Drosophila Synaptotagmin 7 negatively regulates synaptic vesicle release and replenishment in a dosage-dependent manner

Synchronous neurotransmitter release is triggered by Ca2+ binding to the synaptic vesicle protein Synaptotagmin 1, while asynchronous fusion and short-term facilitation is hypothesized to be mediated by plasma membrane-localized Synaptotagmin 7 (SYT7). We generated mutations in Drosophila Syt7 to determine if it plays a conserved role as the Ca2+ sensor for these processes. Electrophysiology and quantal imaging revealed evoked release was elevated 2-fold. Syt7 mutants also had a larger pool of readily-releasable vesicles, faster recovery following stimulation, and intact facilitation. Syt1/Syt7 double mutants displayed more release than Syt1 mutants alone, indicating SYT7 does not mediate the residual asynchronous release remaining in the absence of SYT1. SYT7 localizes to an internal membrane tubular network within the peri-active zone, but does not enrich at active zones. These findings indicate the two Ca2+ sensor model of SYT1 and SYT7 mediating all phases of neurotransmitter release and facilitation is not applicable at Drosophila synapses.


Introduction
Neurotransmitter release from presynaptic terminals is the primary mechanism of synaptic communication and is mediated by fusion of synaptic vesicles (SVs) with the plasma membrane at specific sites known as active zones (AZs) (Katz, 1969;Südhof, 2013;Zhai and Bellen, 2004). A highly conserved protein machinery composed of the SNARE complex drives fusion between the SV and AZ lipid bilayers (Littleton et al., 1998;Sö llner et al., 1993;Sutton et al., 1998;Tucker et al., 2004). Ca 2+ influx through voltage-gated Ca 2+ channels functions as the trigger to activate the fusion process (Borst and Sakmann, 1996;Katz and Miledi, 1970;Katz and Miledi, 1967;Schneggenburger and Rosenmund, 2015;Südhof, 2012). The majority of SVs fuse during a synchronous phase that occurs within a few milliseconds of Ca 2+ entry (Borst and Sakmann, 1996;Goda and Stevens, 1994;Llinás et al., 1981;Sabatini and Regehr, 1996;Yoshihara and Littleton, 2002). Many synapses also have an asynchronous component that results in SV release over hundreds of milliseconds (Goda and Stevens, 1994;Hefft and Jonas, 2005;Kaeser and Regehr, 2014;Yoshihara and Littleton, 2002). Asynchronous release normally accounts for less than 5% of SV fusion following single action potentials at Drosophila neuromuscular junctions (NMJs) (Jorquera et al., 2012). This slower phase of release becomes more prominent during high rates of stimulation (Atluri and Regehr, 1998;Lu and Trussell, 2000;Rozov et al., 2019;Zucker and Regehr, 2002) and mediates all SV fusion at some neuronal connections (Best and Regehr, 2009;Peters et al., 2010). Changes in the kinetics and amount of SV fusion also occur during high frequency stimulation, resulting in facilitation or depression depending on the synapse (Zucker and Regehr, 2002). Defining the molecular machinery and Ca 2+ sensors that regulate the distinct modes and kinetics of SV release is essential for understanding synaptic transmission.
The Synaptotagmin (SYT) family of Ca 2+ binding proteins contain key regulators that control the timing of SV release. SYT proteins have a transmembrane domain and two Ca 2+ binding C2 domains termed C2A and C2B (Adolfsen et al., 2004;Adolfsen and Littleton, 2001;Perin et al., 1990;Sugita et al., 2002;Ullrich and Südhof, 1995). Mammals have three SYT family members that localize to SVs (SYT1, SYT2 and SYT9), while Drosophila contains a single member of the SV subfamily (SYT1) (Littleton et al., 1993a;Pang et al., 2006;Xu et al., 2007). These SYT isoforms bind Ca 2+ and activate synchronous fusion of SVs via interactions with membranes and the SNARE complex (Chang et al., 2018;Chapman and Jahn, 1994;Fernández-Chacó n et al., 2001;Geppert et   Phylogenetic tree of SYT1, SYT7 and E-SYT2 from the indicated species generated using the BLOSUM62 matrix with neighbor joining clustering. (C) Comparison of the structure of the C2A and C2B domains of R. norvegicus SYT1 (magenta) with a homology model of D. melanogaster SYT7 (blue). The C2B residues that form the SYT1-SNARE complex primary binding site are highlighted in yellow, with the counterpart changes noted in SYT7. The C2B HB helix in SYT1 is highlighted in green and missing from SYT7. (D) Diagram of the Syt7 genomic locus on chromosome four with coding exons indicated with boxes. Exon 1 (teal) encodes the intravesicular and transmembrane (TM) domains; exons 2 and 3 (white) encode the linker region; exons 4 and 5 encode the C2A domain (dark blue); and exons 6 and 7 encode the C2B domain (light blue). The location of the Syt7 M2 Minos transposon Figure 1 continued on next page at sites of SV fusion (Sugita et al., 2001). If SYT7 were present on endosomes or other internal membrane compartments, it would be more compatible with a role in SV trafficking rather than the fusion process itself. In summary, conflicting studies have generated confusion over how SYT7 contributes to neurotransmission and if the protein plays distinct roles across different neuronal subpopulations or species.
To examine the function of SYT7 in Drosophila, we generated and characterized Syt7 null mutants. The Drosophila NMJ exhibits similar asynchronous release and facilitation properties to those of mammals (Jan and Jan, 1976;Jorquera et al., 2012;Yoshihara and Littleton, 2002), making it a useful system to examine evolutionary conserved functions of SYT7 in neurotransmitter release. We found Syt7 mutants and Syt1; Syt7 double mutants display increased evoked neurotransmitter release, indicating SYT7 negatively regulates SV fusion independent of SYT1. In addition, CRISPR-mediated tagging of the endogenous Syt7 locus indicates SYT7 localizes to a tubular network inside the presynaptic terminal that resides within the peri-active zone (peri-AZ) region, but is not enriched at sites of SV fusion. These data define a role for SYT7 in restricting SV availability and release, and indicate SYT7 is not a major Ca 2+ sensor for asynchronous fusion and facilitation in Drosophila.

