Critical role for Piccolo in synaptic vesicle retrieval

Loss of function of the active zone protein Piccolo has recently been linked to a disease, Pontocerebellar Hypoplasia type 3, which causes brain atrophy. Here, we address how Piccolo inactivation in rat neurons adversely affects synaptic function and thus may contribute to neuronal loss. Our analysis shows that Piccolo is critical for the recycling and maintenance of synaptic vesicles. We find that boutons lacking Piccolo have deficits in the Rab5/EEA1 dependent formation of early endosomes and thus the recycling of SVs. Mechanistically, impaired Rab5 function was caused by reduced synaptic recruitment of Pra1, known to interact selectively with the zinc finger domains of Piccolo. Importantly, over-expression of GTPase deficient Rab5 or the Znf1 domain of Piccolo restores the size and recycling of SV pools. These data provide a molecular link between the active zone and endosome sorting at synapses providing hints to how Piccolo contributes to developmental and psychiatric disorders.


Introduction
Piccolo is a scaffolding protein of the cytomatrix assembled at the active zone (CAZ), a specialized region within the presynaptic terminal where SV fusion takes place (Ackermann et al., 2015;Gundelfinger et al., 2015). It is a multi-domain protein, which is exclusively localized to active zones where it forms a sandwich-like structure enclosing Bassoon (Nishimune et al., 2016).
At present, the function of Piccolo is not well understood. Initial genome-wide association studies have linked missense mutations in Piccolo to psychiatric and developmental disorders including major depressive, bipolar disorder (Choi et al., 2011;Giniatullina et al., 2015;Minelli et al., 2012;Sullivan et al., 2009;Woudstra et al., 2013) and Pontocerebellar Hypoplasia type 3 (PCH3) (Ahmed et al., 2015). How an active zone protein like Piccolo contributes to these diseases is still unclear. Cellular, biochemical and molecular studies indicate a structural role for Piccolo in organizing the active zone (AZ) (Ackermann et al., 2015;Gundelfinger et al., 2015). Loss of Piccolo in the retina disrupts the assembly of photoreceptor cell ribbons (Müller et al., 2019;Regus-Leidig et al., 2014). However, its association with the SV priming factors RIM1 and Munc13-1 (Südhof and Rizo, 2011) implies an additional role in SV priming, though a direct role in regulating neurotransmitter release has not been demonstrated (Mukherjee et al., 2010). Furthermore, structure/function studies have shown that Piccolo directly interacts with a number of actin-binding proteins, including Profilin-2, Daam1, Abp1, and Trio, and is critical for the activity dependent assembly of presynaptic F-actin (Fenster et al., 2003;Leal-Ortiz et al., 2008;Terry-Lorenzo et al., 2016;Wagh et al., 2015;Waites et al., 2011;Wang et al., 1999). Moreover, it was also found to interact with proteins known to regulate endosome membrane trafficking like Prenylated Rab acceptor protein 1 (Pra1) (Fenster et al., 2000), and the Arf GTPase-activating protein 1 (GIT1) (Kim et al., 2003). Their interplay suggests a possible role for Piccolo in the activity dependent SV membrane recycling and trafficking.
So far, endosome membrane trafficking within presynaptic boutons is only poorly understood. Early studies have shown endosome structures at presynaptic boutons (Heuser and Reese, 1973) as well as the presence of small Rab GTPases, including Rab4, 5, 7, 10, 11, 35, 26 and 14 (Pavlos and Jahn, 2011). Members of the small Rab GTPase family are organizers of membrane trafficking tasks, as they function as molecular switches alternating between a GDP bound 'off-state' and a GTP bound 'on-state' (Stenmark, 2009). In their 'active state', Rab GTPases can recruit effector proteins such as EEA1 to specific membrane compartments (Christoforidis et al., 1999;Spang, 2009). Often these effector proteins also bind to phosphoinositides (PIs), which are also present at synapses (Uytterhoeven et al., 2011;Wucherpfennig et al., 2003). Together, active small Rab GTPases and specific PI species define the identity of a compartment (Balla, 2013;Schink et al., 2016;Zerial and McBride, 2001).
Perturbation of endosomal proteins as well as over-expression of constitutive active Rab5 (Rab5 Q79L ) affects synapse function (Hoopmann et al., 2010;Rizzoli and Betz, 2002;Wucherpfennig et al., 2003). Thus endosome membrane trafficking appears to be a fundamental feature of presynaptic boutons, helping them maintain functional fidelity over time.
To better understand the synaptic function of Piccolo, we recently generated a line of rats in which the Pclo gene was disrupted by transposon mutagenesis (Pclo gt/gt ) (Medrano et al., 2018). Our analysis of hippocampal synapses reveals that knockout of Piccolo causes a dramatic decrease in the number of SVs per synaptic bouton in combination with a smaller total recycling pool (TRP) of SVs and defects in the efficient reformation of SVs. Mechanistically, we find that these phenotypes are caused by the reduced recruitment of the Piccolo binding partner Pra1 to presynaptic boutons, the activation of Rab5 as well as the formation of early endosomes. These data provide a molecular link between the active zone protein Piccolo, endosome trafficking and the functional maintenance of SV pools providing a possible molecular mechanism for how Piccolo loss could contribute to PCH3 and major depressive disorders among others.

Results
Generation and characterization of the Piccolo knockout rat Transposon mutagenesis was used to generate rats with a disrupted Pclo gene (Pclo gt/gt ) (Medrano et al., 2018). As shown in Figure 1A, the transposon element was integrated into exon 3 of the Pclo genomic sequence, leading to a premature stop in the reading frame. Genotypes of offspring were determined by PCR from genomic DNA ( Figure 1B). Western blot analysis from postnatal day 2 (P2) rat brains as well as from primary hippocampal neuron lysates confirmed the loss of Piccolo full-length protein (Figure 1C and D). Bands at 560 kD as well as between 70-450 kD were detected in lysates from Pclo wt/wt and Pclo wt/gt brains ( Figure 1C), reflecting the expression of multiple Piccolo isoforms from the Pclo gene (Fenster and Garner, 2002). In brain lysates from Pclo gt/gt animals, the dominant bands at 560 and 450 kD were absent, indicating the loss of the predominant Piccolo isoforms, though a few weaker bands at 300, 90 and 70 kD are still present ( Figure 1C). Notably, these are not present in lysates from primary hippocampal neurons ( Figure 1D). Consistently, Piccolo intensity at synapses (positive for VGlut1) was reduced by more than 70% in Pclo gt/gt neurons ( Figure 1E), confirming the loss of most Piccolo isoforms from synapses. Similarly, Piccolo immunoreactivity was absent in hippocampal brain sections ( Figure 1F).
As Piccolo is a core protein of the AZ, we examined the effect of loss of Piccolo on the expression levels of other synaptic proteins. In Western blot analysis of P2 rat brain lysates, most proteins tested, were not affected ( Figure 1G). Intriguingly, Piccolo deficiency was associated with a slight decrease in the expression of Bassoon and the SV protein Synaptophysin ( Figure 1G). The levels of the early endosome marker EEA1 were most severely affected, its levels were decreased by half in Pclo wt/gt and Pclo gt/gt lysates ( Figure 1G). In contrast, the levels of the small Rab GTPase Rab5 were slightly increased in Pclo gt/gt brain lysates, whereas the levels of Dynamin were increased in both Pclo wt/gt and Pclo gt/gt brain lysates ( Figure 1G). However, these observed changes did not affect synapse formation, as VGlut1 puncta opposing Homer puncta were present in Pclo wt/wt and Pclo gt/gt neurons ( Figure 1H) and the number of Synapsin puncta per unit length of primary dendrite was not significantly different between Pclo wt/wt and Pclo gt/gt neurons ( Figure 1I).
Even though the number of synapses was not changed, the levels of several SV proteins were decreased in Pclo gt/gt neurons (

