A Role for the V0 Sector of the V-ATPase in Neuroexocytosis: Exogenous V0d Blocks Complexin and SNARE Interactions with V0c

V-ATPase is an important factor in synaptic vesicle acidification and is implicated in synaptic transmission. Rotation in the extra-membranous V1 sector drives proton transfer through the membrane-embedded multi-subunit V0 sector of the V-ATPase. Intra-vesicular protons are then used to drive neurotransmitter uptake by synaptic vesicles. V0a and V0c, two membrane subunits of the V0 sector, have been shown to interact with SNARE proteins, and their photo-inactivation rapidly impairs synaptic transmission. V0d, a soluble subunit of the V0 sector strongly interacts with its membrane-embedded subunits and is crucial for the canonic proton transfer activity of the V-ATPase. Our investigations show that the loop 1.2 of V0c interacts with complexin, a major partner of the SNARE machinery and that V0d1 binding to V0c inhibits this interaction, as well as V0c association with SNARE complex. The injection of recombinant V0d1 in rat superior cervical ganglion neurons rapidly reduced neurotransmission. In chromaffin cells, V0d1 overexpression and V0c silencing modified in a comparable manner several parameters of unitary exocytotic events. Our data suggest that V0c subunit promotes exocytosis via interactions with complexin and SNAREs and that this activity can be antagonized by exogenous V0d.


Introduction
V-ATPase is an important player in acidifying intracellular compartments in eukaryotes. The molecular integrity of this enzyme guarantees energy-dependent proton transfer into specific compartments. Many intracellular organelles such as endosomes, trans-Golgi, secretory granules, synaptic vesicles and lysosomes are all critically dependent, for their function, on an acidic pH generated by the V-ATPase [1]. This pH sensor [2] and mechanochemical energy transducer [3] is composed of two reversibly attached intertwined V1 and V0 domains, each having a complex subunit composition [4,5]. The structural organization of this enzyme is intricate, and the reversible association of V0 and V1 regulates the coupling of ATPase activity to proton transport and consequent acidification of membrane compartments [6,7]. Synaptic vesicle acidification is functionally associated with vesicular neurotransmitter uptake and it has been reported that the extra-membranous V1 dissociates from fully loaded vesicles synaptic vesicles [7]. The V1 domain performs the ATPase activity while proton transport takes place through the membrane-embedded V0 domain.
The latter is composed of a tight assembly of several subunits (V0a, V0d, V0e, Ac45 and ATP6AP2), all surrounding a very hydrophobic rotor composed of several copies of V0c and a single V0c" subunit. V0e, Ac45 and ATP6AP2 are single-pass transmembrane proteins, V0c and c" are tetraspans with two cytosolic loops, and V0a possesses 8 transmembrane domains (TMD) and a prominent cytosolic N-terminus. Although the boxing-glove-shaped V0d [8] has no TMD [5,9], it is a stable component of the V0 domain [10,11] and is required for assembly of V-ATPase complex [12]. Yeast V0d participates in the so called "stator" [13] and has been shown to interact with the large N-terminal domain of V0a [14][15][16] following V1 dissociation, which results in an auto-inhibited form of V0, that is no longer capable of proton translocation. In humans, there are two isoforms of V0d: a ubiquitous V0d1 and a second isoform, V0d2, selectively expressed in osteoclasts, lung, kidney and epididymis [17]. Both isoforms interact with the central stalk V1 subunits V1D and V1F as well as with the V0c rotor [5,9,18]. V0d1 knockout in the mouse is embryonically lethal, highlighting its important physiological role [19]. Consequently, it has been proposed that the V0d subunit could couple the cytosolic part of the enzyme to its membrane-embedded component and therefore potentially couple ATP hydrolysis to proton transport [20].
Independently of its critical role in proton translocation, the V0 domain is implicated in exocytosis [21], neurotransmitter release [22,23] and also in numerous membrane fusion events in intracellular compartments [24][25][26][27]. However, genetic inactivation studies cannot be used to examine the role of the V0 domain in membrane fusion events since proton translocation is also needed for vesicle loading with neurotransmitters and hormones. At the molecular level, neurotransmitter release requires the formation of the soluble Nethylmaleimide-sensitive factor attachment protein receptors (SNARE) complex composed of two proteins of the presynaptic plasma membrane, syntaxin and SNAP-25, and a synaptic vesicle membrane protein, VAMP/synaptobrevin [28]. This minimal fusion machinery is regulated by other factors such as the small soluble presynaptic protein complexin, which that binds assembled SNARE complexes and may promote oligomerization of the SNARE complexes and fusion [29,30]. Several studies have shown that V0 interacts with and potentially regulates SNARE proteins [22,24,[31][32][33]. Inhibition of the V0c loop 3.4 interaction with VAMP2 was shown to significantly decrease neurotransmission and thus highlighted the importance of V0c interaction in modulating SNARE-dependent neurotransmission [33]. Chromophore-assisted light inactivation (CALI) later demonstrated that the photo-inactivation of the V0a subunit rapidly impaired neuronal synaptic transmission and catecholamine release from chromaffin cells [7]. In contrast, photo-inactivation of the V1 catalytic subunit A, very much like pharmacological inhibition of proton transport, induced a delayed inhibition of neurosecretion, strongly arguing that V0 regulates exocytosis independently from proton transport [7] without being directly involved in forming a membrane fusion pore [7,34]. More recently, we showed, using CALI experiments, that photo-inactivation of V0c in CA3 pyramidal neurons, rapidly inhibits neurotransmitter release downstream of synaptic vesicle acidification, thus corroborating the importance of the V0 domain subunits in directly modulating neurotransmission [35]. However, the exact function of the V-ATPase in regulating membrane fusion events remains a matter of debate.
