Introduction

Synapses in the mammalian central nervous system undergo activity-dependent modulation of synaptic weight during repetitive stimulation, known as short-term synaptic plasticity (STP). STP is mediated by presynaptic mechanisms that regulate the release probability (p_r), size of the readily releasable pool (RRP), and kinetics of vesicle replenishment of RRP (Zucker and Regehr, 2002; Jackman and Regehr, 2017; Neher and Brose, 2018). Short-term depression has been observed at synapses with a high baseline p_r, primarily because of the rapid depletion of releasable vesicles during high-frequency stimulation (HFS) (Lin et al., 2022). Conversely, vesicle depletion is less pronounced at synapses with a low baseline p_r, and cumulative increases in p_r during HFS are considered to mediate short-term facilitation (Rozov et al., 2001; Pan and Zucker, 2009; Aldahabi et al., 2024). Central synapses exhibit a wide spectrum of STP depending on different combinations of presynaptic factors mediating short-term depression and facilitation.

Previously, facilitation was modeled as an increase in p_r after an action potential (AP) and its slow decay during the inter-spike interval (ISI), resulting in a cumulative increase in p_r during HFS (Markram et al., 1998; Dittman et al., 2000). Recent studies have suggested that the number of docking (or release) sites in an active zone is limited and only partially occupied by releasable vesicles in the resting states (Miki et al., 2016; Pulido and Marty, 2018; Malagon et al., 2020; Lin et al., 2022). Therefore, p_r is determined by the vesicular fusion probability (p_v) and release site occupancy (p_occ). However, distinguishing the individual contributions of p_v and p_occ to p_r is challenging (Silva et al., 2021; Neher, 2024). It is also unclear whether facilitation is mediated by activity-dependent increases in the p_v, p_occ, or both.

Two recent models for vesicle priming and release view the ‘priming’ as a two-step reversible process that leads to a release-ready state. The release-ready vesicles could be vesicles in a tightly docked state (TS) or those occupying specialized docking sites (DS), and are supplied from vesicles in a loosely docked state (LS) or those occupying a distinct replacement sites (RS), which are called LS/TS model ((Neher, 2024); (Aldahabi et al., 2024)) and RS/DS model ((Miki et al., 2016); (Silva et al., 2021)), respectively. Common to both models is the assumption that release-ready vesicles constitute only a fraction of all docked vesicles under resting conditions and that this fraction is up- and down-regulated by synaptic activity.

Synaptotagmin-7 (Syt7) was recently found to play a crucial role in facilitation at different types of central synapses (Jackman et al., 2016; Chen et al., 2017; Martinetti et al., 2022). In addition to its role in facilitation, Syt7 accelerates refilling of the RRP after depletion and mediates asynchronous release (Bacaj et al., 2013; Liu et al., 2014; Jackman et al., 2016; Tawfik et al., 2021). Syt7-dependent synaptic facilitation is interpreted as an activity-dependent increase in p_v (Turecek et al., 2016; Jackman and Regehr, 2017; Turecek et al., 2017; Norman et al., 2023). However, it is unclear whether Syt7 contributes to facilitation through a Ca²⁺-dependent increase in p_occ (i.e., overfilling). Overfilling has been proposed as a mechanism supporting facilitation at the Drosophila neuromuscular junction (Kobbersmed et al., 2020), yet whether this process also occurs at the mammalian central synapses mediated by Syt7 has not been studied.

Facilitation at recurrent excitatory synapses in cortical recurrent networks may play a crucial role in maintaining working memory (Mongillo et al., 2008; Mongillo et al., 2012; Hansel and Mato, 2013). Although layer 2/3 (L2/3) pyramidal cells (PCs) are highly interconnected and represent task-relevant information during a delay period in working memory tasks (Fujisawa et al., 2008; Ozdemir et al., 2020), most studies on local excitatory recurrent synapses in the neocortex have focused on layer 5 (L5) exhibiting short-term depression (Reyes and Sakmann, 1999; Hempel et al., 2000; Yoon et al., 2020). Further, the synaptic properties of excitatory synapses in L2/3 remain poorly understood.

In the present study, we found that HFS of local excitatory synapses in L2/3 of the medial prefrontal cortex (mPFC) induced initial strong depression followed by delayed facilitation, whereas low-frequency stimulation induced slow monotonous facilitation. This unique form of STP could be explained by 1) high p_v of RRP vesicles at rest and throughout a train stimulation; 2) the low resting vesicular occupancy of RRP and its activity-dependent increase. These conditions led to delayed facilitation through progressive overfilling of release sites during HFS with high p_v vesicles beyond the resting occupancy. Syt7 knockdown (KD) in L2/3 PCs completely abolished these synaptic enhancements by inhibiting the Ca²⁺-dependent increase in release site occupancy. Finally, behavioral tests for trace fear conditioning showed that Syt7 KD impaired the acquisition of trace fear memory and reduced c-Fos expression in the mPFC. Collectively, these results support that Ca²⁺- and Syt7-dependent overfilling of release sites mediates synaptic facilitation at L2/3 recurrent excitatory synapses and contributes to temporal associative learning.

Results

Frequency dependence of STP at local excitatory synapses in L2/3 of the prelimbic cortex

We examined STP profiles at recurrent excitatory synapses between L2/3 PCs and at synapses from L2/3-PCs to fast-spiking interneurons (FSINs) in the prelimbic area of the rat mPFC. The PCs and FSINs were identified based on their distinct intrinsic properties and morphologies (Fig. S1). To stimulate presynaptic PCs, L2/3 PCs were transfected with a plasmid encoding oChIEF using in utero electroporation (IUE; Fig. 1A; Lee et al., 2024). Photostimulation of the oChIEF-expressing cell soma at 40 Hz consistently evoked single APs with high fidelity throughout the 600-pulse trains (Fig. S2A-B), which was consistent with little use-dependent inactivation of oChIEF (Lin et al., 2009; Hass and Glickfeld, 2016).

Figure 1.

Given that IUE transfected approximately 10-20% of L2/3 PCs (Fig. 1A), we made a whole-cell patch on a non-transfected L2/3 PC or an FSIN (Fig. 1C). To stimulate a minimal number of excitatory axon fibers, EPSCs were evoked using minimal optical stimulation (Fig. S3A; 470 nm, 3-4 μm in radius and 3-5 ms in duration). EPSCs evoked by minimal stimulation were examined to determine the uniformity of the kinetic properties (Fig. S3B-C). Moreover, we reassessed the kinetics of EPSCs during trains post hoc to ensure uniformity. EPSCs were evoked by 20-pulse trains at 4 different frequencies (5, 10, 20, and 40 Hz) to cover the range of firing rates of PCs observed in vivo during working memory tasks (Baeg et al., 2003). Excitatory synapses from PCs onto both other PCs and FSINs (PC–PC and PC–FSIN) showed strong (∼2-fold enhancement) and lasting (several seconds) short-term facilitation at 20 Hz and lower frequencies (Fig. 1D-E). The PC-PC synapses exhibited stronger facilitation than did the PC-FSIN synapses.

