A novel and specific regulator of neuronal V-ATPase in Drosophila

The V-ATPase is a highly conserved enzymatic complex that ensures appropriate levels of organelle acidification in virtually all eukaryotic cells. While the general mechanisms of this proton pump have been well studied, little is known about the specific regulations of neuronal V-ATPase. Here, we studied CG31030, a previously uncharacterized Drosophila protein predicted from its sequence homology to be part of the V-ATPase family. We found that this protein is essential and apparently specifically expressed in neurons, where it is addressed to synaptic terminals. We observed that CG31030 co-immunoprecipitated with V-ATPase subunits, in particular with ATP6AP2, and that synaptic vesicles of larval motoneurons were not properly acidified in CG31030 knockdown context. This defect was associated with a decrease in quantal size at the neuromuscular junction, severe locomotor impairments and shortened lifespan. Overall, our data provide evidence that CG31030 is a specific regulator of neuronal V-ATPase that is required for synaptic vesicle acidification and neurotransmitter release.

The CG31030 protein is predicted to be part of the InterPro V-ATPase family, but no 175 experimental information is currently available regarding its potential interactors. To determine 176 whether CG31030 could interact, directly or indirectly, with subunits of the V-ATPase complex, 177 we carried out co-immunoprecipitation experiments using an anti-V5 antibody on proteins 178 extracted from heads of CG31030 V5 mutants and w 1118 control flies, followed by nano LC-MS/MS 179 mass-spectrometry analysis of the precipitated proteins. Three independent experiments were 180 performed to increase reliability, and in total, 410 proteins were identified in all three 181 experiments. Among those, only 12 proteins had at least a two-fold abundance difference with 182 the control in all three experiments (Figure 2A and Supplementary Table 2), one of them being, 183 as expected, the co-immunoprecipitation target CG31030. Remarkably, three of the 11 other proteins were identified as subunits of the V-ATPase complex: Vha100-1, VhaAC39-1 and 185 ATP6AP2 ( Figure 2B). 186 Vha100-1 and VhaAC39-1 code for subunits a and d of the V0 domain, respectively 187 (Vasanthakumar and Rubinstein, 2020). The subunit a, coded by five different genes in 188 Drosophila, is the proton port of the pump (Collins and Forgac, 2020). Among the five isoforms, 189 Vha100-1, which co-immunoprecipitated with CG31030 in our experiments, has been shown to 190 be specifically required in neurons and present at the synapse (Hiesenger et al., 2005). Subunit d 191 of V0 is coded by two Drosophila genes: VhaAC39-1 and VhaAC39-2, and only the first co-192 immunoprecipitated with CG31030. According to FlyAtlas, VhaAC39-1 is expressed in many 193 tissues and enriched in the brain, while VhaAC39-2 seems to be mostly found in testis and 194 salivary glands. For both V0 subunits, CG31030 thus co-precipitated with the likely neuronal 195 isoform. The co-immunoprecipitated V-ATPase subunit which appeared to be the most enriched 196 in the CG31030 V5 sample was interestingly the accessory subunit ATP6AP2, suggesting a possible 197 direct interaction between this protein and CG31030 (Supplementary Table 2). These 198 experiments therefore reinforce the hypothesis that CG31030 directly interacts with the 199 neuronal V-ATPase complex, and more specifically with V0 since all detected partners belong, or 200 interact, with this domain. 201  Because CG31030 appeared to be mainly localized in synaptic areas (see Figure 1), we chose to 218 look at the physiological effect of its disruption at the Drosophila larval neuromuscular junction, 219 a model that has contributed to the study of many essential synaptic processes. At synaptic 220 nerve endings, a prominent role of the V-ATPase is to acidify the lumen of synaptic vesicles, the 221 electrochemical gradient generated providing the driving force to load and concentrate the 222 neurotransmitters. Thus, a malfunction of synaptic V-ATPase should induce a decrease of 223 neurotransmitter concentration inside the vesicles, potentially resulting in an altered synaptic 224 transmission. To test this hypothesis, we co-expressed each of the two strongest CG31030 RNAi 225 constructs together with VMAT-pHluorin, a pH-sensitive probe targeted to synaptic vesicles (Wu 226 et al., 2013), in larval motoneurons using the glutamatergic driver OK371-Gal4. Both RNAi1 and 227 RNAi2 induced a lethal phenotype at pupal stage in these conditions. VMAT-pHluorin is an 228 ecliptic pHluorin that is fluorescent at neutral pH, and gets quenched, by protonation, at acidic 229 pH (Miesenböck et al., 1998). Thus, in control condition, VMAT-pHluorin should not be 230 fluorescent in synaptic vesicles, whose pH is at around 5.5, but only when externalized on the 231 presynaptic membrane during exocytosis, and so, in contact with the more neutral synaptic cleft 232 milieu. In the case of an acidification defect of synaptic vesicles, the probe could be fluorescent 233 both in synaptic vesicles, where pH would be abnormally high, and on the presynaptic 234 membrane ( Figure 3A, central panel). 235 In order to evaluate the ratio of the internal fluorescence (from synaptic vesicles) over the 236 external fluorescence (from the presynaptic membrane) at the neuromuscular junction of panel). This operation resulted in only the internal signal being conserved. In controls, this 240 meant that all signal was abolished, as expected because the synaptic vesicles were normally 241 acidified ( Figure 3B, left panel). In contrast and strikingly, a residual signal was still visible in this 242 acidic milieu in both RNAi1 and RNAi2 knockdown larvae ( Figure 3B, right panel). Quantification 243 of the ratio of fluorescence area in acidic milieu over neutral milieu showed that about 37% of 244 the total signal remained visible in the RNAi larvae after external quenching ( Figure 3C and D). 245 To verify that the residual signal seen in knockdown larvae was indeed coming from inside 246 vesicles, the opposite strategy was used: instead of quenching the outside signal, we revealed 247 all the internal one by collapsing the pH gradient of synaptic vesicles ( Figure 3A, right panel). To 248 do so, we replaced the physiological milieu by an ammonium solution, as previously described 249

