1131Research Article

IntroductionOur understanding of synaptic mechanisms has been greatlyenhanced by the use of naturally occurring neurotoxins. Shaped byevolution, neurotoxins act on key elements of synaptic processesin a highly specific manner. Some of these neurotoxins directly affectthe neurotransmitter-release machinery and have been used todecipher key steps in exocytosis by helping to characterize the roleof certain proteins involved in this process (de Paiva et al., 1999;Schiavo et al., 2000). By contrast, very few neurotoxins have beenfound to affect synaptic-vesicle endocytosis (Ceccarelli and Hurlbut,1980). Initially, the venom of the polychaete Glycera convoluta wasstudied because of its ability to stimulate spontaneous quantalacetylcholine (ACh) release at the frog neuromuscular junction(NMJ) (Morel et al., 1983). G. convoluta venom (GCV) producesa long-lasting (several hours) high-frequency discharge of miniatureend-plate potentials (MEPPs). Interestingly, the number ofspontaneous quantal events that occur during hours of GCVstimulation far exceeds the number of synaptic vesicles known tobe present in motor nerve terminals. However, preliminary

ultrastructural examination of GCV-treated motor nerve terminalsdid not reveal any change in the number of synaptic vesicles, whichhas been used historically as evidence supporting the non-vesicularhypothesis (Morel et al., 1983). Later work from Ceccarelli’s groupuncovered synaptic-vesicle recycling at the frog NMJ (Ceccarelliand Hurlbut, 1980; Ceccarelli et al., 1973), which follows synaptic-vesicle fusion events occurring by means of either a ‘kiss-and-run’mechanism (Fesce et al., 1994) or full fusion with the plasmamembrane (Heuser and Reese, 1973). Although evidence for a kiss-and-run type of exocytosis has been found in neurosecretory andmast cells (de Toledo et al., 1993; Fernandez et al., 1984; Neherand Marty, 1982; Xia et al., 2009), much controversy still existsregarding the relevance of such a fast retrieval mechanism at theNMJ and central synapses (Balaji and Ryan, 2007; Dickman et al.,2005; Granseth et al., 2009; Rizzoli and Betz, 2004; Rizzoli andBetz, 2005; Verstreken et al., 2002; Zhang et al., 2009).

We have purified from GCV a glycoprotein of 320 kDa namedglycerotoxin (GLTx). GLTx is its most active component uponsynaptic release and its molecular target was identified as N-type

Sustained synaptic-vesicle recycling by bulkendocytosis contributes to the maintenance ofhigh-rate neurotransmitter release stimulated byglycerotoxinFrederic A. Meunier1,*, Tam H. Nguyen1, Cesare Colasante2, Fujun Luo3, Robert K. P. Sullivan1,4,Nickolas A. Lavidis5, Jordi Molgó6, Stephen D. Meriney3 and Giampietro Schiavo7

1Molecular Dynamics of Synaptic Function Laboratory, Queensland Brain Institute and School of Biomedical Sciences, The University ofQueensland, Brisbane QLD 4072, Australia2Laboratorio de Fisiología de la Conducta, Facultad de Medicina, Universidad de Los Andes, Mérida 5101, Venezuela3Department of Neuroscience, University of Pittsburgh, Pittsburgh, PA 15260, USA4Centre for Microscopy and Microanalysis, The University of Queensland, Brisbane QLD 4072, Australia5School of Biomedical Sciences, The University of Queensland, Brisbane QLD 4072, Australia6CNRS, Institut de Neurobiologie Alfred Fessard – FRC2118, Laboratoire de Neurobiologie Cellulaire et Moléculaire – UPR9040, 1 Avenue de laTerrasse, 91198 Gif sur Yvette, France7Molecular Neuropathobiology Laboratory, Cancer Research UK London Research Institute, Lincoln’s Inn Fields Laboratories, London WC2A 3PX,UK*Author for correspondence (f.meunier@uq.edu.au)

Accepted 4 January 2010Journal of Cell Science 123, 1131-1140 © 2010. Published by The Company of Biologists Ltddoi:10.1242/jcs.049296

SummaryGlycerotoxin (GLTx), a large neurotoxin isolated from the venom of the sea worm Glycera convoluta, promotes a long-lasting increasein spontaneous neurotransmitter release at the peripheral and central synapses by selective activation of Cav2.2 channels. We foundthat GLTx stimulates the very high frequency, long-lasting (more than 10 hours) spontaneous release of acetylcholine by promotingnerve terminal Ca2+ oscillations sensitive to the inhibitor -conotoxin GVIA at the amphibian neuromuscular junction. Although anestimate of the number of synaptic vesicles undergoing exocytosis largely exceeds the number of vesicles present in the motor nerveterminal, ultrastructural examination of GLTx-treated synapses revealed no significant change in the number of synaptic vesicles.However, we did detect the appearance of large pre-synaptic cisternae suggestive of bulk endocytosis. Using a combination of styryldyes, photoconversion and horseradish peroxidase (HRP)-labeling electron microscopy, we demonstrate that GLTx upregulatespresynaptic-vesicle recycling, which is likely to emanate from the limiting membrane of these large cisternae. Similar synaptic-vesiclerecycling through bulk endocytosis also occurs from nerve terminals stimulated by high potassium. Our results suggest that this processmight therefore contribute significantly to synaptic recycling under sustained levels of synaptic stimulation.

Key words: Bulk endocytosis, Glycerotoxin, Synaptic-vesicle recycling, FM1-43, Photoconversion, Neurotransmitter release

Jour

nal o

f Cel

l Sci

ence

1132

Ca2+ channels (Cav2.2) (Meunier et al., 2002; Schenning et al.,2006). Here, we take advantage of the sustained stimulation ofneurotransmitter release, characterized by burst-like episodesreaching more than 100-200 quanta released per second, that istriggered by GLTx to explore the mechanism of fast synaptic-vesicleretrieval at the amphibian NMJ under sustained asynchronousrelease stimulation.

We found that GLTx promotes presynaptic Ca2+ oscillations,driving an increase in MEPP frequency lasting for more than 10hours and therefore largely exceeding the presynaptic content ofsynaptic vesicles. Quantitative electron microscopy analysisdemonstrated that the number of synaptic vesicles present in motornerve terminals was not significantly affected by GLTx despite thenumber of quanta released. We demonstrate that GLTx drasticallyenhances synaptic-vesicle recycling – an effect prevented bypretreatment with the Cav2.2-specific inhibitor -conotoxin MVIIA.Furthermore, electron microscopy analyses upon photoconversionof the styryl dye AM1-43 or horseradish peroxidase (HRP) labelingindicated that the recycling of synaptic vesicles triggered by GLTxstimulation predominantly originates from large presynapticendosomal structures.

ResultsGlycerotoxin elicits a long-term increase inneurotransmitter releaseGLTx is capable of eliciting a dramatic increase in spontaneousquantal neurotransmitter release at the frog NMJ (Meunier et al.,2002). Simple wash by perfusion of GLTx-free Ringer’s buffercompletely reversed the stimulatory effect (Fig. 1A). Re-applicationof GLTx on the same preparation elicited an equivalent increase inMEPP frequency, demonstrating that the effect of GLTx is indeedreversible and does not seem to promote run down of ACh releasefrom motor nerve terminals (Fig. 1A). We have previously reportedthat the stimulatory effect of GLTx can be prevented by pretreatingthe neuromuscular preparation with a Cav2.2-specific inhibitor, suchas -conotoxin GVIA (-CTx-GVIA) (Meunier et al., 2002;Schenning et al., 2006), which does not compete with GLTx forthe same binding site on the Ca2+ channel (Schenning et al., 2006).To test whether -CTx-GVIA, in addition to preventing thestimulatory action of GLTx, could also block neurotransmitterrelease after the activation of Cav2.2 by GLTx, we determinedwhether the GLTx response could be inhibited by post-treatmentwith -CTx-GVIA. -CTx-GVIA added 20 minutes after GLTxcompletely blocked the stimulatory effect within 5 minutes afterits application in the continuous presence of GLTx (Fig. 1B). Thisresult confirms that GLTx binding does not prevent the interactionof -CTx-GVIA with Cav2.2 and that the effect of GLTx is totallyabrogated by selective Cav blockade. Similar results were obtainedusing -CTx-MVIIA, further demonstrating the essential role ofCav2.2 in the stimulatory activity of GLTx (data not shown).

