Ca2+ sensor proteins in spontaneous release and synaptic plasticity: Limited contribution of Doc2c, rabphilin-3a and synaptotagmin 7 in hippocampal glutamatergic neurons

Presynaptic neurotransmitter release is strictly regulated by SNARE proteins, Ca2+ and a number of Ca2+ sensors including synaptotagmins (Syts) and Double C2 domain proteins (Doc2s). More than seventy years after the original description of spontaneous release, the mechanism that regulates this process is still poorly understood. Syt-1, Syt7 and Doc2 proteins contribute predominantly, but not exclusively, to synchronous, asynchronous and spontaneous phases of release. The proteins share a conserved tandem C2 domain architecture, but are functionally diverse in their subcellular location, Ca2+-binding properties and protein interactions. In absence of Syt-1, Doc2a and -b, neurons still exhibit spontaneous vesicle fusion which remains Ca2+-sensitive, suggesting the existence of additional sensors. Here, we selected Doc2c, rabphilin-3a and Syt-7 as three potential Ca2+ sensors for their sequence homology with Syt-1 and Doc2b. We genetically ablated each candidate gene in absence of Doc2a & b and investigated spontaneous and evoked release in glutamatergic hippocampal neurons, cultured either in networks or on microglial islands (autapses). The removal of Doc2c had no effect on spontaneous or evoked release. Syt-7 removal also did not affect spontaneous release, although it altered short-term plasticity by accentuating short-term depression. The removal of rabphilin caused an increased spontaneous release frequency in network cultures, an effect that was not observed in autapses. Taken together, we conclude that Doc2c and Syt-7 do not affect spontaneous release of glutamate in hippocampal neurons, while our results suggest a possible regulatory role of rabphilin-3a in neuronal networks. These findings importantly narrow down the repertoire of synaptic Ca2+ sensors that may be implicated in the spontaneous release of glutamate.

Proteins of the Double C 2 domain family (Doc2a and Doc2b) have been implicated in spontaneous and asynchronous release (Courtney et al., 2018;Gaffaney et al., 2014;Groffen et al., 2010;Pang et al., 2011;Yao et al., 2011). The asynchronous release function is mostly observed in locally stimulated networks but not in autaptic neurons, illustrating some differences between these preparations (Liu et al., 2009;Wierda and Sorensen, 2014) and leaving some debate about the mechanism involved (Bourgeois-Jaarsma et al., 2019;Díez-Arazola et al., 2020;Houy et al., 2017). Doc2 proteins share many structural and functional properties with Syts. They are enriched in synaptosomal membranes, show Ca 2+ -dependent binding to phosphatidylserine-containing liposomes (Kojima et al., 1996;Orita et al., 1995) and interact with the synaptic secretory protein Munc18-1 (Stxbp1) in GST-pulldown and Yeast-Two-Hybrid experiments (Verhage et al., 1997). During development, the expression of Doc2b starts at mouse embryonic day E12, followed by Doc2a expression from E17 onwards (Korteweg et al., 2000). In multiple tissue Northern blots, Doc2a shows brain-specific mRNA expression while Doc2b is additionally expressed in peripheral tissues (Sakaguchi et al., 1995;Verhage et al., 1997). The adult brain expresses both isoforms in a complementary expression pattern with limited overlap (Verhage et al., 1997). Single-cell quantitative RT-PCR measurements in single hippocampal neurons demonstrated that many neurons express all three Doc2 variants together with up to five different Syts (e.g. 4,7,11 and 13) (Bacaj et al., 2013). Other experiments based on RNAScope visualization of Doc2a and Doc2b mRNA indicated, however, that Doc2a is the predominant isoform in adult glutamatergic pyramidal neurons in the CA3 region of the mouse hippocampus, whereas Doc2b is specifically detected in GABAergic neurons of the dorsal striatum (Courtney et al., 2018). In the same study, the two isoforms were functionally redundant in their ability to restore spontaneous release frequencies in Doc2a or Doc2b deficient neurons. Doc2 differs from Syts by the absence of a transmembrane domain (TMD) in the N-terminal region. Instead, both Doc2a and -b are soluble cytoplasmic proteins which can be recruited to the PM in a reversible, activity-dependent manner (Duncan et al., 1999;Groffen et al., 2004). This can occur by two separate pathways: via the N-terminal MID domain or by activation of the C 2 domains.
The Munc13 interaction can drive Doc2 translocation to the PM (Duncan et al., 1999;Groffen et al., 2004). This mechanism is stimulated by the binding of diacylglycerol (DAG) to the C1 domain of Munc13. Endogenously, DAG is produced by hydrolysis of PIP 2 in the membrane, a reaction catalyzed by phospholipase C. Experimentally, the C1 domain can be activated by phorbol-ester compounds as DAG analogs. Although DAG and phorbol-esters also activate other signalling cascades such as the protein kinase C pathway, C1 domain activation of Munc13-1 plays a critical role in presynaptic activity-dependent augmentation of synaptic strength (Rhee et al., 2002).
In contrast to the Munc13-dependent co-translocation of Doc2b, the PM association of Doc2b can also drive that of Munc13-1 as observed in chemically depolarized PC12 cells (Friedrich et al., 2013). In neurons, repetitive firing at frequencies of 2 Hz and higher causes gradual increases in PM association of both Doc2a and -b (Groffen et al., 2006). In cultured rat hippocampal neurons, the recruitment of Munc13-1 to the PM is blocked by mutagenesis of the MID domain in Doc2b (Xue et al., 2018).
Blockade of the Doc2-Munc13 interaction with a synthetic MID peptide in superior cervical ganglion cells (Mochida et al., 1998) and the calyx of Held synapse (Hori et al., 1999) reduced activity-dependent presynaptic potentiation. In adrenal chromaffin cells, Doc2b Ca 2+dependently enhances the size of the RRP of chromaffin granules in a Munc13-2 dependent manner, and mutations in the MID domain block this function (Houy et al., 2017). Another indication for a functional role of the Doc2-Munc13 interaction was obtained in mouse hippocampal neurons where it played a role in synaptic augmentation at 10 Hz (Xue et al., 2018).