In this study, we discovered that V0c interacts directly with complexin as well as with a mature assembly containing complexin and the trimeric SNARE complex. This interaction is mediated by the first cytosolic loop 1.2 of V0c, which also mediates the V0c interaction with V0d1. We show that V0c bound to V0d is no longer available to interact with complexin and the SNARE complex. Similarly to V0c silencing, V0d1 overexpression or its intraneuronal injection inhibited exocytosis. Altogether, these results bring new insight for the role of V-ATPase in exocytosis and reinforce the hypothesis that the V0 sector of the V-ATPase is an important modulator of SNARE-dependent neurosecretion.
Protein purification: Protease inhibitors were present in all homogenisation steps. Recombinant soluble trimeric SNARE complexes were obtained from 2l of induced culture. Purification was performed at room temperature when not otherwise stated. Pelleted bacteria (7 g) were resuspended in 25 mL of buffer H (sodium phosphate 50 mM, 0.5 M NaCl, 20 mM imidazole, pH 8) and lysed in a French press. The homogenate was centrifuged (200,000× g, 30 min, 4 • C), and the supernatant incubated for 1 h at 4 • C with 1.5 mL packed Ni-NTA Agarose beads (Qiagen). Beads were washed with buffer H adjusted to 50 mM imidazole. Purified proteins were eluted in 0.5 mL fractions by increasing imidazole concentration to 250 mM. Peak fractions were dialyzed against 20 mM Tris-HCl, 1.0 mM EDTA pH7.4 and loaded on a 1 mL HiTrap-Q column equilibrated in the same buffer on an AKTA-purifier system. Proteins were eluted with a 10-500 mM NaCl gradient in the dialysis buffer. The main peak fractions eluting at 375 mM NaCl were pooled. Proteins were quantified by Bradford assay and analysed on SDS-PAGE gel and Western blot. Aliquots were stored at −20 • C. Bacterial pellets from 6His-V0c expression, were resuspended in wash buffer (50 mM Tris-HCl, pH 8.0, 1 mM EDTA) supplemented with 0.2 mg/mL lysozyme and subjected to French press. Homogenates were centrifuged 5 min at 2500× g and the membrane fraction was isolated from supernatant by centrifuging at 200,000× g for 37 min. Membrane were washed once by resuspension of the pellet in wash buffer and centrifugation. Membranes (5 mg/mL protein) were solubilized in 50 mM Tris-HCl, pH 8.0, 10 mM β-mercapto-ethanol, 2% FosCholine-12 at 4 • C for 1 h. The buffer was then adjusted to 0.5 M NaCl, 20 mM imidazole, and the recombinant V0c was purified over Ni-NTA beads (QIAGEN).
Bacterial pellets expressing V0d1 constructs were resuspended in 50 mM Tris-HCl pH 8.0, 150 mM NaCl and subjected to French Press. Insoluble material was eliminated by centrifugation 37 min at 200,000× g. GST-V0d1 was purified by batch incubation of supernatant (1 h 4 • C) with 1 mL Glutathione-Sepharose. The suspension was loaded on a disposable 10 mL column. Washing (20 mM HEPES pH7.4, 150 mM NaCl, 0.1% Triton X-100 For purification of 6His-HA-V0d1, the high-speed supernatant was adjusted to 0.5 M NaCl and 20 mM imidazole and batch incubated with 2 mL Ni-NTA-Sepharose beads (1 hr, room temperature). The suspension was loaded on a disposable column, washed with 20 mM HEPES, 150 mM NaCl, in the presence of 20 mM and 40 mM imidazole successively and eluted by fractions in the presence of imidazole 0.5 M.
For both GST-and 6His-V0d1 preparations, protein-containing fractions were pooled, desalted against 10 mM HEPES pH 7.4 on Zeba-spin columns (Thermo scientific), aliquoted, fast frozen in liquid N2 and stored in aliquots at −80 • C. Protein concentrations were determined by parallel Coomassie staining and A280 absorbance. The very high hydrophobicity of V0c renders less accurate its protein assay by classical methods. Therefore, all 6His-HSV-V0c protein preparations were submitted to comparative relative quantification on Western blot using anti HSV antibodies and GST-HSV fusion protein as a standard.
Preparation of rat brain extracts: Rat brain fraction enriched in plasma membrane (LP1) was purified as described [43]. Briefly, proteins were solubilized at 1.5 mg/mL in 25 mM Tris pH 7.4, 150 mM NaCl containing 1.5% CHAPS, 5 mM EDTA and proteases inhibitors before centrifugation 1 h at 14,000× g and filtration through 0.22 µm filters.
V-ATPase mediated proton uptake: Proton uptake experiments were performed as previously described [33]. Briefly, rat brains were homogenized in 0.32 M sucrose, 10 mM HEPES, pH 7.4, and 0.2 mM EGTA in the presence of protease inhibitor cocktail, and the post-nuclear supernatant was layered onto a 0.8 M sucrose cushion and centrifuged 20 min at 257,000× g. Synaptosomal pellet was resuspended in hypotonic buffer (10 mM Tris-HCl pH 8.5, 1 mM PMSF) and centrifuged 20 min at 40,000× g. The synaptic vesicle-enriched supernatant was adjusted to 10 mM Tris-HCl pH 8.5, 60 mM sucrose, 140 mM KCl, 2 mM MgCl 2 , and 50 µM EGTA (assay buffer).
