news and views - Nature

© 2000 Nature America Inc. • http://neurosci.nature.com

news and views

Preassembly and transport of nerve terminals: a new concept of axonal transport Jack Roos and Regis B. Kelly


A morphological study suggests a new view of synapse formation by demonstrating that presynaptic elements are assembled into ‘prototerminals’ before being transported down axons by fast, saltatory movement. These structures stop moving when the axon contacts a dendrite.

While trying to pass a prefabricated house being trucked along a major highway, we may find ourselves questioning the usefulness of this approach to construction. If we think for a moment, however, it is easy to appreciate that prefabrication allows a complex house to be erected by sophisticated plumbers and carpenters far from its final destination. A similar insight comes from the paper by Ahmari, Buchanan and Smith1 in this issue of Nature Neuroscience, in which the authors provide evidence that partially prefabricated nerve terminals are transported along axons. The presynaptic nerve terminal is packed with synaptic vesicles, exocytotic machinery and the coat proteins needed for vesicle recycling; it is likely that several hundred protein types are concentrated at the terminal (Fig. 1). Because presynaptic proteins are by definition absent from dendrites and even from nonsynaptic areas of the axon, they are thought to share sorting motifs that allow them to be segregated from other neuronal proteins in the cell body and selectively targeted to subregions of the axon. Much of our thinking about how axonal proteins are sorted is based on parallels with epithelial cell sorting. The two processes may share similar mechanisms because viral proteins that go to the apical region of epithelial cells also go to axons in neurons. Apically targeted proteins are believed to travel in sphingomyelin-rich lipid rafts, which can be distinguished experimentally because they cannot be solubilized under specific detergent conditions. Similarly, axonal proteins can be recovered selectively The authors are in the Department of Biophysics and Biochemistry, University of California, San Francisco, California 94143-0448, USA. email: [email protected] nature neuroscience • volume 3 no 5 • may 2000

cle proteins dotting the axons of neuin lipid rafts, and removal of sphinrons with a periodicity of a few microns. gomyelin affects neuronal development2. By fusing a synaptic vesicle protein, Research on sorting in epithelial cells VAMP/synaptobrevin, to green fluoreshas focused on identifying ‘sorting sigcent protein (GFP), the authors were nals’, peptide domains that target proable to observe these puncta in real teins selectively, for example, to the time, in living cells. They found that basolateral membrane or the transcytoteach punctum could move as fast as ic pathway. If all nerve terminal proteins 0.5 µm per second, close to the speed share an ‘axonal targeting signal’, howexpected for microtubule-based axonal ever, it has so far eluded discovery, hindering any search for the sorting machinery. An alternative model a (Fig. 1b) is that only a few proteins have sorting signals that allow them to associate with axonal motors; in this model, proteins that lack signals are transported ‘piggyback style’ by associating with proteins that have sorting signals. The data of Ahmari et al. 1 Docking and Kinesin Synaptic Calcium LDSV Coating proteins fusion machinery vesicle channel suggest that fresh thinking is needed about the trafficking of b presynaptic components in neurons. Before their work, conventional wisdom suggested that the various elements of the nerve terminal—synaptic vesicle components, presynaptic plasma Microtubule membrane proteins, large densecore synaptic vesicles (LDSVs) Bob Crimi and cytoplasmic machinery for Fig. 1. Transport of prefabricated prototerminals. The exocytosis and endocytosis— simplified nerve terminal has docked synaptic vesicles were packaged into individual adjacent to calcium channels and one LDSV. The active compartments, each of which zone of exocytosis has docking and fusion machinery (triwas recognized by a special angles) and is surrounded by a zone of endocytotic member of the kinesin family, machinery3. Earlier models (a) predicted that the compomotor proteins that carry cargo nents of the nerve terminal (coating proteins, LDSVs, to the distal ends of axonal membranes containing presynaptic membrane proteins microtubules (Fig. 1a). To inves- and synaptic vesicles) are transported down the axon in tigate this model, Ahmari and individual packets, with each packet recognized and transported by a different member of the kinesin family. The colleagues followed up on an old latest model (b) is that a pre-assembled complex of observation from static synaptic vesicle proteins, calcium channels, endocytotic immunofluorescence images of machinery and LDSVs is transported as a unit at a speed neurons in culture: these images close to that predicted for kinesin-mediated, microrevealed puncta of synaptic vesi- tubule-based transport. 415



