Synaptotagmin-7–mediated activation of spontaneous NMDAR currents is disrupted in bipolar disorder susceptibility variants

Synaptotagmin-7 (Syt7) plays direct or redundant Ca2+ sensor roles in multiple forms of vesicle exocytosis in synapses. Here, we show that Syt7 is a redundant Ca2+ sensor with Syt1/Doc2 to drive spontaneous glutamate release, which functions uniquely to activate the postsynaptic GluN2B-containing NMDARs that significantly contribute to mental illness. In mouse hippocampal neurons lacking Syt1/Doc2, Syt7 inactivation largely diminishes spontaneous release. Using 2 approaches, including measuring Ca2+ dose response and substituting extracellular Ca2+ with Sr2+, we detect that Syt7 directly triggers spontaneous release via its Ca2+ binding motif to activate GluN2B-NMDARs. Furthermore, modifying the localization of Syt7 in the active zone still allows Syt7 to drive spontaneous release, but the GluN2B-NMDAR activity is abolished. Finally, Syt7 SNPs identified in bipolar disorder patients destroy the function of Syt7 in spontaneous release in patient iPSC-derived and mouse hippocampal neurons. Therefore, Syt7 could contribute to neuropsychiatric disorders through driving spontaneous glutamate release.


Introduction
Neurotransmitter release is driven by an increase in intracellular Ca 2+ . Three types of Ca 2+dependent release have been identified: Action potential (AP) evoked fast synchronous release, slow asynchronous release, and AP-independent spontaneous release [1,2]. All these types are triggered by the orchestration of the SNARE complex, consisting of the v-SNARE synaptobrevin-2 (Syb2), the t-SNAREs syntaxin-1A (Stx1a) and SNAP-25, and their respective Ca 2+ sensor proteins [3][4][5][6][7][8][9][10]. The synaptotagmins are a family of proteins that contain a transmembrane domain and 2 tandom C2 domains called C2A and C2B. Synaptotagmin-1 (Syt1), Syt2, and Syt9 have been suggested to be the Ca 2+ sensors for fast synchronous release, and the mechanisms of synaptic vesicle (SVAU : PleasenotethatSV hasbeendefinedassynapticvesicleatitsfirstmen ) fusion have been elucidated during the past decade through investigating the interaction between Syt1 and the SNARE complex [6,11,12]. The Ca 2+ sensor family for the slow asynchronous release contains 2 players, Doc2 and Syt7 [7,8]. Doc2 is a cytosolic protein containing 2 C2 domains without a transmembrane domain [13,14] and likely contacts most SVs that are tightly docked or primed at the active zone (AZ)  Syt1 and Doc2 have been suggested to be Ca 2+ sensors for spontaneous SV release [9,16]. Previously, it has been suggested that Syt7 and Syt1 were redundant Ca2+ sensors for postsynaptic AMPAR trafficking that underlie an important mechanism for the induction of LTP [22]. Moreover, the robust spontaneous release in Syt1-deficient neurons has suggested that spontaneous release is likely triggered by redundant mechanisms [16]. Although it has been reported that Syt7 KO hippocampal neurons have no apparent changes in the rate of spontaneous release [19], we were interested to know whether Syt7 could be a redundant Ca 2+ sensor with Syt1 and Doc2. To this end, we designed a CRISPRi-based gene KD system to achieve multiplex inactivation of Ca 2+ sensors, including Syt1, Doc2a, Doc2b, and Syt7 (Fig 1A). We expressed the CRISPRi system in cultured wild-type (WTAU : PleasenotethatWThasbeendefinedaswild ) hippocampal neurons through lentiviral infection. Quantitative reverse transcription PCR (qRT-PCR) analysis verified that the mRNA expression of these 4 genes was largely abolished. We evaluated the spontaneous SV release by analyzing the AMPAR-mediated miniature excitatory postsynaptic currents (mEPSCs) (Fig 1B). The results indicated that compared to the scrambled neurons, the Syt1/ Doc2a/Doc2b triple KD (tKD) neurons showed an increase in the frequency of mEPSCs; however, compared to the tKD group, neurons with an addition of Syt7 KD (Syt1/Doc2a/Doc2b/ Syt7 quadra KD (qKD)) showed a reduced mEPSC frequency, which was similar to the level of the scrambled neurons (Fig 1C). Moreover, the mEPSC amplitude was unaffected by the KD of genes. These results indicated that Syt7 could trigger spontaneous SV release in the absence of Syt1 and Doc2 and thus was a redundant Ca 2+ sensor for spontaneous glutamate release.

