Fast resupply of synaptic vesicles requires Synaptotagmin-3

Sustained neuronal activity demands quick resupply of synaptic vesicles in order to maintain reliable synaptic transmission. Such vesicle replenishment is accelerated by sub-micromolar presynaptic Ca2+ signals by an as yet unidentified high-affinity Ca2+ sensor1-4. Here we identify a novel presynaptic role for the high-affinity Ca2+ sensor Synaptotagmin-3 (SYT3)5 in driving vesicle replenishment and short-term synaptic plasticity. Synapses in Syt3 knockout mice exhibit enhanced short-term depression, and recovery is slower and insensitive to presynaptic residual Ca2+. During sustained neuronal firing, SYT3 speeds vesicle replenishment and increases the size of the readily releasable pool of vesicles. SYT3 also mediates a second form of short-term enhancement called facilitation, under conditions of low vesicle release probability. Models of vesicle trafficking suggest that SYT3 could combat synaptic depression by accelerating vesicle docking at active zones. Our results reveal a critical role for presynaptic SYT3 in maintaining reliable high-frequency synaptic transmission in neural circuits.


Main
Synaptic transmission faces a problem of resource scarcity: neurotransmitter release is limited by the size of the readily releasable pool of vesicles (RRP), which is rapidly depleted by synaptic activity. Vesicle depletion decreases neurotransmitter release, and the fidelity of neuronal signaling. To overcome this problem, a form of short-term plasticity termed Ca 2+ -dependent recovery from depression (CDR) accelerates vesicle replenishment as synaptic activity increases [1][2][3][4] . Although much is understood concerning the mechanisms that traffic and dock vesicles at release sites, the Ca 2+ sensor for CDR has not been identified. Multiple Ca 2+ -sensing synaptotagmin isoforms regulate vesicle trafficking and fusion 6 . The low-affinity vesicular synaptotagmins SYT1, 2 and 9 serve well-established roles in synchronous vesicle fusion 7,8 , but the functions of other SYT isoforms remain unclear. The high-affinity Ca 2+ sensor SYT3 has long been hypothesized to play a role in neurotransmitter release 9 , but SYT3 does not support synaptic transmission from Syt1 knockout neurons, and thus is not likely to directly drive vesicle fusion 8 . However, SYT3 binds to SNAREs 10,11 , stimulates SNARE-mediated membrane fusion in vitro 12 , and participates in insulin secretion from pancreatic beta cells 13,14 . These observations led to the hypothesis that SYT3 senses sub-micromolar Ca 2+ signals to drive short-term plasticity 5,9 . Despite its neuronal abundance and putative roles in neurotransmitter release, SYT3 has no known presynaptic function. Instead, previous studies reported that postsynaptic SYT3 mediates AMPA receptor internalization and longterm depression 15,16 .

SYT3 localizes to both presynaptic and postsynaptic compartments
Syt3 is expressed broadly throughout the brain, but enriched in the brainstem, cerebellum, and hippocampus 17,18 . Consistent with this expression pattern, we observed strong immunolabeling for SYT3 in the brainstem and cerebellum ( Fig. 1a-b). Labeling was absent in Syt3 knockout (KO) mice, confirming antibody specificity (Extended Data Fig. 1). In the brainstem, SYT3 immunolabeling was present in the medial nucleus of the trapezoid body (MNTB) in the region of calyx of Held nerve terminals, and overlapped with the presynaptic marker VGLUT1 (Fig 1a). Consistent with presynaptic expression of SYT3 at the calyx of Held, Syt3 mRNA is present in bushy cells in the ventral cochlear nucleus 19 . In the cerebellum, SYT3 was abundant in the molecular layer ( Fig. 1b) where it overlapped with VGLUT1 and VGLUT2, markers for cerebellar parallel fibers and climbing fibers, respectively. However, we could not resolve whether SYT3 immunolabeling is presynaptic or postsynaptic using optical microscopy in brain slices. We next generated synaptosomes from the cerebellum and brainstem (Fig. 1c), and labelled them with antibodies against SYT3, VGLUT1, and the postsynaptic marker PSD-95 (Fig. 1d). Although synaptosomes can contain the presynaptic terminal and postsynaptic fragments 20 , SYT3 labeling was present in nearly all synaptosomes that contained both VGLUT1 and PSD-95, and present in a larger fraction of synaptosomes that were positive for only VGLUT1 compared to only PSD-95, suggesting SYT3 may be enriched in presynaptic compartments (Fig. 1e). We also biochemically subfractionated synaptosomes 21 in order to enrich for presynaptic and postsynaptic compartments (Fig. 1f). Subsequent western blot analysis detected SYT3 in fractions enriched for both compartments (Fig. 1g). These results indicate that SYT3 is present in presynaptic, as well as postsynaptic compartments.

