commentary

Synaptic vesicle retrieval: still time for a kiss Ole Kjaerulff, Patrik Verstreken and Hugo J. Bellen

After exocytosis, synaptic vesicles are reformed by slow clathrin-mediated endocytosis. However, evidence also supports the existence of faster retreival mechanisms in neurons, including ‘kiss-andrun’, where vesicles fuse only partially with the presynaptic membrane before being retrieved. New insights in synaptic vesicle dynamics have been obtained from vesicle imaging and from studies with mutant animals. Recently, measurements of capacitance changes induced by the fusion of single synaptic vesicles in synapses corroborate the hypothesis that kiss-and-run operates in neurons. Here, we review the evidence supporting fast vesicle retrieval and evaluate its role in neurotransmitter release. eurons communicate predominantly using chemical synaptic transmission. At the active zone, synaptic vesicles fuse with the presynaptic membrane and release their neurotransmitter into the synaptic cleft. The merging of vesicular and presynaptic membranes establishes a fusion pore that connects the vesicle lumen with the extracellular space. The pore expands rapidly, forcing the vesicle to collapse into the plasma membrane (PM). Such complete fusion can be followed by conventional clathrin-dependent endocytosis near the active zone (reviewed in ref. 1). However, this process, lasting several tens of seconds or even minutes2, is too slow to balance the quick consumption of vesicles at synaptic terminals. Accordingly, evidence of additional faster vesicular cycling pathways has accumulated3. One model of rapid recycling invokes complete fusion. However, in contrast to conventional slow endocytosis, the collapsed vesicular membrane patch is retrieved from the presynaptic membrane within a few seconds (Fig. 1a, red pathway). In another variant of rapid recycling — known as ‘kiss-and-run’ — the fusion pore expands only transiently4 (Fig. 1a, blue pathway). Moreover, the vesicle membrane does not collapse into the PM, but withdraws from it after closure of the fusion pore. Here, we discuss data in favour of rapid vesicular trafficking at the synapse, with special emphasis on kiss-and-run release. Kiss-and-run has previously been documented for large dense-core granules of non-neuronal secretory cells (reviewed in ref. 5). Evanescent wave microscopy of insulin-secreting cells expressing the membrane protein phogrin fused to green fluorescent protein (GFP) demonstrated that some secretory granules preserve their vesicular shape, even after complete cargo release6. Moreover, in some secretory cell types with large secretory granules, fusion and re-uptake of individual granules can be resolved by capacitance measurements. Combining this technique with carbonfibre amperometry to record secretion of vesicle content, it was demonstrated that

N

a

Rapid recycling Short membrane residence time

2

2 2

1

2

1

FM1-43

FM2-10 b

Clathrin-mediated endocytosis Long membrance residence time

1

3

Dye with slow 'off time'

Vesicle membrane

Dye with fast 'off time'

Pre-synaptic membrane

Neurotransmitter

Figure 1 FM dye properties in the unloading paradigm. Step 1, formation of the fusion pore, fusion and release of neurotransmitter; step 2, rapid vesicle recycling, including vesicle collapse and quick re-internalization of the vesicular membrane (red arrowheads) and kiss and run (blue arrowhead); step 3, slower conventional vesicle internalization (clathrin-mediated endocytosis). a, FM dye unloading. The ‘membrane residence time’ is the delay between exocytotic externalization of vesicular membrane and subsequent re-internalization. Vesicles loaded with FM dye with a long off time (FM1-43) require several rounds of fusion to fully unload during rapid cycling (left, green). In contrast, a dye whose off time is shorter than the membrane residence time (FM2-10) will need fewer rounds for unloading12,13 (right, orange). The fusion pore may present a structural barrier for FM dye loading and unloading35 (not illustrated). b, Pink stars represent FM1-43 or FM2-10. When the membrane residence time is longer than the off time of both FM1-43 and FM2-10, both dyes dissociate fully from the membrane in one round of fusion. Hence, the differences in off times between FM1-43 and FM2-10 will not influence the unloading rate, which will be similar for the two dyes.

