Kiss-and-run, Collapse and ‘Readily Retrievable’ Vesicles


Reinhard Jahn,


Two models of synaptic vesicle recycling have been intensely debated for decades: kiss-and-run, in which the vesicle opens and closes transiently, presumably through a small fusion pore, and full fusion, in which the vesicle collapses into the plasma membrane and is retrieved by clathrin-coat-dependent processes. Conceptually, it seems that kiss-and-run would be faster and would retrieve vesicles with greater fidelity. Is this the case? This review discusses recent evidence for both models. We conclude that both mechanisms allow for high fidelity of vesicle recycling. Also, the presence in the plasma membrane of a depot of previously fused vesicles that are already interacting with the endocytotic machinery (the ‘readily retrievable’ vesicles) allows full fusion to trigger quite fast endocytosis, further blurring the efficiency differences between the two models.

Two main models of vesicle recycling: Collapse or kiss-and-run

The first model of synaptic vesicle exocytosis, introduced in 1955 by del Castillo and Katz, imagined synaptic vesicles to fuse transiently with the membrane at a specialized spot (1). The first systematic studies of recycling were performed by electron microscopy (EM), resulting in two seminal papers published in 1973, one by Heuser and Reese and the second by Ceccarelli et al. (2,3). In both studies, frog neuromuscular junctions were stimulated in the presence of markers such as horseradish peroxidase, allowing the identification of the endocytosed organelles.

Heuser and Reese observed that tetanic (10 Hz) stimulation decreased the number of synaptic vesicles in the end plate. In parallel, the plasma membrane dilated by about 20%, the number of coated vesicles tripled and the proportion of large membrane-bound organelles (termed cisternae) increased more than sixfold. These structures slowly disappeared after stimulation and were replaced by newly formed (labeled) synaptic vesicles. Heuser and Reese proposed that vesicles collapse into the plasma membrane during exocytosis and are then retrieved by coated pits and vesicles. The coated vesicles fused with each other and with the cisternae (which thus represent sorting endosomes in modern terminology), which then bud into new synaptic vesicles (Figure 1F).

Figure 1.

Figure 1.

Possible pathways of synaptic vesicle recycling. The speed and fidelity of the recycling process tend to decrease from A to F (see main text for details). A) Kiss-and-run. B) Collapse of one vesicle and retrieval of another one from the ‘readily retrievable’ pool. C) Collapse and (classical) CME. D) Bulk retrieval: strong stimulation followed by formation of infoldings, which are broken up into vesicles by CME. E) Collapse of a vesicle followed by brief dispersal into a few patches, which are recovered by interaction with the CME machinery. F) Collapse, followed by complete dispersal of vesicle molecules. They are retrieved by CME but need endosomal sorting to be made into new vesicles.

In complete contrast, Ceccarelli et al. (3) observed no vesicle depletion and no substantial increases in coated vesicles or large organelles – just an increase in the number of labeled vesicles over time. The model they proposed was essentially identical to Katz's initial model (Figure 1A); it became known much later as kiss-and-run (4). The simple explanation for the difference between the two sets of findings is that Ceccarelli et al. used a low frequency of stimulation (2 Hz), which apparently is not rate limiting for the vesicle reformation machinery, and thus they investigated a system which was at steady state at all times (reviewed in 5). The number of endocytic intermediates under these conditions is probably too low at any time-point to be clearly identified in EM.

With respect to the original Heuser model, it was soon recognized that no real evidence existed for the fusion of the coated organelles with the cisternae – alternatively, a coated vesicle linked to a cisterna may represent budding of the coated vesicle from the larger membrane. Thus, a simplified recycling model was proposed in which the second fusion step was eliminated (6,7). In this scenario, membrane is retrieved in the form of coated vesicles, which form new synaptic vesicles simply by uncoating (Figure 1C). The large organelles are simply infoldings of the plasma membrane (bulk endocytosis). They are broken into vesicles by clathrin-coat-dependent mechanisms (Figure 1D). This variant of the Heuser–Reese model involves only one fusion event – between the vesicle and the plasma membrane – and it has been supported by later studies on many other synapses (5).

Current status of the field

More than 30 years after the original proposals, the debate between supporters of the different models is still going strong. Using increasingly sophisticated methods, evidence for both models is continuously being published – only to be contested later. In the following paragraphs, we will summarize recent pro and con evidence for both models and discuss that not only the kiss-and-run but also the ‘classical’ pathway is capable of explaining fast, stimulus-associated endocytosis.

