Vti1a/b regulate synaptic vesicle and dense core vesicle secretion via protein sorting at the Golgi

The SNAREs Vti1a/1b are implicated in regulated secretion, but their role relative to canonical exocytic SNAREs remains elusive. Here, we show that synaptic vesicle and dense-core vesicle (DCV) secretion is indeed severely impaired in Vti1a/b-deficient neurons. The synaptic levels of proteins that mediate secretion were reduced, down to 50% for the exocytic SNARE SNAP25. The delivery of SNAP25 and DCV-cargo into axons was decreased and these molecules accumulated in the Golgi. These defects were rescued by either Vti1a or Vti1b expression. Distended Golgi cisternae and clear vacuoles were observed in Vti1a/b-deficient neurons. The normal non-homogeneous distribution of DCV-cargo inside the Golgi was lost. Cargo trafficking out of, but not into the Golgi, was impaired. Finally, retrograde Cholera Toxin trafficking, but not Sortilin/Sorcs1 distribution, was compromised. We conclude that Vti1a/b support regulated secretion by sorting secretory cargo and synaptic secretion machinery components at the Golgi.

R egulated secretion of signaling molecules from synaptic vesicles (SVs) and dense-core vesicles (DCVs) is the primary means of neuronal communication. SVs store neurotransmitters and are locally recycled after exocytosis 1-3 , whereas DCVs, which contain neuropeptides and neurotrophins 4,5 , are continuously generated at the trans-Golgi 6,7 . Immature DCVs are proposed to undergo initial fusion reactions, homotypic fusion 8,9 or heterotypic fusion to endosomes 10 , before they are fusion-competent. Fusion requires formation of SNARE complexes consisting of four SNARE protein domains (R, Qa, Qb, and Qc 11 ). In addition to the canonical exocytic SNARE proteins (VAMP2, Syntaxin-1, and SNAP25 1,3 ), other R-and Q-SNAREs are implicated in regulated secretion. However, how these additional SNAREs regulate secretion in relation to the canonical exocytic SNAREs is still poorly understood.
Here, Vti1a/b-null neurons were used to test this hypothesis. We found that the number of DCVs and synapses, and the synaptic levels of proteins that mediate secretion were all reduced. The secretion efficiency of SVs and DCVs was severely impaired. The delivery of DCV-cargo and SNAP25 into axons was reduced. DCV-cargo and SNAP25 accumulated in the Golgi. These phenotypes were rescued by expression of either Vti1a or Vti1b. Furthermore, Golgi cisternae were distended and the export of cargo from the Golgi was impaired. We conclude that Vti1a/b regulate secretion by sorting secretory cargo and proteins of the secretion machinery out of the Golgi.
These results show that Vti1a/b deficiency results in in vitro cell death, and that the surviving neurons are smaller and have fewer synapses and DCVs. Furthermore, expression of either Vti1a or Vti1b rescues virtually all these phenotypes equally well.
Reduced SV and DCV secretion in Vti1a/b-deficient neurons. We investigated how regulated secretion is affected by loss of Vti1a/b expression by quantifying SV and DCV fusion. To study SV fusion, we applied live dual-color imaging of Synaptophysin-pHluorin 27 (SypHy) and FM4-64-labeled SVs (Supplementary Figure 3A), and whole-cell patch clamp. SypHy-fluorescence was quenched under resting conditions ( Fig. 2a; Supplementary Movie 1). High frequency stimulation (HFS, 100 action potentials (APs) at 40 Hz) triggered SV fusion from most synapses (Fig. 2a). A higher percentage of non-responsive synapses was observed in DKO neurons (Fig. 2b). In responsive synapses, the fraction of SVs that fused was 50% lower (Fig. 2c, d). Neither kinetics of fluorescence increase nor fluorescence decay after HFS, a measure for SV endocytosis and re-acidification 27 , were affected in DKO neurons ( Fig. 2e; Supplementary Figure 3E). Consistent with SypHy, DKO neurons also showed reduced FM4-64 uptake upon 60 mM K + depolarization, and reduced release upon HFS (Supplementary Figure 3F-I). These parameters were rescued by expression of either Vti1a or Vti1b ( In whole-cell patch clamp recordings, evoked transmitter release upon single AP or HFS (100APs, 40 Hz) stimulation was decreased by 80-90% (  Figure 3O). After correcting for the decreased SV charge, evoked release remained reduced to a similar extent, 80-90% (Supplementary Figure 3K, N). The frequency of spontaneous fusion events was reduced by 65% (Fig. 2f, i). The readily releasable pool (RRP), quantified as charge transferred during hypertonic sucrose superfusion, was almost seven fold smaller ( Fig. 2j; Supplementary Figure 3P). The release probability, calculated as charge transfer evoked by one AP over the RRP, was 45% lower (Fig. 2k).