Evolutionary conservation and structural comparison of SYT1 and SYT7
Synaptotagmins form one of the largest protein families involved in membrane tracking, with 17 Syt genes encoded in mammals and 7 Syt genes found in Drosophila (Adolfsen and Littleton, 2001;Craxton, 2010;Sugita et al., 2002). Unlike the SV subfamily of SYTs, only a single Syt7 gene is present in vertebrate and invertebrate genomes, making phenotypic comparisons easier. To examine the evolutionary relationship between SYT1, SYT7 and the more distantly related extended-Synaptotagmin (E-SYT) proteins, a phylogenetic tree was generated using the BLOSUM62 matrix and neighbor joining clustering analysis with protein sequences from placozoa (Trichoplax adhaerens), invertebrates (Caenorhabditis elegans, Drosophila melanogaster, Ciona intestinalis) and vertebrates (Danio rerio, Rattus norvegicus, Homo sapiens, Figure 1B). Although Trichoplax lacks neurons, it is the earliest metazoan that encodes Syt genes and contains both a SYT1 and SYT7 homolog (Barber et al., 2009). The phylogenetic tree contains independent clusters that correspond to the SYT1, SYT7 and E-SYT2 protein families. The clustering of SYT1 homologs across evolution correlates with nervous system complexity, with the Trichoplax homolog forming the outlier member of the cluster. Within the SYT7 cluster, C. elegans SYT7 is the most distantly related member, with the Trichoplax homolog residing closer within the cluster. Drosophila SYT7 is more distant from the vertebrate subfamily clade than is Drosophila SYT1 within its subfamily, suggesting SYT7 sequence conservation is not as closely related to nervous system complexity as SYT1. These observations are consistent with SYT7's broader expression pattern and function within neuronal and non-neuronal cells (MacDougall et al., 2018).
To compare SYT1 and SYT7 proteins, we performed homology modeling between Drosophila SYT7 and the published structure of mammalian SYT7 (R. norvegicus SYT7; PBD: 6ANK) (Voleti et al., 2017). Key structural features are highly conserved in the homology model, including the eight-stranded b-barrel and the Ca 2+ binding loops that form the core of C2 domains ( Figure 1C). In contrast to SYT1, both Drosophila and mammalian SYT7 lack the C2B HB helix previously found to have an inhibitory role in SV fusion (Xue et al., 2010). We next performed insertion in exon two is indicated in red. Sequence of the Syt7 M1 CRISPR mutant versus control is shown below with the start codon in green. The guide RNA sequence used to target Syt7 is bolded, with the cleavage site noted by the red arrowhead and the deleted cytosine with a red dash. (E) Western blot of SYT7 protein levels in head extracts of white, CRISPR control, Syt7 M1 , Syt7 M2 and elav C155 -GAL4; UAS-Syt7 (OE SYT7) with anti-SYT7 antisera (top panel). Syntaxin 1 (SYX1) antisera was used as a loading control (bottom panel). SYT7 is overexpressed 2.48 ± 0.4 fold compared to controls (p<0.05, Mann-Whitney unpaired t-test, n = 4). The online version of this article includes the following figure supplement(s) for figure 1: sequence alignment of SYT proteins from H. sapiens, R. norvegicus and D. melanogaster (Figure 1figure supplement 1). Drosophila SYT7 is 59% identical to human SYT7. Comparing the SYT1 and SYT7 subfamilies, the N-terminus encoding the transmembrane domain and linker region has the greatest variability and shares only 21% identity. Within the C2 domains, there is 100% conservation of the negatively charged Ca 2+ binding residues in the C2 loops. A polybasic stretch in the C2B domain that mediates Ca 2+ -independent PI(4,5)P2-lipid interactions is also conserved. These sequence conservations indicate Ca 2+ -dependent and Ca 2+ -independent membrane binding are key properties of both SYT proteins.
Beyond lipid binding, SYT1's interaction with the SNARE complex is essential for its ability to activate SV fusion. Five key C2B residues (S332, R334, E348, Y391, A455) form the primary interaction site that docks SYT1 onto the SNARE complex (Guan et al., 2017;Zhou et al., 2015). Four of the five primary SNARE binding residues are not conserved in Drosophila SYT7 ( Figure 1C, Figure 1figure supplement 1). In addition, Drosophila and mammalian SYT7 contain specific amino acids substitutions at two of these residues that block SNARE binding and abolish SYT1 function in SV fusion (Guan et al., 2017), including C285 (corresponding to Syt1 mutant R334C) and K299 (corresponding to Syt1 mutant E348K). A secondary SNARE complex-binding interface on SYT1 is mediated by conserved basic residues at the bottom on the C2B b-barrel (R451/R452 in Drosophila R388/ R389 in rodents) and is also not conserved in the SYT7 subfamily (Wang et al., 2016;Xue et al., 2010;Zhou et al., 2015). As such, SYT7 is unlikely to engage the SNARE complex via the primary or secondary C2B interface, highlighting a key difference in how the proteins regulate membrane trafficking. Beyond SNARE-binding, 20 nonsynonymous amino acid substitutions are conserved only in the SYT1 or SYT7 subfamilies, suggesting additional interactions have likely diverged during evolution from the common ancestral SYT protein. In summary, SYT1 and SYT7 likely regulate membrane trafficking through distinct mechanisms, consistent with chimeric SYT1/SYT7 rescue experiments in mammals (Xue et al., 2010).

Generation of Drosophila Syt7 mutations
To assay SYT7 function in Drosophila the CRISPR-Cas9 system was used to generate null mutations in the Syt7 locus. Using a guide RNA targeted near the Syt7 start codon, several missense mutations were obtained. To disrupt the coding frame of Syt7, a single base pair cytosine deletion mutant (Syt7 M1 ) located seven amino acids downstream of the start codon was used for most of the analysis, with an unaffected Cas9 injection line as control ( Figure 1D). A Minos transposon insertion in the second coding exon of Syt7 was also identified from the BDGP gene disruption project ) that generates a premature stop codon before the C2A domain, providing a second independent allele (Syt7 M2 ) in a distinct genetic background ( Figure 1D). To characterize the effects of SYT7 overexpression, a UAS-Syt7 transgene was crossed with the neuronal elav C155 -GAL4 driver. Western blot analysis of adult brain extracts with anti-SYT7 antisera confirmed the absence of SYT7 protein in Syt7 M1 and Syt7 M2 mutants and a 2.5-fold increase in SYT7 protein levels in elav C155 -GAL4; UAS-Syt7 ( Figure 1E). Similar to the loss of SYT7 in mice , Drosophila Syt7 null mutants are viable and fertile with no obvious behavioral defects.
The synaptic levels of SYT7 are likely to be rate-limiting for its ability to regulate synaptic transmission since Syt7 M1 /+ heterozygotes displayed an intermediate increase in evoked release compared to Syt7 M1 null mutants. To determine if the effects of SYT7 are dosage-sensitive, SYT7 was overexpressed 2.5-fold by driving a UAS-Syt7 transgene with neuronal elav C155 -GAL4 ( Figure 1E). Overexpression of SYT7 had no significant effect on spontaneous mEJC kinetics or amplitude ( Figure 3A,B), similar to the lack of effect in Syt7 null mutants. However, SYT7 overexpression resulted in a~2 fold decrease in mEJC frequency ( Figure 2C, p<0.05), suggesting elevated levels of SYT7 can reduce spontaneous fusion. Unlike the increased evoked release in Syt7 M1 and Syt7 M1 /+ mutants, SYT7 overexpression caused a striking reduction in eEJC amplitude ( Figure 3D,E) and eEJC charge ( Figure 3F), with only mild effects on SV release kinetics ( Figure 3G  the inhibitory action of SYT7 on SV release is secondary to a presynaptic role, SYT7 was overexpressed postsynaptically using the muscle specific Mhc-GAL4 driver. Overexpression of SYT7 in muscles had no effect on eEJC amplitude or kinetics ( Figure 3-figure supplement 1A, B). We conclude that increased presynaptic SYT7 levels reduce both spontaneous and evoked SV release, indicating SYT7 functions as a negative regulator of neurotransmission.
Analysis of synaptic structure, AZ morphology and presynaptic Ca 2+ influx in Syt7 mutants To determine if enhanced SV release in the absence of SYT7 results from an increase in AZ number or SV docking, synaptic morphology and ultrastructure at the NMJ was analyzed in Syt7 M1 mutants. Motor neurons form en passant synaptic boutons along the axon that contain hundreds of individual AZs marked by a central filamentous T-bar composed of the ELKS/CAST homolog Bruchpilot (BRP) (Ehmann et al., 2014;Wagh et al., 2006). Immunostaining for BRP, the SV-associated protein Complexin (CPX) and a general marker for neuronal membranes (anti-HRP) was performed at muscle 6/7 and muscle 4, the two NMJs analyzed in this study ( Figure 4A-H). There was no change in the total number of synaptic boutons ( Figure 4C,F), AZ number defined by BRP puncta ( Figure 4D,G), or AZ number per muscle surface area ( Figure 4E,H). To examine the AZ T-bar where SVs cluster, high-resolution structured illumination microscopy (SIM) was performed on larval muscle 4 NMJs following anti-BRP immunostaining. Syt7 M1 mutants displayed the normal BRP ring architecture and showed no major difference in morphology compared to controls ( Figure 4I). Individual T-bar size and intraterminal T-bar spacing was quantified in controls and Syt7 M1 mutants on a Zeiss Airyscan confocal. Although BRP ring structure was intact, Syt7 M1 mutants displayed a 25% decrease in the average volume of individual BRP-labeled T-bars ( Figure 4J), but no change in the spacing of T-bars relative to each other ( Figure 4K). We conclude that loss of SYT7 does not disrupt overall AZ morphology or AZ number, though Syt7 M1 mutants display a mild decrease in T-bar volume.
To assay if increased release in Syt7 M1 mutants is secondary to elevated presynaptic Ca 2+ influx, Ca 2+ dynamics at NMJs were analyzed using Fluo-4 AM at 3 rd instar larval Ib motor terminals at segment A3 muscle 6/7 in control and Syt7 M1 . A stimulation paradigm consisting of three epochs of 10 Hz stimulation for 5 s separated by a 5 s rest period was performed ( Figure 4L). Maximum presynaptic Flou-4 AM fluorescence during the stimulation paradigm was significantly greater in control than in Syt7 M1 (control: 10.7 Â 10 6 ± 1.25 Â 10 6 , n = 11 NMJs from eight larvae; Syt7 M1 : 6.52 Â 10 6 ± 0.75 Â 10 6 , n = 9 NMJs from eight larvae, p<0.01, Figure 4L,M). These data indicate SYT7 does not suppress release by acting as a Ca 2+ buffer or a negative regulator of Ca 2+ channel function. Although the mechanism by which presynaptic Ca 2+ influx is reduced in Syt7 mutants is unknown, these data are consistent with the reduced AZ BRP volume ( Figure 4J) and may represent a homeostatic response secondary to the enhanced release in Syt7 mutants.
To determine if enhanced SV docking could increase the number of SVs available for release in Syt7 mutants, SV distribution was quantified at larval muscle 6/7 NMJs in control and Syt7 M1 using transmission electron microscopy (TEM, Figure 5A). No change in overall SV density was observed within Syt7 M1 boutons, indicating SV recycling is largely unperturbed ( Figure 5B). In contrast to the mild decrease in T-bar area ( Figure 4J), there was no change in the length of individual AZs defined by the electron dense synaptic cleft ( Figure 5C, p=0.93; control: 404 ± 34.5 nm, n = 21 AZs from five larvae; Syt7 M1 : 409 ± 28.9 nm, n = 29 AZs from five larvae). To examine docking, SVs in contact with the plasma membrane under the T-bar (within 100 nm, Figure 5D) or just outside the T-bar (100 to 400 nm, Figure 5E) were quantified. No significant change in the number of SVs docked at the AZ plasma membrane was detected ( Figure 5D-F), indicating morphological docking defined by EM is not altered in Syt7 M1 mutants. To quantify SV distribution in the cytoplasm around AZs, SV number was binned into four concentric hemi-circles from 100 to 400 nm radius centered on the T-bar. No significant difference in SV distribution was observed in any bin ( Figure 5G,H), indicating the morphological distribution of SVs around T-bars is intact in the absence of SYT7. We conclude the enhanced release in Syt7 M1 mutants is not due to increased AZ number or docked SVs.