Collapse and clathrin-mediated endocytosis

Collapse of vesicles into the plasma membrane, at least during strong stimulation, has been well established in numerous preparations. Furthermore, there is plenty of evidence that clathrin- and dynamin-dependent retrieval of synaptic vesicles is essential for the vesicle cycle (8,9). Typical experiments supporting full fusion are those in which strongly stimulated preparations are shown to lose some of their vesicles [with or without an enlargement of the plasma membrane (10)]. Perturbing endocytosis, as in the dynamin temperature-sensitive mutant of Drosophila, shibire, also results in empty nerve terminals (11). Endocytic intermediates such as clathrin-coated vesicles and especially infoldings/cisternae have been observed in many preparations (reviewed in 5 and references therein).

There is one flaw with this type of evidence, which has been pointed out for decades. Harsh stimulation is requiredto see such intermediates, and thus they may not be representative for the processes functioning in vivo. Also, one of the most common preparations used for studies of synaptic activity, cultured hippocampal neurons, is resistant to such approaches – extensive stimulation is unable to reduce the number of vesicles significantly, and endocytic intermediates are not frequently observed (12,13; but also see 14). Even in preparations where clathrin-mediated endocytosis (CME) is well established, it is not trivial to correlate quantitatively the number of coated vesicles or pits with the number of released vesicles (15); a possible answer is that endocytic intermediates are normally short lived and thus are difficult to see with mild stimulation.

Full fusion is also supported by the uptake of bulky molecules such as antibodies into synaptic vesicles during stimulation (16,17); even sandwiches of primary and secondary antibodies can be internalized (18), which is hard to imagine to be accomplished through a narrow fusion pore. Furthermore, in an elegant set of experiments, the collapse of vesicles has recently been investigated by capacitance recording (which measures the total nerve terminal membrane area) in conjunction with interference reflection microscopy (19), which measures the size of the footprint of the nerve terminal. The two measurements paralleled each other in the large nerve terminals of goldfish bipolar cells, suggesting that every fusing vesicle also contributes to the increase in the footprint. Interestingly, some rapid endocytosis does persist in presence of clathrin-perturbing agents in this preparation (20), even though bulk endocytosis [abundant here (5)] might be called in rather than kiss-and-run [note that bulk endocytosis can be extremely rapid (7)].

At least in some synapses, molecular interference with CME can block all vesicle retrieval, suggesting that it is the main recycling mechanism. For instance, overexpression of a dominant-negative form of the clathrin adaptor protein (AP) 180, or RNA interference knock down of clathrin heavy chain, blocked all endocytosis following stimulation in hippocampal synapses (21). However, the treated nerve terminals did have vesicles ready to release at the start of stimulation – after having their CME inhibited for hours (during which spontaneous activity must have released a significant number of vesicles) – indicating that either the treatments drastically prolonged CME, but did not block it completely, or there are also other types of endocytosis.


A major problem associated with the kiss-and-run model is that because of its transient and short-lived nature, it is almost impossible to ‘capture’ such events in EM. In synapses, size and packing density of synaptic vesicles render it impossible to image individual vesicles (unless selectively labeled, see below), and EM only provides ‘frozen’ snapshots. Images of recently fused vesicles, immediately after stimulation, have been obtained in quick-freeze EM (22,23). Pores are large, around 20 nm in diameter (for synaptic vesicles of only approximately 50 nm in diameter) and tend to grow larger with time (22), suggestive of full fusion progressing after formation of the initial pore. Furthermore, observing (labeled) vesicles seemingly in the process of getting endocytosed but lacking obvious clathrin coats should not be used as kiss-and-run evidence, especially as imaging clathrin coats is not trivial (24).

Rapid endocytosis (∼1–10 seconds) has been used as evidence for kiss-and-run as CME is expected to take many seconds. Whether CME does need many seconds to proceed is debatable, especially if there are any preformed clathrin-coated vesicles on the membrane waiting to be endocytosed (7,25; see below). Nevertheless, rapid endocytosis has been observed by capacitance recording at the calyx of Held in the auditory pathway (26) with a time constant of approximately 100 milliseconds, far beyond CME ability. However, a subsequent study with the same preparation used neurotoxins to block exocytosis and indicated that the capacitance transient interpreted as kiss-and-run may not have been related to vesicle recycling (27), inducing a re-evaluation of the original data, which suggest that endocytosis is more than an order of magnitude slower than originally thought (28).