Finally, the paired pulse ratio was larger in DKO neurons (Supplementary Figure 3L, Q).
To investigate DCV secretion, we performed live imaging of NPY-pHluorin-labeled DCVs. NPY-pHluorin fluorescence was initially quenched (Fig. 3a; Supplementary movie 2). HFS   Figure 4G), was 85% lower in DKO neurons ( Fig. 3a, b). The DCV fusion probability, calculated as DCV fusion events (Fig. 3b) over total DCV pool ( Fig. 3c), was 3.5-fold lower ( Fig. 3d; Supplementary movie 2). The release kinetics were similar between groups, with~30% events occurring during the These results indicate that DKO neurons have impaired secretion, characterized by smaller readily releasable SV pools, decreased SV release probability, reduced spontaneous SV fusion frequency and amplitude, and decreased DCV fusion probability. Vti1a/b are dispensable for fusion kinetics during stimulation, SV endocytosis and acidification or DCV cargo loading and acidification.
Reduced protein influx into axons in Vti1a/b DKO neurons. Increased protein degradation or reduced delivery to synapses are the two main explanations for the decreased synaptic levels of proteins that mediate secretion. To address the first possibility, we blocked protein translation for 24 h with cycloheximide. Cycloheximide treatment reduced synaptic SNAP25 and Munc13-1 levels to a similar extent in DHZ and DKO neurons (Supplementary Figure 5K-O). To address protein delivery, SNAP25-EGFP transport was monitored in DIV-5 neurons after photobleaching the proximal axon, identified with the axon initial segment marker NaV II/III -mCherry (Fig. 5a). SNAP25 was selected as reporter because its levels in DKO neurons were decreased the most. SNAP25-EGFP puncta were detected entering the axon in all groups, at an average speed of around 1 µm/s (Fig. 5b). In DHZ neurons, 3.9 puncta/min entered the axon, as opposed to 2.2 in DKO neurons (Fig. 5c).
The reduced number of DCVs detected in DKO neurons ( Fig. 1) may be the consequence of a similar inefficient delivery to axons. To address this, we expressed the DCV-reporter NPY-mCherry and NaV II/III -YFP (Fig. 5d). NPY puncta entered the axon in all groups, at a velocity of 1.0-1.3 µm/s (Fig. 5e). However, similar to SNAP25-EGFP, 40% fewer NPY puncta entered the axon in DKO neurons (Fig. 5f). NPY delivery to the axon was rescued by expression of either Vti1a or Vti1b (Fig. 5d, f). These results indicate that Vti1a/b control the efficient delivery of SNAP25 and DCV cargo to axons.
Because of the preferential localization of Vti1a/b to the Golgi, we examined the structure of this organelle in DKO neurons. At DIV-14, the total areas stained for the cis-Golgi marker GM130 and the trans-Golgi marker TGN46 were reduced in DKO neurons ( Fig. 6d-f), but the staining intensity of these markers was increased (Supplementary Figure 8A, B). Expression of either Vti1a or Vti1b rescued these differences ( Fig. 6d-f; Supplementary Figure 8A, B). A smaller GM130-stained area was also observed in acute hippocampal slices of DKO neurons (Supplementary Figure 8C). At the ultrastructural level, distended Golgi cisternae were frequently observed in DKO neurons (Fig. 6G, H), with clear vacuoles in the proximity of the Golgi (Fig. 6g, i). Occasionally, multilayer membrane structures were observed in DKO neurons (Supplementary Figure 8D). Mitochondria appeared normal (Fig. 6g, j). These data indicate that Vti1a/b localize largely to the Golgi, and that the Golgi morphology is abnormal in DKO neurons.
