Surface SV2A-Syt1 nanoclusters act as a sequestration hub that limits dynamin-1 recruitment and targeting to recycling synaptic vesicles

Following exocytosis, the recapture of plasma membrane-stranded vesicular proteins into recycling synaptic vesicles (SVs) is essential for sustaining neurotransmission. Surface clustering of vesicular proteins has been postulated as a ‘pre-assembly’ mechanism for endocytosis – ensuring high-fidelity retrieval. Here, we used single-molecule imaging to examine the nanoclustering of synaptotagmin-1 (Syt1) and synaptic vesicle protein 2A (SV2A) in hippocampal neurons. Syt1 forms surface nanoclusters through interaction of its C2B domain with SV2A, that are sensitive to mutations in this domain (Syt1K326A/K328A) and knocking down SV2A. SV2A co-cluster with Syt1 and blocking SV2A’s cognate interaction with Syt1 (SV2AT84A) also decreased SV2A clustering. Surprisingly, impairing SV2A-Syt1 nanoclustering enhanced plasma membrane recruitment of key endocytic protein dynamin-1, leading to accelerated Syt1 endocytosis, altered intracellular sorting and decreased trafficking of Syt1 to Rab5-positive endocytic compartments. SV2A-Syt1 surface nanoclusters therefore negatively regulate the rate of their own re-entry into recycling SVs by controlling the recruitment of the endocytic machinery.


Introduction
Synaptic vesicle (SV) recycling involves a balance between fusion (exocytosis) and retrieval (endocytosis) of SVs from the plasma membrane (PM) at nerve terminals during neurotransmission. Both exocytosis and compensatory endocytosis involve the coordinated actions of proteins and lipids to ensure high fidelity of vesicular protein retrieval during high rates of SV fusion. To sustain neurotransmission, nascent recycling SVs need to recapture essential vesicular machinery that are stranded at the PM. However, the mechanisms through which neurons retrieve essential vesicular proteins from the PM are not well defined. As certain SV proteins lack canonical recognition motifs for endocytic adaptor molecules, interactions between vesicular cargoes may facilitate recruitment of proteins from the PM into SVs, thus preserving vesicle protein stoichiometry (Takamori et al., 2006;Wilhelm et al., 2014) during neurotransmission. For example, vesicle-associated membrane protein 2 (VAMP2) is a soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) family member that regulates the fusion of SVs with the PM (Südhof and Rothman, 2009), whose internalisation is facilitated in part, via its interaction with synaptophysin (Gordon and Cousin, 2013;Gordon et al., 2011;Harper et al., 2021;Harper et al., 2017). Similarly, vesicular glutamate transporter 1 (vGlut1) facilitates the recruitment of multiple SV proteins from the PM into SVs (Pan et al., 2015). Thus, interactions between vesicular molecules are theorised to improve the fidelity of endocytic uptake and allow SVs to retain their discrete protein content during multiple rounds of fusion.
One mechanism through which protein interactions improve the fidelity of endocytosis is by forming nanoclusters. Following exocytosis, vesicular proteins cluster at the PM through protein-protein interactions. Notably, VAMP2 disperses following exocytosis and subsequently re-clusters via interactions with endocytic proteins, particularly AP180 and clathrin assembly lymphoid myeloid leukemia (CALM) (Gimber et al., 2015). The endocytic machinery therefore has the potential to initiate clustering of surface-stranded vesicular proteins. However, it is not clear what factors control the clustering of other vesicular proteins, such as synaptotagmin-1 (Syt1). Syt1 is a transmembrane SV molecule that is involved in calcium (Ca 2+ )-dependent exocytosis (Geppert et al., 1994) and forms clusters at the PM Willig et al., 2006). Syt1 binds to the PM phospholipid phosphatidylinositol-4,5-bisphosphate in a Ca 2+ -dependent manner through its cytosolic C2A and C2B domains (Bai et al., 2002;Schiavo et al., 1996;Stein et al., 2007) to mediate exocytosis. Syt1 forms a complex with another vesicular transmembrane protein, synaptic vesicle protein 2A (SV2A) (Bennett et al., 1992), which comprises twelve transmembranespanning domains capped by cytosolic C-terminal and N-terminal regions. The SV2A-Syt1 interaction occurs via the cytosolic domains of Syt1 and SV2A: hydrogen bonds are formed between two lysine residues (K326/K328) residing in the polybasic region of Syt1's Ca 2+binding C2B domain (Fernandez et al., 2001) and the T84 epitope on the N-terminus of SV2A upon phosphorylation of the T84 epitope by casein kinase 1 family kinases (Zhang et al., 2015).
For these reasons, SV2A is also a strong candidate as a regulator of Syt1 nanoclustering during SV recycling.
In this study, we investigated the role of protein-protein interactions in controlling the surface clustering of vesicular machinery at the PM, and how these interactions and clustering events facilitate the re-entry of proteins into recycling SVs during endocytosis. We hypothesised that nanoclustering of vesicular proteins at the PM allows for the generation of a 'readilyaccessible' pool of pre-assembled molecules, forming a depot from which vesicular proteins may be selectively retrieved into nascent recycling SVs. Using super-resolution imaging, we identified the determinants of Syt1 and SV2A nanoclustering by manipulating interactions between Syt1 and SV2A, as well as interactions with endocytic machinery. Syt1-SV2A interaction was shown to be critical for the surface nanoclustering of their respective partner, whilst manipulation of the endocytic machinery had no effect on the nanoclustering of either molecule. Blocking SV2A-Syt1 nanoclustering accelerated Syt1 retrieval during SV endocytosis. This manipulation also led to increased mobility of internalised Syt1 suggesting alterations in recycling SV nanoscale organisation. The findings presented in this study suggest that Syt1 is dynamically sequestered into nanoclusters in an activity-dependent manner through its interaction with the N-terminal tail of SV2A, and that the nanoclustering of Syt1 by SV2A decreases the kinetics of Syt1 endocytic reuptake. Accordingly, we report that SV2A interaction also controls the organisation of Syt1 following internalisation, causing Syt1 entrapment within endocytic pathways associated with early/bulk endosome formation. SV2A therefore plays a critical role in the recycling of Syt1, with implications for the function of Syt1 during multiple rounds of vesicle fusion.

