Synaptotagmin-1 enables frequency coding by suppressing asynchronous release in a temperature dependent manner

To support frequency-coded information transfer, mammalian synapses tightly synchronize neurotransmitter release to action potentials (APs). However, release desynchronizes during AP trains, especially at room temperature. Here we show that suppression of asynchronous release by Synaptotagmin-1 (Syt1), but not release triggering, is highly temperature sensitive, and enhances synchronous release during high-frequency stimulation. In Syt1-deficient synapses, asynchronous release increased with temperature, opposite to wildtype synapses. Mutations in Syt1 C2B-domain polybasic stretch (Syt1 K326Q,K327Q,K331Q) did not affect synchronization during sustained activity, while the previously observed reduced synchronous response to a single AP was confirmed. However, an inflexible linker between the C2-domains (Syt1 9Pro) reduced suppression, without affecting synchronous release upon a single AP. Syt1 9Pro expressing synapses showed impaired synchronization during AP trains, which was rescued by buffering global Ca2+ to prevent asynchronous release. Hence, frequency coding relies on Syt1’s temperature sensitive suppression of asynchronous release, an aspect distinct from its known vesicle recruitment and triggering functions.

www.nature.com/scientificreports www.nature.com/scientificreports/ vesicle recruitment and docking [25][26][27][28] , as well as regulating vesicle endocytosis after vesicle release [29][30][31][32][33] . However, beyond promoting synchronous release and recruiting vesicles, it is well established that, at rest, Syt1, or its functionally-related paralog Syt2, exerts an inhibitory or 'clamping' effect on spontaneous release [34][35][36] and suppresses the asynchronous release component that follows the synchronous spike [37][38][39] . These two phenomena have been related to inhibition of a second Ca 2+ -sensor by Syt1, representing a common mechanism for both 39,40 . Although little is currently known about Syt1's release inhibition during high-frequency activity, given the competition between synchronous and asynchronous release, directed suppression of asynchronous release by Syt1's inhibitory function might play an important role in maintaining synchronous release during sustained high-frequency stimulation.
In this study, we investigate the mechanisms underlying temperature-dependent synchronization of neurotransmission during high-frequency stimulation in hippocampal autaptic excitatory neurons. Our results suggest that increased synchronization of presynaptic Ca 2+ cannot explain increased synchronous release at the end of high-frequency train stimulation. Instead, we characterize a temperature sensitive, inhibitory function of Syt1 acting in maintaining synchronous release during high-frequency activity. We conclude that besides the previously reported triggering and vesicle recruitment functions of Syt1, suppression of asynchronous release is an additional mechanism through which Syt1 synchronizes release. This function appears essential for sustaining frequency coded signalling, thus allowing neurons to transmit information over a broad frequency bandwidth.

Results
Frequency-coding is highly temperature sensitive. At rest or during low-frequencies, synaptic responses are synchronized to APs and relatively insensitive to temperature changes, whereas they rapidly desynchronize during high-frequency stimulation at room temperature (RT). To assess the effect of temperature on frequency-coded information transfer, we compared synaptic responses to a 150-AP train with variable frequencies (40, 20 and 10 Hz) at 22.0 ± 1 °C and 32.5 ± 1 °C (hereafter 22 °C and 32 °C, respectively) in single hippocampal neurons cultured on glia micro-islands (autapses) 41 . With a threshold for post-synaptic spike detection at 30% amplitude of the first evoked response, neurons at 22 °C failed to transmit after the first three pulses. The remaining neurotransmission was dominated by asynchronous release, and homogeneous at different frequencies, constituting a loss of information transfer (Fig. 1a,b). However, when measured at 32 °C, synchronous responses persisted (Fig. 1c), clearly marking the different stimulation frequencies (Fig. 1d). A similar result was reached with a lower threshold at 10%, or a higher threshold at 50% of the first evoked response (Supplementary Fig. S1). Thus, a 10 °C change in temperature critically changes the efficacy of frequency-coding in hippocampal synapses. www.nature.com/scientificreports www.nature.com/scientificreports/  21 pC, n = 9; 32 °C: 28.54 ± 6.83 pC, n = 9), and synchronous (22 °C: 1.135 ± 0.293 pC, n = 9; 32 °C: 13.27 ± 5.134 Hz, n = 9) and asynchronous www.nature.com/scientificreports www.nature.com/scientificreports/ Synchronous responses to single action potentials are not temperature sensitive. We investigated which aspects of synaptic transmission were temperature sensitive. We measured the effect of temperature on spontaneous synaptic release and observed a more than 2-fold higher rate of miniature excitatory post-synaptic currents (mEPSCs) at 32 °C (Fig. 2a,b). Furthermore, mEPSC amplitudes were increased at 32 °C (Fig. 2c), while mEPSC charge remained unchanged (Fig. 2d). Excitatory post-synaptic currents (EPSCs) after a single AP remained synchronous, both at 22 °C and 32 °C (Fig. 2e). In fact, more charge was transferred at 22 °C (Fig. 2f), in line with previous observations 14 , although the amplitude remained unchanged (data not shown). We conclude that synchronous responses evoked by a single AP are temperature insensitive, while spontaneous release is inhibited at 22 °C.