Stronger evidence for kiss-and-run was recently obtained by cell-attached capacitance recording, in the form of capacitance ‘flickers’, in which the signal increases, remains constant between a few milliseconds and about 2 seconds and then abruptly returns to baseline. This paradigm corresponds to the fusion and subsequent reinternalization of single synaptic vesicles and seems to apply for approximately 20% of the vesicles (29; see also 30 for a similar discussion of microvesicles in pituitary nerve terminals). It is highly unlikely that a single vesicle would be retrieved by CME within such a short time period following exocytosis. In these experiments, however, it cannot be fully excluded that the exocytosed and endocytosed vesicles are different – i.e. one vesicle fused and another one was subsequently retrieved (see also discussion of readily retrievable vesicles below).

The strongest (although contested) support for kiss-and-run has been derived from experiments using synaptic preparations with fluorescently labeled synaptic vesicles. Synaptic vesicles can be loaded during recycling with various fluorescent fluid-phase or membrane labels, among which styryl FM dyes are the most common. The release of the fluorescent dye upon a subsequent round of stimulation (destaining) has been used as a direct measure of exocytosis for more than a decade. Some FM dyes are more hydrophilic than others, and it was thought that they would be released preferentially in kiss-and-run events, while collapse would allow all dyes to escape the vesicle equally well. Indeed, the hydrophobic dyes FM 1–84 and FM 1–43 were released more slowly than the hydrophilic FM 2–10 from hippocampal nerve terminals (31). However, the difference between the dyes was not reproduced in a different preparation [frog neuromuscular junction (NMJ); 32], and no difference was later seen in hippocampal synapses at physiological temperature (33).

In a similar experiment, bromophenol blue (BPB) was added to the bath when vesicles preloaded with FM dyes exocytosed. Bromophenol blue quenches all FM fluorescence but is membrane impermeant; however, as a small molecule, it is expected to penetrate into the vesicles through fusion pores that otherwise are too short lived or narrow for FM dyes to escape (34). If vesicles undergo full collapse and lose all their dye, the drop in FM fluorescence during exocytosis should not be affected by BPB. The experiment indicated that a significant proportion of dye was retained in a BPB-sensitive pool, even at physiological temperature. The interpretation of the BPB-sensitive pool as ‘kiss-and-run’ vesicles, however, has been challenged because the dye remaining in the preparation may be trapped in the synaptic cleft and in the pre- and postsynaptic membranes instead of in the synaptic vesicles (21), especially as this type of background contributes significantly to the overall nerve terminal fluorescence (24).

The loss of FM fluorescence from single individual vesicles has also been investigated. In nerve terminals of the bipolar cells from goldfish retina, single vesicles appear to lose all the dye within milliseconds (35), suggestive of collapse. Different results have been obtained in hippocampal neurons when it was attempted to label only a single vesicle per bouton (36). Stimulation resulted in partial loss of dye, indicative of transient fusion, with few boutons losing all fluorescence at once. Subsequently, it was reported that the smaller destaining events were also slower, in good agreement with the loss of dye through a narrow pore (37). However, this interpretation has also been challenged recently by showing that weakly stained boutons released all of their dye within less than 100 milliseconds (38) and that partial release is only observable in more strongly labeled boutons, which would suggest that ‘partial’ release simply reflects the release of one of the few labeled vesicles.

Finally, fusion pores too transient to allow for FM uptake were also proposed. Stimulation with hypertonic sucrose resulted in little FM dye uptake but normal neurotransmitter release (39). In contrast, however, no difference in FM uptake between application of hypertonic sucrose and electrical stimulation was found in neuromuscular preparations (24).

Some of the newest tools – pHluorins – have also been employed in attempting to differentiate between collapse and kiss-and-run. PHluorins are synaptic vesicle proteins fused to a pH-sensitive variant of the green fluorescent protein (GFP). The GFP moiety is linked to an intravesicular domain of the protein, and its fluorescence is quenched by the acidic pH of the vesicle. Upon exocytosis, the fluorescence increases dramatically because of the exposure to the neutral extracellular pH. Endocytosis is studied by measuring the requenching of the pHluorins.