Interaction of the Syt1 C2B domain (K326/K328) with SV2A controls the activitydependent confinement of PM-stranded Syt1 in nerve terminals
First, we investigated the surface mobility and nanoclustering of Syt1 in murine primary cultures of hippocampal neurons using universal Point Accumulation Imaging in Nanoscale Topography (uPAINT). This single-particle tracking technique allows for selective analysis of the nanoscale organisation of surface proteins via labelled ligand tracking (Giannone et al., 2010;Giannone et al., 2013;Joensuu et al., 2016). To visualise Syt1 on the surface of hippocampal neurons, we overexpressed Syt1 tagged with pHluorin (Syt1-pH; Fig. 1A i) -a pH-sensitive green fluorescent protein (GFP) that is quenched in the acidic SV environment and unquenched following exocytosis due to exposure to the neutral, extracellular pH (Miesenböck et al., 1998). Epifluorescence imaging of wild-type Syt1 (Syt1 WT -pH) revealed distinct synaptic boutons that lined the axon of mature neurons (DIV18) (Fig. 1A ii). To track single Syt1-pH molecules at nanoscale level on the PM, we applied atto647N-labelled anti-GFP nanobodies (NBs) (Kubala et al., 2010) in a depolarizing buffer to increase the level of SV fusion with the PM, and therefore the amount of Syt1-pH on the PM available for atto647N-NB binding (Fig. S1). The atto647N fluorophore was excited (647 nm) in total internal reflection fluorescence (TIRF) mode, to selectively image the surface population of atto647N-NB-bound Syt1 WT -pH (16,000 frames, 320 s) (Fig. 1B). Within nerve terminals, we observed the presence of distinct Syt1 nanoclusters . To investigate whether Syt1 undergoes co-clustering with cognate SV2 molecules, dual-colour imaging of Syt1 WT -pH-atto647N-NB with mEos2-tagged wild-type SV2A (SV2A WT -mEos2) was performed. Syt1 and SV2 nanoclusters overlapped along hippocampal nerve terminals (Fig. 1B). To determine whether Syt1 interaction with SV2A regulates the nanoscale organisation of Syt1 at the PM, we expressed a pH-tagged Syt1 mutant that has perturbed SV2A binding (Syt1 K326A/K328A -pH; (Borden et al., 2005). This Syt1 variant contains two lysine (K) to alanine (A) substitutions within the polybasic region of Syt1's C2B domain. A loss of Syt1 subsynaptic clustering was observed upon expression of the K326A/K328A mutant (Fig. 1D iii-v), despite Syt1 WT -pH and Syt1 K326A/K328A -pH being expressed at comparable levels ( Fig. 1E i).
This was quantified by calculating the area under the curve (AUC) of the MSD, and by calculating the ratio of mobile-to-immobile molecules (M/MM) (Fig. 1E iii). Both metrics were significantly decreased following stimulation (p=0.018 and p=0.008 respectively). Further, we used an intraluminal Syt1 nanobody and found that the mobility of overexpressed and endogenous Syt1 were identical (data not shown). These results demonstrate that the lateral entrapment of Syt1 molecules at the PM is an activity-dependent process and not due to increased surface expression. Therefore, to assess the impact of preventing Syt1 from interacting with SV2A, we compared the surface mobility of Syt1 WT -pH and Syt1 K326A/K328A -pH, taking advantage of the unquenching of the pH-tag to identify active presynapses. To control for the impact of Syt1 surface expression on Syt1 mobility, the number of Syt1 WT -pH and Syt1 K326A/K328A -pH detections at the PM were quantified post-stimulation and found to be similar ( Fig. 1E iv). As anticipated, Syt1 WT -pH mobility at the presynapse (Fig. F i) was lower compared to that of the whole axon ( Fig. 1E iii), suggesting that Syt1 is specifically confined at the presynaptic membrane. We next compared Syt1 WT -pH and Syt1 K326A/K328A -pH mobility by plotting their MSD and AUC of the MSD (Fig. 1F i), as well as their frequency distribution and M/MM ratio (Fig. 1F ii). Although the observed positive shift in the M/MM ratio for Syt1 K326A/K328A -pH was not significant (p=0.07; Fig.1F ii), the MSD was significantly higher than that of Syt1 WT -pH (p=0.048; Fig.1F i). These results suggest that the binding of Syt1 via its K326/K328 residues to SV2A controls its confinement at the PM in an activity-dependent manner.
shRNA-mCer, or empty-vector mCer (Fig. 2C). We also tracked Syt1 WT -pH-atto647N-NB in the presence of either empty-vector mCer, SV2A-shRNA-mCer (knock-down) or a bicistronic plasmid encoding both SV2A-shRNA and shRNA-resistant SV2A-mCer (knock-down rescue) ( Fig 2D). Notably, the knockdown of endogenous SV2A increased the surface mobility of Syt1 WT -pH-atto647N-NB at the presynapse, as evident by a significant increase in the AUC of the MSD (Fig. 2E), and a significant decrease in the percentage of immobile molecules (Fig.   2F). For rescued expression, Syt1 WT -pH-atto647N-NB was tracked in cells where SV2A was re-expressed using a bicistronic plasmid that encoded SV2A-shRNA and shRNA-resistant SV2A-mCer ( Fig 2D). This change in Syt1 WT -pH-atto647N-NB mobility in the presence of the SV2A-shRNA was alleviated upon rescue of endogenous SV2A expression using the shRNA- indicating that these effects were Syt1 specific. These results therefore suggest that SV2A plays a key role in sequestering Syt1 into nanoclusters at the PM. Although we cannot rule out that there are other proteins that also contribute to the trapping of Syt1 on the PM, the fact that VAMP2 mobility is unchanged following SV2A knockdown suggests that the entrapment effect of SV2A is specific to Syt1.