Raising temperature increases late-train synchronous release at the cost of asynchronous release. Synchronization of synaptic responses is most likely due to synchronization of the presynaptic vesicle release process. One possible cause for the temperature dependent loss of synchronous responses during high frequency stimulation is increased competition from asynchronous release for the releasable pool of vesicles 22 . To investigate this, we sought an objective way of estimating synchronous and asynchronous release during high-frequency stimulation, under the assumption that most asynchronous release is blocked by buffering global Ca 2+ 22,42,43 . We compared synaptic responses, scaled to their peak amplitude, before (naive) and after application of the slow Ca 2+ buffer EGTA-AM, and noticed that asynchronous release occurred during the decay phase of the EPSC (green area, Supplementary Fig. S2a). Contribution of NMDA currents to this component were considered negligible due to high (4 mM) extracellular Mg 2+ . We determined for each cell the "synchronous time window", defined as the time interval for which the asynchronous release within the interval (green shaded area, Supplementary Fig. S2a) balances the synchronous release outside the interval (red shaded area, Supplementary  Fig. 2a), and found a median value of 22.65 ± 5.65 ms ( Supplementary Fig. S2b, n = 18). To prevent underestimation of the synchronous component and overestimation of the asynchronous component, we set the synchronous time window for our analysis to 25 ms. The standing current due to asynchronous release during high-frequency stimulation, which persists beyond the length of an individual pulse, was subtracted at the start of each pulse before calculating the synchronous component ( Supplementary Fig. S2c). While we noticed that in the presence of EGTA-AM the EPSC width did not change considerably throughout the train, it is possible that the asynchronous contribution within the synchronous window is higher at the end of the train. This would slightly overestimate synchronous release but prevents erroneous conclusions about release desynchronization.
Quantifying synchronous release during 100AP 40 Hz train stimulation at 22 °C we observed that synchronous release rapidly decayed, leaving asynchronous as the dominant form of release (Fig. 2g,h). Synchronous release at the end of the train stimulation (late-train synchronous release), quantified by averaging the charge of the last 20 responses in the train, was negligible (Fig. 2k). In contrast, at 32 °C synchronous release persisted until the end of the train (Fig. 2g,i,k). This more than 10-fold increase was specific for synchronous release as total release only doubled and asynchronous release decreased ( Fig. 2h-k). Expressed as a fraction of total release, the synchronous component (late-train synchronous fraction) was increased almost 7 fold (Fig. 2l). In addition to decreased asynchronous release during late-train stimulation, asynchronous release after the stimulation train (tail release), decayed much faster at 32 °C, as determined by a mono-exponential fit (Fig. 2m,n). All temperature effects were reversible ( Supplementary Fig. S3). Taken together these results indicate an inverse relationship between synchronous and asynchronous release, where an increase in synchronous release at 32 °C seemingly comes at the cost of asynchronous release. This is in line with the previous conclusion that synchronous and asynchronous release compete for the same vesicle pool 22 . As such the loss of late-train synchronous release at 22 °C can be explained by a shift towards asynchronous release.
Buffering global Ca 2+ synchronizes late-train release but reduces total release. Next, we set out to investigate the mechanism by which competition between synchronous and asynchronous release during train stimulation shifts to synchronous release at 32 °C. Previous studies suggested that faster intracellular calcium dynamics play a role, by increasing Ca 2+ clearance, by reducing Ca 2+ influx, or by both 13,14 . To investigate this, we examined release at RT during high-frequency stimulation before and after bath application of EGTA-AM. This slow calcium buffer is known to prevent build-up of global intracellular Ca 2+ 44-46 and synchronize release at RT 22 . Release triggered by a single AP was reduced in EGTA-AM ( Supplementary Fig. S4a,b). However, during 40 Hz stimulation a strong synchronous component remained in contrast to untreated cells ( Supplementary  Fig. S4c-e,g). Meanwhile, late-train asynchronous release decreased ( Supplementary Fig. S4g), leading to an almost 10-fold increase in the synchronous fraction of late-train release ( Supplementary Fig. S4h), similar to our observations at 32 °C (Fig. 2l). We conclude that buffering of global Ca 2+ , suppresses asynchronous release and synchronizes late-train release, in line with previous studies 22,38 .
www.nature.com/scientificreports www.nature.com/scientificreports/ refilling is outweighed by faster Ca 2+ -independent RRP refilling at 32 °C 14 , or a process other than increased Ca 2+ clearance synchronizes release at 32 °C.