The protein most commonly used for such studies is synaptobrevin, the GFP chimera being termed synaptopHluorin. Gandhi and Stevens (25) analyzed the endocytosis behavior after release of single synaptic vesicles and found that a fraction of the events returned to baseline with half-times of less than a second. As this delay covers not only endocytosis but also the ensuing reacidification, the data were interpreted as evidence for kiss-and-run. However, this type of analysis, again, is not trivial. The majority of action potentials did not trigger (measurable) release of vesicles at normal calcium concentrations and were recorded as failures. Other action potentials were followed by rapid upward deflections of the signal, which returned to baseline within one or a few frames, and were recorded as kiss-and-run events (25, Fig. 4e). We agree with the suggestion of Granseth et al. (21) that it cannot be excluded that some of these events represent upward flicking noise, which coincided with stimulation. Also, lateral diffusion of synaptopHluorin after exocytosis may cause rapid loss of signal from the region of interest.

Unlike CME whose molecular details are well worked out (8), it has been difficult to identify proteins functioning in kiss-and-run. Dynamin is generally considered to be required for all endocytosis (27,40). A candidate protein with a specific function in kiss-and-run is endophilin, whose inactivation was proposed to switch vesicles to kiss-and-run (41), although later studies are more in agreement with a slowing of CME in such preparations (42,43).

Kiss-and-run and full fusion: Are they so different in the end?

The major reason for promoting kiss-and-run, despite the still rather circumstantial evidence, is the notion that the slow kinetics and comparably long turnaround times of CME are insufficient to meet the demands of a fast synapse. Kiss-and-run provides an elegant explanation for this problem, and it has the further advantage that the identity of the synaptic vesicle is perfectly maintained, avoiding the need for time- and energy-consuming sorting and recycling mechanisms. However, how strong is the evidence in support of this notion?

Speed of recycling may not be very different under moderate stimulation

Kiss-and-run offers obvious recycling advantages – it is supposedly faster and more energy efficient than full fusion. It could improve neurotransmitter release and vesicle reuse, especially upon strong stimulation (see model by Harata et al. 44). FM dye release revealed that some vesicles can indeed be reused within a few seconds after the first fusion, as if they remained ‘in place’ at the active zone, like kiss-and-run predicts (45). However, they lost this ability relatively fast, by mixing into the general pool of vesicles (45), which may explain why the rapid reuse was not detected later in synaptopHluorin experiments (46).

Single-vesicle imaging revealed that kiss-and-run vesicles did not rerelease under mild stimulation conditions for more than approximately 20 seconds after the first fusion, on average (36). Thus, kiss-and-run may not actually be much faster than full fusion in creating new fusion-competent vesicles. Measured rates of vesicle replenishment are relatively low and are in line with full fusion kinetics (47). Accordingly, a full fusion mechanism of exocytosis is sufficiently fast to balance exocytosis with stimulation as strong as 10 Hz for 30 seconds (33). One may conclude from this type of data that both mechanisms can retrieve vesicles efficiently, at least with moderate stimulation.

Maintaining vesicle identity: Both models can do it

Kiss-and-run recycles vesicles with great fidelity. It was long thought that collapse results in complete mixing of the two membranes, with the vesicle losing its identity, which goes against the fact that the synaptic vesicle has a very specific composition (48), quite different from the plasma membrane. This is not the case: extensive recycling does not result in the overall loss of vesicle components into the plasma membrane (49) or in the enrichment of plasma membrane components in the vesicles (50), even if full fusion plays a significant role in the preparations investigated (5,33).

How would recovery proceed after collapse? The first direct investigation of the process was performed in rapid-freezing and freeze-fracture experiments at the frog NMJ in John Heuser's experiments (7,22). Within milliseconds after stimulation, large membrane particles, closely resembling particles from synaptic vesicles, appeared on the plasma membrane near-active zones (occasionally in small pits seemingly formed by newly fused vesicles). Clusters of particles (indicative of grouped vesicle molecules), as well as single particles, were seen, although the authors themselves noted that assigning clusters was somewhat arbitrary. Within seconds, the particles were found again in new pits on the plasma membrane (generally away from the active zones), which were interpreted as clathrin-coated pits. A first wave of CME was extremely fast, peaking at approximately 1 second after stimulation and ending in approximately 10 seconds.

Thus, vesicles collapsed in the plasma membrane and some molecules may have dispersed into it, but were retrieved within seconds or tens of seconds by clathrin-coated pits. Even if the vesicle lost its components (particles) into the plasma membrane upon fusion, they were recovered before endocytosis, ensuring that a similar new vesicle was formed.

