Development of synaptic junctions in cerebellar glomeruli

23 1

Developmental Brain Research, 8 (1983) 231-245 Elsevier

Development

o f S y n a p t i c J u n c t i o n s in C e r e b e l l a r G l o m e r u l i

DENNIS M. D. LANDIS, LORI A. WEINSTEIN and JOHN J. HALPERIN

Department of Neurology, Massachusetts General Hospital Boston, MA 02114 (U.S.A.) (Accepted November 23rd, 1982)

Key words: synaptogenesis - synapse - cerebellar development - freeze-fracture - membrane structure

In the glomerular synapses of developing mouse cerebellar cortex, 2 components of the synaptic junctions assemble independently in the immature granule cell dendrites, and then combine. 'Initial' junctions between mossy fiber axons and immature granule cell dendrites have presynaptic and postsynaptic electron-dense fuzz and a widened synaptic cleft, but lack the aggregate of intramembrane particles associated with the extracellular half of the postsynaptic membrane which characterizes mature synaptic junctions. In the vicinity of'initial junctions' there are particle aggregates which resemble those at mature synaptic junctions, but which are less densely packed and which are not associated with the other features of a junction. The constituent particles of these aggregates move to the sites of 'initial junctions' to combine with them and form 'immature synaptic junctions'. Many of these immature junctions are larger in area than mature synaptic junctions. The immature junctions accumulate a fairly uniform complement of intramembrane particles, which increase in packing density as the junctions decrease in area and attain smaller, adult size. INTRODUCTION

In an effort to learn more about the manner in which the various components of synaptic junctions are assembled, we have examined with thin-section and freeze-fracture techniques the synaptic junctions between mossy fiber axons and granule cell dendrites in the glomerular synapses of developing mouse cerebellar cortex. These synaptic junctions are like those of many excitatory synapses in the mammalian central nervous system in that the electron-dense fuzz lining the postsynaptic membrane is more prominent than that lining the presynaptic membrane (asymmetric, or Gray's type I synaptic junctions~°), while in freeze-fractured tissue the postsynaptic membrane has an aggregate of intramembrane particles associated with the extracellular half of the fractured membrane 12.18. 20.39. The glomerular synapses are easily identified as they mature over the first two postnatal weeks. There are so many junctions between the mossy fiber axon and its surrounding thicket of granule cell dendrites that one can reasonably expect to detect even fairly short-lived intermediate forms. 0165-3806/83/$03.00 © 1983 Elsevier Science Publishers B.V.

The organization of cerebellar glomeruli and their postnatal development have been thoroughly studied. In adult animals, glomerular synapses in the granular layer consist of an enpassage or terminal varicosity of a mossy fiber axon which is linked to surrounding granule cell dendrites by many small synaptic junctions 6.7. 11.29.30.One or more Golgi II cell axons also form synapses with the same population of granule cell dendrites ~3. These synapses are inhibitory, and the junctions have symmetric, Gray's type II electron-dense membrane specializations ~° which can be readily distinguished from those formed by the mossy fiber axon. The whole glomerulus is invested by an astrocytic sheath. During the first 2 weeks of postnatal development, granule neurons are generated in the external granular layer 25 and migrate across the nascent molecular layer to take up their final positions in the internal granular l a y e I a6-27-34. The first cells to arrive in the internal granular layer abut the white matter, while cells arriving later take up more superficial positions nearer to the Purkinje cell layer. Over this 2-week interval there is an obvious gradient in the maturity of the glomerular synapses with depth in the granular layer:

232 deeper glomeruli, among older granule cells, tire more adult in their configurations than glomeruli among the superficial, recently arrived granule cell neurons. In a single thin section through the granular layer, one can thus examine glomerular synapses at several stages of development. As glomeruli mature, the central axon varicosity becomes larger in volume. Immature granule cell dendrites around primitive glomeruli are larger in diameter than in mature glomeruli; during development the dendrites appear to divide into numerous terminal processes 3s. termed 'digits', each of which establishes synaptic junctions ~.-'2,~`~. A single dendrite thus forms many synaptic junctions in a given glomerular complex. Granule cell neurons send dendrites to several different glomeruli. At present, it is not known whether glomeruli deep in the granular layer in the second postnatal week continue to acquire dendritic input from newly arriving cells, or whether the axon simply acquires more synaptic junctions with dendrites arriving from the immediately surrounding population of granule cells. Previous studies have found that the junctions established between the mossy fiber axons and granule cell dendrites are relatively large in developing glomeruli, but in adult animals the junctions are uniformly quite smalF-2-2~'.-'7. This difference in the size of synaptic junctions between developing and mature animals raises the question of whether the initial, large junctions are replaced by a second set of mature junctions, or whether the initial junctions themselves remodel to the mature configuration. We present here observations made with thin-section and freeze-fracture methods which indicate that 2 discrete components of the synaptic junctions appear in the postsynaptic membrane independently, then become co-extensive, and finally reorganize as the junction becomes smaller in extent. MATERIALS AND METHODS