We subsequently investigated a potential relationship between Golgi defects and the reduced cargo delivery to axons at DIV-5. At DIV-5, the areas stained by GM130, ManII-EGFP, and TGN46 were already smaller in DKO neurons (Supplementary Table 1). Endogenous SNAP25 was found in ManII-EGFP-and TGN46positive areas (Fig. 6k). SNAP25 staining in these areas was greater than in the rest of the soma (Fig. 6k). In DKO neurons, the ratio of SNAP25 expression in ManII-EGFP-or TGN46-positive areas  Figure 8H-J). The expression ratios for SNAP25 and NPY were rescued by expression of either Vti1a or Vti1b (Fig. 6k-p). The expression of the postsynaptic protein PSD95 and TrKB receptors were also decreased and, like NPY-pHluorin and SNAP25, the ratio of expression inside/outside of the Golgi was greater in DKO neurons (Supplementary Figure 9). Hence, in the absence of Vti1a/b, Golgi export appears to be compromised for both presynaptic and postsynaptic proteins, as well as DCV-cargo.

ARTICLE
To further analyze this cargo accumulation, we acquired higher resolution images of the somata. The Golgi was labeled with GM130 and TGN46. Discrete NPY-pHluorin puncta were observed in somata, surrounding the Golgi area (Supplementary Figure 8K). The location of these somatic puncta relative to the Golgi was not altered in DKO neurons (Supplementary Figure 8L). Within the Golgi area, NPY-pHluorin fluorescence was non-homogeneous, with sub-areas of NPY-pHluorin enrichment (Supplementary Figure 8M). In DKO neurons NPY-pHluorin distribution was more homogeneous (Supplementary Figure 8M, N). NPY-pHluorin puncta co-localized to a similar (low) extent with the lysosomal marker Lamp1 (Supplementary Figure 8O-Q).
These data show that Vti1a/b are required for a normal Golgi structure and focal distribution of secretory cargo within subareas of the Golgi. In the absence of Vti1a/b, such cargo, and also SNAP25, accumulate at the Golgi.
Reduced cargo export from the Golgi in Vti1a/b DKO neurons. The accumulation and abnormal distribution of cargo at the Golgi in DKO neurons prompted us to investigate the trafficking to/from this organelle using the RUSH system 29 . NPY-EGFP fused to Streptavidin-binding protein (SBP) was co-expressed in DIV-5 neurons with the ER-resident protein KDEL fused to Streptavidin. The interaction between Streptavidin and SBP traps cargo at the ER 29 (Fig. 7a). Trafficking of the constitutive cargo, GPI, was studied in the same manner. Reversion of the SBP-Streptavidin binding by Biotin application triggered NPY or GPI transport to neurites (Fig. 7b,  c). Prior to biotin application, the distribution of GPI and NPY was dispersed at the somata (Fig. 7d, i), consistent with ER retention 29 . Biotin application triggered GPI and NPY translocation to ManII-ECFP-labeled areas (Golgi) within minutes (Fig. 7d, i), with similar kinetics between groups (Fig. 7d-f, i-k). However, the NPY export kinetics were slower in DKO neurons (Fig. 7d, e) as indicated by the lower fraction that exited the Golgi 64 or 104 minutes after biotin application (Fig. 7g, h). In astrocytes, similar results were obtained (Supplementary Figure 10). For GPI, the export kinetics were initially slower in DKO neurons (32% less cargo exited the Golgi 64 min after biotin application; Fig. 7l), but later converged (Fig. 7m). These data indicate that Vti1a/b are necessary for efficient Golgi export of regulated cargo and, to a lesser extent, constitutive cargo, but not for ER-to-Golgi transport.  