NATURE CELL BIOLOGY VOL 4 NOVEMBER 2002 www.nature.com/naturecellbiology © 2002 Nature Publishing Group

commentary kiss-and-run release occurs in parallel with conventional exocytosis7–9. Evaluating the existence of kiss-and-run at nerve terminals has been more difficult. However, in our opinion, a picture is emerging where kiss-and-run at the synapse is a reality and shares the burden of chemical neurotransmission with other forms of transmitter release. Rapid recycling in hippocampal neurons The study of synaptic vesicle trafficking has been greatly enhanced by the use of ‘FM’ dyes, that is, amphipathic molecules whose fluorescence increases strongly on membrane binding, but which cannot traverse lipid bilayers10,11. When vesicle fusion is evoked by stimulation of nerve terminals in the presence of extracellular FM dye (dye loading), the ensuing endocytosis traps dye in the reforming vesicles, selectively labelling them. Renewed stimulation then allows the trapped dye to escape (unloading). It is useful to define the membrane residence time, which is the delay between vesicle membrane externalization at fusion and subsequent re-internalization of the same membrane patch by endocytosis. A particularly interesting case arises when the residence time of the vesicular membrane is shorter than the time required for the FM dye to dissociate from the lipid bilayer (the ‘off time’). The dye will then not have enough time to fully dissociate from the externalized membrane patch before reinternalization (Fig. 1a, left, ‘rapid recycling’). Consequently, some dye molecules must await subsequent vesicle fusion(s) to escape, so that unloading of dye from the nerve terminal is slowed down. This effect is more pronounced for more ‘sticky’ FM dye variants, that is, variants that departition slowly from lipid membranes (for example, FM1-43; Fig. 1a, left), than for the more hydrophilic variants, which departition faster (for example, FM2-10; Fig. 1a, right). Such differential unloading rates were observed experimentally at hippocampal terminals12,13 and it was concluded that fast vesicle retrieval — lasting less than a couple of seconds — operates at this synapse12. In contrast, slow endocytosis should produce similar unloading rates for dyes of different off times, as the longer membrane residence time allows dye bound to externalized membrane to fully dissociate in a single round of fusion (Fig. 1b, ‘clathrin-mediated endocytosis’; however, see also ref. 14). Rapid reuse of synaptic vesicles Another study further documented the rapid recycling of synaptic vesicles13. At most synapses, the synaptic vesicle pool can be divided into a readily releasable pool (RRP) and a reserve pool (RP). During E246

stimulation, the RRP is released quickly, whereas the RP is released later. In this study, the authors subjected the RRP to a dual-stimulus protocol while simultaneously monitoring the evoked postsynaptic currents and the unloading of FM2-10 (ref. 13). With brief stimulus intervals, the amount of transmitter released was similar during the first and second stimuli. In contrast, dye unloading was severely reduced during the second stimulus when compared with the first. The authors concluded that recently fused RRP vesicles are rapidly (time constant ~1 s) retrieved to the RRP to immediately fuse again. This rapid reuse of RRP vesicles explains why FM2-10 dye unloading was impaired during the second stimulus: the reused vesicles had already unloaded most or all of their dye in response to the first stimulus. Further evidence for the rapid reuse of RRP vesicles was provided by monitoring postsynaptic currents evoked by prolonged intense stimulation of cultured hippocampal terminals (up to 5 min)15. The postsynaptic response was reduced to 10–40% of the initial level after the first 40 s of stimulation. Interestingly, release remained at a constant level for the remaining part of the stimulation period. The plateau level of release was significant, even at immature synapses. This is surprising, as the RP is less well developed at these synapses and cannot contribute substantially to vesicle turnover. Thus, one obvious question concerns how release is sustained. It was proposed that the rapid reuse of RRP vesicles helps sustain prolonged neurotransmission15. Hence, neurons rapidly reuse vesicles during both early and late phases of long-term stimulation. A role for kiss-and-run in rapid recycling The above data suggest that rapid vesicular recycling is active at hippocampal terminals. Another study provided evidence that the underlying mechanism is kiss-and-run, rather than full fusion before rapid re-internalization of membrane16. During exocytosis at hippocampal terminals, a fraction of the synaptic vesicles fail to exchange FM dye between the vesicular lumen and the extracellular milieu. These vesicles, although unable to exchange dye, still release neurotransmitter. The authors estimated that kiss-and-run is responsible for 10–20% of the transmitter release evoked by action potentials16. Genetic studies of endocytic proteins In view of the above data, the case for rapid recycling, including kiss-and-run, seems strong at hippocampal terminals (however, see also ref. 17). But these central synapses contain a small number of vesicles (less than 200 per bouton), and calculations of vesicular supply and demand suggest that