Syt7 is a redundant Ca 2+ sensor for spontaneous release in hippocampal neurons
Previously, it has been suggested that unlike most Syt isoforms including Syt1, Syt7 can interact with Sr 2+ to trigger membrane fusion in vitro [17]. Hence, based on a previously reported Syt7 KO mouse model in which the C2A domain was replaced with a neomycin cassette [24], we employed Sr 2+ as a tool to analyze the role of Syt7 in spontaneous release in the hippocampal CA1 slices (Fig 1D). Following the mEPSC recordings in the presence of 2 mM Ca 2+ , the extracellular Ca 2+ was replaced with Sr 2+ . After 5 min, a second recording was performed in the same neuron in the presence of Sr 2+ . Recordings in the same WT CA1 neuron revealed an increase in the mEPSC frequency when Ca 2+ was replaced by Sr 2+ , whereas in the Syt7 KO neurons, the mEPSC frequency was decreased in response to the Ca 2+ /Sr 2+ transition (Fig 1E). However, the mEPSC amplitude was unchanged (Fig 1F). We repeated the patch clamp recording analysis in the dentate gyrus (DGAU : PleasenotethatDGhasbeendefinedasdentategyru ) neurons of the hippocampal slices and observed similar results (Fig 1G-1I).
To test whether Syt7 triggered spontaneous release through its Ca 2+ binding motif, we expressed the Ca 2+ ligand mutant (Asp225 and 227 in C2A, Asp357 and 359 in C2B, to Asn) of Syt7 (Syt7 CLM ), which abolished the Ca 2+ binding activity of Syt7, in the cultured hippocampal neurons through lentiviral introduction (Fig 1J). We found that unlike hippocampal slices, the cultured WT neurons did not show any significant changes in mEPSC frequency when Ca 2+ was switched to Sr 2+ (Fig 1K). However, consistent with the results of the slice recording, in the Syt7 KO neurons, Sr 2+ resulted in a reduction in mEPSC frequency. We noted that in cultured neurons lacking Syt1 or Doc2, the Ca 2+ /Sr 2+ transition did not induce any obvious changes in mEPSC frequency (S1 Fig). Hence, the Sr 2+ -induced reduction in mEPSC frequency in the Syt7 KO neurons and slices was directly caused by the loss of Syt7. Importantly, this reduction could be rescued by the WT full-length Syt7 (Syt7 FL ) but not by Syt7 CLM (Fig  1K). These results indicated that Syt7 could directly trigger spontaneous SV release through its Ca 2+ binding motif. In addition, we compared the Ca 2+ /Sr 2+ -triggered mEPSCs in the cultured hippocampal neurons with Syt1 KD and Syt1/Syt7 double KD (dKD) and confirmed that the role of Syt7 in spontaneous release was unaffected by the presence of Syt1 (S2 Fig).
Furthermore, we investigated the mEPSCs driven by Ca 2+ concentration gradients (in mM: 0, 0.2, 0.5, 1, 2, 5, and 10) in the hippocampal CA1 slices of Syt7 KO mice (Fig 1L). We found that compared to the WT group, the Ca 2+ sensitivity curve showed a right shift in the Syt7 KO neurons, although the analyses of Ca 2+ cooperativity and affinity did not show statistically significant differences between the WT and Syt7 KO neurons (Fig 1M). Hence, these results indicated that Syt7 probably played a physiological role in spontaneous release.
Together, our results indicated that Syt7 was able to function as a redundant Ca 2+ sensor to trigger spontaneous glutamate release in mouse hippocampal neurons.