How is the vesicle targeted by the endocytic machinery? Signals for CME recognition have proved elusive so far for most vesicle proteins (reviewed by Jung and Haucke in this issue of Traffic 51). One such interaction has been thoroughly characterized with stonin 2 binding both synaptotagmin and the clathrin adaptor complex AP2 (52) and participating in synaptotagmin internalization (53). The fact that synaptotagmin is essential for endocytosis (54) makes for a particularly attractive hypothesis: after fusion, vesicle components remain largely together, with vesicles possibly breaking into a few subdomains or losing a minority of components; the synaptotagmin-containing vesicle patches are targeted by stonin 2, AP2 and the CME machinery, forming a ready-to-internalize vesicle (Figure 1E). The vesicle is retrieved either immediately or upon subsequent stimulation (25).

Readily retrievable vesicles

When a vesicle fuses, is another one retrieved?

As much as 30% of synaptopHluorin may reside in the plasma membrane (21,55). About 8% of synaptophysin–pHluorin (21) and approximately 20% of synaptotagmin–pHluorin molecules (56) are also found here. Significant levels of native protein can be found as well (17,18; detection may not always be trivial, 16). These values are quite significant when compared with the total number of recycling vesicles [approximately 20%; (5)], making the fusion of one vesicle and the endocytosis of another one a real possibility. This was recently investigated using an elegant experimental setup: either nerve terminals containing synaptopHluorin or synaptotagmin–pHluorin are bleached, eliminating surface fluorescence, or their surface pHluorin moieties are proteolytically cleaved (56,57). Then the nerve terminals are stimulated, and the reacidification of the pHluorin is monitored. Both groups found that reacidification is not complete after eliminating surface fluorescence, as if newly fused vesicular material remains on the plasma membrane, while preexisting, bleached molecules are internalized instead. The data argue strongly against kiss-and-run endocytosis. Perhaps more importantly, they indicate that collapse/CME could be quite fast, if vesicles different than the ones just fused are to be taken up (Figure 1, model B); note that pits were found to retrieve vesicles within seconds after stimulation (7).

Can such a ‘readily retrievable’ pool of vesicles even be internalized without exocytosis? The answer seems to be yes: botulinum neurotoxin A (cleaving the vesicle fusion protein SNAP-25) blocked synaptic vesicle exocytosis but not endocytosis of a small pool of vesicles [(58); tetanus neurotoxin, however, which cleaves synaptobrevin, appeared to block both exocytosis and endocytosis].

Do synaptic vesicles lose their components into the plasma membrane during collapse, or do they persist as ‘readily retrievable’ vesicles?

SynaptopHluorin and other variants of synaptobrevin–GFP diffuse out of synapses and into the neighboring axon area during strong stimulation (59,60). Within minutes, the synaptobrevin-associated fluorescence returns to the boutons. Are these free synaptobrevin molecules or packets of molecules (fused vesicles)? The diffusion is fast (seconds or tens of seconds), and thus it was suggested that they are single molecules. This is not an infallible argument, however: movement of complete vesiclesfrom neighboring boutons is sufficiently fast to replace approximately 10% of the vesicles in a bouton in only about 60 seconds (61).

Do all vesicular GFP-tagged molecules disperse into axons following stimulation? Synaptotagmin I seems to, although it is unclear how much (56,57). Synaptophysin–pHluorin did not seem to leave the boutons; this was not because of kiss-and-run as the results were in better agreement with full fusion, followed by diffusion to near-active zone areas (within the same boutons) and reinternalization (21). Why is synaptobrevin–GFP more mobile than synaptophysin–GFP? Different vesicular proteins may have different behaviors (with a larger percentage of synaptophysin localizing to vesicles than to synaptobrevin, see above), but it is also possible that the tag perturbs the normal behavior of synaptobrevin.

When the exocytosed vesicle protein was investigated directly by immunofluorescence with diffraction-unlimited optics, synaptotagmin I was found within clusters containing multiple copies, both at rest and during exhaustive stimulation (17). This indicates that either vesicle components remain within clusters after fusion and are later targeted by the endocytic machinery and retrieved or they disperse very briefly and then recluster upon CME activity. Supporting the first hypothesis, vesicle proteins have been reported to adhere to one another after detergent solubilization (62). If all proteins disperse, as assumed by the second hypothesis, each of them needs to be targeted individually by the CME machinery. However, so far an interaction with the CME system has only been found for few vesicle residents (most prominently synaptotagmin), indicating that at least a major proportion of vesicle proteins must remain associated with such proteins to be retrieved. If they all spread in the plasma membrane, away from proteins like synaptotagmin that do have internalization signals, they may be irretrievably lost.