Ultrastructural analysis of Pclo gt/gt synapses
To more directly assess a role of Piccolo in synapse assembly, we analyzed the ultrastructure of synapses by electron microscopy (EM) using high pressure freezing and freeze substitution. EM micrographs from Pclo wt/wt and Pclo gt/gt neurons revealed the presence of synaptic junctions, with prominent postsynaptic densities (PSDs) formed onto axonal varicosities of Pclo gt/gt neurons, with similar dimensions (length) to Pclo wt/wt synapses (Figure 2A,B,E). At many Pclo gt/gt synapses, SVs (<50 nm diameter) were detected ( Figure 2B), however the number of SVs/bouton was significantly reduced (Figure 2A,B,C), which is in line with the earlier observed reduced levels of Synaptophysin, Synaptotagmin and VGlut1 (Figure 1-figure supplement 1). Even though the total number of SVs was changed, the number of docked SVs at AZs was not altered in Pclo gt/gt synaptic terminals ( Figure 2D,H,I). Intriguingly, lower SV density in Pclo gt/gt boutons was accompanied by a high number of endosome-like structures with diameters larger than 60 nm (Figure 2A,B,F,G). These data suggest that Piccolo is required for the maintenance and/or recycling of SVs.
To test this hypothesis, we performed two types of experiments. Initially, immunocytochemistry was used to measure the total pool of SVs by staining neuronal cultures for the SV protein Synaptophysin. This revealed a reduction of Synaptophysin content per bouton ( Figure 3F and G), indicating that SV loss (as seen in EM micrographs) (Figure 2A and B) is a putative contributor to the reduced functionality of Pclo gt/gt neurons. Next, we performed FM-dye uptake experiments (Smith and Betz, 1996) to determine the size of the total recycling pool (TRP) of SVs. Here, Pclo gt/gt neurons displayed a 30% reduction in FM4-64 intensity, consistent with a smaller TRP of SVs ( Figure 3H and I). Finally, to assess how efficiently SVs are recycled and reused during high frequency stimulation, we determined FM1-43 unloading kinetics. Remarkably, FM1-43 dye was released at a much slower rate from Pclo gt/gt than Pclo wt/wt boutons ( Figure 3J,K). Moreover, the total amount of FM1-43 dye released after 90 s stimulation was significantly reduced in neurons lacking Piccolo ( Figure 3K). As SV exocytosis is normal, these data indicate that the recycling of SVs must be compromised.
Levels of endosome proteins are reduced at Pclo gt/gt synapses Our initial results raised several questions: a) why is SV pool size smaller in boutons lacking Piccolo and b) why are FM unloading rates slowed? Of note, Pclo gt/gt presynaptic terminals do have increased numbers of endosome-like membranes ( Figure 2B and F), which could be explained by defects in the recycling or reformation of SVs.
As an initial test of this hypothesis, we expressed GFP-Rab5 to monitor the presence of endocytic compartments (Spang, 2009;Stenmark, 2009) in Pclo gt/gt neurons. Interestingly, we observed an increase in the number of GFP-Rab5 positive puncta along axons ( Figure 4B and C). To examine whether this was associated with a general up regulation of the endo-lysosomal pathway, we monitored the presence of GFP-Rab7, a marker for late endosomes (Zerial and McBride, 2001) ( Figure 4A). Surprisingly, the number of GFP-Rab7 puncta along axons was reduced in Pclo gt/gt neurons ( Figure 4B and C), indicating that the maturation of membranes within the endosome compartment from Rab5 positive early endosomes towards late Rab7 positive endosomes is affected by the loss of Piccolo ( Figure 4A).
To gain further insights into which maturation step is possibly impaired (Figure 4), we also analyzed the number of EEA1 positive puncta along axons, a marker for early endosomes. EEA1 is a docking factor, which facilitates homotypic fusion between early endocytic vesicles, facilitating the formation of early endosomes (Christoforidis et al., 1999). Indeed, the number of EEA1 immunopositive puncta along axons was reduced in Pclo gt/gt neurons ( Figure 4B and C), further indicating that the loss of Piccolo alters early endocytic membrane trafficking.
As Piccolo is an AZ protein, we next analyzed whether the levels of Rab5, EEA1 and Rab7 were also changed at synapses. Note that more than 80% of the observed Rab5 puncta along axons colocalize with Synaptophysin, indicating that the major fraction of Rab5 puncta is synaptic (Figure 4figure supplement 2I). Quantifying intensities at Synaptophysin positive boutons revealed that while Rab5 levels were not altered ( Figure 4D and G), EEA1 levels were decreased at Pclo gt/gt synapses ( Figure 4, E and H). The same is true for the synaptic levels of mChRab7; they are also reduced in Pclo gt/gt neurons ( Figure 4F and I).
Fewer endosome proteins are recruited to PI3P-positive organelles in Pclo gt/gt neurons A possible explanation for smaller PI3P-positive organelles in Pclo gt/gt boutons is that the transition between small early endocytic vesicles and larger early endosomes is attenuated. To test this hypothesis, we analyzed the recruitment of two endogenous endosome markers, Rab5 and EEA1, towards GFP-2x-FYVE puncta. Here, we analyzed the fraction of GFP-2x-FYVE/Rab5 double positive puncta as a measure for early endocytic vesicles, and the fraction of GFP-2x-FYVE/Rab5/EEA1 triple positive puncta as a measure for early endosomes ( Figure 5A). For technical reasons this analysis was performed along axons without an additional co-staining for synapses. Of note, co-localization studies with Synaptophysin demonstrated that at least 65% of GFP-2x-FYVE positive puncta along axons were synaptic (Figure 4-figure supplement 2D and E), a concept previously reported at the Drosophila NMJ (Wucherpfennig et al., 2003). Therefore our analysis represents a mixture of PI3Ppositive organelles in-and outside synapses.
In Pclo gt/gt axons, the amount of endogenous EEA1 as well as Rab5 at GFP-2x-FYVE puncta was significantly decreased ( Figure 5B,C,D). Of note, although the intensity of Rab5 at GFP-2x-FYVE puncta was decreased in Pclo gt/gt neurons, the overall fraction of puncta that were double positive for GFP-2x-FYVE and Rab5 was not significantly altered in comparison to Pclo wt/wt neurons ( Figure 5E). In contrast, the fraction of early endosomes (GFP-2x-FYVE/Rab5/EEA1) was reduced by about 70% ( Figure 5F). Taken together, these data indicate that the recruitment of EEA1 towards PI3P-positive organelles is affected by the loss of Piccolo, slowing the maturation of early endocytic vesicles into early endosomes.
GTPase deficient Rab5 (Rab5 Q79L ) expression in Pclo gt/gt neurons rescues EEA1 and synaptophysin levels back to Pclo wt/wt amounts As Rab5 S34N expression in Pclo wt/wt neurons mimics Pclo gt/gt phenotypes, we next tested whether the expression of dominant active, GTPase deficient Rab5 (Rab5 Q79L ) can vice versa rescue Piccolo loss of function phenotypes. We therefore expressed mCh-Rab5 Q79L (Stenmark et al., 1994) together with GFP-2x-FYVE in Pclo wt/wt and Pclo gt/gt neurons. Analyzing Rab5 and EEA1 levels at GFP-2x-FYVE puncta revealed a significant rescue of endogenous Rab5 as well as EEA1 levels in Pclo gt/gt neurons ( Figure 7A,B,C,D). Remarkably, the expression of Rab5 Q79L in Pclo wt/wt neurons had the opposite effect, decreasing EEA1 levels ( Figure 7A and D), though total Rab5 levels were only slightly altered ( Figure 7A and C).
To assess whether Rab5 Q79L also rescues SV pool size in Pclo gt/gt neurons, mCh-Rab5 Q79L /GFP-2x-FYVE expressing neurons were immuno-stained for Synaptophysin. Synaptophysin levels increased upon the presence of dominant active Rab5 Q79L ( Figure 7E and F), indicating that the loss of SVs in Pclo gt/gt boutons is linked to altered Rab5 activity.
To assess whether the presence of Rab5 Q79L can also rescue the smaller TRP of SVs and impaired SV cycling seen in Pclo gt/gt neurons, we performed FM1-43 dye uptake and unloading experiments ( Figure 8A). Here, we found that in Pclo gt/gt synapses harboring Rab5 Q79L the amount of FM1-43 dye after 60 mM KCl stimulation was comparable or even higher than in Pclo wt/wt neurons ( Figure 8A and B). In contrast, the presence of Rab5 Q79L did not alter the amount of FM1-43 dye taken up in Pclo wt/wt neurons ( Figure 8A and B). Surprisingly, Rab5 Q79L expression in Pclo gt/gt neurons did not restore FM1-43 unloading kinetics ( Figure 8C and D). In fact, the presence of Rab5 Q79L decreased FM1-43 unloading in Pclo wt/wt neurons to a similar extent as observed in Pclo gt/gt neurons ( Figure 8C and D). The same is true for the total amount of FM1-43 dye that was released. Here, the reduced levels of FM1-43 in Pclo gt/gt boutons were not restored. Furthermore, the portion that was released in Pclo wt/wt neurons was also decreased ( Figure 8D).