Raising temperature does not suppress asynchronous release in Syt1 deficient synapses. In order to establish if at 32 °C low levels of global intracellular Ca 2+ cause suppression of asynchronous release, we investigated the effect of temperature in Syt1 knock-down (KD) synapses. Inhibition of Syt1 expression eliminates synchronous release 23,24 , and previous work has shown that EGTA-AM abolishes the remaining asynchronous www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ release 38 . Therefore, this component is also expected to be reduced at 32 °C in these synapses, if increasing temperature lowers levels of global intracellular Ca 2+ . In Syt1 KD synapses, the effect of temperature on spontaneous release and single EPSCs was comparable to that observed in wild-type synapses (Fig. 3a-d and Fig. 2a-f). Median mEPSC frequency was decreased at 22 °C (Fig. 3a,b), while EPSC charge was increased (Fig. 3c,d). In accordance with previous reports 23,24 , Syt1 KD synapses lacked synchronous release during 40 Hz stimulation (Fig. 3e,f,i). At 32 °C, a small but significant increase in synchronous release was observed, although asynchronous release remained dominant ( Fig. 3e,g,i). As a consequence, the late-train synchronous fraction of release was low (Fig. 3j). Total release at 32 °C was double that at 22 °C (Fig. 3h,i), almost all of which was due to an increase in the asynchronous component. In contrast, after stimulation, asynchronous tail release decayed faster to baseline at 32 °C (Fig. 3k,l). This faster decay at 32 °C in Syt1 KD synapses suggests that Ca 2+ clearance is faster at 32 °C 14,21 . Nevertheless, the increase in asynchronous release suggests that during high frequency stimulation at 32 °C Ca 2+ influx outweighs increased Ca 2+ clearance, and a considerable global intracellular Ca 2+ component remains. Therefore, we conclude that mechanisms other than global intracellular Ca 2+ regulation synchronize release during high frequency stimulation at 32 °C in WT synapses. While likely some endogenous Syt1 remains in Syt1 KD synapses, the reversal in temperature dependence of asynchronous release compared with WT synapses (Fig. 2), indicates a clear role for Syt1 in this process. Syt1 D363N inhibits and delays the onset of asynchronous release. Syt1 and Syt2 have previously been reported to suppress spontaneous [34][35][36] and asynchronous release [37][38][39]51,52 . Given the proposed competition between synchronous and asynchronous release for synaptic vesicles 22 , this may represent a mechanism for Syt1 to establish the temperature dependent shift from asynchronous to synchronous release we observe. To investigate this possibility, we set out to test whether Syt1's release inhibitory function could suppress asynchronous release under conditions of elevated global Ca 2+ induced by our stimulation paradigm. Mutating a single negatively charged amino acid in the Syt1 C2B Ca 2+ -binding loop (D363N) both abolishes synchronous release and suppresses asynchronous release 37,39 . However, this result is not unambiguous as a similar mutation (D309A, D363A, D365A) enhances spontaneous release 53 . To confirm release inhibition by the Syt1 D363N mutation at rest, and examine its effect on Syt1's inhibitory function during high-frequency stimulation, we expressed either Syt1 D363N, or Syt1 WT in hippocampal autapses. To avoid mixing of mutant Syt1 with any residual endogenous Syt1, these constructs were expressed in neurons derived from homozygous Syt1 knock out (KO) mice. Recording at RT, we found that Syt1 D363N reduced mEPSC frequency compared to Syt1 WT rescue (Fig. 4a,b), in contrast to reports with the D309A, D363A, D365A mutations 53 . Furthermore, while Syt1 WT expressing synapses showed typical evoked release, Syt1 D363N strongly reduced both synchronous, asynchronous, and total release during 40 Hz train stimulation ( Fig. 4c-f). Additionally, Syt1 D363N expressing synapses showed reductions in late-train total and asynchronous release of more than 50% compared to Syt1 WT, while late-train synchronous release was negligible for both ( Fig. 4g). Nevertheless, the RRP was similar between Syt1 WT and D363N expressing synapses as measured with 500 mM hypertonic sucrose (Fig. 4h,i). Finally, the remaining asynchronous component in Syt1 D363N expressing synapses was delayed in onset compared to Syt1 KD synapses (Fig. 4j,k), reaching its maximum 20 APs later (Fig. 4l). This indicates that, in addition to its lack of fusion triggering, Syt1 D363N has a net inhibitory effect on vesicle fusion, both at rest and during repetitive stimulation, highlighting a potential native function of Syt1 for suppressing non-synchronous release during high-frequency stimulation, separate from its action in synchronous release triggering.

Normal late-train synchronous release in Syt1 3K expressing synapses. Given the ability of Syt1
to suppress release during high frequency stimulation, we hypothesised that this could contribute to synchronization of late-train release by influencing the competition between asynchronous and synchronous release 22 in favour of the latter. This may constitute a mechanism for temperature dependent synchronization. To investigate this, we examined the Syt1 3K mutant, where three key amino acids of the polybasic region in the C2B domain were neutralised (K326Q, K327Q, K331Q). This is analogous to previously reported mutations in Syt2 that impair its ability to inhibit spontaneous release, leading to the suggestion that these residues are critical to Syt2's ability to inhibit a second Ca 2+ sensor, and implicating a possible role in supressing asynchronous release 39 . In contrast to findings in Syt2 39 , we found that at RT the mEPSC frequency was normal in hippocampal Syt1 KO autapses expressing Syt1 3K (Fig. 5a,b). However, in line with previous reports, the first evoked response was reduced (Fig. 5c,d) 39 .
Next, we assessed the competition between synchronous and asynchronous release during 10 Hz stimulation at RT in Syt1 WT and 3 K expressing synapses. At this frequency, release remained synchronous in Syt1 WT expressing synapses (Fig. 5e-f), thus allowing effects of the 3K mutations on synchronization to be observable. Despite the reduction in the first evoked response, Syt1 3K expressing synapses displayed similar synchronous release during train stimulation as Syt1 WT due to strong facilitation ( Fig. 5e-g,i), and released the same or more in total (Fig. 5h,i). Asynchronous release was not different between Syt1 WT and 3K expressing synapses (Fig. 5i), (i) boxplot of RRP charge estimates for Syt1 WT and D363N expressing synapses (Syt1 WT: 3.039 ± 0.219 nC, n = 3; Syt1 D363N: 1.105 ± 0.923 nC, n = 7). (j) Representative traces of 40 Hz train stimulation in Syt1 D363N rescue and Syt1 KD synapses, scaled to the peak asynchronous current. (k) Asynchronous charge plot (mean ± S.E.M.), normalized to peak mean charge for Syt1 D363N rescue and Syt1 KD synapses. (l) Boxplot of pulse number containing the peak asynchronous charge for Syt1 D363N rescue and Syt1 KD synapses (Syt1 D363N: 31 ± 7, n = 25; Syt1 KD: 11 ± 1, n = 7). All recordings at room temperature (RT), ~22 °C unmonitored. (*p < 0.05, Wilcoxon rank sum test).
