Some aspects of synaptogenesis in the spinal cord of the chick embryo: a quantitative electron microscopic study.

We have quantitatively examined the development of synapses in the ventral part of the lumbar spinal cord of the chick from embryonic day 4 until adulthood. The first synapses occur on day 4 and are of the axo-dendritic type; they are invariably located adjacent to the border between the intermediate and marginal zones. Initially there are more synapses in the presumptive white matter than in the motoneuron neuropil, but this trend is later reversed; however, we found numerous axo-dendritic synapses throughout much of the ventrolateral white matter even in the adult stage. The first axo-dendritic synapses always contain spherical synaptic vesicles and have symmetric membrane specilizations. By day 7 a few of these synapses were found to have mixed populations of spherical and flattened vesicles and asymmetric membrane specilizations. After hatching there are still considerably more axo-dendritic synapses with symmetric membrane specializations. Axo-somatic synapses were first found on embryonic day 6 and were typically located on motoneurons lying adjacent to the marginal zone. These axo-somatic synapses contain a few spherical synaptic vesicles and have symmetric membrane densities. Flattened synaptic vesicles were first found on day 10 and increased throughout development. Although a few axo-somatic synapses with asymmetric membrane specializations were found at practically all stages, the symmetric type was always in the majority. An attempt was made to relate these observations with physiological, behavioral and neuroembryological findings from birds and other forms. For example, the fact that axo-dendritic synapses always appear prior to axo-somatic contacts would seem to rule out the role of somatic synapsesin the initial induction of dendritic growth in the spinal cord.     

Chemosensitive ventrolateral medulla in the cat: the fine structure and GABA-induced cardiovascular effects.

These studies were designed to provide information on both GABA influences on the neuronal structures within the rostral ventrolateral medulla (RVLM) involved in the cardiovascular control, and fine structure of the GABA-sensitive RVLM in cats. By electron microscopy, the S-type asymmetric junctions (believed to mediate excitation) as well as F-type symmetric junctions (thought to provide inhibition) were identified within the RVLM area examined. The axonal terminals thick with spherical electron-transparent synaptic vesicles, measuring about 50 nm in diameter and forming S-type synapses, were distributed within the entire area. Large numbers of axonal terminals forming the F-type synapses and filled with flattened synaptic vesicles with a longitudinal axis of around 60-80 nm were found at the site located 5-6 mm caudal to the trapezoid bodies. Both spherical electron-transparent and flattened synaptic vesicles could be found together with electron-dense vesicles averaging 80-160 nm in diameter. No axonal terminals containing only the latter type of vesicles were found within the area studied. Some axo-axo-dendritic F-S-type synapses were identified within the RVLM. Unilateral injection of GABA into the RVLM site located 2-5 mm caudal to the trapezoid bodies induced a dose-related fall in the systemic arterial pressure and inhibition of the renal nerve sympathetic activity, the most marked responses being found at the site 4-5 mm caudal to the trapezoid bodies. It also resulted in decrease of heart rate and myocardial contractility when injected into a distinct site, located 0-1.5 mm rostral to the outflow of the 12th cranial nerve roots. There was some laterality in GABA effects on heart rate and myocardial contractility following its injections into the RVLM on the left and right side. Injections of bicuculline into the GABA-sensitive RVLM site resulted in an increase in cardio-haemodynamic responses together with the enhancement of renal nerve sympathetic activity in a dose-related fashion. These results suggest that, in cats, GABA-sensitive sympathoexcitatory neuronal structures in the RVLM alter peripheral sympathetic vasomotor and cardiac nerve background activities through GABA action on the bicuculline-sensitive GABAa-receptors.     

EM and EM Golgi study on structure of nucleus rotundus in chicks.