Very few neurotoxins are capable of promoting neurotransmitterrelease at a frequency as high as 100 MEPPs/second. However, -latrotoxin (-LTX) from the venom of the black widow spider andtrachynilysin from the stonefish Synanceia trachynis have beenreported to promote such stimulatory activity (Clark et al., 1970;Colasante et al., 1996; Tsang et al., 2000). These two toxins producea rapid depletion in synaptic vesicles, leading to a block of bothevoked and spontaneous transmitter release (Clark et al., 1970;Colasante et al., 1996; Tsang et al., 2000). In the case of -latrotoxin,this depletion only occurs under Ca2+-free conditions (Ceccarelliand Hurlbut, 1980). We therefore investigated whether GLTx could

also promote long-term blockage of spontaneous release fromintoxicated NMJs. Surprisingly, nerve terminals exposed to GLTxfor 175 minutes still exhibited high MEPP frequency (Fig. 1C),which remained elevated even after more than 10 hours ofcontinuous GLTx stimulation (Fig. 1C), suggesting that synapticvesicles were being actively recycled.

Glycerotoxin promotes a long-term increase in basalintraterminal Ca2+ levelsGLTx has been previously shown to promote spontaneous rises inlevels of intracellular Ca2+ in rat brain synaptosomes (Madeddu etal., 1984) and in the cell body of rat motor neurons in culture; these

Journal of Cell Science 123 (7)

Fig. 1. Long-lasting stimulatory effect of GLTx at the amphibian NMJ.Intracellular recording of spontaneous quantal ACh release from NMJsincubated in the indicated Ringer’s solutions in the absence (empty bars) orpresence (filled bars) of 150 pM GLTx. Sampled traces were collected andanalyzed for mean MEPP frequency 5 minutes before and 20 minutes afteradding GLTx (right panels), and characteristic recordings are shown (leftpanels). (A)GLTx addition induced a large increase in MEPP frequency inCa2+-containing Ringer’s solution. This effect was promptly reversed bysimple wash in a Ca2+-free buffer. Addition of GLTx in a Ca2+-free Ringer’ssolution did not promote any detectable increase in MEPP frequency.Subsequent addition of Ca2+-containing buffer in the continuous presence ofGLTx triggered a large increase in MEPP frequency. (B)GLTx was added to aCa2+-containing Ringer’s solution and a large increase in MEPP frequency wasobserved. The NMJ was then treated with -CTx-GVIA (1M, 20 minutes),which caused a rapid reduction in MEPP frequency. (C)GLTx (150 pM) wasadded to a Ca2+-containing Ringer’s solution, leading to a large increase inMEPP frequency that was still detectable after 10 hours. MEPP quantificationat various time points is indicated in the right panel. Sampled traces wereanalyzed to determine the mean MEPP frequency from 3-9 muscle fibers.Plotted values are representative of at least three independent experiments.

Jour

nal o

f Cel

l Sci

ence

1133GLTx promotes synaptic vesicle recycling

increases were reported to be sensitive to the blockade of Cav2.2channels (Schenning et al., 2006). From this earlier work, it washypothesized that GLTx stimulates neurotransmitter release bydirectly activating Cav2.2 and enhancing Ca2+ entry into the motornerve terminals. To directly test this hypothesis, we imaged thespontaneous fluctuations in intracellular Ca2+ in motor nerveterminals of the frog NMJ after treatment with GLTx. As shownin Fig. 2, the morphological feature of regularly spaced active zones,each of which is about 1 m long, can be exploited to focus theanalysis of Ca2+ fluctuations in individual active zones (Fig. 2A).Using this approach, we collected Ca2+-sensitive fluorescent imageswithin these areas of the NMJ at 0.5 Hz. We detected a GLTx-mediated increase in resting Ca2+ fluctuations (Fig. 2B,C).Strikingly, there were occasional strong Ca2+ fluctuations. After both1 and 5 hours of incubation with GLTx (150 pM), this generalincrease in fluctuations and the strong Ca2+ fluctuations occurredspontaneously at similar rates within these presynaptic active zones(Fig. 2B-E).

To measure the effect of GLTx in terms of promotion ofspontaneous Ca2+ entry, we calculated the frequency of thosefluctuations that exceeded twice the standard deviation of averagefluctuations within these active zones. Following 1 hour of

incubation with GLTx, this frequency was 0.016±0.004/active zone(AZ)/second (54 active zones, 5 terminals) and, following five hoursof exposure, it became 0.013±0.001/AZ/second (37 active zones,3 terminals). There was no significant difference in frequencybetween 1 hour and 5 hours of GLTx incubation (p0.56, unpairedt-test). Considering that each amphibian motor nerve terminalcontains approximately 700 active zones (Soyoun Cho and S.D.M.,unpublished observations), the frequency of spontaneousfluctuations in Ca2+ entry would be about 10/nerve terminal/second.In this light, it is likely that the increase in Ca2+ fluctuations leadsto the reported general elevation of MEPP frequency, and that thestrong Ca2+ entry events might instead trigger the large synchronousbursts in MEPP frequency observed in addition to the generalincrease in asynchronous MEPP frequency (Betz et al., 1992; dePaiva et al., 1999; Henkel et al., 1996; Lin et al., 2005; Meunier etal., 2002; Ribchester et al., 1994; Smith and Betz, 1996).

To confirm that Cav2.2 channels are responsible for the GLTx-evoked Ca2+ fluctuations at the frog NMJ, we examined the effectsof Cav2.2 blockers in this system. Fig. 3A-C shows that thesetransient Ca2+ spikes disappeared after the addition of -CTx-GVIA(2 M). To compare quantitatively the magnitude of Ca2+

fluctuations, we calculated the coefficient of variation (CV) offluorescence within each active zone. One hour (n81 active zones)of GLTx treatment significantly increased this parameter withinindividual active zones (P<0.01, one-way ANOVA) compared with

Fig. 2. GLTx-evoked long-lasting spontaneous Ca2+ oscillations atamphibian NMJ active zones. (A)NMJ preparations were treated withAlexaFluor594--BTX (left panel) and filled with Calcium Green-1 (rightpanel). Predicted active-zone regions of a representative motor nerve terminalwere identified using -BTX labeling of the postsynaptic ACh receptor foldsas indicated. (B)The imaged nerve terminal was treated with 150 pM GLTxfor 60 minutes. The averaged fluorescence image was superimposed on whiteboxes, indicating distinct active zones used for signal analysis in C. (C)Scatterplot of the mean fluorescent signal collected from the selected active zone(white box) of an identified nerve terminal incubated with 150 pM GLTx for60 minutes, showing Ca2+ fluctuations over a period of time. Arrows indicatetime points at which images in B were taken. (D)Spontaneous intracellularCa2+ transients within active regions of the motor nerve terminal were stilldetected after treatment with 150 pM GLTx for more than 5 hours (on thesame identified motor nerve terminal as in A). (E)Scatter plot of meanfluorescent signal within the selected active zone, as shown in the white box.