Quantification of Syt1 nanoclustering at the plasma membrane
The increased mobility of Syt1 in the absence of SV2A expression or interaction suggests that SV2A plays a role in organising Syt1 into nanoclusters at the PM. Based on this, we first defined the dimensions of Syt1 WT -pH nanoclusters using the newly-developed nanoscale spatiotemporal indexing clustering (NASTIC) analysis ( Fig. 3A-D) (Wallis et al., 2021). Under physiological conditions, Syt1 nanoclusters were found to be distributed across the axonal branches of hippocampal neurons (Fig. 3A). To assess whether these Syt1 nanoclusters were associated with endocytic sites, we performed dual colour imaging of Syt1 WT -pH-atto647N-NB using uPAINT in tandem with single-particle tracking photoactivated localization microscopy (sptPALM) of mEos4b-clathrin. This revealed that Syt1 nanoclusters were formed in regions devoid of clathrin, therefore suggesting that Syt1 nanoclustering at the PM occurs independently of clathrin-mediated endocytosis ( Fig. 3B iii-v). Next, to determine whether the increased mobility of the Syt1 SV2A binding mutant was due an alteration in nanoclustering, we used cluster analysis to quantify the size, density, and apparent lifetime of both Syt1 WT -pH and Syt1 K326A/K328A -pH nanoclusters. By doing so, we identified a portion of Syt1 WT -pH molecules that formed nanoclusters (4.68 ± 0.87 %). These nanoclusters had a mean area of Spatiotemporal alignment (xyt) of Syt1 WT -pH nanoclusters in the presynapse, and segregation from mEos4bclathrin clusters. (i) Enhanced view of a single Syt1 WT -pH nanocluster shown in A, with tracks (white), boundary (green dotted outline) and centroids (red). (ii) Lifetime of the nanocluster shown in B i (tracks rotated~90°across xyt) obtained using Nanoscale Spatiotemporal Indexing Clustering (NASTIC). Trajectory centroids of (iii) mEos4b-clathrin and (iv) Syt1 WT -pH-atto647N-NB were (v) merged to show exclusion of Syt1 nanoclusters (green) from clathrin (red). Syt1 nanocluster indicated by arrowheads. Scale bar = 0.5 µm. (C) Nanocluster metrics of Syt1 WT -pH and Syt1 K326A/K328A -pH (WT n=15 neurons and K326A/K328A n=12 neurons), quantified using (i) AUC of the MSD, (ii) rate of clustering (trajectories/s), (iii) apparent lifetime (s), (iv) density (trajectories/µm 2 ), (v) area (µm 2 ) and (vi) radius (µm). (D) Nanocluster metrics of Syt1 WT -pH following SV2A knockdown (mCer n=7 neurons; SV2A-shRNA-mCer n=6 neurons), quantified by (i) the percentage of clustered trajectories, (ii) cluster membership (trajectories/cluster), (iii) area and (iv) apparent lifetime. Statistical significance was determined using a Student's t-test.
Furthermore, although there was only a marginal decrease in the number of trajectories per cluster of Syt1 following knockdown of SV2A (p=0.055; Fig. 3D ii), a significant decrease in both nanocluster area (p=0.006; Fig. 3D iii), and apparent cluster lifetime (p=0.035; Fig. 3D iv) was observed. Taken together, these results demonstrate that expression of (and interaction with) SV2A facilitates the entrapment of Syt1 within nanoclusters at the PM.