they may be forced to use unconventional methods of vesicle trafficking13,18. Hence, it is also interesting to consider the situation in larger synapses, such as the neuromuscular junction (NMJ), which may contain up to 10,000 vesicles. Most recent investigations of vesicle cycling in the Xenopus laevis NMJ have neither confirmed nor refuted the existence of rapid recycling of the kind documented in hippocampal terminals19,20. However, in vivo genetic studies suggest that very fast vesicle retrieval is involved at Drosophila melanogaster NMJs. At low temperature (18 °C), the temperature-sensitive dynamin mutant shibirets undergoes normal vesicle recycling. In contrast, dynamin function is blocked and vesicle recycling abolished at high temperature (30 °C). After stimulation, loss of dynamin function causes a decline, and ultimately a complete failure, of the postsynaptic response. It was observed that in shibirets flies at high temperature, the postsynaptic response declined faster than controls after 20 ms21. This is much too fast to be explained by a defect in clathrin-mediated endocytosis and suggests that dynamin maintains the RRP vesicles by rapid retrieval21. The existence of rapid recycling at the Drosophila NMJ was recently corroborated by the discovery that loss of endophilin produces a very different phenotype than dynamin mutations22. Endophilin, in common with dynamin, is involved in clathrinmediated endocytosis. Biochemical assays and injection of reagents that perturb the binding of endophilin to its molecular partners indicate that endophilin is involved at multiple steps during conventional slow endocytosis23–25. Accordingly, a block in clathrin-mediated endocytosis was observed in Drosophila larvae lacking endophilin. This endocytic block results in enlarged boutons that are severely depleted of synaptic vesicles. However, synaptic vesicle clusters remain at active zones in the mutant terminals22. Electrophysiological recordings at NMJ boutons in endophilin mutants demonstrated that a lack of clathrin-mediated endocytosis does not affect neurotransmitter release at low levels of synaptic activity22. But when stimulated at high frequency, endophilin-null NMJs undergo a strong synaptic depression to 15–20% of the initial level (Fig. 2a, red). This depression is more pronounced than in controls (Fig. 2a, blue), highlighting the importance of clathrinmediated recycling. However, neurotransmitter release in response to intense stimulation is not completely abolished at the endophilin mutant terminals. Instead, neurotransmission remained at the 15–20% level throughout the stimulus, similar to the depression-and-plateau response previously observed at hippocampal terminals15. The vesicles in endophilin mutants are

NATURE CELL BIOLOGY VOL 4 NOVEMBER 2002 www.nature.com/naturecellbiology © 2002 Nature Publishing Group

commentary EJP amplitude (percentage of maximal response)

a 100

Control

80

Endo

60

Shibirets

40 20 0 0

200

400

600

Time of stimulation at 10 Hz (s)

b

c 100 Down-step (aF)

50 aF

0.5 s

50

0 0.5 s

0.5 s

0

50

100

Up-step (aF)

Figure 2 Kiss and run in neurons. a, Postsynaptic response to motor nerve stimulation at 10 Hz, recorded in a Drosophila larval abdominal muscle fibre22. The amplitude of the excitatory junctional potential (EJP) is expressed as a percentage of the maximal response. Blue circles, control; red circles, endophilin mutant; green squares, shibrets at the restrictive temperature. Adapted from ref. 22 with permission from Elsevier Science. b, Capacitance flickers recorded in posterior pituitary neurons. These capacitance changes are evoked by fusion and re-uptake of single small synaptic vesicles and probably represent kiss-and-run events. c, Plot of paired up- and downward capacitance steps separated by less than 2 s. Note that the slope of the regression line is almost one, indicating that after formation of a fusion pore with the presynaptic membrane, the same vesicle is quickly re-internalized. Panels b and c are reproduced from ref. 35.