Because SNARE proteins mediate membrane fusion, the abnormal Golgi structure and cargo export may be the consequence of defective fusion reactions at the Golgi in DKO neurons. Since anterograde transport from the ER was unaltered (Fig. 7), we investigated retrograde transport to the Golgi using Cholera Toxin subunit-B fused to Alexa-488 (CTB-A488) as reporter 30 . In DIV-14 neurons, CTB-A488 expression was predominantly observed at the Golgi 2 h after application (Fig. 8a). This fluorescence was 37% lower in DKO neurons (Fig. 8b) and 27% lower after normalizing to the fluorescence outside this organelle (Fig. 8c, d). Vti1a expression rescued the relative florescence at the Golgi, but Vti1b expression only showed partial rescue ( Fig. 8a-d). Because CTB transport relies on synaptic activity 30 , we repeated the experiment at DIV-5, but no alterations were observed 2 h after incubation (Supplementary Figure 11). Because we previously observed that defects in GPI export from the Golgi were only detected after short, but not long, incubation times ( Fig. 7l-m), we reasoned that a similar time-dependency may affect CTB-A488 and explain the absence of a measurable phenotype 2 h after CTB-A488 incubation in DIV-5 neurons. To explore this possibility, we measured CTB-A488 expression at the Golgi 30 min after incubation and observed that, at this (shorter) time point, the relative fluorescence at the Golgi was decreased in DKO neuron (Supplementary Figure 11A-D). The mammalian VPS10 orthologs Sortilin and Sorcs1 sort regulated cargo [31][32][33] . Therefore, we studied these proteins as candidates to explain altered Golgi export in DKO neurons. However, cellular levels and distribution of these proteins were normal (Supplementary Figure 12A-D), and Vti1a/b and Sortilin or Sorcs1 did not co-immunoprecipitate ( Supplementary Figure 12, E, F), as opposed to Vti1a/b and a Vti interactor, Syntaxin-16 16 , or Vti1p and VPS10 in yeast 34 (Supplementary Figure 12G).
These results indicate that DKO neurons have reduced retrograde transport of CTB-A488, which is fully rescued by Vti1a, and partially by Vt1b. Furthermore, expression or distribution of the VPS10 proteins Sortilin and Sorcs1 provide no explanation for altered Golgi sorting in DKO neurons.

Discussion
We investigated the function of Vti1a/b in regulated secretion. We showed that in the absence of Vti1a/b neurons form less synapses, contain fewer DCVs and have impaired secretion capacity. Furthermore, the synaptic levels of proteins that regulate secretion were lower, the influx of SNAP25 and DCV cargo into the axon was decreased and these molecules accumulated in the Golgi. The expression of either Vti1a or Vti1b rescued all these phenotypes. Finally, Golgi morphology was abnormal, with distended cisternae and clear vacuoles, cargo export from the Golgi was compromised, and retrograde, but not anterograde, trafficking to the Golgi was impaired.
Vti1a and Vti1b are the two mammalian orthologs of the single-yeast Vti1p gene 34 . Vti1a and Vti1b share only 31% homology 35 . In vitro and ex vivo studies have reported separate functions for these proteins in distinct steps of intracellular trafficking [15][16][17][18][19] . In line with this, we observed (partial) colocalization of Vti proteins with endo-lysosomal markers. We also found that Vti proteins co-localize at the Golgi and that many Vti1a/b-null phenotypes in neurons were fully rescued by the expression of either Vti1a or Vti1b, indicative of redundant functions. Furthermore, the fact that DKO mice show neurodegeneration and die at birth, while single null mutants are viable and fertile 12,36 supports shared functions for these proteins. Hence, Vti1a and Vti1b support many cellular functions in neurons in an equally efficient and largely redundant manner. Such redundancy may have remained unnoticed because most of the previous studies selectively addressed single Vti proteins and/ or used systems where only a single-Vti protein may be expressed. Alternatively, redundancy may be specific for neurons, which is supported by the fact that neurons are the most affected cell type in the DKO mice 12 . The reduced DCV pool and impaired CTB delivery to the Golgi were the only two phenotypes more efficiently rescued by Vti1a than Vti1b. This aligns with the decreased number of secretory granules in Vti1a KO chromaffin cells 24 and the Vti1a-dependent transport of Shiga toxin to the Golgi in cell lines 16 . Hence, although few aspects of intracellular trafficking are more efficiently supported by Vti1a, Vti1a and Vti1b generally support cellular functions in a largely redundant manner, at least in neurons.
