Regulation of exocytic mode in hippocampal neurons by intra-bouton calcium concentration

Abstract

Release of neurotransmitters from synaptic vesicles is a central event in synaptic transmission. Recent evidence suggests that synaptic vesicles fuse with the plasma membrane by multiple routes during exocytosis, but the regulation and physiological implications of this choice are unclear. At hippocampal synapses in culture, two modes of synaptic vesicle exocytosis can be distinguished by virtue of the rate and extent of loss of a fluorescent lipid marker (FM1–43). Here we investigate these two modes of exocytosis using fluorescence imaging of FM1–43, combined with quantitative Ca²⁺ imaging using Oregon green BAPTA-1 (OGB1), to examine how the balance of exocytic mode changes during a stimulus train. Our findings are twofold: that the full fusion mode becomes progressively favoured through the course of a 5 or 10 Hz stimulus train, and that this occurs in parallel with presynaptic accumulation of calcium. Blockade of calcium accumulation with AM-EGTA also prevents the conversion of exocytic mode. This conversion of exocytic mode may provide insight as to the mechanisms underpinning short term plasticity.

Introduction

Synaptic communication is critically dependent on the mechanism of exocytosis, the process by which neurotransmitter is released from synaptic vesicles into the synaptic cleft. Recent evidence indicates that this can proceed through multiple routes: full collapse of vesicle membrane resulting in complete, rapid expulsion of vesicular contents, or via the opening of a limited pore connecting the vesicle lumen to the extracellular space (Klingauf et al. 1998; Aravanis et al. 2003; Gandhi & Stevens, 2003; Richards et al. 2005). This latter route is often termed ‘kiss-and-run’ exocytosis.

Mounting evidence indicates that kiss-and-run exocytosis may result in changes in synaptic throughput (Liu et al. 1999; Choi et al. 2000; Renger et al. 2001; Richards, 2009) due to less effective activation of postsynaptic receptors when the neurotransmitter arrives with a slowed time course and consequent lowered peak concentration. This raises the possibility that regulation of fusion mode might underpin certain forms of synaptic plasticity. While there is evidence for presynaptic components of long term plasticity at hippocampal synapses (Choi et al. 2000; Zhang et al. 2006; Ward et al. 2006; Ahmed & Siegelbaum, 2009) short term plasticities such as facilitation, augmentation and depression, which are unequivocally presynaptic in locus, have not been investigated in the context of fusion mode.

Presynaptic facilitation has been extensively investigated and its time course and components are well described (reviewed in Zucker & Regehr, 2002), and are interpreted as the consequence of a build up of residual calcium within the synaptic bouton. The result of this calcium build up is unclear, however. One proposal is that the calcium binds to a high affinity calcium sensor, which then acts in addition to the low affinity calcium sensor that triggers exocytosis at calcium microdomains to enhance overall levels of release. Alternatively, residual calcium might extend the range of microdomains so that more vesicles can be recruited, or it might reflect saturation of endogenous buffers, thus leaving more calcium available for triggering exocytosis. Conceptually at least, another option is a shift in release properties as calcium builds up, and it is this possibility that we address here.

Here, we have examined the dependence of vesicle release mode on time within a very short stimulus train (10 action potentials applied at 1–10 Hz), and independently monitored presynaptic calcium concentration within boutons. We find that there is a progressive conversion of release mode from a predominantly kiss-and-run situation to one where full fusion is favoured. This conversion is frequency dependent and highly correlated with intra-bouton calcium concentration. Combined with a recent study indicating impaired AMPA receptor activation by kiss-and-run mediated exocytosis (Richards, 2009) this provides a potential mechanism for synaptic facilitation that is downstream of the accumulation of residual calcium.

Methods

Cell culture

All animal procedures were carried out according to guidelines laid out by the institutional animal care and use committee of CCHMC, and Drummond (2009).

