Tomosyn associates with secretory vesicles in neurons through its N- and C-terminal domains

The secretory pathway in neurons requires efficient targeting of cargos and regulatory proteins to their release sites. Tomosyn contributes to synapse function by regulating synaptic vesicle (SV) and dense-core vesicle (DCV) secretion. While there is large support for the presynaptic accumulation of tomosyn in fixed preparations, alternative subcellular locations have been suggested. Here we studied the dynamic distribution of tomosyn-1 (Stxbp5) and tomosyn-2 (Stxbp5l) in mouse hippocampal neurons and observed a mixed diffuse and punctate localization pattern of both isoforms. Tomosyn-1 accumulations were present in axons and dendrites. As expected, tomosyn-1 was expressed in about 75% of the presynaptic terminals. Interestingly, also bidirectional moving tomosyn-1 and -2 puncta were observed. Despite the lack of a membrane anchor these puncta co-migrated with synapsin and neuropeptide Y, markers for respectively SVs and DCVs. Genetic blockade of two known tomosyn interactions with synaptotagmin-1 and its cognate SNAREs did not abolish its vesicular co-migration, suggesting an interplay of protein interactions mediated by the WD40 and SNARE domains. We hypothesize that the vesicle-binding properties of tomosyns may control the delivery, pan-synaptic sharing and secretion of neuronal signaling molecules, exceeding its canonical role at the plasma membrane.


Western blot
Lysate for Western blot analysis was prepared by homogenizing cells in denaturing Laemmli sample buffer and boiling for 5 min at 100˚C. Cells had been in culture for 14 days and were infected with lentivirus the day after plating. After SDS-PAGE and wet protein transfer to a PVDF membrane for 2 h at 350 mA at 4˚C, nonspecific antibody binding to the membrane was prevented by incubation with blocking solution (5% w/v milk powder and 0.2% Tween-20 in TBS, pH 7.5) for 1 h at 4˚C. Primary antibody incubation was done overnight at 4˚C. After washing with TBS, the membrane was stained with secondary antibody conjugated with alkaline phosphatase (AP; DAKO, Glostrup, Denmark, 1:5000) for 1 h at 4˚C. After washing again, the AP conjugated antibody was visualized using ECF substrate (GE Healthcare, Little Chalfont, UK). The membrane was scanned with a Fujifilm FLA-5000 Reader.

Immuno-electron microscopy
Hippocampal neuron cultures with or without EGFP-tomosyn-m1 overexpression were fixed (4% formaldehyde, 0.1% glutaraldehyde in phosphate buffer) at room temperature on DIV14 and prepared for cryo-sectioning according by the Tokuyasu method [38]. Briefly, the cells were scraped in gelatin and spun down to a semi-compact pellet. Specimen blocks were cut out, infused with 2.3 M sucrose at 4˚C and mounted on aluminium pins by rapid freezing in liquid nitrogen. 70 nm sections were obtained at -120˚C using a cryo-ultramicrotome (UC6, Leica). Sections were captured on carbon/formvar-coated copper mesh grids. Grids were immuno-labeled using anti-tomosyn-1 (#183103; SySy, Göttingen, Germany, 1:50) and protein-A-gold 10nm (PAG10; CMC Utrecht, The Netherlands) as electron-dense marker. Grids were counterstained by uranyl acetate in methylcellulose before analysis on TEM (Tecnai T12, FEI).