SV2A mobility depends on binding to Syt1 but not to endocytic machinery
Our results demonstrate that the organisation of Syt1 into nanoclusters depends on its interaction with SV2A, and that these Syt1 nanoclusters do not co-localise with clathrin clusters at the PM. Based on this hypothesis, we next investigated whether SV2A nanoclusters follow similar dynamics. To this end, we imaged a Syt1-binding SV2A mutant (SV2A T84A ), and an SV2A mutant that lacks binding to clathrin adaptor AP2 (SV2A Y46A ) (Yao et al., 2010), and examined their nanoscale mobility (Fig. 4A). To validate that binding of Syt1 to SV2A T84A was perturbed, we co-immunoprecipitated SV2A from HEK-293T cells co-expressing Syt1-HA with either SV2A WT -mCer, SV2A T84A -mCer or SV2A Y46A -mCer, by use of an anti-GFP antibody which also recognises ECFP derivative mCer (Fig. 4B). The level of Syt1-HA pulldown was significantly reduced for SV2A T84A -mCer ( Fig. 4B-C). In contrast to this, the opposite effect was observed for SV2A Y46A -mCer AP2-binding mutant, with increased Syt1-HA binding detected ( Fig. 4B-C). This finding suggests that AP2 and Syt1 compete for binding to SV2A, likely due to the proximity of the Syt1 (T84) and AP2 (Y46) interaction sites (Fig.   4A). Next, we performed uPAINT imaging of hippocampal neurons expressing either SV2A WT -pH, SV2A T84A -pH, or SV2A Y46A -pH. SV2A WT -pH and SV2A Y46A -pH formed nanoclusters within presynaptic boutons, which were less prominent for SV2A T84A -pH ( Fig.   4D). Cluster analysis revealed a decrease in the apparent lifetime of SV2A T84A -pH nanoclusters compared to that of SV2A WT -pH ( Fig. 4D i). However, nanocluster area and membership were unchanged ( Fig. 4D ii-iii). Further, SV2A T84A -pH mobility was significantly increased in comparison to SV2A WT -pH ( Fig. 4E i-ii), demonstrating that the SV2A-Syt1 interaction is The pH tag (green) is located between the first and second transmembrane domains. Note the proximity of the N-terminal T84A (red) and Y46A epitopes (green). (B) Co-immunoprecipitation of Syt1-HA with either mCer (control), SV2A WT -mCer, SV2A T84A -mCer or SV2A Y46A -mCer from protein samples derived from HEK-293T cells, using anti-GFP antibody conjugated TRAP beads. Representative blots using total protein lysates (input), and GFP immunoprecipitation of SV2A-mCer (GFP-IP), are shown. (C) Level of Syt1-HA binding by either SV2A T84A -mCer or SV2A Y46A -mCer as normalised to that of SV2A WT -Cer (WT and T84A: n=7, Y46A: n=6 samples per group). (D) Super-resolved SV2A-pH-atto647N-NB (WT, T84A or Y46A) within the presynapse of hippocampal nerve terminals. For D Coeff panels, regions highlighted in warm colours represent points of low mobility. Arrowheads indicate points of nanoclustering. Scale bar = 1 µm. (i) Nanocluster apparent lifetime, (ii) area and (iii) membership of SV2A WT -pH-atto647N-NB (n=221 clusters) and SV2A T84A -pH-atto647N-NB (n=177 clusters). (E) Surface mobility of SV2A-pH-atto647N-NB (SV2A WT n=15; SV2A T84A n=14; SV2A Y46A n=13) at the presynapse, as assessed by (i) MSD and (ii) AUC of the MSD (µm 2 s × 10). (iii-iv) SV2A WT -pH-atto647N-NB mobility following treatment with either DMSO (control; n=16), or Dyngo4A in DMSO (30 µM for 30 min; n=15), as assessed by (iii) MSD and (iv) AUC of the MSD (µm 2 s × 10). Statistical significance was determined by a one-way ANOVA with a Tukey's test for multiple comparisons and a Students' t-test for single comparisons. essential for the trapping of both molecules at the PM. Loss of AP2 binding (SV2A Y46A -pH) however, had no effect on the surface mobility of SV2A (Fig. 4E i-ii), indicating that nanoclustering of SV2A is not regulated by the endocytic machinery. This was further confirmed by treating cells with Dyngo4A (30 µM for 30 min) -an inhibitor of the GTPase activity of dynamin (McCluskey et al., 2013), which failed to elicit a change in the MSD of SV2A WT -pH (Fig. 4E iii-iv). Overall, our results suggest that the nanoclustering of Syt1 and SV2A on the PM is controlled by their mutual direct interaction and is independent from their binding to endocytic machinery.

Activity-dependent retrieval of SV2A is controlled by AP2 but not Syt1
To determine whether the clustering of SV2A is associated with alterations in SV endocytosis, we next examined the kinetics of SV2A-pH retrieval. For this, we expressed either SV2A WT -pH, SV2A T84A -pH or SV2A Y46A -pH in hippocampal neurons and measured the fluorescence decay of the pH-tag following electrical field stimulation (300 action potentials (APs) at 10 Hz), prior to treating neurons with ammonium chloride (NH4Cl) (Fig. 5A). The loss of pH fluorescence is reflective of SV2A-pH retrieval kinetics during SV endocytosis, which is rate limiting in comparison to subsequent SV acidification (Atluri and Ryan, 2006;Granseth et al., 2006). Rapid de-acidification of SVs following NH4Cl treatment results in the total population of SV2A-pH becoming unquenched. These experiments revealed that retrieval of SV2A Y46A -pH was compromised, suggesting that interactions with AP2 are required for its efficient recovery during endocytosis (Fig. 5B i-ii). In comparison to this, impairing the interaction of SV2A with Syt1 (SV2A T84A -pH) did not have an impact on SV2A retrieval ( Fig. 5B i-ii). As there were no differences in the proportion of total SV2A-pH following NH4Cl treatment between SV2A WT -pH, SV2A T84A -pH and SV2A Y46A -pH, the observed delay in SV2A Y46A -pH retrieval was not due to alterations in SV exocytosis (Fig. 5B iii). These findings demonstrate that AP2-mediated endocytosis of SV2A-pH occurs independently of SV2A-Syt1 interaction and nanoclustering.