The role of C 2 domains in Ca 2+ sensor activation
A second mechanism for the Ca 2+ -induced membrane association of Doc2 proteins is encoded by its C 2 domains. Originally identified as the second conserved domain in protein kinase C paralogs (Coussens et al., 1986), C 2 domains were subsequently found in many other enzymes including multiple phospholipases, phosphaditylinositol-3-kinases, the E3 ubiquitin ligase NEDD-4 and several Ras-GTP activating proteins (Leonard, 2013;Nalefski and Falke, 1996). C 2 domains typically fold into a beta sandwich of eight antiparallel strands (Essen et al., 1996;Giladi et al., 2013;Perisic et al., 1998;Sutton et al., 1995;Xu et al., 1998). Flexible loops that emanate out of the beta strands position five conserved acidic residues (numbered 1-5 in Fig. 1B) around a Ca 2+ binding pocket that can cooperatively bind multiple Ca 2+ ions Shao et al., 1996). The Ca 2+ binding causes a strong change in the electrostatic surface of the C 2 domain, causing a reversible recruitment to the protein's target membrane by supporting interactions with phospholipids on the cytoplasmic leaflet (Corbin et al., 2007). Structural differences in the Ca 2+ -binding loops affect the phospholipid selectivity and the affinity and kinetics of Ca 2+ / membrane binding (Corbalan-Garcia and Gómez-Fernández, 2014;Nalefski et al., 2001). Membrane binding can be accompanied by the insertion of lipophilic residues into the hydrophobic phase of the membrane bilayer (Davletov et al., 1998;Frazier et al., 2003). The function of C 2 domains as a Ca 2+dependent membrane association module is considered to be the canonical property of C 2 domains . In fact, a single C 2 domain unit can be sufficient for membrane association (e.g. see Clark et al., 1991).
Many proteins implicated in the organization and function of the presynaptic secretory apparatus contain multiple C 2 domains which can be interspersed (e.g. in Munc13s) or occur as tandem repeats ( members of the Syt family) (Perin et al., 1990;Südhof, 2002). In addition to their interactions with SNAREs, Munc18 and Munc13, the tandem C 2 domain fragments of Syt-1 and Doc2b can induce membrane curvature in cell-free liposome preparations, an activity that depends on Ca 2+ activation and the presence of phosphatidylserine (Brouwer et al., 2015;Groffen et al., 2010;Martens et al., 2007;Sorkin et al., 2019). This effect is thought to result from wedge-shaped insertions into the cytoplasmic leaflet of the bilayer . In tandem C 2 domain proteins, the C 2 A and C 2 B domains play different roles in Ca 2+ sensor activation. This is exemplified by a large difference in the intrinsic Ca 2+ affinity, defined as the affinity of Ca 2+ binding in the absence of phospholipids. In Syt-1, Doc2b and rabphilin-3a, the intrinsic Ca 2+ affinity of the C 2 B domain is much higher (Fernández-Chacón et al., 2002;Giladi et al., 2013;Montaville et al., 2007). By contrast, the C 2 A domain typically has a higher apparent Ca 2+ affinity, which describes the Ca 2+ dependence of membrane association. This difference is likely related to the fact that the C 2 domain does not completely surround bound Ca 2+ ions, but instead, the phospholipids are necessary to complete the full Ca 2+ coordination sphere , although rabphilin-3a may be an exception to this rule (Montaville et al., 2007). The apparent affinity is strongly influenced by the phosphatidylserine concentration as well as that of PIP 2 , the latter bound by a lysine-rich cluster away from the Ca 2+ -binding loops (Bai et al., 2004;Groffen et al., 2010;Montaville et al., 2008). This polybasic stretch also participates in SNARE interactions of Doc2b and Syt-1 (Guillén et al., 2013;Honigmann et al., 2013;Michaeli et al., 2017;Rickman et al., 2004;Sato et al., 2010;Zhou et al., 2017). Substitutions of key aspartates in the C 2 A domain motif strongly affect the Ca 2+dependent phospholipid binding activity. For example, D3/4N mutations in the C 2 A domain of Doc2b induce constitutive Ca 2+ binding and increase spontaneous release rates, whereas the corresponding substitutions in the C 2 B domain do not (Bourgeois-Jaarsma et al., 2019;Courtney et al., 2018;Friedrich et al., 2008;Gaffaney et al., 2014;Groffen et al., 2010;Houy et al., 2017). Interestingly however, the D2N mutation in the C 2 B domain (D291N in Doc2a and D303N in Doc2b) causes a loss of function both in terms of PM association and spontaneous release enhancement (Courtney et al., 2018). Thus, despite the divergent properties of each C 2 domain, they clearly act synergistically in an overall Ca 2+ sensor function, a principle that has also been described for Syt-1 (Bai et al., 2002).