Adult and developing (8, 9, 10 and 12 days postnatal - the day of birth is considered day 0) C57BL/6J mice were anesthetized with intrape-

ritoneal chloral hydrate and pcrfused Ii.,r t5 !nin with a fixative containing 1.25% glutaratdehyde and 1% paraformaldehyde in 0.1 M sodium cacodylate buffer at pH 7.35 and 37'T'. The mice were left undisturbed ibr I h. Then the cerebellum was removed and immersed in the same fixative. For thin section studies, the cerebellum remained in fixative for 3 14 h, and then was sliced in the sagittal plane into 400 ~tm slabs with a Sorvall TC-I tissue chopper. The tissue was rinsed and postfixed in 1% osmium tetroxide in 0.1 M sodium cacodylate buffer for 1 h at 22 "C. Tissue was stained en bloc with uranvl acetate and embedded in Epon. For freeze-fracture studies, the tissue remained in fixative for less than 3 h at 22 ~'C, and then was sliced into 200/tm slabs in the sagittal plane with a Sorvall TC-I tissue chopper, and immersed in 25% glycerol in 0.05 M sodium cacodylate buffer. The tissue was mounted on Balzers gold specimen supports for complementar}, replication, and frozen by immersion in amelting monochlorodifluoromethane ( - 143 "C). Complementary replication was performed in a Balzers 301 freeze-fracture apparatus fitted with electron beam guns and a quartz crystal thin film monitor. For quantitative analysis ofsynaptic size, serial thin sections of cerebellar cortex were obtained from animals at postnatal day 9, at postnatal day 12, and from adults. Sections were cut with similar interference color (silver to gray): Care was taken to align the blocks so that the sections were parallel to the vertical axis of Purkinje cells. The sections included the entire depth of the internal granular layer, from the white matter to the Purkinje cell layer. All mossy fiber granule cell synapticjunctions wholly contained within the section series were photographed. The length of each junction was determined in each of the sections which included it, and these lengths were summed and then multiplied by 60 nm (an arbitrary estimate of section thickness) to obtain a measure of the total area of each synaptic junction. Statistical comparisons were performed using Student's two-tailed t-test.

233 OBSERVATIONS

Glomerular synapses with a central mossy fiber varicosity and surrounding granule cell dendrites are present in developing cerebellar cortex at postnatal day 8 (the earliest age studied). Through postnatal day 12, many of the granule cell dendrites surrounding glomeruli were larger in diameter than in adult animals (compare Figs. 1 and 2); this phase of dendritic maturation probably corresponds to the 'cup-shaped' dendrites described by Larramendi :2. We re-examined the morphology of mature synaptic junctions prior to undertaking an analysis of the developing junctions. Mature synaptic junctions between granule cell dendrites and mossy fiber axons have subtle electron-dense fuzz lining the intracellular surface of the presynaptic membrane, sometimes associated with short, dense projections into the axoplasm (Figs. 3 and 4). There is a widened synaptic cleft containing electron-dense material with laminar substructure. Prominent electron-dense fuzz lines the intracellular face of the postsynaptic membrane, co-extensive with the synaptic cleft. In freeze-fractured preparations the presynaptic membrane has a sparse population of large, 1012 nm particles associated with the cytoplasmic half of the membrane. The postsynaptic membrane contains an aggregate of particles associated with the extracellular half of the fractured membrane; the aggregate is co-extensive with the indentation of the dendritic contour caused by the widened synaptic cleft (Figs. % 11). The particles constituting the aggregate do not have ordered packing, and are not uniform in size or shape. Over the first 2 postnatal weeks, the morphological features of synaptic junctions between mossy fiber axons and granule cell dendrites as visualized in thin-sectioned tissue are similar to those of adult junctions, except for the area of the junctions. There is electron-dense material lining the presynaptic membrane, a widened synaptic cleft with intermediate electron-dense material, and prominent electron-dense fuzz lining the postsynaptic membrane (Figs. 5 and 6). In single thin sections, it was occasionally diffi-

cult to ascertain whether a particular postsynaptic density indeed was apposed to presynaptic specializations, so we examined junctional specializations in serial sections. In all the junctions studied, presynaptic and postsynaptic specializations were present; we did not detect electrondense specializations of the dendritic membrane which were not apposed to an axonal specialization. There is a broad range in the size of individual synaptic junctions between mossy fibers and granule cell dendrites. To obtain an accurate estimate of junction size, we selected synaptic junctions in which the apposed plasma membranes were in cross-section (rather than obliquely sectioned) and which were wholly within the serial sections. An estimate of the area ofsynapticjunctions was obtained by measuring the length of the junction in all of the sections in which it appeared, summing the measurements, and multiplying by 60 n m (an approximation of section thickness). As indicated in Fig. 15, at 9 and 12 days, 19% and 32% (respectively) of the junctions present were larger than any of those in adult animals. The mean area of synaptic junctions was greater at 12 days than 9 days, and both 9- and 12-day animals had a mean synaptic junction size greater than adults (for all comparisons, P ( 0.001, 2-tailed t-test). Having determined the area of junction in serial sections, we attempted to define some other morphological feature of the large synaptic junctions at 9 and 12 days which set them apart from the smaller junctions, or from synaptic junctions in adults. However, the electron-density of the presynaptic fuzz (including dense projections), the cleft material, and the postsynaptic fuzz is similar in large and small junctions in developing animals, and these characteristics are also similar to those of mature synaptic junctions. In contrast, junctions between mossy fiber axons and granule cell dendrites were heterogeneous in developing glomeruli when examined in freeze-fractured preparations. One population of junctions (which we term 'initial junctions') were evident as deformations in the apposed axonal and dendritic membrane contours