Silencing synaptic activity affects synaptic levels of endosomal proteins
At synapses, endocytosis is associated with synaptic activity as part of the fusion and recycling of SVs. Given that cultured hippocampal neurons are intrinsically active (Minerbi et al., 2009), we considered the possibility that this activity creates a pool of Rab5 positive early endocytic vesicles in Pclo gt/gt boutons that subsequently only poorly mature into early endosomes. To test this concept, hippocampal cultures were treated with Tetrodotoxin (TTX) for 24 hr and the synaptic levels of the endosome markers were examined (Figure 8-figure supplement 1). Interestingly, in Pclo wt/wt neurons, the block of synaptic activity causes the levels of EEA1 to drop (Figure 8-figure supplement 1D, F). The levels of Rab5 and Rabex5 were only slightly altered (Figure 8-figure supplement 1A, C, G, I). This is consistent with a role of synaptic activity in the formation of early endosomes. However, silencing synapses in Pclo gt/gt neurons had no effect on EEA1 levels, and only minor effects on the levels of Rabex5, Rab5 (Figure 8-figure supplement 1B, C, E, F, H, I). These data support the concept that proteins involved in the formation of presynaptic early endosomes are recruited into synapses in response to synaptic activity.
Synaptic levels of prenylated rab acceptor protein 1 (Pra1) are reduced in Pclo gt/gt neurons We next asked the question how Piccolo could be linked to the endosome pathway within presynaptic terminals? Piccolo is best known for its role for the dynamic assembly of presynaptic F-actin (Waites et al., 2011). Therefore, we first tested whether the observed endosome phenotype is attributable to Piccolo's essential role in F-actin assembly. Here, we examined the effects of drugs, that either block (Latrunculin A) or enhance (Jasplakinolide) the assembly of F-actin, on the recruitment of Rab5 and EEA1 towards GFP-2x-FYVE punta. We found that the addition of Latrunculin . (E) mCh-Rab5 Q79L expression in Pclo wt/wt neurons increases Synaptophysin intensities. (F) Quantification of (E). Synaptophysin puncta intensity slightly increases upon mCh-Rab5 Q79L expression in Pclo wt/wt neurons. In Pclo gt/gt neurons, mCh-Rab5 Q79L expression rescues Synaptophysin levels higher than Pclo wt/wt levels (Pclo wt/wt = 1 ± 0.02, n = 620 synapses; Pclo gt/gt = 0.86 ± 0.02, n = 526 synapses; Pclo wt/wt (mCh-Rab5 Q79L )=1.07 ± 0.03, n = 473 synapses; Pclo gt/gt (mCh-Rab5 Q79L )=1.26 ± 0.03, n = 446 synapses; three independent experiment). Scale bars represent 10 mm. Error bars in bar graph represent 95% confidence intervals. Numbers given represent mean ± SEM. C, D and H ANOVA with Tukey multi comparison test. ** denotes p<0.01, *** denotes p<0.001 and **** denotes p<0.0001. DOI: https://doi.org/10.7554/eLife.46629.019 The following source data is available for figure 7: Source data 1. This spreadsheet contains the normalized values used to generate the bar plots shown in Figure 7C,D and F. DOI: https://doi.org/10.7554/eLife.46629.020 significantly enhanced the levels of Rab5, but not EEA1 (Figure 8-figure supplement 2A). Consistently, Jasplakinolide was not able to enhance Rab5 and or EEA1 levels at GFP-2x-FYVE punta in boutons lacking Piccolo (Figure 8-figure supplement 2B), indicating that Piccolo's role in F-actin assembly is not important for the formation of early endosomes. Hence the phenotype we observe must be due to a different function of Piccolo.
Interestingly Piccolo has been shown to also interact with Pra1 (Fenster et al., 2000), which is a GDI replacement factor. It is part of the Rab-GTPase activating/deactivating cycle, as it places the GTPase onto its target membrane (Pfeffer and Aivazian, 2004). It is thus tempting to speculate that Pclo gt/gt neurons lack synaptic Pra1, subsequently causing reduced Rab5 activation and impaired early endosome formation. To test this hypothesis, we analyzed the synaptic levels of Pra1 in Pclo wt/wt and Pclo gt/gt neurons. We found that Pra1 localized to presynaptic terminals, co-localizing with Synaptophysin in Pclo wt/wt neurons ( Figure 9A and B). However, in Pclo gt/gt neurons synaptic Pra1 levels were reduced ( Figure 9A,B,C). These data suggest that the reduced formation of Figure 9. The loss of Piccolo leads to diminished levels of synaptic Pra1. (A) Images of Pclo wt/wt and Pclo gt/gt neurons stained with Synaptophysin and Pra1 antibodies. Pra1 is present at Synaptophysin positive synapses in Pclo wt/wt synapses but is reduced in Pclo gt/gt synapses. (B) Detail images of areas indicated in (A). (C) Quantitation of (A). Loss of Piccolo decreases Pra1 intensity at synapses compared to Pclo wt/wt synapses (Pclo wt/wt = 1 ± 0.03, n = 474 synapses; Pclo gt/gt = 0.55 ± 0.02, n = 723; three independent experiments). Scale bar represents 10 mm and 5 mm in zoom image, Error bars in bar graph represent 95% confidence intervals. Numbers given represent mean ± SEM, Student's t -test. **** denotes p<0.0001. DOI: https://doi.org/10.7554/eLife.46629.025 The following source data and figure supplement are available for figure 9: Source data 1. This spreadsheet contains the normalized values used to generate the bar plots shown in Figure 9C. endosomal membranes in boutons lacking Piccolo is due to a reduced synaptic recruitment of its interaction partner Pra1.
Expression of Pclo-Znf1 rescues synaptic Pra1 and EEA1 levels, synaptic vesicle pool size as well as SV cycling As Piccolo interacts with Pra1 via its zinc fingers (Fenster et al., 2000), we hypothesized that this interaction is critical for its synaptic recruitment and subsequent role in Rab5 activation and early endosome formation. This suggests that the synaptic delivery of even one zinc finger could rescue the reduced synaptic levels of Pra1 and EEA1 in Pclo gt/gt neurons. To test this hypothesis, we expressed mCh-tagged Znf1 of Piccolo (Pclo-Znf1-mCh) in Pclo wt/wt and Pclo gt/gt primary hippocampal neurons and stained for Synaptophysin, Pra1 and EEA1. First, we observed that more than 80% of Pclo-Znf1-mCh puncta co-localized with Synaptophysin in Pclo wt/wt neurons (Figure 9-figure supplement 1), demonstrating it is mainly localized synaptically. Second, the expression of Pclo-Znf1-mCh rescued Synaptophysin levels to greater than wildtype levels in both Pclo gt/gt and Pclo wt/wt boutons ( Figure 10A and D). Third, Pclo-Znf1-mCh over-expression also restored Pra1 levels at Pclo gt/gt synapses to even higher levels than in Pclo wt/wt neurons ( Figure 10B and E). Fourth, Pclo wt/wt synapses harboring Pclo-Znf1-mCh displayed a~30% increase in the levels of synaptic Pra1 compared to synapses of un-infected Pclo wt/wt neurons, indicating that the Piccolo Znf1 recruits Pra1 into presynaptic boutons. Also synaptic EEA1 levels in Pclo gt/gt neurons were significantly increased in the presence Pclo-Znf1-mCh ( Figure 10C and F), further indicating that Pclo-Znf1-mCh supports the accumulation of synaptic Pra1 and along with it EEA1.
Intriguingly, Pclo-Znf1-mCh expression in Pclo neurons also increased the ability of synapses to take up FM1-43 dye during 60 mM KCl stimulation. In Pclo wt/wt as well as Pclo gt/gt neurons, FM 1-43 levels were increased in the presence of Pclo-Znf1-mCh ( Figure 11A and B). Considering FM1-43 unloading kinetics, the presence of Pclo-Znf1-mCh in Pclo wt/wt neurons had only a minor effect. Also the total amount of FM1-43 dye that was released during the stimulation was only slightly affected ( Figure 11C and D). In contrast, in Pclo gt/gt neurons, Pclo-Znf1-mCh lead to a faster FM1-43 unloading rate compared to Pclo gt/gt neurons, although the speed seen in Pclo wt/wt neurons was not fully reached ( Figure 11C and D). Finally, the total amount of dye released was increased ( Figure 11D).
Taken together, these data support our central hypothesis that Piccolo via its zinc fingers and its binding partner Pra1 plays critical roles during the recycling and maintenance of SV pools, acting through the activation of Rab5 and EEA1 and the formation of early endosomes ( Figure 11E and F).