news and views

minal’ that travels intact along the axon (Fig. 1b). If this model is correct, it has important implications for how synaptic a material is targeted to the terminal. For examb ple, the sorting domains of many transported proteins may be regions that allow interactions between components of AP-3 coat AP-2/clathrin coat the presynaptic termiBob Crimi nal, rather than interacFig. 2. Synaptic vesicle formation along axons and at synaptic contion with kinesin tacts. LDSVs and irregularly shaped vesicles are determined to be part of the prototerminal by electron microscopy. After exocyto- motors. If this is true, sis, the membranes of LDSVs recycle into a special class of synaptic then the requirements vesicles generated from endosomes by an unusual coating mecha- for synapse assembly nism that is sensitive to brefeldin A. This coat uses a novel adaptor may overlap considercomplex, AP-3, which is widely used to facilitate budding from ably with the requireendosomes, but whose recruitment is sensitive to brefeldin A. ments for axonal Stimulus-dependent release of acetylcholine from axons of imma- transport. This model ture frog motoneurons is sensitive to brefeldin A, suggesting that should soon be testable the transmitter-containing vesicles are made by the AP-3 route (a). in Drosophila, where Release from mature synapses is not sensitive to brefeldin A and probably uses AP-2 and clathrin to recycle synaptic vesicles (b). screens for neuromuscuThe data from Ahmari and colleagues1 suggests that the two types lar phenotypes should of transmitter release may be explained as release from immature lead to the identification of many components prototerminals and mature synapses, respectively. required for the assembly of the presynaptic terminal. Stimulation-evoked release of neurotransport. A major surprise was that transmitter and recycling of synaptic each punctum was too large and had too membrane proteins have been observed much fluorescence to be a single synapalong neuronal processes before synapse tic vesicle; indeed, each punctum could formation 5,6 . The discovery of a procontain as many vesicles as a synapse. Even more surprisingly, using doubletoterminal complex, particularly one that label immunohistochemistry, the contains large, dense-core synaptic vesiauthors showed that each motile unit cles, suggests a possible explanation. also contained calcium channels en When large dense secretory vesicles fuse route to the nerve terminal plasma with the plasma membranes of neuroenmembrane and soluble proteins of the docrine cells, most of their membranes nerve terminal’s endocytotic machinery. are recovered in a small endocytotic vesiAhmari and colleagues 1 went a step cle that resembles a synaptic vesicle 7. further and used electron microscopy to These synaptic-like microvesicles analyze the puncta in some of the cells (SLMVs) are unusual because they are that they had imaged dynamically—a made from endosomes by a coating considerable technological feat. Consismechanism quite different from the one tent with their immunofluorescence normally used to make synaptic vesicles experiments, an array of organelles was (in that it is brefeldin A sensitive)8,9. The discovered in the puncta, including link to the work of Ahmari and colLDSVs and irregularly shaped tubular leagues comes from the observation that organelles. Conspicuously lacking were quantal release from a stimulated axon is classical small synaptic vesicles, consistent brefeldin A sensitive6, whereas that from with earlier work suggesting that synapa mature nerve terminal is not. These tic vesicles are assembled in the nerve terobservations can be integrated into a minals and do not exist in the axon4. model in which prototerminals show a developmentally immature form of vesiThe results of the imaging and EM cle recycling from endosomes (Fig. 2a) experiments are not consistent with the that is replaced during synapse formamodel presented in Fig. 1a. Instead, the tion by a more sophisticated, direct authors reasonably conclude that the elemechanism that is brefeldin A insensitive ments of a mature presynaptic ending are (Fig. 2b). The ‘vesicles’ inside the movassembled into a prefabricated ‘prototer416