localized to active zones and are therefore responsible for the sustained release observed in endophilin mutants. Indeed, the vesicle pool at the active zone is not depleted by repetitive intense stimulation, as judged by transmission electron microscopy. In addition, the terminals of endophilin mutants cannot be loaded with FM1-43 (ref. 22). This suggests that endophilin mutants sustain neurotransmitter release by kiss-and-run. Endophilin and dynamin mutations cause different phenotypes The ability of endophilin mutants to maintain neurotransmitter release contrasts with data obtained from shibirets NMJs at the restrictive temperature. In shibirets, the postsynaptic response to high-frequency presynaptic stimulation is eventually abolished (Fig. 2a, green). The phenotype of endophilin mutants suggests that endophilin is required only for the slow clathrin-mediated pathway, but not for kiss-and-run recycling. In contrast, dynamin must be required for both slow and fast forms of synaptic vesicle recycling,

as shibirets mutants cannot sustain release at high temperature22,26. A role for dynamin in preventing full fusion has been suggested from studies in bovine chromaffin cells27. Fused chromaffin granules are retrieved by a rapid mechanism that is sensitive to antibodies raised against dynamin, but not against clathrin. Perturbing the function of dynamin prolongs catecholamine release from individual vesicles. This would be expected if dynamin normally functions to curtail vesicular release, for example, by closing the fusion pore before progression to complete fusion28,29 (see also Supplementary Information Movie in ref. 30). Interestingly, injection of antibodies against specific dynamin isoforms suggests that clathrindependent endocytosis is mediated by dynamin-2, whereas dynamin-1 mediates rapid endocytosis31. Finally, fluorescent evanescent wave microscopy in pheochromocytoma cells suggests that dynamin clusters ‘sweep’ the PM32. A collision between such a sweeping cluster and a kissing vesicle would bring dynamin into a position to prevent progression to complete fusion and

NATURE CELL BIOLOGY VOL 4 NOVEMBER 2002 www.nature.com/naturecellbiology © 2002 Nature Publishing Group

instead promote transient fusion. Interestingly, other molecules, including complexin, which associates with fusion proteins, have also recently been proposed to regulate constriction of the fusion pore33 and could function together with dynamin in fast retrieval. Capacitance measurements demonstrate kiss-and-run in neurons So far, support for the rapid recycling of synaptic vesicles has come from indirect methods. However, two exciting recent publications34,35 describe presynaptic capacitance measurements indicating that small synaptic vesicles utilize the kiss-andrun mechanism. As the capacitance of a lipid bilayer is proportional to its surface area, fusion of vesicles with the PM increases membrane capacitance. There are two major categories of neurotransmitter-laden vesicles: large dense-core vesicles and small synaptic vesicles. Capacitance measurements have proven extremely valuable in characterizing the release mechanism of large dense-core vesicles5,7–9. Detection of the tiny capacitance changes caused by fusion of individual small synaptic vesicles has, until recently, been considered beyond technical limitations. However, in a recent report, the authors were able to characterize spontaneous fusion of small synaptic vesicles at a presynaptic structure known as the calyx of Held34. The transmitter release associated with these events cause miniature postsynaptic currents (mEPSCs). By improving the signal-to-noise ratio, extensive averaging of capacitance changes time-locked to the mEPSCs allowed measurements of the mean duration of spontaneous fusions34. The duration of the fusion event, equal to the membrane residence time discussed earlier, was only 56 ms. This very short time course suggests that the underlying mechanism of vesicle release corresponds to kiss-and-run, even though an extremely fast collapse into the membrane followed by a very rapid retrieval cannot be excluded in principle. This study also indicates that rapid recycling may be instrumental during lowintensity release. It was found that the residence time of the fusion events became longer when more intense stimulation was used to evoke fusion of vesicles34. Such a slowing of the endocytic rate in response to an increasing exocytotic load is consistent with previously published data14,36 and suggests that synapses can shift from a fast form of retrieval (associated with low levels of release) to a slow rate of endocytosis under intense stimulation. These observations may also explain why the controversy concerning the existence of different forms of endocytosis at synapses has persisted: stimulation paradigms may strongly bias the outcome, and hence the interpretation, of each experiment in different preparations.