Fig. 3. GLTx-induced Ca2+ oscillations are sensitive to -CTx-GVIAtreatment at the amphibian NMJ. NMJ preparations were treated withAlexaFluor594--BTX and Calcium Green-1 as in Fig. 2. (A)The loadedsuperficial nerve terminal was imaged by using epifluorescence microscopy.Representative fluorescence images were collected at the indicated time pointswithin active zones before (control), after treatment with 150 pM GLTx for1 hour (GLTx) and after addition of 1M -CTX-GVIA for 30 minutes(GLTx plus -CTx-GVIA). (B)Mean fluorescent signal within the selectedactive zones (arrowhead shown in the white box in A), normalized to the meanvalue of baseline intensity for each condition. (C)CV of fluorescent intensityover time under the indicated experimental conditions: control (empty bar,n24); GLTx (filled bar, n81) and GLTx plus -CTx-GVIA (filled bar, n57).

Jour

nal o

f Cel

l Sci

ence

1134

active zones in control terminals (n24 active zones) (Fig. 3C).Importantly, there was no significant difference in CV betweencontrol nerve terminals and nerve terminals treated with GLTx plus-CTx-GVIA (n57 active zones; P>0.05, one-way ANOVA) (Fig.3C). Furthermore, the CV of fluorescence intensity of individualactive zones was still significantly increased after 5 hours of GLTxtreatment compared to controls (n37 active zones, data not shown).

Glycerotoxin promotes the long-term upregulation ofendocytosisStyryl dyes such as FM1-43 have been widely used to investigatesynaptic-vesicle recycling at the NMJ (Betz et al., 1992; de Paivaet al., 1999; Henkel et al., 1996; Lin et al., 2005; Ribchester et al.,1994; Smith and Betz, 1996). FM1-43 was applied for 5 minutesin the absence of stimulation before washing the preparation withcold Ca2+-free medium. As shown in Fig. 4A, FM1-43 staining waspredominantly localized on the plasma membrane of motor nerveterminals, as previously shown (Betz et al., 1992). The sameprocedure was then carried out 20 minutes after the initial additionof GLTx. Importantly, GLTx was capable of inducing theinternalization of the styryl dye, as illustrated by the intensity profiletaken across the nerve terminal (Fig. 4B) when compared to theunstimulated control (Fig. 4A). Subsequently, FM1-43-loadedmotor nerve terminals were stimulated with GLTx in a Ca2+-containing Ringer’s solution, rinsed and imaged 5 minutes later.Destaining of the labeled terminals was observed, with translocationof the dye from the intraterminal space to the plasma membrane(Fig. 4C). The GLTx-induced FM1-43 uptake was inhibited by pre-incubation of the preparation with -CTx-MVIIA (Fig. 4D). To testwhether the FM1-43 taken up as a result of GLTx action could bereleased upon electrical stimulation, a destaining experiment was

carried out. Nerve terminals were preloaded with FM1-43 for 5minutes at the end of 30 minutes stimulation using GLTx, washedand imaged before and after stimulation at 5 Hz for 5 minutes. Cleartranslocation of the dye from the intraterminal space to the plasmamembrane was found, suggesting that GLTx-induced recycling ofvesicles can efficiently be reused for neurotransmitter release (Fig.4E). Although most of the internal FM1-43 staining was releasedusing electrical stimulation, few of the FM1-43-labelled internalstructures did not undergo destaining, suggesting that a limitednumber of the endocytic compartments generated by GLTx entereda dead-end pathway. Taken together, these experiments demonstratethat GLTx triggers exocytosis of synaptic vesicles as well aspromoting their active compensatory recycling.

To investigate whether the effect of GLTx on synaptic-vesiclerecycling was long lasting, a short pulse of FM1-43 was appliedduring the last 5 minutes of GLTx treatments of variable duration.This was followed by several washes in cold Ca2+- and GLTx-freeRinger’s solution prior to imaging (Fig. 5A). Interestingly, whenGLTx was co-applied with FM1-43 for only 5 minutes, hotspotsabutting the plasma membrane were clearly visible, indicating sitesof active vesicle recycling (Fig. 5B). Preparations exposed to GLTxfor longer periods (between 30 and 300 minutes) instead exhibitedinternal staining of the motor nerve terminals (Fig. 5B).Quantification of FM1-43 uptake shows that GLTx promotes long-lasting upregulation of synaptic-vesicle recycling (Fig. 5C).Importantly, at longer time points, the internalized FM1-43 stainingwas not uniform throughout nerve terminals nor did it exhibit theclassical banding pattern observed following relatively mildstimulation (5 minutes high K+-induced depolarization, Fig. 6A)(Betz et al., 1992). Rather, donut-like structures displaying intenseFM1-43 staining were detected (Fig. 5B and Fig. 6D). Interestingly,

Journal of Cell Science 123 (7)

Fig. 4. GLTx promotes synaptic-vesicle recycling atthe amphibian NMJ. (A)Control untreated preparationincubated with FM1-43 (5M) for 5 minutes prior to anextensive wash in Ca2+-free Ringer’s solution (see leftpanel). Nerve terminals were imaged by confocalmicroscopy and an intensity profile was run as indicated(right panel). (B)A GLTx-treated (150 pM for 20minutes) preparation was incubated with FM1-43 (5M)for the last 5 minutes prior to an extensive wash in Ca2+-free buffer (see left panel). The intensity profile clearlyshows activity-dependent uptake of the dye. (C)High K+-treated (10 minutes) preparation was incubated withFM1-43 (5M) for the last 5 minutes of stimulation priorto an extensive wash in Ca2+-containing Ringer’s solutionin the presence of GLTx (150 pM for 20 minutes). Notethe GLTx-induced translocation to the plasma membraneof internalized FM1-43. (D)NMJ preparations pretreatedwith 1M -CTx-MVIIA prior to exposure to GLTx andFM1-43 as in B. Note the absence of dye internalization.(E)Motor nerve terminals were stimulated for 30 minuteswith GLTx; FM1-43 (5M) was added during the last 5minutes of stimulation. The preparation was washed fivetimes in normal Ringer’s solution and FM1-43-loadednerve terminals were imaged before and after electricalstimulation at 5 Hz for 5 minutes. Note the electrical-stimulation-induced destaining of internalized FM1-43.Confocal images representative of the results obtainedfrom at least 3-6 independent experiments. Scale bars:10m.

Jour

nal o

f Cel

l Sci

ence

1135GLTx promotes synaptic vesicle recycling

these structures were also induced by long-lasting depolarizationwith high K+ medium (Fig. 6B), suggesting that a switch inendocytosis mode occurs to allow long-term mobilization ofsynaptic-vesicle recycling. These structures were reminiscent oflarge vacuoles, which might appear following sustained stimulationof neurotransmitter release. Importantly, the high intensity of FM1-43 staining of these donut-like structures suggested that the majorityof recycling vesicles could be emanating from these large endocyticstructures.

