SV2B defines a subpopulation of synaptic vesicles

Abstract Synaptic vesicles can undergo several modes of exocytosis, endocytosis, and trafficking within individual synapses, and their fates may be linked to different vesicular protein compositions. Here, we mapped the intrasynaptic distribution of the synaptic vesicle proteins SV2B and SV2A in glutamatergic synapses of the hippocampus using three-dimensional electron microscopy. SV2B was almost completely absent from docked vesicles and a distinct cluster of vesicles found near the active zone. In contrast, SV2A was found in all domains of the synapse and was slightly enriched near the active zone. SV2B and SV2A were found on the membrane in the peri-active zone, suggesting the recycling from both clusters of vesicles. SV2B knockout mice displayed an increased seizure induction threshold only in a model employing high-frequency stimulation. Our data show that glutamatergic synapses generate molecularly distinct populations of synaptic vesicles and are able to maintain them at steep spatial gradients. The almost complete absence of SV2B from vesicles at the active zone of wildtype mice may explain why SV2A has been found more important for vesicle release.

Under the assumption that exocytosis and endocytosis are the only relevant modes of delivery and retrieval of SV2 proteins to and from the peri-AZ membrane, respectively, and that both SV2 paralogues show similar and constant dwell times in the cytoplasmic membrane (A), the number of SIPs in the peri-AZ membrane can yield information about the relative average release rates of SV2A/B+ and SV2A+ vesicles.As we render the synapse in its entity and therefore our counts reflect the underlying stoichiometry our data allows us to estimate this parameter.Every (full) exocytosis event will deliver its vesicular SV2 cargo to the membrane for the period of the dwell time (~20 s, (Soykan et al., 2016)).If the next vesicle is released during this dwell time -i.e. before the previously released material is endocytosed -material will accumulate in the membrane.In fact, the accumulation will increase proportionally to the release frequency -in analogy to the well-known single compartment model describing the rise in intracellular calcium as proportional to the action potential frequency (Helmchen et al., 1996).Our data lacks timing so we cannot determine release rates themselves, but, under our assumption made above, the SIP counts are still proportional to the average release rates and we can make a relative comparison between the synaptic release rates of SV2A+ and SV2A/B+ vesicles.In that simple model, the observed count of SV2A, #A, must be proportional to the sum of the release rate of SV2A+ and SV2A/B+ vesicles, RelRt_A, RelRt_A/B yielding (1).
RelRt_A/B is multiplied by 0.5 as release of vesicles containing SV2A and B delivers only half of SV2A molecules compared to SV2A only containing vesicles.Similarly, we can write (2).
As we know that there are twice more SV2A than SV2B molecules on the membrane we can in (2) substitute 0.5#A for #B, rearrange and substitute into (1) to obtain (3), i. e. we can derive a first rough estimate to predict that the total release rate of SV2A/B+ vesicles is twice that of vesicles containing only SV2A.This represents an upper estimate and needs some corrections for faster modes of endocytosis happening near the edge of the AZ, see below, but it already shows that release of SV2A/B+ vesicles cannot be a very rare phenomenon when compared to SV2A+ vesicles even though SV2A/B+ vesicles are not found near the AZ.
In response to low frequency firing, as it is typical for CA3 cells which provide the presynaptic terminals studied here, a fast, sub-second endocytosis component can be responsible for 1/2 if not 2/3 of vesicle recycling following action potential-triggered release from the AZ (Delvendahl et al., 2016;Soykan et al., 2017)see B).For this reason, the number of membrane SV2A SIPs may underestimate the release rate of SV2A in the calculation above by a factor of 2 to 3. Because we did neither observe SV2B near the AZ domain nor in the AZ membrane, we believe SV2B is not released at the AZ and may not undergo ultrafast endocytosis.Taking fast endocytosis into account, the average release rate of SV2A+ vesicle would be equal or ~50% higher than that of SV2A/B+ vesicles.SV2A was consistently observed on docked vesicles but it was only rarely found in the membrane of the AZ.We counted on average only ~0.25 SV2A SIPs per synapse in the membrane of the AZ (data not shown), roughly 10-fold less than in the membrane of the peri-AZ (cf.Fig. 3).The likelihood to identify a SIP-labelled SV2 at a certain location in an EM image, such as the AZ, also depends on how much time it spends there.If SV2A rapidly moves through one location and rests at a different one, the likelihood it is found at the first and transient position is small (see C)).The fact that it is very likely that SV2A-containing vesicles are released at the AZ but we only rarely detect SV2A in the AZ membrane suggests that the protein is rapidly cleared from the AZ area.The 10-fold difference in SIP counts between AZ and peri-AZ membrane in turn indicates that SV2A stays at least 10-fold longer in the peri-AZ membrane than in the AZ.Assuming a dwell time in the peri-AZ before endocytosis of ~20 s (Soykan et al., 2016) our data would predict SV2A to not stay longer than ~2 s in the AZ membrane following fusion.The residency time could differ even more than 10-fold if some SV2A molecules would not end up in the peri-AZ and but were directly endocytosed at the edge of the AZ, implying that the time spent in the AZ membrane would be even shorter than 2 s.This is consistent with the view and findings that release sites are cleared within 1-2 s from synaptic vesicle material stranded in the membrane after fusion by lateral movement and ultrafast endocytosis at the edge of the AZ (Hosoi et al., 2009;Hua et al., 2011Hua et al., , 2013;;Watanabe et al., 2013;Gimber et al., 2015;Tehran and Maritzen, 2022).Thus, our counting of membrane associated SV2-labelling SIPs structurally supports the view that vesicular proteins are rapidly cleared from the AZ membrane at least 10-fold faster than they are endocytosed.
Further, this counting strongly suggests that upon endocytosis SV2A and B are assigned to new vesicles in a regulated and non-random manner.If stranded SV2 proteins would be randomly integrated into endocytosed vesicles, SV2A+ and SV2A/B+ vesicles would also be generated but the resulting fractions would not match the estimated release rates of SV2A+ and SV2A/B+ containing vesicles: 2/3 and 1/3 of the membrane pool of SV2 proteins are comprised of SV2A and B, respectively, and vesicles contain 5 copies of SV2 (Mutch et al., 2011).This would mean that endocytosis of synaptic vesicles carrying 5 SV2A molecules would happen with a probability of at most (2/3)^5 ~ 13% and 5 SV2B with a probability of at most (1/3)^5 ~ 0.4% and SV2A/B+ vesicles would occur in the remaining ~86% of cases.
In other words, the random model would generate 6-7-fold more SV2A/B+ vesicles than SV2A+ vesicles which does not match the estimate that SV2A+ and SV2A/B+ are very likely released at similar rates if not SV2A+ vesicles are released more often (see above).To match those most probable release rates, endocytosis should be specific with regard to the equipment of endocytosed vesicles with SV2A and SV2B proteins.Indeed, mechanisms have been proposed potentially allowing for such preference like clustering of vesicular protein content post-fusion, keeping a pool of pre-assorted proteins next to the AZ ready for endocytosis, and a surface pool of stranded vesicles (Wienisch and Klingauf, 2006;Hua et al., 2011).