Hippocampi were removed from 0–2 day postnatal rat pups that had been chilled by ice bath and killed by decapitation. These were then coarsely minced before being incubated for 20–30 min at 37°C in Dulbecco's modified Eagle's medium (DMEM) containing 1 mg ml⁻¹ trypsin. The tissue suspension was then washed in DMEM containing 1% serum and triturated through a series of Pasteur pipettes of decreasing tip diameter. The resulting cell suspension was centrifuged at 1000 g to pellet the cells before resuspension in culture medium. The cells were initially plated in a small drop of medium at a density of approximately 40,000 per 16 mm coverslip, which had previously been coated with poly-l-lysine, then left for 2 h to settle, at which point DMEM containing 10% fetal bovine serum (FBS) was added to fill the culture dish. After 48 h this medium was replaced by the following medium: 40 ml DMEM, 45 ml minimal essential medium (MEM), 10 ml Ham's F12 medium, 1 ml FBS, 0.25 g bovine serum albumin, 50 μm glutamine, 0.5 ml insulin/transferrin/selenium, and 1 ml Di Porzio mixture (comprising progesterone 1.25 mg ml⁻¹, hydrocortisone 2 mg ml⁻¹, triiodothyronine 1 mg ml⁻¹, putrescine 20 mg ml⁻¹, superoxide dismutase 20,000 units ml⁻¹, and transferrin 20 mg ml⁻¹). The l-arginine content of the tissue culture medium was 0.66 mm. To minimize glial growth 0.7 mg 5-fluoro-2-deoxyuridine and 1.6 mg uridine were also added. Cultures were maintained for ∼14 days prior to use to allow full maturation. Pyramidal neurons were identified by their morphology and appearance under phase contrast microscopy (phase bright cell bodies possessing long spiny neurites with few recurrent branches).

Imaging

Imaging was carried out as described previously (Richards, 2009) on a Zeiss Axiovert 200M inverted microscope with harmonic focus drive, integrated into the Marianas workstation from Intelligent Imaging Innovations (Boulder, CO, USA). The camera was a Roper Cascade 512B, viewing the coverslip through a 100× 1.4 NA objective (Zeiss), and illumination was provided by a Lambda DG4 ultra high speed wavelength switcher (Sutter Instrument Co., Novato, CA, USA) through a scrambled light pipe to provide even illumination of the specimen. Imaging frequency was 20 Hz, with 30 ms exposure time. Subsequent analysis was carried out using ImageJ (NIH, Bethesda, MD, USA). The bathing saline had the following composition: 140 mm NaCl, 5 mm KCl, 2 mm CaCl₂, 1.5 mm MgCl₂, 5.5 mm glucose, 20 mm Hepes buffered to pH 7.3 using NaOH. Release was evoked by action potentials initiated via a presynaptic patch pipette. The amplitude of fluorescence drops was assessed by comparing fluorescence intensity prior to action potential stimulation with the intensity once the fluorescence had reached a new steady state. Imaging during high frequency trains caused a photobleach of approximately 8%, and image intensities were corrected accordingly using standard curves obtained without stimulation

Calcium imaging

Oregon green BAPTA-1 (OGB1) was used at 40 μm. Calcium concentrations were obtained using the method of Maravall et al. (2000). Using this approach, only F_max and R_f needed to be measured. Values used were R_f= 7.3 and K_D= 207 nm. The procedure was as follows: we patched the neuron with OGB1 in the pipette, we waited for filling to be complete (∼10 min), and we then stimulated at 20 Hz for 1 s to get F_max. We incubated the cells (while still patching) with AM-BAPTA (100 μm), washed them, waited ∼10 min, and measured F_min in zero external Ca²⁺. AM-EGTA was bath applied at 25 μm prior to break-in, and incubated for 20 min. For these experiments 25μM EGTA was also added to the pipette solution.

