Tetanus insensitive VAMP2 differentially restores synaptic and dense core vesicle fusion in tetanus neurotoxin treated neurons

The SNARE proteins involved in the secretion of neuromodulators from dense core vesicles (DCVs) in mammalian neurons are still poorly characterized. Here we use tetanus neurotoxin (TeNT) light chain, which cleaves VAMP1, 2 and 3, to study DCV fusion in hippocampal neurons and compare the effects on DCV fusion to those on synaptic vesicle (SV) fusion. Both DCV and SV fusion were abolished upon TeNT expression. Expression of tetanus insensitive (TI)-VAMP2 restored SV fusion in the presence of TeNT, but not DCV fusion. Expression of TI-VAMP1 or TI-VAMP3 also failed to restore DCV fusion. Co-transport assays revealed that both TI-VAMP1 and TI-VAMP2 are targeted to DCVs and travel together with DCVs in neurons. Furthermore, expression of the TeNT-cleaved VAMP2 fragment or a protease defective TeNT in wild type neurons did not affect DCV fusion and therefore cannot explain the lack of rescue of DCV fusion by TI-VAMP2. Finally, to test if two different VAMPs might both be required in the DCV secretory pathway, Vamp1 null mutants were tested. However, VAMP1 deficiency did not reduce DCV fusion. In conclusion, TeNT treatment combined with TI-VAMP2 expression differentially affects the two main regulated secretory pathways: while SV fusion is normal, DCV fusion is absent.

Soluble N-ethylmaleimide-sensitive factor attachment receptor (SNARE) complex formation is essential for secretion in the two main regulated secretory pathways in neurons, synaptic vesicle (SV) and dense core vesicle (DCV) exocytosis 1,2 . While the SNARE complex that typically drives SV fusion, consisting of syntaxin-1, synaptosomal associated proteins of 25 kDA (SNAP-25) and vesicles associated membrane protein 2 (VAMP2 or synaptobrevin 2), has been studied in great detail 2,3 , the cognate SNARE proteins for DCV fusion are still poorly characterized.
The seven genes of the VAMP family all encode proteins that contain a SNARE and transmembrane domain 2,4,5 . All VAMP proteins, except VAMP5, are reported to form functional SNARE complexes 5 . VAMP2 is the most abundant and widely distributed VAMP protein in the brain 6,7 . In hippocampal neurons, deletion of VAMP2 expression (VAMP2 knock out (KO)) impairs calcium-dependent SV fusion in most but not all neurons 8,9 . Inhibiting VAMP2 expression in cortical neurons using short-hairpin RNA reduces DCV exocytosis 10 . VAMP1 (synaptobrevin 1) is highly expressed in the spinal cord and less abundant in the brain compared to VAMP2 11 , but may be the main VAMP isoform in certain brain areas 7,12 or neuronal subtypes 9,13 . In a subset of VAMP2 KO neurons which highly express VAMP1, SV fusion is less impaired 9 . Loss of VAMP1 reduces synaptic transmission at the neuromuscular junction 11 . Selective knockdown of VAMP1 in rat trigeminal neurons reduces release of calcitonin gene-related peptide (CGRP), a dense core vesicle cargo peptide, whereas selective Scientific RepoRtS | (2020) 10:10913 | https://doi.org/10.1038/s41598-020-67988-2 www.nature.com/scientificreports/ cleavage of VAMP2 with botulinum neurotoxin B (BoNT/B) does not, suggesting VAMP1 to be the main isoform in these neurons 14,15 . Secretory granule exocytosis is only mildly affected in VAMP2 KO chromaffin cells and gene-inactivation of VAMP3 (cellubrevin) has no effect. However, granule exocytosis is severely impaired in the absence of both VAMP2 and VAMP3, suggesting functional redundancy 16 . VAMP3 expression is undetectable in neurons 8,17 but highly expressed in glial cells where it mediates secretion of NPY containing vesicles 18 . Taken together, although VAMP2 is the major isoform for SV and DCV exocytosis, VAMP3 can partly take over its function in chromaffin cells and VAMP1 drives at least some SV fusion