Evidence for unknown Ca 2+ sensors in synaptic release
The different release types are thus thought to rely on different Ca 2+ sensors, each showing different Ca 2+ affinities and potentially competing with each other for vesicular fusion, although other types of interaction are also possible (Walter et al., 2011). Doc2a and Doc2b have a very high apparent Ca 2+ affinity (EC 50 value of 450 and 175 nM respectively) as determined by membrane association in living cells (Groffen et al., 2006). Syt-1, on the other hand, is more closely localized to the site of Ca 2+ entry during synaptic depolarization (reviewed in Bornschein and Schmidt, 2019). A further difference between the mode of action is likely caused by the membrane association and dissociation kinetics, both showing a slower timescale for Doc2 proteins than Syt-1 (Yao et al., 2011). The removal of Syt-1 driven release causes a significant increase in spontaneous release ('unclamping'), but this phenomenon does not require Doc2a or Doc2b (Díez-Arazola et al., 2020). Thus, the way in which multiple sensors interact with each other is incompletely understood, although a picture emerges where the presence of faster sensors appears to dominate the overall release kinetics of synapses with multiple co-expressed sensors (Turecek and Regehr, 2019).
Various lines of evidence suggest that one or more Ca 2+ -sensors other than Syt-1 and Doc2a/b account for residual spontaneous release. First, hippocampal neurons lacking both Doc2a and -b still have a remaining spontaneous release component and the frequency of these residual events is sensitive to the Ca 2+ -chelator BAPTA (Groffen et al., 2010). Similarly, mouse cortical neurons in which four Doc2-related genes (Doc2a, b, c & Rph3a) were knocked-down by RNA interference displayed residual spontaneous release, measured as miniature excitatory postsynaptic currents (mEPSCs) and inhibitory ones (mIPSCs), which still appeared Ca 2+ -sensitive (Pang et al., 2011). Also in Syt-1 knockout (KO) neurons, significant quantal release persisted after quadruple knock-down (Pang et al., 2011). The effect of single Doc2c or Rph3a gene suppression in Doc2a, b knock-down neurons was not tested in this study. Also in triple KO mice lacking Syt-1, Doc2a and Doc2b, excitatory hippocampal neurons still exhibited residual spontaneous secretion (Díez-Arazola et al., 2020). In the calyx of Held synapse, where Syt-2 abolishes synchronous neurotransmission, the remaining release rate shows a dependence over 3 orders of magnitude of [Ca 2+ ] i , suggesting the presence of additional Ca 2+ -sensitive components other than Doc2 proteins (Kochubey et al., 2011). Taken together, there is clear evidence for the existence of additional Ca 2+ -dependent mechanisms for AP-independent neurotransmitter release, but the involved Ca 2+ sensors remain enigmatic.
In this study, we investigated the potential role of three C 2 domain proteins based on their high similarity with Doc2a and b: Doc2c, rabphilin-3a and Syt-7 ( Fig. 1A) in both networks and autapses.

Doc2c
Doc2c is the least studied member of the Doc2 family. It is widely expressed in the adult mouse brain, including in the olfactory bulb, cortex, hippocampus, cochlear nucleus, trapezoid body and cerebellum and also in peripheral tissues such as the heart (Fukuda and Mikoshiba, 2000). Hippocampal neurons mostly co-express Doc2a, b and c at substantial levels (Bacaj et al., 2013). Its domain organization closely resembles that of Doc2a and -b, including an N-terminal MID and two C 2 domains. Despite its high amino acid sequence identities with Doc2a and b (45.6% and 43.2% respectively) the C 2 A domain lacks Ca 2+ -dependent lipid binding capacity (Fukuda and Mikoshiba, 2000) and does not associate with the PM in response to Ca 2+ influx (Fukuda and Mikoshiba, 2000;Groffen et al., 2006). This is undoubtedly due to the loss of conserved acidic residues in the C 2 A domain (Fig. 1B). Instead, Loop 3 in the C 2 A domain of Doc2c contains a unique basic insertion RLRRRRRR, which could support electrostatic interactions with negatively charged phospholipids or may act as a nuclear localization signal Groffen et al., 2006). Despite its non-canonical C 2 A domain, the five Ca 2+ binding Fig. 1. Domain structure and alignment of functional motifs in the investigated C 2 domain proteins. A. Doc2c contains a Munc13-interacting domain (MID) on its Nterminal and two C 2 domains (C 2 A and C 2 B) on its C-terminal side. Rabphilin-3a consists of a Rab3/Rab27 binding domain (RBD) and a C 2 tandem. Syt-7 has an Nterminal transmembrane domain (TM) and two C-terminal C 2 domains. Numbers indicate amino acids. B. CLUSTAL Omega v1.2.4 alignment of the MID and Ca 2+binding loops in the C 2 A and C 2 B domain of mouse Doc2a (GenPept accession number NP_001355285.1), Doc2b (NP_031899.2), Doc2c (NP_068563.2), Rabphilin-3A (NP_035416.1), Syt-1 (NP_033332.1) and Syt-7 (NP_061271.1). Within the MID, bold residues are critically involved in Munc13 binding which is blocked by substitution with the sequence YKDWAF (Díez-Arazola et al., 2020; Groffen et al., 2004;Hori et al., 1999;Houy et al., 2017;Mochida et al., 1998;Xue et al., 2018). In C 2 domains, beta strand numbers are indicated in grey; Ca 2+ -binding loops containing the five acidic residues that define the C 2 domain motif are numbered 1-5 (reviewed in Nalefski and Falke, 1996). C. Homologous gene targeting strategy for Doc2c null mice. TV, targeting vector. Boxes indicate exons 1-11; grey boxes are coding sequences. Exon phases are indicated below each intron in the wildtype gene. A neomycin resistance gene for positive selection of the ES cells was flanked by Frt sites (red triangles). LoxP sites are indicated by green triangles. Recombination by Flp and Cre results in the deletion of exons 3 to 7, causing a frame shift (fs) of downstream exons. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) aspartates in the C 2 B domain are fully conserved (Fig. 1B). Also the MID domain of Doc2c is completely conserved with Doc2a/b, providing a potential mechanism for Munc13-dependent association with the PM in response to DAG (Duncan et al., 1999). A null model for Doc2c has not yet been described.

