Due to its transient nature, SV exocytosis has been difficult to characterize morphologically. A number of questions remain partially unresolved to this date. In particular, it has been suggested that following Ca^2+^ entry, the insertion of synaptotagmin-1 into the membrane induces an increase in membrane curvature, which lowers the energy barrier of fusion. Such membrane deformations have been observed in biochemically reconstituted models of exocytosis but have not yet been reported in functional synapses [@doi:10.1016/j.cell.2010.02.017;@doi:10.1002/embr.201337807]. Moreover, it is not clear whether the membrane deformation occurs subsequently to Ca^2+^ influx or if primed SVs and their PM counterpart present such deformation [@doi:10.1002/embr.201337807]. The optimal sample preservation delivered by cryo-ET makes it possible to investigate the role of tethers located between SVs and the AZ PM and the function of inter-SV connectors. Combining cryo-ET with spray-mixing plunge-freezing enabled us to investigate the morphological changes occurring immediately after depolarization.
It should be noted that our study has some uncertainties and limitations. One uncertainty concerns the delay between stimulation and freezing. In future studies, this uncertainty could be reduced by stimulating exocytosis with a flash of light, on samples either expressing channelrhodopsin or containing caged calcium. Nonetheless, accessing sub-millisecond delays, which is required to observe the early stages of SV exocytosis would be technically quite challenging with light stimulation. The identity of the proteins composing the tethers and connectors represents another uncertainty. While stimulated synaptosomes were sprayed with a depolarizing solution, control synaptosomes were not sprayed. This may raise the concern that the observed changes are due the mere spraying and not to depolarization. It should be noted that plunge freezing is typically done in an environment at 100% relative humidity to prevent sample drying. Consequently, water droplets likely also form on the sample surface when the sample is not sprayed. Furthermore, earlier studies have shown that the spray does not affect the near-atomic structure of purified synaptic membranes [@doi:10.1016/j.jmb.2004.12.031;@doi:10.1016/j.jmb.2012.07.010]. It is therefore unlikely that the observed changes are due to the mere spraying. Another limitation is common to all cryo-ET studies: to date, cytoplasmic protein labelling for cryo-ET remains extremely challenging. Nevertheless, several laboratories, including ours, are working towards solving this limitation [@doi:10.1101/2022.04.10.487799]. Finally, we could only analyze a restricted number of specimens. This was mainly due to two factors. Samples were relatively thick, and finding suitably thin synapses required extremely time-consuming screening. In future studies, we might resort to cryo-focused ion beam (cryo-FIB) milling to prepare thin lamellas [@doi:10.1038/nmeth1014;@doi:10.1016/j.jsb.2016.07.010]. Manual segmentation of SVs was possibly the most serious bottleneck. In the future, deep-learning based segmentation procedures might reduce this burden.
Depolarization through spraying droplets of KCl solution on synaptosomes milliseconds before freezing allowed us to capture snapshots of exocytosis (Figure {@fig:membrane_morphology}B1-B3). In spite of the uncertainty on the exact delay between stimulation and freezing, our approach allowed access to shorter delays than any other cryo-EM technique. Since the complete collapse of SVs occurs somewhere between 20 and 50 ms after stimulation [@doi:10.1083/jcb.88.3.564], we could observe early and late stages of exocytosis on the same spray-mixed and plunge-frozen synaptosome grid. The temporal sorting of observed exocytosis snapshots was done in the most parsimonious way. Nonetheless, it should be kept in mind that the annotation in early and late stages is a working hypothesis, which is supported by the literature [@doi:10.1083/jcb.88.3.564].