Synaptic vesicles were preincubated with 2 µM acridine orange (AO) before triggering proton uptake by the addition of 500 µM Mg-ATP. ATP-dependent proton transport was monitored, using a Biotek Sirius HT injector plate reader, by the quenching of AO fluorescence at 525 nm with data acquisition every 15 s. ELISA: HSV-V0c was probed with anti-HSV antibody (1/4000), GST-V0d1 and GST with goat anti-GST polyclonal antibody (1/4000). VAMP2, syntaxin and complexin were, respectively, probed with monoclonal antibodies 6F9 [33], 10H5 and SP33 at 1.0-1.2 µg/mL. Proteins were immobilized overnight at 4 • C on Maxisorb™ (Nunc) 96 wells microplates at 1 µg/well in 100 µL NaHCO 3 0.1 M pH 8.9. After a 1 h blocking step at 37 • C in 200 µL binding buffer (BB: 50 mM Tris-HCl pH 7.5, 150 mM NaCl, 3% BSA w/v), incubations with interacting proteins (100 µL) at the indicated concentrations were performed overnight at 4 • C in BB supplemented with 0.1% Triton X-100. For sequential binding, incubation with the first interactant was performed for 6 h at 4 • C, and then the solution was replaced by a fresh one containing both the first and the second interactant (V0c or V0d), and the plates were further incubated overnight at 4 • C. After interactions, plates were washed three times with 200 µL/well of washing buffer (WB: 50 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.1% BSA w/v, 0.1%Triton X-100) and incubated 1hr at 4 • C with 100 µL of specific primary antibodies in BB supplemented with 0.1% Triton X-100. Wells were washed (3 × 200 µL/well WB) and incubated for 1 hr at 4 • C with 100 µL horse radish peroxidase-coupled anti-IgG (either donkey anti-rabbit, goat anti-Mouse or rabbit anti-goat, depending on the primary antibody). After three washes, plates were revealed with TMB solution standard (Uptima, Interchim, Montluçon, France) according to supplier specifications. Colour development was stopped with 100 µL of 1.0 M H 2 SO 4 and absorbance read at 450 nm. Measurements were systematically performed in triplicates. Unless otherwise stated, blanks were always subtracted in the represented histograms. GST was used as a control binding partner when GST-V0d1 binding was measured. Signal in GST wells was either subtracted from GST-V0d1 signal or shown in parallel.
Surface plasmon resonance analysis: SPR experiments were performed at 25 • C using Biacore 3000 (GE healthcare) or Biacore T200 (Cytiva). The running buffer was HBS (10 mM HEPES pH 7.4, 150 mM NaCl) supplemented or not with CHAPS 0.2%. Nonspecific binding on control flow cells was automatically subtracted from experimental measurements to yield the specific signal. Proteins were coupled on sensor chips CM5 (Cytiva) or CMDP from Xantec (2D carboxylmethyldextran). About 25 fmoles of GST, Complexin1-GST or GST-V0d were covalently immobilized using amine coupling chemistry at pH 5. Analytes were injected at 10 µL/min, and blank run injections of running buffer were performed in the same condition permitting to yield double-subtracted sensorgrams using Biaevaluation 4.2 software (GE Healthcare).
SPR detection of native SNAREs/complexin on immobilized recombinant V0c: Solubilized proteins from LP1 were diluted 5-fold in running buffer (25 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.2% CHAPS) and injected over 2 flow cells (CMDP chip) in series including a control flow cell functionalized with irrelevant mouse antibodies and another one with anti HSV antibodies that previously captured HSV-V0c.
Superior cervical ganglion neurons and synaptic transmission recordings: 6-8 week cultures, EPSP recording and injection of recombinant His-V0d (6.5 µM) or BSA (20 µM) were performed as described previously [44]. EPSPs were recorded at 0.1 Hz. The peak amplitudes were normalized to the values before injection. The averaged and smoothed values (Origin 7.5) obtained from an eight-point moving average algorithm were plotted against recording time with t = 0 corresponding to the start of 3 min presynaptic injection of His -tagged V0d. Data are mean ± SEM and statistical significance was evaluated using a two-tailed Mann-Whitney U test.
Chromaffin cell culture and catecholamine release recordings: Freshly dissected primary bovine chromaffin cells were cultured in DMEM in the presence of 10% fetal calf serum, 10 µM cytosine arabinoside, 10 µM fluorodeoxyuridine and antibiotics as described previously [45]. Plasmids and siRNAs were introduced into chromaffin cells (5 × 10 6 cells) by Amaxa Nucleofactor systems (Lonza) according to manufacturer's instructions and as described previously [46]. At 48 h/96 h after transfection, catecholamine secretion was evoked by applying K + (100 mM) in Locke's solution without ascorbic acid for 10 s to single cells by mean of a glass micropipette positioned at a distance of 30-50 µm from the cell. Electrochemical measurements of catecholamine secretion were performed using 5 µm diameter carbon-fibre electrodes (ALA Scientific, Farmingdale, NY, USA) held at a potential of +650 mV compared with the reference electrode (Ag/AgCl) and approached closely to the transfected cells essentially as described previously [47]. Amperometric recordings were performed with an AMU130 (Radiometer Analytical, Villeurbanne, France) amplifier, sampled at 5 kHz, and digitally low-pass filtered at 1 kHz. Analysis of amperometric recordings was performed as described previously [48], allowing automatic spike detection and extraction of spike parameters. The number of amperometric spikes was counted as the total number of spikes with an amplitude >5 pA. Data were analysed using SigmaPlot 13 software. In the figure legends, n represents the number of cells analysed. Statistical significance has been assessed using t-test as data fulfilled requirements for parametric tests.
