privat, A., A possible mechanism for the resorption of attachment plates in the growing rat brain, Brain Research, 69 (1974) 125-129. 43 Purves, D. and Lichtman, J. W., Elimination of synapses in the developing nervous system, Science, 210 (1980) 153-157. 44 Rakic, P., Synaptic specificity in the cerebellar cortex: study of anomalous circuits induced by single gene mutations in mice, CoM Spr. Harb. Symp. quant. BioL, 40 (1976) 333-346. 45 Raisman, G. and Field, P. M., A quantitative investigation of the development of collateral reinnervation after partial deafferentation of the septal nuclei, Brain Research, 50 (1973) 241-264. 46 Raisman, G., Field, P. M., Ostberg, A.-J. C., Iversen, L. L. and Zigmond, R. E., A quantitative ultrastructural and biochemical analysis of the process of reinnervation of the superior cervical ganglion in the adult rat, Brain Research, 71 (1974) 1-16. 47 Redfern, P. A., Neuromuscular transmission in newborn rats, J. Physiol. (Lond.), 209 (1970) 701-709. 48 Reese, R. P., Bunge, M. B. and Bunge, R. P., Morphological changes in the neuritic growth cone and target neuron during synaptic junction development in culture, J. cell Biol., 68 (1976) 240-263. 49 Ruffalo, R., Eisenbarth, G. S., Thompson, J. M. and
50
51
52
53
54
55
56
57
58
59
Nirenberg, M., Synapse turnover: a mechanism for acquiring synaptic specificity, Proc. nat. Acad. Sci. U.S.A., 75 (1978) 2281-2285. Slack, J. R., Hopkins, W. G., Williams, M. N., Nerve sheaths and motoneurone collateral sprouting, Natur6 (Lond.), 282 (1979) 506-507. Smolen, A. J., Postnatal development of ganglionic neurons in the absence of preganglionic input: morphological observations in synapse formation, Develop. Brain Res., 1 (1981) 49-58. Sotelo, C., Anatomical, physiological and biochemical studies of the cerebellum from mutant mice. II. Morphological study of cerebellar cortical neurons and circuits in the Weaver mouse, Brain Research, 94 (1975) 19-44. Sotelo, C., Synaptic remodeling in mutants and experimental animals. In F. Vital-Durand (Ed.), Aspects of neural plasticity, Vol. 43 pp. 167-190 and Jeannerod, M. INSERM 11-12, 1975. Sotelo, C. and Arsenio-Nunes, M. L., Development of Purkinje cell in absence of climbing fibers, Brain Research, 111 (1976) 389-395. Sotelo, C., Hillman, D. E., Zamora, A. J. and Llinas, R., Climbing fiber deafferentation : Its action on Purkinje cell dendritic spines, Brain Research, 98 (1975) 574-58l. Stelzner, D. J., Martin, A. H. and Scott, G. L., Early stages of synaptogenesis in the cervical spinal cord of the chick embryo, Z. Zellforseh., 138 (1973) 475-488. Walberg, F., Role of normal dendrite in removal of degenerating terminal boutons, Exp. NeuroL, 8 : 112-124, 1963. Westrum, L. E., Electron microscopy of degeneration in the lateral olfactory tract and plexiform layer of the prepyriform cortex of the rat, Z. Zellforsch., 98 (1969) 157 187. Westrum, L. E. and Black, R. G., Fine structural aspects of the synaptic organization of the spinal trigeminal nucleus (pars interpolaris) of the cat. Brain Research, 25 (1971) 265-287.