Steady-state EPSCs at the PC-PC synapses were not significantly different from 5 to 20 Hz (Fig. 1F). Frequency invariance at the PC-PC synapses suggests that Ca²⁺-dependent synaptic facilitation may counteract vesicle depletion in a frequency-dependent manner (Turecek et al., 2017; Lin et al., 2022). Despite facilitation at all frequencies, the paired pulse ratio (PPR) rapidly decreased as the ISI decreased (Fig. 1G, I). At 40 Hz stimulation, both PC-PC and PC-FSIN synapses initially underwent strong depression then followed by facilitation in a few stimuli (Fig. 1Dd, Ed). Such delayed facilitation after paired pulse depression (PPD) has not been previously observed in central synapses except for a slight increase in EPSC at the cerebellar synapses under artificially depolarized conditions (Pulido and Marty, 2018).

We confirmed that optically evoked EPSCs were strictly dependent on AP generation (Fig. S4A), suggesting that EPSCs are evoked by optical stimulation of presynaptic axon fibers rather than by direct depolarization of presynaptic boutons. Additionally, dual patch-clamp techniques showed that the STP of optically evoked EPSCs did not differ from the electrically evoked STP of EPSCs at the PC-FSIN synapses (Fig. S4B). Finally, we confirmed that AMPA receptor (AMPAR) desensitization was not responsible for PPD at 40 Hz, ruling out the possible effects of postsynaptic factors on STP (Fig. S5A), and that AMPAR saturation was not significant during facilitation (Fig. S5B).

Delayed facilitation results from slow activation of Ca²⁺-dependent vesicle replenishment at a constantly high vesicular fusion probability

The strong PPD at 40 Hz suggests a high vesicular fusion probability (p_v) and likely tight coupling of the releasable vesicles to a Ca²⁺ source. To test this hypothesis, we examined the effects of EGTA, a slow calcium chelator, on synaptic transmission at the PC-PC synapses. The baseline EPSC amplitude (EPSC₁) was not affected by incubation of the slice in the bath solution containing 50 μM EGTA-AM for longer than 20 min, supporting tight coupling. Moreover, the PPR was decreased to 0.19, and subsequent facilitation slowed but was still distinct (Fig. 2A-B). These findings indicate that the initial p_v should be higher than 0.81 considering vesicle recruitment during ISI. The effects of EGTA-AM were confirmed in the same cell by measuring the baseline EPSC and PPR before and after applying EGTA-AM to the bath. EPSC₁ remained unchanged, while PPR was reduced by half (Fig. S6).

Figure 2.

Notably, the application of EGTA-AM markedly slowed facilitation at 40 Hz (Fig. 2A,C). The frequency invariance of steady-state EPSCs (Fig. 1F) and the slower facilitation at 40 Hz in the presence of EGTA (Fig. 2A) imply a Ca²⁺-dependent increase in the refilling rate of the RRP and/or an increase in the p_v of reluctant vesicles in light of previous studies (Hosoi et al., 2007; Liu et al., 2014; Ritzau-Jost et al., 2014; Turecek et al., 2016; Ritzau-Jost et al., 2018; Kusick et al., 2020). Moreover, facilitation was accelerated at higher stimulation frequencies. The plot of the rate constant for the increase in EPSCs (denoted as k_STF, the rate constant for the development of short-term facilitation) as a function of the stimulation frequency (f_stim) revealed a linear relationship (Fig. 2C). k_STF in the presence of EGTA-AM was distinctly below this relationship, supporting the Ca²⁺-dependent acceleration of facilitation.

k_STF was also accelerated when the stimulation frequency was increased from 5 Hz to 40 Hz (Fig. 2D; 1.27 ± 0.18/s to 24.27 ± 3.44/s). Notably, the second 40-Hz train stimulation led to strong PPD, followed by accelerated facilitation (Fig. 2D,E). The strong PPD suggests that the vesicular fusion probability kept high during the 5-Hz stimulation, arguing against the possibility for an increase in p_v. Both two-step sequential priming models, LS/TS and RS/DS models, are suitable to interprete our data and we will refer to the different populations of vesicles as TS and LS vesicles, respectively (Taschenberger et al., 2016; Neher and Brose, 2018; Lin et al., 2022; Neher, 2024). Within this framework, vesicles are are released from TS with high p_v both in the resting state and during facilitation, and the delayed facilitation is seen as a result of Ca²⁺-dependent progressive increase in the occupancy of TS vesicles on docking sites.

Release sites have low baseline occupancy and these are increased during facilitation and post-tetanic augmentation

Given that p_v was greater than 0.8, the two-fold increase in EPSC during 20 pulse train stimulations could not be explained by an increase in p_v. Recent studies have shown that vesicle release sites (N) are limited in number and are partially occupied at rest (Miki et al., 2016; Pulido and Marty, 2018; Malagon et al., 2020; Lin et al., 2022). Under such a high p_v, two-fold synaptic facilitation should be attributed to an increase in p_occ. The vesicle replenishment rate is accelerated during HFS in a Ca²⁺-dependent manner (Fig.2A-C; Hosoi et al., 2007). This may lead to the overfilling of release sites beyond their basal vesicular occupancy, which in turn may increase EPSC under high p_v. To test this hypothesis, we determined the quantal parameters at the L2/3 local excitatory synapses. We first measured the quantal size from asynchronous release events after replacing extracellular Ca²⁺ with Sr²⁺, as previously described (Yoon et al., 2020) (Fig. S7A-B). The mean and coefficient of variance of the quantal sizes were 24 pA and 0.25 at PC-PC synapses, respectively, and were 30 pA and 0.31 at PC-FSIN synapses, (Fig. S7C).

Next, we performed a variance-mean (V-M) analysis of phasic EPSCs at excitatory synapses during train stimulation at 5 Hz and 40 Hz (Scheuss et al., 2002). Applying multiple-probability fluctuation analysis (MPFA) using intra-site quantal variability of 0.2 (Valera et al., 2012), the V-M plot was best fitted by the number of release sites (N) of 5.3 and the quantal size (q) of 25 pA at the PC-PC synapses (Fig. 3A). The p_r for EPSC₁ was calculated as 0.32. We performed the same analysis for the PC-FSIN synapses and found N = 5.3, q = 35 pA, and p_r for EPSC₁ = 0.31 (Fig. 3B). Our estimate for N was smaller but comparable to the estimate (6.9) at excitatory synapses in L2/3 of the mouse barrel cortex (Holler et al., 2021).

Figure 3.

The p_r estimate was clearly smaller than the possible minimum value for p_v (0.81) estimated from the PPR data (Fig. 2B). Given that p_r is a product of p_occ and p_v (Neher, 2024), a p_r smaller than p_v indicates the partial occupancy of release sites in the resting state. If synaptic facilitation were mediated by the recruitment of new release sites, it would be accompanied by an increase in the variance of augmented EPSCs (Valera et al., 2012; Kobbersmed et al., 2020). However, the augmented EPSCs during facilitation followed the same parabola in the V-M plot (Fig. 3A,B). Therefore, facilitation appears to be mediated by the progressive overfilling of a finite number of release sites beyond the basal occupancy. In the next section, we show that p_v is close to unity based on a synaptic failure analysis (Fig. S8). Assuming this, the baseline p_r at the first pulse estimated from the fitted parabolic curve can be largely regarded as a baseline p_occ (c.a. 30%), and the two-fold increase in EPSCs during facilitation can be attributed to an increase of up to 60% in p_occ.