Presynaptic SYT3 is required for Ca 2+ -dependent recovery from depression
The calyx of Held is a powerful one-to-one glutamatergic synapse that is highly amenable to quantitative studies of synaptic transmission 22,23 . To measure neurotransmitter release at the calyx of Held, we performed whole-cell voltage-clamp recordings in MTNB neurons in acute brain slices from adult (P30-P60) animals (Fig.  2a). Spontaneous excitatory postsynaptic currents (sEPSCs) had similar amplitudes in wildtype (WT) and Syt3 knockout animals (Extended Data Fig. 2), indicating that the loss of SYT3 does not affect postsynaptic receptor density under basal conditions. sEPSC frequency was similar in both genotypes (Extended Data Fig. 2e), suggesting that SYT3 does not affect initial release probability. EPSCs evoked by electrically stimulating presynaptic axons had similar amplitudes and kinetics in WT and Syt3 knockouts, also indicating that SYT3 does not contribute to basal release properties (Extended Data Fig. 2f-g).
To probe short-term plasticity, we stimulated calyces with trains of 100 stimuli. During 1 Hz stimulation, both genotypes depressed mildly to the same steady-state amplitudes. However, at frequencies ≥10 Hz WT synapses depressed far less than Syt3 knockout synapses (Fig. 2b). As a result, charge transfer (integrated steady-state EPSC amplitudes per second) increased almost linearly as a function of frequency from 10-200 Hz in WT, but not in Syt3 knockouts (Fig. 2c). Enhanced synaptic depression in Syt3 knockout animals could be the indirect result of changes to presynaptic Ca 2+ channel expression, or function. Larger presynaptic Ca 2+ currents would raise the initial probability of release and accelerate depletion of the RRP, while use-dependent inactivation of Ca 2+ channels could reduce release probability during train stimulation. We assessed these possibilities by performing voltage-clamp recordings from presynaptic calyces of young animals (P11-13) 24 . We found that both the amplitudes and use-dependence of Ca 2+ currents induced by 1 ms depolarizations at 100 Hz were unaltered in KO calyces (Extended Data Fig. 2h-k). Taken together, these data suggest that SYT3 is required to maintain release during high-frequency firing.
At the calyx of Held, CDR plays a critical role in accelerating vesicle replenishment and maintaining neurotransmitter release during high-frequency stimulation. The defining characteristic of CDR is that recovery from depression is slowed when the exogenous Ca 2+ buffer EGTA is introduced into presynaptic terminals 1,2 . We probed recovery from depression at varying intervals after inducing depression with 100 stimuli at 200 Hz. Syt3 knockout synapses recovered with a time constant ~2-fold slower than WT synapses ( Fig. 2e-g). Bath application of 100 µM EGTA-AM slowed recovery in WT animals to the degree observed in Syt3 knockouts. In contrast, EGTA-AM had no effect on recovery in Syt3 knockouts (Fig. 2f,g). This indicates that SYT3 is required for CDR and sustained neurotransmitter release at the calyx of Held.
To determine whether SYT3 affects CDR by acting presynaptically in a cell-autonomous manner, we rescued presynaptic Syt3 expression in knockout animals by injecting a bicistronic adeno-associated viral (AAV) vector into the ventral cochlear nucleus to express SYT3 and GFP. 4-5 weeks after injection, a subset of transduced calyces could be visually identified by GFP fluorescence, and post-hoc immunohistochemistry confirmed the presence of SYT3 in GFP-positive presynaptic terminals (Fig. 3b). GFP/SYT3-positive synapses exhibited reduced depression and a faster recovery, comparable to WT synapses. GFP/SYT3-negative calyces exhibited high-frequency depression, and slow recovery, similar to non-injected Syt3 knockouts (Fig. 3c-e). To determine whether Ca 2+ binding by SYT3 is required for CDR, we expressed a Ca 2+ -binding deficient mutant SYT3 (Syt3 D/N ). Syt3 D/N expression failed to ameliorate depression or speed recovery from depression. Taken together, these data suggest that CDR requires Ca 2+ binding to presynaptic SYT3.

