Synapsin 2a tetramerisation selectively controls the presynaptic nanoscale organisation of reserve synaptic vesicles

Neurotransmitter release relies on the regulated fusion of synaptic vesicles (SVs) that are tightly packed within the presynaptic bouton of neurons. The mechanism by which SVs are clustered at the presynapse, while preserving their ability to dynamically recycle to support neuronal communication, remains unknown. Synapsin 2a (Syn2a) tetramerization has been suggested as a potential clustering mechanism. Here, we used Dual-pulse sub-diffractional Tracking of Internalised Molecules (DsdTIM) to simultaneously track single SVs from the recycling and the reserve pools, in live hippocampal neurons. The reserve pool displays a lower presynaptic mobility compared to the recycling pool and is also present in the axons. Triple knockout of Synapsin 1-3 genes (SynTKO) increased the mobility of reserve pool SVs. Re-expression of wild-type Syn2a (Syn2aWT), but not the tetramerization-deficient mutant K337Q (Syn2aK337Q), fully rescued these effects. Single-particle tracking revealed that Syn2aK337QmEos3.1 exhibited altered activity-dependent presynaptic translocation and nanoclustering. Therefore, Syn2a tetramerization controls its own presynaptic nanoclustering and thereby contributes to the dynamic immobilisation of the SV reserve pool.

1.It is not entirely clear what is the fraction of synaptic vesicles being labeled with K+ loading and tracked in real time?Although it is a valid approach, it does raise some concern if such approach may yield results that are not necessarily relevant to more physiological stimuli such as field stimulation.Prolonged K+ depolarization may label different sub-pools of SVs and even induce activity-dependent plasticity that does not reflect the mobile dynamics of SVs among different pools under physiological conditions.
2. In Fig 6-7, Re-introduction of syn 2a WT seem to have over-suppressed the motility of reserved pool while its mobility with K337Q mutant was more comparable to that of , raising the question how meaningful such experiments are or whether these are artifacts from over-expression.
3. Synapsin 1-3 may form oligomers in native synapses so it might be too hasty to conclude that tetramerization of synapsin 2a is the core mechanism for presynaptic nanocluster of SVs.In fact, it would be ideal to use an approach to acutely perturb tetramerization, e.g.peptide, to destablise tetramer or other forms of oligomerization in WT synapses to strengthen such a conclusion.
Reviewer #2 (Remarks to the Author): Longfield et al present an advanced optical microscopy study on the live-cell dynamics of synaptic vesicles (SV) and their modulation by the protein synapsin2a.Through an optimized single-particle tracking algorithm, the authors were able to differentiate for the first time between recycling and reserve pools of SVs and how these are influenced by synapsin2a and especially the latter's the multimerization.This is a very clever and impressive approach with interesting results, which in principle deserve publication.Unfortunately, I find the story at parts difficult to follow due to several reasons, mainly since detailed explanations are missing.-I was not convinced that the authors really proved that they could distinguish between the two different pools by just changing the temporal pattern of the labelling protocol.How can they be completely sure that the SVs are really 100% recycling or reserve?Maybe I did not get the details here.
-Concrete quantifications of mobility are missing throughout the text -maybe state comparing numbers of diffusion coefficients?-The authors should discuss their results in light of the LLPS model.Can any links be drawn between both models, e.g. between multimerization and condensation?Can it not be interpreted in both ways or congruently?-What does AUC give us other than a qualitative measure of general mobility?Any quantitative interpretation?-The authors included some controls to check for biasing diffusion of unbound/free At565Nb.Maybe include more controls like the dye alone without Syt1pH?-End page 8, beginning page 9: I got confused here.On one hand the authors state that the absence of synapsin did not dramatically alter the mobility patterns, on the other hand they state that synapsins selectively lower the mobility of the reserve pool.A slight rephrasing and indication of concrete numbers (diffusion coefficients) would help.
- The reviewer is correct that high potassium stimulation does not mimic all the events that occur during physiological stimulation, as it elicits high-frequency spontaneous release likely to promote various types of endocytosis.However, it is considered a well-established paradigm to elicit neurotransmitter release, used for many years in a multitude of neuronal and neurosecretory cell types.As requested, we have now successfully labelled the reserve pool of SVs using a more physiological stimulation paradigm -high-frequency field stimulation (50 Hz, 300 AP).To investigate the effect of electrical stimulation on the reserve pool of SVs and confirm the efficacy of our labelling technique, our lab had to introduce an entirely new stimulation procedure.This involved importing specialised equipment, building the system, adapting it to our super-resolution acquisition procedure, and solving a multitude of technical and methodological problems.In addition, to successfully label and image the reserve pool of SVs, we had to adapt our new set-up to the sterile conditions needed to avoid contamination and preserve neuronal culture viability across the experimental duration.
These new experiments are now included in the new Supplementary Figure 4.These experiments demonstrate that there is no difference in the mean square displacement (MSD) between the high potassium and electrical stimulation paradigms 48 hours after labelling.This analysis was carried out both in the presynapses and axons and no differences were found in either compartment.We have also amended both the Method and Results sections to highlight these results:

