To analyze the morphological changes occurring in the presynapse shortly after stimulation, we pursued a time-resolved cryo-electron tomography approach similar to the one introduced by Heuser, Reese, and colleagues [@doi:10.1083/jcb.81.2.275]. Whereas Heuser et al. used an electrical stimulus, we chose to trigger exocytosis by spraying a depolarizing solution containing 52-mM KCl onto the specimen a few milliseconds before freezing for two reasons. First, spraying allows cryo-ET imaging because it is compatible with plunge freezing directly on an EM grid, which is not the case of electrical stimulation. Second, this method allows catching synapses at delays between stimulation and freezing lower than a millisecond, as explained below. Some of the delays achieved here are shorter than those attained by electrical stimulation. The solution was sprayed with an atomizer and droplets hit the EM grid a few milliseconds before freezing. The spray-mixing plunge freezing setup was custom-built based on a system introduced by Berriman and Unwin [@doi:10/ctwp5j].
The spray droplet size was optimized by cutting a 1-ml pipette tip to a diameter matching an EM grid (3 mm) and fixed to the atomizer glass outlet to disperse the spray (Figure {@fig:workflow}Ai). Furthermore, to achieve the shortest possible delay between spraying and freezing, the nozzle was set 1-2 mm above the liquid ethane container. This generated many small spray droplets spread throughout the grid (Figure {@fig:workflow}Aii-Aiv, Figure @{fig:spray_workflow}A-D). Even if sprayed droplets were well distributed throughout the grid, not all synaptosomes were in contact with exocytosis-triggering KCl solution. Synaptosomes located on the landing spot of a droplet were stimulated instantly and therefore were frozen for a delay equal to the time between the grid crossing the spray and hitting the cryogen, which was typically set at 7 or 35 ms (Figure @{fig:spray_workflow}B). However, synaptosomes situated at a distance from a landing spot were only stimulated when the KCl concentration rose due to diffusion reached a threshold triggering voltage-gated calcium channel opening (Figure @{fig:spray_workflow}C and D). Through this process, we were able to trap stimulated synapses at the very earliest stages of exocytosis. Given the very low throughput of cryo-electron tomography, we followed a correlative light and electron microscopy approach. By cryo-fluorescence microscopy, we identified areas where synaptosomes fluorescently labeled by calcein blue and spray droplets labeled by fluorescein were colocalized. Additionally, phase contrast imaging enabled quality control of the frozen EM grid with respect to ice contamination and ice cracks, as shown previously [@doi:10.1016/j.jsb.2007.07.011]. 9 control and 9 stimulated synaptosome tomograms were analyzed (Figure {@fig:suppl_tomogram_slices}A and B, Movie EV1, Table @tbl:synaptosome-tomograms). We restricted our analysis to synaptosomes that possessed a smooth PM, free of signs of rupturing and that had a mitochondrion, as we considered these factors essential for synaptosome function.
{#fig:workflow width="15cm"}
{#fig:spray_workflow width="15cm"}
In addition, we manipulated the electrostatic state of the SNARE complex through mutated SNAP-25 protein introduced using lentiviral vectors into primary SNAP-25 knockout neurons grown on EM grids [@doi:10.1016/j.celrep.2019.01.103] (Figure {@fig:workflow}Bi-Biv). The “4E” mutation contains four glutamic acid substitutions, whereas the “4K” contains four lysine substitutions; all mutations are placed in the second SNARE-motif of SNAP-25 and were shown to decrease and increase the rate of spontaneous miniature release, respectively [@doi:10.1016/j.celrep.2019.01.103]. Optimization of primary neurons culturing conditions allowed us to establish a protocol, which provides functional synapses thin enough for direct imaging by cryo-ET. Astrocytes were added to 12 well plates and were grown for 2 days. After 2 days, the medium was exchanged to a medium that favors neuronal growth and impedes astrocyte growth. At the same time a droplet of the neuronal suspension was added onto a flame sterilized EM grid and incubated for 30 min at 37 °C, hereafter the grids were placed into the 12 well plates containing the astrocytes. Neurons were grown for 10-14 days until plunge freezing and were then analyzed at a Titan Krios by cryo-ET (Figure {@fig:suppl_tomogram_slices}C and D, Figure @{fig:suppl_segmentation}A-C, Movie EV2, Table @tbl:neuron-tomograms). Thereby, we could image chronically overactive or inactive synapses and relate presynaptic architectural modifications to different functional states.