Electrophysiology

To allow for the elicitation of action potentials in the presynaptic cell, pyramidal cells in primary culture were patch-clamped in whole cell voltage clamp mode. Patching was carried out under visual guidance. The cells on their coverslips were mounted in a glass-bottomed incubation chamber on an inverted microscope (Zeiss Axiovert 200M), and the final approach was viewed with Nomarski optics using a 100× objective. Pipettes (5–7 MΩ) were pulled on a Sutter Instrument Co. microelectrode puller. Pipette tips were coated with wax containing 1:1 (by weight) paraffin:mineral oil to reduce noise due to tip capacitance. The internal solution was (in mm): CsMeSO₃ 130, NaCl 8, Na₂ATP 3, MgSO₃ 3, Hepes 10; pH 7.3. Patch seal resistances were between 2 and 10 GΩ, and zero current potentials (resting membrane potential) were more negative than −65 mV. The signal was amplified by an Axopatch 200 amplifier (Molecular Devices, Sunnyvale, CA, USA), low-pass filtered at 2–5 kHz and digitized at 20 kHz in real time to a computer hard drive using pCLAMP 9 software (Molecular Devices).

Results

Kiss-and-run is the dominant release mode at cultured hippocampal synapses in response to single somatically induced action potentials

Hippocampal neurons maintained in culture were patched at the cell body. A small subset (∼6) of synaptic vesicles were labelled with 8 μm FM1–43 as described previously (Richards et al. 2005; Richards, 2009). Following complete wash, neurons were patched at the soma, to allow for precise temporal control of exocytosis. Ten minutes later, a single action potential was elicited by current injection, while imaging at a high rate (20 Hz). Individual boutons were analysed separately, and showed one of three behaviours. In more than half of them, no change in FM1–43 fluorescence was observed – this corresponds both to presynaptic failures, as seen in quantal analysis experiments, and to the situation where an unlabelled vesicle is released, or possibly reflects fusion events releasing no FM dye, as described by Henkel & Betz (1995). Where loss of FM1–43 fluorescence was observed, it showed two types of behaviour: in some cases, there was an abrupt loss of fluorescence broadly corresponding to our estimates of the fluorescence of a single vesicle, while in others there was a smaller, slower loss of fluorescence, which we have previously interpreted as fusion pore mediated release (Richards et al. 2005; Richards 2009). Figure 1A shows typical traces indicating failure as well as full fusion and kiss-and-run type fusion events. Figure 1B shows the relative frequency of these events (n= 8). Figure 1C shows a comparison of amplitude histograms with and without stimulation. For both conditions, image analysis was carried out as though stimulation (1 Hz) was present. Amplitudes were measured before and after each action potential (virtual in the case of the ‘no stimulation’ condition), and amplitudes plotted.

Figure 1. FM1–43 efflux from vesicles reveals two modes of exocytosis.

A, five representative traces of fluorescence over time, indicating no response (top) or either small amplitude slow responses (kiss-and-run) or large amplitude rapid responses (full fusion). B, if all responses to a single action potential are measured and pooled, the results indicate approximately twice as many kiss-and-run responses as full fusion responses. Data from 5 coverslips, 3 culture preparations, 552 events. C, left, fluorescence changes due to ‘mock’ stimulation (i.e. times where stimulation would have been carried out if it had been an active experiment); 12 s of imaging without stimulation. Right, fluorescence changes from each stimulation in 1 Hz train of 10 shocks. Data are 200 events in each case.

Change of release properties during a stimulus train

Presumed full fusion events reach completion within around 100 ms, which allows us to investigate release mode during moderate frequency (<10 Hz) stimulation. Using the experimental method described above, we imaged presynaptic boutons while eliciting a short train of 10 action potentials 200 ms apart (5 Hz). Figure 2A shows the average response to the first action potential (where a drop in fluorescence was seen), compared to the average drop in response to the 10th and last action potential. The fluorescence drop following stimulus 10 is both faster and larger in amplitude than that seen in response to stimulus 1. Figure 2B and C shows the change in mean amplitude and tau as a function of stimulus number during a 5 Hz train. Note that this is an average trace; inspection of individual traces reveals clear ‘full fusion’ and ‘kiss-and-run’ type events as are shown in Fig. 1A.