www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ and no significant difference was observed in the synchronous fraction of late-train release at 10 Hz (Fig. 5j). The results remained similar when increasing the stimulation frequency to 20 Hz, although here Syt1 3K autapses showed a slight increase in the synchronous fraction ( Supplementary Fig. S5f). Application of EGTA-AM prevented facilitation of the EPSC during train stimulation in Syt1 3K expressing synapses ( Supplementary Fig. S5i), with a strong trend towards lower total release compared to Syt1 WT with EGTA-AM ( Supplementary Fig. S5l,m). Taken together, these findings indicate that the 3K mutations produce an overall reduction in Syt1 triggered release, that can be compensated by global intracellular Ca 2+ , rather than impaired inhibition of spontaneous and asynchronous release as shown for Syt2 39 . As such, we conclude that the Syt1 3K mutations are not suitable for testing for a possible role of Syt1's release inhibition in temperature dependent synchronization. Instead, the observed characteristics of Syt1 3 K mimic the phenotype of a Syt1 mutant with similar mutations in the C2B polybasic region (K325A, K327A) that have previously been shown to regulate vesicle tethering to the plasma membrane in preparation for synchronous release 28 . Impaired release inhibition in Syt1 9Pro neurons desynchronizes late-train release. As an alternative candidate for investigating the role of inhibition by Syt1 in temperature dependent synchronization of late-train release, we selected the Syt1 9Pro mutant, in which the flexible linker between the C2 domains is replaced with a rigid nine-residue proline segment. This mutant was reported to show impaired inhibition of spontaneous release while the first evoked was unaffected 54,55 . Although effects from these mutations on asynchronous release have not yet been investigated, the specific orientation of Syt1's C2 domains enforced by the rigid proline linker in Syt1 9Pro is believed to affect Syt1's ability to clamp fusion in general 55 , making this a possible location for the competitive advantage of synchronous release observed at higher temperatures. In line with this, we observed significantly higher mEPSC frequencies in hippocampal Syt1 KO autapses expressing Syt1 9Pro than in Syt1 WT rescue recorded at RT (Fig. 6a,b). The first evoked response was not different (Fig. 6c,d), as shown previously 54,55 . Interestingly, during 10 Hz trains, synchronous release decreased to a greater extent in Syt1 9Pro expressing synapses, and asynchronous release increased concomitantly ( Fig. 6e-g,i). As a consequence, the synchronous fraction of late-train release was significantly reduced compared to Syt1 WT (Fig. 6j), without affecting total release during 10 Hz stimulation (Fig. 6h,i). Desynchronization in Syt1 9Pro expressing synapses became even more pronounced at 20 Hz (Fig. 7a-c,e,f). Hence, although a similar number of vesicles is released during train stimulation in Syt1 WT and 9Pro expressing synapses, release in Syt1 9Pro is less synchronous and more asynchronous.
Besides increased competition from asynchronous release, poor synchronization in Syt1 9Pro expressing synapses might be due to an overall deficit in triggering of late-train synchronous release. To investigate this, we suppressed asynchronous release at RT by buffering global intracellular Ca 2+ with EGTA-AM. After incubation with EGTA-AM, release completely synchronized, similar to Syt1 WT expressing synapses with EGTA-AM ( Fig. 7g-i,k,l). Application of EGTA-AM reduced the first evoked response and the total release during the train to a similar extent for Syt1 WT and Syt1 9Pro expressing synapses (Supplementary Fig. S7; Fig. 7d,j). We conclude that late-train synchronous release in Syt1 9Pro expressing synapses is only impaired when asynchronous release is present. Likely, in Syt1 9Pro expressing synapses, decreased suppression of asynchronous release allows it to out-compete synchronous release for synaptic vesicles.
Given a role for Syt1 in suppressing asynchronous release, this would likely also affect the decay of asynchronous tail release after stimulation trains. However, asynchronous tail release after 100 pulse trains at 10 Hz recorded at RT was similar in Syt1 9Pro and Syt1 WT expressing synapses ( Supplementary Fig. S6). However, in Syt1 9Pro expressing synapses the decay of tail release slowed significantly when the stimulation frequency was increased up to 40 Hz, but not in Syt1 WT (Supplementary Fig. S6b). Possibly, only when intracellular Ca 2+ remains high for an extended period (as after 40 Hz), can an effect from Syt1's release inhibitory function be detected in asynchronous tail release.

Suppression of asynchronous release by Syt1 is required for temperature dependent synchronization.