The analysis of EM structure of nucleus rotundus completes the results got by Golgi study. The fine structure of neurons and neuropil of the nucleus and the synaptic relations were studied by EM. The fine structural details of principal neuron were described. Several synapses of symmetrical type with flattened vesicles in large terminals contacted the cell body and also the origin and proximal part of the main dendrites. In the neuropil synaptic junctions were formed by terminals that contained (1) spherical vesicles with occasionally very few dense core vesicles, (2) flattened synaptic vesicles. Terminals that contained spherical vesicles were associated with asymmetric synaptic densities, and terminals that contained flattened vesicles formed symmetric junctions. Synapses of asymmetric type associated mostly with terminal sections of dendrites forming glomerular-like structure. Synapses of symmetric type with flattened synaptic vesicles contacted the branching areas of dendritic terminals and side-branches, the origin of main dendrites and the cell surface of principal neuron.     

Morphology of afferent and efferent synapses in hearing organ of the goldfish

The sacculus, the hearing organ, of the goldfish was fixed with three different kinds of fixatives: osmium tetroxide, glutaraldehyde and formaldehyde. Three kinds of synapses are encountered in the sensory epithelium: the synapse between the hair cell and the afferent nerve ending, the synapse between the efferent nerve ending and the hair cell, and the synapse between the hair cell and the afferent nerve ending, there are synaptic vesicles about 370-430 Å in average diameter in the hair cell side of the junction. The vesicles are always round in dependent of the fixative used. In the other two kinds of synapses, the efferent ending contains smaller vesicles about 310-360 Å in average diameter. The shape of the vesicles changes according to the fixative used; round in osmium tetroxide-fixed materials, and elongated in aldehydefixed materials. The synapse between the hair cell and the afferent nerve ending in excitatory in nature, while the other two types of synapses are thought to be inhibitory in nature.     

Fornix afferents to the anteroventral thalamic nucleus: An EM study in the rat

LENN, N. J. Fornix afferents to the anteroventral thalamic nucleus: An EM study in the rat. BRAIN RES. BULL. 3(6) 589–593, 1978.—Three types of synapses occur in the anteroventral thalamic nucleus (AVN). Type 1 consists of small (0.5–0.8 μm) axonal endings densely packed with spherical synaptic vesicles. They form markedly asymmetrical synaptic contacts with distal portions of dendrites. Degenerative changes in these axons following destruction of the fornix identify them as the endings of the subicular projection to AVN. Type 2 synapses consist of large (1.0–1.5 μm) axonal processes containing spherical vesicles which form asymmetrical synapses on more proximal dendrites. Type 3 endings consist of large unidentified processes containing spherical, and occasionally flattened, synaptic vesicles forming symmetrical contacts with the largest stem dendrites. Neither of these synaptic types were modified by fornix lesions. The synaptic arrangements within AVN are simpler than other thalamic nuclei in that serial synapses and synaptic glomeruli are not present.     

Heterogeneity of synaptic vesicles in the squid giant fibre system

Four synaptic contacts in the squid giant fibre system (from the afferent boutons on the first order giant cell in the brain through the giant synapses in the palliovisceral lobe and the stellate ganglion), as well as neuromuscular junctions in the fin, contain populations of agranular (electron transparent) synaptic vesicles. A classification of these vesicle populations is possible on the basis of the average vesicle diameter and the reactivity with zinc iodide-osmium (ZIO). We distinguish 4 synapses with significantly different average vesicle sizes, 3 synapses with ZIO-positive vesicles and two synapses with ZIO-negative vesicles. In two consecutive links of the giant fibre system there are always structurally different synaptic contacts.     