Immunoprecipitation of tomosyn-1 protein complexes
A hippocampal P2 + microsome fraction (Li et al., 2012) was mixed with an equal volume of extraction buffer (2% Triton X-100, 150 mM NaCl, 50 mM HEPES pH 7.4) and protease inhibitor (Roche Applied Science, Penzberg, Germany), and rotated at 4˚C for 1 h. After centrifugation at 20,000 × g for 20 min, the pellet was re-extracted with extraction buffer (1% Triton X-100, 150 mM NaCl; 50 mM HEPES pH 7.4 and protease inhibitor). Supernatants from both extractions were pooled, centrifuged at 20,000 × g for 20 min and served as IP input. Typically 5 mg P2 + M and 10 μg antibodies were used for each IP experiment. After overnight incubation, 50 μl slurry of Protein A/G PLUS-Agarose beads (Santa Cruz Biotechnology, Dallas, TX) was added for each IP and rotated at 4˚C for 1 h. The beads were spun down at 1000 × g for 1 min and washed four times in ice-cold extraction buffer (as above but with 0.1% Triton X-100). Beads with bound protein complexes were mixed with 2x SDS-PAGE loading buffer, and heated to 98˚C for 5 min. Five μl 30% acrylamide was added and the mixture was incubated at RT for 30 min. Proteins were resolved on a 10% SDS polyacrylamide gel, fixed overnight and stained with Coomassie Blue for 30 min. Each sample lane was cut into five slices for in-gel digestion and peptide recovery as described [40].

LC-MS-MS characterization of tryptic peptides
Peptides were re-dissolved in 20 μl 0.1% acetic acid and analyzed by an LTQ-Orbitrap mass spectrometer (Thermo Electron, San Jose, CA, USA) equipped with an HPLC system (Eksigent, Redwood City, CA). Samples were trapped on a 5 mm Pepmap 100 C18 (Dionex, Sunnyvale, CA) column (300 μm ID, 5 μm particle size) and then analyzed on a 200 mm Alltima C18 homemade column (100 μm ID, 3 μm particle size). Separation was achieved by using a mobile phase consisting of 5% acetonitrile, 94.9% H 2 O, 0.1% acetic acid (solvent A) and 95% acetonitrile, 4.9% H 2 O, 0.1% acetic acid (solvent B), with a linear gradient from 5 to 40% solvent B in 40 min at a flow rate of 400 nl/min. Eluted peptides were electro-sprayed into the LTQ-Orbitrap operated in a data-dependent mode. Mass spectrometric data was searched against the Uniprot proteomics database (version 2013-01-06) with MaxQuant software (version 1.3.0.5) to obtain peptides and proteins identified in each experiment, as well as their label-free abundance. Search parameters were: MS accuracy 6 ppm, MS-MS accuracy 0.5 Da, fixed modification of cysteine alkylation with acrylamide, variable modification of methionine oxidation and protein N-terminal acetylation, digestion with trypsin, protein hits containing at least one unique peptide, and false discovery rates of both peptides and proteins within 0.01.

Statistical analysis
Statistical analysis was performed using SPSS (Version 20.0, Armonk, NY). Since data were not normality distributed (Kolmogorov-Smirnov), Mann-Whitney tests for two independent samples were applied.