We first analyzed the morphology of SVs fusing with the AZ PM. As explained above, synaptosomes of a single grid have not all been stimulated for the same duration (Figure @fig:spray_workflow). The time interval between triggering exocytosis and freezing ranged between 0 ms and the interval between spray droplets hitting the grid and freezing (maximum 35 ms; see [@doi:10/ctwp5j]). This offered the unique possibility to observe SV exocytosis events immediately after their initiation, and even before membranes have started to mix.
Synaptosomes from both control and sprayed grids were thoroughly analyzed for signs of exocytosis, which consisted of morphological changes of the AZ PM and the SV membrane occurring upon stimulation. These signs were only detected in synaptosomes from sprayed grids and neither non-sprayed tomograms acquired specifically for this study nor in past studies containing hundreds of SVs at the active zone.
In non-sprayed synaptosomes, proximal SVs were spherical and the opposite plasma membrane was smooth (Figure {@fig:membrane_morphology}A). Among the synapses that showed bent lipid membranes, we distinguished early and late fusion stages. The early stage events were defined as those that do not show open fusion pore. We observed two different types of these events, as follows. (i) Both the vesicle membrane and the PM were slightly bent towards each other, without making a contact (Figure {@fig:membrane_morphology}Bi-Biii; orange arrows; n=8). These structures, which have previously been reported in liposomes but not in synapses, have been referred to as membrane curvature events [@doi:10.1016/j.cell.2010.02.017]. Control synaptosomes (i.e. not sprayed) on the other hand, had a straight PM, and no SV membrane was buckled (Figure {@fig:membrane_morphology}A). (ii) A contact was formed between the vesicle and the PM bilayer where both membranes lose their clear contours (Figure {@fig:membrane_morphology}Ci & Cii; pink arrows; n=3), or where they establish a point contact (Figure {@fig:membrane_morphology}D-E; blue arrows; n=5). Similar events were observed previously only by cryo-ET in chronically stimulated synaptosomes, but could not be firmly associated with a particular SV exocytosis state [@doi:10.1083/jcb.200908082].
The late fusion stage events were defined as those where SV lumenal and the extracellular regions were in direct contact. These events include small SV fusion pore opening (Figure 3F; blue arrows; n=3) wide fusion pore opening (Figure 3G; n=1), and cases where SVs collapsed almost completely and only a small bump on the PM remained visible (Figure 3H; n=11). Using high-pressure freezing and freeze-substitution, Imig et al. observed very similar structures in mouse hippocampal organotypic slices [@doi:10.1016/j.neuron.2014.10.009]. Nevertheless, very fine lipid membrane deformations such as those in Figure {@fig:membrane_morphology}Bi-iii and SVs located at a few nanometer distance to the AZ PM were not observed, possibly because of the application of dehydration and heavy metal staining.
{#fig:membrane_morphology width="15cm"}
The notion that the SV fusion pore opening is the last step before the full SV collapse is well established, and it was hypothesized that lipid membrane perturbation and/or lipid point contact precede pore opening [@doi:10.1002/pro.3445]. Therefore, our temporal assignment of the early and late fusion stages is dictated by the current understanding of neurotransmitter release. Regarding the temporal assignment of events comprising the early and the late stage, it is very likely that lipid membrane perturbation events precede establishing the contact between SV and PM lipids (Figure {@fig:membrane_morphology}B-E). Also, it is expected that in the full-collapse fusion, the SV pore progressively widens (Figure {@fig:membrane_morphology}F-H). These establish the most likely temporal sequence of the events we observed.
Importantly, in the analysis that follows, only the early and the late fusion stages are distinguished. Furthermore, while the fusion stages were defined based on the observed lipid deformations, the further analysis concerns features that were not used to define the fusion stages, namely SV-bound protein complexes and the position of SVs within the presynaptic terminal.