Syt3 increases the readily releasable pool of vesicles
Presynaptic Ca 2+ increases both the rate of vesicle replenishment 1 , and the size of the RRP 25-27 . To determine whether SYT3 affects the RRP through its role in vesicle replenishment, we used EPSCs driven by 200 Hz train stimulation to calculate RRP size. We applied multiple, independent methods to estimate RRP size: backextrapolation of cumulative release corrected to account for vesicle replenishment early in the train 27 , a forwardextrapolation method well-suited to estimate the RRP at high frequencies 28 , and an exponential fit of the initial depression within a train 29 . Across all the analysis methods, RRP was significantly increased in WT synapses compared to Syt3 knockout synapses (Extended Data Fig. 3).
Synaptic ultrastructure suggests that vesicles at the active zone exist in two states: tightly docked and lying within < 2nm of the plasma membrane, or a loosely docked state further from the membrane (~5-10 nm) 30 . Tightly docked vesicles are hypothesized to constitute much of the RRP, but loosely docked vesicles may reversibly transition to tight docking and contribute to the RRP 31 . To explore how SYT3 could produce the synaptic properties we observed at the calyx of Held, we constructed models of vesicle docking and fusion. In these models, vesicles are first recruited to a loosely docked state, from which they can reversibly transition to a tightly docked state. Loosely docked vesicles could be refilled by an infinite reserve pool. To minimize the number of free parameters, we purposefully chose the simplest scenario, where only tightly docked vesicles could release, with a uniform release probability. We fit the model to EPSC trains at all frequencies, and recovery from depression using Pearson's χ 2 test. After first fitting the model to Syt3 knockout data, we modeled WT synapses by introducing SYT3 as a single, activity-dependent parameter (Extended Data Fig. 4). Models in which SYT3 modulated vesicle release probability of tightly docked vesicles failed to reproduce the behavior of WT synapses. Instead, we found that WT depression and recovery was best fit by a model in which SYT3 increased the transition rate from loose to tight docked states (Extended Data Fig. 4i-j). Such a model is supported by recent evidence for activity-dependent, transient docking of vesicles 32 . In summary, while there is debate as to the identity of vesicles that constitute the RRP, these results show that more vesicles are releasable when SYT3 is present in presynaptic terminals.

Syt3 accelerates vesicle replenishment at multiple synapses
CDR is also prominent in cerebellar climbing fibers, a powerful glutamatergic synapse onto Purkinje cells 2 . Our immunohistochemistry suggested colocalization of SYT3 with VGLUT2, the presynaptic marker for climbing fibers (Fig. 1b), and Syt3 RNA is expressed in the inferior olive where climbing fibers originate 33 . We performed voltage-clamp recordings from Purkinje cells while electrically stimulating climbing fiber axons in the granule cell layer (Fig. 4a). As with the calyx of Held, we observed no difference in the amplitudes or kinetics of EPSCs between WT and Syt3 knockout climbing fibers, suggesting that SYT3 does not contribute to initial release (Fig. 4b-c). Climbing fibers have a high initial release probability, such that a single stimulus significantly depletes the RRP and causes synaptic depression 2 . CDR can therefore be evaluated by probing recovery at varying interstimulus intervals (Δt) after a single, depressing stimulus. Climbing fibers in Syt3 knockouts exhibited significantly greater depression and recovered ~2-fold slower than WT synapses ( Fig. 4ef,i). Bath application of 100 µM EGTA-AM slowed recovery in WT animals to the degree observed in Syt3 knockouts. EGTA-AM had no effect on recovery in knockouts ( Fig. 4g-i). Another synapse where recovery from depression is slowed by EGTA-AM is the cerebellar mossy fiber-to-granule cell terminal 4 . To test whether SYT3 accelerates vesicle replenishment in mossy fibers, we performed voltage-clamp recordings from cerebellar granule cells and induced synaptic depression with trains of 200 Hz stimulation and probed the time course of recovery of EPSCs (Extended Data Fig. 5). Similar to calyx of Held and climbing fiber synapses, the time-course of recovery from depression was significantly slower in mossy fibers from Syt3 knockouts (WT, τW = 0.66 ± 0.19 s; Syt3 KO, τW = 2.3 ± 0.59 s). Taken together, these results demonstrate that SYT3 is required for CDR and fast vesicle replenishment in multiple brain regions, at synapses with widely varying release properties.