Rabphilin
Rabphilin-3A, encoded by the gene Rph3a, was first identified and isolated as a putative target of Rab3a , a small Gprotein involved in vesicle trafficking and neurotransmitter release (Numata et al., 1994;Shirataki et al., 1994;Takahashi et al., 1995). Electron microscopy, subcellular fractionation and imaging revealed that rabphilin-3a is transiently recruited to presynaptic vesicles (Mizoguchi et al., 1994) where it reversibly interacts with Rab3a (Stahl et al., 1996). The role of rabphilin-3a in the Ca 2+ -dependent secretion pathway is still elusive as it seems to act at multiple stages of neurosecretion. It is composed of two main functional domains (Fig. 1). The first is the N-terminal Rab3a binding domain (RBD), responsible for association with Rab3 (McKiernan et al., 1996) and Rab27 (Fukuda et al., 2004b). This domain also contains phosphorylation sites for cAMPdependent protein kinase (PKA), protein kinase C (PKC) and Ca 2+ / Calmodulin-dependent kinase II (CaMKII) (Foletti et al., 2001;Fykse et al., 1995). Phosphorylated rabphilin-3a has a reduced affinity for membranes (Foletti et al., 2001), suggesting that the membrane association of rabphilin-3a is modulated by those kinases as part of the synaptic vesicle cycle, potentially affecting neurotransmission and synaptic plasticity. The second main functional domain is formed by two tandem C 2 domains at the C-terminus, which mediate Ca 2+ -dependent binding to phospholipids  and SNAP-25 binding via a polybasic stretch (Ferrer-Orta et al., 2017;. Both regions of rabphilin-3a associate with synaptic vesicles (SVs) (Senbonmatsu et al., 1996).
Rabphilin-3a binds Ca 2+ with a high affinity in the submicromolar range (Pinheiro et al., 2016) and binds phospholipids in a Ca 2+ -dependent manner. Its C 2 domains show a high homology with those of Doc2a and Doc2b (61% and 73% identity respectively) (Brose et al., 1995;Verhage et al., 1997). Moreover, rabphilin-3a represents the evolutionary ancestor in D. melanogaster (Lloyd et al., 2000) which lacks Doc2 proteins but where spontaneous release is still present. Numerous studies have suggested a role for rabphilin-3a as a regulator of SV cycling, docking and exocytosis, notably in adrenal chromaffin cells (Chung et al., 1995), PC12 cells (Komuro et al., 1996) and neurons (Deák et al., 2006;Stanic et al., 2015). However, electrophysiological investigation of Rph3a null mice did not reveal obvious physiological impairments (Schlüter et al., 1999). Others reported a positive function of rabphilin-3a in the recovery rate after intense neuronal stimulation (Deák et al., 2006). Altogether, literature supports the hypothesis that rabphilin-3a is involved in neurotransmitter release but the precise mechanism remains undefined. Due to its high similarity with Doc2 proteins and its Ca 2+ -binding properties, we hypothesized that rabphilin-3a might play a role in spontaneous release.
In neurons, Syt-7 is involved in asynchronous release (Bacaj et al., 2013). Syt-7 localization is still debated as it appeared anchored on the PM (Sugita et al., 2001) but also enriched on vesicles (Fukuda et al., 2004a). Conflicting observations were made in synaptic vesicles vs dense core vesicles. PM targeting is presumably caused by the absence of an N-glycosylation site in the luminal domain of Syt-7, which in Syt-1 is essential for vesicular targeting (Han et al., 2004). The C 2 A and C 2 B domains of Syt-7 exhibit a ~10-fold higher Ca 2+ affinity than the corresponding domains of Syt-1 (Sugita et al., 2002). In cell-free assays, phospholipid binding by Syt-7 occurs near resting and residual [Ca 2+ ] concentrations (EC 50 of 1.5 μM and 2.5 μM Ca 2+ for C 2 A and C 2 B respectively). Its dissociation from phospholipids is extremely slow (Hui et al., 2005) (K 1 ~ 19.7 s − 1 and K 2 ~ 8 s − 1 for a double exponential fit). These properties are consistent with the established role of Syt-7 as a sensor for asynchronous release, but also make it conceivable that it could trigger spontaneous release. In synapses co-expressing both sensors, Syt-7 mediated asynchronous release is normally occluded by fast Syt-1-induced release (Bacaj et al., 2013;Turecek and Regehr, 2018), suggesting that both types of release compete with each other. Syt-7 is implicated in synaptic facilitation in various hippocampal (Jackman et al., 2016) and granule cell synapses (Turecek and Regehr, 2018). Moreover, Syt-7 driven asynchronous exocytosis seems to make a significant contribution to high frequency neurotransmission in the Calyx of Held: tonic release slightly depolarizes the postsynaptic cell, which helps to maintain high-fidelity spike transmission (Luo and Südhof, 2017).
Thus, together with Syt-1 or Syt-2, Syt-7 generates AP-dependent secretion but its role in spontaneous release is still unclear. Syt-7 was shown to have no effect on spontaneous release upon its ablation in wildtype neurons (Liu et al., 2014). On the other hand, Syt-7 overexpression in Syt-1 KO hippocampal neurons decreased mIPSC frequency (Bacaj et al., 2013) and Syt-7 removal in Syt-2 KO mice (Syt-7 KO vs Syt-2/7 double KO) augmented mEPSCs at the Calyx of Held (Luo and Südhof, 2017), suggesting a clamping function. Yet, the effect of Syt-7 on spontaneous release is related to the presence or absence of fast Syts and its putative role in this process was not investigated in absence of Doc2a and -b.