Optical quantal mapping in Syt7 mutants
Given quantal size ( Figure 2B), AZ number ( Figure 4D,G) and SV docking ( Figure 5H) are unchanged in Syt7 mutants, increased release probability (P r ) at individual AZs is a candidate Scale bar = 20 mm for large panels and 2 mm for boxed regions. Synaptic morphology was quantified for 3 rd instar muscle 6/7 (C-E) and muscle 4 (F-H) in controls and Syt7 M1 mutants. No significant differences were detected in synaptic bouton number (C, F); muscle 6/7: p=0.78; control: 81.87 ± 5.301, Figure 4 continued on next page mechanism to mediate the elevated quantal content during single stimuli. We previously developed a quantal imaging approach to map AZ P r at Drosophila NMJs by expressing myristoylated GCaMP6s in muscles (Akbergenova et al., 2018;Melom et al., 2013). Using this approach, P r maps for evoked release were generated for all AZs from Ib boutons at muscle 4 NMJs in control and Syt7 M1 mutants ( Figure 6A). Similar to controls, AZs formed by single motor neurons in Syt7 M1 displayed heterogeneous P r ( Figure 6B). However, P r distribution was strikingly different between the genotypes, with a greater number of high P r and fewer low P r AZs at Syt7 M1 NMJs ( Figure 6C). Syt7 M1 NMJs also had fewer silent AZs that showed no release (control: 19.9%; Syt7 M1 : 4.6%). Overall, mean P r was increased 2-fold ( Figure 6D, p<0.01; control: 0.063 ± 0.002, n = 1158 AZs; Syt7 M1 : 0.12 ± 0.004, n = 768 AZs). In contrast, the maximum AZ P r in the two genotypes was unchanged ( Figure 6D, control: 0.61; Syt7 M1 : 0.63), indicating an upper limit on release strength for single AZs that is similar between controls and Syt7 M1 . We conclude that the enhanced release in the absence of SYT7 results from an increase in average P r across the AZ population.

Loss of SYT7 enhances SV release in Syt1 null mutants
Drosophila Syt1 null mutants have dramatically reduced synchronous SV fusion and enhanced asynchronous and spontaneous release (Jorquera et al., 2012;Lee et al., 2013;Yoshihara et al., 2010;Yoshihara and Littleton, 2002). We generated Syt1; Syt7 double mutants to determine if SYT7 mediates the residual asynchronous release present in Syt1 nulls. A complete loss of asynchronous release in Syt1; Syt7 double mutants should occur if SYT7 functions as the sole asynchronous Ca 2+ sensor, while a reduction in release is expected if it is one of several sensors mediating the residual synaptic transmission in Syt1. Animals lacking SYT1 were obtained by crossing an intragenic Syt1 deletion (Syt1 N13 ) with a point mutant containing an early stop codon (Syt1 AD4 ), an allelic combination referred to as Syt1 Null . Loss of SYT1 results in lethality throughout development, although some Syt1 Null mutants survive to adulthood when cultured under special conditions (Loewen et al., 2001). Surviving Syt1 Null adults are severely uncoordinated and die within several days. Quantification of survival rates demonstrated 45.3% of Syt1 Null mutants survived from the 1 st instar to the pupal stage, with 44.1% of mutant pupae surviving to adulthood (n = 5 groups with >40 starting animals each). In contrast, 5.6% of Syt1 Null ; Syt7 M2 double mutants (referred to as Double Null ) survived from the 1 st instar to the pupal stage, and 6.6% of mutant pupae survived to adulthood (n = 6 groups with >80 animals each). Western blot analysis confirmed loss of both proteins in Double Null mutants and demonstrated no change in expression of SYT1 or SYT7 in the absence of the other family member in individual null mutant backgrounds ( Figure 7A). Although loss of both SYTs caused synergistic defects in survival, residual synaptic transmission must exist given some Double Null mutants survive.