SV2A knockdown increases dynamin1 recruitment at the plasma membrane
The accelerated endocytosis of Syt1 in the absence of SV2A suggests that compensatory recruitment of endocytic machinery is taking place. For this reason, we examined the effect of perturbing SV2A expression on dynamin1 recruitment. Due to the reduced surface area and thin architecture of neurons, which would impede interpretation of these results, these experiments were instead conducted in pheochromocytoma (PC12) neurosecretory cells which have a larger surface area and have previously been used for SV2A-shRNA knockdown (Dong et al., 2006). Using TIRF microscopy, we conducted time-lapse imaging of GFP-tagged dynamin1 (Dyn1-GFP) (Xue et al., 2011). We analysed the effect of SV2A knockdown on the activity-dependent recruitment of dynamin1 to the PM, by transfecting PC12 cells with either SV2A-shRNA-mCer (knockdown) or empty vector mCer (knockdown control; Fig. 6A-C).
Stimulation of PC12 cells with BaCl (2 mM) resulted in a substantial increase in dynamin1 recruitment to the PM, as indicated by a significant increase in Dyn1-GFP fluorescence intensity (FI; Fig. 6B-C). Importantly, this effect was potentiated by knocking down SV2A ( Fig. 6B-C). Subsequently, we assessed the effect of SV2A knockdown on the clustering and mobility of dynamin1 on the PM at the presynapse, by performing sptPALM on hippocampal neurons co-expressing Dyn1-mEos2 with either mCer or SV2A-shRNA-mCer (Fig 6D-F).
Knockdown of endogenous SV2A led to an increase in synaptic clustering (Fig. 6D) as evidenced by a large decrease in the AUC of the MSD of Dyn1-mEos2 (Fig. 6E), and a shift in the M/MM ratio of Dyn1-mEos2 towards decreased mobility (Fig. 6F). These findings suggest that increased dynamin1 recruitment occurs in response to de-clustering of stranded Syt1 at the PM. To determine whether these effects also occurred for Syt1 in the absence of dynamin1 interaction, the mobility of a phospho-inhibitory mutant of Syt1 tagged with pH (Syt1 T112A -pH) was examined. This mutation is present within the juxtamembrane region of Syt1 (between the C2 domains; Fig. 6G i) and has been reported to decrease interaction with dynamin1 (De Jong et al., 2016;McAdam et al., 2015). No change in the surface mobility of Syt1 in either the axons or nerve terminals of hippocampal neurons was observed for this dynamin1-binding Syt1 mutant (Fig 6H-I). Further, pharmacological inhibition of dynamin's enzymatic activity using Dyngo4A treatment (30 µM for 30 min) also failed to elicit a change in Syt1 mobility (Fig. 6I). Coupled with the previous observation that SV2A mobility was unaffected by AP2 interaction and dynamin inhibition (Fig. 4), these results suggest that the endocytic machinery has a limited role in regulating the nanoscale organization of Syt1 and SV2A at the PM. Rather, Syt1-SV2 nanoclustering acts upstream of endocytosis by controlling the recruitment of the endocytic machinery.

SVs
The accelerated retrieval of Syt1 that was observed in the presence of the Syt1-binding SV2A mutant (SV2A T84A ) raises the question of whether SV2A controls the endocytic targeting of Syt1 to recycling SVs. If this is the case, interfering with SV2A may also causes intracellular missorting of Syt1. To address this question, we performed the sub-diffractional Tracking of Internalised Molecules (sdTIM) technique (Joensuu et al., 2017;Joensuu et al., 2016) which enables the imaging of internalised Syt1. Syt1-pH-positive neurons were pulsed (56 mM K + with anti-GFP atto647N-NB for 5 min) before being washed and chased under resting conditions (5.6 mM K + for 5 min). To selectively image the recycling pool of SVs containing Syt1 with greater accuracy, we cleaved the Tobacco-Etch Virus (TEV) sequence present between Syt1 and the pH tag, using an active TEV protease (60 U for 15 min), thereby removing the pH-atto647N-NB tag from Syt1-pH-atto647N-NB that remained stranded at the surface (Fig. 7A i-iii) (Gimber et al., 2015;Hua et al., 2011;Nair et al., 2013;Wienisch and Klingauf, 2006). As expected, we observed a decrease in Syt1 WT -pH and Syt1 K326A/K328A -pH fluorescence in the presence of the active TEV protease compared to the inactivated boiled TEV control (95 °C for 10 min; Fig. 7A iv) and were able to track distinct recycling SVs containing Syt1-pH-atto647N-NB across both the total axon and in nerve terminals following active TEV digestion (Fig. 7B). As a control, we assessed Syt1 WT -pH and Syt1 K326A/K328A -pH

Figure 7. Syt1 intracellular sorting is altered in the absence of SV2A interaction. (A)
Schematic depicting the removal of the pH-atto647N-NB tag from the surface fraction of Syt1-pH-atto647N-NB via active TEV protease digestion (60 U; 15 min), which enables the specific imaging of the internalised fraction of Syt1-pH-atto647N-NB. (i) Atto647N-NB binds to Syt1-pH present on the PM during stimulation (high K + ; 5 min). (ii) Syt1-pH-atto647N-NB is internalised following a chase step (low K + ; 5 min) into acidic vesicles which quenches the fluorescence of the pH tag. (iii) Active TEV protease cleaves pH-atto647N-NB from the remaining Syt1 surface fraction. (iv) Fold change (%) in Syt1-pH fluorescence in the presence of the inactive (boiled) and active TEV protease for Syt1 WT -pH (n=6 inactive, n=9 active) and for Syt1 K326A/K328A -pH (n=4 inactive, n=5 active). (B) Loss of Syt1 WT -pH surface fluorescence induced by active TEV protease which allows selective tracking of internalised Syt1 WT -pH-atto647N-NB within the recycling pool of SVs. Arrowheads indicate the trajectories of individual Syt1-positive recycling vesicles in a nerve terminal. Scale bar = 4 µm (axon) and 2 µm (presynapse). (C-D) Mobility of Syt1 WT -pH-atto647N-NB and Syt1 K326A/K328A -pH-atto647N-NB in the presence of (C) inactive and (D) active TEV, with the MSD and AUC of the MSD shown for (i-ii) the total axon and (iii-iv) the presynapse. (E) Hippocampal neurons were co-transfected with Rab5-mRFP and either Syt1 WT -pH (n=14) or Syt1 K326A/K328A -pH (n=17), fixed and imaged by 3D-structural illumination microscopy (3D-SIM). Scale bar = 2 µm (axon), 0.1 µm (zoom). Quantification showing (F) the number of Syt1-pH-atto647N-NB localizations per Rab5-positive vesicle surface, (G) Syt1-pH-atto647N-NB density within Rab5-mRFP surfaces (#spots/µm 3 ), and (H) the volume of Rab5-mRFP colocalised with Syt1-pH-atto647N-NB (µm 3 ). Statistical significance was determined using Student's t test. mobility in the presence of either inactive (Fig. 7C) or active (Fig. 7D) TEV protease in both the total axon (Fig. 7C i-ii and 7D i-ii) and at the presynapse (Fig. 7C iii-iv and 7D iii-iv). In the absence of surface digestion, Syt1 WT -pH and Syt1 K326A/K328A -pH had a comparable mobility across both the entire axon and within nerve terminals (Fig. 7C). This mobility was reduced upon surface digestion with active TEV (Fig. 7C-D), confirming that Syt1-pH-atto647N-NB was specifically imaged within the recycling pool of SVs. Although Syt1 WT -pH and Syt1 K326A/K328A -pH had similar mobilities across the whole axon following surface digestion ( Fig. 7D i-ii), surprisingly, Syt1 K326A/K328A -pH was significantly more mobile than Syt1 WT -pH at the presynapse (Fig. 7D iii-iv). This suggests that loss of SV2A binding causes intracellular missorting of Syt1 specifically in nerve terminals. It is unlikely that the observed difference in mobility between Syt1 WT -pH and Syt1 K326A/K328A -pH is due to stranding of Syt1-pH at the PM, as no significant differences were observed in the absence of digestion of the surface fraction of Syt1-pH. Therefore, loss of SV2A binding may cause Syt1 to enter an alternative endocytic compartment, leading to differences in intracellular sorting.