Aim of this study
Taken together, we hypothesized that Doc2c, rabphilin-3a or Syt-7 might regulate spontaneous release and investigated spontaneous excitatory release in Doc2a and -b deficient hippocampal neurons lacking either Doc2c, rabphilin-3a or Syt-7. Neither Doc2c nor Syt-7 were implicated in spontaneous release in neuronal networks and autapses. In absence of Syt-7, we detected no impairment in synchronous or asynchronous release but did observe a role in synaptic facilitation. Rabphilin-3a inhibited mEPSCs in networks but not autapses, suggesting that it does not function as a Ca 2+ sensor but may regulate synaptic release by a mechanism specific to networks.

Mouse lines
Animals were housed, bred and handled in accordance with Dutch and EU governmental guidelines. Protocols were approved by the VU University Animal Ethics and Welfare Committee (approval number FGA 11-06). All mouse lines were extensively backcrossed and maintained on C57BL/6J genetic background. Doc2a, b DKO mice were previously described (Groffen et al., 2010) and maintained as homozygous lines. For Doc2c KO mouse line generation (Caliper Life Sciences,), a gene targeting vector was constructed with homology arms from the Doc2c locus on chromosome 19, using BAC clone RP23-34E20. The 5 ′ homology arm (~5.6 kb), 3 ′ homology arm (~3.5 kb) and the conditionally deleted region (~1.4 kb) were all generated by PCR. These fragments were cloned sequentially in the FtLoxNwCD vector. Exons 3-7 were flanked by loxP sites to allow Cre-mediated excision of amino acids 82 -232, causing a frame shift of the downstream exons (see Fig. 1). The Neo expression cassette (for positive selection of the ES cells) was flanked by Frt sequences for subsequent removal of the Neo cassette. A Diphteria toxin A expression cassette was used for negative selection against ES cells originating from random integration events. The final vector was confirmed by sequence analysis. NruI was used for linearizing the final vector prior to electroporation of ES cells from C57BL/6 mice. Correctly targeted clones were identified by PCR, southern blotting and karyotyping. After blastocyst injection, male chimeras were bred with wildtype C57BL/6 females to generate heterozygotes. The strain was sequentially crossed to a Flp-and Cre deleter strain (Schmidt-Supprian et al., 2007;Wu et al., 2009), both maintained on a C57BL/6 inbred background to generate a Doc2c +/− colony that was maintained by congenic mating with C57BL/6 inbred mice for several generations. The Rph3a KO mouse line was kindly donated by Prof. Thomas Südhof and maintained on a C57BL/6 background as described (Schlüter et al., 1999). Rph3a KO mice were mated with Doc2a, b DKO mice to create Doc2a − /− , b − /− , Rph3a +/− mice, giving Doc2a − /− , b − /− DKO and Doc2a − /− , b − /− , Rph3a − /− TKO littermates. Syt-7 mutant mice were previously described