Short-term facilitation does not require SYT7
Although these results indicate SYT7 is a not a key asynchronous Ca 2+ sensor in Drosophila the protein has also been implicated as the Ca 2+ sensor for facilitation (Chen et al., 2017;Jackman et al., 2016;Turecek and Regehr, 2018), a short-term form of presynaptic plasticity that results in enhanced SV fusion during closely-spaced stimuli. To examine facilitation, [Ca 2+ ] was lowered from 2 mM to 0.175 mM or 0.2 mM to identify conditions where the initial P r was matched between control and Syt7 mutants. In 0.175 mM Ca 2+ , controls displayed an 11% failure ratio in response to single action potentials, while Syt7 M1 had no failures ( Figure 8A). In 0.2 mM Ca 2+ , neither genotype had failures ( Figure 8A), although evoked release was increased 3-fold in Syt7 M1 ( Figure 7B,C, p<0.01, control: 7.73 ± 1.5 nA, n = 9; Syt7 M1 : 23.72 ± 6.2 nA, n = 9). In contrast, EJC amplitude was not   statistically different between control in 0.2 mM Ca 2+ (7.73 ± 1.5 nA, n = 9) and Syt7 M1 in 0.175 mM Ca 2+ (8.70 ± 1.6 nA, n = 9). Facilitation was assayed in these conditions where initial P r was comparable. Control and Syt7 M1 mutants displayed robust facilitation to paired pulses separated by 10 or 50 ms at both Ca 2+ concentrations ( Figure 8D). A modest reduction in paired-pulse ratio was observed in Syt7 M1 at 0.175 Ca 2+ compared to control at 0.2 mM Ca 2+ ( Figure 8E,F, p<0.05; 10 ms interval: 31% decrease; 50 ms interval: 22% decrease). These data indicate SYT7 is not the sole effector of facilitation. The mild defect in Syt7 mutants could be due to a partially redundant role for SYT7 in facilitation or secondary to differences in Ca 2+ available to activate the true facilitation sensor. Given [Ca 2+ ] was lowered in Syt7 M1 to match initial P r between the genotypes, the latter hypothesis is more likely.
To determine if short-term facilitation could be elicited in the absence of both SYT1 and SYT7, a 10 Hz stimulation train in 2.0 mM Ca 2+ was given to Double Null mutants and eEJC responses were compared to Syt1 Null mutants alone. Similar to the increased quantal content to single action potentials, Double Null mutants displayed larger facilitating responses during the early phase of stimulation ( Figure 8G-I; cumulative average release for 10 stimuli: Syt1 Null (n = 12): 87 ± 7.0 quanta; Double Null (n = 13): 109 ± 9.9 quanta; 20 stimuli: Syt1 Null : 209 ± 13.8 quanta; Double Null : 261 ± 22.6 quanta; 50 stimuli: Syt1 Null : 594 ± 34.5 quanta; Double Null : 745 ± 56.2 quanta, p<0.03). These results indicate short-term facilitation can occur in the absence of both SYT1 and SYT7, and is enhanced during the early phases of stimulation, consistent with SYT7 negatively regulating SV fusion with or without SYT1.
Syt7 mutants have access to a larger pool of fusogenic SVs but maintain a normal rate of SV endocytosis at steady-state Enhanced SV release in Syt7 mutants could reflect increased fusogenicity of the entire SV population or conversion of a non-fusogenic SV pool into one capable of release in the absence of SYT7. To test whether SYT7 normally renders a pool of SVs non-fusogenic, 1000 stimuli at 10 Hz were applied in 2 mM Ca 2+ at 3 rd instar muscle 6 NMJs to deplete the readily releasable pool (RRP) and drive SV cycling to steady-state. The total number of released SVs and the SV recycling rate was then measured. Both control and Syt7 M1 eEJCs depressed during the stimulation train. However, SV release in Syt7 M1 mutants remained elevated over much of the initial stimulation ( Figure 9A) and the integral of release during the train was greater than controls ( Figure 9B), indicating Syt7 nulls have access to more fusogenic SVs. SV release rate in Syt7 M1 eventually reached the same level as control following depletion of the RRP ( Figure 9C, control quantal content: 131.5 ± 10.7, n = 7; Syt7 M1 quantal content: 123.1 ± 10.5, n = 8). We conclude that SV endocytosis and recycling rate is SYT7independent at steady-state, although Syt7 M1 mutants contain a larger RRP available for fusion.
To further examine SV recycling, FM1-43 dye uptake and release assays were performed in control and Syt7 M1 mutants at 3 rd instar muscle 6/7 NMJs. At low stimulation rates (0.5 Hz), Syt7 M1 mutants took up significantly more FM1-43 dye than controls ( Figure 9D,F), consistent with the increased SV release observed by physiology. In contrast, no significant difference in FM1-43 uptake was found following high frequency 10 Hz stimulation for 500 stimuli ( Figure 9E,G). These data suggest previously exocytosed SVs re-enter the RRP more often in the absence of SYT7 given the normal recycling rate ( Figure 9C). Consistent with this hypothesis, no change in FM1-43 release was detected with high [K + ] stimulation following 10 Hz loading ( Figure 9H). Together with the electrophysiology data, we conclude Syt7 mutants have a larger RRP, but no changes in SV endocytosis. (E) Quantification of facilitation (P2/P1) at 10 ms interval for the indicated genotypes (0.2 mM Ca 2+ : control, 1.93 ± 0.095, n = 9; Syt7 M1 , 1.28 ± 0.12, n = 9; 0.175 mM Ca 2+ : Syt7 M1 , 1.47 ± 0.11, n = 12). (F) Quantification of facilitation (P2/P1) at 50 ms interval for the indicated genotypes (0.2 mM Ca 2+ : control, 1.64 ± 0.043, n = 9; Syt7 M1 , 1.23 ± 0.056, n = 9; 0.175 mM Ca 2+ : Syt7 M1 , 1.34 ± 0.054, n = 12). Statistical significance was determined using one-way ANOVA (nonparametric) with post hoc Tukey's multiple comparisons test for panels A-F. (G) Average eEJC quantal content determined from mEJC charge in 2 mM Ca 2+ during a 10 Hz stimulation paradigm (30 stimuli at 0.5 Hz, 500 stimuli at 10 Hz, and return to 0.5 Hz) in Syt1 Null (black) and Double Null (red). (H) Average quantal content for the last four responses of 0.5 Hz stimulation and the first 14 responses during 10 Hz stimulation in Syt1 Null (black) and Double Null (red). P1 denotes the 1 st response and P2 the 2 nd response to 10 Hz stimulation. (I) Quantification of P2/P1 ratio in Syt1 Null (black, 1.15 ± 0.089, n = 12) and Double Null (red, 1.55 ± 0.22, n = 13) at onset of 10 Hz stimulation. Statistical significance was determined with a Mann-Whitney unpaired t-test for panels H and I. Syt7 mutants have enhanced refilling of the readily-releasable SV pool independent of endocytosis To probe how SYT7 regulates SV cycling and the transition between distinct SV pools, eEJC recovery kinetics following high frequency stimulation were characterized. A paradigm consisting of 30 stimuli at 0.5 Hz, 500 stimuli at 10 Hz and a final 50 stimuli at 0.5 Hz was given to Syt7 M1 mutants, Syt7 M1 /+ heterozygotes and controls in 2 mM Ca 2+ ( Figure 10A). During 0.5 Hz stimulation, Syt7 M1 and Syt7 M1 /+ displayed elevated levels of release. Following the onset of high frequency stimulation, Syt7 M1 and Syt7 M1 /+ synapses depressed while controls displayed a mild facilitation before quickly transitioning to depression ( Figure 10B). Remarkably, Syt7 M1 and Syt7 M1 /+ displayed an extremely rapid recovery of eEJC amplitude and quantal content during the 2 s interval following termination of the 10 Hz train compared to controls ( Figure 10C). A similar rapid recovery was observed in Syt7 M1 after 2000 stimuli were given at 10 Hz to fully deplete the RRP and normalize release rates to control levels (Figure 10-figure supplement 1A-C). These observations suggest SYT7 also functions to reduce SV entry into the RRP, while negatively regulating release of newly regenerated SVs. The enhanced refilling of the RRP did not require SYT1 function, as Double Null mutants also displayed larger eEJCs than Syt1 Null alone during the recovery from a 10 Hz stimulation train ( Figure 8G).
The partial elevation in RRP refilling rate at Syt7 M1 /+ synapses indicates the amount of SYT7 in the presynaptic terminal regulates SV entry into the releasable pool. To determine if RRP refilling is dosage-sensitive, the stimulation paradigm above (0.5 Hz/10 Hz/0.5 Hz) was applied to SYT7 overexpression larvae (elav C155 -GAL4; UAS-Syt7) in 2 mM Ca 2+ . Presynaptic overexpression of SYT7 had the opposite effect of Syt7 mutants and Syt7/+ heterozygotes, not only reducing eEJC amplitude at 0.5 Hz, but greatly limiting the ability of SVs to re-enter the RRP following termination of the 10 Hz stimulation train (Figure 10-figure supplement 2A-C). We conclude that SYT7 limits release in a dosage-sensitive manner by negatively regulating the number of SVs available for fusion and slowing recovery of the RRP following stimulation.
To determine if increased RRP refilling in Syt7 M1 requires an enhanced rate of SV endocytosis or is mediated through refilling from a pre-existing SV pool, recordings were repeated in the presence of the proton pump inhibitor bafilomycin. Bafilomycin blocks neurotransmitter reloading of newly endocytosed SVs and should eliminate the enhanced refilling of the RRP if recycling is essential. Alternatively, if SVs are recruited more rapidly from pre-existing pools, bafilomycin would not abolish the enhanced recovery. The same 0.5 Hz/10 Hz/0.5 Hz paradigm was applied in three successive epochs in the presence of 4 uM bafilomycin or DMSO (control) in the bath solution. As expected, bafilomycin progressively reduced eEJC amplitude throughout the experiment and eliminated most evoked responses during the 3 rd stimulation epoch ( Figure 10D). Syt7 M1 mutants displayed a similar fold-enhancement in the recovery of the RRP in the presence of bafilomycin, though the absolute numbers of SVs re-entering the pool decreased following the 2 nd 10 Hz stimulation as the number of neurotransmitter-containing SVs declined ( Figure 10E,F). We conclude that the rapid refilling of the RRP can occur from pre-existing SV pools. In addition to reducing fusogenicity of SVs already docked at the AZ, these data indicate SYT7 regulates transition kinetics between vesicle pools by reducing the number of SVs moving from the reserve pool to the RRP.