commentary Furthermore, in a study of nerve terminals from posterior pituitary neurons, the authors went one step beyond averaging35. By reducing the level of noise in cellattached capacitance recordings, they managed to directly capture capacitance changes elicited by fusion and retrieval of single small vesicles (Fig. 2b). Recordings of spontaneous activity contained both capacitance up-steps (vesicle fusion) and downsteps (vesicle membrane retrieval) with an amplitude that was expected from the fusion of small synaptic vesicles (microvesicles). Around 5% of the up-steps were closely followed by a down-step of the same amplitude (Fig. 2c). The authors interpreted these amplitude-matching up- and down-steps as capacitance flickers associated with kiss-and-run35. The average duration of the kiss-and-run events was 310 ms. Capacitance flickers reflecting kiss-and-run events of large dense-core vesicles were also observed in these recordings, confirming previous data regarding fusion of this vesicle type7,9. Perspectives and summary The capacitance data just discussed35 provide strong evidence that synaptic vesicles can use kiss-and-run. However, the fusion pore diameter, which could be estimated for some of these events, was only 0.3 nm, too narrow for neurotransmitters to efficiently pass through the pore. However, the authors pointed out that there are several caveats to this conclusion. First, 0.3 nm is a rough estimate of the fusion pore size. Second, it is an inherent weakness of the technique used in their study that larger fusion pores tend to escape detection, that is, there is bias towards narrow pores. Moreover, fast fluctuations in fusion pore diameter, which would allow efficient

E248

transmitter efflux, cannot be resolved. Finally, there may be mechanisms that speed up transmitter extrusion35. The question of efficient transmitter release through fusion pores therefore remains unanswered. At hippocamapal terminals, it was found that FM dyes are unable to access vesicles that engage in kiss-and-run16. The small size of the fusion pore estimated in one of the capacitance studies35 may explain this finding. FM dyes are considerably larger than small neurotransmitters and may therefore be unable to pass through the fusion pore of synaptic vesicles. In conclusion, the available data suggests that synaptic vesicles are able to employ the kiss-and-run mode of release in systems as diverse as hippocampal neurons, the Drosophila NMJ, the calyx of Held and posterior pituitary neurons. More studies will be required to elucidate the molecular requirements and the regulation of this extreme form of rapid recycling. Ole Kjaerulff is in the Division of Neurophysiology, Department of Medical Physiology, University of Copenhagen, DK-2200 Copenhagen, Denmark. Patrik Verstreken is in the Program in Developmental Biology, Baylor College of Medicine, Houston, TX 77030, USA Hugo J. Bellen is in the Program in Developmental Biology and the Howard Hughes Medical Institute, Department of Molecular and Human Genetics, Division of Neuroscience, Baylor College of Medicine, Houston, TX 77030, USA e-mail: [email protected] 1. Heuser, J. Cell Biol. Int. Rep. 13, 1063–1076 (1989). 2. Slepnev, V. I. & De Camilli, P. Nature Rev. Neurosci. 1, 161–172 (2000). 3. Palfrey, H. C. & Artalejo, C. R. Neuroscience 83, 969–989 (1998). 4. Fesce, R., Grohovaz, F., Valtorta, F. & Meldolesi, J. Trends Cell Biol. 4, 1–4 (1994). 5. Artalejo, C. R., Elhamdani, A. & Palfrey, H. C. Curr. Biol. 8, R62–R65 (1998). 6. Tsuboi, T., Zhao, C., Terakawa, S. & Rutter, G. A. Curr. Biol. 10, 1307–1310 (2000).