Ultrastructural characterization of synaptic terminals formed on newly generated neurons in a song control nucleus of the adult canary forebrain

The fine structure of synaptic terminals contacting neurons generated in the forebrain of adult male canaries was investigated by autoradiography and electron microscopy. The procedure for labeling the new neurons included pretreating adult canaries with 3H-thymidine and sacrificing them 23-45 days later. Neurons were identified as newly generated by the presence of 3H-thymidine in the cell nucleus. The new neurons in the nucleus hyperstriatum ventralis, pars caudalis (HVc) were identified by autoradiography and light microscopy and examined with electron microscopy. Several types of synaptic terminals contacted the cell body and proximal dendrites of the newly formed neurons. Synaptic junctions were formed by terminals that contained spherical, agranular vesicles, large dense-core vesicles and spherical, agranular vesicles, and pleomorphic or flattened synaptic vesicles. Terminals that contained spherical vesicles were most often associated with asymmetric synaptic densities, and terminals that contained pleomorphic or flattened vesicles formed symmetric junctions. New neurons were also contacted by small terminals that contained few vesicles and had little pre- or postsynaptic density associated with the junction; these terminals may be a special type or may be in the process of developing their synaptic contact with the new neuron. In addition, rare terminals that appeared to be degenerating or to contain debris from other degenerating neural elements contacted new neurons. In summary, these data indicate that the new neurons, which are known to be inserted into existing neural networks, receive synaptic input from at least three different sources.     

Observations on the organization of the dendritic processes and receptor terminations in the abdominal muscle receptor organ of homarus

The dendritic processes and terminals of the abdominal muscle receptor organ of the lobster have been studied with the electron microscope. In both the “fast” and “slow” adapting components of the muscle receptor organ, dendritic processes ramify within the body of the stretch muscle in a zone of “relative muscle exclusion,” where the continuity of the muscle bundle is interrupted by a zone of connective tissue. The cell bodies and larger dendritic processes are ensheathed in alternate layers of connective tissue and Schwann cells, but the smallest dendritic twigs are entirely divested of their Schwann cell envelope. The membrane of the dendritic terminations is intimately applied to either connective tissue of the zone of relative muscle exclusion or to the sarcolemma of the stretch muscle. This apposition is interpreted as morphological specialization for transduction or mechanoelectric conversion. It is suggested that these two types of dendritic terminals may represent two distinct sites for transduction and may also mediate different modalities of input. The distribution of dendritic processes and terminals differs in the fast and slow systems. In the fast system the dendrites quickly ramify into a glomerular burst of processes near the junction of stretch muscle and connective tissue. In the slow system, the dendrites ramify more gradually and lie more nearly parallel to the long axis of the stretch muscle. In both the fast and slow systems, but especially in the dendritic terminals of the fast neuron, collections of vesicles 400–2000 Å in diameter have been observed. These vesicles are clearly different in fine structure from those found in the neuromuscular junction or axo-dendritic synapses. An effort is made to relate neurophysiological characteristics of this system with the fine structural observations, with particular attention to the mechanism of transduction, or mechanoelectric conversion.     

Neuronal gap junctions and morphologically mixed synapses in the spinal cord of a teleost, Sternarchus albifrons (gymnotoidei)

Spinal cord motoneurons in the gymnotid, Sternarchus albifrons, were studied electron microscopically with special reference to the freeze-fracture method. Two types of motoneurons were identified. Electromotor neurons are monopolar and are located in a midline column dorsal to the ventral gray. These cells have a small fraction of their surface covered by synapses from descending axons, often at nodes. The synapses have multiple gap junctions, but few presynaptic vesicles or other correlates of chemical transmission. The gap junctions have an ordinary appearance in freeze-fracture replicas and exhibit a highly ordered substructure. The not infrequent appositions between the cell bodies of electromotor neurons exhibit no junctional specializations. Ordinary motoneurons are multipolar and densely covered with axosomatic and axodendritic synapses. In thin sections these synapses can be divided into two groups according to whether the vesicles are spherical or flattened. Gap junctions occur only at the first type, thus forming ‘morphologically mixed’ synapses. In freeze-etch replicas of motoneurons, the gap junctions are often found near clusters of postsynaptic E face particles elsewhere associated with excitatory chemical transmission. In addition, vesicle attachment sites occur in the presynaptic membranes of some synapses with gap junctions. The morphological observations are consistent with dual chemical and electrical transmission at these particular synapses, i.e. electrical excitation across gap junctions and chemical excitation at active zones with spherical vesicles and post-synaptic E face particles.     