235 caused by a widened intercellular cleft; when cross-fractured, the widened intercellular space contained granular material, like that of a mature synaptic cleft. The overall shape of these junctions is usually circular or ovoid (see B in Figs. 7 and 8). In several instances, the crossfractured dendritic cytoplasm adjacent to these deformations was seen to have a subtle substructure which probably corresponds to electrondense fuzz as visualized in thin-sectioned preparations. These deformations of membrane contour were present only where granule cell dendrites were apposed to mossy fiber axons; they were not encountered in granule cell dendrites adjacent to other dendrites or to glia. We distinguish a second population of junctions (and termed these 'immature junctions') which share with initial junctions the feature of cleft material and indented membrane contours, but which manifest in addition an aggregate of particles associated with the extracellular half of the fractured dendritic membrane, co-extensive with the indentation into the dendritic contour (see C in Figs. 7 and 8). The area of dendritic membrane occupied by an individual immature junction is variable, and the number of associated particles is also variable; very large immature junctions tend to have few associated particles, those of intermediate size have more, up to a m a x i m u m of about 60 particles. In some smaller immature junctions, this approximate n u m b e r of particles was more densely packed. In addition to the initial junctions (without associated particles) and immature junctions (with a continuum of particle n u m b e r and particle packing), there are aggregates of particles asso-

ciated with the extracellular half of the fractured dendritic membrane, in a region of the dendrite apposed to a mossy fiber axon, but not co-extensive with a junctional cleft (see A in Figs. 7 and 8). The particles in these aggregates are not uniform in shape or size, but resemble the particles associated with immature junctions and those at mature synaptic junctions. At postnatal days 9 and 12, immature granule cell dendritic membranes have all 3 specializations in close proximity to one another. In adult animals, granule cell dendrites do not manifest indentations without associated particles (initial junctions) or aggregates of particles at non-junctional sites. We wanted to be certain that the several specializations of dendritic contour and particle distributions in fact represented synapticjunctions, and so examined with serial thin-section and freeze-fracture techniques 3 other classes of junctions in glomeruli: Golgi II cell axon-togranule cell dendrite synapses, puncta adhaerentia between granule cell dendrites, and a different sub-class of adhaerens-type junction between mossy fiber axons and granule cell dendrites which we term 'attachment plaques'. Synaptic junctions between Golgi II cell axons and granule cell dendrites in the periphery of adult glomeruli have indistinct electron-dense material lining the intracellular aspect of the presynaptic and postsynaptic membranes (Fig. 3). The junctional cleft is widened, and contains electron-dense material with a distinct substructure. We have not recognized in freeze-fractured preparations any specialization of intramembrane particle distribution in either presynaptic or postsynaptic membrane at these junctions.

Fig. 1. Developing glomerulus in the internal granular layer of the developing cerebellar cortex of a 9-day-old mouse. A central mossy fiber axon (MF) containing many synaptic vesicles is surrounded by several granule cell dendrites (G). At least 2 distinct types of junctional specializations have been formed between the axonal varicosity and the dendrites. There are several synaptic junctions (black arrows), characterized by electron-dense material associated with the postsynaptic membrane, a widened synaptic cleft with intermediate electron-dense material, and comparatively sparse electron-dense material associated with the presynaptic membrane. Another specialization, which we refer to as an 'attachment plaque' (black-on-white arrow), has electrondense fuzz symmetrically disposed on both the dendritic and axonal membranes. Synaptic vesicles do not cluster toward the axona! side of the attachment plaque. Granule cell dendrites are linked to one another by a third class of junction: puncta adhaerentia (*). These also have electron-dense material associated with both of the apposed membranes, but the junctional cleft material tends to be more ordered than at attachment plaques. (26,000 x .) Fig. 2. Glomerulus in the cerebellar cortex of a mature mouse. There are many synapticjunctions established between the mossy fiber axon (MF) and surrounding granule cell dendrites (G). Two Golgi II cell axon boutons (GC-1I) are in the periphery of the glomerulus, each establishing symmetric synaptic junctions with granule cell dendrites. ( 18,000 x .)

236

Fig. 3. Synaptic junctions in a mature cerebellar glomerulus. At the left there is a typical asyminetric synaptic junctitm established by the mossy fiber axon with a granule cell dendrite (G). The junction has electron-dense material associaled with the postsynaptic membrane, a widened synaptic cleft with intermediate electron-dense material, and relativel,~ subtle elecu~m-der!~e