Discussion
Our study has shown that Piccolo plays a critical role in the activity dependent recycling of SVs. Specifically, we find that Piccolo loss of function reduces SV numbers as well as the recycling of SVs, without affecting the activity dependent docking and fusion of SVs. Mechanistically, we find that boutons lacking Piccolo accumulate endocytic vesicles that fail to mature into early endosomes due to an impaired Rab5 activation and recruitment of EEA1 towards PI3P-positive organelles. This appears to be a consequence of reduced presynaptic levels of Pra1, subsequently impacting total SV pool size and SV cycling dynamics, reflected in decreased FM-dye unloading kinetics. The observed phenotypes can be restored by the over-expression of Rab5 Q79L as well as the Znf1 domain of Piccolo, which interacts with and restores synaptic Pra1 levels, SV pool sizes and SV cycling dynamics in Piccolo deficient synapses.

Ultrastructural changes in boutons lacking Piccolo
Piccolo and its isoforms are encoded by a large 380 kb gene (Pclo) (Fenster and Garner, 2002). Using transposon mediated mutagenesis (Izsvák et al., 2010), a rat line was created that lacks high and low molecular weight isoforms of Piccolo. However not all isoforms are gone, indicating that the rat model is not representing a complete knockout model (Figure 1). It thus cannot entirely be excluded that truncated versions of Piccolo contribute to the here-observed phenotypes.

Synaptic transmission in boutons lacking Piccolo
A fundamental question is how the loss of Piccolo and the reduced number of SVs adversely impacts synapse function. Our electrophysiological data indicate that the evoked release of neurotransmitter is not affected (Figure 3). This is somewhat surprising as Piccolo has been shown to physically interact with AZ proteins including RIM, RimBP, VGCC and Munc13 (Gundelfinger et al., 2015), which are known to be involved in SV priming and calcium mediated SV fusion (Breustedt et al., 2010;Davydova et al., 2014;Girach et al., 2013;Schoch et al., 2002). However, a function in synaptic release could easily be masked by complementary roles played by other AZ proteins such as Bassoon (Gundelfinger et al., 2015).
Although RRP, Pvr and PPR were not altered (Figure 3), we did observe an enhanced rundown of EPSC amplitudes during 10 Hz stimulation in boutons lacking Piccolo (Figure 3). Conceptually, this could be related to a faster depletion of a smaller reserve pool of SVs or defects in the retrieval of SVs from the plasma membrane. Reduced levels of FM4-64 dye after KCl stimulation support the idea that the TRP of SVs is smaller in Pclo gt/gt neurons ( Figure 3). Furthermore, the larger number of endocytic-like vesicles ( Figure 2) suggests that loss of Piccolo impairs efficient recycling of SVs, reducing the pool of SVs. Accordingly, FM1-43 unloading in boutons lacking Piccolo was dramatically slowed indicating that SVs are not efficiently regenerated within this time frame (Figure 3).