In order to determine the role of Syt1 mediated suppression of asynchronous release in the temperature dependent change in late-train synchronous release, we recorded hippocampal Syt1 KO autapses expressing Syt1 WT or 9Pro at 22 °C and 32 °C. Basal release parameters in both Syt1 WT and 9Pro expressing synapses responded similarly to temperature shifts. Both showed a 4-fold lower mEPSC frequency at 22 °C, while frequencies overall were higher for Syt1 9Pro (Supplementary Fig. S8a,b). We observed no significant effects on first evoked responses ( Supplementary Fig. S8c,d). At RT, during 10 Hz and 20 Hz stimulation Syt1 9Pro expressing synapses showed a clear defect in maintaining synchronous release compared to WT (Figs 6j and 7f). At 40 Hz, both Syt1 WT and 9Pro expressing synapses almost completely desynchronized (Fig. 8a,b,f) and similar late-train synchronous fractions were observed at 22 °C (Fig. 8j). However, at 32 °C, Syt1 9Pro expressing synapses did not synchronize release to the same extent as Syt1 WT, and asynchronous release remained dominant (Fig. 8a,c,g). Although both Syt1 WT and 9Pro expressing synapses showed clear temperature-dependent increases in late-train synchronous release (Fig. 8e,i), the synchronous fraction of late-train release in Syt1 9pro expressing synapses was significantly lower than Syt1 WT at 32 °C (Fig. 8j). Finally, after stimulation ended, the decay of asynchronous tail release at 22 °C was slower (Fig. 8k,m), but not at 32 °C (Fig. 8l,m). This is in line with our previous findings ( Supplementary Fig. S6), and supports faster Ca 2+ clearance at 32 °C 14,21 . We conclude that Boxplot with fraction of late-train charge transfer during the final 3 pulses released synchronously (Syt1 WT: 0.3834 ± 0.0616, n = 33; Syt1 3K: 0.4038 ± 0.0522, n = 24). All recordings at room temperature (RT), ~22 °C unmonitored. (*p < 0.05, Wilcoxon rank sum test).
www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ the Syt1 9Pro mutations inhibit temperature dependent synchronization. Impaired release inhibition in Syt1 9Pro expressing synapses likely leads to premature depletion of vesicles by asynchronous release. This leaves insufficient releasable vesicles for late-train synchronous release. We propose that suppression of asynchronous release by Syt1 constitutes a temperature sensitive mechanism for synchronization of release during high-frequency stimulation, and enables frequency coding.

Discussion
Synchronization of vesicle release to AP firing is a prerequisite for frequency coding at the synapse. We show that lowering the temperature or reducing the inhibitory function of Syt1 by rendering the C2-domain linker region inflexible, impairs synchronous release at high-frequencies. At the same time, asynchronous release is enhanced, in line with the notion that both modes of release compete 22,42 . Based on these findings we propose that the main role of Syt1's release inhibitory function is to synchronize synaptic transmission by suppressing asynchronous release in between APs in a highly temperature sensitive manner. This novel role of the inhibitory function of Syt1 is independent of its role in synchronizing the first evoked response and contributes together with its known vesicle recruitment and triggering functions to synchronous transmission over a broad frequency bandwidth.
Our data confirmed previously observed differences in the temperature sensitivity of synchronous release in response to a single AP and during high-frequency stimulation in hippocampal synapses 14 . While the first evoked response remained synchronous at 22 °C, with unchanged amplitude and increased width, late-train synchronous release was almost completely abolished ( Fig. 2e-g). Based on previous reports, this difference between initial and late-train release cannot be explained by post-synaptic receptor desensitization at RT 15 , or failure of action potential propagation, which has been shown to be extremely reliable at high-frequencies and different temperatures [56][57][58] . Furthermore, no evidence of impaired action potential propagation at RT is present when release is synchronized through incubation with EGTA-AM ( Supplementary Fig. S4).
In the Calyx of Held, late-train release can remain synchronous at RT during frequencies beyond 100 Hz, and both first evoked and late-train EPSCs increase in amplitude and decrease in width after raising temperature 13,15 . An increase in EPSC width is most likely caused primarily by presynaptic processes, such as broadening of the Ca 2+ peak 59,60 due to an increased AP width 15 and reduced buffering and extrusion of Ca 2+ at 22 °C 14,21 . The fact that at the same time late-train synchronous release is reduced at 22 °C suggests that an additional mechanism is at work to synchronize release during high-frequency stimulation, which becomes less efficient at lower temperatures. In the Calyx this process might be more efficient at RT or these synapses may exploit other principles to synchronize release during trains.
Temperature dependent changes in calcium currents 13,15,20 , and/or calcium clearance 14,21 have been proposed as possible mechanisms for temperature sensitive synchronization of late-train release. Broader AP induced calcium peaks and slower calcium buffering/extrusion may increase global intracellular Ca 2+ at 22 °C, thus promoting asynchronous release 43 . Consequently, given competition between synchronous and asynchronous release for the same vesicle pool 22,42 , synchronous release would be reduced. Indeed, we found that both buffering global intracellular Ca 2+ with EGTA-AM and raising temperature achieved similar increases in late-train synchronous release (Supplementary Fig. S4; Fig. 2). However, the two conditions differed in that increasing temperature increased total release whereas EGTA-AM lowered it. Presumably lower levels of global intracellular Ca 2+ reduced refiling of the vesicle pool in the latter case [47][48][49][50] . Furthermore, in Syt1 KD synapses, where release depends entirely on global Ca 2+ 38 , raising the temperature did not suppress but instead increased asynchronous release during stimulation (Fig. 3). At the same time, Syt1 KD synapses showed faster decay of asynchronous tail release after stimulation, indicative of faster Ca 2+ clearance at 32 °C. These observations suggest that during high-frequency stimulation, Ca 2+ influx outcompetes Ca 2+ clearance and global intracellular Ca 2+ reaches levels high enough to support the increase in total release we observed in Syt1 KD synapses at 32 °C. Therefore, we conclude that faster Ca 2+ dynamics are not the dominant factor in temperature dependent synchronization in hippocampal synapses.