Catecholamine content measurement: Chromaffin cells were incubated for 30 min with 0.5 µM of Neurosensor 510 (NS510) at 37 • C and then washed to remove excess sensor before fixation. Labelled cells were visualized using a Leica SP5II confocal microscope and quantification of NS510 signal was performed in individual cells.

V0c loop 1.2 Binds Complexin
We have previously shown that the cytosolic loop (L3,4) linking transmembrane regions 3 and 4 of V0c binds VAMP2 and that this interaction modulates neurotransmission [33]. We therefore screened for other potential V0c binding partners. Recombinant V0c was immobilized on the chip of an SPR apparatus and a detergent extract of a lysed rat brain synaptosomal fraction was injected into the flow cell ( Figure S1A). Bound proteins were then detected by injecting monoclonal antibodies against proteins involved in exocytosis ( Figure S1B). In addition to the previously reported SNARE proteins (VAMP2, syntaxin 1 and SNAP25 [33,49]), complexin1 and 2 were specifically identified in association with V0c ( Figure S1B). As complexin could be indirectly associated with V0c due to its association with the SNARE complex [49], we used purified recombinant proteins to investigate whether a direct interaction occurred. Further investigations of complexin/V0c interactions were performed using complexin 1, hereafter referred to as complexin. When complexin was immobilized on ELISA plates, a strong binding of V0c (20 nM) was detected ( Figure 1A), while up to 1 µM of V0c did not generate significant binding to BSA-blocked ELISA wells in the absence of complexin ( Figure 1A and Figure S2). V0c binding to complexin was not inhibited in the presence of VAMP2 suggesting that VAMP2 and complexin interact with V0c through distinct domains ( Figure 1A). In order to identify the molecular determinants on V0c that are important for complexin binding, we probed the implication of both cytosolic V0c loops that link, respectively, V0c TMDs 1 and 2, and 3 and 4. For this purpose, we perturbed the integrity of V0c loops 1.2 (L 1.2s ) or 3.4 (L 3.4s ) in the full-length protein by scrambling their linear sequences in the full-length protein, and we monitored interactions using an SPR-based method. V0c binding to immobilized complexin dissociates very slowly, indicating a strong interaction ( Figure 1B). Binding of V0c containing a scrambled loop 3.4 to immobilized complexin was nearly identical to that of wild-type V0c ( Figure 1B). However, V0c with scrambled loop 1.2 failed to bind complexin ( Figure 1B). The implication of V0c loop1.2 in complexin binding was confirmed by the experiment showing that a peptide corresponding to loop 1.2 totally inhibited V0c binding to immobilized complexin, while loop 3.4 peptide had no effect ( Figure 1C).
In accordance with the potentially late implication of V0c in exocytosis [33,35], we investigated whether V0c interacts with the SNARE complex in the presence of complexin. The three SNARE proteins (devoid of any TM domains) were co-expressed in bacteria and SDS-resistant purified SNARE complexes were obtained ( Figure S3). SNARE complexes were immobilized on ELISA plates, and binding of V0c was assayed with or without preincubation with complexin. As shown in Figure 1D, V0c interacts with the SNARE complex in the presence of complexin. In summary, V0c interacts with complexin through the cytosolic loop 1.2 and the presence of complexin does not inhibit V0c binding to VAMP2 or to the SNARE complex.

V0d Interaction with V0c Loop 1.2 Precludes V0c Association with Complexin and the SNARE Complex
Although V0d1 is a soluble protein, it remains stably associated with the V0 sector after dissociation of the V1 sector [14,15,50]. In order to assess the implication of V0d1 in the interaction of V0c with other molecular partners, we started by characterizing recombinant V0d1 binding to a series of immobilized proteins. As shown in Figure 2A, ELISA experiments revealed that GST−V0d1 bound to immobilized V0c, but did not interact with VAMP2, the dimeric syntaxin/SNAP−25 t−SNARE complex, the trimeric SNARE complex, complexin or complexin associated with the SNARE complex. In accordance with the potentially late implication of V0c in exocytosis [33,35], we investigated whether V0c interacts with the SNARE complex in the presence of complexin. The three SNARE proteins (devoid of any TM domains) were co-expressed in bacteria and SDS-resistant purified SNARE complexes were obtained ( Figure S3). SNARE complexes were immobilized on ELISA plates, and binding of V0c was assayed with or without preincubation with complexin. As shown in Figure1D, V0c interacts with the SNARE complex in the presence of complexin. In summary, V0c interacts with complexin through the interaction of V0c with other molecular partners, we started by charac recombinant V0d1 binding to a series of immobilized proteins. As shown in Fig  ELISA experiments revealed that GST−V0d1 bound to immobilized V0c, but interact with VAMP2, the dimeric syntaxin/SNAP−25 t−SNARE complex, the SNARE complex, complexin or complexin associated with the SNARE complex. We then used SPR to identify the domain of V0c that interacts with V0d. As in Figure 2B, purified recombinant V0c bound stably to immobilized V0d1 and no apparent dissociation, in accordance with previous data showing that V0d st partitions with V0 in detergent-resistant membranes [11]. The specificity interaction was corroborated by the absence of V0c binding when it had be incubated with an excess of V0d1 ( Figure 2B). Binding of V0c that has a scramble We then used SPR to identify the domain of V0c that interacts with V0d. As shown in Figure 2B, purified recombinant V0c bound stably to immobilized V0d1 and showed no apparent dissociation, in accordance with previous data showing that V0d stably copartitions with V0 in detergent-resistant membranes [11]. The specificity of this interaction was corroborated by the absence of V0c binding when it had been pre-incubated with an excess of V0d1 ( Figure 2B). Binding of V0c that has a scrambled linear sequence of loop 3.4 to immobilized V0d1 was nearly undistinguishable from wild-type V0c; however, V0c with a scrambled linear sequence of loop 1.2 failed to bind V0d1 ( Figure 2B). The involvement of V0c loop 1.2 in V0d binding was supported by the observation that the addition of free L 1.2 peptide totally inhibited V0c binding to immobilized V0d1, while free L 3.4 peptide addition had no effect ( Figure S4). The involvement of V0c loop 1.2 in interaction with both V0d1 and complexin was confirmed by the observation that co-injecting a molar excess of complexin inhibited V0c binding to immobilized V0d1 ( Figure 2C).