Methods:
'Electrical field stimulation was performed to validate the efficacy of the high K + stimulation protocol (Supplementary Fig 4).To label the reserve pool of SVs, neurobasal medium containing 400 pM At565Nb was added to Syt1-pH transfected hippocampal neurons seeded in NuncTM 35 mm glassbottom dishes with 10 mm micro-wells (Thermo Scientific, #150680).The neurons were challenged with a train of 300 action potentials (APs) delivered at 50Hz (100 mA and 1 ms pulse width) using a 35mm dish insert with field stimulation electrodes Warner Instruments,Holliston,MA) and single channel digital stimulator (Warner Instruments, Holliston, MA, Panlab, #LE12106).After stimulation, unbound Atto565Nbs were washed off with fresh neurobasal medium, and neurons were then chased for 48 hrs at 37°C in their original conditioning medium.'

Results:
'Lastly, to test whether the depolarising stimulus (high K + ) captured the dynamics of the reserve pool vesicles, we tested a more physiological stimulation paradigm.We performed high-frequency field stimulation (50 Hz, 300 action potentials (AP)) in sterile conditions in mature hippocampal neurons expressing Syt1pH (as above) in the presence of anti-GFP-At647Nbs and chased for 48h.The mobility of the labelled vesicles was indistinguishable from those observed following high K + stimulation (Supplementary Fig. 4), thereby validating our high K + stimulation protocol.'Figures 6 and 7 do not relate to the mobility of the reserve pool of synaptic vesicles.They examine the mobility of Synapsin 2a-mEos2 itself in resting and stimulated conditions.The reviewer may have been referring to Figure 8, where the mobility of the reserve pool of vesicles was investigated in SynTKO neurons following rescue conditions.During the revision of this manuscript, we realised that these data were underpowered.As a result, we imported more SynTKO hippocampi and repeated these experiments (increasing the number of neurons analysed).We have now also included a new set of analyses focusing on the mobility of reserve SVs within the axons (Figure 9h, i).Importantly, we found that the mobility of axonal reserve SVs in neurons rescued with KQ mutant was equivalent to SynTKO.

SynTKO hippocampal neurons. (i) Comparisons of the ratio of detections in the axons and presynapses
from reserve SVs in SynTKO neurons (black), reserve SVs when Syn2a K337Q -mEos3.2 is expressed (magenta), and reserve SVs when Syn2a WT -mEos3.2 is expressed (cyan).Data are displayed as mean ± SEM.Values were obtained from n ≥ 33 (presynapses) and n ≥ 7 (axons) from at least 12 neurons in f and i.Data was obtained from ≥ 3 biological replicates.Statistical comparisons were performed using one-way ANOVA and Dunnett's or Tukey's multiple comparisons test.
The reviewer is right that the effect of Synapsin 2a WT re-expression may be stronger than expected.
Our interpretation of this response is that as Syn2a is the only isoform able to fully rescue the SynTKO phenotype in excitatory neurons (Gitler et al., 2008), it likely overcompensates when expressed in SynTKO in isolation.It is important to note that this effect was much reduced with the tetramerisation mutant, suggesting a key role of synapsin tetramerisation in immobilising the reserve pool of vesicles.
We have now included a discussion paragraph on the potential consequences of synapsin homo-and hetero-dimers.