We investigated the time course of recovery from facilitation and post-tetanic augmentation (PTA) by applying dual-train stimulations (30 pulses at 40 Hz) separated by different inter-burst intervals (IBIs; Fig. 3C,D). The initial EPSCs in the second train were enhanced by more than twofold, and followed by a strong PPD, suggesting that fusion probability stays high in the IBIs and thus synaptic transmission is carried by TS vesicles. The decay time course of PTA was characterized by two phases (initially fast and later slow; Fig. 3E), which may reflect a two-phase Ca²⁺ decay or two distinct recovery processes that occur sequentially. The variance of potentiated EPSCs measured at IBIs of 0.1, 0.2, and 0.5 s was not different from that of EPSCs at the peak of facilitation (filled circles, Fig. 3A-B). Given that the same RRP mediates both facilitation and augmentation and that post-tetanic EPSCs is carried by TS vesicles similar to the baseline EPSCs, this plot suggests that PTA, similar to short-term facilitation, is mediated by an increase in the TS vesicle occupancy.

Failure rate analysis suggests that vesicular fusion probability is closed to one

The vesicle dynamics of STP has been explained using a simple refilling model (Hosoi et al., 2007; Neher and Sakaba, 2008). This model assumes that the vesicles are reversibly docked to a limited number of release sites (N) with forward and reverse rate constants (denoted as k₁ and b₁, respectively). Under the framework of the simple refilling model (see Materials and Methods), the rate constant for the refilling of the RRP after depletion by the first pulse is the sum of the baseline k₁ (k₁) and b₁. This is estimated as 23/s from the plot of PPR as a function of ISIs (Fig. S8A). From the baseline occupancy of 0.3 and p_occ = k₁ / (k₁ + b₁), we estimated k₁ to be 6.9/s.

When paired APs were applied with an ISI of 20 ms, the failure rate at 1^st pulse, P(F₁), was 10.6%, and the probability of two consecutive failures, P(F₁, F₂), was 6.2% (Fig. S8A). From the equation P(F₁) = (1 – p_r)^N, p_r was calculated as 0.312 when N = 6, comparable to p_r (= 0.32) estimated from the V-M analysis (Fig. 3A). We calculated P(F₁, F₂) under the framework of a simple refilling model, with p_v and k₁ set as free parameters and other parameters set according to the following relationships: P(F₁) = (1 – p_r)^N, p_r = p_v ·p_occ, and p_occ = k₁ / (k₁ + b₁).

The calculated values for P(F₁, F₂) and their difference from the observed value (6.2%) are shown in the plane of k₁ vs. p_v in Figure S8B-C. As previously mentioned, k₁ was greater than 6/s. Under this high k₁ condition, the difference between the calculated and observed values of P(F₁,F₂) was minimal when p_v = 1 (Fig. S8C).

Pharmacological manipulations reveal specific molecular identities underlying vesicle loading processes

Our results suggest that the baseline occupancy of TS vesicles is low and that the activity-dependent progressive overfilling of docking sites with TS vesicles is responsible for short-term facilitation and augmentation. To elucidate the specific molecular link between synaptic enhancement and Ca²⁺-dependent overfilling, we examined the effects of several drugs known to affect vesicle dynamics. We first examined the roles of phospholipase C (PLC) and diacylglycerol (DAG) by applying 5 μM U73122, a PLC inhibitor, or 20 μM 1-oleoyl-2-acetyl-sn-glycerol (OAG), a DAG analog. The STP at 40 Hz and PTA at 0.5-s interval were examined before and after applying each drug to the bath. For both drugs, the facilitation and augmentation were reduced (Fig. 4A-B). U73122 did not affect the basal EPSC amplitude (EPSC₁) but resulted in a stronger PPD, slower facilitation, and lower augmentation (Fig. 4A). These effects of U73122 suggest the involvement of Ca²⁺-induced PLC activation in progressive overfilling. In contrast, OAG increased EPSC₁ while maintaining a pronounced PPD (Fig. 4B), indicating an increase in the TS vesicle pool size. Meanwhile, OAG slowed k_STF and reduced augmentation, indicating occlusion of activity-dependent synaptic facilitation. Phorbol esters and DAG are thought to accelerate priming through Munc13 activation (Rhee et al., 2002; Taschenberger et al., 2016; Aldahabi et al., 2022); they also shorten the bridges that interlink docked vesicles with the active zone (AZ) membrane (Papantoniou et al., 2023). Collectively, these suggest that PLC- and DAG-mediated signaling play a role in the overfilling of docking sites underlying short-term facilitation and augmentation.

Figure 4.

Although the acceleration of Ca²⁺-dependent vesicle refilling supports rapid facilitation at 40 Hz, the later part of the delayed facilitation underwent a slight depression, which was more pronounced during the second burst (Fig. 3D, Fig. 4). Sustained synaptic transmission is limited by the spatial availability of docking sites because it is undermined by accumulation of exocytic remainings in the presynaptic AZ during HFS (Neher, 2010; Haucke et al., 2011). To verify this, we examined the effects of dynasore (100 μM), a specific inhibitor of endocytosis (Macia et al., 2006). The treatment with dynasore did not influence EPSC₁ or the early phase of facilitation, but it exacerbated depression during the later phase and attenuated PTA (Fig. 4C). This suggests that late synaptic transmission during the 40-Hz train is limited by site clearance.

Given the involvement of actin polymerization in the multiple steps of vesicle docking and recycling, we examined the effects of 20 μM latrunculin B (LatB), an inhibitor of actin polymerization. The effects of LatB were complicated: (1) EPSC₁ was decreased, whereas PPR was increased (Fig. 4Da) and (2) facilitation and augmentation were reduced (Fig. 4Db). The latter effect may result from defects in vesicle replenishment rates, either through the impaired physical movement of vesicles or disrupted site clearance (Sakaba and Neher, 2003; Lee et al., 2012; Hallermann and Silver, 2013; Miki et al., 2016), as Ca²⁺-dependent vesicle refilling mediates facilitation and augmentation (Figs. 2–3). Meanwhile, the former effect might be attributed to a shift in equilibrium in the priming states towards a loosely docked state (LS) at rest. It is unknown whether actin affects vesicular fusogenicity in neurons, although it facilitates multiple steps of vesicle fusion by enhancing membrane tension in neuroendocrine cells (Wu and Chan, 2022).

Short-term facilitation at both types of local excitatory synapses in L2/3 is abolished by Syt7 knockdown

We examined whether Syt7 mediates facilitation at the PC-PC and PC-FSIN synapses in L2/3 of the mPFC. To this end, oChIEF and shRNA against Syt7 (shSyt7) were co-expressed in L2/3 PCs using IUE. Syt7 KD resulted in a complete loss of facilitation at all tested stimulation frequencies (Fig. 5A-B). Short-term depression and paired-pulse depression were more pronounced as the stimulation frequency increased, and the frequency invariance of steady-state EPSC was abolished, implying little contribution of Ca²⁺-dependent vesicle recruitment (Fig. 5C-F). The lack of facilitation could not be attributed to the failure of APs in presynaptic axons, as optical stimulation elicited reliable APs in Syt7 KD PCs during the 40-Hz train (Fig. S2C). Syt7 KD had little effect on the basal properties, including the peak amplitude, rise/decay time, and time-to-peak, of single AP-induced EPSCs (Fig. S3D-E). Moreover, Syt7 KD did not significantly affect the intrinsic properties of neurons (Fig. S9).