Syt3 underlies multiple forms of short-term plasticity
When initial release probability is lowered by decreasing the concentration of extracellular Ca 2+ ([Ca 2+ ]e), cerebellar climbing fibers and the calyx of Held exhibit a second form of short-term plasticity known as facilitation 34,35 . At mammalian synapses, facilitation is primarily driven by SYT7, another high-affinity synaptotagmin isoform 36,37 . However, low [Ca 2+ ]e facilitation persists in Syt7 knockout animals at both the calyx of Held 38 and climbing fibers 39 . To test the possibility that SYT3 mediates this unexplained form of facilitation, we probed short-term plasticity in low [Ca 2+ ]e. As previously reported, WT climbing fibers and calyces exhibited short-lived paired-pulse facilitation ( Fig. 5a-c, g-i). However, low [Ca 2+ ]e facilitation was completely absent in Syt3 knockouts. During train stimulation, EPSCs facilitated throughout the train in WT, but not KO synapses ( Fig. 5d-f, j-l). This results is somewhat surprising, because CDR and facilitation have traditionally been associated with different mechanisms: CDR is thought to result from accelerated resupply of vesicles, whereas facilitation has been attributed to increased vesicle release probability 9,40 (although see 30 ). However, the role of SYT3 in both CDR and facilitation suggests that both forms of short-term plasticity could manifest from the same underlying steps in vesicle trafficking.

Discussion
Despite decades of study, most synaptotagmin isoforms lack well-defined functions. Here we identify the first role for SYT3 in presynaptic terminals, in accelerating vesicle resupply to sustain robust synaptic transmission. Our results help to explain the observed Ca 2+ -dependence of vesicle replenishment 3 , recovery from depression 1,2 , and the size of the RRP 27 . Genetic ablation of Syt3 lead to slow, Ca 2+ -insensitive vesicle replenishment, without affecting basal release properties or presynaptic Ca 2+ currents. Our data are best fit by a biophysical model where Ca 2+ -bound SYT3 accelerates vesicle docking. SYT3 differs from other synaptotagmin isoforms in its Ca 2+ -binding properties: It has a high affinity for Ca 2+ , making it well-suited to sense the sub-micromolar Ca 2+ signals that drive CDR and other forms of short-term plasticity 5 . However, following a drop in Ca 2+ , SYT3 dissociates from lipid membranes with fast kinetics similar to the low-affinity isoforms SYT1 and 2 41 . Thus, SYT3 could respond rapidly to changing Ca 2+ levels in order to match vesicle resupply to the demands of ongoing synaptic transmission.
Syt3 is expressed broadly, but heterogeneously, in the mammalian brain 17,33 . In this study we focused on multiple excitatory synapses where Ca 2+ -dependent vesicle replenishment has been well-characterized. Loss of Syt3 produced similar deficits in synapses despite their widely varied release probabilities, morphology, and in vivo firing patterns, indicating a conserved role in vesicle replenishment. Syt3 RNA is enriched in inhibitory neurons 42 , suggesting that SYT3 may also sustain the release of GABA, and thereby affect the balance of excitation and inhibition in neural circuits. Consistent with this possibility, Syt3 mutations have been reported in patients with epilepsy 43 and autism spectrum disorder 44 . In future studies, cell type-specific removal of Syt3 could reveal how activity-dependent vesicle replenishment contributes to neural circuit function, and neuropsychiatric disease.

Methods
No statistical methods were used to predetermine sample size, and experiments were not randomized. Initial electrophysiological recordings at all synapses were performed blind to genotype, but blinding was abandoned after genotypes could be readily identified by physiology. No other blinding was performed.

Animals and viruses.
All mice were handled in accordance with NIH guidelines and protocols approved by Oregon Health & Science University's Institutional Animal Care and Use Committee. Syt3 KO mice were generously provided by Dr. Camin Dean. KO and WT animals of either sex from homozygous or heterozygous breeding pairs (littermates) were used. Viral plasmids were obtained from GenScript with permission from Dr. Camin Dean. AAV9-EGFP-p2A-Syt3 WT and AAV9-EGFP-p2A-Syt3 D/N were produced by the Molecular Virology Core at Oregon Health & Science University. For rescue experiments, 500-900 nl of AAV was injected into P2 animals as previously described 45 . Briefly, animals were anesthetized hypothermically, and virus was injected unilaterally into the left cochlear nucleus. Injection coordinates were (from lambda): 1.4 mm lateral, 2.7 mm caudal, 2.9 mm ventral.