Large endocytic structures make a major contribution tosynaptic-vesicle recycling under sustained stimulationWe next investigated the ultrastructural changes produced by GLTxtreatments of various lengths. A number of parameters wereexamined, including the number of synaptic vesicles, coatedvesicles, large dense core vesicles and enlarged endosomes, and

the area of the presynaptic nerve terminals (Table 1). Mostparameters remained unchanged by GLTx treatment at all of thetime points investigated. Importantly, the number of synapticvesicles was not affected, demonstrating that the recycling processwas indeed capable of maintaining an intact pool of synapticvesicles. Another line of evidence that active recycling was takingplace upon GLTx treatment was the finding that the number oflarge endosomes was significantly increased by GLTx treatmentat longer time points (Table 1). To investigate the nature andlocalization of recycling vesicles, we used the photoconvertiblestyryl dye AM1-43 to identify AM1-43-loaded recycling vesiclesin GLTx-treated preparations by electron microscopy. We nextincubated preparations in the presence or absence of GLTx for 60and 120 minutes prior to applying a pulse of AM1-43 for 5 minutes,followed by extensive washing in cold Ca2+- and GLTx-freeRinger’s solution. Confocal images of the stained nerve terminalwere taken (Fig. 6D) before carrying out photoconversion andprocessing of the preparation for electron microscopy. Unstimulatednerve terminals did not exhibit any noticeable AM1-43 precipitate(Fig. 6E). Examination of GLTx-treated nerve terminals revealedlarge endosomal structures (Fig. 6F,G). Importantly, the vastmajority of photoconverted vesicles were either abutting or in closeproximity to these large structures (Fig. 6E,F). In some instances,a number of these large structures were found to be either veryclose to or fused with each other (see serial sections in Fig. 6G,inserts).

The lack of detectable photoconverted AM1-43 dye inside thelumen of the large endosomal structures in GLTx-treated nerveterminals was surprising. However, this is not unprecedentedbecause it has been previously suggested that FM dyes do notconvert well inside endosomal compartments (de Lange et al.,2003). We therefore used the classical fluid-phase markerhorseradish peroxidase (HRP) to probe the formation of the largeendosomes under long-term GLTx stimulation. NMJ preparations,in either the presence or absence of GLTx, were exposed toHRP for 5 hours. In preparations treated with GLTx,electrophysiological recordings of MEPP frequency wereperformed to confirm the activity of GLTx prior to addition ofHRP (data not shown). After 5 hours, samples were fixed andprocessed for electron microscopy. In control unstimulatedpreparations treated with HRP, no endosomes were detected andthe HRP precipitate was confined to the synaptic cleft andaround the perisynaptic Schwann cells (Fig. 7A). In contrast, inGLTx-stimulated preparations, HRP was taken up into bulkendosomes (Fig. 7B-D). HRP-positive tubular structuresseemingly in contact with the plasma membrane were alsodetected (Fig. 7B and E). Remarkably, the electron-denseprecipitate generated by HRP in our experiment did not fill theentire endosome, but was mainly found decorating its membrane(Fig. 7B-G). This is consistent with previous results obtained usingthe same technique (Clayton et al., 2008; Clayton et al., 2009;Meshul and Pappas, 1984). Notably, HRP-filled vesicles weredetectable in close proximity to these endosomes (Fig. 7F). Oneexample is shown in Fig. 7F (arrow): one can see clear continuityin the HRP staining between the HRP precipitate lining theendosomal membrane and an omega-shaped vesicle seeminglybudding from the bulk endosome. Interestingly, we also observedC-shaped bulk cisternae and multilamellar endosomes in GLTx-treated nerve terminals (Fig. 7H and I). Endosomes with vesiclesseemingly in the process of budding off were frequently observedin GLTx-treated nerve terminals (Fig. 7J-M). Occasionally, large

Fig. 5. Time-course of GLTx-induced synaptic-vesicle recycling at the frogNMJ. (A)Nerve-muscle preparations were incubated with GLTx (150 pM) forvarious periods of time and a pulse of FM1-43 (5M) was applied in the last 5minutes of GLTx incubation. Samples were subsequently washed in a Ca2+-free Ringer’s solution and imaged by confocal microscopy. Ttime of GLTxtreatment. (B)Confocal images highlighting the change in GLTx-mediatedFM1-43 uptake over the indicated times. Confocal images are representativeof the results obtained from 2-6 independent experiments. Scale bars: 10m.(C)Sampled images were analyzed to determine the mean FM1-43 uptakefrom 3-4 motor nerve terminals. Plotted values are representative of 3-4independent experiments.

Jour

nal o

f Cel

l Sci

ence

1136

membrane invaginations with vesicles seemingly pinching offfrom these intermediate structures were also observed (Fig. 7N).Altogether, our data suggest that GLTx promotes an activerecycling process stemming from bulk endosomes that contributesto the maintenance of neurotransmitter release at the NMJ.

DiscussionIn this study, we characterized the morphological and functionalchanges caused by GLTx at the amphibian NMJ. We found thatGLTx elicits long-term stimulation of asynchronous quantaltransmitter release without noticeable changes in the number of

Journal of Cell Science 123 (7)

Table 1. Estimates of morphological parameters per nerve terminal of amphibian cutaneous pectoris nerve-musclepreparations following GLTx exposure

TerminalsSynapticvesicles

Nerve terminalarea (µm2)

Large dense corevesicles Coated vesicles

Endosomalstructures

Endosome area(%)

Control (2 hours) 60 142±16 2.05±0.11 1.60±0.19 3.01±0.50 3.02±0.26 2.5±1.4Control (8 hours) 65 149±11 1.94±0.09 1.32±0.13 2.60±0.72 2.51±0.33 3.0±0.6GLTx (2 hours) 60 134±12 2.20±0.15 2.25±0.11 3.48±0.37 4.76±0.50 18.6±2.4GLTx (4 hours) 40 140±17 2.36±0.19 1.93±0.21 2.55±0.29 4.08±0.62 15.3±1.2GLTx (6 hours) 50 150±15 2.21±0.10 2.20±0.20 3.70±0.32 3.92±0.37 12.9±3.1GLTx (8 hours) 65 137±9 1.95±0.12 1.37±0.10 2.99±0.45 4.56±0.40 9.56±1.2

Data given are mean±s.e.m.

Fig. 6. Ultrastructural analysis of synaptic-vesicle recycling under sustained stimulationat the amphibian NMJ. Nerve-musclepreparations were stimulated by high K+

depolarization (A-C) or GLTx (D) for theindicated times and a pulse of AM1-43 wasapplied in the last 5 minutes of stimulation. Thepreparation was subsequently washed in a Ca2+-free Ringer’s solution and imaged by confocalmicroscopy (A,B,D) or electron microscopy (C).Scale bars: 10m (A,B,D). Control (E) andGLTx-stimulated (F,G) preparations wereprocessed for photoconversion and electronmicroscopy. The majority of the photoconvertedvesicles are located in the immediate vicinity ofthe large endosomal limiting membrane. Insertsin G show a serial section through a largeendosome, highlighting the complexity of thismembrane network. The arrows indicate theclose apposition of several large vacuoles withsmaller endosomes, which seem to emerge fromlarger ones. Confocal images and electronmicrographs are representative of resultsobtained from 3-6 independent experiments.

Jour

nal o

f Cel

l Sci

ence

1137GLTx promotes synaptic vesicle recycling

synaptic vesicles. GLTx was therefore used as a tool to examinethe modality of synaptic-vesicle recycling under such conditionsof elevated neurotransmitter release. Using FM1-43, we found thatGLTx greatly increased the synaptic-vesicle recycling rate, whichwas upregulated for over 5 hours. Photoconversion experimentscarried out at various time points after GLTx application clearlyshow that most recycling vesicles are localized on or near largeendosomal compartments.