SYT7 localizes to an internal membrane network within the peri-AZ that resides in proximity to multiple presynaptic compartments
Defining the subcellular localization of SYT7 could help elucidate how it modulates SV dynamics. SYT7 could be a resident protein of the SV pool it regulates or reside on an alternative compartment that exerts control over a subset of SVs. To examine the subcellular localization of SYT7, an RFP tag 10 Hz stimulation in control (black, 131.54 ± 10.71, n = 7) and Syt7 M1 (blue, 123.05 ± 10.47, n = 8). Statistical significance for B and C was determined with a Mann-Whitney unpaired t-test. (D) FM1-43 loading in control and Syt7 M1 larvae at muscle 6/7 NMJs in 2 mM Ca 2+ following 150, 300 or 600 stimuli delivered at 0.5 Hz. (E) FM1-43 loading with 500 stimuli at 10 Hz in 2 mM Ca 2+ and FM1-43 unloading with high K + (90 mM) in control and Syt7 M1 larvae at muscle 6/7 NMJs. (F) Quantification of FM1-43 loading following 150, 300 or 600 stimuli delivered at 0.5 Hz. (G) Quantification of FM1-43 loading after 500 stimulati at 10 Hz. (H) Quantification of FM1-43 unloading with high K + (90 mM). Statistical significance was determined with Student's t-test for F-H. Scale bar = 5 mm. Representative eEJC traces of P1 and P2 for control (black) and Syt7 M1 (blue) are shown on the right. (C) Quantification of P531/P530 ratio (P530 is the last response to 10 Hz and P531 is the 1 st response to 0.5 Hz stimulation delivered 2 s after P530) in control (black, 0.93 ± 0.06, n = 8), Syt7 M1/+ (green, 1.33 ± 0.04, n = 12) and Syt7 M1 (blue, 1.91 ± 0.09, n = 8). Representative eEJC traces of P530 and P531 for control (black) and Syt7 M1 (blue) are shown on the right. (D) Representative eEJC traces for control with DMSO (black) or 4 mM bafilomycin (gray) and Syt7 M1 with DMSO (dark blue) or 4 mM Figure 10 continued on next page was introduced at the 3'-end of the endogenous Syt7 locus using CRISPR ( Figure 11A). This approach generated a SYT7 RFP C-terminal fusion protein expressed under its endogenous enhancers to avoid any overexpression that might trigger changes in its normal localization. The RFP C-terminal fusion did not abolish SYT7 function, as eEJC amplitude in 2 mM Ca 2+ was not significantly different between control and SYT7 RFP (control: 198.9 ± 8.8 nA, n = 14; SYT7 RFP , 227 ± 11.3 nA, n = 14, p=0.1). A sfGFP version (SYT7 GFP ) was also generated with CRISPR that showed the same intra-terminal expression pattern as SYT7 RFP (Figure 11-figure supplement 1A). Western blot analysis with anti-RFP identified a single band at the predicted molecular weight (73kD) of the fusion protein in SYT7 RFP animals ( Figure 11B), indicating a single SYT7 isoform is expressed in Drosophila. Immunostaining of 3 rd instar larvae with anti-RFP antisera revealed SYT7 RFP was enriched in presynaptic terminals and formed an expansive tubular network near the plasma membrane that extended into the center of the bouton ( Figure 11C,D). Neuronal knockdown of Syt7 with two independent RNAi lines (elav C155 -GAL4; UAS-Syt7 RNAi) dramatically reduced SYT7 RFP on western blots ( Figure 11B) and eliminated expression of SYT7 RFP at the NMJ ( Figure 11-figure supplement 2), indicating the signal is specific to SYT7 and localizes predominantly to the presynaptic compartment.
SYT7 localization was widespread within the peri-AZ region, with SYT7 RFP tubules in close proximity to other labeled membrane compartments, including endosomes, lysosomes, and the plasma membrane ( Figure 12-figure supplement 1). To determine if the SYT7 compartment required endosomal trafficking for its assembly or maintenance, a panel of dominant-negative, constitutivelyactive or wildtype endosomal UAS-RAB proteins (Zhang et al., 2007) were expressed with elav C155 -GAL4 in the SYT7 RFP background. Manipulations of RAB5 (early endosomes), RAB7 (late endosomes) or RAB4 and RAB11 (recycling endosomes) did not disrupt the abundance or morphology of the SYT7 tubular network (Figure 12-figure supplement 2). Similarly, no change in the distribution of several compartment markers were found in Syt7 M1 mutants, including the early endosomal marker RAB5, the late endosomal/peri-AZ marker RAB11 and the peri-AZ protein Nervous Wreck (NWK) (Figure 12-figure supplement 3). In addition, no defect was observed in trans-synaptic transfer of the exosomal protein SYT4 to the postsynaptic compartment, indicating SYT7 does not regulate

Frequency Low High
Y-axis X-axis Y-axis X-axis Merge Merge Figure 12. Localization of SYT7 in presynaptic terminals. Immunostaining for the indicated proteins in each panel was performed at 3 rd instar larval muscle 6/7 NMJs. Staining for all panels except A were done in the SYT7 RFP endogenously tagged background using anti-RFP to label the SYT7 compartment, with the merged image shown on the right. The Pearson correlation coefficient (r) calculated from the cytofluorogram co-localization plots is shown on the upper right. All images are from single confocal planes. (A) Co-localization of the SV proteins SYT1 (left, magenta, anti-SYT1 antisera) and nSYB (middle, green, endogenous nSYB GFP ) as a positive control. The remaining panels show boutons co-stained for SYT7 RFP (left, magenta, anti-RFP antisera) and the indicated compartment marker (    exosome trafficking as described for several other peri-AZ proteins (Walsh et al., 2019). Although no sub-compartment overlapped completely with SYT7, the protein is positioned within the peri-AZ to interact with SVs, endosomes and the recycling machinery to negatively regulate the size of releasable SV pools (Figure 12-figure supplement 4). We conclude that SYT7 does not localize to SVs and is not enriched at AZs, consistent with SYT7 negatively regulating SV release through an indirect mechanism that does not require its presence at sites of SV fusion.