Early endosome trafficking is impaired in boutons lacking Piccolo
An important question is how Piccolo loss of function impairs SV regeneration/recycling? Interestingly, the loss/inactivation of proteins regulating SV endocytosis, like Endophilin, Synaptojanin, Dynamin and the small GTPase Rab5, cause similar phenotypes to those observed in Pclo gt/gt neurons (Hayashi et al., 2008;Milosevic et al., 2011;Schuske et al., 2003;Wucherpfennig et al., 2003). However, our data with GFP-2x-FYVE argues against a block in early events of endocytosis as a) GFP-2x-FYVE levels are not significantly altered in Pclo gt/gt neurons (Figure 4-figure supplement  2) and b) PI3P is only generated on already pinched off vesicles (Harris et al., 2000;Mani et al., 2007;Verstreken et al., 2003). Importantly, GFP-2x-FYVE puncta are smaller in Pclo gt/gt boutons (Figure 4-figure supplement 2). This indicates that while the initial uptake of SV membranes is not affected, the subsequent formation of the early endosome compartment is disturbed. Of note, our EM analysis revealed the accumulation of round vesicles with a diameter >60 nm, which could represent early endocytic vesicles (Figure 2).
In line with this concept, we observed that loss of Piccolo differentially affected the appearance of markers for early and late endosomes. Rab5 puncta accumulate along Pclo gt/gt axons, whereas EEA1 and Rab7 puncta numbers are decreased (Figure 4), indicating that the loss of Piccolo leads to an early pre EEA1 block in the maturation of the endosome compartment. Consistently, the Figure 11 continued FM1-43 intensities in Pclo wt/wt and Pclo gt/gt boutons either positive (arrow) or negative (arrowhead) for Pclo-Znf1-mCh. Right: Images depicting FM1-43 unloading kinetics in Pclo wt/wt and Pclo gt/gt boutons either positive (arrow) or negative (arrowhead) for Pclo-Znf1-mCh. (D) Quantification of (C). The presence of Pclo-Znf1-mCh is sufficient to partially rescue slowed FM1-43 unloading kinetics and total FM1-43 dye released in Pclo gt/gt boutons (Pclo wt/wt = 52.64 ± 1.04, n = 213 synapses; Pclo gt/gt = 40.23 ± 1.55, n = 144 synapses; Pclo wt/wt + Pclo-Znf1-mCh = 48.7 ± 0.96, n = 257 synapses; Pclo gt/gt + Pclo-Znf1-mCh = 48.74 ± 0.87, n = 201 synapses; four independent experiments). (E) Model illustrating the contribution of Piccolo in the recycling of synaptic vesicles. Left panel: Piccolo regulates Rab5 function and subsequently early endosome formation through its interaction with Pra1. In the absence of Piccolo (right panel) less Pra1 is localized within the presynaptic terminal. Less synaptic Pra1 negatively impacts Rab5 function, slowing early endosome formation with the consequence that SVs pools and synaptic activity are decreased over time. Scale bars represent 5 mm. Error bars in bar graph represent 95% confidence intervals. Numbers given represent mean ± SEM. ANOVA with Tukey multi comparison test. * denotes p<0.05, *** denotes p<0.001 and **** denotes p<0.0001. DOI: https://doi.org/10.7554/eLife.46629.030 The following source data is available for figure 11: Source data 1. This spreadsheet contains the normalized values used to generate the bar plots shown in Figure 11B and D. DOI: https://doi.org/10.7554/eLife.46629.031 synaptic levels of the endosome proteins EEA1 and Rab7 are decreased in Pclo gt/gt neurons (Figure 4), supporting the hypothesis that Piccolo is important for recruitment of endosome proteins towards synapses. As Rab5-positive early endocytic vesicles are not able to mature into EEA1 positive early endosomes, they accumulate at the synapse and eventually leave it without gaining an early endosome identity, indicated by increased numbers of GFP-Rab5-positive puncta along Pclo gt/gt axons (Figure 4). Additionally, lower synaptic levels of mchRab7 as well as reduced numbers of GFP-Rab7 positive puncta along axons indicate that trafficking of proteins through late endosomal structures could be affected. This would subsequently affect the efficient endo-lysosomal degradation of defective molecules.
PI3P as well as Rab5-GTP are necessary to promote the formation of early endosomes (Spang, 2009). PI3P levels are not altered in Pclo gt/gt neurons (Figure 4-figure supplement 2), suggesting a possible defect in the activation of Rab5. It is activated, amongst others, via its GEF Rabex5 (Horiuchi et al., 1997). Interestingly, levels of Rabex5 at GFP-2x-FYVE puncta are reduced in Pclo gt/gt neurons ( Figure 5-figure supplement 1). There are also fewer organelles harboring a complex of Rabex5 and Rab5 ( Figure 5-figure supplement 1). Importantly, this complex is necessary for the activation of Rab5, the recruitment of EEA1 and the formation of early endosomes (Murray et al., 2016;Simonsen et al., 1998;Stenmark et al., 1995). Alterations in Rabex5 could lead to less active Rab5 and fewer early endosomes, a concept further supported by the fact that the expression of a GDP-locked Rab5 (Rab5 S34N ) in Pclo wt/wt neurons also causes a reduction in EEA1 at GFP-2x-FYVE puncta and thus fewer early endosomes ( Figure 6). However Rab5 S34N expression does lead to more GFP-2x-FYVE/Rab5 double positive organelles indicating that the inactive GTPase still becomes associated with PI3P-positive organelles. This is different in Pclo gt/gt neurons, where already less Rab5 is present at PI3P-positive organelles, indicating that the loss of Piccolo may cause less efficient recruitment of Rab5 onto endosome membranes (Figures 5 and  6). Consistently, a constitutive active Rab5 (Rab5 Q79L ) rescued EEA1 level at GFP-2x-FYVE puncta (Figure 7). It also rescued Synaptophysin levels back to WT levels, indicating that Rab5 dependent early endosome formation contributes to SV pool size at excitatory vertebrate synapses (Figure 7). This notion is consistent with studies in Drosophila, where Rab5 dominant negative constructs also block the maturation of early endosomes at the neuromuscular junction (NMJ), while the overexpression of WT Rab5 increases SV endocytosis and quantal content (Wucherpfennig et al., 2003).

The formation of early endosomes depends on synaptic activity
Although the molecular mechanisms regulating SV endocytosis have been studied in great detail (Saheki and De Camilli, 2012), less is known about the subsequent endocytic steps and their relationship to synaptic transmission and SV recycling. Important open questions include whether the early endosome compartment is part of the SV cycle and thus activity dependent? Hoopmann and colleagues could show that SV proteins travel through an endosome compartment, while they are recycled (Hoopmann et al., 2010). Furthermore, it has been shown at the Drosophila NMJ that the synaptic FYVE-positive compartment disappears upon block of synaptic activity (Wucherpfennig et al., 2003). Consistently, we could show in rat hippocampal neurons that synaptic levels of EEA1 significantly decrease upon TTX treatment (Figure 8-figure supplement 1), indicating that early endosome formation in vertebrate synapses is also activity dependent.