Diffuse and punctate distribution of tomosyn immunoreactivity in axons and dendrites
Intracellular tomosyn-1 distribution was assessed using immunocytochemistry on primary cultures of mouse hippocampal neurons. Specificity of the custom-made antibody was confirmed by the absence of staining in Tom-1 KO/KO neurons (S1 Fig). As expected for a synaptic protein [17,20,28], punctate localization of tomosyn-1 was observed ( Fig 1A). Interestingly, puncta were present both in axons and dendrites, supporting the previous notion that neuronal tomosyn is not confined to presynaptic sites [28]. In addition, as also reported before [20,30,31], diffuse tomosyn expression was observed. Lentiviral expression of a N-terminal EYFP-tagged splice variant of tomosyn-1 (EYFPtomosyn-m1; Fig 1B) yielded a similar distribution (expression levels were 3.6 ± 1.25 times higher than endogenous tomosyn-1 mean ± s.e.m.; n = 5; see typical immunoblot in Fig 1C). Expression of an EYFP-tagged splice variant of tomosyn-2 (EYFP-tomosyn-xb2) also resulted in a diffuse and punctate distribution, overlapping with endogenous tomosyn-1 ( Fig 1D). Notably, EYFP-tomosyn-xb2 and endogenous tomosyn-1 did not strictly co-localize in all neurite extensions. Thus, in line with several previous observations, both tomosyn isoforms localized both in the cytosol and in clusters along neurites.
To investigate the nature of tomosyn puncta, we performed co-localization experiments with markers for various organelles involved in synaptic function and secretory trafficking (Fig 2). Tomosyn-1 puncta co-localized with the SV proteins VAMP2 and synapsin-1 ( Fig 2E  and 2F), the synaptic marker bassoon ( Fig 2C) and the DCV cargo protein chromogranin B ( Fig 2H). However, none of these markers showed complete overlap with tomosyn-1 puncta, suggesting that tomosyn-1 expression was not restricted to any single type of organelle. The degree of co-localization between total EYFP-tomosyn-m1 (both diffuse and punctate) and the various markers was quantified by Pearson's correlation [36] and Manders' coefficients [37], producing the highest scores for syntaxin-1 and synaptotagmin-1 (Syt-1; Fig 2K-2M). The colocalization with VAMP2 was also observed for endogenous tomosyn-1 (S2 Fig). To achieve a higher spatial resolution, we also analyzed cultured hippocampal neurons by immune-electron microscopy. In line with the findings from light microscopy, tomosyn immunoreactivity was enriched in synaptic boutons (N-P for endogenous and 2Q for overexpressed tomosyn-1 in Fig 2) where it was either dispersed in the cytosol (Fig 2N) or associated with small clear vesicles ( Fig 2O) or DCVs ( Fig 2P). All in all, these results suggest that tomosyn-1 is localized not only to the cytosol, but also to (clusters of) SVs and LDCVs.

Tomosyn-1 and -2 co-migrate with synapsin and NPY in living neurons
Tomosyn localization to neuronal secretory vesicles is conceivable given its association with secretory granules in INS-1E cells [32], SVs from rat brain [16] and C. elegans DCVs [21] as well as the direct interaction of rat tomosyn with the vesicular proteins Syt-1 [41] and Rab3 [29]. To differentiate between immobile synapses and mobile organelles, we performed live imaging of EYFP-tomosyn-m1 puncta and observed that many tomosyn-1 puncta moved along the neurite (typical example in Fig 3A-3C). A kymograph representation shows bidirectional movement of these puncta ( Fig 3C). In some cases, new puncta emerged from existing ones (stable or moving; open arrowheads), suggesting the segregation of vesicles from a cluster. Within 30 s, 30.3 ± 0.02% of puncta changed movement direction (mean ± s.e.m.; n = 21 cells). Since vesicular trafficking seems to be activity-dependent, puncta mobility upon neural stimulation was tested [42]. Without stimulation, the mean velocity of moving tomosyn puncta was 0.34 ± 0.01 μm/s (n = 896 puncta from 25 cells). Upon high-frequency field stimulation the speed was slightly reduced to 0.30 ± 0.01 μm/s (Fig 3C and 3D; the stimulation period is depicted by black/white inversion in Fig 3C; Mann-Whitney U = 308464.5, z = 2.823, p = 0.005, r = 0.070; n = 749 puncta from 25 cells).

