Synaptic vesicle biogenesis, docking and fusion: a molecular description.

Calakos, Nicole, and Richard H. Scheller. Synaptic Vesicle Biogenesis, Docking, and Fusion: A Molecular Description. Physiol. Rev. 76: 1-29, 1996. -- Secretion of neurotransmitter is the primary means of intercellular communication within the nervous system. This process is regulated by a highly orchestrated cycle of membrane trafficking within the presynaptic nerve terminal. Characterization of proteins localized to the synaptic vesicle and the subsequent studies of their properties have led to a model for the biochemical pathway that underlies vesicle docking, activation, and fusion. The proteins found to function in the synapse are related to those in yeast and other organisms, demonstrating that the mechanisms that mediate vesicle trafficking are conserved in all eukaryotic species.

I. INTRODUCTION

Neurons communicate with their targets through the regulated release of chemical neurotransmitters. In 1897, Sherrington (332) coined the term synapse to describe the site of this communication. The concept of the synapse fit well with Ramon y Cajal's (63) view of neurons as individual cellular units with pre- and postsynaptic sites, which he developed from detailed observations of neuronal morphology. In 1921, the chemical nature of neurotransmission between a nerve and its target (the vagus nerve and heart) was definitively demonstrated by Loewi (229). Important insights into the nature of neurotransmitter release accumulated rapidly during the 1950s and 1960s due largely to the efforts of Katz and co-workers (119, 194). Their extensive work on the frog neuromuscular junction established that the release of neurotransmitter occurs in discrete units called quanta (199,194). In 1956, with the advent of electron microscopy, visualization of the abundant vesicular structures termed synaptic vesicles in presynaptic terminals led to the "vesicle hypothesis," which proposed that each vesicle contained a single quantum of neurotransmitter (101). During the 1970s, evidence for the vesicular nature of neurotransmitter release was solidified using freeze-fracture electron microscopy to capture images of synaptic vesicles in the process of exocytosis (69, 171).

Recently, the biochemical purification of synaptic vesicles in large quantities has greatly facilitated the identification of synaptic vesicle proteins. The cDNAs encoding synaptic vesicle proteins and additional synaptic vesicle-associated proteins have been cloned. This has enabled genetic and biochemical studies that have uncovered the roles of individual proteins in neurotransmission. The characterization of interactions between synaptic proteins has revealed a number of protein complexes that are thought to mediate neurotransmission. The focus of this review is our current understanding of the molecular mechanisms by which synaptic vesicles release their contents at the synapse. After an action potential, calcium influx into presynaptic terminals triggers synaptic vesicle exocytosis and subsequent release of neurotransmitters into the synaptic cleft. The neurotranmitters traverse the synaptic cleft, bind receptors on the postsynaptic cell, and trigger a postsynaptic response. The molecular basis of neurotransmitter release is of interest not only because it is the predominant mode of neuronal intercellular communication but also because it may serve as a key site for the modulation of synaptic transmission efficacy. Changes in synaptic efficacy are thought to underlie the processes of learning and memory.

In addition, the molecular mechanisms underlying intracellular membrane trafficking events like synaptic vesicle exocytosis are thought to be conserved. Through genetic and biochemical approaches, homologues of many of the synaptic proteins have been identified in a diversity of organisms. In yeast, mutations in the genes encoding these homologues produce phenotypes with defects in the constitutive secretory pathway (35, 122). Neurotransmitter release by synaptic vesicles, while utilizing the common mechanisms of vesicle docking and fusion, represents a unique vesicular trafficking event in that it is activated by calcium influx, the kinetics of activation are extremely rapid, and it may undergo modulation. The identification of the molecular machinery that imparts these unique characteristics to synaptic vesicle release represents an important challenge for future studies.