In addition to triggering synchronous release 23,24 Syt1 suppresses asynchronous release [37][38][39]54,55 . This function is evident when Syt1's fusion triggering function is selectively impaired in the D363N mutant with reduced Ca 2+ -binding to the C2B domain, displaying that inhibition persist even during high-frequency stimulation (Fig. 4). Suppression of asynchronous release by Syt1 could provide a competitive advantage to synchronous release, and thereby represent the mechanism for temperature-dependent synchronization. Until now, studies of Syt1's inhibitory function have focussed on spontaneous release 39,54,55 and suggested a role, either in maintaining the synaptic vesicle pool by preventing "leakage" of vesicles 26,61 , or in suppressing biological noise 12 . In line with these reports, we confirmed that impaired inhibition in the Syt1 9Pro mutant leads to increased spontaneous release 54,55 . However, this did not cause a defect in the first evoked response or total release during high-frequency stimulation (Fig. 6). As such, a role for the inhibitory function in maintaining the synaptic vesicle pool is unlikely. Interestingly, during high-frequency stimulation, Syt1 9Pro expressing synapses showed faster depression of synchronous release, while, concomitantly, asynchronous release increased (Figs 6-8), and persisted for longer ( Supplementary Fig. S6). In accordance with a competition model, the defect in synchronization could be completely rescued by suppressing asynchronous release with EGTA-AM (Fig. 7), bypassing the inhibitory role of Syt1, and excluding a possible defect in Ca 2+ -dependent release triggering due to the 9Pro mutations. However, rescue was only partial when raising recording temperature from 22 °C to 32 °C (Fig. 8). As we showed global separately. (j) Boxplot with fraction of late-train charge transfer during the final 3 pulses released synchronously (Syt1 WT: 0.5409 ± 0.0360, n = 27; Syt1 9Pro: 0.3654 ± 0.0755, n = 28). All recordings at room temperature (RT), ~22 °C unmonitored. (*p < 0.05, Wilcoxon rank sum test).
www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ Ca 2+ persists at 32 °C (Fig. 3), and the 9Pro mutations impair temperature dependent synchronization, it is likely that raising temperature requires Syt1's release inhibitory function to synchronize release. In contrast, spontaneous release increased with temperature, regardless of the 9Pro mutations ( Supplementary Fig. S8a,b). This may indicate a basal increase in fusogenicity, independent of Syt1's suppression of asynchronous release. Overall, we propose that preventing asynchronous release during stimulation is the main purpose of Syt1's release inhibitory function. While suppression of spontaneous release undoubtedly has relevance, it is likely secondary to the increased maintenance of synchronous release achieved by inhibition of the asynchronous component, which directly affects the neurons signalling capabilities.
Syt1 mediated vesicle recruitment, or plasma membrane attachment, has also been shown to aid in synchronization of release, both in a constitutive and Ca 2+ -dependent manner 28 . This notion fits within a proposed sequential two-step release model, where a pre-release docking step separates synchronous release from asynchronous release [62][63][64] . The same residues required for constitutive recruitment 28 are targeted in the Syt1 3K mutant used in this study. While we originally wished to employ the Syt1 3K mutant to test to role of Syt1's inhibitory function in temperature dependent synchronization of release, we did not find any evidence of impaired release inhibition due to these mutations, in contrast to previous results in the Calyx of Held 39 . We did observe a reduced first evoked response in Syt1 3K compared to Syt1 WT expressing synapses (Fig. 5), in line with a role in recruitment. The defect in synchronous release after a single AP was rescued during high-frequency stimulation (Fig. 5). This suggests that Ca 2+ -dependent vesicle recruitment remains intact in Syt1 3K expressing synapses and contributes to late-train synchronous release during high-frequency activity. However, accelerated vesicle recruitment alone is not sufficient to maintain synchronous release during high-frequency stimulation. Firstly, vesicles recruited during high-frequency stimulation tend to release asynchronously due to global intracellular Ca 2+ 22,42,43 . Secondly, reducing Ca 2+ -dependent vesicle recruitment by buffering global intracellular Ca 2+ actually leads to an increase in synchronous charge in WT synapses ( Supplementary Fig. S4) 22,42 , instead of a reduction.