Non-sprayed rat synaptosomes as well as WT-SNAP-25 mouse cultured neuron synapses showed typical SV distribution, as observed in previous cryo-ET studies (Figure {@fig:vesicle_distribution}) [@doi:10.1083/jcb.200908082]. Vesicle occupancy in WT-SNAP-25 synapses was 0.13 in the proximal zone (0-45 nm from the AZ PM), dropped to 0.09 in the intermediate zone (45-75 nm), rose to 0.16 in the distal I zone (75-150 nm), decreased to 0.14 in the distal II zone (150-250 nm), and decreased further to 0.11 in the distal III zone (450-900 nm, Figure {@fig:vesicle_distribution}A).
{#fig:vesicle_distribution width="15cm"}
The absolute values differ between WT cultured mouse neurons and non-stimulated rat synaptosomes, but the SV occupancy distribution follows the same pattern. The difference in absolute values can likely be attributed to the different experimental and animal models used. Sprayed synaptosomes that were showing early signs of exocytosis had a nearly identical SV occupancy pattern as non-sprayed synaptosomes (Figure {@fig:vesicle_distribution}B, dark blue and gray, respectively). However, when SV full collapse figures were apparent, SV occupancy in the proximal zone was reduced (P<0.08 with Benjamini-Hochberg correction), whereas SV occupancy further away from the AZ PM was unchanged. This is similar but less pronounced than the redcution of the proximal SV occupancy observed after prolonged stimulation (30 to 60 s), which we reported earlier [@doi:10.1083/jcb.200908082]. Thus it is consistent with some membrane proximal SVs having engaged in exocytosis, while none of the recycling and reserve pool SVs have.
In order to investigate the consequences of chronic high or low synaptic activity, we investigated the 4E and 4K mutants (Figure {@fig:vesicle_distribution}A, green and gold, respectively). In the proximal zone, SV were significantly less concentrated in the constitutive active 4K mutant than in the WT (P<0.05, ANOVA test with Benjamini-Hochberg correction). This can be readily attributed to the high probability of spontaneous exocytosis generated by the additional positive charges of the SNARE bundle. Furthermore, the proximal SVs of both mutants were located significantly further away from the active zone plasma membrane (P<0.05 in both cases, ANOVA test with Benjamini-Hochberg correction, Figure {@fig:vesicle_distribution}C). The larger mean distance between proximal SVs and plasma membrane in the 4E mutant might result from the repulsion between negative charges present in SNAP-25 and on the plasma membrane. In the case of the 4K mutant, the larger mean distance may be due to the high spontaneous exocytotic activity of this mutant. Indeed, SVs located in close proximity of the plasma membrane have a very high probability of fusing, which then leaves proximal SVs located a higher distance to the PM. Of note, no significant difference in the mean distance between proximal SVs and the plasma membrane was observed following stimulation in rat synaptosomes (Figure {@fig:vesicle_distribution}D). In the most distal zones of neuronal synapses, SV occupancy in both mutants mostly got lower than in the WT, although the differences were not significant (Figure {@fig:vesicle_distribution}A).
We investigated the tethering state of proximal SVs (i.e. the SVs whose center is located within 45 nm of the AZ PM; Figure {@fig:tethers}A-D). We first compared the situation in synaptosomes prior to and following stimulation. In non-sprayed synaptosomes, 54% of the proximal vesicles were tethered, which is in agreement with previous results (Figure {@fig:suppl_histograms}B) [@doi:10.1083/jcb.200908082]. Interestingly, in the early fusion group the fraction of tethered proximal vesicles significantly increased to 80% (P<0.05, χ2 test with Benjamini-Hochberg correction). In the late fusion group, however, 53% of the proximal vesicles were tethered, which was not significantly different from the non-sprayed group. The average number of tethers per proximal SV followed the same pattern. Proximal SVs had 0.89 ± 0.12 tethers in the non-sprayed group (Figure {@fig:tethers}D). This parameter significantly rose to 2.09 ± 0.33 in the early fusion group (P<0.001, ANOVA test with Benjamini-Hochberg correction), while it returned to 1.00 ± 0.20 in the late fusion group.