GLTx has unique features among excitatory neurotoxinsOther excitatory neurotoxins, such as -LTX or trachynilysin, havebeen shown to promote the selective depletion of small synapticvesicles (Colasante et al., 1996; Matteoli et al., 1988). These twoneurotoxins stimulate asynchronous neurotransmitter release at asimilar rate to GLTx, but this rapidly results in a full blockade ofneurotransmitter release caused by the depletion of synaptic vesicles(Colasante et al., 1996; Tsang et al., 2000). It therefore appears thatthe mechanism of GLTx action differs from that of the above-mentioned pore-forming neurotoxins (Davletov et al., 1998).Although the stimulatory effect of trachynilysin is strictly dependenton external Ca2+, the synaptic-vesicle depletion elicited by -LTXin the absence of external Ca2+ (Ceccarelli et al., 1973) has beenattributed to the mobilization of intracellular Ca2+ from mitochondria(Tsang et al., 2000). One of the major differences between the effectsof -LTX and GLTx is that the stimulatory effect of GLTx isdependent on external Ca2+ and the intracellular Ca2+ signals elicitedby the two toxins are not identical in nature. Whereas -LTXpromotes a long-lasting high-amplitude rise in intracellular Ca2+,GLTx gives rise to an overall small increase in the basal Ca2+ level(low amplitude), accompanied by occasional Ca2+ spikes

(Fig. 2C,E). In this light, the Ca2+ signal generated by GLTx isreminiscent of that elicited by electrical stimulation and this mightexplain why GLTx does not overwhelm the endocytic mechanism.Alternatively, pore-forming toxins such as trachynilysin and -LTXmight cause a run-down of the energy level of the nerve terminals,contributing to the blockade of endocytosis and vesicle recycling.GLTx targets motor nerve terminals through a direct interaction withCav2.2, resulting in channel upregulation (Meunier et al., 2002;Benoit et al, 2002; Schenning et al., 2006). Similarly, GLTx hasbeen shown to promote Ca2+ influx in Cav2.2-expressing HEK andbovine chromaffin cells (Meunier et al., 2002) and, more recently,in cultured motor neurons (Schenning et al., 2006). The single largeCa2+ transient detected in these cells was hard to reconcile with thelong-lasting stimulatory effect of GLTx observed at the amphibianNMJ. Our results demonstrate that GLTx promotes subtle long-lasting changes in the basal level of intraterminal Ca2+ (Fig. 2),consistent with a general increase in asynchronous release.Moreover, the intermittent larger Ca2+ transient could account forthe burst of MEPPs also observed in GLTx-treated preparations.

Number of vesicles released by long-term activityThe ability of neurons to sustain high levels of neurotransmitterrelease under high-frequency stimulation is crucial to maintainingmotor and more integrative cerebral functions. The level of GLTx-induced asynchronous neurotransmitter release from motornerve terminals is comparable to that seen in high-frequency nervestimulation (~100 synaptic vesicles/second; Fig. 1). In 10 hoursof GLTx treatment, an estimated more than 3 million synapticvesicles underwent fusion. In our longest GLTx incubation(24 hours), the nerve terminals were still firing at

Fig. 7. GLTx promotes the uptake of HRP into large endosomes.NMJs were treated with HRP (10 mg/ml) for 5 hours in eitherthe absence (A) or presence of GLTx (B-N), fixed and processedfor electron microscopy. No endosomes were detected in controlunstimulated NMJ preparations, with HRP precipitate confinedto the synaptic cleft and Schwann cells (A, arrows). Inset from(B) shows the accumulation of HRP precipitate inside a GLTx-induced bulk endosome (C,D) and tubular structures (E). TheHRP signal is present in vesicles seemingly budding from thebulk endosomes (F,G, arrow). GLTx-treated nerve terminalsdisplayed distinct ultrastructural features, such as C-shapedcisternae (H) and multilamellar endosomes (I). Numerousexamples of vesicles in the process of budding off from the bulkendosomes were observed (J-M, black arrows). Occasionally,regions of membrane invagination were observed, with omega-shaped vesicles seemingly budding from these large membraneinvaginations still attached to the plasma membrane (N, blackarrow). PM, plasma membrane; white asterisk, intraterminalspace.

Jour

nal o

f Cel

l Sci

ence

1138

100 MEPPs/second, suggesting that, in this case, over 8 millionvesicles were released. Considering that the total number ofsynaptic vesicles at the frog presynaptic nerve terminal has beenestimated to be 300,000 (Couteaux and Pecot-Dechavassine,1970; Heuser and Reese, 1981; Van der Kloot and Molgo, 1994),the number of released vesicles after 10 hours of GLTx actionexceeded the actual total number of synaptic vesicles by morethan an order of magnitude. This suggests that the synaptic-vesiclerecycling mechanism is not only highly upregulated, but alsorobust enough to sustain high activity for many hours, with noevidence of synaptic-vesicle depletion (Table 1). The total numberof readily releasable vesicles per nerve terminal at a given timepoint has been estimated to be ~25,000. GLTx action wouldtherefore deplete this vesicle pool in just over 4 minutes,suggesting that, over 10 hours of stimulation, the readily releasablepool has been completely replenished over 150 times.

Other studies have used high-frequency stimulation conditionsof 30 Hz for 1 minute, which corresponds to 1800 stimulations(Richards et al., 2000). Assuming that each stimulation results in380 quanta being released (Van der Kloot and Molgo, 1994), thetotal number of quanta released is 684,000. Our study is perhapsmore comparable to an earlier study, which describes 3 millionquanta being released over 24 hours using a 2 Hz nerve-evokedsimulation paradigm (Lynch, 1982). This type of stimulation wouldproduce a higher rate of release (700-800 quanta/second), whichwas shown to induce significant depletion of synaptic vesicles(Lynch, 1982). To the best of our knowledge, our study establishesthat the machinery for synaptic-vesicle recycling at the frog NMJcan cope with the release of over 100 but less than 700vesicles/second.

Mode of endocytosis under GLTx stimulationTwo main hypotheses have been put forward to explain synaptic-vesicle recycling: clathrin-mediated endocytosis occurring outsidethe active zones; and a ‘kiss-and-run’ mechanism involving thetransient fusion and immediate retrieval of synaptic vesicles inthe active zone (Ceccarelli et al., 1973; Colasante et al., 1996;Colasante and Pecot-Dechavassine, 1996; Heuser and Reese,1973). Both modes of endocytosis have been shown to playimportant roles in maintaining cellular communication at varyingfrequencies in a range of synapses (Colasante and Pecot-Dechavassine, 1996; Harata et al., 2006). At the frog NMJ, coatedpits and coated vesicles were only detected far from the activezones upon electrical stimulation in the presence of the K+ channelblocker 4-aminopyridine (Colasante and Pecot-Dechavassine,1996; Heuser and Reese, 1981), whereas during simple high K+

stimulation, neither coated pits nor synaptic-vesicle depletion weredetected (Colasante and Pecot-Dechavassine, 1996); an alternativemechanism for synaptic-vesicle recycling, such as kiss-and-run,has thus been inferred.

The use of fluorescent styryl dyes has allowed the characterizationof distinct vesicle pools that differ in their dynamic properties(Rizzoli and Betz, 2005). Low-frequency neurotransmission mainlyrelies on the readily releasable pool of synaptic vesicles, whereashigh-frequency stimulation depletes the reserve pool (Gaffield etal., 2006; Richards et al., 2000). We have used GLTx as a tool toinvestigate which mechanism is responsible for the replenishmentof synaptic vesicles under conditions of sustained high-level releaseof neurotransmitters at the NMJ. Our photoconversion experimentdemonstrates that recycling occurs via synaptic vesicles that arelikely to bud from large endosomal structures.