Loss of interaction with SV2A alters Syt1 trafficking, by sequestering it from Rab5endosomes
Syt1 has previously been shown to be localized in early and recycling endosomes following internalisation 20 minutes post-fusion (Diril et al., 2006). It is becoming increasingly apparent that at physiological temperatures, both clathrin-dependent and clathrin-independent cargo sorting occurs at the level of internalised endosomes (Ivanova et al., 2021;Kononenko et al., 2013;Watanabe et al., 2014). Therefore, differences between Syt1 WT -pH and Syt1 K326A/K328A -pH mobility upon internalisation may stem from mutant or wild-type Syt1 being either sorted back to the recycling SV or to the endo-lysosomal system. To address this, we examined the co-localisation of internalised Syt1 WT or Syt1 K326A/K328A with Rab5 -a GTPase associated with early endosomes (Bucci et al., 1992) and bulk endosomes (Kokotos et al., 2018).
Supplementary figure 2 (Figure 7). Hypothetical model: SV2A-bound Syt1 nanoclusters control the selective targeting of Syt1 to recycling synaptic vesicles. (A) Schematic showing (i) the SV2A-Syt1 interaction that forms between the K326/K328 epitopes at the polybasic region of the C2B domain of Syt1 WT -pH and the T84 epitope at the N-terminal tail of SV2A WT -pH, and (ii) disruption of the SV2A-Syt1 interaction via mutation (Syt1 K326A/K328A or SV2A T84A ). (B) Schematic depicting the effect of the SV2A-Syt1 interaction on the nanoclustering and subsequent internalisation of Syt1. (i) Syt1 forms nanoclusters on the PM that are highly immobile and may be preferentially sorted into Rab5-positive early/bulk endosomes. (ii) Unclustered Syt1 is more mobile and can recruit dynamin1 to promote its internalisation into recycling SVs, which occurs at a faster rate.

Discussion
Interactions between vesicular proteins on the PM have recently been proposed to help maintain the organisation, stoichiometry, and composition of SVs by enhancing the fidelity of endocytic events (Gordon and Cousin, 2013;Harper et al., 2021;Harper et al., 2017). In this study, we provide evidence that the nanoclustering of SV2A-Syt1 at the PM controls the targeting of stranded vesicular proteins into recycling SVs. We demonstrate that SV2A controls the nanoclustering of Syt1 at the PM through interactions between its cytoplasmic Nterminus and the K326/K328 epitopes within the polybasic region of the C2B domain of Syt1. Importantly, this mechanism also works in reverse -SV2A is also sequestered into nanoclusters through its interaction with Syt1. On the other hand, genetic manipulation of the interaction with the endocytic machinery (AP2 and dynamin), as well as pharmacological inhibition of endocytosis (Dyngo4a), have no impact on either SV2A surface mobility or nanocluster formation. Syt1-SV2A co-clustering acts upstream of endocytosis and controls the kinetics of Syt1 endocytosis. This was made evident by the increase in recruitment to the PM, and immobilisation of dynamin1 in response to SV2A-Syt1 cluster inhibition. Therefore, this negative regulation of the endocytic machinery by SV2A-Syt1 nanocluster formation ultimately controls the rate and selective targeting of these surface stranded SV proteins into recycling SVs.

Mechanisms regulating SV2A-Syt1 nanoclustering
The finding that SV2A controls Syt1 surface nanoclustering through interaction with the K326/K328 residues of Syt1 fits with previous observations that mutating these residues to alanines decreases Syt1 oligomerisation (Chapman et al., 1998). Here, we found that the reverse is also true -whereby SV2A nanoclustering is regulated via its interaction with these Syt1 residues. This is similar to what has been shown for other presynaptic molecules, such as syntaxin1A Lang et al., 2001;Sieber et al., 2006) and Munc18 (Kasula et al., 2016), which form nanoclusters dependent on molecular interactions and play key roles in exocytosis. Importantly, SV2A-Syt1 nanoclustering is not regulated by the endocytic machinery (AP2 or dynamin1), unlike VAMP2 which does form endocytic machinery-mediated nanoclusters (Gimber et al., 2015). Vesicular proteins are therefore pre-assembled into nanoclusters on the PM following exocytosis, either by direct binary interaction or via interaction with the endocytic machinery. This process mediates the fidelity of their re-uptake into recycling SVs. In the case of SV2A, this hypothesis is supported by the observation that Syt1 and AP2 compete for interaction with SV2A, and that the endocytic kinetics of SV2A is unaffected by Syt1 interaction (SV2A T84A ) and is instead dependent on AP2 binding (SV2A Y46A ). Unclustered Syt1 may be more accessible to the endocytic machinery, as the absence of SV2A potentiates dynamin recruitment and subsequent uptake of Syt1 into recycling SVs. Furthermore, the lack of bidirectional control of the retrieval of Syt1 and SV2A may be due to both molecules engaging different sets of endocytic adaptor molecules, which may lead to differing patterns of surface nanoclustering. For example, Syt1 has a unique interaction with stonin2, an endocytic adaptor that appears to exclusively shepherd Syt1 into clathrin-coated pits in concert with AP2 (Diril et al., 2006). Similarly to what we observed following SV2A knockdown, loss of stonin2 also accelerates the rate of Syt1 endocytosis (Kononenko et al., 2013).