Figure 5.

To examine whether Syt7 KD had any nonautonomous effects on postsynaptic cells, we measured the quantal size (q) from asynchronous events in the presence of Sr²⁺ and found that the mean quantal size remained unaltered at Syt7 KD synapses (Fig. S7C). This indicated that Syt7 did not influence postsynaptic parameters, including the density and kinetic properties of AMPARs. To test the specificity of shSyt7, we examined whether co-expression of the shRNA-resistant form of Syt7 restored synaptic facilitation in both excitatory synapse types. Presynaptic overexpression of shRNA-resistant Syt7 under the CAG promoter rescued the effect of shRNA at both PC-PC and PC-FSIN synapses (Fig. S10). These results underscore the crucial role of presynaptic Syt7 in mediating short-term facilitation at local excitatory synapses in L2/3 of the mPFC.

Syt7 KD synapses exhibit complementary changes in the number of release sites and their vesicle occupancy

To estimate the quantal parameters in Syt7 KD synapses, we performed V-M analysis of EPSCs evoked by train stimulation (Fig. 6A-B). Fitting a parabola to the V-M plot revealed higher baseline p_occ in Syt7 KD synapses than in wild-type (WT) synapses for both synapse types (KD, 0.68 and 0.42; WT, 0.32 and 0.31 for PC-PC and PC-FSIN synapses, respectively). The P_occ monotonously decreased during a train stimulation, as expected from short-term depression. Moreover, the number of release sites (N) was markedly lower in Syt7 KD synapses (N = 3.5) than in WT synapses (N = 5.3; Fig. 3A). Next, we examined the PTA induced by the same protocol as in Figure 3 (two 40-Hz train stimulations separated by variable IBIs) at Syt7 KD synapses (Fig. 6C). The 40 Hz train-induced synaptic depression quickly recovered during IBIs, but no PTA was observed at Syt7 KD synapses, indicating its crucial role in augmentation (Fig. 6D-E).

Figure 6.

TS vesicle recovery following depletion is accelerated by Syt7

Syt7 is recognized not only for its crucial role in facilitation, but also for mediating Ca²⁺- dependent replenishment of releasable vesicles (Liu et al., 2014; Tawfik et al., 2021). Given that facilitation at L2/3 excitatory synapses is driven by progressive overfilling (Figs. 2-3), the loss of facilitation at Syt7 KD synapses in the present study implies that Syt7 plays a key role in the Ca²⁺-dependent acceleration of refilling and overfilling of the docking sites with TS vesicles. To test this hypothesis, we examined the recovery kinetics of EPSCs after TS vesicles were depleted by 3 pulses at 40 Hz in wild-type (WT) and Syt7 KD synapses (Fig. 7). Strong PPD was observed not only in the first burst, but also in the second burst. This suggested that EPSCs in the second burst were mediated by TS vesicles, similar to those in the first burst (Fig. 7A,D). Therefore, EPSC₁ in the second burst can be interpreted as a release from TS vesicles recovering during a given IBI. The recovery time course of EPSC₁ in the second burst indicated full recovery of TS vesicles within 100 ms in both excitatory synapse types in WT rats (Fig. 7C,F). However, this recovery was remarkably slower in Syt7 KD synapses, requiring 5 s for full recovery, indicating that TS vesicle recovery was greatly accelerated by Syt7. Overall, synaptic facilitation and augmentation at local excitatory synapses in L2/3 can be ascribed to progressive activity-dependent increases in the occupancy of TS vesicles (Figs. 2-3), in which Syt7 plays a key role.

Figure 7.

Behavioral consequences of Syt7 KD in L2/3 PCs of the mPFC

We previously proposed that the disparity in STP between the PC-PC and PC-IN synapses results in activity-dependent changes in the ratio of synaptic weights at the PC-PC and PC-IN synapses (J_ee/J_ie). This may in turn have a profound influence on persistent activity in a recurrent network by biasing the E-I balance (Yoon et al., 2020). Syt7 KD not only eliminated facilitation, but also the activity-dependent increase in the J_ee/J_ie ratio in the L2/3 network (Fig. S11). This suggests the possibility that Syt7 contributes to the persistent activity during working memory tasks in the L2/3 recurrent network. Trace fear conditioning (tFC) requires associative learning between an auditory cue (conditioned stimulus, CS) and temporally separate aversive events (unconditioned stimulus, US). The formation of trace fear memory requires the prelimbic area of the mPFC and hippocampus. mPFC activity during the trace interval is thought to temporarily hold CS information, enabling the network to associate temporally discontinuous events (i.e., temporal associative learning) (Gilmartin et al., 2013). Synaptic facilitation and augmentation are implicated in temporary memory retention in recurrent networks (Mongillo et al., 2008; Mongillo et al., 2012).

Given that Syt7 is crucial for synaptic facilitation (Fig. 5) and augmentation (Fig. 6) in prelimbic L2/3 recurrent excitatory synapses, we tested whether trace fear memory formation was impaired in rats with depleted Syt7 transcripts, specifically in the L2/3 PCs of the mPFC. Syt7-targeted shRNA (shSyt7) or scrambled shRNA (Scr) constructs were transfected bilaterally into the L2/3 PCs of the mPFC using IUE (Fig. 8A). Both the Scr-transfected control and shSyt7-transfected rats underwent tFC training (Fig. 8B). The following day, the rats were subjected to a tone test in which trace fear memory formation was assessed from the freezing response to a tone alone in a distinct context. Compared to the control group, the shSyt7-transfected group exhibited significantly lower levels of freezing behavior in response to auditory cues during the first four trials, suggesting an impairment in trace memory formation (Fig. 8C).

Figure 8.

The tone test was repeated the following day to test for extinction memory formation. Syt7 KD animals exhibited much less freezing to the CS (+1d) than did WT animals. Considering the Rescorla-Wagner model (Yau and McNally, 2023), our results suggest that the strength of CS-US associative memory was weaker in KD animals than in WT animals; thus, the acquisition of trace fear memory was impaired by Syt7 KD in the L2/3 PCs of the mPFC.

Next, we assessed the locomotor activity and anxiety levels of rats using the open field test (OFT) and elevated plus maze test (EPM). Scr controls and Syt7 KD rats showed similar exploratory patterns, displaying comparable total distance moved and time spent in the periphery of the OFT or the open/closed arms of the EPM (Fig. S12). Previous studies have suggested that the acquisition of trace fear memory requires elevated trace activity in prelimbic PCs (Gilmartin et al., 2013). Given that c-Fos, an immediate early gene, is upregulated following patterned activities in neurons, its expression is considered a reliable indicator of recent neuronal activation (Fields et al., 1997; Guzowski et al., 2005). To investigate the effect of Syt7 on neuronal activity in vivo during tFC, animals were sacrificed at the time of peak c-Fos protein expression, approximately 90 min after completion of the behavioral task (Arime and Akiyama, 2017).