We observed that the curvature of some PM regions facing some SVs increased following depolarization. The SV facing such a PM buckling also seemed to get kinked. These deformations were not seen in non-sprayed synaptosomes. This indicates that in functional synapses exocytosis starts with a Ca^2+^-dependent membrane deformation, which is supported by a wealth of in vitro biochemical data [@doi:10.1016/j.cell.2010.02.017;@doi:10.1146/annurev-biophys-111821-104732]. Deformation may be caused in part by the intercalation of synaptotagmin-1 C2A and C2B domains between membrane head groups. A recent biophysical study indicated that C2A and C2B preferably insert in SV membrane and PM, respectively [@doi:10.1016/j.bpj.2019.06.016]. It may also be due to the tension/force induced by SNARE-complex zippering [@doi:10.1126/science.1224492]. Subsequent snapshots showed a fuzzy contact point between the SV and the PM, which likely corresponds to lipid splaying or the merging of the two membranes. Membrane fusion then occurred and yielded classical Ω-figures with variable neck diameters. Finally, nearly fully collapsed SVs were imaged. Overall our observations support the standard model of full collapse SNARE-dependent membrane fusion [@doi:10.1073/pnas.1816495115; @doi:10.1002/cbic.201100020] and reveal details of exocytosis early stage, prior to actual membrane fusion.
SV local concentration - a.k.a. SV occupancy - is tightly correlated with the distance from the AZ PM. Under resting conditions, SV occupancy showed a minimum in the intermediate region (45-75 nm away from the AZ) (Figure {@fig:vesicle_distribution}A-B), in agreement with previous reports [@doi:10.1083/jcb.200908082]. By definition, all SVs in the proximal region are directly facing the PM. Their high concentration can be attributed to the fact that more than 50% of them are tethered to the PM. On the other hand, the number of connectors per SV and SV connectivity is generally higher in the distal regions (Figure {@fig:connectors}A-D). This increased value correlates with the higher SV occupancy. Thus, we may hypothesize that SV local concentration is a function of their level of tethering to the PM and of connection with other SVs. Interestingly, under short stimulation of a few milliseconds, SV occupancy only decreased in the proximal region, most likely as a consequence of the fusion of SVs with the PM (Figure {@fig:vesicle_distribution}B).
In order to further assess the relation between SV tethering, connectivity, and occupancy, we analyzed synapses expressing either WT SNAP-25, a more positively charged mutant (4K), or a more negatively charged mutant (4E) [@doi:10.1016/j.celrep.2019.01.103]. The 4K mutant has a decreased energy barrier to membrane fusion and causes constitutive active exocytosis, whereas the 4E mutant shows a decreased exocytotic activity because of a higher energy barrier to membrane fusion. Our data show that the 4K mutant had a significantly decreased proximal SV occupancy, while there was no significant difference in the case of the 4E mutant (Figure {@fig:vesicle_distribution}A). The decrease was probably due to the high frequency of spontaneous exocytosis observed in the 4K mutant [@doi:10.1016/j.celrep.2019.01.103]. In the intermediate, and distal I regions, the occupancy of both mutants were very similar to the one of the WT. In more distal regions, the variability in occupancy between individual active zones strongly increased and made intergroup comparisons difficult. Our data show that strong disturbances in exocytotic activity lead to profound differences in SV connectivity. We note that a correlation exists between SV connectivity and concentration. Future studies will be necessary to assess whether SV concentration depends on the SV connectivity and to decipher the molecular mechanism influencing these parameters.
Previously, we showed that the number of tethers of a SV defines whether its exocytosis can be induced by treatment with a hyperosmotic sucrose solution, which corresponds to a definition of the RRP [@doi:10.1083/jcb.200908082;@doi:10.1016/j.sbi.2019.01.008]. We reported that SVs with at least 3 tethers belong to the RRP, according to this definition. In order to further assess this model, we analyzed synapses of neurons expressing the SNAP-25 mutants. 17% of the WT proximal SVs had 3 tethers or more. Critically, the 4K mutant had none such SV. As the RRP (assessed with hyperosmotic sucrose treatment) in this mutant was formerly shown through functional assays to be depleted, our present observation further supports our morphological definition of the RRP [@doi:10.1016/j.celrep.2019.01.103]. 15% of the proximal SVs had 3 tethers or more in the 4E mutant, which is very similar to the WT situation, while this mutant was shown to possess a normal-sized RRP. Our observations are also consistent with a number of studies that have concluded that SV exocytosis requires a minimum of three SNARE complexes [@doi:10.1074/jbc.m109.047381;@doi:10.1126/science.1193134;@doi:10.1126/science.1214984].