SV2A-mediated nanoclustering regulates the targeting of Syt1 into recycling vesicles
Several factors may account for the accelerated kinetics of unbound Syt1 in the absence of SV2A. One possible explanation for fast-paced internalisation of unclustered Syt1 into SVs may be a result of the increased recruitment of endocytic machinery to the PM. In support of this, our results demonstrate that the activity-dependent recruitment of dynamin to, and immobilization on, the PM is potentiated in the absence of SV2A. Another potential explanation is that the accelerated re-entry of Syt1 into recycling SVs may occur via ultrafast endocytosis -a mode of SV recruitment in which dynamin1 plays a critical role (Imoto et al., 2022;Watanabe et al., 2013). SV2A-Syt1 nanoclusters may therefore act as a sequestration hub, that rate-limits endocytosis by controlling the level of dynamin1 recruitment. As such, surface nanoclustering may act as a transition state for Syt1, allowing neurons to fine-tune endocytosis and titrate Syt1 internalisation back into recycling SVs. This sequestration process may facilitate titration due to steric limitations owing to the size of Syt1 nanoclusters, which are significantly larger than both clathrin-coated pits (0.065-0.125 µm) (Kirchhausen and Harrison, 1981) and recycling SVs (0.040 µm in diameter) (Zhang et al., 1998). Removing AP2 binding (SV2A Y46A ) increased the interaction of SV2A with Syt1, raising the possibility that Syt1 interaction restricts SV2A's access to the endocytic machinery. Endocytic events are also controlled by Syt1 (Li et al., 2017;McAdam et al., 2015;Nicholson-Tomishima and Ryan, 2004;Poskanzer et al., 2003), as the C2 domains of Syt1 regulate the kinetics of vesicle internalisation in a Ca 2+ -dependent manner (Yao et al., 2012). As the binding of SV2A to the C2B domain of Syt1 is negatively regulated by Syt1's interaction with Ca 2+ (Schivell et al., 2005), SV2A may compete with Ca 2+ for C2B binding, thereby acting as a clamp. Therefore, it is tempting to speculate that the pace of Syt1-mediated endocytosis may be decreased through competitive interaction with Ca 2+ and nanoclustering with SV2A.
The potentiated recruitment of dynamin1 to the PM upon dispersal of Syt1 from nanoclusters suggests that unclustered Syt1 associates with the endocytic machinery to speed up the rate of Syt1 internalisation. However, the presence of alternative trafficking routes, of differing endocytic kinetics, may also account for the fast-paced recruitment of Syt1 into recycling SVs.
We demonstrate that SV2A-Syt1 interaction controls the intracellular sorting of Syt1 to Rab5positive early or bulk endosomes, with clustered Syt1 WT preferentially associating with Rab5 following stimulation, in comparison to the SV2A binding mutant (Syt1 K326A/K328A ). Therefore, Syt1 nanoclustering likely restricts its access to smaller endocytic pits, instead redirecting it into Rab5-positive early endosomes (Wucherpfennig et al., 2003). This rerouting may play a physiological role in preventing the build-up of SV2A-Syt1 nanoclusters stranded on the PM during repetitive rounds of SV fusion. This mechanism likely favours activity-dependent internalisation of SV2A-Syt1 into bulk endocytic compartments that are positive for Rab5 (Kokotos et al., 2018) and large enough to engulf these nanoclusters. Bulk endosomes, have been shown to act as a sorting station for SV cargoes in which proteins can either be rerouted back to the reserve pool of SVs (Cheung et al., 2010) or trafficked to the endo-lysosomal system (Ivanova et al., 2021). Therefore, activity-dependent bulk endocytosis of SV2A-Syt1 during sustained neurotransmission may act as an intermediate step in the reformation of SVs.

Functional consequences of SV2A-Syt1 nanoclustering on neurotransmission
The role of the SV2 family of proteins in neurotransmission is not well understood. Studies suggest that SV2A controls the size of the readily releasable pool of SVs (Custer et al., 2006), primes Ca 2+ -dependent SV fusion and regulates short-term synaptic plasticity (Chang and Südhof, 2009). Thus, as nanoclustering is theorised to determine the cellular function of proteins, our findings suggest that SV2A regulates Syt1 function by promoting its nanoclustering in both recycling vesicles and at the PM, ultimately controlling neurotransmitter release and plasticity. Understanding the molecular steps involved in this process provides insight into the role of SV2A and Syt1 during synaptic dysfunction, neurological disorders and potential avenues for therapeutic treatment. The critical importance of SV2A in vesicular targeting is described in a related study, which demonstrates that SV2 acts as a gateway for the entry of botulinum neurotoxin type-A (BoNT/A) into neurons, whereby BoNT/A hijacks SV2A-Syt1 nanoclusters to promote its own internalisation into SVs (Joensuu et al., 2022).
In summary, our results demonstrate that following exocytosis, PM-stranded vesicular proteins Syt1 and SV2A interact with each other to form nanoclusters independently of the endocytic machinery. These SV2A-Syt1 nanoclusters act upstream of endocytosis by rate-limiting the recruitment of the endocytic machinery, thereby controlling the rate of their retrieval into recycling SVs.