We then assessed if V0d could modulate the V0c interaction with VAMP2 and the SNARE/complexin complex, using ELISA competition experiments. As shown in Figure 2D, the presence of GST-V0d1 prevents V0c from binding to VAMP2 (inhibited by 71.3% ± 3.2) and to in situ reconstituted SNARE complex (inhibited by 100%). The blocking effect of V0d is also observed in the presence of complexin for V0c binding to VAMP2 (inhibited by 90.7% ± 1) and to the SNARE complex (inhibited by 96.1% ± 6.2). This interference probably results from V0d interaction with V0c loop 1.2 inducing a steric hindrance of VAMP2 access to loop 3.4.

Presynaptic V0d1 Injection Inhibits Neurotransmission in SCG Neurons
The importance of V0 subunits in the proton pump activity of the V-ATPase, and therefore in cell viability, compromises interpretation of data from gene inactivation and mRNA silencing to address the direct implication of these subunits in exocytosis. A multimer of V0c subunits constitutes the V-ATPase rotor, associated with a single V0d subunit. Although it does not have a membrane anchor, V0d behaves like an extrinsic membrane protein, and V0d1 is not detected in cytosolic fractions ( Figure S5). Based on our biochemical results indicating that V0d inhibits V0c interactions with VAMP2, SNARE complex or complexin, we reasoned that introduction of soluble V0d into the cell would stably occupy free binding sites on the multimeric V0c rotor, impeding association with exocytotic proteins. Superior cervical ganglion neurons (SCG) in culture have been widely used to explore the dynamics of presynaptic protein-protein interactions and understand their implication in neurotransmitter release [51]. As VAMP2 binding to V0c is involved in modulation of neurotransmission [22] and in order to test the hypothesis that V0d1 may regulate V0c interactions with SNARE proteins, we injected purified recombinant His-tagged V0d1 protein (6.5 µM) into presynaptic SCG neurons and monitored neurotransmitter release through postsynaptic recordings. As shown in Figure 3A (left), neurotransmission started to decrease only a few minutes after injection and reached a plateau shortly after 20 min, with a mean inhibition ratio of 28% ± 3.9 p < 0.01 at 9 min after injection and a maximal inhibition of 35% after 35 min. Of note, the control BSA injections did not lead to any significant decrease in neurotransmission ( Figure 3A, right). As any perturbation of the V-ATPase function could drastically impact enzymatic activity and neurotransmitter uptake into synaptic vesicles, we verified whether the presence of excess V0d could impact proton pumping and, consequently, if neurotransmission decrease does not result from a defect in acidification of the vesicle lumen, which may lead to reduced neurotransmitter loading. We therefore monitored synaptic vesicle V-ATPase activity in purified synaptic vesicles in the presence or absence of an excess of V0d1, by monitoring the intra-vesicular accumulation of fluorescent protonated Acridine Orange. As shown in Figure 3b, the addition of up to 9 µM of V0d did not inhibit vesicular acidification, in contrast to application of the proton pump inhibitor bafilomycin A1, which resulted in total inhibition of proton pump activity, and completely prevented Acridine Orange (AO) accumulation in synaptic vesicles ( Figure 3B). This clearly demonstrates that excess V0d does not perturb vesicular acidification. Hence, a presynaptic V0d overload, which prevents V0c from binding to complexin and the SNARE complex, rapidly diminished acetylcholine release by a mechanism independent of the proton pump activity of the V-ATPase.

V0d Overexpression Perturbs Catecholamine Release in Chromaffin Cells
To investigate the effects of V0d1 on catecholamine exocytosis, we overexpressed myc-tagged V0d1 from a bicistronic mRNA co-expressing EGFP in chromaffin cells. Using carbon-fibre amperometry [52,53], we monitored catecholamine release triggered by KCl depolarization and analysed several exocytosis parameters (Figure 4). Catecholamine release from EGFP-expressing cells was not different in any of the measured parameters compared to un-transfected cells (data not shown). Among all the parameters analysed, four were significantly decreased upon V0d overexpression: spike number (57 ± 4% of control), the total amount of currents during the spikes Imax (# 74 ± 2 % of control), the

V0d Overexpression Perturbs Catecholamine Release in Chromaffin Cells
To investigate the effects of V0d1 on catecholamine exocytosis, we overexpressed myc-tagged V0d1 from a bicistronic mRNA co-expressing EGFP in chromaffin cells. Using carbon-fibre amperometry [52,53], we monitored catecholamine release triggered by KCl depolarization and analysed several exocytosis parameters (Figure 4). Catecholamine release from EGFP-expressing cells was not different in any of the measured parameters compared to un-transfected cells (data not shown). Among all the parameters analysed, four were significantly decreased upon V0d overexpression: spike number (57 ± 4% of control), the total amount of currents during the spikes I max (# 74 ± 2% of control), the quantal size Q (70 ± 4% of control) as well as the pre-spike foot (PSF) charge Q (# 73 ± 6% of control). The kinetic parameters of individual exocytosis events T Half and the time to peak were not significantly affected. The percentage of spikes with a detectable foot (about 25% in the two conditions) was not modified, and no significant variation in the pre-spike or foot parameters were observed. Altogether, these results showed that V0d overexpression dramatically reduced the number of exocytotic events and reduced the amount of catecholamine released in the remaining events. of control). The kinetic parameters of individual exocytosis events T Half and the time to peak were not significantly affected. The percentage of spikes with a detectable foot (about 25% in the two conditions) was not modified, and no significant variation in the pre-spike or foot parameters were observed. Altogether, these results showed that V0d overexpression dramatically reduced the number of exocytotic events and reduced the amount of catecholamine released in the remaining events.