Fine structure of rat locus coeruleus

Locus coeruleus of the rat was studied in material prepared by aldehyde-osmium fixation. Cell bodies of locus coeruleus neurons possess large nuclei with a prominent nucleolus, a homogeneous karyoplasm of moderate density, and occasional indentations of the nuclear membrane. The cytoplasm is rich in organelles, including an extensive network of endoplasmic reticulum which forms well organized Nissl bodies. The highly developed Golgi apparatus surrounds the nucleus and extends into large dendritic trunks. In coronal section, cell bodies appear elongated along an approximate dorso-ventral axis, and most dendrites as well as axons appear in cross-section. In parasagittal sections the cells are very elongate, with dendrites and axons in the neuropil mostly cut longitudinally. Thus, locus coeruleus neurons possess disc-shaped dendritic fields parallel to the anterior-posterior axis of the brainstem, with predominantly longitudinal axo-dendritic synaptic configurations. Presynaptic profiles in locus coeruleus neuropil were classified according to the characteristics of their vesicle populations and other features. The most frequently encountered synaptic ending was characterized by small, round, densely packed synaptic vesicles, and comprised approximately 41% of the total sample of 775 synapses. Another group having large, rounded synaptic vesicles, which could be traced in a number of instances to large myelinated axons, accounted for 20% of the sample. Synaptic endings having large, flattened vesicles were also numerous, comprising 23% of the total. Another category of presynaptic endings was identified as those possessing numerous, small, flattened vesicles and comprising about 11% of the sample. Presynaptic endings having many vesicles of mixed sizes accounted for 2% of the total, and another group of the same proportion having small, rounded synaptic vesicles but also an unusually large number of larger, dense-cored vesicles was also present. Two other categories of synaptic endings were encountered, each comprising less than 1% of the total. One of these was derived from small, unmyelinated axons and contained clusters of pleomorphic synaptic vesicles. The other consisted of dendro-dendritic synapses between locus coeruleus neurons and also displayed small clusters of pleomorphic synaptic vesicles near the zone of synaptic apposition. Quantitative analysis revealed that most afferents to the nucleus synapse onto dendrites ranging between 0.5 and 2.5 micrometers in diameter and onto spine-like appendages derived from somata and dendrites. There were no significant differences between different categories of afferent terminals and their spatial distribution onto various postsynaptic targets of locus coeruleus neurons.     

Development of the human cervical spinal cord with reference to synapse formation in the motro nucleus

Synaptic development in the motor neuropil of the cervical spinal cord was quantitatively studied by light and electron microscopy in human embryos and fetuses ranging from five to 19 weeks of ovulation age. The numbers of axodendritic synapses increase substantially at the end of the eighth week. However, axosomatic synapses rapidly proliferate between 10.5 and 13 weeks of ovulation age. The increases in both types of synapses generally coincide with the behavioral chages in human fetuses that have been reported by other investigators. Only synaptic boutons containing spherical vesicles (S-type synapses) were found in the motor neuropil throughout the stages examined; no synapses with flattened vesicles (F-type synapses) were encountered. The majority of these synaptic boutons contain only a small number of synaptic vesicles (fewer than 20), although the number tends to increase with maturation. There is no significant maturational change in the size of synaptic vesicles. The present study suggests that synapse formation in the motor neuropil of the human fetus cervical spinal cord may continue up to 19 weeks of ovulation age because immature types of synapse are found in all fetuses. Myelin formation probably begins by the 11th week.     