Molecular mechanism of vesicular tomosyn-1 targeting
The vesicular accumulation of tomosyn could involve various molecular interactions. The proteinaceous surface of synaptic vesicles purified from rat brain has been thoroughly characterized [16] and contains four known tomosyn-1 interactors: SNAP25, syntaxin-1, Syt-1 [25,41,44,45] and Rab3 [29]. The interaction with SNAP25 and syntaxin-1 involves the C-terminal coiled-coil (CC) domain of tomosyn which can engage in a stable four-helical bundle [25]. The other two interactions are both mapped to the large N-terminal domain [41,46]  interactions offer plausible possibilities for vesicle binding, we explored the synaptic interactome for potential novel interactions and performed a series of immunoprecipitation (IP)  Vesicle targeting of tomosyn by redundant interactions experiments from mouse brain synaptosomes, followed by mass spectrometry (MS) to identify each interactor. To consider the most robust interactions, we focused on interactions that were confirmed in a reciprocal experiment (i.e. with swapping the bait and prey proteins). IP-MS supported the previously established tomosyn-1 interaction with syntaxin-1a (Stx1A), SNAP25 and Syt-1 in multiple independent experiments [25,41,44,45]. Reverse IP-MS analysis of these proteins confirmed the presence of tomosyn-1 (see Table 1, summarizing the intensity-based quantification or iBAQ values of interactors identified by mass spectrometry). Despite clear evidence from previous studies, our approach did not detect the Rab3 interaction, possibly because this interaction is dependent on GTP activation. Furthermore, this approach did not identify novel interactions. In view of the strong co-localization with Syt-1 ( Fig 2D) and its important role in secretion via both SVs [34,47] and DCVs [48,49], we first tested whether the vesicular co-localization of tomosyn depends on the presence of Syt-1. Co-localization and co-migration of EYFP-tomosyn-m1 with vesicular markers was assessed in Syt-1 deficient (Syt-1 KO/KO ) hippocampal neurons. These neurons completely rely on Syt-1 for synchronous synaptic transmission [34]. The amount of co-localization between tomosyn-1 and VAMP2 or chromogranin B was unaffected (Fig 7A-7E). Moreover, co-migration of tomosyn-1 puncta with synapsin-mCherry ( Fig 7F) and NPY-mCherry (Fig 7G) was still observed in absence of synaptotagmin-1: in Syt-1 WT/WT neurons, 78 ± 4.1% moving NPY puncta contained tomosyn (n = 9 cells), compared to 76 ± 4.2% in Syt-1 KO/KO neurons (n = 8 cells). Thus, even though Syt-1 is a known tomosyn-1 interactor, it is not essential for its vesicular targeting. Likewise, we probed whether tomosyn-1 interaction with the SNAREs syntaxin-1 and SNAP25 is essential for its vesicular targeting. This interaction depends on tomosyn's C-terminal SNARE domain [44]. Lentiviral expression vectors for various tomosyn fragments carrying C-terminal deletions (Fig 8A) co-migrated normally with SVs and DCVs as shown by live imaging (see constructs "WD40-Tail" and "WD40" in Fig 8; see S2 and S3 Files), suggesting that SNARE interactions are not essential for vesicular targeting of tomosyn.
To further investigate the potential role of interactions involving the tomosyn WD40-1 and WD40-2 domains (Rab3 and Syt-1), we also investigated the effect of N-terminal or internal deletions on tomosyn's vesicular co-localization. All tested EYFP-tomosyn constructs were observed to co-migrate with synapsin-mCherry and NPY-mCherry (Fig 8B) to some extent. Quantitative analysis showed a reduced percentage of EYFP-positive synapsin-mCherry puncta in cells that expressed the tomosyn fragments named "Tail-CC" and "CC" (Fig 8C; Mann-Whitney U = 350; z = -4.662 for WT vs Tail-CC and U = 245; z = -5.239 for WT vs CC; n = 42 cells for WT, 41 for Tail-CC and 37 for CC), suggesting that tomosyn's N-terminal domain contributes to synaptic vesicle binding. Nevertheless, the SNARE domain fragment (residues 1047-1116 of tomosyn; CC) still co-localized with more than 50% of the synapsin-mCherry puncta (see arrowheads in Fig 8B; see also S4 and S5 Files). According to structural models, three loops are predicted to emanate out of the N-terminal domain structure [31]. Deletion of each of these loops separately (Δloop1, Δloop2, Δloop3), as well as mutation of the SUMOylation site located in loop 2 at residue K730 (K730R), did not block vesicle targeting. Similar results were obtained for large dense-core vesicles labelled by NPY-mCherry, except that the constructs Tail-CC and CC did not show a reduced percentage of tomosyn-positive mCherry puncta (Fig 8D). In a next experiment, we tested whether the vesicular accumulation could be the result of two redundant interactions: one via Syt-1 and another via a SNARE-pairing mechanism on tomosyn's C-terminal domain. Even in Syt-1 KO/KO neurons, the lack of the C-terminal Tail and CC domains did not abolish the co-migration of tomosyn fragments with synapsin-mCherry (Fig 9B, see arrowheads). Quantitative analysis showed that the degree of co-localization with synapsin-mCherry was still higher than 80% in all tested constructs (Fig 9C, data from 9 cells with 369 puncta for the WT construct, 303 for WD40-Tail, 364 for Tail-CC, 241 for CC and 290 for WD40). In this smaller experiment, the Tail-CC construct showed a similar trend towards a lower degree of co-localization as in Syt-1 wildtype cells (compare with Fig 8C), where the "CC"  N-and C-terminal domains. A) Wild type and mutant EYFP-tomosyn-1m constructs were co-expressed with synapsin-mCherry or NPY-mCherry using lentiviral vectors. mCherry-labelled puncta were observed by live imaging during 60s at 1 frame/s. Previously mapped interaction domains with Syt-1, Rab3, SNAP25 and construct showed near-complete SV association. Taken together, these data show that tomosyn binds to secretory vesicles by a mechanism that does not require Syt-1.