7. Alvarez de Toledo, G., Fernandez-Chacon, R. & Fernandez, J. M. Nature 363, 554–558 (1993). 8. Neher, E. Nature 363, 497–498 (1993). 9. Ales, E. et al. Nature Cell Biol. 1, 40–44 (1999). 10. Betz, W. J. & Bewick, G. S. Science 255, 200–203 (1992). 11. Cochilla, A. J., Angleson, J. K. & Betz, W. J. Annu. Rev. Neurosci. 22, 1–10 (1999). 12. Klingauf, J., Kavalali, E. T. & Tsien, R. W. Nature 394, 581–585 (1998). 13. Pyle, J. L., Kavalali, E. T., Piedras-Renteria, E. S. & Tsien, R. W. Neuron 28, 221–231 (2000). 14. Ryan, T. A., Smith, S. J. & Reuter, H. Proc. Natl Acad. Sci. USA 93, 5567–5571 (1996). 15. Sara, Y., Mozhayeva, M. G., Liu, X. & Kavalali, E. T. J. Neurosci. 22, 1608–1617 (2002). 16. Stevens, C. F. & Williams, J. H. Proc. Natl Acad. Sci. USA 97, 12828–12833 (2000). 17. Sankaranarayanan, S. & Ryan, T. A. Nature Neurosci. 4, 129–136 (2001). 18. Harata, N. et al. Trends Neurosci. 24, 637–643 (2001). 19. Wu, L. G. & Betz, W. J. Neuron 17, 769–779 (1996). 20. Richards, D. A., Guatimosim, C. & Betz, W. J. Neuron 27, 551–559 (2000). 21. Kawasaki, F., Hazen, M. & Ordway, R. W. Nature Neurosci 3, 859–860 (2000). 22. Verstreken, P. et al. Cell 109, 101–112 (2002). 23. Gad, H. et al. Neuron 27, 301–312 (2000). 24. Ringstad, N. et al. Neuron 24, 143–154 (1999). 25. Schmidt, A. et al. Nature 401, 133–141 (1999). 26. Delgado, R. et al. Neuron 28, 941–953 (2000). 27. Artalejo, C. R., Henley, J. R., McNiven, M. A. & Palfrey, H. C. Proc. Natl Acad. Sci. USA 92, 8328–8332 (1995). 28. Elhamdani, A., Palfrey, H. C. & Artalejo, C. R. Neuron 31, 819–830 (2001). 29. Graham, M. E., O’Callaghan, D. W., McMahon, H. T. & Burgoyne, R. D. Proc. Natl Acad. Sci. USA 99, 7124–7129 (2002). 30. Galli, T. & Haucke, V. Sci. STKE [online] (cited 26 June 2001) http://stke.sciencemag.org/cgi/content/full/sigtrans;2001/88/re1 (2001). 31. Artalejo, C. R., Elhamdani, A. & Palfrey, H. C. Proc. Natl Acad. Sci. USA 99, 6358–6363 (2002). 32. Tsuboi, T. et al. J. Biol. Chem. 277, 15957–15961 (2002). 33. Archer, D. A., Graham, M. E. & Burgoyne, R. D. J. Biol. Chem. 277, 18249–18252 (2002). 34. Sun, J. Y., Wu, X. S. & Wu, L. G. Nature 417, 555–559 (2002). 35. Klyachko, V. A. & Jackson, M. B. Nature 418, 89–92 (2002). 36. Sankaranarayanan, S. & Ryan, T. A. Nature Cell Biol. 2, 197–204 (2000). ACKNOWLEDGEMENTS We thank E. Seto, A. Fayyazuddin, J. Hounsgaard and P. Saggau for their comments on the manuscript. O.K is supported by the Novo Nordisk Foundation and the Lundbeck Foundation. H.J.B. is supported by the Howard Hughes Medical Institute.

NATURE CELL BIOLOGY VOL 4 NOVEMBER 2002 www.nature.com/naturecellbiology © 2002 Nature Publishing Group