Vti1a is detected in SV fractions 14,37 and local roles in endosomal sorting of SVs 38 and spontaneous SV fusion 20,39 have been proposed. This latter proposal was supported by the spontaneous fusion of Vti1a-labeled acidic organelles (which were distinct from Syntaxin6-or transferrin receptor-labeled organelles) and by a reduced spontaneous SV fusion upon single Vti1a or combined Vti1a/VAMP7 knockdown 20,39 . We observed that SV and    10 µm (a, d), 500 nm (g), 5 µm (j, m). Expression ration represent reporter intensity in Golgi or TGN over rest of the soma DCV secretion were severely impaired in Vti1a/b-deficient neurons. Because SNAP25 supplies the Qb domain in the SNARE complex that drives SV exocytosis, also spontaneous 22,23 , it seems improbable that the Qb-SNAREs Vti1a/b have a direct role in SV fusion with the plasma membrane. Our data demonstrates that Vti1a/b maintain normal levels of, at least, SNAP25, Munc13-1, RIM, Bassoon, and Synaptotagmin-1 and, to a lesser extent, Syntaxin-1, Synaptobrevin-2, and Munc18-1, and that at least SNAP25 accumulates at the Golgi. Deficiency for most of these proteins inhibits SV fusion 3,40,41 and DCV fusion 21,25,42 , and haploinsufficiency also affects synaptic transmission for at least, SNAP25 or Munc18-1 43,44 . Hence, the combined reduction of the  levels of these proteins provides a plausible explanation for the impaired secretion in the absence of Vti1a/b. Although our data cannot exclude local synaptic functions for Vti1a/b, we conclude that Vti1a/b support synaptic secretion by sorting and targeting components of the secretion machinery to synapses. Fewer DCVs were observed in synaptic micrographs of DKO neurons and staining for DCV-cargo was reduced, whereas DCV cargo loading per vesicle was unaltered. This indicates that DKO neurons express fewer DCVs with probably unaltered cargo composition. In contrast, the number of SVs per synapse was unaffected, while staining intensity for synaptic markers was decreased in DKO synapses. This suggests that, unlike DCV biogenesis, SV biogenesis is unaffected by the loss of Vti1a/b, despite the reduced levels of synaptic proteins. We conclude that SV biogenesis is coupled to the rate of exocytosis, as suggested 2 , and does not depend on Vti1a/b. Furthermore, given the reduced staining of synaptic proteins, SVs most likely contain a reduced number of native proteins per vesicle in DKO neurons. This provides a plausible explanation for the impaired synaptic transmission in Vti1a/b neurons.
The influx of DCV-cargo and proteins of the secretion machinery into the axon was decreased in DKO neurons, but colocalization of these proteins with lysosomes or their cellular stability were unaltered. Therefore, the reduced levels of DCVs and proteins that drive secretion are most likely explained by decreased axonal influx, not by aberrant protein turnover. The two main cellular pathways that can explain the decreased influx into axons are impaired maturation of transport vesicles (by homotypic or heterotypic fusion) or impaired protein export from the Golgi. The Vti partner Syntaxin-6 mediates homotypic fusion of secretory organelles ex vivo 9 , arguing for the first scenario. However, the somatic distribution of DCVs and the cargo content of DCVs were normal in Vti1a/b-deficient neurons, suggesting unaffected maturation. A different Qb-SNARE may support vesicle maturation or, alternatively, this process may not be essential in neurons. In contrast, we observed independent lines of evidence for the second scenario, impaired Golgi export. The Golgi was smaller and cisternae were abnormal, the normal non-homogeneous distribution of DCV-cargo inside the Golgi was lost, cargo accumulated at the Golgi and the transit time of labeled cargo through this organelle was abnormally long. In addition, the export of labeled cargo was also delayed in astrocytes, indicating that Vti1a/b do not only support this function in neurons. Therefore, we conclude that Vti1a/b are required for normal protein export from the Golgi.