Lead contact
Information and requests for reagents, materials and resources should be directed to and will be fulfilled by Professor Frédéric A. Meunier (f.meunier@uq.edu.au).
Site-directed mutagenesis was performed to introduce T112A into Syt1 WT -pH backbone using

Super-resolution microscopy
For live single particle tracking, neurons were placed in low K + imaging buffer (

Fluorescence imaging
Timelapse recordings of pHluorin (pH) were performed as previously described (Harper et al., 2020 Cells were solubilised in HEPES buffer (50 mM HEPES pH 7.5, 0.5 % Triton X-100, 150 mM NaCl, 1 mM EDTA, 1 mM PMSF, protease inhibitor cocktail) for 1 h prior to centrifugation (17,000 g for 10 min), from which the resulting supernatant was isolated and treated with GFP-Trap beads (Chromotek, Germany) and rotated at 4 °C for 2 h, followed by additional wash (3X) in HEPES buffer. Samples were incubated in SDS sample buffer (10 min at 65 °C) and loaded on an SDS-PAGE gel for western blotting, which was carried out in accordance with previous studies (Anggono et al., 2006). Primary antibodies used were anti-GFP rabbit (Abcam, ab6556, 1:4000) and anti-HA rabbit (ICLlab, RHGT-45A-Z, 1:20000). IRDye secondary antibody (800CW anti-rabbit IgG (#925-32213, 1:10000) and Odyssey blocking PBS buffer (#92740000) were from LI-COR Biosciences (Lincoln, Nebraska, USA). Blots were visualised using a LI-COR Odyssey fluorescent imaging system (LI-COR Biotechnology, Cambridge, UK). Band densities were determined using LI-COR Image Studio Lite software (version 5.2). The amount of Syt1-HA co-immunoprecipitated was normalised to the amount of input protein. These values were then normalised to the amount of immunoprecipitated SV2A-mCer.

Image processing
For single particle tracking, image processing was carried out in PALMTracer, a customwritten software that operates in MetaMorph 7.7.8 (Molecular Devices, CA, USA) (Kechkar et al., 2013). Regions of interests (ROIs) were drawn around nerve terminals defined as hotspots of increased pHluorin fluorescence. As exocytic fusion occurs exclusively at the level of the synapse, we have systematically used pHluorin unquenching to delineate the presynapse.
Trajectories lasting a minimum of eight frames were selected and reconstructed. The mean square displacement (MSD) was calculated by fitting the equation MSD(t)=a+4Dt (where D is diffusion coefficient, a is y intercept and t is time), with MSD quantified over a 200 ms period.
The diffusion coefficient was calculated and divided into mobile and immobile populations with a diffusion coefficient of log10 > -1.45 µm 2 s -1 considered as mobile (Constals et al., 2015;Joensuu et al., 2016). A SharpViSu tool was used to perform drift correction for cluster analysis (Andronov et al., 2016). A custom-built python tool was used to perform NAnoscale SpatioTemporal Indexing Clustering (NASTIC) on our track files to determine the size, density, and apparent lifetime of Syt1 and SV2A nanoclusters. NASTIC generates a series of overlapping, spatiotemporal bounding boxes around trajectories used to determine cluster formation (Wallis et al., 2021). Nanoclusters were thresholded at a radius of 0.15 µm, with anything greater excluded from analysis.
Offline data processing of pH-transfected neurons was performed using Fiji Is Just ImageJ (Fiji) software (Schindelin et al., 2012). A script based on background thresholding was used to select nerve terminals, which placed ROIs of identical size over those responding to stimulation. Average fluorescent intensity was measured over time using the Time Series Analyzer plugin before screening ROIs using a customised Java program that allows for visualisation of the fluorescent responses and removal of aberrant traces from the data.
Subsequent data analyses were performed using Microsoft Excel, Matlab (Cambridge, UK) and GraphPad Prism 6.0 (La Jolla, CA, USA) software. The change in activity-dependent pHluorin fluorescence was calculated as F/F0 and normalised to the peak of stimulation.
3D-SIM processing and channel alignment were performed using Zen 2012 SP2 Black (version 11.0, ZEISS). Rab5-mRFP clusters were observed along the axon of each neuron.
Colocalization of points and surfaces was carried out in IMARIS (version 9.6.0). A series of 3D surfaces were defined for Rab5 cluster points (561 nm) across each z-stack based on the signal intensity across the neuron, with neuronal morphology defined based on Syt1-pH signal.
For each surface, the points corresponding to internalised Syt1-pH-atto647N-NB molecules (642 nm) were identified.

Statistics and data analysis
A Students t-test was performed for comparison between two groups ( Fig. 1-2, Fig. 3A, Fig.   3C iv-vi, D i-iv, Fig. 4C i-iii, Fig. 4D iv, Fig. 6-7). A one-way ANOVA was performed followed by a post-hoc Tukey's test for multiple comparison data with a gaussian distribution of residuals (Fig. 3C i-iii, Fig. 4B, Fig. 4D ii, Fig. 5B). For non-parametric analysis assuming no gaussian distribution, a Kruskal-Wallis test with Dunn's test for multiple comparisons was carried out (Fig. 5C-D). The level of significance was set to p<0.05. Error bars represent standard error of the mean (SEM). All n values correspond to independent cells/neurons unless specified (e.g., Fig. 4D i-iii where n corresponds to individual nanoclusters).

Data and code availability
A custom-built Python tool for NASTIC analysis was can be found at (Wallis et al., 2021) with the Python code available at https://github.com/tristanwallis/smlm_clustering. Analysis of pHluorin fluorescence in nerve terminals was carried out using a custom-made script based on background thresholding. This was used to select nerve terminals, which placed regions of interest of identical size over those responding to stimulation (Harper et al., 2020).