Discussion:
'Synapsins have been shown to form oligomeric structures both in vitro and in vivo 22,48,68 .Synapsins structural domain (C-domain) mediate its assembly into tetramers.Previous studies have identified key residues in the C-domain of synapsin that mediate its assembly into tetramers, that are essential for SV-tethering.One missense mutation known to perturb tetramer assembly, used in our study (K337Q) was unable to rescue reserve vesicle mobility and the SV density at the presynapse in SynTKO neuronal cultures.This suggests that homo-tetramerization of Syn2a is sufficient to rescue reserve pool mobility dynamics within the pre-synapse.Hetero-tetramerization, and/or synergistic/opposing roles of Synapsin1, in the background of Syn2a function, is likely to be important in regulating the SV cluster.It has been suggested that domain E (present in Synapsin1a, 2a and 3a isoforms) may adopt a different conformation, potentially during oligomerisation or within BMCs.In addition, such intramolecular interactions could regulate the targeting of Synapsin isoforms into different presynaptic or axonal subcellular compartments.' 3. Synapsin 1-3 may form oligomers in native synapses so it might be too hasty to conclude that tetramerization of synapsin 2a is the core mechanism for presynaptic nanocluster of SVs.In fact, it would be ideal to use an approach to acutely perturb tetramerization, e.g.peptide, to destabilise tetramer or other forms of oligomerization in WT synapses to strengthen such a conclusion.
It is unquestionable that synapsin can generate biomolecular condensates (BMCs) in vitro and in neurons (Zhang and Augustine, 2021;Gitler et al., 2004;Hosaka and Südhof, 1999).However, the mechanism by which the reserve pool of synaptic vesicles is immobilised in nerve terminals is still an open question.Our results suggest that the tetramerization of Synapsin 2a could be involved in the cross-linking of synaptic vesicles.It is still unclear how this mechanism only targets the reserve pool of SVs and whether synapsin tetramerization is involved in modulating the ability of other Synapsins to undergo phase separation, forming BMCs in the synapse.More work is therefore needed to distinguish between these two possibilities.
We thank the reviewer for their suggestions.The mutant form of Synapsin 2a (Syn2a K337Q ) that we used has been reported to selectively prevent the tetramerization of this Synapsin (Gitler et al, 2008;Song and Augustine, 2023).We do not currently have tools to acutely perturb oligomerization.We agree with the reviewer that our results should be interpreted carefully, in light of various recent papers including (Hoffmann et al., 2023;Song and Augustine, 2023).We have therefore, carefully amended the discussion to reflect this discrepancy in our interpretation and to lay the basis for future studies.Discussion: "Synapsin isoforms have been shown to form oligomeric structures both in vitro and in vivo22,48,68.

Synapsins structural domain (C-domain) mediate its assembly into tetramers. Previous studies have
identified key residues in the C-domain of synapsin that mediate its assembly into tetramers, that are essential for SV-tethering.One missense mutation known to perturb tetramer assembly, used in our study (K337Q) was unable to rescue reserve vesicle mobility and the SV density at the presynapse in SynTKO neuronal cultures.This suggests that homo-tetramerization of Syn2a is sufficient to rescue reserve pool mobility dynamics within the pre-synapse.Hetero-tetramerization, and/or synergistic/opposing roles of Synapsin1, in the background of Syn2a function, is likely to be important in regulating the SV cluster.It has been suggested that domain E (present in Synapsin1a, 2a and 3a isoforms) may adopt a different conformation, potentially during oligomerisation or within BMCs.In addition, such intramolecular interactions could regulate the targeting of synapsin isoforms into different presynaptic or axonal sub-cellular compartments."