in neurons as well as CGRP release from trigeminal ganglionic neurons. It is unknown if all DCV fusion in the brain is mediated by VAMP2 or if functional redundancy between isoforms occurs. Here, we studied the role of VAMP proteins in neuronal DCV exocytosis in single hippocampal neurons. Neurons were treated with tetanus neurotoxin (TeNT), which specifically cleaves VAMP 1, 2 and 3 but not the other four VAMP proteins 10,[19][20][21][22][23] . TeNT disrupted both SV and DCV fusion. Expression of a tetanus-insensitive (TI) version of VAMP2 (VAMP2 Q76V, F77W 24 ) restored SV fusion. However, expression of TI-VAMP1, TI-VAMP2 or TI-VAMP3 did not restore DCV exocytosis while both TI-VAMP1 and TI-VAMP2 travel with DCVs through neurites. To study the potential role of VAMP1 in DCV exocytosis we used vamp1lew (lethal wasting, hereafter referred to as vamp1 -/-) mouse that lacks VAMP1 25 . DCV fusion in vamp1 -/neurons was unaffected. Hence, TeNT treatment combined with TI-VAMP2 expression differentially affects SV and DCV fusion and may be used as a tool to selectively inhibit DCV fusion, leaving SV fusion unchanged.
TeNT insensitive (TI)-VAMP2 efficiently restores SV exocytosis upon TeNT treatment. To determine whether VAMP2 is sufficient for regulated exocytosis, we introduced TeNT insensitive (TI)-VAMP2 in TeNT treated neurons. TeNT cleaves the bond between glutamine 76 and phenylalanine 77 in VAMP2 23 . Mutation of these sites to valine and tryptophan, respectively, renders VAMP2 resistant to TeNT cleavage ( Fig. 2A) 24,33,34 . TI-VAMP2 (VAMP2 Q76V, F77W) was N-terminal tagged with mCerulean 24 which enabled detection of infected neurons during live cell experiments.
We first confirmed the functionality of TI-VAMP2 in SV exocytosis using synaptophysin-pHluorin 35 . Neurons were infected with synaptophysin-pHluorin and with TeNT, TeNT and TI-VAMP2, or with a control construct. Lentiviral infection of TeNT efficiently cleaved endogenous VAMP2 but not TI-VAMP2, which was expressed at a similar level as endogenous VAMP2 (Fig. 2B). Upon the same trains of action potentials as used before for DCV exocytosis (16 trains of 50 AP at 50 Hz), SV exocytosis was detected as an increase in fluorescence at puncta (Fig. 2C). TeNT treatment abolished SV fusion (Fig. 2D, red line) as shown before 23,36,37 . Co-infection of TI-VAMP2 restored SV exocytosis (Fstim max control: 0.46 ± 0.05, TeNT: 0.01 ± 0.0, TeNT + TI-VAMP2: 0.36 ± 0.05; Fig. 2E). This indicates that TI-VAMP2 is functional and sufficient to support SV exocytosis.
As a consequence, TeNT may still bind to the TI-VAMP constructs and potentially block DCV fusion via steric hindrance. In addition, TeNT expression is known to activate transglutaminases which could indirectly affect DCV fusion 40,41 . To test the potential effects of steric hindrance or transglutaminase activation we designed a proteolytically inactive TeNT construct by substituting glutamic acid 234 for glutamine (TeNT-E234Q) 21,42 .
The construct was N-terminally tagged with an HA-tag to validate expression and placed in an IRES-mCherry vector to visualize expression during live-cell experiments. Hippocampal neurons infected with NPY-pHluorin and a control, TeNT or TeNT-E234Q construct were stained for the HA-tag and VAMP2 (Fig. 7A). We found no loss of VAMP2 staining after lentiviral infection with TeNT-E234Q (Fig. 7A). These results confirm TeNT-E234Q to be proteolytically inactive 21,42 . Next, we studied DCV fusion in neurons co-infected NPY-pHluorin and a control, TeNT or the TeNT-E234Q construct. We found no difference in DCV fusion events between control or TeNT-E234Q infected neurons (DCV fusion events control: 181.1 ± 57.6, TeNT: 0 ± 0, TeNT-E234Q: 187.1 ± 50.1, Fig. 7B-D). To conclude, these results suggest that TeNT does not block DCV fusion via steric hindrance or transglutaminase activation.