Coated vesicle morphology and sub-populations at the neuromuscular junction

Serial sections of frog cutaneous pectoris neuromuscular junctions were examined to determine if isolated coated vesicles in one section are connected to infoldings of the presynaptic membrane in adjacent sections or are truly pinched off from the plasmalemma. Twenty percent of the coated vesicles examined serially were isolated from plasmalemma. In addition, two populations of coated vesicles were observed: those the size of synaptic vesicles and a smaller population (8% of total) of larger diameter (100 nm).     

Observations on the synaptology of intracellularly injected spinothalamic tract neurons in the cat

Single spinothalamic tract cells in the deep lumbar group have been injected with horseradish peroxidase and their dendritic trees examined on the electron microscope. Large dendrites are surrounded by rosettes of from 5 to 12 synaptic terminals. Most terminals contain round vesicles but some terminals contain flattened clear vesicles and others contain membranous tubes or cisterns. Large (mean diameter 100 nm) round, dense-core vesicles are found in both terminals containing clear round vesicles and terminals containing flattened vesicles. As yet none of those terminals on STT cell dendrites have been found participating in axo-axonic synapses, suggesting that presynaptic inhibition or facilitation plays no role in modulating the responses of these cells to presynaptic activity.     

Coated vesicles and synaptogenesis. A developmental study in the cerebellar cortex of the rat

The distribution of coated vesicles was examined by electron microscopy in the cerebellar cortex of rats aged 0, 3, 5, 7, 10, 12, 15, 21 and 30 days. Vesicles with external coverings were identified in closed (spherical), open (flask-like) and flattened forms. Closed coated vesicles were seen in large numbers in the vicinity of the Golgi apparatus, and sometimes in direct continuity with its cisterns, in rapidly growing regions of immature neurons, such as the apical cones of Purkinje cells in 3–7-day-old animals, and in the dendrites of Purkinje cells in the 7–10-day-old animals. Closed, open and flat coated vesicles were seen on the surface of cells, and in neurons these were frequently opposite cell processes with which synapses are formed, such as in the dendrites of Purkinje cells opposite parallel fibers. Occasionally, these open coated vesicles were situated opposite cell processes which had dense membranes, and sometimes they were continuous with the dense membranes of early attachment sites or presumptive synapses. In the mossy axon terminals in the glomeruli, open granular coated vesicles were common in the 21- and 30-day-old rats.

It was postulated that the coated vesicles may be involved in the formation of early attachment sites or of the dense membranes of synapses, and the following sequence of events was suggested. Through a budding process spherical, coated dense membranes are formed from the membranes of the cisterns of the Golgi apparatus. These spherical vesicles migrate to the surface of cells and after attachment to the membrane open up and unfold over its surface. When this unfolding becomes complete the flat dense membranes contribute material to attachment sites or synaptic membranes. Coated vesicles may also contribute to the expansion of the dense membrane surfaces of already formed synapses. The relation between these coated vesicles and micropinosomes was discussed and examples were illustrated where parts of the parallel fiber actually protruded into the cavity of the Purkinje cell coated vesicles.     

Ultrastructural quantitative analysis of glutamatergic and GABAergic synaptic terminals in the phrenic nucleus after spinal cord injury