The accumulation of DCV-cargo in the Golgi and its abnormally long transit time are consistent with the reduced number of secretory granules in Vti1a-null chromaffin cells 24 and the accumulation of tumor necrosis factor-α in a Golgi-like compartment in macrophages after Vti1b inhibition 45 . Our study shows that targeting of cargo presumably not sorted to the regulated secretory pathway, like SNAP25, GPI, PSD95, and TrKB receptors, was also impaired in DKO neurons. This is in line with the impaired delivery to the plasma membrane of Ca 2+ channels in Vti1a-null chromaffin cells 24 and Kv4 K + channels in HeLa cells deficient for Vti1a or the Vti SNARE partner VAMP7 46 , and indicates that the exit of many proteins from the Golgi relies on Vti1a/b.
Neurons degenerate in the absence of Vti1a/b 12 . Because neuronal death was already abundant at DIV-4, when regulated secretion is minimal 21 , and because abolishing regulated secretion perse does not trigger neuronal death 47 , the degeneration observed in DKO neurons is probably not a direct consequence of impaired exocytosis. However, we cannot exclude that the combined loss of (many) membrane proteins contributes to the degeneration observed. Despite the accelerated cell death of DKO neurons, many cellular functions were normal. The distribution of synapses and DCVs was unaffected, DCV trafficking speed (around 1 µm/s) was in the range of reported values 26,48 ; and SV endocytosis, calcium influx and efflux and vesicular acidification were similar to DHZ neurons, confirming that Vti1a/b DKO neurons can cope with energy-demanding tasks. Furthermore, protein stability, mitochondrial morphology and ER-to-Golgi cargo trafficking were unaffected. This is in line with previous observations that many intracellular trafficking pathways, including recycling of transferrin receptors, delivery of endocytosed cargo to lysosomes and autophagy, were fully functional in Vti1a/b DKO fibroblasts 12 . Golgi abnormalities in the absence of Vti1a/b were also observed in acute brain slices, indicating that this is not a feature unique for cultured neurons. Furthermore, the levels of synaptic proteins and SV fusion were rescued by expression of Vti1a or Vti1b after most DKO neurons died, which confirms that these features are not intrinsic of the surviving neurons. Hence, we conclude that loss of Vti1a/b function does not trigger generalized cellular dysfunctions, but selectively affects regulated secretion by impaired Golgi export of relevant proteins, which is not a unique feature for surviving neurons or culture conditions.
Molecules are delivered to the Golgi anterogradely from the ER, or retrogradely from the plasma membrane or intracellular organelles 49 . ER-Golgi trafficking was normal in Vt1a/b-deficient neurons and astrocytes, in line with the fact that other Qb-SNAREs mediate this anterograde pathway 50,51 . In contrast, retrograde transport from the plasma membrane to the Golgi was impaired in Vti1a/b-deficient neurons, consistent with the decrease in Shiga toxin transport in cell lines upon interference with Vti1a-function using antibodies 16 . The preferential localization of Vti1a/b to the Golgi aligns with a role in retrograde transport to this organelle. Vti1b localized to the Golgi in neurons to a larger extent than what was reported in fibroblast cell lines 52 , which may indicate cell type differences. Vti1a/b deficiency had a larger impact on retrograde transport at DIV-14 than at DIV-5. This can generally be explained by a progressive nature of this defect: components that regulate this trafficking probably become progressively missorted over time. Alternatively, different Qb-SNARE may be involved in similar pathways during different developmental stages 21 , or synaptic activity influences CTB retrograde trafficking, as suggested 30 . Expression of either Vti1a or Vti1b in DKO neurons increased CTB delivery to the Golgi, but only Vti1a fully rescued it. Retrograde transport of Shiga toxin to the Golgi depends on EpsinR 53 , a Vti1b interactor 54 , which confirms the relevance of Vti1b in plasma membrane-to-Golgi transport. In addition to plasma membrane-Golgi transport, additional retrograde transport routes may be affected in DKO neurons. Indeed, Vti1a mediates late endosome-to-Golgi