Figure 2. Average unitary FM1–43 efflux increases during a 5 Hz stimulus train.

A, average response to a single action potential taken from either the first action potential of a stimulus train (continuous line) or the 10th action potential (dashed line). Non-responders were excluded, and the mean fluorescence drop can be seen to be larger at the end of the 5 Hz train than at the beginning. B, the average amplitude of fluorescence drop following each stimulus in a short 5 Hz train is plotted. No record was included for those occasions where no measurable drop occurred. C, individual traces of responders to each action potential in a stimulus train were fitted to a single exponential, and the average value of tau for these fits is plotted against stimulus number. Data from 7 coverslips, 4 culture preparations, 34 boutons.

One explanation for the change in average could be a gradual change in exocytic properties such as multi-vesicular release. We investigated this by plotting the maximum amplitude drop seen at each time point, over time. Multivesicular relase is seen at the calyx of Held (He et al. 2009), and promotion of this fusion mode would result in an increase in amplitude of single fluorescence drops. As shown in Fig. 3A, however, the maximum amplitude of the drop in fluorescence did not change during a stimulus train. One might think that this could be explained by recruitment of more exocytosis; however our analysis was confined to those time points where there was a drop in fluorescence, and so is unaffected by any change in the number of detectable events.

Figure 3. Kiss-and-run and full fusion responses can be distinguished while stimulating at 5 Hz.

A, the maximal drop in fluorescence amplitude was monitored for 34 boutons stimulated at 5 Hz. Average values are plotted as means ± s.e.m., and show no change through the course of a stimulus train. B, two example traces at 5 Hz are shown, and one at 1 Hz, with the fluorescence drops classified as kiss-and-run (KR) or full fusion (FF). C, if events are classified as full fusion or kiss-and-run on the basis of amplitude and kinetics, the probability of any given response being full fusion can be determined. This value is plotted against stimulus number for a 5 Hz train. Data from 7 coverslips, 4 culture preparations, 34 boutons.

An alternative explanation for the increase in average fluorescence drop might be progressive favouring of the full fusion release mode at the expense of kiss-and-run. We investigated this by assigning individual responses to the classes of either kiss-and-run or full fusion on the basis of both amplitude and fall time. Three raw fluorescence traces are shown in Fig. 3B, with their classification indicated (KR for kiss-and-run and FF for full fusion). Note that in all three cases, both fusion modes are seen to operate within the same bouton. The 1 Hz trace also demonstrates that the reduced efflux seen in kiss-and-run type events does result in dye retention, as described previously (Aravanis et al. 2003; Richards et al. 2005). Figure 3C shows the progressive favouring of the full fusion classification as the stimulus trains wore on, suggesting that this is the best explanation. This also indicates that the gradual change seen in Fig. 2B and C need not be interpreted as a graded response, but can also be interpreted as a population shifting between two modes.

Measuring presynaptic calcium accumulation during short stimulus trains

As short term presynaptic plasticity has been long established to rely upon the accumulation of presynaptic calcium (usually termed the ‘residual calcium hypothesis’) this was an obvious place to look for a signal that might regulate the conversion of fusion mode. We patched neurons with 40 μm Oregon green BAPTA1 (OGB1) in the pipette, and dialysed it into the cell over the course of 10–15 min. Action potentials were then elicited presynaptically while the fluorescence of nerve boutons (identified as small swellings along the axon) were studied. As shown in Fig. 4A, stimulation at 5 Hz resulted in discretely identifiable episodes of calcium entry, with each peak decaying somewhat prior to the next action potential, and resulting in significant accumulation of broadly presynaptic calcium (in other words, this does not reflect calcium within microdomains around the calcium channels). Stimulation at 20 Hz resulted in much more rapid calcium accumulation, which could not be easily resolved into individual episodes of calcium entry; this also resulted in apparent dye saturation. Figure 4B shows calcium signals from boutons with action potentials elicited at 1 Hz – robust signals were detected, but there was little to no calcium accumulation seen through the train of 10 action potentials.