Syt7-triggered spontaneous glutamate release activates GluN2B-NMDARs
We then asked why the synapses employed so many Ca 2+ sensors to drive spontaneous release; namely, did these diverse subtypes of release have different physiological functions? Recently, we have found that Syt7 KO mice showed bipolar-like behavioral abnormalities through a mechanism involving the hypoactivity of GluN2B-NMDARs [15,23]. Hence, we performed electrophysiological recordings on the hippocampal slices of the WT and Syt7 KO mice to test whether Syt7-triggered spontaneous glutamate release could activate the GluN2B-mediated NMDARs compared to the other 2 Ca 2+ sensors (Fig 2A). To facilitate the recognition of NMDAR-mEPSC events, we generated the co-occurrence of AMPAR activation during the recordings and analyzed the AMPAR/NMDAR-mEPSCs. We applied Ro25-6981, a GluN2Bspecific antagonist, to the recording chamber to evaluate the GluN2B activity; AP5 was used to evaluate all NMDAR activity. We found that in both the WT and Syt7 KO neurons, the application of Ro25-6981 did not induce any obvious changes in the frequency or amplitude of the AMPAR/NMDAR-mEPSCs (Fig 2B and 2C). When we analyzed the charge and kinetics of the mEPSCs (Fig 2D), we found that in the WT neurons, the Ro25-6981 treatment could significantly reduce the decay time and charge transfer compared to the untreated control neurons (Fig 2E and 2F). This result indicated that in the WT neurons, spontaneous glutamate release could efficiently activate GluN2B-NMDARs. In contrast, in the Syt7 KO slices, we observed that the acute application of Ro25-6981 failed to significantly affect the decay time or charge transfer of the AMPAR/NMDAR-mEPSCs, which could still be reduced by AP5 (Fig  2E and 2F). This result indicated that in the Syt7 KO mice, where the spontaneous release was predominantly triggered by Syt1/Doc2, the GluN2B activity was attenuated during spontaneous neurotransmission. We also tested the GluN2B activity in cultured hippocampal neurons with Syt1 KD or Syt1/Syt7 dKD and observed that the Ro25-6981-induced attenuation of GluN2B activity was detected in the Syt1 KD neurons but not in the Syt1/Syt7 dKD neurons (S3 Fig), indicating that Syt7 could trigger spontaneous GluN2B activity when Syt1 was absent. We then investigated whether this attenuation was caused by the absence or mislocalization of GluN2B-NMDARs at postsynaptic spines. Immonoblot and qRT-PCR analyses revealed that the expression of GluN2B was up-regulated in the Syt7 KO neurons compared to the WT controls (S4A Fig). Moreover, the previously reported developmental switch of NMDARs from GluN2B to GluN2A was not affected by the Syt7 deficiency (S4B Fig) [25,26]. We also carried out super-resolution stochastic optical reconstruction microscopy (STORM) analysis and observed that the Syt7 deficiency did not affect the synaptic distribution of GluN2B in the hippocampal neurons (S4C and S4D Fig). These observations are consistent with our previous study [15]. We therefore concluded that Syt7-triggered spontaneous glutamate release was an important source of GluN2B activation in the mouse brain, which distinguished it from that triggered by Syt1 or Doc2.

Retargeting Syt7 to the central AZ triggers non-GluN2B spontaneous transmission
Previously, we have demonstrated that Syt7 and GluN2B were localized to the peripheral synaptic region, with the former in the peripheral AZ membrane and the latter in the peripheral PSD [15]. To help understand the role of Syt7 in spontaneous SV release, we employed a protein chimera Syt7 GAP43 , in which the C2 domains of Syt7 were conjugated to a short N-terminus 20 amino acid peptide of GAP43, to retarget Syt7 to the whole AZ. We carried out immunoelectron microscopy (immuno-EM) analysis to compare the localization of Syt7 and Syt7 GAP43 in the presynaptic AZ membrane. As the commercial Syt7 antibodies cannot label Syt7 GAP43 , we added an HA tag to the C-terminus of WT full length Syt7 (Syt7 FL -HA) or Syt7 GAP43 (Syt7 GAP43 -HA) to facilitate the immuno-EM experiments. qRT-PCR and immunoblot analyses revealed that HAconjugated Syt7 was expressed in the hippocampal neurons and could be detected by the HAspecific antibody (Fig 3A and 3B).