In the kiss-and-run mode of endocytosis, recycled vesicles areexpected to localize at or very near the plasma membrane (Rizzoliand Betz, 2004). Our results show that, under GLTx-inducedsustained stimulation of exocytosis, only a few recycled vesiclescan be detected in this area (Fig. 6). This is in good agreement withwork demonstrating that recycled vesicles were not enriched nearthe active zone (Rizzoli and Betz, 2004). In our experiments, mostof the photoconverted vesicles were localized either abutting or inthe proximity of large vacuolar structures. These compartments werenot detected upon short-term high-rate stimulation; smaller oblong-shaped cisternae were mostly observed in this case (Richards et al.,2000). Similarly, low-rate chronic stimulation (24 hours at 2 Hz)has been shown to increase the number of these small cisternae atthe frog NMJ (Lynch, 1982). One could speculate that the largeendosomal structures highlighted in our study might result from theaccumulation of plasma membrane derived from the hot-spotsdetected on the plasma membrane after 5 minutes exposure to GLTx(Fig. 5B). Our HRP-labeling approach confirmed the resultsobtained by AM1-43 photoconversion by showing that HRP wastaken up by these GLTx-induced bulk endosomes (Fig. 7) and thatvesicles budding off from these structures also contained somedetectable precipitate (Fig. 7).

Our findings suggest that the same bulk endocytic mechanismis involved in short- and long-lasting stimulation, and that thepathway triggered by GLTx is actively involved in recycling, asdemonstrated by the destaining experiment (Fig. 4E). At the frogNMJ, GLTx yields the same ultrastructural changes seen upon short-duration electrical stimulation (Richards et al., 2000), such as theappearance of C-shaped endosomes (Fig. 7H) and larger-diameterendocytic vacuoles, such as those recently described by Claytonand co-workers (Clayton et al., 2008; Clayton et al., 2009). Wecannot rule out from our data that excess membrane could also beredirected to dead-end compartments, as suggested by the presenceof GLTx-induced FM1-43 staining that was resistant to electricalstimulation (Fig. 4E). Consistent with this, together with bulkendosomes surrounded by recycling vesicles, a small number ofmultilamellar structures were also observed using HRP labeling (Fig.7I). At present, we know neither the nature nor the precisecontribution of this pathway to the overall process. However, it istempting to hypothesize that GLTx promotes the recycling ofsynaptic vesicles through bulk endocytosis and that the excessmembrane generated by upregulated synaptic-vesicle fusion isinternalized in a dead-end pathway, as previously described(Heerssen et al., 2008; Kasprowicz et al., 2008).

In conclusion, GLTx represents a powerful research tool toinvestigate endocytic pathways at the nerve terminal. More detailedmolecular characterization of the large vacuolar compartments willbe necessary to further understand the precise function(s) of bulkendosomes and their precise contribution to synaptic-vesiclerecycling.

Materials and MethodsReagents, drugs and toxins-CTx-GVIA and -CTx-MVIIA were purchased from Alomone (Jerusalem, Israel).FM1-43 was purchased from Molecular Probes (Carlsbad, USA) and AM1-43 waspurchased from Biotium (Hayward, USA). All other reagents were from Sigma (CastleHill, Australia).

Purification of GLTxG. convoluta specimens were obtained from the Marine Biological Station, Roscoff(France). GLTx was prepared as previously described and tested for its ability tostimulate catecholamine release from bovine adrenal chromaffin cells (Meunier etal., 2002). Protein concentration was determined using the Bradford assay (BioRad,Hercules, USA).

Journal of Cell Science 123 (7)

Jour

nal o

f Cel

l Sci

ence

1139GLTx promotes synaptic vesicle recycling

Tissue preparationAmphibian NMJ preparations were used. In these preparations, N-type (Cav2.2) Ca2+

channels are responsible for eliciting Ca2+-induced neurotransmitter release (Van derKloot and Molgo, 1994). Electrophysiology was carried out using Rana esculenta.Ca2+ imaging was done on Rana pipiens and the FM1-43 study was carried out onBufo marinus (School of Biomedical Sciences Teaching Facility, University ofQueensland, Australia). Adult frogs (R. pipiens) were anesthetized in a 0.1% tricainemethane sulfonate solution and killed by double pithing. The cutaneous pectoris (CP)muscles with their motor nerve were dissected and bathed in normal frog Ringer’ssolution (NFR; in mM: 116 NaCl, 2 KCl, 1.8 CaCl2, 1 MgCl2, 5 HEPES-OH, pH7.4). To load the Ca2+-sensitive dye Calcium Green-1 (3000 Da dextran conjugate;Molecular Probes), the motor nerve was cut approximately 0.5 mm from the edgeof the muscle. The cut end of the nerve was immersed in a drop of 30 mM CalciumGreen-1 for 7-8 hours at room temperature. The preparation was then washed in NFRand incubated at 4°C for 2 hours before being pinned down on an elevated Sylgardplatform (Dow Corning, Midland, USA) at the centre of a recording chamber. Thenerve was stimulated through a suction electrode at five times the threshold intensityrequired to cause muscle twitching. Labeling of postsynaptic nicotinic ACh receptorswith 2 g/ml AlexaFluor594-conjugated -bungarotoxin (-BTX) (Molecular Probes)for 10 minutes was used to focus on the active zone and evaluate the z-axis drift overthe course of data collection. Muscle contractions were blocked by adding 10 MD-tubocurarine. Superficially positioned nerve terminals in a single focal plane werechosen for study.

Calcium imaging and analysisThe wavelengths and timing of the laser illumination (Krypton-Argon laser; Innova70 Spectrum, Coherent, Hilton, Australia) were selected using an acousto-optic tunablefilter (AOTF; ChromoDynamics, Chicago, IL, USA). The laser was fiber coupled tothe illumination port of an upright fluorescence microscope (Lumplan/FL IR,Olympus). A 100� water-immersion objective (1.0 NA) was used and provided aspatial resolution of ~333 nm. Calcium Green-1 was excited with the 488 nm laserline and emitted light was collected through a 530±20 nm filter. AlexaFluor594–-BTX was excited with the 567 nm line and light was collected through a 620±30 nmemission filter. All images were collected using a liquid-nitrogen-cooled, back-thinnedCCD camera (LN1300B, Roper Scientific, Tucson, USA), which enabled the high-sensitivity detection of fluorescence signal with low noise. Using 20 Hz electricalnerve stimulation, only those terminals with a strong nerve-stimulation-evoked Ca2+

signal were chosen for further experiments. To detect spontaneous Ca2+ oscillations,50-100 fluorescence images were collected at 0.5 Hz for each nerve terminal in theabsence of nerve stimulation. For each image, the laser illuminated the calcium-sensitive dye for 1 ms. All image processing and analyses were performed usingMATLab. Before analysis, images were coregistered to correct for slight movementof the preparation during data collection. To measure GLTx-induced spontaneousCa2+ fluctuations within the active-zone regions of the adult NMJ, the average pixelintensities over the entire active zone were calculated. The region of interest for eachactive zone was defined by correlating the banding pattern of -BTX labeling withthe average Ca2+ image (see Fig. 2A). To assess background Ca2+ oscillations withinindividual active zones before GLTx treatment, a histogram of fluorescence intensitywas plotted and fitted with a Gaussian distribution. The mean of the 95% range offluorescence intensity (±2 s.d.) was taken as baseline fluorescence. Differences frombaseline were determined for individual images by subtracting this baselinefluorescence. The resulting ‘differential images’ represented the relative fluorescencechanges, computed as F/F0(F–F0)/F0, and were displayed in pseudo-colors. TheCV in the fluorescence within each active zone was calculated as the ratio of standarddeviation in F divided by mean fluorescence intensity. The magnitude of Ca2+

oscillations was determined by comparison with the mean value of CV. The frequencyof spontaneous Ca2+ fluctuations following GLTx treatment was calculated by countingthe number of fluorescence intensity excursions that exceeded 2�s.d. of the baselinefluorescence.