We compared SV tethering before and shortly after depolarization. Our observations are schematically summarized in Figure {@fig:model}. Interestingly, the fraction of proximal SVs that were tethered increased by 50% shortly after stimulation. Simultaneously, the number of tethers per proximal SV more than doubled, and the fraction of proximal SVs with 3 or more tethers tripled. All these measurements returned to pre-stimulation values in presynaptic terminals presenting more advanced stages of exocytosis (late fusion stage). These data indicate that immediately after the onset of stimulation, a quick and massive increase in tethering occurs. This phenomenon was resolved in our measurements because the spraying of synaptosomes with an intermediate K^+^-concentration made it possible to detect synaptosomes in an early stage of fusion, which would have been missed during either strong or chronic stimulation, because these would deplete primed vesicles.
The phenomenon of rapid, depolarization-induced tethering leads to some free proximal SVs becoming tethered to the AZ PM, while some previously single- or double-tethered SVs gained the additional tether(s) thus rendering them releasable according to our definition of the RRP (as triple-tethered vesicles) [@doi:10.1083/jcb.200908082]. There are several important implications of this finding. First, the increase in the number of tethers during the initial membrane contact - in excess of the three tethers formed during priming - might help overcome the fusion barrier. Functional reconstruction led to the suggestion that SNARE-complexes primarily form downstream of Ca^2+^-influx [@doi:10.1038/nsmb.2061], whereas mutagenesis studies in cells supported the notion that SNARE-complexes had already formed before arrival of the Ca^2+^-trigger, i.e. during priming [@doi:10.1083/jcb.200907018]. In fact, both notions might be partly correct, as the formation of a low number of SNARE-complexes might lead to a stable primed state, defined by a valley in the energy landscape due to the dual inhibitory/stimulatory features of the SNARE-complex [@doi:10.1038/emboj.2010.130;@doi:10.1016/j.celrep.2019.01.103 ], whereas more SNARE-complexes might form dynamically after triggering, during membrane fusion itself. Accordingly, in in vitro fusion assays, additional SNARE-complexes, in addition to those required for fusion pore formation, cause to fusion pore stabilization [@doi:10.1126/science.1214984;@doi:10.1038/nature25481]. Second, vesicles that have not formed three tethers before stimulation might fuse with delayed kinetics during triggering, which accounts for the variable exocytosis kinetics among SVs [@doi:10.1073/pnas.1606383113;@doi:10.1073/pnas.1314427110;@doi:10.1016/j.neuron.2015.08.038;@doi:10.7554/eLife.51032]. Superprimed vesicles are expected to have formed the largest number of tethers before stimulation [@doi:10.1073/pnas.1314427110;@doi:10.1073/pnas.1606383113]. Third, functionally overlapping protein complexes might be involved in priming and triggering, depending on the timing of their formation. Accordingly, triggering that stimulates tether-formation might also stimulate priming for those vesicles that were not tethered before stimulation. Indeed, a number of recent publications have suggested that some SVs can get primed extremely quickly in response to Ca^2+^ influx [@doi:10.3389/fnsys.2019.00030;@doi:10.1016/j.tins.2021.04.003;@doi:10.1038/s41593-020-00716-1;@doi:10.1016/j.neuron.2016.07.033;@doi:10.7554/eLife.51032].
{#fig:model width="15cm"}
Our study revealed fine morphological changes occurring in the presynaptic terminal immediately after the onset of exocytosis, as well as in chronically active or inactive synapses. It indicates that the rise in presynaptic Ca^2+^ induced increased SV tethering prior to SV fusion, which potentially corresponds to SV superpriming. We also detected modifications of proximal SV interconnections in response to evoked exocytosis, as well as more drastic modifications of distal SV interconnections in chronically active synapses and in inactive synapses. These changes likely affect SV mobility and recruitment at the AZ.