Links between Piccolo and the endosome pathway
Our data show that Rab5 activity and subsequently the recruitment of EEA1 towards PI3P containing membranes, depends on the presence of Piccolo ( Figure 5). A still open question is how Piccolo can influence Rab5 function and the formation of early endosomes. So far a prominent role for Piccolo for presynaptic F-actin assembly has been described (Waites et al., 2011). Yet neither stabilizing F-actin in Pclo gt/gt neurons rescued the recruitment of Rab5 or EEA1 to GFP-2x-FYVE puncta (Figure 8-figure supplement 2) nor destabilizing F-actin assembly in Pclo wt/wt neurons mimicked endosome phenotypes observed in Pclo gt/gt neurons (Figure 8-figure supplement 2). This suggests that the Piccolo associated endosome phenotype is not due to its role in F-actin assembly, though we cannot rule out that it contributes in some parts to the SV recycling phenotype. Further studies are necessary to elucidate this question.
Interestingly, early interaction studies have shown an interaction between Piccolo and the GDI replacement factor (GDF) Pra1 (Fenster et al., 2000), a protein known to promote Rab5 recruitment to endosome membranes where it becomes activated (Abdul-Ghani et al., 2001;Fenster et al., 2000;Hutt et al., 2000;McLauchlan et al., 1998). Pra1 nicely localizes at synapses (Figure 9) and is thus a good candidate to regulate synaptic Rab5 activity. In line with this concept, synaptic levels of Pra1 are reduced in Pclo gt/gt synapses (Figure 9). Importantly, the expression of a single Piccolo Znf1 domain, which binds Pra1 and localizes to presynaptic boutons ( Figure 10, Figure 9-figure supplement 1) (Fenster et al., 2000), was sufficient to restore presynaptic levels of both Pra1 and EEA1 ( Figure 10). These data indicate that Piccolo promotes the localization and stabilization of Pra1 within presynaptic boutons and thus Rab5/EEA1 dependent formation of early endosomes.

Defects in early endosome formation lead to impaired SV cycling at synapses
Our study of Piccolo loss of function neurons reveals consequences to endosome compartments as well as SV cycling. A remaining question is whether the observed defects in the formation of early endosomes are causal for the observed defects in SV cycling. The expression of inactive mCh-Rab5 S34N in Pclo wt/wt neurons also reduced the numbers of early endosomes ( Figure 6) and mimicked functional defects in SV cycling, seen in Pclo gt/gt neurons. The fact that inactive Rab5 negatively impacts early endosome numbers and SV cycling supports the hypothesis that the formation of early endosomes and SV cycling are functionally linked ( Figure 11). In this case, one would also expect that rescuing synaptic levels of endosome proteins through the expression of Rab5 Q79L and/ or Pclo-Znf1-mCh ( Figure 8) would be sufficient to restore functional SV recycling. In fact, both are able to restore the size of the TRP of SVs in Pclo gt/gt neurons (Figure 8). In contrast to Rab5 Q79L , Pclo-Znf1-mCh is also able to speed up FM1-43 unloading kinetics in Pclo gt/gt boutons ( Figure 11). Surprisingly, dominant active Rab5 Q79L does not have such a positive effect on SV cycling (Figure 8). The differences between Rab5 Q79L and Pclo-Znf1-mCh in Pclo gt/gt neurons could be due to the fact that Rab5 Q79L cannot cycle anymore. It has been shown that active cycling of Rab5 is necessary for vesicle fusion (Wucherpfennig et al., 2003). Taken together our functional data support the hypothesis that the dynamic formation of early endosome compartments at synapses is functionally important for efficient SV cycling and that Piccolo, Pra1 and Rab5 are key regulators.
Finally, it is worth considering the disease relevance of the observed changes at Pclo gt/gt synapses. As mentioned above, Piccolo has been implicated in psychiatric, developmental and neurodegenerative disorders, though the molecular mechanism and its contribution to these disorders is unclear. The current study provides insights into one possible underlying mechanism. Specifically, the observed defects in early endosome formation are anticipated to affect the normal trafficking of SV proteins and membranes. This could adversely affect the functionality of synapses throughout the brain, by impairing synaptic transmission, the maintenance of SV pools and the integrity of synapses. Anatomically, these alterations could destabilize synapses, leading for example, to neuronal atrophy during aging, a concept consistent with recent studies showing that Piccolo along with Bassoon regulates synapse integrity through the endo-lysosome and autophagy systems (Okerlund et al., 2017;Waites et al., 2013). Future studies will help clarify these relationships.

Hippocampal cell culture preparation
Micro-island cultures were prepared from hippocampal neurons and maintained as previously described (Arancillo et al., 2013). All procedures for experiments involving animals, were approved by the animal welfare committee of Charité Medical University and the Berlin state government. Hippocampi were harvested from Pclo wt/wt and Pclo gt/gt (Wistar) P0-2 rats of either sex. Neurons were plated at 3000 cells/35 mm well on mouse astrocyte micro-islands to generate autaptic neurons for electrophysiology experiments. For live cell imaging and immunocytochemistry, hippocampal neuron cultures were prepared using the Banker culture protocol (Banker and Goslin, 1988;Meberg and Miller, 2003;Tanaka, 2002). Astrocytes derived from mouse P0-2 cortices were plated into culture dishes 5-7 d before adding neurons. Nitric acid treated coverslips with paraffin dots were placed in separate culture dishes and covered with complete Neurobasal-A containing B-27 (Invitrogen, Thermo Fisher scientific, Waltham, USA), 50 U/ml Penicillin, 50 mg/ml streptomycin and 1x GlutaMAX (Invitrogen, Thermo Fisher scientific, Waltham, USA). Pclo wt/wt and Pclo gt/gt hippocampi were harvested from P0-2 brains in ice cold HBSS (Gibco, Thermo Fisher scientific, Waltham, USA) and incubated consecutive in 20 U/ml papain (Worthington, Lakewood, USA) for 45-60 min and 5 min in DMEM consisting of albumin (Sigma-Aldrich, St. Louis, USA), trypsin inhibitor (Sigma-Aldrich, St.Louis, USA) and 5% FCS (Invitrogen, Thermo Fisher scientific, Waltham, USA) at 37˚C. Subsequent tissue was transferred to complete Neurobasal-A, and triturated. Isolated cells were plated on coverslips with paraffin dots at a density of 100,000 cells/35 mm well and 50,000 cells/20 mm well. After 1,5 hr coverslips were flipped upside down and transferred to culture plates containing astrocytes in complete Neurobasal-A. When necessary, neurons were transduced with lentiviral constructs 72-94 hr after plating. Cultures were incubated at 37˚C, 5% CO 2 for 10-14 days before starting experiments.