ing prototerminals could be more heterogeneous in shape because they are SLMVs, not classical synaptic vesicles. As suggested by the authors, this might explain the heterogeneity of quantal size during axonal secretion. If such prototerminals can move rapidly along the axon, there must be a mechanism to stop them when they reach a synaptic contact. Ahmari and colleagues found that when an axon contacted a dendritic projection, the bolus of moving membrane became static. Somehow, adhesion to the dendrite must be triggering a cytoplasmic change within the axon that can be recognized by the prototerminal; when it stops, the presynaptic plasma membrane components and associated fusion machinery are inserted into the plasma membrane. A link between adhesion and membrane insertion was also discovered by Richard Scheller and James Nelson, who found that the site of adhesion between epithelial cells also became a site of exocytosis. In their experiments, the cytoplasmic face of adhesion sites accumulated the proteinaceous ‘exocyst’ complex, which is known from work on yeast to mark sites of membrane fusion10. This finding has been linked to synapse formation; in neurons, exocysts accumulate transiently in axonal varicosities before the appearance of stimulus-dependent synaptic vesicle recycling 11. These results suggest that exocysts might be involved in guiding the prototerminal to adhesion sites at a synaptic contact. Both epithelial adhesion sites and nerve terminals contain adhesion proteins of the cadherin family12 with associated filamentous actin. Sites of exocytosis in yeast are also marked by the presence of actin patches. The link between exocyst accumulation and prototerminal capture seems ripe for further exploration. Earlier studies of axonal transport have assumed steady rates of movement. Observing axonal transport in real time now reveals that the movement of prototerminals can be saltatory, with periods of rest (0.5 to 4.5 min) interspersed between rapid movements (0.5 µm per second). This discovery raises new issues for cell biologists. Does the arrest of prototerminal movement along the axon use the same mechanisms as arrest through contact with a dendritic extension? For example, are the exocysts involved? One attractive explanation for transport arrest comes from studies of melanosome movement14. Melanosomes nature neuroscience • volume 3 no 5 • may 2000



news and views

can switch from making long movements along microtubule highways to making short ones—sometimes stopping altogether—along actin backroads 13,14 , a process also detected in extracts of squid axoplasm 15. The saltatory movements observed in axons could be due to pausing at actin-rich side roads or rest stops along the microtubule highway. Fortunately, predictions such as these will be relatively easy to test with technologies now available; for example, one can compare the distributions of actin and actin motors with that of the prototerminals. The work of Ahmari and colleagues dramatically changes how we should think about neuronal polarity, and it suggests many further experiments. The results

also illustrate that the first technology of cell biologists, merely looking at cells, remains one of the best; this is especially true now that we can follow a single type of fluorescent protein in a living cell. 1. Ahmari, S. E., Buchanan, J. & Smith, S. J. Nat. Neurosci. 3, 445–451 (2000). 2. Ledesma, M. D., Brugger, B., Bunning, C., Wieland, F. T. & Dotti, C. G. EMBO J. 18, 1761–1771 (1999). 3. Roos, J. & Kelly, R. B. Curr. Biol. 9, 1411–1444 (1999). 4. Hirokawa, N. Science 279, 519–526 (1998). 5. Kraszewski, K. et al. J. Neurosci. 15, 4328–4342 (1995). 6. Zakharenko, S., Chang, S., O’Donoghue, M. & Popov, S. V. J. Cell Biol. 144, 507–518 (1999).

Sodium channels develop a tyrosine phosphatase complex Michael W. Salter and Yu Tian Wang A new study shows that sodium channels interact with receptor-like tyrosine phosphatase β, which increases sodium currents and may also help localize the channels. Phosphorylation is a ubiquitous posttranslational mechanism for rapid, reversible modulation of proteins, increasingly recognized as a principal means for regulating the function, expression and localization of ion channels1. Phosphorylation at a particular site is determined by the balance between phosphate addition by protein kinases and its removal by phosphatases. Protein kinases and phosphatases fall into two main types: those that act on serine and threonine residues and those that act on tyrosine residues. We have an extensive understanding of ion channel regulation by numerous serine/threonine protein kinases and phosphatases1. Tyrosine phosphorylation2 also regulates both ligand3 and voltage-gated4 channels. Implicit in these studies is the opposing modulation of ion channel The authors are in the Programme in Brain and Behaviour, University of Toronto, Hospital for Sick Children, 555 University Ave., Toronto, Ontario M5G 1X8, Canada. e-mail: [email protected] or [email protected] nature neuroscience • volume 3 no 5 • may 2000