Figure 4. Calcium accumulation during a high frequency train.

A, OGB1 was dialysed into neurons via the same patch pipette used to elicit action poentials. Following 20 min to allow diffusion of indicator, fluorescence was monitored during stimulation at either 5 or 20 Hz. Data are expressed as ΔF/F normalized to 100%. Data in A and B are the average of three successive trials on the same bouton with 10 min recovery time between each trial, and are representative of the larger population. B, OGB1 was dialysed into neurons via patch pipette, and fluorescence monitored. Little to no calcium accumulation was seen during stimulation at 1 Hz. C, pooled data showing calcium accumulation during stimulus trains. Raw fluorescence data were converted to calcium concentration and plotted against stimulus number for stimulation at 1 (triangle), 5 (filled square) and 10 Hz (open circle). Data are from 98 boutons.

Previous studies have developed methods for reasonable estimates of calcium concentration from single wavelength dyes such as OGB1 (Maravall et al. 2000; Jackson & Redman 2003), and we have used this approach to convert our raw fluorescence data into approximate calcium concentrations. Figure 4C shows the time course of calcium accumulation (from multiple experiments) during stimulation at 1, 5, or 10 Hz. The data at 5 and 10 Hz indicate an increasing slope, suggesting buffer saturation (Jackson & Redman 2003).

Relationship between intra-bouton calcium and exocytic mode

With data on the changes of exocytic mode during a short train of action potentials, and data on the presynaptic calcium concentration, we were able to look for a correlation between the two. Using the approach described in Fig. 1, we elicited 10 action potentials at 1 Hz, 5 Hz or 10 Hz, and quantified the release mode as the probability of full fusion exocytosis, given a response. Support for the hypothesis that calcium might regulate release mode was provided by the fact that 1 Hz stimulation – which does not show any calcium accumulation during a short train (see Fig. 4B) – did not show any change over the course of the 10-action-potential train. We investigated this further by plotting intra-bouton calcium for each stimulus number, against the probability of full fusion at that point during either a 5 or a 10 Hz train (Fig. 5). Not only is there a strong apparent calcium dependence under both conditions, but they fall on the same curve.

Figure 5. The probability of full fusion correlates with intracellular calcium accumulation.

If the calcium data and the release mode datasets are combined, one can plot the dependence of release mode against intracellular calcium. This was done for both 5 Hz (circles) and 10 Hz (squares). It reveals strong apparent calcium dependence for release mode.

AM-EGTA blocks the enhancement of full fusion without affecting release

Although we have demonstrated a clear correlation between fusion mode and presynaptic calcium concentration, a more direct linkage between these factors is desirable. To that end, we investigated the use of EGTA, a slow calcium chelator, as a means of separating intracellular calcium from stimulus number. As illustrated in Fig. 6A, intracellular calcium showed greatly attenuated accumulation in AM-EGTA loaded cells compared to controls, when each was stimulated at 5 Hz. In keeping with the hypothesis that residual calcium accumulation acts to drive the conversion from kiss-and-run to full fusion release mode, we found that AM-EGTA loading prevented this conversion during moderate frequency stimulation. (Fig. 6B).

Figure 6. AM-EGTA prevents facilitation of full fusion exocytosis during 5 Hz stimulation.

A, calcium accumulation during 5 Hz stimulation in the presence and absence of AM-EGTA (25 mm). Representative trace (average of 3 trials) showing fluorescence data for a control bouton (0 mm AM-EGTA) and one where 25 mm AM-EGTA was loaded into cells prior to patch. B, the average probability of full fusion for each stimulus in a train is plotted against stimulus number. No significant increase in full fusion probability is seen during the 5 Hz train, in contrast with the conversion to almost complete full fusion seen in cells lacking AM-EGTA. Exocytosis itself was not blocked, however.