To elucidate the activation patterns of specific neuronal populations, we performed immunohistochemical co-labeling of c-Fos and GAD67 (a GABAergic interneuron marker) upon tFC acquisition. The density of c-Fos(+) cells and the percentage of GAD67(+) cells among c-Fos(+) cells in all prelimbic layers were significantly lower in Syt7 KD rats than in control rats (Fig. 8F-G). Additionally, analysis of the proportion of active cells among transfected L2/3 PCs (c-Fos+/GFP+) revealed that Syt7 KD L2/3 PCs were significantly more active than Scr-transfected PCs (Fig. 8H). This unexpected hyperactivity of Syt7 KD PCs may be attributed to the high reciprocal connectivity between PCs and FSINs in the neocortex (Holmgren et al., 2003; Otsuka and Kawaguchi, 2009), as PCs with short-term depression would transmit less input to FSINs, which in turn would exert less feedback inhibition on reciprocally connected PCs. Collectively, these findings show that the slow activation of vesicle refilling supports various forms of facilitation at excitatory synapses in PFC L2/3, which is essential for neuronal activity during temporal associative learning.

Discussion

The present study characterized STP at local excitatory synapses in the L2/3 network of the rat mPFC at physiological extracellular [Ca²⁺]. These synapses displayed initial strong depression and then delayed facilitation at 40 Hz, while monotonous slowly developing facilitation was observed at lower frequencies. Our study suggests that such delayed facilitation after a brief depression results from a high p_v (Figs. 2-3 and Fig. S8), along with Ca²⁺-dependent delayed acceleration of vesicle refilling and overfilling (Figs. 2–3). The two-fold increase in EPSC during 20 pulse train stimulations under the condition of such a high p_v suggests a Ca²⁺-dependent increase in the vesicle replenishment rate during a train; this then leads to overfilling of release sites beyond their incomplete basal occupancy (Fig. 3A,B). The high p_v and strong PPD did not result from artificial bouton stimulation, AMPAR desensitization, or any potential distortion caused by optogenetic methods (Figs. S4 and S5). Additionally, similar to short-term facilitation, PTA could be explained by an increase in the TS vesicle occupancy (Fig. 3C-E), consistent with previous studies showing post-tetanic increases in the p_occ and/or vesicle pool size at various types of synapses (Lee et al., 2010; Vandael et al., 2020; Tran et al., 2023; Silva et al., 2024). These findings suggest that release sites are partially occupied at rest and that a progressive overfilling of release sites with TS vesicles underlies facilitation and augmentation at intracortical L2/3 excitatory synapses.

Progressive overfilling of docking sites

Conventional models for facilitation assume an increase in p_v after an AP of RRP vesicles, which had been in a low_p_v or reluctant state at rest (Dittman et al., 2000; Turecek et al., 2016). This facilitation model is unlikely to explain our results, because TS vesicles are predominantly released not only at rest but also in the facilitated state during or after train stimulation, as evidenced by the strong PPD (ISI, 25 ms) during a 5-Hz train (Fig. 2D-E) and at different intervals after a 40-Hz/30 pulse trains (Fig. 3D). Moreover, the PPD (ISI, 25 ms) remained pronounced during the early phase of facilitation (Fig. 7), arguing against the possibility of slowly increasing the p_v of reluctant vesicles. These results suggested that facilitation was mediated by an activity-dependent progressive increase in occupancy of TS vesicles (TS occupancy). Facilitation through an increase in TS occupancy has been termed frequency facilitation, differentiating it from paired pulse facilitation that relies on a residual calcium-dependent transient increase in the p_v of releasable vesicles (Neher, 2024).

Assuming an increase in TS occupancy, to attain a two-fold increase in EPSC (Fig. 1), the baseline occupancy of the docking sites by TS vesicles should be low, as estimated from the V-M plot (c.a. 30% in Fig. 3). However, it is unclear whether the remaining docking sites that are not occupied by TS vesicles are truly vacant or are occupied by reluctant LS vesicles that are not released by the AP or AP trains. TS vesicles can be supplied by the transformation of LS to TS vesicles, both of which reside at the same docking site (LS/TS model; reviewed in Neher, 2024). Alternatively, TS vesicles can be supplied from a distinct replacement pool (RS/DS model; reviewed in Silva et al., 2021). Further studies, including ultrastructural analyses of active zones, are required to address this issue.

Given the crucial role of Syt7 in facilitation (Fig. 5), the facilitation mechanisms are closely related to how Syt7 mediates synaptic facilitation. Based on the slow kinetics of Syt7, previous computational modeling studies have proposed that Syt7 may mediate short-term facilitation through an activity-dependent increase in p_v (Turecek et al., 2016; Jackman and Regehr, 2017; Turecek et al., 2017; Norman et al., 2023). The present study, however, found that both facilitation and augmentation were mediated by overfilling of the release sites (Figs. 2–3). Moreover, Syt7-KD synapses displayed slower recovery of TS vesicles (Fig. 7). These results suggest that Syt7 is essential for activity-dependent overfilling of docking sites with TS vesicles. Whereas the baseline EPSC was not altered (Fig. S3), and complementary changes in the number of docking sites and their baseline occupancy were observed in Syt7 KD synapses (Fig. 6). These results raise a possibility that Syt7 may play a role in providing additional vacant docking sites that can be overfilled during facilitation. It remains to be elucidated whether the decrease in the number of docking sites at Syt7 KD synapses is related to the subcellular localization of Syt7 to the plasma membrane of the active zone (Sugita et al., 2001; Vevea et al., 2021).

STP model in light of known Ca²⁺ binding kinetics of Syt7

Syt7 acts as an upstream Ca²⁺ sensor that facilitates activity-dependent replenishment (Liu et al., 2014; Bacaj et al., 2015; Chen et al., 2017; Tawfik et al., 2021). The present study shows that EPSCs are mediated by TS vesicles both at rest and during activity-dependent facilitation. Releasable vesicles are homogeneous, and Syt7-dependent overfilling is responsible for the activity-dependent enhancement of EPSCs. Thus, we tested whether slowly developing facilitation could be replicated within the framework of a simple refilling model based on the known Ca²⁺ binding kinetics of Syt7 (Brandt et al., 2012). To this end, we made following assumptions: (1) a Ca²⁺-dependent increase in the forward refilling rate (k₁) requires a full Ca²⁺-bound form of Syt7, to which multiple Ca²⁺ ions can bind (two or three Ca²⁺ ions on each C2A and/or C2B domain) (Brandt et al., 2012; Chon et al., 2015; Voleti et al., 2017); (2) sequential Ca²⁺ binding steps are synergistic through a cooperativity factor, b; (3) Syt7 senses AP-induced local [Ca²⁺] similar to that estimated at the calyx of Held (a Gaussian function with a full width at half maximum value of 0.2 ms and a peak of 40 μM) (Wang et al., 2008); and (4) residual [Ca²⁺] follows a mono-exponential function with an amplitude of 1 μM and decay time constant of 50 ms (Jackson and Redman, 2003).

Under these assumptions and the framework of the simple refilling model, slow and delayed facilitation could be replicated by modeling k₁ as k₁ = k_1,b + K_1,max × [Syt7: n Ca²⁺], where [Syt7: n Ca²⁺] represents the fraction of fully Ca²⁺-bound Syt7 (Fig. 9A-C). Syt7 was slowly activated over the course of the stimulus train, suggesting that the cooperative binding of multiple Ca²⁺ ions progressively increased the putative active form of Syt7 because of its slow membrane binding/unbinding kinetics. Nevertheless, the late phase of facilitation at high frequencies was predicted to be higher in the simulation than in the experiment. As expected, based on the notion that sustained release during HFS is limited by site clearance in the active zone (Fig. 4C), the model/data discrepancy became more pronounced at higher frequencies. Incorporating endocytic terms into the model may mitigate late-phase discrepancies.