There are three types of synaptic vesicles that are distinguished morphologically: small synaptic vesicles (SSVs), catecholamine-containing small dense-core vesicles (SDCVs), and neuropeptide-containing large dense-core vesicles (LDCVs) The SSVs contain the classical neurotransmitters [acetylcholine, glycine, [Gamma]-aminobutyric acid (GABA), and glutamate] and when visualized by electron microscopy are electron lucent with a uniform diameter of 50 nm. Dense-core vesicles (DCVs), so named by their high electron density under electron microscopy, have larger diameters (80-200 nm) with the neuropeptide-containing vesicles (LDCVs) being larger than the catecholamine-containing vesicles (SDCVs). Biogenesis of SSVs and DCVs significantly differs (reviewed in Refs. 25, 302). Dense-core vesicles are loaded with neuropeptides at the trans-Golgi network (TGN) in the cell body and must be replenished by synthesis of new vesicles at this site. Although the membrane proteins of SSVs are synthesized in the cell body, SSVs are loaded with neurotransmitter in the nerve terminal and are recycled locally.

Many proteins implicated in synaptic vesicle docking and fusion are common to SSVs and DCVs. However, differences in docking and the Idnetics of release distinguish SSVs from DCVs. For example, in electron microscopic studies, DCVs are not seen docked at specialized release sites called "active zones" as are SSVs. Morphologically, "docked" synaptic vesicles are defined as those that are directly apposed to the presynaptic plasma membrane. Docked vesicles are thought to be ready for immediate release upon calcium-dependent stimulation. Docking may also be described functionally as the process that positions a vesicle adjacent to the appropriate target membrane with which it is to fuse. By this definition, all intracellular membrane trafficking events have a docking step. Small synaptic vesicle secretion is therefore specialized due to the accumulation of vesicles in the docked state. The ability of SSVs to accumulate in a "release ready" docked state may underlie the rapid exocytotic response of SSVs to a secretion stimulus. Small synaptic vesicles release their contents within 200 [Mu]s after calcium influx, whereas the DCVS release orders of magnitude more slowly (50 ms) (for examples, see Refs. 4, 79, 226, 360). A molecular difference in the release machinery of SSVs is supported by the observation that in the frog neuromuscular junction, [Alpha]-latrotoxin induces extensive exocytosis of cholinergic synaptic vesicles but not the dense-core peptidergic vesicles (237). Thus, while SSVs and DCVs may utilize common proteins for docking and fusion, these vesicle types have unique regulatory mechanisms of this common process.

This review primarily focuses on the SSVs. Small synaptic vesicles were the source for purification of many of the proteins now implicated in synaptic vesicle docking and fusion. An important area for future studies is the molecular basis of the differences in regulation of docking and fusion between SSVS and DCVs.

II. SYNAPTIC VESICLE BIOGENESIS

The nature of synaptic vesicle biogenesis and recycling is still a subject of interesting debate (reviewed in Refs. 25, 196, 302). The following summary represents a model currently supported by the majority of data (Fig. 1). Synaptic vesicle proteins are synthesized in the cell body and transported to the nerve terminal in transport vesicles (Fig. 1, Ia and Ib). These transport vesicles may undergo rounds of constitutive fusion and endocytotic recycling in the cell body and along the axon. The synaptic vesicle membrane proteins may undergo additional cycles of constitutive exo- and endocytosis in the nerve terminal before they are segregated into mature synaptic vesicles (Fig. 1, Ic). Vesicles are loaded with neurotransmitter within the nerve terminal via vesicular neurotransmitter transporters (Fig. 1, IIa). In the nerve terminal, clusters of synaptic vesicles a short distance from the plasma membrane appear to be associated with the cytoskeleton, perhaps via synapsin in a "reserve pool" (Fig. 1, IIb) (145, 148, 204). A small portion of vesicles is docked at release sites on the presynaptic plasma membrane (the active zone) (Fig. 1, IIIa) (83). Docked vesicles are thought to represent the pool of vesicles immediately available for fusion with the plasma membrane after stimulation. Recycling of synaptic vesicles occurs locally at the synapse to replenish the vesicle pool (Fig. 1, IIIb and IIIc). In addition, anterogradely transported synaptic vesicle precursors help to maintain adequate pools. Synaptic vesicle proteins may eventually be retrogradely transported along the axon and degraded (Fig. 1, IV).