Reviewer #2 (Remarks to the Author):
Longfield et al present an advanced optical microscopy study on the live-cell dynamics of synaptic vesicles (SV) and their modulation by the protein synapsin2a.Through an optimized single-particle tracking algorithm, the authors were able to differentiate for the first time between recycling and reserve pools of SVs and how these are influenced by synapsin2a and especially the latter's multimerization.This is a very clever and impressive approach with interesting results, which in principle deserve publication.Unfortunately, I find the story at parts difficult to follow due to several reasons, mainly since detailed explanations are missing.
We thank the reviewer for their positive remarks about our approach.We have now increased the critical details of our methodology to improve the readability and interpretation of our manuscript.
-I was not convinced that the authors really proved that they could distinguish between the two different pools by just changing the temporal pattern of the labelling protocol.How can they be completely sure that the SVs are really 100% recycled or reserve?Maybe I did not get the details here.
Our protocol is based on previously published work done by Truckenbrodt et al. (2018).Here, the authors studied the aging of synaptic vesicle proteins using antibodies directed against the luminal domain of synaptotagmin 1 (syt1) to reveal the localization and activity status of vesicle molecules over time.They showed that (1) antibody-labelled syt1 proteins gradually decrease in synapses over several days (0-10 days with 48hrs standing out as a key time point).( 2) Antibody-labelled syt1 molecules become rapidly inactive in terms of exocytosis over time, that many labelled molecules still remain within synapses but do not participate in release under normal network activity.Further, that these molecules can be induced to release by strong stimulation.Finally, (3) Newly synthesized proteins are preferentially incorporated into actively recycling vesicles compared to inactive vesicles and that inactivated vesicles do not return to the actively recycling pool under normal network activity.Data all suggesting that as syt1 SV proteins age, they incorporate into a population of vesicles that are no longer fusogenic under normal network activity and do not co-localise with new recycling vesicles (aka the reserve pool).
As all the work done by Truckenbrodt et al. (2018) was carried out in fixed preparations, we were able to expand on this research in our investigation of SV incorporation into the reserve pool of live hippocampal neurons (Figure 1).Our results clearly show a significant and progressive decrease in the mobility of labelled vesicles following longer chasing times, as expected from the reserve pool.Further, we assessed the mobility of this vesicular pool in response to restimulation and, as expected from the reserve pool, no change in mobility was detected.This suggests that they do not respond to stimulation and undergo fusion with the plasma membrane as previously found with the recycling pool of SVs (Joensuu et al., 2016;Figure 3a-f).Additional context as to the choice of pulse chase intervals has been included in the Results section and we have more clearly referred the important body of work that has helped shape our new experimental paradigm.

Results:
'These images qualitatively show the lower mobility of Syt1pH/At565Nb trajectories at 48 hrs, solidifying the importance of this time point in SV maturation, and complimenting previous Syt1 experiments performed by Truckenbroldt et al, (2018).In this study the authors demonstrated that Syt1 labelled 48 hours earlier, no longer co-localised with newly labelled SV proteins and were no longer fusogenic, suggesting that these 'old' Syt1-tagged vesicles were segregated from the recycling pool and transitioned into the reserve pool 43 .Further quantification of all time points showed a progressive decrease in SV mobility over time, specifically within the presynaptic terminal (Fig. 1f, g), but not in the axonal compartment (Fig.1h,i).The lowest SV mobility in the presynaptic compartment was exhibited at 48 hrs, confirming that by this time recycling SVs had transitioned into the reserve pool.' To further investigate the differences between the reserve and recycling population of SVs we have set up a collaboration with David Holcman's Group (Group of Data Modelling and Computational Biology, IBENS, Ecole Normale Superieure, 75005 Paris, France) to perform high-throughput statistical analysis on our data.This approach is based on computing diffusion and density maps as well as identifying a high-density regions of SV trajectories in live hippocampal neurons.Here, we found that the reserve pool of SVs had a significantly altered probability of generating sub-micron regions that act as 'sinks' with higher energy (kT) and residence time.These new data are presented in the new Figure 4 and we have included our interpretation of this information in our Results and Discussion sections.