237

Fig. 5. Synaptic junction and an attachment plaque in a glomerulus of a 9-day-old mouse. The synaptic junction on the left has subtle electron-dense material associated with the presynaptic membrane of the mossy fiber (MF), a widened synaptic cleft, and prominent electron-dense material in the postsynaptic membrane of the granule celt dendrite (G). The attachment plaque at the right, between vertical arrows, has electron-dense material associated with the membranes in both the dendrite and the axon. There is no clustering of vesicles near the attachment plaque specialization. (86,000 × .) Fig. 6. Synaptic junctions and attachment plaque in the glomerulus of a 9-day-old mouse. A long synaptic junction typical of developing glomeruli is evident at the left. An attachment plaque is present at the right, between the arrows. Immediately to the right of the attachment plaque there is a membrane specialization which resembles that ofa synapticjunction, with subtle presynaptic electron-dense material, more prominent postsynaptic electron-dense material, and a widened cleft. This is apparently a synaptic junction, but there is no clustering of synaptic vesicles toward the presynaptic component of the junction. In the serially sectioned preparations, we frequently encountered attachment plaques immediately adjacent to comparatively large synaptic junctions, such as this. Mossy fiber (M F), granule cell dendrite (G). (86,000 × .) material associated with the presynaptic membrane. In this instance there is definite clustering of synaptic vesicles toward the presynaptic component of the junction. In the right of the illustration, a Golgi-lI cell axon (GC-II), with small, pleomorphic synaptic vesicles, establishes a symmetric synaptic junction with a granule cell dendrite. This symmetric synaptic junction has relatively subtle electron-dense material associated with pre- and postsynaptic membranes, but a definitely widened synaptic cleft. (89,000 x ). Fig. 4. Attachment plaque, synaptic j unction, and puncta adhaerentia in a mature cerebellar glomerulus. At the left of the illustration, a mossy fiber (MF) has formed an attachment plaque with a granule cell dendrite (G). The attachment plaque has electron-dense fuzz symmetrically disposed beneath the junctional membranes, and a slightly widened cleft with intermediate dense material. At the top is a synaptic junction between the mossy fiber axon (MF) and a granule cell dendrite., with synaptic vesicles clustered toward the junction. The presynaptic electron-dense material is in the shape of 2 dense projections, and can be distinguished from the more homogeneous electron-dense material in the axonal component of the attachment plaque. At least 2 puncta adhaerentia (asterisks)join granule cell dendrites. These junctions appear similar to attachment plaques in that they have electron-dense material lining the membrane of both dendrites and a widened cleft, but their appearance in freeze-fractured preparations is quite different (see text). (89,000 x .)

f

i

~S~II~

i ~

! !

~

.i~'~~~

239 Granule cell dendrites in mature and developing glomeruli establish junctions with other granule cell dendrites, termed 'puncta adhaerentia'. In thin sections, these junctions have a

widened intercellular cleft containing amorphous electron-dense material and electrondense fuzz subjacent to both of the plasma membranes, co-extensive with the widened cleft (Fig. 4). In freeze-fractured preparations, these puncta adhaerentia have aggregates of small, irregular particles associated with the extracellular half of the fractured membrane, co-extensive with the widened cleft (Figs. 12 14). When the plane of fracture split one dendritic membrane at the site of a punctum, cross-fractured the cleft and then split the apposed dendritic membrane, it was apparent that both participating membranes had similar particle aggregates. In developing and mature animals there are junctions between granule cell dendrites and mossy fiber axons which have symmetric specializations in thin sections, and which are different in their appearance from synaptic junctions. We term these 'attachment plaques', but it should be emphasized that their function is entirely unknown. They have distinct electrondense fuzz symmetrically disposed subjacent to both participating membranes, co-extensive with a slightly widened junctional cleft; the fuzz extends deeper into the cytoplasm than does the postsynaptic electron-dense fuzz. There is no clustering of synaptic vesicles in the mossy fiber axons at the sites of these junctions. Inspection of serially-sectioned glomeruli at 9 and 12 days revealed that these attachment plaques commonly are nearby, but not continuous with, synaptic junctions. We were unable to identify in freeze-fractured preparations a specialization of intramembrane particle distribution in either granule cell dendritic membrane or mossy fiber axonal membrane which correlates with the site

Fig, 8. Granule cell dendritic membrane structure in the glomerulus of a 9-day-old mouse. On the extracellular half of the fractured granule cell dendritic membrane (G), there are particle aggregates not associated with deformations of membrane contour (left of A), particle aggregates associated with deformations of m e m b r a n e contour (above C), and indentations into the membrane contour without associated particle aggregates (below B). The arrow indicates a crossfractured widening of the extracellular space, at the site of an indentation into the contour of the granule cell dendrite without associated particles. An asterisk designates the cytoplasmic membrane half of the mossy fiber axon. (68,000 × .)