Primary culture of hippocampal neurons
All mouse lines were viable. Neurons were dissociated from pups at embryonic day 18 (E18) or postnatal day 1 (P1). Immediately after decapitation, tails were clipped for speed genotyping which was done in parallel during tissue dissection and dissociation of neurons. Homozygous wildtype and KO mice from the same litter were used. For Doc2c, in case no wildtype homozygotes were present, we used heterozygotes and kept those experiments separate for comparison with wildtype homozygotes. Because the heterozygous and homozygous wildtypes showed the same phenotype, the data were combined into a single control group. Hippocampi were isolated after brain dissection in Hanks buffered salt solution (HBSS, Sigma) buffered with 1 mM HEPES (Invitrogen). Neuronal tissue was incubated for 20 to 30 min in 0.25% trypsin (Invitrogen) at 37 • C and washed three times in DMEM (Invitrogen). Cells were dissociated by trituration with a fire-polished Pasteur pipette and spun down during 5 min at 1000 rpm. The pellet was resuspended in Neurobasal (Gibco) prewarmed at 37 • C and cells were counted in a Fuchs-Rosenthal chamber. Neurons were plated in prewarmed Neurobasal medium supplemented with 2% B-27, 1.8% 1 M HEPES, 0.25% glutamax and 0.1% Pen-strep (all from Invitrogen) as previously established (Wierda et al., 2007). Electrophysiology experiments were performed in network or autaptic cultures. For network cultures, hippocampal neurons were plated at a density of 25 K cells per well in 12wells plates on etched glass coverslips containing a confluent layer of rat astrocytes (Wierda et al., 2007). For autaptic cultures, 1.5 K cells per 12well were plated on coverslips with astrocyte micro-islands (Wierda et al., 2007). Cultured neurons were incubated at 37 • C and 5% CO 2 and used for experiments between day-in-vitro 14 and 21 (DIV 14-21).
The patch pipettes were made of borosilicate and pulled using a multi-step filament puller (P-1000, Sutter Instruments, Novato, USA) to reach a pipette resistance from 3 to 5 MΩ. In whole-cell configuration, neurons were voltage clamped at − 70 mV with an Axopatch 200B or Multiclamp 700B amplifier (Molecular Devices). Signal was low-pass filtered at 4 kHz and digitized at 10 kHz with a Digidata 1440A or 1550 (Molecular Devices). Neurons with a series resistance (Rs) exceeding 15 MΩ, with an Rs increase beyond 20% of the initial value or with a leakage below − 500 pA were excluded. Rs was compensated to 70%. EPSCs were elicited by depolarizing the cell to 0 mV for 1 ms. Standard stimulation paradigms comprised spontaneous activity recording, paired pulse stimuli with intervals of 1000, 500, 200, 100, 50 and 20 ms, two trains of each 100 action potentials at 5 Hz and 40 Hz. Each train was followed by a single stimulus at 2 s after the last depolarization to test synaptic recovery. The first EPSC is equal to the first part of the 1000 ms paired pulse recording.
Miniature EPSCs were detected using Mini Analysis 6.0 (Synaptosoft Inc.), using a threshold amplitude of 7 to 10 pA depending on recording quality. Evoked release events were analyzed using an in-house routine in the MATLAB® environment (He et al., 2017;Mathworks) to calculate the paired pulse ratio, the EPSC charge and amplitude. In AP-induced burst EPSCs, synchronous and asynchronous components were isolated from total response charge as follows. Stimuli artefacts were removed and data points surrounding the artefact were connected by a straight line. A second line was calculated to connect the EPSC values preceding each stimulus. The area below the first and second line was considered as synchronous, the area between the second line and baseline was considered as asynchronous charge. Both components were calculated using cubic interpolation. Averaged trains traces were obtained using a custom written script in Igor 6.37 (Wavemetrics).

Statistical analysis
All statistical analysis was performed using SPSS v.25.00 (IBM Corp., Armonk, NY, USA). All data are reported by boxplots together with single data points. The number of measurements "n", indicate the number of cells per group and the number of independent observations "N", represent independent experimental weeks. One experimental week generally corresponds to one pup for each genotype. If multiple pups were used for a single genotype, the dissociated neurons were pooled before seeding. One to three cells were recorded from a single coverslip. Data were checked for normality assumption using Shapiro-Wilk and Kolmogorov-Smirnov tests. Homogeneity of the variance was assessed with Levene's test. Considering that groups were independently acquired from one another, independent samples tests were performed. When assumptions were met, independent t-tests were performed. For non-normally distributed data, Mann-Whitney tests were executed. For each experiment, the p-value alpha significance threshold was adjusted for multiple testing. The effect size was calculated for the independent samples t-test as d = t ̅̅̅̅̅̅̅̅̅̅̅ N1+N2 N1N2 √ and for Mann-Whitney U test as ƞ 2 = Z2 N− 1 .

Results
To identify potential Ca 2+ sensors responsible for the remaining spontaneous release component in Doc2a (DKO) neurons, we selected three C 2 domain containing proteins as putative candidate sensors: Doc2c, rabphilin-3a and Syt-7.

Doc2c does not contribute to spontaneous release in absence of Doc2a & b
To examine possible functions of Doc2c in spontaneous release, we generated Doc2c null mice (see details in Methods section) which were viable and fertile. The mice were mated to previously described Doc2a − / − , b − /− mice to produce DKO (Doc2a − /− , b − /− , c +/+ ) and TKO (Doc2a − / − , b − /− , c − /− ) littermates. Hippocampal neurons in primary network culture were analyzed by whole cell voltage-clamp recordings to measure mEPSCs in presence of tetrodotoxin (TTX) and Gabazine (Gz). Homozygous wildtypes (Doc2c +/+ ) and heterozygotes (Doc2c +/− ) did not show differences for all parameters tested (data not shown) and were therefore pooled together in one control group. The removal of Doc2c did not affect the rate of spontaneous release in network cultures ( Fig. 2A-B). The amplitude, charge, rise time and decay were also not significantly altered (Fig. 2B, Table 1), dismissing a postsynaptic effect. To confirm these results, we performed the same experiment in autaptic neurons ( Fig. 2C-D). In autapses the resting membrane potential was set to − 70 mV, allowing to record mEPSCs in absence of TTX. Again, no modification in the frequency or postsynaptic attributes of mEPSCs was found in absence of Doc2c. We thus conclude that Doc2c does not contribute to the spontaneous release events in DKO neurons.