V0c Silencing Reduces Catecholamine Release in Chromaffin Cells
Our biochemical analysis suggests that the effect of V0d1 is exerted through a reduction in the number of V0c interaction sites available for complexin/SNARE complexes. In order to corroborate this interpretation, we monitored catecholamine release from chromaffin cells and analysed the parameters of exocytosis after siRNA silencing of V0c ( Figure 5A,B). V0c silencing leading to a reduction in V0c levels by 40-

V0c Silencing Reduces Catecholamine Release in Chromaffin Cells
Our biochemical analysis suggests that the effect of V0d1 is exerted through a reduction in the number of V0c interaction sites available for complexin/SNARE complexes. In order to corroborate this interpretation, we monitored catecholamine release from chromaffin cells and analysed the parameters of exocytosis after siRNA silencing of V0c ( Figure 5A,B). V0c silencing leading to a reduction in V0c levels by 40-50% without affecting V0a1 and V1A expression levels (data not shown) resulted in an approximately 40% decrease in the quantal content ( Figure 5), presumably due to a perturbation of catecholamine loading into secretory granules. Interestingly, very similar to the effects of V0d1 overexpression, V0c silencing, using two distinct siRNA, significantly decreased spike number/cell (# 50% of control) and I max (# 70% of control) ( Figure 5). These results suggest that V0c availability is implicated in determining the number of fusion events. In addition, and in contrast to V0d1 overexpression, V0c silencing led to a significant decrease in several additional PSF parameters: the number of PSF per cell (# 40% of control), the amount of current per PSF (# 80% of control) and the quantal size Q of the foot (# 60% of control). In order to ascertain the specificity of the observed V0c siRNAs effects, we performed rescue experiments and we reintroduced wild-type V0c insensitive to the siRNA. As shown in Figure 5, rescue experiments restore wild-type recording parameters demonstrating that the observed changes using the siRNA are due to a specific decrease in V0c.
To assess more specifically the effect of V0d1 overexpression and V0c silencing on catecholamine content in secretory granules, we used the selective noradrenalin and dopamine fluorescent indicator NS510, which specifically accumulates in chromaffin secretory granules [54]. As illustrated in Figure S6, overexpression of V0d1 and silencing of V0c reduced by 16% and 30%, respectively, the mean intensity of the NS510 staining. These observations suggest that long term overexpression of V0d1, as well as a reduction in the level of V0c expression, moderately reduced catecholamine content in secretory granules that might result from a reduction in secretory granule acidification. It is, however, unlikely that this moderate effect on catecholamine loading could explain the potent reduction in the number of individual exocytotic events, arguing for the importance of the V0d/V0c interaction in the modulation of exocytosis.  To assess more specifically the effect of V0d1 overexpression and V0c silencing catecholamine content in secretory granules, we used the selective noradrenalin dopamine fluorescent indicator NS510, which specifically accumulates in chroma secretory granules [54]. As illustrated in Figure S6, overexpression of V0d1 and silenc of V0c reduced by 16% and 30%, respectively, the mean intensity of the NS510 stain These observations suggest that long term overexpression of V0d1, as well as a reduc

Discussion
It has been reported that the V1 domain of the V-ATPase is absent from fully loaded as well as docked synaptic vesicles [6,7] and that conformational rearrangements in V0a and V0d subunits take place upon V1 dissociation leading to a so-called auto-inhibited V0 domain [5,[14][15][16]. Several reports already showed that independently of its participation in the proton pumping activity of the V-ATPase, the V0 domain is implicated in exocytosis [21][22][23] as well as in membrane fusion of intracellular compartments [24][25][26][27]. In addition, the direct interactions of the V0 domain subunits with the minimal membrane fusion machinery, i.e., the SNARE complex or individual SNARE proteins, have been reported with functional consequences [22,31,33,55]. Complexin is a major regulatory partner of the SNARE complex [49]. In vertebrates, among the four isoforms, complexins 1 and 2 are mainly neuronal, and complexin 3 and 4 are retina specific [56]. Despite its small size, complexin has a very intricate function with distinct adjacent and functionally different domains. Each of these domains can separately exert opposing properties as both an inhibitor or a facilitator of synaptic vesicle fusion modulating evoked and spontaneous release [29,49,[57][58][59]. It interacts through its central α-helix domain with the SNARE complex [29,60,61] and is implicated in organizing high-order fusogenic SNARE/synaptotagmin complexes [29,62]. Structural analysis has shown that complexin is associated with the SNARE complex in an anti-parallel orientation, leaving the C-terminal domain oriented towards the N-terminal domains of syntaxin and VAMP in the assembled SNARE complex [60,63]. Additionally, the interaction of the C-terminal domains of complexins 1 and 2 with lipids [64] and with the highly curved membranes of synaptic vesicles was suggested to be important for its inhibitory function [65][66][67].