Discussion
Tomosyn is a cytosolic inhibitor of secretion that localizes both pre-and postsynaptically in diverse model systems [17,20,28]. Despite the absence of a membrane anchor, some evidence points to an association of tomosyn with secretory vesicles [16,21,32]. Here we studied the localization of tomosyn in cultured hippocampal neurons. Besides a diffuse distribution in neurites and accumulation at synapses, tomosyn co-localized with moving SVs and DCVs in living neurons. The presence of at least a third type of tomosyn-containing transport organelles was suggested by fast-moving tomosyn puncta that did not co-migrate with synapsin-or NPY-mCherry (Fig 5E and 5F). In line with the broad distribution of neuronal secretory vesicles [42,50], tomosyn puncta were observed in both axons and dendrites. The association of tomosyn with secretory vesicles is apparently driven by multiple redundant interactions in the N-and C-terminal domains. The observation that the SNARE domain alone is sufficient for co-migration with secretory vesicles suggests a contribution of the reported SNARE interaction [25,44,45,51]. In addition however, the isolated N-terminal domain is also able to bind to vesicles. This activity is not attributable to Syt-1 binding [41], leaving Rab3 as the most likely candidate [29]. Interestingly, the yeast tomosyn ortholog Sro7p also associates with secretory vesicles through an interaction with the Rab GTPase Sec4p [46,52]. This interaction is GTPdependent and was mapped to the boundary between the two WD40 propellers of Sro7p, suggesting a highly conserved role in vesicular trafficking. Compared to full length tomosyn, reduced co-migration of tail-CC and CC tomosyn-m1 fragments with synapsin puncta, but not NPY puncta, was observed, suggesting that the tomosyn binding modes may differ between SVs and DCVs.