Discussion

If neurons exhibit distinct mechanisms of vesicle fusion, as neuroendocrine cells are now well accepted to do, it seems likely that cells possess some means to regulate this. This could be on the basis of the specific complement of presynaptic proteins that are expressed, but could equally be regulated by intracellular messengers. In this work, we have examined the role that intra-bouton calcium concentration might play in this process. We have demonstrated that while the efflux of FM1–43 from vesicles to single action potentials appears to be predominately subquantal – an indication of fusion pore mediated release – this balance shifts to favour apparent full fusion during the first few action potentials of a high frequency train. This change in exocytic balance tracks the build up of intracellular calcium, and is blocked when the build up of calcium is prevented by AM-EGTA. These findings are broadly in agreement with previous work where bromophenyl blue uptake into vesicles undergoing kiss-and-run exocytosis was found to change between low frequency and high frequency stimulation (Harata et al. 2006). Work using quantum dot loading of synaptic vesicles has also investigated this (Zhang et al. 2009), and found a progressive favouring of full fusion exocytosis through a 10 Hz stimulus train. The present work is the first to investigate the role of intracellular calcium in this process.

Vesicle pools and release mode

Work in many different preparations has demonstrated that vesicle cycling kinetics are dependent on the vesicle pool (Koenig & Ikeda, 1996; Kuromi & Kidokoro 1998; Richards et al. 2000, 2003; Voglmaier et al. 2006; He et al. 2006). The experiments described in this article investigate the modulation of exocytic mode during the first 10 action potentials of a stimulus train, which corresponds to three to six vesicles, depending on stimulation frequency. This is much smaller than most estimates of the cycling vesicle pool (∼30 vesicles; Schikorski & Stevens, 2001; Harata et al. 2001); consequently, it is entirely possible that the exocytic mode conversion described here is vesicle pool specific, corresponding either to the standard readily releasable pool (RRP), or even to an ultra-releasable pool, potentially corresponding to vesicles docked at the plasma membrane (Schikorski & Stevens, 2001). Repopulation of the readily releasable pool in hippocampal boutons by rapid endocytosis has been described previously (Pyle et al. 2000). In that work, the authors detected kiss-and-run exocytosis via differential retention of FM1–43 and FM2–10, and found evidence suggesting that kiss-and-run was very much favoured by vesicles within the RRP. Our present findings are broadly in line with these points, with the additional element that fusion mode appears to change even in the course of a very brief train of stimuli insufficient to deplete the RRP by itself. The same article also described vesicular re-use on relatively brief time scales facilitated by kiss-and-run (1–2 s); however, the experiments in the present work were not designed to test for this.

Short-term plasticity and exocytic mode

Synapses possess well-characterized forms of plasticity in response to activity, and the short-term forms of plasticity in particular are well established in possessing a presynaptic locus. Synaptic depression is generally accepted as tracking the overall availability of vesicles within the nerve terminal (see for example Richards et al. 2003), although other factors may also play a role. Synaptic facilitation has been shown to possess calcium dependence, and is characterized by several phases, identified by their rate of development, and the time course of decay following the end of stimulation. At its simplest, this is often grouped into three or four stages: facilitation, which decays on a time course of milliseconds to seconds and is often considered as having two phases within it (F1 and F2), post-tetanic potentiation, which has a decay of around 5–10 min, and augmentation, which falls between the two (Kalkstein & Magleby 2005). These processes show different levels of calcium dependence, and the observation that augmentation and post-tetanic potentiation have a multiplicative relationship suggests that they operate through independent mechanisms. Facilitation appears to arise from calcium acting at a site with millisecond kinetics at the crayfish neuromuscular junction (Kamiya & Zucker, 1994) and a few tens of milliseconds at cerebellar synapses (Atluri & Regehr, 1996). The two components of facilitation observed at other synapses may reflect separate sites of action in F1 and F2 facilitation, or alternatively, intra-bouton calcium may decay non-exponentially, for example by diffusion away from active zones or clusters of active zones, with non-exponential kinetics (Issa & Hudspeth, 1996; Tang et al. 2000). It seems that a detailed description of the interplay between forms of facilitation and depression needs to be undertaken on a synapse by synapse level. The results presented here provide one locus at which observed release probability can be modulated, which is downstream of calcium influx. Whether this means such a mechanism is responsible for facilitation is unclear since hippocampal neurons in culture often show a mixture of facilitation and depression (Kaplan et al. 2003), or facilitation only of asynchronous release (Cummings et al. 1996; Kaplan et al. 2003), potentially reflecting the age-dependent maturation seen in acute slice (Scullin et al. 2010). Once the molecular basis for this ‘exocytic switch’ is determined, it should be possible to design experiments to directly address these questions.