We then carried out EM analysis using the HA antibody. We observed that the Syt7 FL -HA and Syt7 GAP43 -HA groups of neurons showed a similar AZ length of 250 to 300 μm (Fig 3C). Moreover, Syt7 GAP43 -HA was largely present within a distance of 150 μm from the AZ center, whereas the distribution curve of Syt7 FL -HA was shifted to the outer region (Fig 3C). Hence, Syt7 FL -HA was largely absent from the central AZ, while Syt7 GAP43 -HA was retargeted to the whole AZ. We also evaluated the distribution of Syt7 at the global level using an immunofluorescence-based Duolink in situ proximity ligation assay (PLA) (S5A Fig). Compared to the association of Stx1a to SNAP-25, the Syt7-SNAP-25 Duolink puncta showed lower fluorescence density and smaller area, indicating that Syt7 had very limited contact overall with SNAP-25 compared to Stx1a (S5B and S5C Fig). We then compared the DUOlink puncta of Syt7 FL -HA and Syt7 GAP43 -HA and observed that Syt7 GAP43 -HA showed much greater contact with the SNARE proteins (S5D and S5E Fig), indicating that Syt7 GAP43 -HA was widely distributed in the AZ.
Next, we carried out patch clamp recording experiments to analyze the effects of Syt7 GAP43 on AMPAR-mediated mEPSCs (Fig 3D). We employed the acute Ca 2+ /Sr 2+ transition protocol and found that similar to our previous observation, the Syt7 KO neurons showed an obvious reduction in mEPSC frequency in response to the Ca 2+ /Sr 2+ switch; moreover, this reduction was rescued by both Syt7 FL and Syt7 GAP43 (Figs 3E and S6A). Hence, Syt7 GAP43 was able to trigger spontaneous release. We noted that the mEPSC amplitude was unchanged in all groups (Figs 3F and S6B). We then assayed the GluN2B activity during the Syt7 GAP43 -triggered AMPAR/NMDAR-mEPSCs (Fig 3G). We observed that the Ro25-6981 treatment induced an obvious attenuation in the decay time and charge transfer of the mEPSCs in the WT neurons but not the Syt7 KO neurons (Figs 3H and 3I and S6C and S6D), which is consistent with our previous observations. Importantly, the GluN2B hypoactivity in the KO neurons was rescued by Syt7 FL , but it could not be significantly changed by Syt7 GAP43 , indicating that Syt7 GAP43 could not function to effectively activate the GluN2B-NMDARs. In addition, in all groups of neurons, the Ro25-6981 treatment did not affect the frequency; however, it led to a decrease in the mEPSC amplitude when Syt7 or Syt7 FL was present (S6E and S6F Fig). Together, our results showed that modifying the localization of Syt7 in the AZ still allowed Syt7 to drive spontaneous release, but the positional noncorrespondence largely diminished the activation of GluN2B-NMDARs, which supported the unique function of Syt7-triggered release.

Human Syt7 variants induce GluN2B hypoactivity in human iPSC-derived and mouse hippocampal neurons
Previously, we have identified several Syt7 SNPs in BD patients and demonstrated that the Syt7 gene was associated with susceptibility to this disease [15]. Moreover, the Syt7 SNPs, L227M and Q247H, could attenuate the Syt7-triggered asynchronous and short-term plasticity forms of glutamate release. Both mutations resided on exon 6 of the human SYT7 gene, which encodes the juxtamembrane linker region of the Syt7 protein and showed damage potential on the structure of the Syt7 protein by Harmfulness evaluation analysis with the Polyphen2_H-DIV and Polyphen2_HVAR algorithms. In the mice, Syt7 isoform 4 had these 2 mutation sites. Thus, we compared the performances of isoform 4 of mouse Syt7 (Syt7 IF4 ) and its L152M/Q172H mutant corresponding to human L227M/Q247H (Syt7 IF4mut ) in spontaneous release. In cultured WT neurons expressing the Syt7-specific CRISPRi rather than Syt7 KO neurons, we expressed Syt7 IF4 and Syt7 IF4mut through lentiviral infection and investigated the spontaneous GluN2B activity (Fig 4A). Analysis of AMPAR/NMDAR-mEPSCs revealed that compared to the scrambled neurons, the Syt7 KD neurons and KD neurons expressing Syt7 IF4 or Syt7 IF4mut showed similar frequencies and amplitudes of mEPSCs; moreover, the Ro25-6981 treatment did not affect either the mEPSC frequency or amplitude in these 2 groups of neurons, whereas AP5 could reduce the mEPSC amplitude (Fig 4B and 4C). Consistent with our observations in the Syt7 KO neurons, the decay time and charge transfer of AMPAR/ NMDAR-mEPSCs were unaffected by Ro25-6981 in the Syt7 KD neurons. Importantly, Ro25-6981 could significantly reduce the decay time and charge transfer of the mEPSCs in the Syt7 IF4 -expressing KD neurons, but it failed to affect the Syt7 IF4mut -expressing KD neurons (Fig 4D-4F). Therefore, in mouse hippocampal neurons, the 2 Syt7 SNPs could not efficiently activate GluN2B-NMDARs via spontaneous glutamate release.