ElectrophysiologyMEPPs were recorded at room temperature (22°C) with an intracellular glass capillarymicroelectrode filled with 3 M KCl (8–12 ) using conventional techniques. Themotor nerve of the isolated NMJ was stimulated via a suction microelectrode. Afteramplification, electrical signals were digitized, displayed on an oscilloscope andsimultaneously recorded on videotape with the aid of a modified digital audioprocessor (Sony PCM 701 ES) and a video-cassette recorder (Sony SLC 9F). Datawere collected and analyzed with SCAN software kindly provided by John Dempster(University of Strathclyde, UK) running on a PC equipped with an analogue-digitalconverter (Model DT2821, Data Translation, Marlboro, USA). Statistical analysiswas performed using the Student’s t-test (two tailed, unpaired). Values are expressedas mean±s.e.m. and data were considered significant at P<0.05.

FM-143 imagingA 5 minute pulse of FM1-43 (10 M) was applied to toad NMJ preparations in NFRsolution in either the absence or presence of GLTx or high K+ (30 mM). Preparationswere then washed five times with cold, Ca2+-free, EGTA Ringer’s solution. NMJswere visualized and images acquired on a Zeiss 510 meta laser scanning confocal

microscope. For destaining experiments, the NMJ preparations were washed withNFR solution five times prior to electrical stimulation.

HRP accumulationToad NMJ preparations were treated with GLTx and MEPP frequencies were recordedto confirm toxin activity. 10 mg/ml HRP (Sigma, type VI) in Ringer’s solution wasapplied to preparations and incubated for a further 5 hours. Muscles were brieflywashed in Ringer’s solution and then fixed in 4% paraformaldehyde, 2.5%glutaraldehyde in 0.1 M phosphate buffer for 2 hours at room temperature. Thepreparations were then washed in 0.1 M PBS three times for 20 minutes, followedby washing in acetate buffer (pH 6.0) for 5 minutes. Sections were incubated for 15minutes in a 2% NiSO4 solution (in sodium acetate buffer) containing 2 mg/ml D-glucose, 0.4 mg/ml NH4Cl and 0.025% diaminobenzidine (DAB). Glucose oxidase(0.2 l/ml) was added to the nickel-DAB solution, resulting in the production ofH2O2 and the deposition of nickel-DAB. Sections were washed in sodium acetatebuffer to stop the reaction and then washed thoroughly in PBS (three times for 15minutes).

PhotoconversionOur protocol was adapted from previous studies (Harata et al., 2001; Richards et al.,2003). AM1-43-labeled preparations were fixed in PBS containing 2% glutaraldehydefor 2 hours. Muscles were then quickly washed twice in PBS, quenched in PBScontaining 100 mM glycine pH 7.4 for 1 hour, soaked in PBS containing 100 mMammonium chloride for 5 minutes, and further washed in PBS for 30 minutes. Sampleswere treated with 1.5 mg/ml DAB diluted in PBS for 30 minutes in the dark. Thepreparation was then illuminated under UV light through a 10� air objective (CarlZeiss, Oberkochen, Germany) from a 100 W mercury lamp for 45-70 minutes,depending on the rate of formation of the brown precipitate.

Electron microscopyMuscle preparations were fixed in 2.5% glutaraldehyde, post-fixed in 1% osmiumtetroxide, followed by ethanol dehydration and Epon resin embedding (ProSciTech,Thuringowa, Australia). Cross-sections (60 nm thick) were stained with uranyl acetateand Reynold’s lead citrate. NMJs were examined using a JEOL 1010 transmissionelectron microscope.

This work was supported by a grant from the Australian ResearchCouncil (F.A.M. and T.H.N.), a fellowship to F.A.M. from the NationalHealth and Medical Research Council of Australia, by Cancer ResearchUK (G.S.) and by US National Institutes of Health R01 NS043396 toS.D.M. Deposited in PMC for release after 12 months.

ReferencesBalaji, J. and Ryan, T. A. (2007). Single-vesicle imaging reveals that synaptic vesicle

exocytosis and endocytosis are coupled by a single stochastic mode. Proc. Natl. Acad.Sci. USA 104, 20576-20581.

Benoit, E., Mattei, C., Ouanounou, G., Meunier, F. A., Suput, D., Le Gall, F., Marquais,M., Dechraoui, M. Y. and Molgó, J. (2002). Ionic mechanisms involved in the nodalswelling of myelinated axons caused by marine toxins. Cell. Mol Biol. Lett 7, 317-321.

Betz, W. J., Mao, F. and Bewick, G. S. (1992). Activity-dependent fluorescent stainingand destaining of living vertebrate motor nerve terminals. J. Neurosci. 12, 363-375.

Ceccarelli, B. and Hurlbut, W. P. (1980). Ca2+-dependent recycling of synaptic vesiclesat the frog neuromuscular junction. J. Cell Biol. 87, 297-303.

Ceccarelli, B., Hurlbut, W. P. and Mauro, A. (1973). Turnover of transmitter and synapticvesicles at the frog neuromuscular junction. J. Cell Biol. 57, 499-524.

Clark, A. W., Mauro, A., Longenec, H. E. J. and Hurlbut, W. P. (1970). Effects of blackwidow spider venom on frog neuromuscular junction. Effects on fine structure of frogneuromuscular junction. Nature 225, 703-705.

Clayton, E. L., Evans, G. J. and Cousin, M. A. (2008). Bulk synaptic vesicle endocytosisis rapidly triggered during strong stimulation. J. Neurosci. 28, 6627-6632.

Clayton, E. L., Anggono, V., Smillie, K. J., Chau, N., Robinson, P. J. and Cousin, M.A. (2009). The phospho-dependent dynamin-syndapin interaction triggers activity-dependent bulk endocytosis of synaptic vesicles. J. Neurosci. 29, 7706-7717.

Colasante, C. and Pécot-Dechavassine, M. (1996). Ultrastructural distribution ofsynaptophysin and synaptic vesicle recycling at the frog neuromuscular junction. J.Neurosci. Res. 44, 272-282.

Colasante, C., Meunier, F. A., Kreger, A. S. and Molgó, J. (1996). Selective depletionof clear synaptic vesicles and enhanced quantal transmitter release at frog motor nerveendings produced by trachynilysin, a protein toxin isolated from stonefish (Synanceiatrachynis) venom. Eur. J. Neurosci. 8, 2149-2156.

Couteaux, R. and Pécot-Dechavassine, M. (1970). [Synaptic vesicles and pouches at thelevel of “active zones” of the neuromuscular junction]. C. R. Acad. Sci. Hebd. SeancesAcad. Sci. D 271, 2346-2349.

Davletov, B. A., Meunier, F. A., Ashton, A. C., Matsushita, H., Hirst, W. D., Lelianova,V. G., Wilkin, G. P., Dolly, J. O. and Ushkaryov, Y. A. (1998). Vesicle exocytosisstimulated by alpha-latrotoxin is mediated by latrophilin and requires both external andstored Ca2+. EMBO J. 17, 3909-3920.

de Lange, R. P., de Roos, A. D. and Borst, J. G. (2003). Two modes of vesicle recyclingin the rat calyx of Held. J. Neurosci. 23,10164-10173.

Jour

nal o

f Cel

l Sci

ence

1140

de Paiva, A., Meunier, F. A., Molgó, J., Aoki, K. R. and Dolly, J. O. (1999). Functionalrepair of motor endplates after botulinum neurotoxin type A poisoning: biphasic switchof synaptic activity between nerve sprouts and their parent terminals. Proc. Natl. Acad.Sci. USA 96, 3200-3205.

de Toledo, G. A., Fernandez-Chacon, R. and Fernandez, J. M. (1993). Release ofsecretory products during transient vesicle fusion. Nature 363, 554-558.