Discussion
In this study, we assessed the role of VAMP proteins in neuronal DCV exocytosis in mouse hippocampal neurons. TeNT treatment, which cleaves VAMP1, 2 and 3, abolished SV and DCV fusion, as expected 10,19,20,23,36,37 . TI-VAMP2 restored SV fusion, but not DCV fusion in TeNT infected neurons. TI-VAMP1 and TI-VAMP3 also failed to restore DCV exocytosis in TeNT treated neurons. Hence, TeNT treatment combined with TI-VAMP2 expression differentially affects DCV and SV fusion. We explored two possibilities for this unexpected difference between the two secretory pathways. First, we hypothesized that the cleaved transmembrane part of VAMP2, which could remain on the vesicle after TeNT cleavage, prevents targeting of enough TI-VAMP2 molecules to support DCV fusion despite sufficient targeting  www.nature.com/scientificreports/ to SVs. However, overexpression of the cleaved transmembrane VAMP2 fragment did not block DCV fusion in WT neurons (Fig. 6). We therefore conclude that the lack of rescue of DCV fusion by TI-VAMP2 after TeNT cleavage cannot be explained by steric hindrance of this fragment. Secondly, we explored the possibility that DCVs require both VAMP1 and VAMP2 in a sequential scenario involving an upstream step before fusing with the plasma membrane. We considered VAMP1 to play a role in DCV fusion because VAMP1 deficiency reduces spontaneous and evoked synaptic transmission in the mouse neuromuscular junction 11 . In VAMP2 KO neurons, residual synaptic transmission correlates with VAMP1 expression levels, and silencing of VAMP1-expression further reduces synaptic transmission 9 . Hence, VAMP1 is expressed in neurons and functions in synaptic transmission in the peripheral and central nervous system. Furthermore, VAMP1 is localized on DCVs in the rat spinal cord 49 and involved in release of CGRP in rat trigeminal ganglionic neurons 14,15 . In addition, we found that overexpressed VAMP1 and VAMP2 travel together with DCVs to a similar extent in hippocampal neurons www.nature.com/scientificreports/ (Fig. 5). However, DCV fusion in vamp1 -/hippocampal neurons was unaffected (Fig. 8), indicating that VAMP1 is not essential for DCV exocytosis. VAMP3 is undetectable in neurons 8,17 (Fig. 1) and was therefore not considered in a sequential scenario. In addition, acute knockdown of VAMP1 or VAMP3 did not reduce BDNF release in cortical neurons 10 . Hence, because (1) VAMP1 is dispensable for DCV fusion; (2) VAMP3 is not detected in hippocampal neurons; and (3) TeNT only cleaves VAMP1, 2 and 3, we conclude that VAMP2 is the only known vSNARE suitable to support DCV fusion in these neurons. However, TI-VAMP2 could not rescue DCV fusion in TeNT treated neurons despite similar expression levels as endogenous VAMP2 and rescuing SV fusion (Fig. 2). This could be due to a number of reasons. First, TeNT cleaves VAMPs that are present on the vesicles. SVs contain on average 75 VAMP molecules 50 and in liposomes and chromaffin cells, vesicles are estimated to require one to three SNARE complexes for fusion 51,52 . The number of VAMP molecules per DCV is unknown but DCV fusion pores in chromaffin cells are estimated to contain six to eight SNARE complexes 53,54 . Hence, DCVs likely require more VAMP molecules for fusion than SVs. Secondly, SVs locally recycle after fusion, increasing the likelihood of incorporating TI-VAMP2 into their membranes. DCVs do not recycle after fusion and the lifetime of a DCV after budding from the TGN is unknown. Therefore, incorporation of TI-VAMP2 molecules is likely slower. Although TI-VAMP1 and TI-VAMP2 travelled together with DCVs (Fig. 5), the number of TI-VAMP2 molecules per vesicle may be insufficient to support DCV fusion but sufficient for SV fusion after TeNT treatment. Taken together, loading of relatively less TI-VAMP2 on DCV and a higher demand for TI-VAMP2 molecules to support DCV fusion could contribute to the inability of TI-VAMP2 to rescue DCV fusion after TeNT treatment.