Lentivirus production
Lentivirus was produced as previously described (Haferlach and Schoch, 2002). In brief, an 80% confluent 75 cm 2 flask of HEK293T cells was transfected with 10 mg shuttle vector and mixed helper plasmids (pCMVd8.9 7.5 mg and pVSV-G 5 mg) using XtremeGene 9 DNA transfection reagent (Roche Diagnostics, Mannheim, Germany). After 48 hr, cell culture supernatant was harvested and cell debris was removed by filtration. Aliquots of the filtrate were flash frozen in liquid nitrogen and stored at À80˚C until use. Viral titer was estimated by counting cells in mass culture WT hippocampal neurons expressing GFP or mCherry (mCh) as fluorescent reporter. Primary hippocampal cultures were infected with 80 ml of the viral solution (0.5À1 Â 10 6 IU/ml) 72-96 hr post-plating.

Western blot analysis
Brains from P0-P2 animals as well as hippocampal neurons 14 days in vitro (DIV) were lysed in Lysis Buffer (50 mM Tris-HCl, 150 mM NaCl, 5 mM EDTA, 1% TritonX-100, 0.5% Deoxycholate, protease inhibitor pH 7.5) and incubated on ice for 5 min. Samples were centrifuged at 4˚C with 13.000 rpm for 10 min, supernatant was transferred into a fresh tube and the protein concentration was determined using a BCA protein assay kit (Thermo Fisher scientific, Waltham, Massachusetts, USA Electrophysiological recordings were analyzed using Axograph X (Axograph, Berkley, USA), Excel (Microsoft, Redmond, USA), and Prism software (GraphPad, La Jolla, USA). EPSC amplitude was determined as the average of 5 EPSCs at 0.1 Hz. RRP size was calculated by measuring the charge transfer of the transient synaptic current induced by a 5 s application of hypertonic solution (500 mM sucrose in extracellular solution). Pvr was determined as a ratio of the charge from evoked EPSC and the RRP size of the same neuron. Short-term plasticity was analyzed evoking 50 synaptic responses at 10 Hz. PPR was measured dividing the second EPSC amplitude with the first EPSC amplitude elicited with an inter-pulse-interval of 25 ms.

Electron microscopy
For electron microscopy, neurons were grown on 6 mm sapphire disks and processed as earlier described (Watanabe et al., 2013). In brief, cells were sandwiched between spacers and fixed through high-pressure freezing. Following freeze substitution with 1% osmium and 1% uranylacetate in acetone, sapphire discs were washed four times with acetone and postfixed with 0.1% uranylacetate for 1 hr. Following four washing steps with acetone, cells were infiltrated with plastic (Epon, Sigma-Aldrich). Cells were incubated consecutively with 30% Epon in acetone for 1-2 hr, 70% Epon in acetone for 2 hr and 90% Epon in acetone o.n at 4˚C. The next day, cells were further infiltrated with pure Epon for 8 hr. Epon was changed three times. Later the plastic was polymerized at 60˚C for 48 hr. 70 nm sections were cut using a microtome (Reichert Ultracut S, Reichert/Leica, Wetzlar, Germany) and collected on 0.5% formvar grids (G2200C, Plano, Wetzlar Germany). The sections were stained with 2.5% uranyl acetate in 70% methanol for 5 min prior to imaging. Sections were imaged on a TM-Zeiss-900 electron microscope equipped with a 1 k slow scan CCD camera (TW 7888, PROSCAN, Germany). Presynaptic structures were identified by morphology. Image analysis was performed using custom macros (ImageJ). Briefly, vesicle and endosome structures were outlined with the 'freehand' selection tool and area and length was measured. In addition a line was drawn along the PSD and measured. Docked vesicles were counted manually. Vesicles were defined as docked when the vesicle membrane touched the plasma membrane. Statistical significance was determined using Student's t-test with GraphPad Prism.

Immunocytochemistry (ICC)
Hippocampal neurons DIV14-16 were prepared for ICC. In brief, cells growing on coverslips were washed in PBS and subsequently fixed in 4% paraformaldehyde (PFA) for 5 min at RT. After washing in phosphate buffered saline (PBS, Thermo Fisher scientific, Waltham, USA), cells were permeabilized with 0.1% Tween20 in PBS (PBS-T). Afterwards cells were blocked in blocking solution (5% normal goat serum in PBS-T) for 30 min and incubated with primary antibodies in blocking solution overnight at 4˚C. The following antibodies were used: Synaptophysin (1:1000; mouse; synaptic

Rat perfusion
Adult rats of different ages were first bemused in Isoflurane (Abbott GmbH and Co. KG, Wiesbaden, Germany) and then deeply anesthetized with a mix of 20 mg/ml Xylavet (CO-pharma, Burgdorf, Germany) and 100 mg/ml Ketamin (Inresa Arzneimittel GmbH, Freiburg, Germany) in 0.9% NaCl (B/ BRAUN, Melsungen, Germany). Afterwards the heart was made accessible by opening the thoracic cavity. Subsequently a needle was inserted into the protrusion of the left ventricle and the atrium was cut open with a small scissor. Thereby the blood could be exchanged with PBS. Following the exchange of blood, the animals were perfused with freshly made 4% PFA. Following fixation, the brain was dissected from the head and further incubated in 4% PFA for 24 hr at 4˚C. Subsequently the brain was transferred to 15% sucrose at 4˚C for cryo-protection. After the tissue sank down to the bottom of the tube, it was transferred into 30% sucrose and incubated again until sinking to the bottom. Afterwards the brains were frozen using 2-methylbutane (Carl-Roth, Karlsruhe, Germany) cooled with dry ice to À60˚C and stored at À20˚C until cut with a cryostat.
Brain sectioning 20 mm thin sections were cut from frozen brains using a Leica cryostat. Brains were attached to a holder in sagittal as well as coronal orientation and cut at -20˚C. Single sections were transferred from the blade of the knife onto a cooled slide. Slides were stored at À20˚C.

Immunohistochemistry (IHC)
20 mm thin Pclo wt/wt or Pclo gt/gt brain sections were taken out of the freezer and dried for 2 hr at RT. Subsequently sections were surrounded with a liquid barrier marker (AN92.1, Carl-Roth, Karlsruhe, Germany) to prevent solutions from overflowing and washed 3 times for 10 min in TBST (20 mM Trisbase, 150 mM NaCl, 0.025% Triton X-100). Subsequently, sections were blocked for 2 hr in TBS plus 10% normal goat serum and 1% BSA and incubated with primary antibodies in TBS plus 1% BSA overnight in a humidity chamber at 4˚C. The next day, sections were washed 3 times for 5 min with TBST before adding the secondary fluorophore-conjugated antibody in TBST 1% BSA for 1 hr at RT. After three washing steps in TBS, sections were counterstained for 30 min (1:1000 DAPI in TBS) and washed 2 times for 10 min in TBS. Afterwards, sections were mounted in ProLong Diamond Antifade Mountant (Thermo Fisher scientific, Waltham, USA).
Super-resolution imaging using structured illumination SIM imaging was performed on a Deltavision OMX V4 microscope equipped with three water-cooled PCO edge sCMOS cameras, 405 nm, 488 nm, 568 nm and 642 nm laser lines and a Â 60 10.42 numerical aperture Plan Apochromat lens (Olympus). Z-Stacks covering the whole cell, with sections spaced 0.125 mm apart, were recorded. For each Z-section, 15 raw images (three rotations with five phases each) were acquired. Final super-resolution images were reconstructed using softWoRx software and processed in ImageJ/FIJI.