Discussion
To characterize the location and function of SYT7 in Drosophila we used the CRISPR-Cas9 system to endogenously label the protein and generate null mutations in the Syt7 locus. Our findings indicate SYT7 acts as a negative regulator of SV release, AZ P r , RRP size, and RRP refilling. The elevated P r across the AZ population in Syt7 mutants provides a robust explanation for why defects in asynchronous release and facilitation are observed in the absence of the protein, as SYT7 levels set the baseline for the amount of evoked release. SYT7's presence on an internal tubular membrane network within the peri-AZ positions the protein to interface with the SV cycle at multiple points to regulate membrane trafficking. In addition, increased SV release in animals lacking both SYT1 and SYT7 indicate the full complement of Ca 2+ sensors that mediate the distinct phases of SV release remain unknown.

Syt7 mutants have increased P r at Drosophila NMJs
Using a combination of synaptic physiology and imaging approaches, our findings indicate SYT7 acts to reduce SV recruitment and release. Minor defects in asynchronous release and facilitation were identified in Drosophila Syt7 mutants, as observed in mouse and zebrafish models (Bacaj et al., 2013;Chen et al., 2017;Jackman et al., 2016;Turecek and Regehr, 2019;Turecek and Regehr, 2018;Weber et al., 2014;Wen et al., 2010). However, we attribute these defects to reduced SV availability as a result of increased P r in Syt7 mutants. Indeed, a key feature of facilitation is its critical dependence on initial P r (Neher and Brose, 2018;Zucker and Regehr, 2002). Low P r synapses increase SV fusogenicity as Ca 2+ levels rise during paired-pulses or stimulation trains, resulting in short-term increases in P r for SVs not recruited during the initial stimuli. In contrast, depression occurs at high P r synapses due to the rapid depletion of fusion-capable SVs during the initial response. Prior quantal imaging at Drosophila NMJs demonstrated facilitation and depression can occur across different AZs within the same neuron, with high P r AZs depressing and low P r AZs facilitating (Peled and Isacoff, 2011). Given the elevated P r in Syt7 mutants, the facilitation defects are likely related to differences in initial P r and depletion of fusion-competent SVs available for release during the 2 nd stimuli. A similar loss of SVs due to elevated P r in Syt7 mutants would reduce fusogenic SVs that are available during the delayed phase of the asynchronous response. Syt1; Syt7 double mutants continue to show asynchronous fusion and facilitation, demonstrating there must be other Ca 2+ sensors that mediate these components of SV release. The predominant localization of endogenous SYT7 to an internal tubular membrane compartment at the peri-AZ also places the majority of the protein away from release sites where it would need to reside to directly activate SV fusion. As such, our data indicate SYT7 regulates SV release through a distinct mechanism from SYT1.
We can also conclude that the remaining components of asynchronous fusion and facilitation must be mediated by an entirely different family of Ca 2+ -binding proteins than Synaptotagmins (or through Ca 2+ -lipid interactions). Of the seven Syt genes in the Drosophila genome, only 3 SYT proteins are expressed at the motor neuron synapses assayed in our study -SYT1, SYT4 and SYT7 (Adolfsen et al., 2004). For the remaining SYTs in the genome, SYT-a and SYT-b are expressed in neurosecretory neurons and function in DCV fusion (Adolfsen et al., 2004;Park et al., 2014). SYT12 and SYT14 lack Ca 2+ binding residues in their C2 domains and are not expressed in motor neurons (Adolfsen et al., 2004). In addition, SYT4 is found on exosomes and transferred to postsynaptic cells, where it participates in retrograde signaling (Adolfsen et al., 2004;Harris et al., 2016;Korkut et al., 2013;Walsh et al., 2019;Yoshihara et al., 2005). Syt1; Syt4 double mutants display the same SV fusion defects found in Syt1 mutants alone, indicating SYT4 cannot compensate for SYT1 function in SV release (Barber et al., 2009;Saraswati et al., 2007). As such, SYT1 and SYT7 are the only remaining SYT isoforms that could contribute to SV trafficking within Drosophila motor neuron terminals.
A prior study from our lab using a Syt7 exon-intron hairpin RNAi we generated did not result in an increase in evoked release (Saraswati et al., 2007). Although a reduction in ectopic expression of SYT7 in muscles could be seen with Mhc-GAL4 driving the UAS-Syt7 RNAi, our anti-SYT7 antisera does not recognize the endogenous protein in neurons using immunocytochemistry, preventing a determination of presynaptic SYT7 levels following neuronal RNAi. To further examine this issue, we performed western blot analysis with this RNAi and compared it those used in the current study. Our results confirmed that the RNAi line failed to reduce endogenous GFP-tagged SYT7 (data not shown), although the two commercial RNAi lines we used here were highly effective ( Figure 11B). Based on these data, we conclude that the previous Syt7 UAS-RNAi line was ineffective in knocking down endogenous SYT7. Given the Syt7 M1 and Syt7 M2 alleles result in early stop codons and lack SYT7 expression by western blot analysis and display elevated levels of fusion, our data indicate SYT7 normally acts to suppress SV release as demonstrated by electrophysiology and optical P r imaging. SYT7 overexpression reduces SV release even more, further confirming that the levels of SYT7 set the baseline amount of SV fusion at Drosophila NMJ synapses.

SYT7 regulates the recruitment and fusion of SVs in a dose-dependent manner
Although our data indicate SYT7 is not the primary asynchronous or facilitation Ca 2+ sensor in Drosophila, we found it inhibits SV release in a dosage-sensitive manner. The reduction in SV release is not due to changes in the Ca 2+ cooperativity of fusion or enhanced presynaptic Ca 2+ entry, ruling out the possibility that SYT7 normally acts as a local Ca 2+ buffer or an inhibitor of presynaptic voltage-gated Ca 2+ channels. The reduction in release is also not due to altered endocytosis, as Syt7 mutants have a normal steady-state rate of SV cycling following depletion of the RRP. Instead, SYT7 regulates SV fusogenicity at a stage between SV endocytosis and fusion. Given the rapid enhanced refilling of the RRP observed in Syt7 mutants, and the suppression of RRP refilling following SYT7 overexpression, our data indicate SYT7 regulates releasable SVs in part through changes in SV mobilization to the RRP. Ca 2+ is well known to control the replenishment rate of releasable SVs, with Calmodulin-UNC13 identified as one of several molecular pathways that accelerate RRP refilling in a Ca 2+ -dependent manner (Dittman et al., 2000;Dittman and Regehr, 1998;Junge et al., 2004;Lipstein et al., 2013;Ritzau-Jost et al., 2018). Our findings indicate SYT7 acts in an opposite fashion and slows RRP refilling, providing a Ca 2+ -dependent counter-balance for SV recruitment into the RRP. Although such an effect has not been described for mammalian SYT7, defects in RRP replenishment have been observed when both SYT1 and SYT7 are deleted or following high frequency stimulation trains Durán et al., 2018;Liu et al., 2014).
SYT7's role in restricting SV fusion and RRP size also affects spontaneous release. Although Syt7 mutants alone do not show elevated mini frequency, Double Null mutants have a 2-fold increase in spontaneous release. Similar increases in spontaneous release were observed at mammalian synapses lacking both SYT7 and SYT1 (or SYT2), with the effect being attributed to a dual role in clamping fusion in the absence of Ca 2+ (Luo and Südhof, 2017;Turecek and Regehr, 2019). Our results indicate the elevation in spontaneous release at Drosophila synapses is a result of an increase in releasable SVs rather than a clamping function for SYT7. Following overexpression of SYT7, there is a reduction in the number of fusogenic SVs available for both evoked and spontaneous release. The dosage-sensitivity of the various phenotypes indicate SYT7 abundance is a critical node in controlling SV release rate. Indeed, mammalian SYT7 levels are post-transcriptionally modulated by g-secretase proteolytic activity and APP, linking it to SV trafficking defects in Alzheimer's disease (Barthet et al., 2018).