In this study, we report a direct interaction of the V0c subunit of the V-ATPase with complexin as well as the tetrameric complexin/SNARE complex. We show that the interaction of complexin with V0c is mediated by the loop 1.2. As the loop 3.4 of V0c binds VAMP2 and V0c binding to complexin is not competed by VAMP2, one may assume that V0c can interact concomitantly with two essential components of the exocytotic machinery. Moreover, we demonstrate that the loop 1.2 of V0c mediates an extremely stable interaction with V0d1 in line with the existence of contact sites between V0d and V0c rotor in yeast [15] and mammalian [18] V-ATPase. Our biochemical experiments indicate that an excess of V0d1 hinders V0c interaction with VAMP2, complexin and the complexin/SNARE complex. Since V0c interacts with complexin and VAMP2 via two distinct loops (loop 1.2 and loop 3.4, respectively) and complexin does not inhibit V0c/VAMP2 interaction, the inhibitory effect of V0d on V0c binding to VAMP2 may result from a steric hindrance due to the higher molecular weight of V0d compared to complexin. In order to gain an insight into the physiological implication of V0c loop 1.2 in exocytosis, we reasoned that introducing soluble V0d would impede V0c loop 1.2 interaction with complexin and the SNARE complex. In a first approach, we monitored excitatory postsynaptic potentials (EPSPs) in connected superior cervical ganglion (SCG) neurons after presynaptic injection of bacterially expressed and purified V0d1. SCG neurons are a well-established culture system suited for the study of neurotransmitter release mechanisms. The very short axonal connections provide a setup where somatically injected effectors rapidly reach nerve terminals. Upon V0d1 microinjection in SCG neurons, we observed rapid inhibition of neurotransmission similar to that observed upon injection of a peptide corresponding to V0c L3.4 loop [33] and other agents that affect late steps in exocytosis [68]. As V0d is a functional component of the V-ATPase complex, we addressed the possibility that introducing free V0d could modify intramolecular interactions between V-ATPase subunits and consequently inhibit proton pumping, which would then compromise neurotransmitter loading into synaptic vesicles. In this case, vesicles devoid of neurotransmitter might still fuse without generating an EPSP. To explore this possibility, we studied the effect of an excess of V0d1 on synaptic vesicle acidification using an Acridine Orange-based assay. Although the V-ATPase inhibitor bafilomycin produced an immediate and complete block of proton accumulation in synaptic vesicles, V0d1 concentrations higher than those injected into SCG neurons failed to decrease acidification. These data strongly suggest that an excess of cytosolic V0d1 is unlikely to perturb synaptic vesicle loading with neurotransmitters. Similarly to inhibition of acetylcholine release in SCG neurons, catecholamine secretion was altered after V0d1 overexpression in chromaffin cells, a well-defined cellular model to study neuroendocrine secretion [69]. In addition to a significant decrease in the number of spikes ( Figure 4A,C) and maximal current reduction ( Figure 4D), the quantal content of granules was affected upon V0d1 overexpression ( Figure 4D) without affecting time to peak kinetics. Apart from the pre-spike foot (PSF) charge which was decreased, V0d1 overexpression did not affect any other PSF parameter ( Figure 4E). It is, however, of note that under these conditions of V0d1 overexpression, a reduction in catecholamine content of approximatively 16% was observed using the fluorescent indicator NS510 ( Figure S6). The observed decrease in the quantal content by roughly 30% may thus partly be due to the long-term genetic nature of V0d1 overexpression. It is possible that co-translational abundance of V0d1 with endogenous native levels of other V-ATPase subunits may perturb V-ATPase assembly and therefore granule acidification and catecholamine loading. On the other hand, the modest effect observed on catecholamine loading under these conditions is unlikely to explain the robust differences in spike number, Imax and charge observed when V0d1 is overexpressed, arguing for a role of the V0d1-V0c interaction in the regulation of exocytosis.
In an earlier report, we have shown that inhibiting V0c interaction with VAMP2 significantly inhibits neurotransmission [33]. Previous data showed that the native multimeric V0c rotor binds to a single V0d1 subunit [1,5,18]. Our electrophysiological recordings in SCG neurons as well as the biochemical results suggest that inhibition of exocytosis upon V0d1 injection is likely due to binding of the exogenous V0d1 to V0c subunits in the V0 rotor that are not in contact with the single intrinsic V0d1 subunit. Consequently, this could prevent V0c interactions with complexin and SNARE proteins. In order to corroborate this hypothesis in chromaffin cells and compare the effect of V0d1 overexpression and V0c availability on catecholamine secretion, we performed siRNA V0c downregulation ( Figure 5A,B) and monitored catecholamine secretion parameters. As expected, the granule quantal content was significantly decreased, but catecholamine secretion, although diminished, still took place ( Figure 5C). Both V0c siRNAs triggered a decrease in V0c expression and led to an inhibition of the number of individual release events as well as maximal current that reached very similar levels of inhibition that were observed upon V0d1 overexpression. In contrast to the effects of V0d1 overexpression, a reduced V0c expression level affected the number of PSF per cell, the spike shape (Imax and half width), the footImax highlighting the importance of V0c availability in modulating release. Previous structural studies of the yeast V-ATPase showed that, in the auto-inhibited V0 domain, V0d interacts with the N-terminal domain of V0a upon V1 dissociation and engages interactions with the V0c rotamer subunits [5,[14][15][16]. In bovine V-ATPase, V0d shows stable interactions with 4 c-subunits and 1 c" in the assembled V-ATPase [18]. However, how the exogenous V0d incorporates into auto-inhibited V0 is still to be determined at the structural level.