Results:
'To further characterize the differences between recycling and reserve pools of vesicles, we used a high throughput statistical approach based on computing the diffusion and density maps as well as identifying the high-density regions 44-46 .The underlying biophysical model (equation 2 in Methods) assumes that vesicles can either move according to Brownian motion or interact with the local environment that can stabilize them in given sub-micrometre subregions.The density and diffusion maps are constructed from the ensemble of trajectories by estimating the local density and diffusion coefficients in a grid map decomposed into small bins (see Methods for the statistical estimators eqs. 9 -5).The results are shown in Fig. 4a-j.The density map reveals areas (red) characterized by local highdensity regions (HDRs) the recycling and reserve SV populations of live hippocampal neurons (Fig. 4g-j illustrates examples of potential well or SV traps in these pools).The diffusion map reveals a more uniform distribution of SV mobility within the axonal compartment (Fig. 4e-f).To analyse these HDRs, we use the potential paradigm framework 44 under which such regions result from long-range force interaction.We first observed that high-density regions are much more frequent for reserve (n=77) than recycling vesicles (n=4) for the same neuronal region.
We further characterized these HDRs by extracting their boundaries and associated energy using an automated classification algorithm 47 .We report here that the size of the wells associated with the reserve pool was larger (0.18 ± 0.07 µm) than those for recycling (0.1 ± 0.01 µm) (Supplementary Fig. 5).In addition, the stability of the reserve pool measured by the energy of the well shows that it is higher (E = 2.15kT) than for recycling (E = 0.94kT) (Fig 4k).Lastly, we determined the residence time in confinement versus Brownian diffusion of trajectories within the wells.We found that the potential wells were able to confine reserve SVs for at least 3x times longer than SVs of the recycling pool (Fig. 4l-m).The reserve pool of SVs is therefore characterized by a much larger number of high-density regions defined as potential wells.These wells are larger in size (Supplementary Fig. 5) and their energy is also larger compared to that of the recycling pool (Fig. 4k).This (1) confirms that the two populations of DsdTIM labeled SVs are indeed different (as they display unique mobility patterns) and ( 2) suggests that the reserve pool is much more stable, and the associated mechanism of SV trapping involves a greater force than that of the recycling pool.' Discussion: 'In depth analysis of single SV trajectories reveals that the most striking difference between the reserve and recycling pools stemmed from the number and size of high-density regions, that were much more prominent for the reserve pool SVs.These high density regions were both larger and more stable for the reserve pool, suggesting that the two types of vesicles interact differently with their nanoscale presynaptic environment.Surprisingly, high density regions were also found along the axon which could indicate either that these are silent synapses 58

Figure 5 :
This quantification must include the labelling degree, otherwise it might be biased.Any comment/control?-Page 10, middle: What stimulation?Name it.-Page 10, middle: What is meant by low MSD? -Page 10, middle: What is meant by "regulates the nanoscale organisation"?Did not get the argument here.Point-by-point response to referees Reviewer #1 (Remarks to the Author): Using dual-pulse sub-diffractional Tracking of Internalised Molecules (DsdTIM) to simultaneously track SVs from the recycling and reserve pools, Longfield et al studied the mobility of SVs tagged by two fluorophores in the axon and boutons of cultured hippocampal neurons.They found that SVs in the reserve pool shows a lower mobility than those in the recycling pool in boutons, but similar mobility in axons.The mobility of SVs in the reserve pool is preferentially elevated in triple knockout of synapsins (Synapsin 1-3 TKO).This phenotype can be rescued by reintroduction of WT Synapsin2a but its K337Q mutant devoid of the tetramerization failed to do so.These observations led the authors to conclude that Synapsin2a tetramerization allows presynaptic association of SV clustering to reduce their mobility in the reserve pool of presynaptic boutons.Overall this study provide some interesting observations on the dynamics of SVs in two pools distributed in boutons and axons respectively.The results are acquired with advanced optic technology and innovative imaging and analysis paradigms, yielding solid results.However, the conceptual novelty of this study is rather marginal, given that synapsin2a tetramerization has already been highlighted but the other referred paper.The author may want to consider the perturbations that promote or inhibit tetramerization to give this paper new twist.Major comments: 1.It is not entirely clear what is the fraction of synaptic vesicles being labelled with K+ loading and tracked in real time?Although it is a valid approach, it does raise some concern if such approach may yield results that are not necessarily relevant to more physiological stimuli such as field stimulation.Prolonged K+ depolarization may label different sub-pools of SVs and even induce activity-dependent plasticity that does not reflect the mobile dynamics of SVs among different pools under physiological conditions.
either high K+ or electrical field stimulation.(a, c) Average MSD of reserve SVs labelled using either a high K+ buffer (blue) or train of 300 APs (50Hz for 6s; orange), within the (a) presynapses and (c) axons.(b, d) Area under the MSD curve (AUC; µm2 s) for (b) presynapses and (d) axons.Data are displayed as mean ± SEM.Values were obtained from n ≥ 15 (presynapses) and n ≥ 7 (axons) from at least 7 neurons in a to d.Data was obtained from 1 independent neuronal culture.Statistical comparisons were performed using Student's t test or Mann-Whitney U test.
Fig 6-7, Re-introduction of syn 2a WT seem to have over-suppressed the motility of reserved pool while its mobility with K337Q mutant was more comparable to that of WT (MSD values in Fig 3-4), raising the question how meaningful such experiments are or whether these are artifacts from over-expression.