Control condition
Fluorescence from presynaptic membrane and SVs

SV acidification defect
Fluorescence from presynaptic membrane and SVs

Control condition
Fluorescence from presynaptic membrane  This effect was not simply due to a difference in larval size, as the average length and width of 300 recorded larvae were not significantly different (Supplementary Figure 3). 301

SV acidification defect
In contrast, the stride size actually showed little or no difference, depending on the RNAi, 302 between knockdown larvae and controls ( Figure 4G and H). Instead, we found that the stride 303 duration, which is the time necessary to accomplish one peristaltic wave, was significantly 304 longer in knockdown animals compared to controls for both RNAi ( Figure 4I and J). These results 305 suggest that the knockdown larvae may need about twice as much time as controls to reach the 306 required level of muscular contraction to accomplish one stride. This could be a compensatory 307 mechanism to adapt to a lower amount of neurotransmitter released in response to motor 308 nerve stimulation, as would be expected if synaptic vesicles are less filled with neurotransmitter 309 in CG31030 knockdown context. 310 RNAi1 and RNAi2 larvae appeared lower than controls, this effect was not statistically significant 345 ( Figure 5E). This result suggests that CG31030 knockdown did not significantly increased the 346 number of unacidified vesicles empty of neurotransmitter. 347 In this study, we investigated the hypothesis that the previously uncharacterized Drosophila 368 protein CG31030 is a specific regulator of the neuronal V-ATPase. At variance with its broadly 369 expressed paralog VhaAC45, we have shown that CG31030 is found mainly, if not only, in 370 neurons. We also provide evidence that CG31030 interacts with two constitutive subunits and 371 one accessory subunit of the V-ATPase, the constitutive ones being also enriched in neurons, 372 and that it is required to have properly acidified synaptic vesicles. This implies that CG31030 is 373 an essential protein for nervous system functioning in Drosophila.
In yeast, all V-ATPase subunits are coded by a single gene, with the exception of the V0 subunit 377 a. The knock-out of any of the single-gene subunits all present a similar phenotype: the inability 378 to survive in a neutral pH environment (Nelson, 2003). For subunit a, the same phenotype was 379 only achieved in a double-mutant of both isoforms (Manolson et al., 1994).  Table 1) and the missing two-thirds likely died at an early 388 developmental stage as no larval lethality was observed. This could be explained by the fact that 389 CG31030 protein re-expression in rescued knock-out flies first required the expression of Gal4 390 under regulation of the elav promoter, which starts to express rather late in embryos, followed 391 by activation of the UAS sequence upstream the CG31030 insert. So, it is possible that part of 392 the mutant embryos did not survive the delay inherent to this ectopic expression process. 393 We found that CG31030 transcripts follows the repartition of the nervous system, being mainly 394 expressed in the head. Moreover, pan-neuronal expression of CG31030 RNAi3 , the RNAi construct 395 with the weakest effect on fly survival, was sufficient to decrease by more than 80% its level in 396 the brain of adult escapers. These results indicate that CG31030 expression is mainly neuronal, 397 if not even entirely restricted to neurons. In addition, CG31030 cellular localization shows 398 similarity with a synaptic pattern. Cell bodies were also marked, although generally less 399 strongly. Synaptic protein complexes can be assembled in the cell bodies before being 400 The sequence of a V5 tag was inserted in frame after the coding sequence of the CG31030 gene, 531 using a homology-directed repair CRISPR-Cas9 method (see Figure 1C). The following guide RNA 532 sequence: 5'-TTCACCGTACAGGAGTAAGG-3' was cloned into the BbsI site of pCFD3: (TTX) 10 -6 M, so that no spike could occur, from the ventral longitudinal abdominal muscle 6 in 651 segment A3. Quantal analysis was performed following the theoretical background described in 652 (Kuno, 1971) and (Castellucci and Kandel, 1974)