We then aimed to ascertain whether the deficits in spontaneous release existed in the BD patient neurons, which have insufficient Syt7 expression [23]. To this end, we analyzed the AMPAR/NMDAR-mEPSCs in the hippocampal DG-like neurons derived from the iPSCs of 3 sporadic BD-I patients and 3 healthy controlsAU : PerPLOSstyle; subjectsshouldnotbeusedforhumanpat [27], as well as patient neurons with lentiviral overexpression of Syt7 FL or Syt7 IF4mut . The neurons differentiated from the iPSCs were mostly Prox1 + DG granule cells and showed normal AP firing (S7A and S7B Fig). The results indicated that the decay time and the charge transfer of the mEPSCs were not significantly different between the 4 groups (Figs 4G-4I and S7C-S7J). The Ro25-6981 treatment induced an obvious reduction in the mEPSC decay time and charge transfer of the healthy control neurons, whereas the diseased neurons were only slightly affected. Moreover, the diseased neurons overexpressing Syt7 FL showed a significantly reduced decay time or charge transfer in response to the Ro25-6981 treatment, whereas the neurons overexpressing Syt7 IF4mut failed to change the slight response of the diseased neurons to the drug treatment. These results indicated that the hippocampal neurons derived from the iPSCs of these BD patients showed Syt7 deficit induced GluN2B dysfunction and could be restored by the exogenous introduction of Syt7 but not by the Syt7 SNPs of BD patients.
Together, our data indicated that Syt7 deficit-induced defects in spontaneous release could induce GluN2B dysfunctions, which might contribute to the etiology of the investigated BD cases.

Discussion
Candidate Ca 2+ sensors for spontaneous release included Syt1, Doc2, and Syt7. The former two have been verified to function to drive this type of release [9,16]. Recently, it has been reported that Syt7 could participate in a pathological stress-induced augmentation of spontaneous release [28]. In the present study, we demonstrated that Syt7 acted as a redundant Ca 2+ sensor to trigger spontaneous glutamate release that contributed to the activation of postsynaptic GluN2B-NMDARs. The physiological relevance of Syt7 to spontaneous release has long been unrecognized, because the mEPSC frequency failed to show changes in the Syt7-deficient neurons or in the Syt1/Syt7 double-deficient neurons [8,19]. In recent years, accumulating evidence has revealed that Ca 2+ -dependent vesicle exocytosis can have redundant mechanisms. For instance, Syt1 and Syt7 are redundant Ca 2+ sensors that mediate AMPAR exocytosis during LTP [22]. In Syt1-deficient neurons, it seemed that at least one sensor with greater Ca 2+ binding affinity worked to induce the enhancement of the spontaneous release. However, it has been found that simply inactivating Doc2 or Syt7 alone in neurons lacking Syt1 could not address the redundancy of spontaneous release [8,16,29]. In the present study, we utilized the multiplex CRISPRi technique to abolish the expression of all 3 candidate Ca 2+ sensors, including Syt1, Syt7, and Doc2a/2b, and found that the redundantly enhanced spontaneous release in Syt1-deficient neurons could be largely diminished. This result indicated that Syt7 and Doc2 probably played co-redundant roles in triggering spontaneous release in the Syt1-deficient neurons. We note that the occurrence of spontaneous release events is often regulated by AP firing-relevant homeostatic plasticity [30]. Hence, it is necessary to consider the factors involving chronic homeostasis, in addition to redundancy, when evaluating the function of AP-related proteins in spontaneous release. This is supported by the fact that Syt1 drives a majority of the AP-triggered SV release, and in Syt1-deficient neurons, the rate of spontaneous release is increased. As Doc2 could affect asynchronous release in neurons, which contributes to neural network excitability and synaptic plasticity, it is also possible that Doc2 can indirectly affect the mEPSC frequency through modulating the baseline neural network activity. In the present study, we