transport 17 , also in yeast 55 . Defective transport of sorting molecules between endo-lysosomal organelles and the Golgi may also help to explain Golgi export defects in Vti1a/b DKO neurons, and would fit with the role of endo-lysosomal proteins in sorting of regulated secretory cargo 10,56,57 . Several cellular phenotypes show striking similarities to those observed for Vti1a/b deficiency. First, blocking expression of the Golgi-Associated Retrograde Protein (GARP)-complex subunit Vps54 in cell lines resulted in inefficient transit of GPI from the Golgi to the cell surface 58 . Second, disrupting either the Conserved Oligomeric Golgi (COG) or the GARP tethering complexes, which function in recycling to or within the Golgi, resulted in distended Golgi cisternae and vacuolization 53,[59][60][61] . These similarities suggest that Vti1a/b may work together with one or both these complexes to regulate transport to or within the Golgi and that impaired delivery of molecules that control protein sorting may explain the cellular phenotypes described in our study (Fig. 8e). Interestingly, Vti1a indeed interacts with COG and GARP subunits 62,63 , but the functional implications of these interactions remains unknown.
In yeast, Vti1p binds to the sorting receptor Vps10 and a Vti1p mutation causes cargo missorting 34 . In mammals, the VPS10 receptors Sortilin and Sorcs1 bind and sort regulated cargo [31][32][33][37][38][39] , and would provide a plausible link between retrograde trafficking defects and other cellular phenotypes observed in Vti1a/b DKO neurons. However, the levels and distribution of these receptors were unaltered in DKO neurons, and binding to mammalian Vti1a or Vti1b was not detected. Additional molecules have been implicated in protein sorting at the Golgi, including calcium transporters 64 , proton pumps 65 , HID-1 66 , CAB45 67 , and lipids 68,69 . Cytosolic proteins, such as actin regulators 64 , adapter complexes 56 , enzymes 70 , and Arf members 71,72 , also contribute to Golgi sorting, but the correct localization of these soluble proteins probably does not rely directly on Vti1a/bdependent membrane trafficking. An overarching model for protein sorting at the Golgi is lacking and the link between molecules that function in this process and Vti1a/b remains unclear. However, the current study clearly indicates that Vti1a/b function is central to coordinate Golgi import/export, with major downstream consequences if this aspect is not functional.

Methods
Laboratory animals and primary cultures. Vti1a/1b null mice, generated by replacing a 6.5 Kb fragment in Vti1a exon 6 and a 7 Kb fragment in Vti1b exon 4 with neomycin resistance cassettes, have been described before 12,36 . Embryonic day (E) 18.5 embryos were obtained by cesarean section of pregnant females from timed mating of Vti1a± and Vti1b ∓ −/ + mice, or vice versa. Mouse hippocampi were dissected from these embryos in HBSS (Sigma), digested in 0.25% trypsin (Life technologies) for 15 min at 37°C and dissociated with fire-polished Pasteur pipettes. Dissociated neurons were resuspended in Neurobasal supplemented with 2% B-27, 18 mM HEPES, 0.25% Glutamax, 0.1% penicillin/streptomycin (Life technologies), and plated on pre-grown glia cells (25,000 to 50,000 neurons/well for DHZ and DKO, respectively). For astrocyte micro-islands, 1500 or 4000 neurons/ well for DHZ and DKO, respectively, were plated. A higher density was used for DKO to compensate for the neuronal death. DHZ were used as controls due to the low probability to obtain DKO and double-WT mice within the same litter. Neurons were maintained at 37°C, 5% CO 2 for 13-15 days (referred as DIV-14) or 4-6 days (referred as DIV-5).Rat and mouse astrocytes were cultured from P1-2 pup cortices. For functional experiments, astrocytes were plated in T75 flasks, grown in DMEM with 10% FCS, 1% NEAA, 1% penicillin/streptomycin (Life Technologies) at 37°C, 5% CO 2 till confluent, and subsequently replated on glass coverslips and grown for 1 week. For micro-islands, astrocytes were prepared by plating 6000 rat glia cells per agarose-coated glass coverslip stamped with 0.1 mg/ ml poly-D-lysine (Sigma) and 0.7 mg/ml rat tail collagen (BD Biosciences). Animals were housed and bred according to institutional and Dutch guidelines.