A. Protein Synthesis

Most synaptic vesicle membrane proteins are presumably translated similarly to other transmembrane proteins of their type by utilizing signal receptor protein (SRP) and translocation proteins for targeting and insertion into the endoplasmic reticulum (ER) during protein synthesis. However, the synaptic vesicle protein VAMP (vesicle-associated membrane protein), also known as synaptobrevin, and the plasma membrane protein syntaxin pose an interesting problem for the usual cotranslational mechanism of translocation to the ER. The membrane anchor of these proteins is comprised of amino acid residues at the extreme COOH-terminal. Thus synthesis of the protein would be completed by the ribosome before the transmembrane domain was recognizable by translocation machinery. This type of protein may be inserted posttranslationally into membranes (164, 203). Studies of VAMP synthesis in vivo and expression experiments in vitro determined that VAMP was indeed inserted into membranes posttranslationally independent of the SRP and translocation machinery (203). Vesicle-associated membrane protein was not directly inserted into its target organelle, the synaptic vesicle, but was detected in ER membranes and observed to be routed through the Golgi apparatus. Although the membrane insertion of VAMP is SRP independent, it does appear to rely on a proteinaceous component and ATP hydrolysis (203). It will be interesting to further characterize the machinery responsible for its membrane insertion.

The steady-state intracellular localization of many synaptic vesicle proteins is concentrated at synaptic varicosities. Golgi complex-localized immunostaining is also apparent for many synaptic vesicle proteins, suggesting that they are routed through this complex during biosynthesis (28, 126, 186,, 258, 260, 362). Synapsins and rab3A are notable exceptions in that they do not localize to the Golgi complex and are thought to associate with membranes at a later stage (126, 238). The recently identified cysteine string proteins may also associate with membranes at a later stage because they are thought to bind to membranes via fatty acylation of their cysteine residues (150). There are two synaptic plasma membrane proteins that are thought to associate with membranes via fatty acylation of cysteine residues: 25-kDa soluble N-ethylmaleimide-sensitive factor (NSF) attachment protein (SNAP-25) and 43-kDa guanosinetriphosphatase (GTPase)-activating protein (GAP-43) (167, 340).

B. Localization

There are two possible sites for formation of the synaptic vesicle organelle: in the cell body (for example, by selective budding from the TGN) and in the nerve terminal (for example, by selective budding off of early endosomes or the plasma membrane). To date, the majority of data support the latter hypothesis, although this is an important issue that deserves further clarification.

Some of the earliest support for the idea that SSVs are formed in the nerve terminal comes from an axonal ligature experiment (370). The anterogradely transported vesicles that accumulated on the proximal side of the ligature did not have the uniform appearance or diameter characteristic of SSVs. Synaptic vesicle biogenesis has been proposed to diverge from the constitutive secretion pathway (66, 189, 215, 303). In this scenario, synaptic vesicle membrane proteins follow the constitutive secretory pathway in transport vesicles from the TGN to the plasma membrane. After fusion with the plasma membrane, they are proposed to be internalized with other markers of receptor-mediated endocytosis. Synaptic vesicle membrane proteins would then be sorted away from the endocytotic pathway, possibly at the level of the early endosome, to become mature synaptic vesicles. Evidence for an involvement of the constitutive secretory pathway in synaptic vesicle biogenesis has come from studies that have identified synaptic vesicle membrane proteins in endocytotic vesicles. The synaptic vesicle membrane protein synaptophysin is found in such endocytotic vesicles in both neuroendocrine cells where it is endogenously expressed and in nonneuronal cells after transfection with synaptophysin cDNA (66, 189, 215). However, these studies are limited because they analyze the steady-state distribution of synaptophysin. In another study, synaptophysin localization was followed after biosynthetically labeling with [35.sup.S]sulfate (303). Synaptophysin was separated from a marker of secretory granules after exiting the TGN and comigrated with a marker of constitutive secretory vesicles in PC-12 cells (303). Eventually, synaptophysin could be detected at the cell surface in the absence of stimulating regulated secretion. Surface-labeled synaptophysin comigrated with an early endosome population before finally being isolated in synaptic-like microvesicles (SLMVs). The delay between the period when synaptophysin was first seen on the plasma membrane and when it could be detected in SLMVs was sufficiently long (3 h) to suggest that it may cycle multiple times between the plasma membrane and early endosome. Although these data are from neuroendocrine cells, SLMVs share a number of molecular and physical properties with synaptic vesicles (302, 361).