Quantitative analysis of electron microscopic postembedding immunochemically stained material indicates that 48% of all terminals in the rat phrenic nucleus are glutamatergic and 33% are γ-aminobutyric acid (GABA) ergic. Three distinct types of glutamatergic terminals were observed in the rat phrenic nucleus: terminals characterized by large, loosely arranged spherical synaptic vesicles (SI) or small, compact spherical synaptic vesicles (Ss) and elongated terminals containing spherical synaptic vesicles with neurofilaments (NFs). All three types of glutamatergic terminals display asymmetrical synaptic membrane densities with postsynaptic dense bodies being present in some of the S-type terminals. The GABAergic immunoreactive terminals in the phrenic nucleus most closely resemble F-type terminals. They are characterized by flattened or pleomorphic synaptic vesicles and symmetric synaptic membrane densities. Among the 48% glutamatergic terminals, 27% are SI, 65% are Ss, and 8% are NFs, respectively. Significantly fewer glutamate, GABA, and unlabeled terminals per unit area are present in the phrenic nucleus 30 days after a C2 spinal cord hemisection as compared to nonhemisected controls. The average number of active zones per terminal, however, is greater in the hemisection group (1.45 ± 0.03) than in the control group (1.34 ± 0.03), with the active zones in the glutamate terminals mainly accounting for this difference. Moreover, the length of the active zones in the glutamate terminals was significantly longer in the hemisection group (0.37 ± 0.013 μm) as compared to the controls (0.24 ± 0.008 μm). In addition, the mean length of synaptic active zones in GABAergic terminals was also found to be longer in the hemisection group (0.36 ± 0.022 μm) as compared to controls (0.28 ± 0.014 μm). Finally, there is also a significantly higher ratio of synaptic active zones to the total number of glutamate-labeled terminals after injury (1.73 ± 0.08) as compared to controls (1.41 ± 0.04). The number of double/multiple synapses, the percentages of Sl, Ss, and NFs-type terminals, and the percentages of synaptic active zones contacting either distal dendrites or proximal dendrites/somata do not change significantly 30 days after injury. These results are important for a more complete understanding of the synaptic plasticity that occurs in the phrenic nucleus after spinal cord injury and to show how the plasticity may relate to the unmasking of latent bulbospinal respiratory connections which restore function to the hemidiaphragm paralyzed by an ipsilateral spinal cord hemisection. © 1996 Wiley-Liss, Inc.     

Laminar analysis of the synaptic organization of the frog optic tectum]

Distribution of axo-axonic and axo-dendritic synapses, nervous terminals and neuron somata in the depth of the optic tectum was studied in frog Rana temporaria L. in norm and 6-9, 60 and 134 days after contralateral eye enucleation. In outer plexiform layer 9 density of synapses was found to be maximal near the surface of the tectum; it decreased in deeper levels. In the outer zone of layer 9 (about 0.30 micrometer from the surface) many axo-axonic synapses were found. Terminals of myelinated optic fibres of large diameter ("dark" terminal degeneration) were widely distributed in layer 9. Density of axo-dendritic synapses in deep plexiform layer 5 was similar to that in layer 9. Numerous nervous terminals containing not only light synaptic vesicles, but also granular vesicles were seen in layer 5 and in adjacent zones.     

Projection of olfactory bulb efferents to layer I GABAergic neurons in the entorhinal area. Combination of anterograde degeneration and immunoelectron microscopy in rat

Glutamic acid decarboxylase (GAD) immunoelectron microscopy in combination with anterograde degeneration was applied in rats to study the synaptic targets of olfactory bulb afferents to the lateral subdivision (LEA) of the entorhinal area (EA). Immunoreactive neurons and terminals are scattered throughout all layers of LEA. After olfactory bulb resection, terminal degeneration occurs in layer Ia of EA. Using the electron microscope we examined serial thin sections of 12 and 14 immunoreactive neurons sampled from layer Ia of the dorsal (DLEA) and ventral (VLEA) subdivisions of LEA, respectively. The morphology of all these neurons is similar: they are small (short axis 5-9 micron, long axis 7-12 micron) and possess eccentrically located, indented nuclei provided with filamentous nuclear rodlets. The immunoreactive neurons have thin, smooth dendrites which usually emerge abruptly from the somata. We observed a single cilium on 5 of the immunoreactive neurons. In layer Ia of both DLEA and VLEA, the somata of the immunoreactive neurons are contacted by degenerating, non-immunoreactive boutons showing asymmetric synaptic junctions. In addition to these boutons, 4 other categories of axo-somatic terminals can be distinguished: normal, non-immunoreactive boutons forming asymmetric synapses and containing spherical synaptic vesicles; normal, non-immunoreactive boutons with symmetric synapses and pleomorphic synaptic vesicles; normal, non-immunoreactive boutons with asymmetric synapses, containing dense-cored vesicles in addition to spherical synaptic vesicles; and normal, immunoreactive boutons with symmetric synapses and pleomorphic synaptic vesicles. It is suggested that the GAD-immunoreactive neurons which receive olfactory bulb input correspond to local circuit neurons with intralaminar axons which innervate each other as well as the distal segments of the apical dendrites of projection neurons with cell bodies in layers II and III. Thus, the olfactory input in EA seems to be wired not only for excitation of layers II and III pyramidal neurons but also for feed-forward inhibition using GABAergic intermediary neurons, strategically located in the area of termination of olfactory bulb fibers.     