Doc2c is not involved in AP-dependent release from DKO neurons
The autaptic culture preparation allows to concomitantly investigate any possible effects on AP-dependent neurotransmitter release. We applied a stimulation paradigm to assess the first evoked EPSC, paired pulse plasticity and repetitive stimulation trains (Fig. 3). The first AP response and the paired-pulse ratio (PPR) in TKO and DKO neurons were not significantly different (Fig. 3A-D and Table 1). Likewise, synaptic release rundown occurred normally during trains of 100 AP with pulse intervals of 5 Hz (Fig. 3E-K) or 40 Hz (Fig. 3L-R). The proportion of synchronous and asynchronous release was unchanged (Fig. 3J,Q).
Few Ca 2+ -sensors can potentiate exocytosis after tetanic burst (Habets and Borst, 2005;Lee et al., 2010;Nanou et al., 2016). To explore if Doc2c has such a role in post-burst potentiation (PTP), we calculated the recovery ratio of the initial EPSC of the train and a single pulse given 2 s after the end of the train. Again, no significant difference was noted for both the 5 Hz (Fig. 3K) and the 40 Hz (Fig. 3R) stimulation. We conclude that Doc2c does not function in spontaneous or evoked release in glutamatergic hippocampal neurons lacking Doc2a and -b.

Rabphilin inhibits spontaneous release in hippocampal networks but not autaptic neurons
We continued our investigation with rabphilin-3a, another double C 2 domain protein which is the closest ortholog of Doc2 proteins in D. melanogaster, has an extremely high Ca 2+ sensitivity and is involved in multiple steps of the SV cycle (Deák et al., 2006;Komuro et al., 1996). Rabphilin-3a KO mice are viable and fertile. They carry a frameshift deletion of two exons encoding residues 74-146 of rabphilin-3a (Schlüter et al., 1999). The mice were mated with Doc2a,b DKO to produce TKO mice (Doc2a − /− , b − /− , Rph3a − /− ).
We first recorded spontaneous release in Doc2a − /− , b − /− , Rph3a +/+ DKO versus Doc2a − /− , b − /− , Rph3a − /− TKO hippocampal neurons cultured in networks (Fig. 4A-B). The mEPSC release rate significantly increased (Fig. 4B) upon rabphilin-3a deletion (from 3.45 ± 0.72 Hz in DKO to 7.50 ± 1.63 Hz in TKO neurons, Table 1). To exclude a confounding effect of experimental variation between individual datasets, we also calculated the normalized mEPSC frequency (i.e. the relative frequency to that of the control group within each independent experiment). This showed a similar significant difference (Fig. 4B). The decay was slightly but significantly faster (exponential time constant of 1.57 ± 0.04 ms in DKO vs. 1.29 ± 0.09 ms in TKO neurons, Fig. 4B). The amplitude and rise time remained unchanged. In autaptic neurons however (Fig. 4C-D), the frequency and kinetics of mEPSCs were unchanged. Thus, based on the results in network cultures, rabphilin-3a might inhibit spontaneous release, but we found no evidence for such a function in autaptic neurons.

Rabphilin-3a has no major function in evoked release in Doc2a, b DKO neurons
We next investigated evoked release properties in DKO versus TKO autaptic neurons lacking rabphilin-3a. Similar as described above for Doc2c, we stimulated self-innervating autaptic neurons to assess synaptic strength in resting (first EPSC) and repetitively stimulated neurons (Fig. 5). When comparing experiments obtained in different experimental weeks (e.g. Fig. 3D vs. 5D), there was a variation in the degree of short-term plasticity. This may be due to many experimental variables (e.g. quality of glia or many other parameters). In our experimental design, we aimed to directly compare genotypes within the same experiments, thereby preventing confounding effects of such experimental variation.
When comparing DKO vs TKO in the same experiment, we found no significant difference in the amplitude or shape of the first EPSC, nor in short-term plasticity (Fig. 5A-D). During repetitive low frequency stimulations, both groups showed a similar synaptic fatigue (Fig. 5E-K). The synchronous component during 5 Hz stimulation showed similar rundown (Fig. 5G & J). During 40 Hz stimulation, the asynchronous release component appeared slightly smaller in absence of rabphilin-3a (Fig. 5N). However, the cumulative asynchronous charge or other release parameters resulting from 40 Hz trains (Fig. 5O-Q) were not significantly reduced. EPSC recovery during a 2 s rest interval after the burst was also unchanged in both 5 Hz and 40 Hz stimulation paradigms (Fig. 5K,R). Moreover, the RRP size and the replenishment rate remained similar among DKO and TKO autapses (data not shown). In conclusion, rabphilin-3a ablation in absence of Doc2a and -b had no significant effect on AP-dependent exocytosis in autaptic neurons.

Syt-7 does not contribute to spontaneous release
Previous investigations revealed no direct function of Syt-7 in spontaneous release (Bacaj et al., 2013;Luo and Südhof, 2017) but its overexpression in Syt-1 KO hippocampal neurons and ablation in Syt-2 KO neurons suggested a clamping function (Bacaj et al., 2013;Luo and Südhof, 2017). However, the involvement of Syt-7 in spontaneous release in absence of Doc2a and -b was not yet investigated.
In neuronal networks, 5KO neurons presented a normal mEPSC frequency and we observed no significant difference in any parameters evaluated compared to cells from QKO littermates (Fig. 6A-B). Measurements in autaptic neurons confirmed that the lack of Syt-7 had no effect on mEPSCs (Fig. 6C-D). These findings indicate that Syt-7 has no role in spontaneous release even in absence of members of the Doc2 / rabphilin-3a family, extending previous observations in wildtype neurons (Liu et al., 2014).