In neurons, there is evidence that synaptic vesicle proteins participate in constitutive cycles of exo- and endocytosis. An antibody to the luminal domain of the synaptic vesicle protein synaptotagmin was used to monitor exposure of synaptotagmin to the cell surface (239). The antibody-labeled synaptotagmin appears to undergo rounds of exo- and endocytosis in the absence of stimulation or calcium in the dendrites and axons of developing cultured hippocampal neurons. During synaptogenesis, the synaptotagmin antibody staining redistributes to varicosities from being uniform along the axon. Restriction of other synaptic vesicle antigens to synaptic sites during synaptogenesis has also been reported (231). During the formation of synapses, a regulatory switch must redirect membrane flow in the nerve terminal from constitutive addition to the growing axon to local recycling that forms mature synaptic vesicles capable of docking and calcium-regulated exocytosis (404, 419). Perhaps the calcium-independent cycles of exo- and endocytosis observed before synaptogenesis in developing hippocampal neurons (239) and neuromuscular junctions (404) reflect the early stages of synaptic vesicle biogenesis in mature neurons. An important area for future research is to address the location at which synaptic vesicle proteins are finally segregated from other membrane proteins in this exo-/endocytotic cycle: the plasma membrane or the early endosome?

C. Axonal Transport

The anterograde movement of synaptic vesicles occurs via fast axonal transport (reviewed in Refs. 89, 173). Several groups have provided evidence that kinesin is required for the anterograde movement of membranous organelles (90, 174, 254, 255, 327). However, in Drosophila kinesin heavy chain mutants, the nerve terminals are not depleted of synaptic vesicles, suggesting that kinesin may not be the motor responsible for the transport of this organelle (142).

The kinesin homologue unc-104 in Caenorhabditis elegans is a more likely candidate for the molecule that mediates synaptic vesicle transport (153). The unc-104 mutant phenotype has fewer synapses, the nerve terminals have fewer synaptic vesicles, and there is an abundance of small 30- to 50-nm vesicles in the cell body. Knowledge of the nature of the accumulated vesicles in the cell body of these mutants win help in determining the biosynthetic origin of synaptic vesicles. Notably, other membrane-bound organelles were not affected in the unc-104 mutants, including the secretory granules in endocrine cells. There may be multiple kinesin-like molecules within a cell that each direct traffic of distinct organelles

In support of this hypothesis, many new kinesin-like molecules have been identified, several of which are brain specific (5, 173, 200, 264, 285). It will be interesting to determine whether each of these motor proteins has a distinct trafficking role in the neuron and to what degree functional redundancy comes into play.

D. Protein Sorting

During synaptic vesicle biogenesis, there must be mechanisms for determining which proteins are synaptic vesicle constituents and which are to be excluded from this organelle. To date, comparisons of the primary amino acid sequences do not reveal a common signal that might be used for this purpose. Synaptic vesicle membrane proteins associate as a high-molecular-weight multimeric complex in detergent-solubilized synaptic vesicles (32). This complex could represent a functionally important microdomain of membrane proteins that aids in the segregation of synaptic vesicle proteins either from endosomes, the plasma membrane, or the TGN.

Several studies identify protein domains needed for correct localization of synaptic proteins. Experiments with transgenic mice demonstrate that an N[H.sub.2]-terminal domain of synapsin Ib is sufficient for its targeting (139). In Drosophila synaptotagmin mutants, a truncated synaptotagmin protein which lacks the COOH-terminal C2 domain is targeted to the synapse-rich neuropil of the spinal cord as it is in wild-type flies, suggesting that this domain is not essential for its localization (105). Studies of the biosynthesis of VAMP indicate that the transmembrane domain does not direct VAMP to its target organelle, suggesting that a targeting signal exists in its cytoplasmic domain (203). It is of interest that studies of Drosophila sed5 (the mammalian syntaxin 5 homologue implicate both membrane and cytoplasmic domains in the targeting of this protein to Golgi membranes (19). Sed5 is thought to be the target membrane receptor for transport vesicle docking and, like VAMP, has a COOH-terminal membrane anchor.