Instead, we propose a three-step mechanism by which Syt1 supports synchronous release during low-and high-frequency activity. First, Syt1 recruits vesicles to the plasma membrane to establish or refill the synaptic vesicle pool. Second, Syt1 suppresses global Ca 2+ triggered asynchronous release. Third, Syt1 triggers synaptic vesicle release upon peak Ca 2+ , coupled to the AP. Under near-physiological conditions (WT; 32 °C), Syt1 aids in providing a sufficient vesicle pool at rest (Fig. 9a), capable of producing a synchronous response upon peak Ca 2+ after an AP ( Fig. 9a; inset). During high-frequency stimulation, asynchronous release is prevented (Fig. 9b left; inset) due to Syt1's release inhibitory function (clamped vesicles; Fig. 9b left), reserving these vesicles for synchronous release with peak Ca 2+ upon the next AP (Fig. 9b right). When temperature is lowered (WT; 22 °C), release at rest is not affected (Fig. 9c), however, due to the temperature dependent slowing of the on-rate of Syt1's inhibitory function, during late-train release, newly primed vesicles are released asynchronously with global Ca 2+ before they can be clamped (Fig. 9d left). This depletes the available vesicles for the next peak Ca 2+ (Fig. 9d right), leaving insufficient vesicles for synchronous release (Fig. 9d right; inset). Similarly, when Syt1's inhibitory function is impaired (Syt1 9Pro), global Ca 2+ depletes the available vesicles asynchronously (Fig. 9f), and a synchronous peak cannot be generated during the late train (Fig. 9f right; inset). However, Syt1's inhibitory function does not affect release from a state of rest, where the only Ca 2+ trigger available is peak Ca 2+ synchronous with the AP (Fig. 9e). In contrast, impairing Syt1's constitutive vesicle recruitment function reduces the available pool for synchronous release at rest (Fig. 9g), leading to a decrease in the synchronous response ( Fig. 9g; inset). However, during high-frequency stimulation, increases in Ca 2+ -dependent vesicle recruitment driven by global Ca 2+ , provide a strong supply of vesicles (Fig. 9h), supporting synchronous release upon peak Ca 2+ (Fig. 9h  right; inset). In conclusion, we argue that Syt1's inhibitory function acts independently, but in cooperation with its Ca 2+ -dependent release triggering 23,39 and vesicle recruitment functions 25,27,28 , to maintain synchronous release and enable frequency coding over a broad frequency bandwidth.
Until now, little is known about the molecular mechanism of release inhibition by Syt1. Recent structural studies have exposed two separate interfaces of the Syt1 C2B domain with the primed SNARE complex 65,66 , both important for fusion. One of these, the tripartite interface, was additionally implicated in establishing a 'lock' on release 66 . One possibility is that in the Syt1 9Pro mutation, fixing the orientation of the C2B domains impairs the inhibitory function by preventing optimal interaction with the SNAREs 55 . Increased release inhibition by Syt1 at 32 °C could be due to temperature dependent increases in the association and dissociation constants for the tripartite interface with the SNAREs. This would not affect the steady state of the inhibited, or "clamped", pool of vesicles (Fig. 9a), but would speed up the clamping of newly primed vesicles after late-train synchronous release of previously clamped vesicles.
Interestingly, several other C2B domain containing proteins can associate with SNAREs at the tripartite interface 66 . Among them, Synaptotagmin-7 (Syt7), which has been implicated in triggering asynchronous release 52,[67][68][69] . This is in line with a model where distinct Ca 2+ -sensors compete for SNARE-binding, thus defining the balance between synchronous and asynchronous release 66,70 . Indeed, a recent study shows that neurons control the relative contributions of synchronous and asynchronous release, and thereby the ability to encode spike trains, by regulating expression of fast Synaptotagmin isoforms 71 . However, other pathways of suppression, such as direct inhibition of a second Ca 2+ sensor for asynchronous release cannot be excluded 39 .
www.nature.com/scientificreports www.nature.com/scientificreports/  69,[71][72][73][74][75] . In the hippocampus, inhibitory cholecystokinin (CCK) interneurons display substantial asynchronous release after burst of activity, presumably to prolong release inhibition 72 . Similarly, in the cortex, asynchronous GABA release is thought to decrease generalized synchronous firing, preventing epileptic activity 75 . Furthermore, asynchronous release triggered by Syt7 is proposed to help in boosting post-synaptic firing probability during high-frequency stimulation 68 . Interestingly, Syt7 has also been described to contribute to synchronous release as a sensor for short-term plasticity 69,76,77 , and establishing amplitude frequency invariance 77,78 . One way to explain the involvement in these two processes is the ability of Syt7 or other slow Ca 2+ sensors to reduce the energy barrier for vesicle fusion when activated by global intracellular Ca 2+ . Activation of a slow Ca 2+ sensor by itself will trigger asynchronous release, thus desynchronizing release. However, when it is activated simultaneously with Syt1, their additive effect on the energy barrier could lead to multiplicative effects on the vesicle fusion rate, requiring only little activation of the slow sensor to boost Syt1 action 77,79,80 . In this scenario, Syt1 performs a balancing act by allowing just enough activation of a second Ca 2+ sensor to enhance synchronous release during AP's without inducing substantial premature release of vesicles in between AP's.
In conclusion, we identified a novel aspect of the inhibitory function of Syt1 in maintaining synchronous transmission over a wide range of frequencies by suppressing asynchronous release. This function appears to be independent of its fusion promoting, and vesicle recruitment functions, and presumably acts by (partially) inhibiting slow Ca 2+ sensors. All these functions of Syt1 work together to optimize synchronous vesicle release, thus broadening the bandwidth for frequency coded information transfer.

Materials and Methods
Animals. Neuronal cultures were prepared from embryonic day 18 (E18) pups of both sexes, obtained by caesarean section of pregnant female mice. For WT neurons, the C57BL/6 mouse line was used (Figs 1-3), and a previously described Synaptotagmin-1 knockout 23 line was used for Syt1 KO neurons (Figs 4-8). New-born pups (P0-P1) from Winstar rats were used for glia preparations. Animals were housed and bred according to institutional and Dutch governmental guidelines, and all procedures are approved by the ethical committee of the Vrije Universiteit, Amsterdam, The Netherlands (Dierexperimentencomissie (DEC) license number: FGA11-03).