How does SYT7 negatively regulate recruitment and fusion of SVs?
The precise mechanism by which SYT7 reduces release and slows refilling of the RRP is uncertain given it is not enriched at sites of SV fusion. Although we cannot rule out the possibility that a small fraction of the protein is found at AZs, SYT7 is predominantly localized to an internal membrane compartment at the peri-AZ where SV endocytosis and endosomal sorting occurs (Coyle et al., 2004;Koh et al., 2004;Marie et al., 2004;Rodal et al., 2008;Sone et al., 2000). SYT7 membrane tubules are in close proximity and could potentially interact with peri-AZs proteins, endosomes, lysosomes and the plasma membrane. Given its primary biochemical activity is to bind membranes in a Ca 2+ -dependent manner, SYT7 could mediate cargo or lipid movement across multiple compartments within peri-AZs. In addition, it is possible SYT7 tubules could form part of the poorly defined SV recycling endosome compartment. However, we observed no change in SV density or SV localization around AZs, making it unlikely SYT7 would be essential for endosomal trafficking of SVs. The best characterized regulator of the SV endosome compartment in Drosophila is the RAB35 GAP Skywalker (SKY) (Uytterhoeven et al., 2011). Although Sky mutations display some similarities to Syt7, including increased neurotransmitter release and larger RRP size, Syt7 lacks most of the welldescribed Sky phenotypes such as behavioral paralysis, FM1-43 uptake into discrete punctated compartments, cisternal accumulation within terminals and reduced SV density. In addition, we found no co-localization between SKY-GFP and SYT7 RFP within presynaptic terminals.
By blocking SV refilling with bafilomycin, our findings indicate the fast recovery of the RRP can occur via enhanced recruitment from the reserve pool and does not require changes in endocytosis rate. The phosphoprotein Synapsin has been found to maintain the reserve SV pool by tethering vesicles to actin filaments at rest (Akbergenova and Bykhovskaia, 2007;Bykhovskaia, 2011;Hilfiker et al., 1999;Milovanovic and De Camilli, 2017;Shupliakov et al., 2011). Synapsin interacts with the peri-AZ protein Dap160/Intersectin to form a protein network within the peri-AZ that regulates clustering and release of SVs (Gerth et al., 2017;Marie et al., 2004;Winther et al., 2015). Synapsin-mediated phase separation is also implicated in clustering SVs near release sites (Milovanovic et al., 2018;Milovanovic and De Camilli, 2017). SYT7 could similarly maintain a subset of SVs in a non-releasable pool and provide a dual mechanism for modulating SV mobilization. Phosphorylation of Synapsin and Ca 2+ activation of SYT7 would allow multiple activity-dependent signals to regulate SV entry into the RRP. As such, SYT7 could play a key role in organizing membrane trafficking and protein interactions within the peri-AZ network by adding a Ca 2+ -dependent regulator of SV recruitment and fusogenicity.
Additional support for a role for SYT7 in regulating SV availability through differential SV sorting comes from recent studies on the SNARE complex binding protein CPX. Analysis of Drosophila Cpx mutants, which have a dramatic increase in minis (Buhl et al., 2013;Huntwork and Littleton, 2007;Jorquera et al., 2012), revealed a segregation of recycling pathways for SVs undergoing spontaneous versus evoked fusion (Sabeva et al., 2017). Under conditions where intracellular Ca 2+ is low and SYT7 is not activated, spontaneously-released SVs do not transit to the reserve pool and rapidly return to the AZ for re-release. In contrast, SVs released during high frequency evoked stimulation when Ca 2+ is elevated and SYT7 is engaged, re-enter the RRP at a much slower rate. This mechanism slows re-entry of SVs back into the releasable pool when stimulation rates are high and large numbers of SV proteins are deposited onto the plasma membrane at the same time, allowing time for endosomal sorting that might be required in these conditions. In contrast, SVs released during spontaneous fusion or at low stimulation rates would likely have less need for endosomal re-sorting. Given SYT7 restricts SV transit into the RRP, it provides an activity-regulated Ca 2+ -triggered switch for redirecting and retaining SVs in a non-fusogenic pool that could facilitate sorting mechanisms.
Beyond SV fusion, presynaptic membrane trafficking is required for multiple signaling pathways important for developmental maturation of NMJs (Harris and Littleton, 2015;McCabe et al., 2003;Packard et al., 2002;Piccioli and Littleton, 2014;Rodal et al., 2011). In addition, alterations in neuronal activity or SV endocytosis can result in synaptic undergrowth or overgrowth (Akbergenova et al., 2018;Budnik et al., 1990;Dickman et al., 2006;Guan et al., 2005;Koh et al., 2004). We did not find any defect in synaptic bouton or AZ number, indicating SYT7 does not participate in membrane trafficking pathways that regulate synaptic growth and maturation. However, a decrease in T-bar area and presynaptic Ca 2+ influx in Syt7 mutants was found. Although it is unclear how these phenotype arise, they may represent a form of homeostatic plasticity downstream of elevated synaptic transmission (Frank et al., 2020). There is also ample evidence that SV distance to Ca 2+ channels plays a key role in defining the kinetics of SV release and the size of the RRP (Bö hme et al., 2016;Chen et al., 2015;Neher, 2015;Neher and Brose, 2018;Wadel et al., 2007), suggesting a change in such coupling in Syt7 mutants might contribute to elevations in P r and RRP refilling. Further studies will be required to precisely define how SYT7 controls the baseline level of SV release at synapses.
Sequence alignment, phylogenetic tree construction and molecular modeling NCBI BLAST was used to identify homologs of SYT1, SYT7 and ESYT-2 in the genomes of C. elegans, C. intestinalis, D. rerio, M. musculus, H. sapiens, R. norvegicus and T. adherens. Jalview was used to align SYT1 and SYT7 protein sequences from D. melanogaster, M. Musculus and H. sapiens with the T-coffee multiple sequence alignment algorithm. Jalview and Matlab were used to generate a phylogenetic tree using BLOSUM62 matrix and neighbor joining clustering. The SWISS model server (https://swissmodel.expasy.org) was used for homology modeling of Drosophila SYT7 from R. norvegicus SYT7 (PBD: 6ANK) (Waterhouse et al., 2018). The PyMOL Molecular Graphics System (Version 2.0 Schrö dinger, LLC) was used to visualize SYT1 and SYT7 protein structures.
Immunoreactive proteins were imaged on either a Zeiss Pascal Confocal (Carl Zeiss Microscopy, Jena, GERMANY) using a 40x or 63X NA 1.3 Plan Neofluar oil immersion objective or a ZEISS LSM 800 microscope with Airyscan using a 63X oil immersion objective. For AZ volume and AZ proximity measurements, samples were imaged on a Zeiss Airyscan microscope and BRP labeling was analyzed in Volocity 6.3.1 software (Quorum Technologies Inc, Puslinch, Ontario, CAN). AZs clusters larger than 0.2 mm 3 were rarely found, but could not be resolved into single objects by the software. To ensure such clusters did not affect AZ size analysis, all AZs larger than 0.2 mm 3 were excluded from the analysis.

Electrophysiology
Postsynaptic currents from the indicated genotypes were recorded from 3 rd instar muscle fiber 6 at segment A3 using two-electrode voltage clamp with a À80 mV holding potential in HL3.1 saline solution (in mM, 70 NaCl, 5 KCl, 10 NaHCO3, 4 MgCl2, five trehalose, 115 sucrose, 5 HEPES, pH 7.2) as previously described (Jorquera et al., 2012). Final [Ca 2+ ] was adjusted to the level indicated in the text. The Ca 2+ cooperativity of release was determined from the slopes of a linear fit of a double logarithmic plot of evoked responses in the linear range (0.175 to 0.4 mM Ca 2+ ). Inward currents recorded during TEVC are labeled as positive values in the figures for simplicity. For experiments using bafilomycin, 4 mm bafilomycin (LC Laboratories, Woburn, MA, USA) was dissolved in dimethyl sulphoxide (DMSO, Sigma, St. Louis, MO, USA) in HL3.1 and bath applied to dissected larvae. DMSO containing HL3.1 was used for control. Data acquisition and analysis was performed using Axoscope 9.0 and Clampfit 9.0 software (Molecular Devices, Sunnyvale, CA, USA). mEJCs were analyzed with Mini Analysis software 6.0.3 (Synaptosoft, Decatur, GA, USA). Motor nerves innervating the musculature were severed and placed into a suction electrode. Action potential stimulation was applied at the indicated frequencies using a programmable stimulator (Master8, AMPI; Jerusalem, Israel).