Many synaptic vesicle proteins have been expressed in nonneuronal cells to investigate targeting signals. The subcellular targeting in this exogenous environment may provide clues to the sorting signals of synaptic vesicle proteins. However, neuron-specific sorting machinery may drastically alter any targeting patterns observed in these nonneuronal cells. Many neuron-specific clathrin coat and adaptor proteins have been identified (292). These molecules are good candidates for the specific recognition of synaptic vesicle constituents.

Three synaptic vesicle proteins, synaptophysin, synaptotagmin, and SV2, have been expressed in nonneuronal cells to study their subcellular targeting (121). Synaptophysin colocalized with endocytotic vesicles containing the transferrin receptor as has been previously observed (66, 189, 215, 303; however, see also Ref. 208 for contrasting localization in a hepatoma cell line). A COOH-terminal domain of synaptophysin was essential for this localization (214). In contrast to the localization of synaptophysin, a significant fraction of synaptotagmin was localized to the plasma membrane. Synaptotagmin has been shown to bind the adaptor protein AP-2 with high affinity in in vitro studies (412). An interaction with AP-2 might mediate its endocytosis into the synaptic vesicle pool in neurons, while this interaction is absent in Chinese hamster ovary (CHO) cells. The synaptic vesicle protein SV2 partially localized to a vesicular population that was not endocytotic. Combinations of the three synaptic vesicle proteins were coexpressed in CHO cells to determine whether the coexpression could alter their targeting. The three proteins all targeted similarly to their individual expression patterns, and a new vesicle population was not seen.

E. Neurotransmitter Loading

In the last few years, the neurotransmitter transporter proteins for catecholamines and acetylcholine have been characterized and the cDNAs encoding them have been cloned. Neurotransmitter transporters are exciting potential targets for drugs that may be useful in neuropsychiatric disorders, for example. Neurotransmitter uptake proteins on the plasma membrane of cells are important for terminating the postsynaptic response (reviewed in Ref. 326). This section focuses on synaptic vesicle membrane transporters whose function is to load the vesicle with neurotransmitter to be released during exocytosis (112). This step is required during synaptic vesicle biogenesis, but is also crucial for the reloading of recycled synaptic vesicles.

The proton pump resident in the synaptic vesicle is an essential part of the machinery required to load vesicles with neurotransmitter (reviewed in Ref. 262). With the use of the energy derived from ATP hydrolysis, this multisubunit complex creates an acidic lumen and electrochemical gradient. Vesicular chloride channels may dissipate some fraction of the membrane potential (298, 403). There are four classes of neurotransmitter with respect to uptake properties. biogenic amines, acetylcholine, glutamate, and GABA/glycine. The bioenergetic requirements for the transport of each of these classes are reviewed in detail elsewhere (242). Briefly, glutamate uptake is driven largely by the membrane potential difference; amine and acetylcholine uptake requires the pH gradient; and GABA uptake requires both the pH and potential gradients.

The cDNAs encoding several vesicular neurotransmitter transporters have been cloned. A strategy using N-methyl-4-phenylpyridinium toxicity resulted in the cloning of cDNAs encoding vesicular biogenic amine transporters (VMAT-1 and VMAT-2) (221, 222). The cDNA for an amine transporter was also cloned using a genetic strategy (116). In C. elegans, unc-17 mutants have a paralyzed phenotype, and the gene is predicted to encode a 12-transmembrane domain protein typical of transporters (11). Additional cDNAs that are predicted to encode proteins with 48-50% amino acid identity to unc-17 have been cloned from Torpedo and rat (311, 380). A rat cDNA, isolated by virtue of its homology to the Torpedo clone, has been functionally identified as encoding a vesicular acetylcholine transporter (117). In this study, the cDNA encoding the human homologue (94% identical to rat) was also identified.