Constructs and lentiviral infection. The Synaptotagmin-1 mutations Syt1 D363N, and Syt1 3 K (K326Q, K327Q, K331Q), were generated using Quickchange (Stratagene) and verified by sequencing, Syt1 9Pro (residues 264-272 replaced with nine proline residues) as previously described 54,55 , and was kindly provided by Dr. Edwin Chapman, (Howard Hughes Medical Institute, Madison, WI, USA). For Syt1 knockdown, we used anti-Syt1 shRNA described previously 52 , kindly provided by Thomas Südhof (Howard Hughes Medical Institute, Stanford, CA, USA). For expression in neurons, constructs were subcloned into synapsin-promotor-driven-pLenti vectors, and viral particles were produced as described 81 . For Syt1 knockdown, WT neurons were infected with viral particles at DIV 2, and for rescue experiments, Syt1 KO neurons were infected at DIV4. In experiments regarding the effect of temperature, cells were controlled at 22 ± 1 °C and 32.5 ± 1 °C using bath perfusion with warm or cold ACSF. For these experiments, bath temperature was monitored continuously using two temperature sensors (TS-100; npi Electronic GmbH), one of which was placed near the bath outflow and provided feedback to an inline solution heater while the other sensor was positioned near the patched cell single exponential fits of asynchronous tail release in Syt1 WT and 9Pro neurons at 22 °C and 32 °C (WT 22 °C: τ = 0.4956 ± 0.127 s, n = 19; 9Pro 22 °C: τ = 0.6869 ± 0.160 s, n = 21; WT 32 °C: τ = 0.1534 ± 0.0879 s, n = 16; 9Pro 32 °C: τ = 0.1762 ± 0.0743 s, n = 21). (*p < 0.05, Wilcoxon signed-rank test for paired and Wilcoxon rank sum test for independent samples).
www.nature.com/scientificreports www.nature.com/scientificreports/ and used to register the precise temperature during recordings. All other recordings were performed at room temperature (RT) 22 ± 2 °C unmonitored. For experiments including the cell permeable Ca 2+ chelator EGTA-AM, cells were recorded at RT before or after 10 min incubation with 50 µM EGTA-AM in bath. This concentration was chosen to be sufficiently high to avoid buffer saturation during train stimulation, while keeping suppression of EPSC amplitude to a minimum. RRP size was assessed by hypertonic sucrose application (500 mM for 7 s) 79,82 , using a piezo-controlled barrel application system (Perfusion Fast-Step, Warner Instruments) and fitted as described previously 79 .  Figure 9. Syt1's release inhibitory function synchronizes release during high-frequency stimulation. Working model for the effect of Syt1's release inhibitory and vesicle recruitment functions on release coming from rest (First evoked release; left column) and during high-frequency stimulation (right column), split between global Ca 2+ induced release in the interpulse-interval (i.e. 75 ms), and peak Ca 2+ induced release right after the AP (i.e. 25 ms). With vesicles not subject to Syt1's release inhibitory function in green (unclamped), and clamped vesicles in red. (a,c,e,g) First evoked release after a period of rest during peak Ca 2+ in WT synapses at 32 °C (a), in WT synapses at 22 °C (c), in Syt1 9Pro synapses at 32 °C with a shift towards unclamped vesicles due to impaired Syt1 release inhibition (e), and in Syt1 3K synapses at 32 °C with a decreased number of available vesicles due to impaired constitutive vesicle recruitment (g). (b,d,f,h) Late-train release during high-frequency stimulation, due to global Ca 2+ (left), and peak Ca 2+ (right), in WT synapses at 32 °C (b) where rapid transition of newly primed vesicles to the clamped (red) state suppresses asynchronous release due to global Ca 2+ (b, left), preserving vesicles for synchronous release during peak Ca 2+ (b, right). In WT synapses at 22 °C (d), where a slower on-rate of release inhibition leaves vesicles unclamped causing asynchronous release (d, left), and leaving insufficient vesicles for a synchronous response (d, right). In Syt1 9Pro synapses at 32 °C (f) where impaired release inhibition leaves vesicles unclamped (f, left), leaving too few vesicles for synchronous release (f, right). And in Syt1 3 K synapses at 32 °C (h) where global Ca 2+ drives increased Ca 2+ -dependent recruitment, increasing the available pool for synchronous release (h, right). In all panels insets show representations of the post-synaptic response based on the depicted fusion events, with synchronous responses in dark grey, asynchronous responses in light grey, and the baseline represented by the dashed line.
www.nature.com/scientificreports www.nature.com/scientificreports/ Data analysis. Offline analysis was performed using custom-written software routines in Matlab R2018b (Mathworks). In all figures, stimulation artefacts have been removed. For evoked release, total charge was calculated by integrating the current from the end of the stimulation until the start of the next pulse. Synchronous release was separated from asynchronous release by subtracting the standing current at the start of the EPSC and integrating the first 25 ms of the stimulation, referred to as the synchronous window ( Supplementary Fig. S2). This window was chosen based on a comparison of release per pulse before and after incubation with EGTA-AM, under the assumption that most asynchronous release is blocked by buffering global Ca 2+ , and provided an objective estimate of the synchronous component (see main text). Asynchronous release per stimulation interval was determined by subtracting the synchronous charge from the total charge. Statistical significance was determined using Wilcoxon signed-rank tests and Mann-Whitney U tests to compare paired-and independent measurements, respectively, p-values below 0.05 were considered significant. All statistical tests were performed in Matlab (Mathworks).