FM-dye uptake
Functional presynaptic terminals were labeled with FM4-64 dye (Invitrogen, Thermo Fisher scientific, Waltham, USA) as described earlier (Waites et al., 2013). In brief, Pclo wt/wt , Pclo wt/wt expressing mChRab5 S34N or mChRab5 Q79L or Pclo-Znf1-mCh as well as Pclo gt/gt , Pclo gt/gt expressing mChRab5 S34N or Rab5 Q79L or Pclo-Znf1-mCh neurons (12-16 DIV) were mounted in a custom-built imaging chamber and perfused with Tyrode's saline solution (25 mM HEPES, 119 mM NaCl, 2.5 mM KCl, 30 mM glucose, 2 mM CaCl 2 , 2 mM MgCl 2 , pH 7.4) at 37˚C. An image of the basal background fluorescence was taken before the addition of the FM dye. To define the total recycling pool of SV, neurons were incubated with Tyrode's buffer containing 1 mg/ml FM4-64 or FM1-43 dye and 60 or 90 mM KCl for 90 or 60 s. Subsequently, unbound dye was washed off and images were acquired from different areas of the coverslip.

FM-dye unloading
DIV 21 Pclo wt/wt or Pclo wt/wt expressing mChRab5 S34N or Rab5 Q79L or Pclo-Znf1-mCh as well as Pclo gt/gt or Pclo gt/gt expressing mChRab5 S34N or Rab5 Q79L or Pclo-Znf1-mCh primary hippocampal neurons were mounted in a field stimulation chamber (Warner Instruments, Hamden, USA) and perfused with Tyrode's saline solution (25 mM HEPES, 119 mM NaCl, 2.5 mM KCl, 30 mM glucose, 2 mM CaCl2, 2 mM MgCl2, pH 7.4) at 37˚C. At first, an image was taken to obtain background fluorescence. Afterwards, Tyrode's saline solution was exchanged with Tyrode's saline solution plus 1 mg/ ml FM1-43 dye and neurons were stimulated with 900 AP 10 Hz using a field stimulator (Warner Instruments, Hamden, USA), which was controlled by pCLAMP Software (Molecular Devices, Sunnyvale, USA). Following stimulation, unbound dye was washed away with at least 10 ml Tyrode's saline solution passing through the chamber. Following washing, five images were taken to document FM1-43 dye uptake after stimulation. Afterwards cells were stimulated again with 900 AP and 10 Hz. 120 images were taken every second to document unloading of the FM1-43 dye during stimulation.

Image processing
For image processing ImageJ/FIJI (Schindelin et al., 2012) and OpenView software (written by Dr. Noam Ziv, Technion Institute, Haifa, Israel) was used. Piccolo intensity in VGlut1 puncta as well as FM4-64 dye uptake in Pclo wt/wt and Pclo gt/gt neurons was measured using a box routine with Open-View software. To measure Synaptophysin, Synaptotagmin, VGlut1 and Synapsin intensities in Pclo wt/wt and Pclo gt/gt neurons, 9-pixel regions of interest (ROIs) positive for Synaptophysin were manually selected in ImageJ, subsequently the mean intensity within these ROIs was measured using a customized ImageJ script. To determine the number of synapses, the number of Synapsin puncta along dendrites was calculated from randomly picked MAP2 positive primary dendrites. Puncta per unit length of dendrite were counted manually. The number of GFP-Rab5, GFP-Rab7 and EEA1 positive puncta along axons was determined from randomly picked axon sections. Numbers of puncta per unit length of axon were counted manually. The intensity of various endosome proteins (Rabex5, Rab5, Rab7, EEA1, Pra1) at synapses was measured using a customized ImageJ script. 9-pixel ROIs were manually picked based on Synaptophysin staining. Subsequently, corresponding fluorescence intensities were measured in all active channels. The background fluorescence was subtracted from all ROIs before the average intensity of the endosome proteins was calculated from all selected ROIs. The fraction of synapses positive for endosome proteins was calculated with a defined intensity value as threshold. It was defined as the standard derivation intensity of the corresponding endosome protein measured in Pclo wt/wt synapses. The size of GFP-2x-FYVE positive compartments imaged with SIM was measured using ImageJ. The 'freehand' tool was used to mark and measure the area of GFP-2x-FYVE puncta. The intensity of endosome proteins (Rabex5, Rab5, EEA1) at GFP-2x-FYVE puncta was measured using a customized ImageJ script. 9-pixel ROIs were picked manually based on GFP-2x-FYVE staining along axons. Subsequently, corresponding fluorescence intensities were measured in all active channels. The background fluorescence was subtracted from all ROIs before the average intensity of the endosome protein was calculated from all selected eGFP-2x-FYVE puncta. The fraction of double or triple positive GFP-2x-FYVE vesicles was calculated using a defined intensity value as threshold. This was defined as the standard derivation intensity of the corresponding protein intensity measured in Pclo wt/wt synapses. Synaptophysin intensity in the presence or absence of Rab5 Q79L or Rab5 S34N was determined in ImageJ. 9-pixel ROIs were picked along axons positive or negative for GFP-2x-FYVE. Subsequently, Synaptophysin intensity within the defined ROIs was measured using a customized ImageJ script. The intensity of different endosome proteins at the cell soma was measured using ImageJ. The 'free-hand' tool was used to surround the soma. Subsequently the fluorescence intensity within this area was measured and depicted as intensity per soma area. The intensity of different endosome proteins along dendrites was measured using ImageJ. 9-pixel ROIs were picked manually along dendrites marked by MAP2. Subsequently, corresponding fluorescence intensities were measured using a customized ImageJ script.
FM1-43 unloading kinetics in mCh-Rab5 S34N , mCh-Rab5 Q79L or Pclo-Znf1-mCh positive or negative synapses were measured in FM1-43 single or FM1-43 -mCh-Rab5 S34N /mCh-Rab5 Q79L /Pclo-Znf1-mCh double positive puncta from the same coverslip in ImageJ. ROIs were picked manually and subsequently FM1-43 intensities within the defined ROIs were measured throughout the time series using a customized ImageJ script. The percentage of FM1-43 dye released during 90 s field stimulation was calculated analyzing the remaining FM1-43 dye intensity after 92 s. The difference between the beginning of the stimulation and the end was than calculated as % released FM1-43 dye.
Post-processing of automatically measured image data used 'python' and the 'pandas' data analysis package (McKinney, 2010). Statistical analysis was calculated in GraphPad Prism. Data points were plotted using python, matplotlib and seaborn (Hunter, 2007).

Statistical analysis
All values are shown as means ± 95% confidence interval. Statistical significance was assessed using Student's t test or ANOVA for comparing multiple samples. For all statistical tests, the 0.05 confidence level was considered statistically significant. In all figures, * denotes p<0.05, ** denotes p<0.01, *** denotes p<0.001 and **** denotes p<0.0001. also thank the Advanced Light Microscopy core facility at the Radium Hospital in Oslo for access to an OMX super-resolution microscope. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.