Potential mechanisms of exocytic mode conversion

Over the years, several mechanisms have been proposed for the ability of presynaptic calcium build up to enhance neurotransmission. These vary from explanations routed in the ability of the nerve terminal to handle calcium, to more molecular explanations. Examples of calcium handling effecting neurotransmission include saturation of endogenous buffers, leading to supralinear accumulation of calcium, and calcium-induced calcium release, which also provides a mechanism for supralinear calcium build up. The results described here present an additional piece of the puzzle; while non-linear calcium accumulation may explain the shape of the release mode curve, the observation that release mode changes with accumulating calcium indicates that there is an intrinsic non-linearity of the secretory response to calcium located within the secretory machinery itself. This may reflect a secondary process such as phosphorylation of SNARE proteins, or it may be a fundamental property of the synaptotagmin–SNARE complex.

It is interesting that we do not find an abrupt threshold type effect for the conversion between exocytic modes (i.e. full fusion becomes progressively favoured rather than a situation where vesicles are initially all kiss-and-run and the switch to the full fusion mode). This may provide clues for the potential mechanism responsible for the shift. For example, if the calcium sensor responsible for the exocytic mode ‘switch’ were present at the active site, one might expect an all or nothing conversion. Conversely, the graded response seen suggests that some population dynamic is present within the nerve terminal. This is likely to be mediated by either a vesicular or a cytosolic factor presumably influenced by the number of calcium ions bound (if in direct response to calcium accumulation) or perhaps the number of phosphorylation sites activated by calcium-dependent kinases (Klingauf et al. 1998).

In this work, kiss-and-run and full fusion have been discussed as separate mechanisms. This is not necessarily so; two other mechanistic possibilities present themselves. The first alternative is that the fusion pore diameter and opening duration could be influenced by calcium in distinct fashions. Thus, partial slow release of FM dye could be due to a rapidly closing small diameter fusion pore, whereas the more rapid, complete loss of dye would be due to a longer duration, larger diameter fusion pore. This would be analogous to the described effect of fusion pore diameter on glutamate efflux (Choi et al. 2000), and is fully compatible with the data reported here. Another possibility is the model described for the fusion pores of dense-cored vesicles in neuroendocrine cells (Wang et al. 2003) where a small fusion pore opens and then either closes again (kiss-and-run) or proceeds with enlargement and collapse (full fusion). In order to distinguish these proposed modes from each other and the kiss-and-run/full fusion dichotomy presented here, a technique such as that used by Kasai et al. (2005) to measure fusion pore diameters in astrocytes will be required.

The demonstration that neurons have two modes of exocytosis implies that some cellular factor must set this preference. Here we have demonstrated that the accumulation of calcium during moderately high frequency firing acts to promote the full fusion release mode. This has significant implications for models of short term plasticity at hippocampal synapses, since facilitation, augmentation and post-tetanic potentiation are all driven by calcium accumulation.

Glossary

Abbreviations

OGB1

Oregon green BAPTA-1

AM-EGTA

acetoxy-methyl ester of EGTA

Author contributions

D.A.R. designed, carried out and analyzed the experiments described herein. D.A.R. also wrote the paper and approved the final version for publication. Work was carried out at CCHMC.