TeNT is also known to activate transglutaminases which is shown to have minor effects on synaptic transmission by itself 40,41 . However, transglutaminase activity could have a stronger negative effect on DCV fusion. Alternatively, the fact that TI-VAMP2 restores SV fusion, but not DCV fusion after TeNT treatment, might also be explained by steric hindrance of DCVs selectively. The residues required for VAMP2 recognition by TeNT 39 were not affected by the mutations in the TI-VAMP2 construct. Therefore, TeNT still recognizes TI-VAMP2 and could affect the efficiency of fusion via steric hindrance upon binding TI-VAMP2. The effective window of TeNT proteolysis is before docking, before a SNARE complex is formed (reviewed by 19 ). Since DCVs are typically not predocked at the plasma membrane 26 , TeNT might constantly bind VAMP molecules and prevent fusion via steric hindrance. In line with the potential effects of steric hindrance, our SV fusion experiments show a slower, though not significant, rise in sypHy signal during extensive stimulation in TeNT treated neurons (Fig. 2), possibly reflecting a negative role of TeNT binding to TI-VAMP2 on newly recruited vesicles. Alternatively, these results could be explained by transglutaminase activation 40,41 . However, treating neurons with TeNT-E234Q, which is still able to bind VAMPs and activate transglutaminases 21,[40][41][42]55 , did not change the number of DCV fusion events (Fig. 7). These results exclude that transglutaminase activation or steric hindrance by TeNT binding alone can account for the lack of DCV fusion rescue by TeNT insensitive VAMP2.
In conclusion, the failure of TI-VAMP2 to restore DCV fusion after TeNT treatment indicates a novel mechanistic difference between the SV and DCV secretory pathways and can be used as a tool to selectively target DCV fusion leaving SV fusion unaffected.

Materials and methods
Biosafety. All experiments procedures were performed according to the local guidelines of the VU University/ VU University Medical Centre. For lentiviral work we followed all safety measures according to European legislation (ML-II, permit number: IG16-223-IIk).
DCV exocytosis events were detected manually as sudden appearance of NPY-pHluorin positive puncta using ImageJ. For experiments with infection of multiple constructs only neurons that expressed all indicated constructs (based on expression tags) were used for analysis. NPY-pHluorin events were considered a fusion event if the maximal fluorescence was at least twice the SD above noise. Custom written MATLAB scripts were used to calculate the number and timing of fusion events. Quantification of moving puncta and co-trafficking was done manually. Kymographs were generated with ImageJ of a neurite stretch positive for both NPY-mCherry and VAMP. Kymograph lines were analyzed for stationary (vertical lines) and moving puncta (diagonal lines). Co-trafficking was determined by overlapping lines of moving puncta. SV exocytosis was measured with ImageJ in manually placed regions where NH 4 + increased fluorescence, where the Fstim max was the maximal response during stimulation relative to the maximal response during NH 4 + superfusion.

Statistics.
Normal distributions for all datasets were assessed first using Shapiro-Wilk normality tests. To test more than 2 groups, we used one-way analysis of variance (ANOVA) followed by a post-hoc Tukey test to compare conditions when data was normally distributed or the Kruskal-Wallis test for non-parametric data followed by Dunn's multiple comparisons test to compare conditions. To compare two groups, we used an unpaired Student's t-test in the case of normal distributed data (only for the total number of NPY-pHluorin labeled DCVs and the number of DCV fusion events between control and vamp1 -/neurons, Fig. 6) or Mann-Whitney U tests for non-parametric data (all other cases). Data is represented as average with standard error of the mean (SEM). Dots in bar graphs indicate individual data points of single neurons. All data and statistical tests used are summarized in detail in Supplementary Table S1.