employed an acute Ca 2+ /Sr 2+ switching approach and demonstrated that Syt7 could have an instant impact on the rate of spontaneous release. Previously, in vitro reconstitution studies have revealed that Syt7 could efficiently interact with Sr 2+ to induce robust membrane fusion, whereas Syt1 and most other Syt isoforms could not [17]. In physiological tests, it was found that Sr 2+ could promote AP evoked asynchronous SV release [31,32]. Particularly, even in the Syt1-deficent neurons, the application of extracellular Sr 2+ would attenuate the fast synchronous phase of SV release, while the asynchronous phase release remained unaffected compared to the application of Ca 2+ , leading to a deceleration in the decay of the overall SV release [33]. All these findings support that compared to Ca 2+ , Sr 2+ has a weakened interaction with Syt1 and other synchronous Ca 2+ sensors (e.g., Syt2) to induce fast SV release, while the asynchronous release remains unaffected. Hence, it was appropriate to employ Sr 2+ as a tool to compare the roles of Syt7 and Syt1 in spontaneous release. Moreover, the Syt7 KO neurons showed alterations in the Ca 2+ sensitivity of the spontaneous SV release. In addition, we confirmed that the Syt7 deficiency-or mutant-induced changes in mEPSC frequency were not caused by abnormalities in the synapse density, readily releasable SV pool (RRP) size or protein mislocalization (S8 Fig). These results, together with previous findings [9,16], indicated that spontaneous release is a heterogeneous phenomenon involving multiple mechanisms that can have redundant and homeostatic adjustment, and Syt7 is one of the Ca 2+ sensors involved. We note that the Sr 2+ application increased the mEPSC frequency in the WT slices but not in the WT neuronal cultures (Fig 1). This discrepancy could be caused by several reasons. First, the protein machinery for SV fusion might have some differences in the brain slices and cultured neurons (e.g., different splice variants of SNAREs) and thus have slightly different Sr 2+ response kinetics. Second, the intracellular Sr 2+ buffering system might have some differences in the brain slices and cultured neurons, and thus the intracellular Sr 2+ concentration might be different. Our results indicated that the Syt7-triggered spontaneous release has a unique function compared to that triggered by Syt1 or Doc2, that it could be an important source for the activity of GluN2B-NMDARs in the brain. It has been widely shown that GluN2B is likely an important NMDAR subunit involved in mental illnesses including BD. Importantly, this dysfunction was not only induced by the Syt7 deficiency but was also shown by the BD susceptibility variants of Syt7. In the present study, the role of Syt7 in GluN2B activation was further verified by the Syt7 GAP43 mutant experiments, which targeted Syt7 to the broad AZ. Previously, it has been suggested that posttranslational modifications in the N-terminal segment, which could direct the protein trafficking and participate in scaffolding interactions, were the determinants for the different trafficking routes of Syt7 and Syt1 [34]. Using Syt7 GAP43 and Syt7 GAP43 -HA, we not only removed the peripheral AZ trafficking potential of Syt7 but also endowed Syt7 with the ability to broadly distribute in the AZ.
Conventionally, it has been thought that NMDARs are likely saturated by quantal size spontaneous glutamate release, based on the electrophysiological analysis of NMDAR current kinetics [35]. However, this opinion has been challenged in the past more than 10 years. For instance, it has been suggested that in the postsynaptic region, spontaneous and evoked release may activate different populations of NMDARs [36], leading to differential responses to ketamine, the prominent NMDAR-targeted antidepressant [37]. These findings provide direct evidence that NMDARs were unlikely to be saturated by spontaneous release. Here, our data provided further evidence to support that separated subpopulations of NMDARs showed correspondence to different presynaptic SV pools.