Electrophysiology.
Recordings were performed at 34±0.5 °C, except low [Ca 2+ ]e and presynaptic recordings, which were performed at room temperature to aid comparison with previous studies 34,35 . Slices were continuously perfused with oxygenated ACSF at ~3 ml/min. To isolate AMPAR-driven responses, ACSF contained (in µM): 100 Picrotoxin, 5 CGP, 1 AM-251, 20 CPP. For MNTB recordings, 0.5 µM Strychnine was added. For calyx of Held and climbing fiber recordings in 2 mM [Ca 2+ ]e, 1 mM of kynurenic acid was added to mitigate AMPA receptor saturation 22,46 . Whole-cell recordings were obtained using a EPC10/2 patch-clamp amplifier (HEKA Elektronik). For sEPSC recordings in the MNTB, health of calyces was verified with brief train stimulation. For AAV rescue experiments, GFP-positive calyces were visualized using a Thorlabs M470L2 LED and a Carl Zeiss GFP filter set. For EGTA-AM experiments, 100µM of EGTA-AM was washed onto slices for 15 min and washed-out for at least 10 minutes prior to recording. MNTB neurons were patched with borosilicate glass electrodes (1.5-2.5 MΩ). Cells were voltage-clamped at -70 mV, and bipolar platinum/iridium stimulation electrodes were placed contralaterally around the afferent fiber tract close to the midline. Purkinje cells were patched with 1.2-2 MΩ pipettes, held at -40 mV to inactivate voltage-gated sodium channels, and ACSF-filled monopolar glass stimulation electrodes (2-3 MΩ) were placed in the granule cell layer 50-100 µm from the targeted cell. To ensure no contamination from parallel fiber inputs, stimulation intensity was gradually increased to produce an all-or-none climbing fiber response. For mossy fiber recordings, cerebellar granule cells were patched with 6-8 MΩ pipettes, held at -70 mV and inputs were stimulated in the white matter 100 µm from the targeted cell with monopolar glass electrodes. For MNTB and granule cell recordings, the pipette solution contained (in mM): 130 Cs-gluconate, 10 CsCl, 5 EGTA, 10 HEPES, 10 TEA-Cl, 5 Na-phosphocreatine, 4 Mg-ATP, 0.5 Na-GTP, 3 QX-314. For Purkinje cell recordings, the pipette solution contained (in mM): 35 CsF, 100 CsCl, 10 EGTA, 10 HEPES, 5 QX-314. Presynaptic recordings in P11-13 calyces were performed using 7-9 MΩ pipettes containing (in mM): 130 Cs-gluconate, 15 CsCl, 20 TEA-Cl, 5 Na2-phosphocreatine, 10 HEPES, 0.2 EGTA, 4 ATP-Mg, and 0.5 GTP-Na., and 1 µM TTX, 10 mM TEA, and 100 µM 4-AP were added to ACSF to isolate Ca 2+ -currents. Calyces were held at -80 mV and 1 ms steps to 0 mV were used to open voltage-gated Ca 2+ -channels and elicit tail currents 24 . Data were acquired with a sampling rate of 50 kHz, low-pass Bessel-filtered at 2.9 kHz, and liquid junction potential was corrected. In MNTB, calyx of Held, and Purkinje cell recordings, series resistance was compensated by at least 80%.
Data in figures represent mean ± S.E.M.. Paired-pulse responses were fit exponentially, and train responses were fit double-exponentially in IgorPro 8 (Wavemetrics). Stimulation inter-train intervals in all synapses were 30 s. Unless otherwise stated, statistical significance was assessed by initial one-way ANOVA if more than 2 samples were compared. Normal distribution was tested using a single-sample Kolmogorov-Smirnov test. Normally distributed samples were tested for homoscedasticity and compared using unpaired two-tailed Student's t-tests. If one or both samples were not normally distributed, Mann-Whitney u-tests were used instead. All statistical analyses were performed using Microsoft Excel or IgorPro 8.