Figure 9.

Moreover, removal of the catalytic function of Syt7 in the model effectively accounted for the complete loss of facilitation observed in Syt7 KD (Fig. 9D).

Possible mechanisms underlying the depression-facilitation sequence

Cooperative binding of multiple Ca²⁺ ions to Syt7 predicts delayed activation and a progressive increase in vesicle refilling during train stimulations, resulting in delayed facilitation. Combining this with a high p_v could replicate not only the slow facilitation at 5 Hz, but also the unique depression-facilitation sequence observed at 40 Hz (Fig. 9). As simulated by Pulido and Marty (2018), delayed facilitation can be reproduced by a high p_v in combination with a low baseline occupancy of replacement sites and subsequent Ca²⁺-dependent refilling. However, whether this model is also able to replicate slow facilitation at low frequencies remains to be tested. An alternative, but not exclusive, possibility for delayed facilitation is that the presynaptic global Ca²⁺ itself may undergo a slow buildup during a train due to the saturation of endogenous Ca²⁺-binding proteins.

Facilitation requires not only Syt7, but also activation of PLC-DAG signaling

To gain insight into the specific molecular mechanisms underlying the vesicle-loading processes, we examined the effects of pharmacological inhibitors targeting PLC and DAG (Fig. 4). Both a PLC inhibitor and a DAG analog reduced facilitation and augmentation. This suggests that the progressive overfilling of docking sites with TS vesicles is supported by the PLC-DAG signaling pathway and probably downstream Munc13 (Rhee et al., 2002; Lou et al., 2008). Phorbol esters have been recently shown to increase the TS fraction of docked vesicles by enhancing vesicle priming without affecting Ca²⁺ transients or p_v (Taschenberger et al., 2016; Aldahabi et al., 2022; Papantoniou et al., 2023; Aldahabi et al., 2024). In line with these findings, signaling involving PLC and DAG may support Ca²⁺-dependent overfilling of TS vesicles by enhancing molecular priming processes.

Roles of facilitation and augmentation in working memory and related tasks

Previous theoretical studies have proposed that working memory in a recurrent network can be maintained in the form of augmentation at synapses (Mongillo et al., 2008) and/or persistent activity of neurons in a memory ensemble. This is supported by the facilitation of recurrent excitatory synapses or disparity in STP between excitatory and inhibitory synapses. However, whether these short-term enhancements in synaptic efficacy contribute to working memory-related behaviors has not been examined. tFC is a temporal associative learning task requiring temporary retention of CS information during a trace period. The mPFC and hippocampus are crucial for tFC, and tFC is interfered with by a load of working memory and attention distraction, implying that the tFC shares the same neural substrate with other working memory tasks. In this study, tFC in WT and Syt7 KD animals revealed that Syt7 deficiency in the L2/3 pyramidal neurons of the mPFC impaired the acquisition of trace fear memory and reduced c-Fos expression associated with tFC training (Fig. 8). Given that synaptic enhancements in prelimbic L2/3 recurrent excitatory synapses can be attributed to an increase in p_occ, these behavioral results suggest that presynaptic high Ca²⁺ transients are temporally memorized in the presynaptic terminal as an increase in TS occupancy.

Conclusions

Progressive overfilling of synaptic vesicles with a high fusion probability supports the facilitation and augmentation of recurrent excitatory synapses within the L2/3 network of the mPFC and is essential for temporal associative learning.

Materials and methods

Animals/Subjects

All experiments were carried out on Sprague-Dawley rats of either sex at p28-43. Rats were maintained in temperature-controlled rooms on a constant 12 h light/12 h dark cycle and were given access to water and food ad libitum. All animal procedures were approved and performed in accordance with the Institutional Animal Care and Use Committee at Seoul National University (SNU200522-1).

In utero electroporation (IUE)

Pregnant Sprague Dawley rats on embryonic day (E) 17.5 were deeply anesthetized with 5% (v/v) isoflurane for the duration of surgery. The uterine horns were exposed by laparotomy and wet with warm sterile phosphate-buffered saline (PBS). The plasmids (0.5-2 µg/µl) together with the 0.01% fast green dye were injected into either the left or the right lateral ventricle or bilateral ventricles of embryos through a thin glass capillary (Narishige PC-10). The embryo’s head was carefully held between the forceps-shaped electrodes of 10 mm diameter (CUY65010; NEPAGENE, Japan). Electroporation pulses (50 V for 50 ms) were delivered five times at intervals of 150 ms using a square-wave generator (ECM830; BEX, United States). The uterine horns were placed back into the abdominal cavity and abdominal muscles and skin were sutured. After the surgery, the animals were recovered under the infrared lamp. For knockdown experiments in slice electrophysiology, the shRNA constructs were co-electroporated with the pCAG-oChIEF-tdTomato, because of the consistent co-expression of the two constructs in a significant proportion of the transfected cells in vivo (Bony et al., 2013). Pups with robust fluorescence signals in the prefrontal cortex of both hemispheres were used for behavior experiments.

Preparation of vector constructs

For an optic stimulation, pCAG-oChIEF-tdTomato was transfected by IUE. Short hairpin RNA (shRNA) against Syt7 sequence (GATCTACCTGTCCTGGAAGAG) was expressed under the control of an U6 promoter of pAAV-U6-GFP vector (Cell Biolabs, #VPK-413). This shRNA was shown to effectively knockdown Syt7 in cell cultures and in vivo by previous reports from more than one research group (Bacaj et al., 2013; Li et al., 2017). Scrambled control sequence (TCGCATAGCGTATGCCGTT) was designed by shuffling the recognition region sequences and a BLAST analysis confirmed that the control shRNA sequence had no target gene. For rescue experiments, the cDNA for rat Syt7 (NM_021659) rendered resistant to the shRNA were inserted in the KD vector and their expression was driven by the CAG promoter; the vector also contained a self-cleaving P2A peptide sequence followed by EGFP, which ensures visualization of transfected cells. All the constructs above were verified by DNA sequencing. Plasmid DNA was purified and concentrated under endotoxin free conditions (Qiagen EndoFree Maxi Kit).

Slice preparation

Acute brain slices were prepared by using a VT1200S Vibratome (Leica Microsystems) after isoflurane anesthesia and decapitation. Coronal sections of the mPFC (300 μm in thickness) were cut in ice-cold artificial cerebrospinal fluid (aCSF) composed of (in mM) 125 NaCl, 3.2 KCl, 25 NaHCO₃, 1.25 NaH₂PO₄, 20 D-glucose, 2 NaPyr, 1 MgCl₂, and 1.3 CaCl₂ (Yoon et al., 2020), bubbled with 95% O₂ and CO₂ (pH = 7.3; 310 mOsm). Slices were allowed to recover for a minimum of 30 minutes at 36 °C, and subsequently maintained at room temperature until recording. Composition of aCSF was identical throughout the preparation, incubation, and recording periods unless otherwise stated. For the recordings, the mPFC slices were continuously perfused with oxygenated aCSF at a flow rate of 1.5 ml/min, and the temperature was maintained at 31–32 °C using an in-line temperature controller (Warner Instruments).