240

Fig. 9. Synaptic junction in a mature glomerulus. The plane of fracture has exposed the cross-fractured axoplasm ol the moss\ fiber axon (MF), a portion of the cytoplasmic half of the axonal membrane (arrow), a widened synaptic cleft (asterisk i. and an aggregate of particles on the extracellular half of the dendritic membrane (G). The aggregate is clearly co-extensive with the widened synaptic cleft. ( 132,000 x.) Fig. 10. Synapticjunction in a mature glomerulus. The plane of fracture has cross-fractured the axoplasm oflhe mossx fiber bouton (MF), exposed the cytoplasmic half of the axonal membrane (*), has cross-fractured the widened synaptic cleft (between arrowheads), exposed a portion of the extracellular half of the dendritic membrane, and then entered the cytoplasm of the dendrite (G). On the extracellular half of the dendritic membrane there is an aggregate of particles which is co-extensive with the widened cleft and also co-extensive with structure evident in the etched granule cell dendritic cytoplasm. ( 132,000 ×./ Fig. 11. Aggregate of intramembrane particles at a synaptic junction in a mature glomerulus. The aggregate of particles on the extracellular half of the fractured dendritic membrane (G) is typical of aggregates at the site of synapticjunctions in mature glomeruli, and is composed of particles with varying shapes and sizes. ( 132,000 x .) Fig. 12. Particle specialization at the site of a puncture adhaerens in a mature glomerulus. The plane of fracture has exposed the cytoplasmic half of the one granule cell dendritic membrane (*) and the extracellular half of another granule cell dendritic membrane (G). There is an aggregate of particles on the extracellular half of the membrane which is co-extensive with a slightly widened junctional cleft and with fibrillar material in the junctional cleft. The particles composing this aggregate are smaller and more angular than the particles associated with synapticjunctions. ( 132,000 x .) Fig. 13. Membrane structure at the site ofa punctum adhaerens in a mature cerebellar glomerulus. The plane of fracture ha', exposed the cytoplasmic half of one granule cell dendritic membrane (*) and the extracellular half of the adjacent granule cell dendritic membrane (G). The aggregate of small, angular particles is co-extensive with a widened cleft. On the cytoplasmic half of the membrane there is a subtle pit, presumably complementary to an aggregate of particles in that membrane (arrow). (132,000 x .) Fig. 14. Membrane structure associated with a punctum adhaerens in a mature cerebellar glomerulus. The plane of fracture has exposed the cytoplasmic half of one granule cell dendrite (*) and the extracellular half of another granule cell dendrite (G). The widening of the cleft co-extensive with the particle aggregate is evident, and one can appreciate a subtle indentation into the contour of the cytoplasmic half of the granule cell dendrite which is presumably caused by the widened junctional cleft (between the arrowheads). ( 132,000 x.)

241 of an attachment plaque. Mossy fiber axons do not have aggregates of particles associated with the extracellular half of the membrane in mature or developing animals, and so the axonal membrane at the site of an attachment plaque does not resemble the specialization of granule cell dendrite membrane at a punctum adhaerens (we chose to identify this class of junction as an 'attachment plaque' to emphasize that while this junction and puncta adhaerentia between granule cell dendrites appear to be similar in thinsectioned tissue, they are different in freezefractured preparations). We cannot exclude the possibility that some of the smaller deformations of membrane contour without associated particle aggregates in developing granule cell dendrites, which we have referred to as 'initial junctions', actually represent attachment plaques. In serially thin-sectioned preparations of developing glomeruli, attachment plaques are much less common than synaptic junctions. The area of individual attachment plaques is usually less than 6 x 104 nm 2, smaller than most synaptic junctions in developing glomeruli. It seems much less likely that either of the specializations of particle distribution in immature granule cell dendrites apposed to mossy fiber axons (immature junctions or non-junctional particle aggregates) represent attachment plaques because the axonal membrane lacks such structure. Since the membrane specializations at attachment plaques appear in thin sections to be identical in presynaptic and postsynaptic membranes, one would have expected the corresponding structures in freeze-fractured preparations also to be similar in axon and dendrite. The attachment plaques in cerebellar glomeruli, as visualized in thin-sectioned preparations, are apparently identical to those linking primary auditory afferent axons and principal cell plasmalemma in the synapses of the end bulbs of Held in the anteroventral cochlear nucleus. There, too, no corresponding structure has been found in freezefractured preparations ]2. It is difficult to recognize specializations ofintramembrane particle distribution in the presynaptic membrane region of mossy fiber axons. In developing and mature animals there are

I Z 0 .J

8 DAY

211

I1. o 0. U. 0 1J

"i .... i"

" " ieb- " " i l -

AREA

U ~E DAY

0 111 0

.... i .... / ....

i

" " iali

" " in6-

AREA

Z 0 I-4 I-~ u -1

~DUL'r

0 u. o 111

[mH AREA

G~ 1 ~

~ !.

r~.)

Fig. 15. Histograms of synaptic junction size in (top to bottom) 9 day, 12 day, and adult mice. The synaptic size, plotted on the abscissa as square nanometers, was obtained by mea; suring the length of the synaptic junction in all of the serial sections in which it appeared, summing the measurements, and multiplying by 60 nm (an estimate of section thickness). In each histogram, the ordinate is the percentage of synapses in animals of that age which falls within the indicated size range.

242 clusters of sparse, large particles associated with the cytoplasmic membrane half, which probably represent presynaptic specializations. In adult animals, the membrane domain of these larger particles is roughly the size of a synaptic junction, and the contour of the axonal membrane often is evidently embossed by the widened synaptic cleft. Commonly there are sites of coated vesicle formations at the periphery of these presynaptic active zones, and less frequently one can find synaptic vesicle fusion sites among the large particles of the presynaptic active zone. Unfortunately, suboptimal aldehyde fixation can cause an artifactual reticular (as contrasted to uniform) distribution of particles, and distinguishing the subtle cluster of large particles on the presynaptic cytoplasmic half from artifact can be difficult. There are no aggregates of particles associated with the extracellular half of fractured mossy fiber axon membranes, except for the 3 5 particles clustered at sites of coated vesicle formation.