Fig. 1 .
Fig. 1.Tracking the reserve pool of SVs in live hippocampal neurons.(a) Graphical representation of the sdTIM protocol optimization for the reserve pool of SVs: DIV19 hippocampal neurons expressing Synaptotagmin1-pHluorin (Syt1pH) were stimulated with high K + medium containing anti-GFP Atto565-tagged nanobodies (At565Nbs) (red) for one minute.Following stimulation, the excess

Fig 3 .
Fig 3. Quantification of the reserve and recycling pool mobilities.(a, c) Average MSDs of the resting reserve pool (RP) of SVs (black), RP SVs after stimulation (red) and the recycling pool of SVs (green) within the (a) presynapses and (c) axons.(b, d) Area under the MSD curve (AUC; µm 2 s) for (b) presynapses and (d) axons.(e, f) Average diffusion coefficient of the RP (black), RP SVs after stimulation (red) and the recycling pool of SVs (green) within the (e) presynapses and (f) axons.(g, h) Comparisons of the density of detections the axons and presynapses from (g) recycling and (h) reserve vesicles tagged with anti-GFP At565Nbs or At647Nbs respectively, normalised by the area (traj/µm 2 ).Data are displayed as mean ± SEM.Values were obtained from n ≥ 28 (presynapses) and n ≥ 11 (axons) from at

Fig. 4
Fig. 4 Density and diffusion maps describing the differences between reserve recycling SVs.(a, b) Representative neuron where recycling (a) and reserve (b) SVs displayed as individual trajectories.(c, d) Density map of recycling (c) and reserve (d) SVs.(e, f) Diffusion maps of recycling (e) and reserve (f) SVs where the high-density regions are delimited by an ellipse (red).(g-j) High density regions of recycling (g, h) and reserve (i, j) SVs are characterized by converging arrows (4 colours corresponding to 4 main directions).The ellipse corresponds to the boundary found by automated algorithms 47 and the energy of the wells is in the unit of kT.(k) Energy (in kT) of the associated recycling and reserve SV potential wells.(i) Residence time (s) inside a potential well of the associated recycling and reserve SVs.Data are displayed as mean ± SEM.Values were obtained from n = 4 (recycling pool potential wells) and n = 77 (reserve pool potential wells) from at least 5 neurons in panels k and l.Data was obtained from ≥ 3 independent neuronal cultures.Statistical comparisons were performed using Kolmogorov Smirnov test.
and/or that the reserve pool has an intrinsic ability to generate clusters in axons.The analysis of the reserve pool SV trajectories in synapsin TKO neurons revealed a clear role of synapsin in shaping both the size of these high density regions and the strength of the interactions between vesicles of the reserve pool.It remains unclear how these high-density regions are generated and what mechanism(s) determine their size, which can extend for up to two hundred nanometres.In particular, it is unclear how these long-range forces are generated via synaptic protein self-assembly mechanisms and/or interactions with cytoskeletal elements of the presynapse.How the distinct mechanisms underpinning the clustering of reserve and recycling SV pools cross-talk to filter recycling SV to the active zone and restrict the reserve vesicles from accessing this critical zone will require further investigation.Further, we established the role of synapsin 2a tetramerization in dynamically anchoring the reserve pool of SVs at the presynapse.The conventional understanding of how SVs cluster at the presynapse is currently under scrutiny, particularly as we make the technological advances necessary to study protein interactions at this scale.Ultrastructural analysis of SynTKO neuronal synapses revealed dispersed SVsand altered SV clusters adjacent to the active zone 24 .Synapsins were shown to undergo LLPS and mediate membraneless compartments called biomolecular condensates (BMCs) at the presynapses 24 .Synapsin BMCs could underpin the formation of the potential wells described herein and characterized by long-range interactions.'-Concrete quantifications of mobility are missing throughout the text -maybe state comparing numbers of diffusion coefficients?