{#fig:tethers width="15cm"}
We then analyzed whether the decreased occupancy in the late fusion group was associated with a decreased number of triple-tethered SVs (defined as SV with at least three tethers), which as mentioned in the introduction are suggested to belong to the RRP. In resting, non-sprayed synapses, 8% of the proximal SVs were triple-tethered (Figure {@fig:tethers}B). Surprisingly, the fraction of triple-tethered proximal SVs drastically increased to 29% in the early fusion group (P<0.01, χ2 test with Benjamini-Hochberg correction). The fraction decreased to 13% in the late fusion group. This suggests that upon stimulation, some proximal SVs very rapidly acquire new tethers. Using our definition of the RRP (vesicles that are triple-tethered) this would indicate that the RRP rapidly increases after stimulation and more vesicles become primed for exocytosis. Furthermore, the lower proximal vesicle occupancy in the late fusion group indicates that under our stimulation conditions, replenishing vesicles to the proximal zone is slower than their release.
The situation in the WT-SNAP-25 neurons was similar to unstimulated synaptosomes. 53% of the proximal SVs were tethered and 17% of the proximal SVs belonged were triple-tethered (Figure {@fig:suppl_histograms}A and Figure {@fig:tethers}A). On average, proximal SVs had 1.17 ± 0.23 tethers (Figure {@fig:tethers}C). The corresponding values for the 4E mutants were not significantly different (15% and 0.96 ± 0.18, respectively). However, in all 4K mutant datasets there was not a single SV that was part of the RRP, i.e. triple-tethered. Similarly, the number of tethers per proximal SV was significantly lower in the 4K mutant than in the WT (Figure {@fig:tethers}C, P<0.05, ANOVA test with Benjamini-Hochberg correction). These results are in line with physiological measurements that have shown that the RRP is depleted in the chronically spontaneously active 4K mutant, and they provide additional evidence that RRP-vesicles have at least 3 tethers [@doi:10.1016/j.celrep.2019.01.103].
The majority of SV are linked to other SVs via molecular bridges previously termed connectors [@doi:10.1083/jcb.200908082;@doi:10.1016/j.sbi.2019.01.008;@doi:10/d3bx8g;@doi:10.1083/jcb.108.1.111]. The function and composition of connectors are not clear yet. It was earlier proposed that connectors limit SV dispersion and allow SV mobilization for release. It is generally assumed that synapsin is involved in connector formation and may be one of its components. It has been suggested that connectors reduce SV mobility and maintain a locally high SV concentration in the presynapse. The connectivity level of an individual SV might be one of the factors defining the pool to which the SV belongs. To shed some light on the role of connectors, we analyzed SV connectivity in our datasets. We focused our analysis to the SVs located at distance of the AZ PM lower than 250 nm in synaptosomes and lower than 900 nm in neurons (Figure {@fig:connectors}A-H). Furthermore, we defined 6 distance groups: proximal (0-45 nm), intermediate (45-75 nm), distal I (75-150 nm), distal II (150-250 nm), distal III (250-450 nm), and distal IV (450-900 nm) similarly to previous studies [@doi:10.1083/jcb.200908082;@doi:10.1083/jcb.201206063].
In non-sprayed synaptosomes datasets, approximately 70% of the proximal and intermediate SVs were connected to other vesicles. In distal I and II regions, this value rose to 84% (P<0.05 χ2 test with Benjamini-Hochberg correction) and 87% (P<0.001), respectively (Figure {@fig:connectors}D). Similarly, the number of connectors per vesicle significantly increased from the proximal region (1.63 ± 0.17) to the distal I region (2.39 ± 0.10, P<0.001 ANOVA test with Benjamini-Hochberg correction) and the distal II region (2.92 ± 0.10, P<0.001, Figure {@fig:connectors}B).