Modeling
Biophysical models of synaptic vesicle trafficking and fusion were based on previous studies 4, 47 . For simplicity, release was driven by a single pool of tightly docked vesicles (N td ), which was refilled from a fusionincompetent loosely docked state (N ld ) 30 . Loosely docked vesicles were filled from an infinite reserve pool at a rate k1, and the maximal refill rate was inversely proportional to the fill-state of N ld . Loosely docked vesicles could return to the reserve pool at a rate k-1. N ld vesicles matured to N td at a rate of k2, inversely proportional to the fill-state of N td , and return to N ld with the rate k-2 inversely proportional to its fill-state. Upon stimulation, N td vesicles were released with a uniform release probability p. Transition rates between loose and tight docking were described by: 20 s of inactivity were simulated before each stimulation pattern to allow fill-states to reach equilibrium. In the model where SYT3 drove a transient increase in release probability, p was described by: In the models where SYT3 transiently increased k1 or k2, they were described by: Best fits to experimental KO data were performed using Pearson's χ 2 tests across all frequency trains (1 Hz, 10 Hz, 20 Hz, 50 Hz, 100 Hz, 200 Hz) and recovery from depression after 200 Hz. These parameters were then fixed, and WT experimental data were subsequently fit by introducing parameters representing a putative SYT3 contribution. Baseline sampling rate was 1 ms, and lowering it to 100 µs only marginally changed the best fit parameters. All modeling and fitting were performed with custom IgorPro 8 functions.

Synaptosome preparation
6 mice were euthanized, and the brainstem and cerebellum were dissected in ice-cold ACSF. Pooled tissue from each brain region was suspended in 6 ml of 0.32 M sucrose homogenization buffer (4 mM HEPES, 0.1 mM CaCl2, 1 mM MgCl2, plus Roche protease inhibitor tablet) and homogenized with a Teflon homogenizer using 10 strokes at 900 rpm. Nuclear material was removed by centrifuging at 1000 g for 10 mins at 4 °C. The supernatant was collected and centrifuged at 17,500 g for 15 min at 4 °C. The pellet was suspended in 2.5 ml of homogenization buffer, and homogenized with manually with 6 strokes. The homogenate was brought to 1.25 M sucrose by addition of 12 ml of 2 M sucrose homogenization buffer, and 5 ml of 0.1 mM CaCl2, then divided into 2 ultracentrifuge tubes and carefully overlaid with 2.5 ml of 1 M sucrose homogenization buffer. The gradient was centrifuged at 100,000 g for 3 h at 4 °C. The hazy layer of synaptosomes was collected from the 1/1.25 M sucrose interphase, and diluted 1:10 with 0.32 M sucrose homogenization buffer, and centrifuged at 20,000 g for 30 min at 4 °C. The pellet (synaptosomes) was resuspended with 2 ml of 0.32 M sucrose homogenization buffer. For western blots, 100 µl of synaptosomes were centrifuged at 11,000 g for 11 mins, and the pellet resuspended in 5% SDS. 50 µl of aliquoted synaptosomes were stored at -80 ⁰C for immunostaining, and the remainder was used for biochemical fractionation.
Biochemical fractionation of synaptosomes was performed as previously described 21 with minor modifications. Purified synaptosomes were lysed osmotically by addition of 5 ml of 0.1 mM CaCl2 and homogenized with 3 strokes at 2000 rpm. 5 ml 20 mM MES (pH 6.0) was added and the mixture was incubated on ice for 30 min, then centrifuged at 40,000 g for 30 min. The resulting supernatant, considered to be the vesicle-rich fraction, was acetone precipitated and centrifuged at 18,000 g at 4 °C for 30 min to collect synaptic vesicle proteins. The pellet, containing presynaptic and postsynaptic membranes, was washed with 20 mM Tris-HCl, pH 8.0, 1% Triton X-100, 0.1 mM CaCl2, resuspended and centrifuged at 201,000 g to collect the postsynaptic fraction. The supernatant was acetone precipitated overnight to obtain the presynaptic fraction. Samples were stored at -80 °C until processed for western blotting.

Extended data
Extended Data Fig. 1 48 . d) Best fit of the model shown in c (black) to data (gray) for WT calyces of Held. e) A model where SYT3 increases p immediately after each action potential, after which p decreases exponentially back to its baseline value. f) Best fit of the model shown in e to WT data. g) A model where SYT3 transiently increases k1. After each action potential k1 increases by a fixed amount, after which k1 decreases exponentially back to its baseline value. h) Best fit of the model shown in g to WT data. i) A model where SYT3 transiently increases k2. After each action potential k2 increases by a fixed amount, after which k2 decreases exponentially back to its baseline value. j) Best fit of the model shown in i to WT data. Note that this model produced the lowest χ 2 values among all models tested, supporting a scenario where SYT3 accelerates the transition of vesicles to a tightly docked state.
Extended Data Figure