For immunoprecipitation, Vti1a and 1b were fused to Myc tags, and SorCS1 (gift from J. de Wit, VIB Center for the Biology of Disease, Leuven, Belgium) and Sortilin (gift from C. M. Petersen, Aarhus Universiteit, Denmark) were fused to Flag tags. These plasmids were expressed under CMV promoters and transfected into HEK293T cells.
Image analysis. A custom-written algorithm in MATLAB (Mathworks), SynD 73 , was used to quantify dendritic and axonal length, and SV and DCV puncta from maximum projections of confocal images. For SV and DCV fusion analysis, traces were expressed as fluorescence change (ΔF) over initial fluorescence (F 0 , obtained by averaging the first 10-s frames). For SV fusion, 4×4 pixel ROIs were placed on all synapses, identified as non-mobile SypHy puncta with a ΔF/F 0 > 3 standard deviation (sd) upon NH 4 + superfusion. For DCV fusion, 3×3 pixel ROIs were manually placed on all NPY-pHluorin puncta that appeared upon HFS stimulation. Only when ΔF/F 0 > 3 sd, ROIs were counted as DCV fusion events. The total DCV pool was quantified from the maximum projection during NH 4 + superfusion. SynD was used for DCV puncta detection. The detected puncta may contain several DCVs due to resolution limitations and, hence, pool underestimation was avoided by applying puncta intensity correction. Intensity correction considers that the modal puncta intensity per neuron corresponds to the intensity of individual DCVs that fuse, as demonstrated 25 . For synaptic DCV events, binary masks were built in ImageJ (National Institutes of Health) based on Synapsin-ECFP. DCV fusion events with >50% pixels within the binary mask were considered synaptic. For FRAP experiments, the number and speed of NPY-mCherry or SNAP25-EGFP puncta that entered the axon from the soma were calculated using kymographs (ImageJ). For visualization purposes, brightness, and contrast of representative examples was adjusted in a linear and equal fashion between groups.
For Western blots, samples were separated on 10% SDS-polyacrylamide gels and transferred to Nitrocellulose or PVDF membranes. Blots were blocked in 2% milk and 0.5% bovine serum albumin (BSA) for 1 h at 4°C and incubated with primary and secondary antibodies at 4°C overnight and 1 h at RT, respectively. Blots were scanned using a Fuji Film FLA 5000. Blocking and antibody solutions were prepared in PBS containing 0. 1% Tween-20.
Electron microscopy. Neurons were fixed for 90 min at RT with 2.5% glutaraldehyde in 0.1 M cacodylate buffer, pH 7.4. Cells were washed and postfixed for 1 h at RT with 1% OsO 4 /1% KRu(CN) 6 . After dehydration through a series of increasing ethanol concentrations, cells were embedded in Epon and polymerized for 48 h at 60°C. Ultrathin sections (80 nm) were cut parallel to the cell monolayer, collected on single-slot formvar-coated copper grids, and stained in uranyl acetate and lead citrate in Ultra stainer LEICA EM AC20. Neuronal sections were randomly selected at low magnification and imaged using a JEOL1010 transmission electron microscope at 60 kV.
Statistics. Shapiro and Levene's were used to test distribution normality and homogeneity of variances, respectively. When assumptions of normality or homogeneity of variances were met, parametric tests were used: t-test or ANOVA (Tukey as post hoc). Otherwise, non-parametric tests were used: Mann-Whitney U or Kruskal-Wallis test (Holm as post hoc). R was used as software. No predetermined sample size or randomization were used. No data was excluded. Data was analyzed blindly.
Code availability. The custom code used for data analysis during the current study is available from the corresponding author on reasonable request.
Data availability. The data sets generated and analyzed during the current study are available from the corresponding author on reasonable request.