Supplementary
Figure S4: SV2B KO mice exhibit an increased seizure threshold in the maximal electroshock model but are inconspicuous in other behavioral tests (A) Significantly increased average minimal current necessary to trigger tonic seizures in SV2B KO mice with the maximal electroshock (MES) model.N = 10 (SV2B WT) and 9 (SV2B KO) animals.(B) Unaltered average minimal current necessary to trigger partial seizures in SV2B KO mice with the 6 Hz seizure model.N = 14 (SV2B WT) and 13 (SV2B KO) animals.(C) SV2B KO mice behave in the Open field (actimetry) test like SV2B WT mice with same exploration distance (left) and number of rearings (right).N = 15 (SV2B WT) and 16 (SV2B KO) animals.(D) SV2B KO mice behave in the Y-maze test like SV2B WT mice with same number of entries (left) and alternation probability (right).N = 15 (SV2B WT) and 16 (SV2 BKO) animals.(E) Passive avoidance test measuring the latency to enter the chamber before and after a foot shock did not reveal any difference between SV2B WT and KO mice.N = 15 (SV2B WT) and 16 (SV2B KO) animals.(F-H) Motor performance assessed by the rotarod on three consecutive days did not differ between SV2B WT and KO mice.(F) first and (G) last trial, (H) best performance of all trials.N = 15 (SV2B WT) and 16 (SV2B KO) animals.(A-H) Data are shown as mean ± SEM.