Finally, we note that in the current study, the mEPSC amplitude analysis showed discrepant results in our experiments. For instance, in Figs 2C and 4C, Ro25-6981 and AP5 did not affect the mEPSC amplitude; however, in Figs 4C and S2B, these 2 drugs could induce decreases in the mEPSC amplitude. There are 2 possible reasons for this discrepancy. First, the experiments in Fig 2 were performed on brain slices, while those in Figs 4 and S2 were performed on cultured neurons. In different neuronal systems, the rising kinetics of NMDAR-mEPSC might have some differences, which might be sufficient to affect the mEPSC amplitude when combined with AMPAR-mEPSCs. Second, as the time period for our study spanned several years and several generations of animals were used for the electrophysiological recordings in these figures, the later offspring could show accumulative impacts of some unknown epigenetic modifications. In this case, the functional phenotype of Syt7 in spontaneous release persisted throughout these animal generations, indicating that Syt7 is a compelling candidate Ca 2+ sensor for spontaneous release.

Animals and plasmids
Syt7 KO mice were kindly provided by E.R. Chapman (Madison, Wisconsin, USA) with permission from N.W. Andrews (College Park, Maryland, USA). All the animal experiments were conducted under the guidance and with the approval of the Institutional Animal Care & Use Committee of Tsinghua University and the Animal Welfare and Ethics Committee of Tsinghua University (Ethics approval number: F16-00228; A5061-01), and they conformed to the National Institutes of Health guidelines on the ethical use of animals. For the lentiviral experiments, a bicistronic lentiviral vector system, pLox Syn-DsRed-Syn-GFP (pLox), was used by substituting either the DsRed or GFP coding sequence or both with the target cDNA sequence. For the Syt7 experiments, the cDNAs include mouse full-length or HA-tagged mouse Syt7 isoform α cDNA (Syt7 FL ; NM_018801.3), isoform 4 cDNA (Syt7 IF4 ; NM_001373944.1), and isoform 4 cDNA carrying human SYT7 L227M/Q247H (mouse L152M/Q172H) mutations (Syt7 IF4mut ). In the mouse Syt7 isoform α cDNA, Asp Ca 2+ ligands 225, 227 (in C2A), 357 and 359 (in C2B) were mutated to Asn in Syt7 CLM ; cDNA for the C2AB domain (a.a. 134-403) was conjugated to the 1-20 amino acid sequence of GAP43 to generate Syt7 GAP43 . The lentiviral CRISPRi system was described previously [38]. The sgRNAs were designed to target the DNA region from −50 to 300 bp relative to the TSS of Syt1, Syt7, and Doc2a/b [39]. The sgRNA sequences are listed in S1 Table. Mouse primary neuronal culture Mouse primary neuronal culture was performed as previously described [15]. Briefly, hippocampal neurons were dissected from newborn WT and homozygous Syt7 KO mice in the dark phase and incubated in 0.25% trypsin-EDTA (LifeAU : PleaseprovidemanufacturerlocationdetailsforLif Technologies) for 15 min at 37˚C. After washing with Hank's Buffered Salt Solution plus 5 mM HEPES (Life Technologies), 20 mM Dglucose and 2% fetal bovine serum (FBS) (GibcoAU : PleaseprovidemanufacturerlocationdetailsforGibc ), the neurons were mechanically dissociated in culture medium and plated on poly-D-lysine-coated glass coverslips at a density of 50,000 to 100,000 cells/cm 2 . Cells were grown in Neurobasal-A medium (Life Technologies) supplemented with 2% B-27 (Life Technologies) and 2 mM glutamax (Life Technologies). Cultures were maintained at 37˚C in a 5% CO 2 -humidified incubator.