DISCUSSION In thin-sectioned tissue, the morphological characteristics of the synaptic junctions between mossy fiber axons and granule cell dendrites appear to be similar in developing and mature animals except for overall size. Many of the synaptic junctions in cerebellar glomeruli during the first 2 postnatal weeks are larger in area than any of the synaptic junctions in adult cerebellar glomeruli:22<27. It is not obvious in thin-sectioned tissue whether larger junctions in developing glomeruli shrink to the adult size, whether they divide, or whether they are replaced. In freeze-fractured tissue, however, developing and adult junctions are quite different in the characteristics of the associated particle aggregate. The change in the appearance of the postsynaptic particle aggregate during development is most consistent with the idea that the junctions remodel to adult size by shrinking. In freeze-fractured tissue, developing granule cell dendrites have at least 3 distinct specializations of membrane contour and intramembrane particle distribution where they face mossy fiber

axons. We would like to propose that these specializations may correspond to different stages in synaptogenesis. One of these specializations, which we have termed an "initial junclion', appears as an indentation into the dendritic and axonal contour where junctional cleft material has embossed the apposed membranes. These initial junctions probably have electron-dense material in the junctional cleft and lining the axonal and dendritic membranes when viewed in thin sections. A second specialization m developing dendritic membrane structure appears as an aggregate of particles associated with the extracellular half of the fractured dendritic membrane, located in a region of the dendrite not embossed by a widened junctional cleft. Since we have not found electron-dense material lining the dendritic membrane in the absence of cleft or presynaptic specializations, these particle aggregates probably do not have corresponding electron-density in thin sections. In adult tissue, we found neither indentations without particle aggregates ('initial junctions') nor particle aggregates (see also refs. 16, 18, 30) without junctions. We suppose that the constituent particles of these aggregates become associated with the 'initial junctions' to form immature synapticjunctions. These have the full morphological complement of electron-dense specializations and co-extensive particle aggregates. There is a continuum in the appearance of the immature synaptic junctions, ranging from those with large areas and loosely packed particles to those with smaller areas and more tightly packed particles. This suggests to us that immature synaptic junctions acquire a fairly uniform number of particles, and the area of the junction subsequently decreases, causing an increase in particle packing density. It is difficult to exclude the possibility that some large immature junctions actually subdivide: however, if this does happen, it must occur before the particle packing density has approached adult values. Unfortunately, it is difficult to be certain that the particles aggregated in immature junctions are the same as those in mature synaptic junctions. The particles in these freeze-fractured granule cell dendrites almost certainly represent

243 proteins which partially or entirely span the hydrophobic interior of the membrane 32.33.In aldehyde-fixed, glycerol cryo-protected tissue, the appearance of particles associated with a particular membrane-associated protein may be distinctive: one can, for example, easily distinguish between the particles at the sites of acetylcholine receptors and the particles at sites of gap junctions. It is likely that the particles aggregated in smooth portions of immature granule cell dendrites, the particles associated with the indented portions of the immature dendrites, and the particles in mature synaptic junctions are all the same because they have the same general shape and size; however, we have no direct evidence concerning the actual composition of the proteins represented by the particles. We have emphasized the observation that particle aggregates form in developing granule cell dendrites at non-junctional sites. We do not know whether this phase of dendritic differentiation is dependent on some presynaptic influence. The constituent particles of non-junctional aggregates ultimately are incorporated into junctions, and so appear to be participating in an interaction with the axon at this later stage. Studies of the development of the parallel fiber axon synapse with Purkinje cell dendritic spines reveal even more extensive dissociation between presynaptic and postsynaptic maturation. In w e a v e r 14.15.17A9.36.37.4°.41 and reeler mice ~7,35, Purkinje cell dendritic spines form and acquire postsynaptic particle aggregates co-extensive with 'cleft' material despite a complete lack of parallel fibers in the region of the spine. In mice homozygous for the Purkinje cell degeneration (pcd) mutation, synaptogenesis between parallel fibers and Purkinje dendritic spines proceeds apparently normally through postnatal day 182~. Then, however, otherwise unremarkable postsynaptic specializations are found in aberrant positions on the shafts of distal dendrites, without associated axons. In this mutant and in the reeler, the dissociation between presynaptic element and maturation of postsynaptic membrane structure does not represent direct mutant gene action, because the Purkinje cells are able to form normal synapses elsewhere in the dendritic arbor.

In a recent study of synaptogenesis at photoreceptor-to-horizontal cell synapses in Xenopus retina, Nagy and Witkovskf 8 provide evidence that the presynaptic specialization is assembled before postsynaptic structure becomes apparent. They argue that the particles of the postsynaptic site probably become associated with the membrane elsewhere, and become trapped at the junctional site. They did not find aggregates of junction-like particles at sites other than synaptic junctions. The findings of McGraw and colleagues23.24 also contrast with those presented here. In the optic tectum, they detected a successive accretion of particles and enlargement of aggregates at postsynaptic sites, and did not encounter non-synaptic aggregates. It seems likely that the details of synaptogenesis will be found to vary at different synapses, perhaps in concert with the nature of the neurotransmitter. Synaptic junction maturation in cerebellar glomeruli appears to have certain features in common with the formation of neuromuscular junctions. Acetylcholine receptors insert into the muscle membrane and aggregate prior to the arrival of the afferent axon 2-s. When axons contact the muscle surface, they do not seek out the sites of clustered receptors; instead, aggregates of receptors from beneath the axons2,3-~-9,probably by the accumulation of receptors which have moved in the plane of the membrane from extra-junctional aggregates42. Similarly, the evidence presented here indicates that the synapserelated proteins in immature granule cell dendrites insert into the membrane and can aggregate at a site remote from that ofa synapticjunction; these correspond to the particle aggregates without associated membrane indentations. Subsequently, the synapse-related proteins singly or in aggregate translocate to the site of the junctional cleft. Does this sequence in synaptic junction formation hold any implications for the function of the various components? Particles associated with the extracellular half of the postsynaptic membrane have been found to be a feature of excitatory synapses in the cerebellar cortex ~6,~8, olfactory bulb 2°, and anteroventral cochlear nucleus ~2. However, the exact nature of the protein