We then compared the number of connectors per SV between non-sprayed synaptosomes and early fusion or late fusion synaptosomes. In the proximal group there were significantly more connectors in the early fusion group than in the non sprayed group (1.63 ± 0.17 and 2.69 ± 0.43, P<0.05 ANOVA test with Benjamini-Hochberg correction) and this number significantly dropped to 0.95 ± 0.18 in the late fusion group (P<0.05, Figure {@fig:connectors}B). Consistently, the number of connectors per non triple-tethered proximal SV significantly increased from 1.64 ± 0.17 in the non-sprayed group to 2.69 ± 0.54 in the early fusion group (P<0.05 ANOVA test with Benjamini-Hochberg correction) and dropped to 0.91 ± 0.19 in the late fusion group (P<0.05, Figure {@fig:connectors}F). The number of connectors per triple-tethered proximal SVs showed the same pattern, but these results did not reach significance. When considering non-tethered versus tethered proximal SVs, we observed no significant difference in the number of connectors per SV between the experimental groups (Figure {@fig:connectors}H). We compared the connected fraction of proximal SVs between non-sprayed and sprayed synatopsomes and found no significant difference (Figure {@fig:connectors}D, Figure {@fig:suppl_histograms}D and F).
Taken together, our observations indicate that following depolarization, the number of connectors per proximal SV with less than 3 tethers (i.e. non-RRP) first increases and then decreases to a value lower than the initial one. We have seen earlier that the fraction of tethered proximal SVs does not differ between non-sprayed and late fusion synaptosomes (Figure {@fig:suppl_histograms}B). Thus, our data suggest that establishing connectivity is a slower process than tethering. We hypothesize that given the free space made in the proximal region after some SVs have fused, vesicles from the intermediate region diffuse to the proximal zone and become tethered to the AZ PM, reaching non-stimulated levels within the time frame of our experiments. However, during this time, the number of connectors did not reach the non-stimulated levels. Furthermore, we have observed that connectors remained present between fusing SV and neighbor SV (Figure @{fig:suppl_connected_slices}A,B). Thus, in addition to passive diffusion, pulling towards the plasma membrane of SV connected to fusing SV can contribute to replenishing the RRP.
{#fig:connectors width="15cm"}
We then analyzed SNAP-25 neurons. For SNAP-25-WT, similarly to non-sprayed synaptosomes, the fraction of connected SVs was significantly higher in the distal II and III regions than in the proximal region (p<0.05 and P<0.01, respectively, χ2 test with Benjamini-Hochberg correction), albeit the absolute values were overall lower than in synaptosomes (Figure {@fig:connectors}C). Consistently, the number of connectors per SV in SNAP-25-WT synapses increased from 1.95 ± 0.38 in the proximal region to 3.23 ± 0.21 in the distal II region (Figure {@fig:connectors}A, P<0.05, ANOVA test with Benjamini-Hochberg correction). The fraction of connected SVs in the distal II region was significantly lower in the 4E and 4K mutant than in the WT (p<0.05, χ2 test with Benjamini-Hochberg correction, Figure {@fig:connectors}C). This was supported by a significantly lower number of connectors per SV in the distal II region for the 4E mutant versus the WT (P<0.001, ANOVA test with Benjamini-Hochberg correction) as well as for the 4K mutant versus the WT (P<0.001, Figure {@fig:connectors}A). Furthermore, this number was also significantly lower in the distal III and IV regions for the 4E mutant vs the WT (P<0.001). Consistently the fraction of connected SVs was lower in the distal II, III, and IV regions for the 4E mutant vs the WT (P<0.05, P<0.001, and p<0.05, respectively, χ2 test with Benjamini-Hochberg correction, Figure {@fig:connectors}C). The fraction of connected SVs was not different in the 4K mutant versus the WT, except in the distal IV region where it was higher (P<0.05 χ2 test with Benjamini-Hochberg correction). The number of connectors per proximal SV and the connected fraction of proximal SVs were not affected by the mutations (Figure {@fig:connectors}A, E and G, Figure {@fig:suppl_histograms}A, C, and E). These results indicate that prolonged abnormal exocytotic activity, such as those caused by functional alterations of SNAP-25, are correlated with severe changes in intervesicular connectivity in the distal region.