Differentiation of iPSC into DG-like neurons
The forebrain neural progenitor cells (NPCs) derived from BD patients and healthy people were characterized as previously described [27,40]. The information about the approving committee, informed consent, and clinical trial registration number of the iPSC studies was described in the original articles [27,40]. To obtain hippocampal DG-like neurons, NPCs were differentiated in DMEM/F12 supplemented with N2 (Life Technologies), B27 (Life Technologies), 20 ng/ml BDNF (PeprotechAU : PleaseprovidemanufacturerlocationdetailsforPeprotechandRDinthesentenceToobt ), 1 mM dibutyrl-cyclicAMP (Sigma), 200 nM ascorbic acid (Sigma), 1 μg/ml Laminin, and 620 ng/ml Wnt3a (R&D) for 3 to 4 weeks. Wnt3a was removed after 3 weeks. All cells used in the present study were verified as mycoplasma contamination free.

Immunoelectron microscopy
Cells were fixed in 4% paraformaldehyde (PFA) and 0.2% glutaraldehyde in phosphate buffered saline (PBS) at room temperature for 10 min. Cells were permeabilized in PBS containing 0.01% saponin for 5 min, blocked with PBS buffer containing 10% FBS, 10% goat serum, and 0.1% cold fish gelatin at room temperature for 40 min, and then exposed overnight to primary antibodies in blocking solution. The primary antibodies included rabbit polyclonal anti-HA antibody (1:200, Easybio, #BE2008). Samples were incubated with Flouronanogold anti-rabbit Fab'AleaxFlour 488 (1.4-nm diameter, Nanoprobes, #7204, 1:200) for 2 h and washed with PBS containing 0.1% cold fish gelatin for 10 min for 6 times. Then, the signal was intensified with a gold enhancement kit (GoldEnhance EM, Nanoprobes, #2114) at room temperature for 2 min. The samples were postfixed in 1% OsO4 containing 1.5% potassium ferrocyanide, then dehydrated in a series of graded ethanol solutions and embedded in epoxy resin. Ultrathin sections were collected and stained with uranyl acetate and lead citrate and observed and analyzed using a Hitachi H-7650B transmission electron microscope.

Reverse transcription PCR
Total RNA was isolated using Trizol (Life Technologies) according to the manufacturer's instructions. cDNA was synthesized using SuperScript III Reverse Transcription Kit (Life Technologies). The primer sequences for PCR are listed in S2 Table. qRT-PCR was performed on a Bio-Rad CFX96 thermal cycler using SYBR green supermix (BioAU : Pleaseprovidemanufacturerlo -Rad) and gene-specific primers. Quantitative analysis was performed employing the ΔΔCT method and the GAPDH as the endogenous control.

Duolink in situ proximity ligation assay (PLA)
The Duolink in situ PLA was performed using a kit purchased from Sigma (Cat# DUO92008, 92004 and 92002) according to the manufacturer's instructions. Briefly, neurons were fixed for 10 min in 4% PFA in PBS, followed by 10 min in permeabilization buffer containing 0.1% Triton X-100. After blocking for 1 h in PBS plus 0.1% Triton X-100 and 1% BSA, cells were incubated for 1 h with primary antibodies and washed 3 times with PBS. The PLA plus and minus probes were added, and the cells were incubated for 1 h at 37˚C. After that, ligation and amplification were performed. Cells were stained with DAPI (0.1 μg/ml, Sigma) in PBS for 10 min at room temperature. All images were collected using an Olympus FV1200 Confocal microscope (OlympusAU : PleaseprovidemanufacturerlocationdetailsforOlympusinthesentenceAllimageswerecollec ) with a 60X objective (NA 1.40). Images were acquired using Fluoview software version 1.6 (Olympus) and analyzed using ImageJ software (NIH). The primary antibodies included rabbit polyclonal anti-Stx1a antibody (1:500, Abcam, #ab41453), mouse monoclonal anti-SNAP25 antibody (1:500, Synaptic Systems, #111011), and rabbit polyclonal anti-Syt7 antibody (1:200, Synaptic Systems, #105173).

Statistical analysis
Data are shown as mean ± SEM of the values from at least 3 independent experiments. The numerical data are included in S1 Data. For experiments in Fig 2D-2I, statistical significance was evaluated using two-tailed paired Student t test at P < 0.05. For experiments in Fig 3, ANOVA was used. For all other experiments, statistical significance was evaluated using twotailed unpaired Student t test at P < 0.05.