Functional implications
The ability to associate with the vesicle membrane should be taken into account in functional models for tomosyn-dependent secretory regulation. As suggested by previous findings, tomosyn could contribute to organelle trafficking in several ways: (i) by regulating vesicle motility through interactions with motor proteins, or (ii) by directing SNARE-dependent vesicle tethering/docking to their correct target sites.
The first hypothesis is supported by studies in yeast, where the tomosyn orthologue Sro7p [56] interacts with the actin-binding motor protein myosin Va, implicated in polarized exocytosis [57]. In neurons, Myosin Va regulates retrograde axonal transport of DCVs [53] as well Vesicle targeting of tomosyn by redundant interactions as local movement of SVs [58]. In our experiment using synaptosomes however, type V myosin did not co-immunoprecipitate with tomosyn-1 or -2. Anterograde transport of SV proteins is mediated by the neuron-specific kinesin motor protein KIF1A [59,60]. In C. elegans ventral cord motor neurons, axonal targeting of tomosyn is regulated by the KIF1A homolog Unc-104 [20,61]. Again in our synaptosome preparation, KIF1A did not co-IP with tomosyn-1 or -2. We also did not detect an effect of overexpressed tomosyn on the velocity of vesicles (Fig 5G). Thus, whereas the essential role of motor proteins in vesicle trafficking is undisputed, our collective data suggests no major role for tomosyn in regulating the activity of motor proteins in mammalian hippocampal neurons.
According to the second hypothesis, tomosyn may co-migrate with secretory organelles to inhibit release at off-target sites and support their delivery at the correct destination. In mature synapses, recycling SVs are shared between release sites [5][6][7], a process that could contribute to synaptic plasticity during repetitive stimulation. In our study we frequently observed the stopping of moving tomosyn puncta at an immobile fluorescent structure. This, as well as the observed co-localization with bassoon, suggests that the immobile structures are likely synapses. We also observed events where a single moving fluorescent structure split two or more fluorescent structures, or vice versa (see a few examples in Fig 8). This supports the existing idea that vesicles can be co-transported in clusters [7]. In a recent study in insulin-secreting INS-1 cells, tomosyn tightly associated with NPY-GFP labeled DCVs where it remained associated until near the time of vesicle fusion and then diffused away [62]. Taken together, these observations support the hypothesis that tomosyn contributes to the trafficking and delivery of secretory vesicles.

Potential mechanisms for tomosyn-mediated cargo delivery
Which mechanism could be responsible for capturing secretory vesicles at their correct destinations? First, the interaction of VAMP2, the vesicular SNARE, is thought to contribute to this specificity by forming trans-complexes with its cognate t-SNAREs, syntaxin-1 and SNAP25 [63]. However, trans-SNARE complex formation may be preceded by molecular interactions of Syt-1 with the t-SNARE complex [64,65]. If tomosyn associates with both Syt-1, syntaxin-1 and SNAP25 during their vesicular transport, it is conceivable that the higher concentrations of t-SNAREs at release sites [13][14][15][16] may induce tomosyn-mediated capturing of secretory vesicles by forming a tethering complex between Syt-1, tomosyn itself and the t-SNARE complex. Such complexes have indeed been detected in the LP2 fraction from rat cerebral cytosol [66] and in direct pulldown experiments [41]. Subsequently, tomosyn may aid transitioning to an exocytic SNARE complex [67].
If tomosyn interacts with proteins on the destination site to deliver vesicles at their release sites, this would presumably involve significant conformational changes. Structural and functional studies of tomosyn and its yeast orthologue Sro7p have suggested a model where the Nterminal WD40 domains can support intramolecular interactions with the C-terminal domain [31,52,56]. Intramolecular rearrangements likely affect the accessibility of the SNARE domain for t-SNARE pairing and for tomosyn's activity in regulating neurotransmission [68]. Posttranslational changes in tomosyn or its binding partners, such as the phosphorylation of tomosyn by PKA [17], by Cdk5 via interaction with the GTP-bound state of Rab3A [29] or of syntaxin-1 by the Rho/ROCK pathway [69] could affect the likelihood of such structural rearrangements.
In conclusion, tomosyn is historically thought to be a soluble inhibitor of vesicular transmitter release that hampers vesicle fusion by v-SNARE competition. In this study, tomosyn localization to secretory and/or transport vesicles was observed, which challenges this classical view. Besides co-transport with other proteins engaged in secretion, vesicular tomosyn might be involved in spatial restriction of vesicle fusion and synaptic capturing of secretory vesicles.