All these data reinforce our understanding of the implication of the V-ATPase V0 subunits in modulating SNARE-dependent neurotransmission and corroborate the importance of V0c in SNARE-dependent neurosecretion. In a fully assembled V-ATPase, a V0c multimer accommodates only one V0d subunit [1,5]. Moreover, we show that the V0d subunit is absent from cytosol. We speculate that, upon V1 dissociation, V0c subunits that are not linked to V0d1 would be free to bind VAMP2 as well as the cytosolic complexin [70]. The V0c subunits of the V0 sector associated with several VAMP2 molecules would interact with membrane t-SNAREs and engage the formation of a SNARE rosette around the V0c rotor [71]. Upon overexpression of V0d1, V0d1-free V0c subunits in a V0 rotor may become sequestered and hindered from binding to VAMP2 [33] and complexin, resulting in an inhibition of formation of the SNARE rosette and thereby neurotransmitter release. Using molecular dynamics [72], it has recently been suggested that confinement of a rosette of SNARE complexes [71,73] is essential for rapid fusion pore formation, and fusion pore expansion is accompanied by a release from this confinement. A potential consequence of the presence of an excess of V0d1 would be that V0c would fail to form the SNARE complex rosette (Figure 6), leading to a decrease in the frequency of membrane fusion events. Whether the single endogenous V0d1 subunit per V-ATPase is implicated in modulating exocytosis has not yet been addressed, and future investigations are needed to clarify this point.
fusion pore formation, and fusion pore expansion is accompanied by a release from this confinement. A potential consequence of the presence of an excess of V0d1 would be that V0c would fail to form the SNARE complex rosette (Figure 6), leading to a decrease in the frequency of membrane fusion events. Whether the single endogenous V0d1 subunit per V-ATPase is implicated in modulating exocytosis has not yet been addressed, and future investigations are needed to clarify this point. Supplementary Materials: The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/cells12050750/s1, Figure S1. V0c captures SNARE proteins and complexin from rat brain extract. V0c captures SNARE proteins and complexin from rat brain extract. (A) Example of specific immobilization of solubilized rat brain proteins on V0c using SPR: recombinant HSV-tagged V0c was captured (1000 RU) on a sensor chip using anti-HSV antibodies. LP1 rat brain extract was injected three times (arrows) at 1 µL/min into the flow cell. This injection yields about 600 RU of stable and specific binding compared to a control flow cell loaded with only anti-HSV. Background in the absence of V0c has been subtracted from the presented data. (B) Specific signal resulting from injection of non-immune (NIS), anti-syntaxin1 (10H5), anti-SNAP25 (Br05), anti-complexin (LP27, SP13) and anti-VAMP2 (6F9) antibodies over brain extract proteins Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/cells12050750/s1, Figure S1. V0c captures SNARE proteins and complexin from rat brain extract. V0c captures SNARE proteins and complexin from rat brain extract. (A) Example of specific immobilization of solubilized rat brain proteins on V0c using SPR: recombinant HSV-tagged V0c was captured (1000 RU) on a sensor chip using anti-HSV antibodies. LP1 rat brain extract was injected three times (arrows) at 1 µL/min into the flow cell. This injection yields about 600 RU of stable and specific binding compared to a control flow cell loaded with only anti-HSV. Background in the absence of V0c has been subtracted from the presented data. (B) Specific signal resulting from injection of non-immune (NIS), anti-syntaxin1 (10H5), anti-SNAP25 (Br05), anti-complexin (LP27, SP13) and anti-VAMP2 (6F9) antibodies over brain extract proteins captured on immobilized V0c as illustrated in A. Detection levels (RU) are represented in histograms. Open and black bars are two independent experiments.; Figure S2. Background levels of V0c binding in ELISA plates. V0c binding to BSA-saturated ELISA wells was probed using increasing concentrations of V0c (10-1000 nM). Binding was detected using anti-HSV antibodies. Note that binding signals were equivalent to a condition in which V0c was not added. Results are means of triplicate determination ± SD. Figure S3. Quality control of SNARE complex formation. An amount of 4 µg of the purified SNARE complex was either heated or not at 95 • C in sample buffer and analysed by SDS PAGE and Coomassie blue staining (left). The molecular identity of the components of this complex was verified by Western blot using anti-VAMP2 (6F9), anti-syntaxin1(10H5) and anti-SNAP-25 (6C11) antibodies (right). Note that the SNAP-25 epitope recognized by 6C11 is mostly hidden in the assembled SNARE complex. Positions of the SNARE monomers and the SNARE complex are indicated, respectively, by (*) and (**). Figure S4. Characterisation of V0c/V0d1 interaction. An amount of 50 nM of recombinant wild-type V0c was injected over consecutive flow cells of an SPR sensorchip pre-functionalised with GST or GST-V0d1 in the presence of an excess of either V0c loop 1.2 (L1.2) or loop 3.4 peptides (L3.4). Binding signals on precoated GST flow cells have been subtracted from represented sensorgrams. Figure S5. Distribution of V0d1 in different membrane fractions. The distribution of V0d1 was analysed by Western blot in different membrane fractions of mice brain lysates (30 µg/well). (* = actin; ** = V0d1). Note the absence of V0d1 from the cytosolic fraction. Figure S6. Effect of V0d overexpression and V0c silencing on catecholamine loading in secretory granules of bovine chromaffin cells. Chromaffin cells overexpressing V0d and TurboRFP (A) or silenced for V0c (B) were incubated with NS510. Quantification of accumulated NS510 signal was performed from individual cells (n = 53) from three independent experiments for each condition.