Electrophysiology

Whole-cell patch-clamp recordings were performed from pyramidal cells and fast-spiking interneurons in layer 2/3 of the prelimbic mPFC. Neuronal subtypes were initially selected based on morphology under visual guidance by using an upright microscope equipped with infrared differential interference contrast (IR-DIC) optics (BX51WI, Olympus), and were further identified by their electrophysiological properties (see Supplementary Fig. 1; van Aerde and Feldmeyer, 2015; Tremblay et al., 2016). To analyze these intrinsic properties in current-clamp experiments, we measured the following parameters: (1) resting membrane potential (RMP), (2) input resistance (R_in), (3) sag ratio, and (4) the F-I curve, which plots firing frequencies (F) against the amplitude of injected currents (I). Both cell types exhibited a low sag ratio below 0.1, consistent with findings from other studies (van Aerde and Feldmeyer, 2015; Yoon et al., 2020). PCs and FSINs were easily distinguished by their unique firing frequency, Rin, spike adaptation, RMP and morphological characteristics including the shape of soma and the thickness of apical dendrites. For experiments involving oChIEF-expressing neurons in WT or Syt7 KD rats, whole-cell recordings were performed from the fluorescently labeled pyramidal neurons in layer 2/3 of the prelimbic cortex using suitable filter sets. Patch pipettes (tip resistance of 2–3 MΩ) were pulled from borosilicate glass and filled with (in mM) 130 K-gluconate, 8 KCl, 10 HEPES, 0.2 EGTA, 20 Na₂Pcr, 0.3 Na₂GTP, 4 MgATP. Intracellular solutions were adjusted to pH 7.25 and 300–310 mOsm. Neurons were voltage clamped at −78 mV and series resistance (Rs) were continuously monitored during the EPSCs recordings. Experiments were discarded before analysis if the Rs changed by 20% or larger deviation of baseline value. Measurements were not corrected for Rs compensation, bridge balance, or liquid junction potential. Electrophysiological data were low-pass-filtered at 1 kHz (Bessel) and acquired at 20 kHz using a Multiclamp 700B amplifier paired with Digidata 1440A digitizer and Clampex 10.2 software (Molecular Devices).

Synaptic stimulation

IUE resulted in sparse expression of channelrhodopsin (10-20%) (Fig. 1A). Exploiting the sparse expression, we employed a minimal optical stimulation method to activate a single or very few excitatory synapses in layer 2/3. The optic minimal stimulation was performed by our published protocol with minor modifications (Yoon et al., 2020). This approach aimed to minimize the risk of synaptic contamination, which could arise from engaging extra terminals that were initially inactive (depolarized below threshold) but became active by temporal summation during high-frequency trains. It also helped prevent the buildup of depolarization at the synaptic terminals, which might influence the probability of release, as noted by (Jackman et al., 2014). To achieve this, we employed a collimated digital micromirror device (DMD)-coupled LED (Polygon400; Mightex Systems) to confine 470 nm blue light to a small area with a radius of approximately 3–8 μm (typically 3–4 μm), as measured at the focal plane of a 60× water immersion objective [numerical aperture (NA) = 1.0; LUMPlanFL, Olympus] (Fig. S3A). To find the minimal stimulation area, the radius of illumination area was increased from 2 μm by 50% at each step (Fig. S3A), and selected the smallest illumination area that elicited EPSCs. Setting the duration of illumination between 3 and 5 ms (typically 5 ms), photostimulation onto oChIEF-expressing cells reliably induced action potentials (APs) during 600-pulse trains (see Fig. S2).

STP at various stimulus frequencies (5, 10, 20, and 40 Hz) was examined using 20-pulse trains. Trials of train stimulation were repeated with a prolonged interval of 30 s to avoid change in baseline properties. We observed no systematic changes in the amplitudes of the first EPSC of each train, nor did we find significant differences in the paired-pulse ratio (PPR) across stimulation trials in the same cell. For each stimulus frequency, approximately ten to twenty consecutive traces were collected and then averaged across trials to yield a mean EPSC trace. For recovery experiments, two consecutive trains of 3 or 30 stimuli at 40 Hz, with varying time intervals (0.1-10 s), were applied every 10 s or 60 s, respectively, at both excitatory synapses. For 600-stimulus trains at 10Hz, the trial was delivered only once for each cell to avoid possible after-effects caused by prolonged synaptic stimulation (Hirsch and Crepel, 1990). Prior to the stimulus trains, the EPSCs were normalized to the mean amplitude of the baseline. For the baseline measurements, paired pulses with various ISIs (20-200 ms) were delivered every 3 s (1/3 Hz) at least 20 times.

For all recordings, picrotoxin (PTX, 10 μM) was included in the recording solution to isolate glutamatergic responses. mPFC PC terminals were entirely glutamatergic as the application of 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX, 20 μM), an AMPA receptor antagonist, completely abolished the monosynaptic events evoked by photostimulation. A low concentration of PTX was employed to minimize the relieving effects on spike suppression caused by blockade of GABAa receptors. We confirmed that no synaptic response was elicited by electrical stimulation in the presence of both 10 μM PTX and 20 μM CNQX.

In Fig. S4A, 1 µM TTX (1 mM stock solution) and 0.1 mM 4-AP (500 mM stock solution) were added to the bath solution to block action potentials and restore presynaptic glutamate release, respectively (Fig. S4A). In Fig. S5A, 50 µM of cyclothiazide, a positive allosteric modulator of the AMPA receptors, was included to minimize rapid desensitization of the postsynaptic receptors. In Fig. S5B, 0.5 mM kynurenate was added to bath solution to prevent possible saturation of AMPA receptors, especially during maximal synaptic facilitation. For pharmacological manipulations of synaptic vesicle dynamics (Fig. 3), drugs related to vesicle supply including U73122, OAG, LatB, EGTA-AM, and dynasore at indicated concentrations were added to the recording solution during baseline measurements or PTA/recovery experiments (IBI = 0.5 s). Chemicals were from Sigma-Aldrich, Merck, or Tocris.

Measurements of quantal size

Experiments for measuring asynchronous release were performed as previously described (Yoon et al., 2020). In brief, we replaced extracellular Ca²⁺ entirely with 2 mM Sr²⁺ to induce asynchronous release. A single stimulus was delivered at 3-second intervals over 100 repetitions, and asynchronous release events were detected during a 100-ms window starting 25 ms after stimulus onset for each trial. The same procedure was also executed under sham conditions (0% light intensity) to serve as a control, thus mitigating contamination from nonspecific spontaneous events and potential errors associated with in the detection algorithm. q was determined as the mean value from the first Gaussian function after fitting the EPSC distribution with either a single or double Gaussian function (Fig. S8).