Direct observations of synapses between GABA-immunoreactive boutons and identified spinocervical tract neurons in the cat's spinal cord.

Three spinocervical tract neurons in adult cats were physiologically characterized and intracellularly labelled with horseradish peroxidase. The neurons were reconstructed and examined with the light microscope and were prepared for postembedding immunochemical analysis by using an antiserum which specifically recognizes GABA in glutaraldehyde-fixed tissue. Semithin sections were tested and examined with the light microscope. Somata, proximal, and distal dendrites of all three cells were associated with numerous punctate GABA-immunoreactive structures. Immunoreactive perikarya of small neurons in the vicinity of spinocervical tract cells were also observed. Ultrastructural analysis, with the immunogold technique, revealed that somata and proximal dendrites of all three neurons received synaptic contacts (about 37% of total synapses) from GABA-immunoreactive boutons and that distal dendrites were also associated with substantial numbers of immunoreactive structures (about 27% of synapses). Immunoreactive boutons were small (about 1 micron in diameter), contained irregularly shaped agranular vesicles, and formed symmetrical synaptic junctions with identified neurons. An additional group of immunoreactive boutons was observed to be associated with one of the cells only; these contained many large dense-core vesicles in addition to small agranular vesicles. Boutons containing round agranular vesicles and flattened agranular vesicles were not observed to be immunoreactive. The evidence supports the idea that much of the postsynaptic inhibition observed in spinocervical tract neurons is mediated by GABA and that even the most distal dendrites of these neurons receive inhibitory inputs.     

Methods of fixation and the morphology of synaptic vesicles

Cerebellar mossy, parallel, basket and Golgi axons were used to study the variables contributing to vesicle shape. It has been found that, when aldehydes are used as a primary fixative, synaptic vesicles can be either small round vesicles (less than 400 Å in diameter), large round vesicles (400–500 Å in diameter) or flat, with a ratio of large to small diameters of 2 or greater (500–600 Å × 250 Å). The proportion of round vesicles as well as the presence and the number of flat vesicles in each axon depend basically upon the osmotic pressure of the buffer. Concentration of aldehydes, duration of the fixation and dehydration, temperature of the fixative, etc. do not affect the synaptic vesicle shape. With buffers (phosphate and sodium cacodylate) of low osmolarity, mossy and parallel axons exhibit a round synaptic vesicle population and no flat vesicles are seen at all. With the same low osmolarity buffers, basket and Golgi axons have a synaptic vesicle population that is mainly round with a low proportion of flat vesicles. As the osmolarity of the buffers is increased, the number of flat vesicles in basket and Golgi axons increases. In mossy and parallel axons the flat vesicles are present with highest osmolarities but always in a lower proportion than in basket and Golgi axons. Quantitative observations indicate that each axon type, under given conditions of fixation, exhibits a mixed vesicle population with a characteristic vesicle ratio. Although this vesicle ratio is modified significantly with changes in the osmotic pressure of the buffer, each axon type exhibits a different and characteristic modification of its vesicle ratio. These observations show that the vesicle ratio and its changes in different conditions of fixation are useful parameters for classifying axon.