Dickman, D. K., Horne, J. A., Meinertzhagen, I. A. and Schwarz, T. L. (2005). Aslowed classical pathway rather than kiss-and-run mediates endocytosis at synapseslacking synaptojanin and endophilin. Cell 123, 521-533.

Fernandez, J. M., Neher, E. and Gomperts, B. D. (1984). Capacitance measurementsreveal stepwise fusion events in degranulating mast cells. Nature 312, 453-455.

Fesce, R., Grohovaz, F., Valtorta, F. and Meldolesi, J. (1994). Neurotransmitter release:fusion or ‘kiss-and-run’? Trends Cell Biol. 4, 1-4.

Gaffield, M. A., Rizzoli, S. O. and Betz, W. J. (2006). Mobility of synaptic vesicles indifferent pools in resting and stimulated frog motor nerve terminals. Neuron 51, 317-325.

Granseth, B., Odermatt, B., Royle, S. J. and Lagnado, L. (2009). Comment on “Thedynamic control of kiss-and-run and vesicular reuse probed with single nanoparticles”.Science 325, 1499.

Harata, N., Ryan, T. A., Smith, S. J., Buchanan, J. and Tsien, R. W. (2001). Visualizingrecycling synaptic vesicles in hippocampal neurons by FM 1-43 photoconversion. Proc.Natl. Acad. Sci. USA 98, 12748-12753.

Harata, N. C., Choi, S., Pyle, J. L., Aravanis, A. M. and Tsien, R. W. (2006). Frequency-dependent kinetics and prevalence of kiss-and-run and reuse at hippocampal synapsesstudied with novel quenching methods. Neuron 49, 243-256.

Heerssen, H., Fetter, R. D. and Davis, G. W. (2008). Clathrin dependence of synaptic-vesicle formation at the Drosophila neuromuscular junction. Curr. Biol. 18, 401-409.

Henkel, A. W., Lubke, J. and Betz, W. J. (1996). FM1-43 dye ultrastructural localizationin and release from frog motor nerve terminals. Proc. Natl. Acad. Sci. USA 93, 1918-1923.

Heuser, J. E. and Reese, T. S. (1973). Evidence for recycling of synaptic vesicle membraneduring transmitter release at the frog neuromuscular junction. J. Cell Biol. 57, 315-344.

Heuser, J. E. and Reese, T. S. (1981). Structural changes after transmitter release at thefrog neuromuscular junction. J. Cell Biol. 88, 564-580.

Kasprowicz, J., Kuenen, S., Miskiewicz, K., Habets, R. L., Smitz, L. and Verstreken,P. (2008). Inactivation of clathrin heavy chain inhibits synaptic recycling but allowsbulk membrane uptake. J. Cell Biol. 182, 1007-1016.

Lin, M. Y., Teng, H. and Wilkinson, R. S. (2005). Vesicles in snake motor terminalscomprise one functional pool and utilize a single recycling strategy at all stimulusfrequencies. J. Physiol. 568, 413-421.

Lynch, K. (1982). The effects of chronic stimulation on the morphology of the frogneuromuscular junction. J. Neurocytol. 11, 81-107.

Madeddu, L., Meldolesi, J., Pozzan, T., Cardona Sanclemente, L. E. and Bon, C. (1984).Alpha-latrotoxin and glycerotoxin differ in target specificity and in the mechanism oftheir neurotransmitter releasing action. Neuroscience 12, 939-949.

Matteoli, M., Haimann, C., Torri-Tarelli, F., Polak, J. M., Ceccarelli, B. and De Camilli,P. (1988). Differential effect of alpha-latrotoxin on exocytosis from small synaptic vesiclesand from large dense-core vesicles containing calcitonin gene-related peptide at the frogneuromuscular junction. Proc. Natl. Acad. Sci. USA 85, 7366-7370.

Meshul, C. K. and Pappas, G. D. (1984). The relationship of pinocytosis and synapticvesicles at the frog neuromuscular junction. Brain Res. 290, 1-18.

Meunier, F. A., Feng, Z. P., Molgó, J., Zamponi, G. W. and Schiavo, G. (2002).Glycerotoxin from Glycera convoluta stimulates neurosecretion by up-regulating N-typeCa2+ channel activity. EMBO J. 21, 6733-6743.

Morel, N., Thieffry, M. and Manaranche, R. (1983). Binding of a Glycera convolutaneurotoxin to cholinergic nerve terminal plasma membranes. J. Cell Biol. 97, 1737-1744.

Neher, E. and Marty, A. (1982). Discrete changes of cell membrane capacitance observedunder conditions of enhanced secretion in bovine adrenal chromaffin cells. Proc. Natl.Acad. Sci. USA 79, 6712-6716.

Ribchester, R. R., Mao, F. and Betz, W. J. (1994). Optical measurements of activity-dependent membrane recycling in motor nerve terminals of mammalian skeletal muscle.Proc. Biol. Sci. 255, 61-66.

Richards, D. A., Guatimosim, C. and Betz, W. J. (2000). Two endocytic recycling routesselectively fill two vesicle pools in frog motor nerve terminals. Neuron 27, 551-559.

Richards, D. A., Guatimosim, C., Rizzoli, S. O. and Betz, W. J. (2003). Synaptic vesiclepools at the frog neuromuscular junction. Neuron 39, 529-541.

Rizzoli, S. O. and Betz, W. J. (2004). The structural organization of the readily releasablepool of synaptic vesicles. Science 303, 2037-2039.

Rizzoli, S. O. and Betz, W. J. (2005). Synaptic vesicle pools. Nat. Rev. Neurosci. 6, 57-69.

Schenning, M., Proctor, D. T., Ragnarsson, L., Barbier, J., Lavidis, N. A., Molgo, J.J., Zamponi, G. W., Schiavo, G. and Meunier, F. A. (2006). Glycerotoxin stimulatesneurotransmitter release from N-type Ca2+ channel expressing neurons. J. Neurochem.98, 894-904.

Schiavo, G., Matteoli, M. and Montecucco, C. (2000). Neurotoxins affectingneuroexocytosis. Physiol. Rev. 80, 717-766.

Smith, C. B. and Betz, W. J. (1996). Simultaneous independent measurement ofendocytosis and exocytosis. Nature 380, 531-534.

Tsang, C. W., Elrick, D. B. and Charlton, M. P. (2000). -Latrotoxin releases calciumin frog motor nerve terminals. J. Neurosci. 20, 8685-8692.

Van der Kloot, W. and Molgó, J. (1994). Quantal acetylcholine release at the vertebrateneuromuscular junction. Physiol. Rev. 74, 899-991.

Verstreken, P., Kjaerulff, O., Lloyd, T. E., Atkinson, R., Zhou, Y., Meinertzhagen, I.A. and Bellen, H. J. (2002). Endophilin mutations block clathrin-mediated endocytosisbut not neurotransmitter release. Cell 109, 101-112.

Xia, X., Lessmann, V. and Martin, T. F. (2009). Imaging of evoked dense-core-vesicleexocytosis in hippocampal neurons reveals long latencies and kiss-and-run fusion events.J. Cell Sci. 122, 75-82.

Zhang, Q., Li, Y. and Tsien, R. W. (2009). The dynamic control of kiss-and-run andvesicular reuse probed with single nanoparticles. Science 323, 1448-1453.

Journal of Cell Science 123 (7)

Jour

nal o

f Cel

l Sci

ence