244 represented by the particles is unknown. We would point out that their position in the postsynaptic membrane and their probable behavior during the formation of the synaptic junction are exactly like those of neurotransmitter receptor sites in vertebrate striated muscle.

REFERENCES 1 Altman, J., Postnatal development of the cerebellar cortex in the rat: maturation of the components of the granular layer, J. comp. Neurol., 145 (1972) 465 514. 2 Anderson, M. J. and Cohen, M. W., Nerve-induced and spontaneous redistribution of acetylcholine receptors on cultured muscle cells, J. Physiol. (Lond.), 268 (1977) 75"7 773. 3 Anderson, M. J., Cohen, M, W. and Zorychta, E., Effects ofinnervation on the distribution of acetylcholine receptors on cultured muscle cells, J. Physiol. (Lond.), 268 (1977) 731-756. 4 Bevan, S. and Steinbach, J. H., The distribution ofalphabungarotoxin binding sites on mammalian skeletal muscle developing in vivo, J.Phvsiol. (Lond.), 267 (1977) 195 213. 5 Burden, S., Development of the neuromuscular junction in the chick embryo: the number, distribution, and stability of acetylcholine receptors, Develop. Biol., 61 (1977) 79 85. 6 Fox, C. A., Hillman, E. E., Siegesmund, K. A. and Dutta, C. R., The primate cerebellar cortex: a Golgi and electron microscopic study. In C. A. Fox and R. Snider (Eds.), The Cerebellum Progress in Brain Research, Vol. 25, Elsevier, Amsterdam, 1967, pp. 174 255. 7 Fox, C. A., Siegesmund, K. A. and Dutta, C. R., The Purkinje cell dendritic branchlets and their relation with the parallel fibers: light and electron microscopic observations. In M. M. Cohen and R. S.Snider (Eds.), Morphological and Biochemical Correlates of Neural Activity, Harper and Row, NewYork, 1964, pp. 112 141. 8 Fertuck, H. C. and Salpeter, M. M., Quantitation of junctional and extrajunctional acetylcholine receptors by electron microscope autoradiography after 125-l-alphabungarotoxin binding at mouse neuromuscular junctions, J. CellBiol.. 69(1976) 144 158. 9 Frank, E. and Fischbach, G. D., Early events in neuromuscular junction formation in vitro, J. Cell Biol., 83 (1979) 143 158. 10 Gray, E. G., Axo-somatic and axo-dendritic synapses of the cerebral cortex: an electron microscopic study, J. Anat., 93 (1959)420 433. I 1 Gray, E. G., The granule cells, mossy synapses and Purkinje spine synapses, J. A nat., 95 ( 1961 ) 345- 356. 12 Gulley, R. L., Landis, D. M. D. and Reese, T. S., Internal organization of membranes at end bulbs of Held in the anteroventral cochlear nucleus, J. comp. Neurol., 180 (1978) 707 742. 13 Hamori, J. and Szentagothai, J., Participation of Golgi neuron processes in their cerebellar glomeruli: an electron microscopic study, Exp. Brain Res., 2 (1966) 35 48.