Fluo-4 AM imaging
Fluo-4 AM (F14201; ThermoFisher) loading was performed as previously described (Dawson-Scully et al., 2000). During incubation, neuronal membranes were labeled with DyLight 649 conjugated anti-HRP at 1:1000 (#123-605-021; Jackson Immuno Research, West Grove, PA, USA). NMJs of Ib motoneurons at muscle 6/7 were identified and motor nerves were stimulated in HL3 saline with 20 mM MgCl 2 and 1.1 mM Ca 2+ for 5 s at 10 Hz for three epochs, each with a 5 s rest period between stimulation. Imaging of Fluo-4 AM fluorescent signal was performed on a Zeiss Axio Imager two equipped with a spinning-disk confocal head (CSU-X1; Yokagawa, JAPAN) and ImagEM X2 EM-CCD camera (Hamamatsu, Hamamatsu City JAPAN). 5 mm stacks from synaptic boutons were imaged at a frame rate of 1.25 Hz and mean Fluo-4 AM fluorescent intensity was determined during the stimulation protocol for each trial.
Optical quantal imaging and P r mapping P r mapping was performed on a Zeiss Axio Imager equipped with a spinning-disk confocal head (CSU-X1; Yokagawa, JAPAN) and ImagEM X2 EM-CCD camera (Hamamatsu, Hamamatsu City JAPAN) as previously described (Akbergenova et al., 2018). Myristoylated-GCaMP6s was expressed in larval muscles with 44H10-LexAp65 (provided by Gerald Rubin). Individual PSDs were visualized by expression of GluRIIA-RFP under its endogenous promoter (provided by Stephan Sigrist). An Olympus LUMFL N 60X objective with a 1.10 NA was used to acquire GCaMP6s imaging data at 8 Hz. 3 rd instar larvae were dissected in Ca 2+ -free HL3 containing 20 mM MgCl 2 . After dissection, preparations were maintained in HL3 with 20 mM MgCl 2 and 1.0 mM Ca 2+ for 5 min. A dual channel multiplane stack was imaged at the beginning of each experiment to identify GluRIIA-positive PSDs. Single focal plane videos were then recorded while motor nerves were stimulated with a suction electrode at 1 Hz. GluRIIA-RFP PSD position was re-imaged every 25 s during experiments. The dual channel stack was merged with single plane images using the max intensity projection algorithm from Volocity 6.3.1 software. The position of all GluRIIA-RFP PSDs was then spliced with the myr-GCaMP6s stimulation video. GluRIIA positive PSDs were detected automatically using the spot finding function of Volocity and equal size ROIs were assigned to the PSD population. In cases where the software failed to label visible GluRIIA-RFP PSDs, ROIs were added manually. GCaMP6s peak flashes were then detected and assigned to ROIs based on centroid proximity. The time and location of Ca 2+ events were imported into Excel or Matlab for further analysis. Observed GCaMP events per ROI were divided by stimulation number to calculate AZ P r . FM1-43 uptake and release assays 3 rd instar wandering larvae were dissected in Ca 2+ -free HL3.1 and axons were severed from the CNS. Axon bundles were stimulated with a suction electrode in 1.5 mM CaCl 2 HL3.1 solution containing 2 mM of the lipophilic dye FM 1-43FX (F35355; Thermo Fisher Scientific, Waltham, MA, USA). Dye loading was performed at 10 Hz for 50 s (500 events) or at 0.5 Hz for 300 s (150 events), 600 s (300 events) and 900 s (600 events) as indicated. After stimulation, samples were washed for 2 min in Ca 2+ free HL3.1 containing 100 mM Advacep-7 (Sigma; A3723) to help remove non-internalized FM 1-43 dye. Image stacks from muscle 6/7 at segment A3 were obtained using a spinning disk confocal microscope. FM1-43 unloading was done with a high K + (90 mM) HL3.1 solution for 1 min, followed by washing in a Ca 2+ free HL3.1 solution for 1 min. An image stack at segment A3 muscle 6-7 was obtained on a Zeiss Axio Imager two equipped with a spinning-disk confocal head with a 63X water immersion objective. Mean FM1-43 intensity at the NMJ was quantified using the Volocity 3D Image Analysis software (Quorum Technologies Inc, Puslinch, Ontario, CAN).

Electron microscopy
Syt1 M1 and control 3 rd instar larvae were dissected in Ca 2+ -free HL3.1 solution and fixed in 1% glutaraldehyde, 4% formaldehyde, and 0.1 m sodium cacodylate for 10 min at room temperature as previously described (Akbergenova and Bykhovskaia, 2009). Fresh fixative was added and samples were microwaved in a BioWave Pro Pelco (Ted Pella, Inc, Redding, CA, USA) with the following protocol: (1) 100W 1 min, (2) 1 min off, (3) 100W 1 min, (4) 300W 20 secs, (5) 20 secs off, (6) 300W 20 secs. Steps 4-6 were repeated twice more. Samples were then incubated for 30 min at room temperature with fixative. After washing in 0.1 M sodium cacodylate and 0.1 M sucrose, samples were stained for 30 min in 1% osmium tetroxide and 1.5% potassium ferrocyanide in 0.1 M sodium cacodylate solution. After washing with 0.1 M sodium cacodylate, samples were stained for 30 mins in 2% uranyl acetate and dehydrated through a graded series of ethanol and acetone, before embedding in epoxy resin (Embed 812; Electron Microscopy Sciences). Thin sections (50-60 nm) were collected on Formvar/carbon-coated copper slot grids and contrasted with lead citrate. Sections were imaged at 49,000 Â magnification at 80 kV with a Tecnai G2 electron microscope (FEI, Hillsboro, OR, USA) equipped with a charge-coupled device camera (Advanced Microscopy Techniques, Woburn, MA, USA). Type Ib boutons at muscle 6/7 were analyzed. All data analyses were done blinded.
For SV counting, T-bars at Ib boutons were identified and a FIJI macro was used to draw four concentric circles with 100 nm, 200 nm, 300 nm or 400 nm radius. The concentric circles were drawn with the T-bar at the center. To quantify vesicle density, FIJI was used to measure the area of the bouton and quantify the total number of vesicles within it. Final analysis was performed in Matlab and Excel.

Co-localization analysis and 3D reconstruction
The JaCOP FIJI algorithm (Bolte and Cordelières, 2006) was used to obtain cytofluorogram plots of bouton image stacks that were probed for RFP and a 2 nd labeled compartment in SYT7 RFP 3 rd instar larvae. Automatic thresholding was used to identify pixels above background for both channels. To obtain an average Pearson correlation, cytofluorograms from boutons obtained from three animals were analyzed in Matlab. All data analyseis were done blinded. 3D reconstruction was performed using the 3D Viewer plugin in FIJI (Schmid et al., 2010). The bouton stack was displayed as a surface and labeled with SYT7 RFP in magenta and HRP in black.

Statistical analysis
Statistical analysis and graphing was performed with either Origin Software (OriginLab Corporation, Northampton, MA, USA) or GraphPad Prism (San Diego, CA, USA). Statistical significance was determined using specific tests as indicated in the text. Appropriate sample size was determined using GraphPad Statmate. Asterisks denote p-values of: *, p 0.05; **, p 0.01; and ***, p 0.001. All histograms and measurements are shown as mean ± SEM.