function by phosphotyrosine phosphatases (PTPs), but the identity of specific enzymes has been elusive. Thus the report by Catterall and colleagues5 fills an important void in our understanding by identifying the first example of a PTP regulating an ion channel. The authors show that voltage-gated sodium channels are associated with and regulated by receptorlike tyrosine phosphatase β (RPTPβ). PTPs comprise a large family of enzymes6, of which more than 80 have been identified7. PTPs are grouped primarily into receptor-like, or transmembrane, PTPs and cytoplasmic PTPs. Receptor-like PTPs have a single membrane-spanning domain, and most contain tandem intracellular PTP domains, with the one closer to the membrane typically having greater or sometimes all the catalytic activity. Extracellular domains are variable in the PTP family, but most contain structural motifs suggesting roles in cell–cell or cell–matrix adhesion. RPTPβ (also known as PTPζ8) is unique in the PTP family in that three forms are generated by alternative splicing: a long

7. Strasser, J. F., Arribas, M., Blagoveshchenskaya, A. D. & Cuter, D. F. Mol. Biol. Cell 10, 2619–2630 (1999). 8. Faundez, V., Horng, J. T. & Kelly, R. B. Cell 93, 423–432 (1998). 9. Shi, G., Faundez, V., Roos, J., Dell’Angelica, E. C. & Kelly, R. B. J. Cell Biol. 143, 947–955 (1998). 10. Grindstaff, K. K. et al. Cell 93, 731–740 (1998). 11. Hazuka, C. D. et al. J. Neurosci. 19, 1324–1334 (1999). 12. Uchida, N., Honjo, Y., Johnson, K. R., Wheelock, M. J. & Takeichi, M. J. Cell Biol. 135, 767–779 (1996). 13. Rodionov, V. I., Hope, A. J., Svitkina, T. M. & Borixy, G. G. Curr. Biol. 8, 165–168 (1998). 14. Rogers, S. L. & Gelfand, V. I. Curr. Biol. 8, 161–165 (1998). 15. Kuznetsov, S. L., Langford, G. M. & Weiss, D. G. Nature 356, 722–725 (1992).

and short transmembrane form, both of which contain the catalytic domains, and a secreted form of the protein, which lacks the transmembrane and intracellular portions. The secreted form and the long catalytic splice form have a large extracellular insert rich in glycosaminoglycan side chains, rendering both proteins chrondroitin sulfate proteoglycans. For clarity, only splice variants containing phosphatase domains are referred to as RPTPβ; the secreted form is commonly known as phosphacan, named as an abundant brain proteoglycan9. Voltage-gated sodium channels are fundamental for neural functioning, as they mediate the sodium current responsible for the rapid rising phase of action potentials10. In the brain, sodium channels are composed of three glycoprotein subunits: the α subunit, which forms the channel pore and is sufficient to produce a functional channel on its own, and the β1 and β2 subunits, which modulate channel gating. Like other ion channels, sodium channel function is modulated by serine/threonine11 and tyrosine kinases12. In this issue5, Catterall and colleagues5 show that sodium channels isolated from rat brain interact with RPTPβ extracellularly and intracellularly. The extracellular portion of RPTPβ has two prominent amino (N)-terminal domains: a carbonic anhydrase-homology (CAH) domain and a fibronectin type III (FN) repeat. Catterall and colleagues found in experiments with fragments of the extracellular region of RPTPβ that the CAH domain but not the FN domain binds to the extracellular region of the sodium channel. Thus, sodium channels may act as an extracellular ligand for the receptor por417