The synaptic vesicle protein SV2 has long been characterized as a synaptic vesicle antigen (56, 127). The SV2 CDNA predicts a protein structure containing 12 transmembrane domains (18, 120). The structure and sequence of SV2 are most homologous to a class of bacterial transport proteins that cotransports protons with sugars, citrates, or drugs (18, 165, 166). Apart from this homology, there is no further indication of the substance which SV2 might transport, nor can the direction of transport (into or out of the vesicle) be inferred from the protein sequence of SV2. Future studies are needed to determine whether SV2 transports a neurotransmitter or another substance.

F. Cytoskeletal Association

Many demands for speed, reliability, and regeneration are placed on the transmitting system of the neuron. One way that neurons are thought to accomplish this specialization is in the clustering of synaptic vesicles near release sites. While some vesicles are docked at the active zone, many more are seen in close proximity in a meshwork of filaments. This population of vesicles has been termed the reserve pool (148). The reserve pool of vesicles is thought to represent vesicles being held nearby, ready for docking at the release sites as they are needed.

There is abundant evidence that the synaptic vesicle protein synapsin I serves as a reversible link between synaptic vesicles and this cytoskeletal meshwork (reviewed in Refs. 148, 378). For example, phosphorylation of synapsin I decreases its association with actin filaments and synaptic vesicles (17, 325, 337, 363). In addition, dephosphorylated synapsin I inhibits neurotransmitter release when injected into the squid giant synapse (225). These and other studies suggest that the phosphorylation of synapsin I facilitates neurotransmitter release by dissociating synaptic vesicles from the cytoskeleton. Calcium/ calmodulin-dependent protein kinase II (Ca/CaMKII) is both the kinase that modulates synapsin I activity by phosphorylation and the synaptic vesicle attachment site for synapsin I (31, 325).

Despite the working hypothesis that synapsin I has a unique role in tethering and releasing synaptic vesicles from the cytoskeleton, mouse knockouts of synapsin I appeared normal and morphological defects in their nervous system were not observed (312). The only identified functional difference in neurotransmission was an increase in paired-pulse facilitation, a mechanism of short-term plasticity. Because synapsin II was not altered in the synapsin I knockouts, a partial functional redundancy among synapsins might explain the mild phenotype of the mice. Synapsin II also binds synaptic vesicles and actin filaments (76, 355). Although synapsin II does not have a phosphorylation site for Ca/CaMKII, both synapsins I and II have sites for Ca/CaMKI and adenosine 3',5'-cyclic monophosphate Kinase (148). The functional importance of these phosphorylation sites on either synapsin I or II is not known. The defect in plasticity might reflect a specific Ca/CaMKII modulation-sensitive role of synapsin I that does not overlap with synapsin II. As with many of the synaptic vesicle proteins, careful analysis of the role of multiple isoforms is necessary.

In addition to synapsin-mediated release of synaptic vesicles from the filamentous cytoskeleton, disassembly of actin filaments may also facilitate vesicle exocytosis (37, 74, 240, 382). Studies in chromaffin cells and pancreatic acinar cells show that the filamentous actin cytoskeleton inhibits vesicular access to the plasma membrane and exocytosis (257, 382). There are several actin-depolymerizing proteins that could mediate the filament disassembly, including gelsolin, scinderin, and actin-depolymerizing factor (257, 367). It will be intriguing to determine the physiological regulation of this process. Synaptic vesicle availability at the active zone could thus be determined by two processes: synapsin-mediated release from the cytoskeleton and filamentous actin disassembly.

G. Endocytosis and Local Recycling

1. Location

Both coated and uncoated endocytotic structures have been described in nerve terminals. With the use of electron microscopic analysis, the most abundant endocytotic structures have been noted lateral to the active zone, membrane rather than coincident with this region (168-170, 191, 247). This observation suggests that synaptic vesicle membrane proteins must diffuse away from the exocytotic release zone before being internalized. Some endocytotic structures were observed in the active zone, although these structures were generally uncoated invaginations (70, 71, 247). In electron microscopic studies, it can be difficult to determine whether an uncoated invagination represents an exocytotic or endocytotic process. For further analysis, the reader is referred to excellent reviews of…