If vesicle proteins remain together after exocytosis, the question arises whether these represent separate entities in the plasma membrane, with no intermixing and forming a ‘readily retrievable’ pool, or whether they break apart and internalization then retrieves vesicles formed from mixed pieces. A firm answer in either direction is difficult to give at present. The photobleaching experiments discussed above (56,57) gave some indications: the kinetics of endocytosis seems to be slower after bleaching, as if quenched (pre-existing) vesicles were favored during endocytosis (56, Fig. 3a). Also, the lower the stimulation, the higher the proportion of pre-existing vesicles that was retrieved; with 40 action potentials, essentially none of the newly released vesicle molecules were retrieved (57). This indicates that there is indeed a pool of ‘readily retrievable’ vesicles, with limited mixing with the newly fused vesicles. What is the mechanism for preferential retrieval of ‘older’ molecules? It might be through preformed clathrin coats, which wait for stimulation to endocytose (7,63).

The bleaching experiment has been performed at the level of single vesicles (25). Nerve terminals expressing synaptopHluorin were bleached to approximately 10% of the initial fluorescence levels and then stimulated. Increases in signal of about 1 quantal unit (one vesicle) were typically followed by returns to baseline, more or less rapidly. Some vesicles remained stranded, as if no endocytosis of the particular vesicle occurred, while occasionally 1-quantum reductionsin signal were observed, interpreted as the stimulated endocytosis of a vesicle, even though no exocytosis took place at the particular bouton. The results do suggest that vesicles act as quanta of membrane, without breaking apart; however, when one vesicle fuses, another one may be retrieved. It needs to be borne in mind, however, that because of the low signal-to-noise in these experiments, other mechanisms involving partial breakup might have escaped detection.

Does endosomal sorting occur in nerve terminals, and does it function as an intermediate in synaptic vesicle recycling?

Of all recycling models, only Heuser and Reese's can deal with this situation. If partly assembled vesicles are generated by CME, they will need to fuse with sorting endosomes before functional vesicles are formed through endosomal budding. This is in line with the fact that synaptic vesicles are highly enriched in endosomal proteins (48), including the membrane organizer Rab5. Two recent reports further support an involvement of endosomes. First, newly endocytosed vesicles isolated from synapses were able to fuse with each other as well as with early endosomes from neuroendocrine cells, with fusion requiring SNAREs specific for endosomal fusion rather than for exocytosis (64). Second, in experiments monitoring vesicular glutamate transporter 1-pHluorin recycling, it was found that a (most likely endosomal) pathway mediated by AP3 is involved in the recycling of at least a fraction of newly released vesicles (65).


While the debate about kiss-and-run versus CME is not going to end soon, CME is catching up by displaying advantages that used to be the realm of the kiss-and-run mechanism. Clathrin-mediated endocytosis involves invaginating membrane, recruiting and assembling the clathrin coat, pinching off and finally disassembling the coat, overall lasting about 30 seconds to 1 min. The presence of a ‘readily retrievable’ pool reduces the process just to pinch-off and coat disassembly. A wave of vesicle pinch off peaks within 1 second after stimulation (7). If the subsequent uncoating is as fast as that measured in dendrites (66), which has a time constant of approximately 4 seconds, CME allows for endocytosis to take place about as fast as the kiss-and-run mechanism (time scale of a few seconds instead of tens of seconds), with the only difference being that in the former case the identity of the endocytosed vesicle is different from that of the exocytosed vesicle (although the two would be fairly similar synaptic vesicles). Thus, while many pathways of clathrin-mediated recycling can be imagined (Figure 1), acting with different degrees of speed and fidelity, the combination of collapse and ‘readily retrievable’ pool gets to more or less level terms with kiss-and-run. New experiments and perhaps new tools (67) are needed to further test whether kiss-and-run is indeed the better model or whether the two models serve the same purpose with fairly similar efficiency.


S. O. R. acknowledges a long-term fellowship from the International Human Frontier Science Program Organization. We thank Sina-Victoria Barysch, Ioanna Bethani and Ulf Geumann for critically reading the manuscript. We hope our colleagues will understand that, because of space limitations, only a fraction of the relevant work could be discussed.