ACKNOWI.EDGEMENTS

We thank Mr. L. Cherkas for photographic help and Ms. Darlene Jackson for preparing the manuscript. This work was supported by a N1H postdoctoral fellowship 1F32 NS06638-01 (J.J.H.), and a Teacher-lnvestigator Award NS 00353 (D.M.D.L.) and NIH Grant NS 15573. 14 Hanna, R.B., Hirano, A. and Pappas, G. D., Membrane specializations of dendritic spines and glia in the weaver mouse cerebellum: a freeze-fracture study, J ('ell Biol.; 68 (1976) 403 410. 15 Hirano, A. and Dembitzer, H. M., Cerebellar alterations in the weaver mouse, J. Cell Biol.. 56 (1963) 478 486. 16 Korte, G. E. and Rosenbluth, J., Freeze-fracture study of the post-synaptic membrane of the cerebellar mossy fiber synapse in the frog, J. comp. Neurol.. 193 (1980) 689 700. 17 Landis, D. M. D. and Landis, S. C., Several mutations in mice that affect the cerebellum. In R. A. P. Kark, R N. Rosenberg and L. J. Schut (Eds.), Advances in Neurology. Vol. 2l. Raven, New York, 1978, pp. 85 105. 18 Landis, D. M. D. and Reese, T. S., Differences in membrane structure between excitatory and inhibitory synapses in the cerebellar cortex, J. comp. ~eurol,. ]55 (1974)93 126. 19 Landis, D. M. D. and Reese, T. S., Structure of the Purkinje cell membrane in staggerer and weaver mutant mice,./, cornp. Neurol., 171 (1977)247 260. 20 Landis, D. M. D., Reese, T. S. and Raviola, E., Differences in membrane structure between excitatory and inhibitory components of the reciprocal synapse in the olfactory bulb, J. comp. Neurol., 155 (1974) 67 92. 21 Landis. S. C. and Mullen, R. J., The development and degeneration of Purkinje cells in pcd mutant mice. J. comp. Neurol., 177(1978) 125 144. 22 Larramendi, L. M. H., Analysis ofsynaptogenesis m the cerebellum of the mouse. In R. Llinas (Ed.), Neurobiotogv of Cerebellar Evolution and Development. American Medical Association, Chicago, 1969, pp. 803 843. 23 McGraw, C. F. and McLaughlin, B. J., Fine structural studies of synaptogenesis in the superficial layers of the chick optic tectum.J. Neurocvtol., 9 (1980) 79 93. 24 McGraw, C. F., McLaughlin, B. J. and Boykins, L. G,, A freeze-fracture study of synaptic junction development in the superficial layers of the chick optic rectum. ,l. Neurocvtol., 9 (1980) 95 106. 25 MiMe, I. L. and Sidman, R. L, An autoradiographic analysis of histogenesis in the mouse cerebellum, Exp. Neurol., 4(1961) 277 296. 26 Mugnaini, E., Maturation of nerve cell populations and establishment of synaptic connections in the cerebellar cortex of the chick. In R. Llinas (Ed.), Neurobioloxv o{" Cerebellar Evolution and Development, American Medical Association, Chicago, 1969, pp. 749 782. 27 Mugnaini, E. and Forstronen, P. F., Ultrastructural studies on the cerebellar histogenesis. I. Differentiation of granule cells and development of glomeruli in the chick embryo, Z. Zellforsch.. 77(1967) 115 143. 28 Nagy, A. R, and Witkovskv, P., A freeze fracture study of

245

29 30 31 32 33

34

35

synaptogenic contacts in the outer plexiform layer of the retina, J. Neurocvtol.. 10 ( 1981) 897- 919. Palay, S. L., The electron microscopy of the glomeruli cerebellosi. In Cvtology of Nervous Tissue, Taylor and Francis, London, 1961, pp. 8~84. Palay, S. L. and Chan-Palay, V., Cerebellar Cortex: Cytology and Organization, Springer, New York, 1973. Peters, A.. Palay, S. L. and Webster, H. de F., The Fine Structure of the Nervous System, W. B. Saunders, Philadelphia, pp. 6, 126, 138, 144~ 145. Pinto da Silva, P., Translational mobility of the membrane intercalated particles of human erythrocyte ghosts, J. CellBiol., 53(1972) 777 787. Pinto da Silva, P., Interpretation of freeze-fracture and freeze-etch images: morphology and realism. In J. E. Rash and C. S. Hudson, (Eds.), Freeze-Fracture: Methods, Artifacts, and Interpretations, Raven Press, New York, 1979, pp. 185-193. Rakic~ P., Neuron glia relationship during granule cell migration in developing cerebellar cortex. A Golgi and electronmicroscopic study in Macacus rhesus, J. comp. Neurol., 141 (1971)283 312. Rakic, P., Synaptic specificity in the cerebellar cortex: study of anomalous circuits induced by single gene mutations in mice. In The Synapse, CoM Spring Harbor Symposia on Quantitative Biology, Iiol. XL, 1976, pp. 333346.

36 Rakic, P. and Sidman, R. L., Organization of cerebellar cortex secondary to deficit of granule cells in weaver mutant mice, J. comp. Neurol., 152 (1973) 133 162. 37 Rakic, P. and Sidman, R. L., Sequence of developmental abnormalities leading to granule cell deficit in cerebellar cortex of weaver mutant mice, J. comp. Neurol., 152 (1973) 103- 132. 38 Ramon y Cajal, S., Histologie du Systeme Nerveux de I'Homme et des Vertebres, Maloine, Paris, 1911, reprinted by Consejo Superior de lnvestigaciones Cientificas, Madrid, 1955. 39 Sandri, C., Akert, K., Livingston, R. B. and Moor, H., Particle aggregations at specialized sites in freeze-etched postsynaptic membranes, Brain Res., 41 (1972) 1- 16. 40 Sotelo, C., Permanence and fate of paramembranous synaptic specializations in 'mutants' and experimental animals, Brain Res., 62 (1973) 345-351. 41 Sotelo, C., Dendritic abnormalities of Purkinje cells in the cerebellum of neurologic mutant mice (weaver and staggerer). In G. W. Kreutzberg (Ed.),Advances in Neurologv, Iiol. 12, Physiology and Pathology of Dendrites, Raven Press, New York, 1975, pp. 335-351. 42 Weinberg, C. B., Reiness, C. G. and Hall, Z. W., Topographical segregation of old and new acetylcholine receptors at developing ectopic endplates in adult rat muscle, J. CellBioL, 88 (1981)215 218.