Calculation of probability for double failure

The failure rate at the 1st pulse when initial RRP size (n₁) = i can be calculated as: P(F₁ | n₁=i) = (1 - p_v)ⁱ. The probability that initial RRP size is i was calculated as: P(n₁=i) = C(N, i) * p ⁱ * (1-p_occ)^(N-i), where p_occ = mean release site occupancy, and N = total number of release sites. Under the conditions of no vesicle release at the 1st pulse (F₁) and n₁ = i, the probability for the second failure is

, where n₂ = number of vesicles just before the 2nd pulse. When forward and backward rate constants are k₁ and b₁, respectively, P(n₂|n₁=i, F1) is calculated as follows:

, where Δt = ISI. By the Bayes rule,

Thus, the probability for two consecutive failures can be calculated as:

P(F₁) and N are observed values. The k₁ and p_v were set free variables. Other parameters were set according to following relationships: P(F₁) = (1 – p_r)^N, p_r = p_v p_occ, and p_occ = k₁ / (k₁ + b₁).

Behavioral tests

Trace fear conditioning

For trace fear conditioning (tFC) (Gilmartin et al., 2013), individual rats were placed in the training chamber (30.5 x 25.4 x 30.5 cm; Coulbourn Instruments) consisting of metal grid floor connected to an electrical stimulator (H10-11R-TC-SF, Coulbourn Instruments) for delivery of a footshock. After a 5 min habituation period, rats received five pairings of a 20 s tone conditioned stimulus (CS, 2.5 kHz, 75 dB), followed by empty 20s trace period and a unconditioned stimulus (US; 1 s, 0.7 mA footshock) with a pseudorandom inter-trial interval (ITI) of 240 ± 20 s. During this training session, rats learned to associate the auditory CS with the shock US. The next day, the rats were tested for memory of CS-US trace association (called tone test). The acquisition of fear learning, often indicated by the extent of freezing to the CS, was assessed in a novel chamber (a white hexagonal enclosure) in a dimly lit room. During this tone (or retrieval) test, rats were first allowed to explore the new context for 5 min (habituation), followed by twelve 20 s CS presentations with a variable ITI of 120 ± 20 s. This entire session was repeated on the next day for assessment of extinction learning. Identities of the chambers were also determined with the presence of distinct odors; the training and testing chambers were cleaned with 70% ethanol and 1% acetic acid respectively before and after each session. For all experimental sessions, the activity of animals in the chambers was recorded at 30 frames per second using the EthoVision XT (Noldus Information Technology, Wageningen, Netherlands), and stored as a video file. The freezing behavior, defined as behavioral immobility other than respiratory movement, was assessed using an open-source video analysis pipeline ezTrack (Pennington et al., 2019). The percentage of freezing was calculated as a total duration of freezing divided by the total duration of observation.

Elevated Plus Maze

Elevated plus maze (EPM) test was performed using a plus-shaped apparatus elevated 60 cm above the ground. The maze consisted of two open arms (50 x 10 x 0.5 cm (H)), two closed arms (50 x 10 x 40 cm (H)) and a center area (10 x 10 cm). Rats were individually placed on the center area, facing an open arm, and the path of the animal was recorded with a video camera for 5 min. We analyzed the total distance traveled for 5 min and time spent in each arm. Animals were tracked offline using the open-source tracking system ezTrack (Pennington et al., 2019).

Open field test

Open field exploration test (OFT) was performed in an open field apparatus in a dimly lit room. The open field consisted of a white plastic board (1.20 m in diameter) surrounded by white plastic walls (50 cm in height). Individual rats were placed in a center zone (0.6 m in diameter), and allowed to freely explore the apparatus for 10 min. Time spent in the center zone and total distance traveled were analyzed for initial 5 min. Animals were tracked offline using the open-source tracking system ezTrack (Pennington et al., 2019).

Histology and immunohistochemistry

At the end of the behavioral experiments, rats were anesthetized using isoflurane inhalation and transcardially perfused with PBS followed by 4% paraformaldehyde (PFA, T&I, Korea). Brains were extracted and post-fixed in 4% PFA solution for 24 h. Perfused brains were sliced at 100 μM for estimation of GFP expression or 50 μM for immunohistochemistry. For analysis of c-Fos expression, rats were returned to their home cages immediately after the training session and perfused approximately 90 minutes later (Arime and Akiyama, 2017). Brains were then cryopreserved, frozen, and sectioned coronally. Immunofluorescence staining was conducted on free-floating slices using a rabbit monoclonal anti c-fos (1:1000, Cell Signaling Technology, cat# 2250s) and mouse monoclonal anti-GAD67 (1:500, Millipore, cat# mab5406). Fluorescent secondary antibodies included Cy5-conjugated goat anti-rabbit IgG (1:1000, Abcam, cat# ab97077) and Cy3-conjugated goat anti-mouse IgG (1:1000, Abcam, cat# ab97035). Brain slices were permeabilized and blocked with PBS containing 0.5% Triton X-100 and 5% normal goat serum (NGS) for 1h at room temperature. Slices were incubated overnight at 4 °C with primary antibodies, followed by 4 h incubation with secondary antibodies at room temperature. Slices were then mounted with Antifading Mounting medium (Abcam, cat# ab104139). Confocal images were acquired using a Leica TCS SP8 microscope set to the same laser intensity and acquisition parameters to compare the immunosignals in sections from Scr and Syt7 KD groups. For cell counting and colocalization, fluorescent protein-positive cells were automatically quantified with spot-detection algorithms in Imaris 9.5 (Oxford instruments).

Data analysis

Data were analyzed and presented using ClampFit (Molecular Devices), Igor Pro (Wavemetrics), MATLAB (Mathworks), and Prism (GraphPad). Data points and error bars represent the mean ± standard error of the mean. All statistical analyses were conducted using two-tailed comparisons. Sample sizes and statistical tests for each group are detailed in the figure legends.

Numerical integration of the simple refilling model

This model posits reversible docking of vesicles at finite release sites (N_max) with forward (k₁) and reverse rate constants (b₁) (Hosoi et al., 2007; Neher and Sakaba, 2008).

According to this scheme, dn/dt = k1 u - b1 n - p_v n δ(t - t_AP), where u is the number of unoccupied sites (= N_max - n); t_AP is the timing of AP firing; and δ is the Dirac delta function. The p_v was assumed to be unity (Fig. S8). The sum of basal k₁ (= k_1,b) and b₁ was be estimated as 23/s from the dependence PPR on ISIs (Fig. S8Ab). From the baseline occupancy of 0.3 and p_occ = k₁ / (k₁ + b₁), we could estimate baseline k₁ (k_1,b) and b1 as 6.9/s, and 16.1/s, respectively. Ca²⁺ and Syt7-dependent increase of k was modeled as k = k + K [Syt7: n Ca²⁺], where [Syt7: n Ca²⁺] represents the fraction of full Ca²⁺-bound Syt7. For deterministic simulations, we time-integrated a set of differential equations for each model using Euler methods with a time step of 1 ms. All calculations were performed on the platform of Matlab (R2022b, Mathworks, USA).

Acknowledgements

We thank Dr. E. Neher and Dr. A. Marty for critical reading of this manuscript and invaluable comments.

Additional information

Author contributions

All experiments were carried out at cell physiology lab under Y.K and S.H.L’s supervision. Conception or design of the work: J.S. and S.H.L. Acquisition, analysis or interpretation of data for the work: J.S, Y.K. and L.S.H. Drafting the work or revising it critically for important intellectual content: J.S., S.Y.L, Y.K. and L.S.H.

Funding

This study was supported by grants from the National Research Foundation of Korea (RS-2024-333669 to S-H. Lee; 2021R1I1A1A01059646 to Y. Kim), and Seoul National University Hospital (2024).