Syt-7 removal affects short-term plasticity in presence of Syt-1
We next investigated evoked fast and delayed release in the QKO versus 5KO autapses (Fig. 7). Syt-7 removal had no effect on first EPSC amplitude or charge (Fig. 7A, C). However, 5KO neurons showed a reduced PPR (Fig. 7B, D) with a more pronounced short-term depression (STD) in line with previous investigations (Jackman et al., 2016). For 5 Hz and 40 Hz trains we observed a normal rundown of EPSC amplitude, a normal proportion of synchronous and asynchronous release and a normal recovery rate (Fig. 7E-R). The deeper STD induced in Syt-7 KO was not associated with changes in evoked release during burst stimuli. Unlike in the fast spiking Calyx of Held synapse (Luo and Südhof, 2017), the asynchronous defect was not detectable in presence of the fast sensor (in this case, Syt-1). The RRP size and the pool recovery were similar between the two genotypes (data not shown). Our results confirm that Syt-7 is specifically involved in short-term facilitation in presence of Syt-1 in hippocampal synapses.

Discussion
The synaptic mechanism responsible for the residual spontaneous secretion in absence of Doc2a and b remains unknown. Various lines of evidence indicate a role for Ca 2+ in regulating the spontaneous release rate (Emptage et al., 2001;Ermolyuk et al., 2013;Groffen et al., 2010;Llano et al., 2000) but no direct link has been found between Ca 2+ signalling, Ca 2+ -sensors and residual minis. Here, we selected three candidate genes encoding C 2 domains based on their homology to Doc2b and/or high Ca 2+ -sensitivity.
Despite their high resemblance with Doc2a and -b, removal of Doc2c and Syt-7 did not influence spontaneous release in excitatory hippocampal neurons. For Doc2c, this may be expected in view of the loss of the conserved five-aspartate motif in the C 2 A domain (Fig. 1B). Furthermore, in the human genome, the exon structure of mouse Doc2c is not conserved and the corresponding sequence is annotated as a pseudogene (DOC2GP) which is transcribed but not translated into a protein product (Craxton, 2010). In mice, on the other hand, Doc2c encodes a protein of similar length as Doc2a/b and contains a well conserved MID and C 2 B domain, which theoretically could allow an alternative activation mode of the protein in response to DAG production by phospholipase C. Our results now clearly rule out a role of Doc2c in spontaneous glutamate release, but do not fully preclude potential roles in other secretory processes.
Rabphilin-3a seemed to inhibit spontaneous release frequency in networks. Without additional experiments, it is uncertain if the observed effect depends on the Ca 2+ -binding ability of rabphilin-3a. Importantly, the inhibitory activity did not appear in autaptic cultures. Similar contradictory results between autaptic and dissociated network cultures were already reported for Syt-1 KO neurons (Liu et al., 2009;  Sorensen, 2014). This effect is likely caused by an interdependence of glutamatergic and GABAergic innervation for spontaneous release regulation because networks contain mixed glutamatergic/GABAergic inputs. It is established that mEPSCs are negatively correlated with mIPSCs (Wierda and Sorensen, 2014) and that the Syt-1 KO phenotype was unveiled only in mixed glutamatergic/GABAergic pairs but not in homotypic culture (Wierda and Sorensen, 2014). Given the similarities with our results, we hypothesize that the same mechanism is at play for rabphilin-3a, giving rise to the discrepancy between network and autaptic culture. Another possibility would be that rabphilin-3a inhibits spontaneous release by a mechanism that requires neurotrophic factors that are lacking or diminished in autaptic culture. Most studies could not attribute a specific function to rabphilin-3a. While Rab3A alone causes a significant synaptic phenotype (Castillo et al., 1997;Geppert et al., 1994aGeppert et al., , 1997 and deletion of all four Rab3 isoforms is lethal (Schlüter et al., 2004), the deletion of rabphilin-3a in mice (Schlüter et al., 1999) or C. elegans (Staunton et al., 2001) does not produce a major phenotype. Deák and colleagues investigated the effect of single Rph3a gene deletion on spontaneous release and reported that this accelerated synaptic recovery both in single-knockout mice and in combination with a synaptobrevin-2 (Syb-2) null allele (Deák et al., 2006). In the Rph3a single KO, minis were not affected but rabphilin-3a deletion in Syb-2 null mice significantly reduced the mini frequency. Although a direct comparison is complicated by different genetic backgrounds and the fact that they investigated evoked release in high density cultures, our observations are not consistent with these findings. All considered, it seems unlikely that rabphilin-3a plays an important role in enhancing spontaneous release in a manner similar to the Doc2a, b proteins. Its regulation of mEPSC frequency might instead be via a mechanism specifically active in networks but not autapses.
The alteration of Syt-7 expression levels in Syt-1 or -2 KO neurons changed the frequency of mEPSCs (Luo and Südhof, 2017) and mIPSCs (Bacaj et al., 2013), suggesting a clamping function. However, upon its ablation in wildtype neurons, no effect in spontaneous release was found (Liu et al., 2014). Our results confirmed that Syt-7 does not function in spontaneous release in absence of the main sensors for spontaneous release (Doc2a,b) and the potential Ca 2+ -sensor candidates Doc2c and rabphilin-3a. One explanation could be that the presence of fast Syts masks the function of Syt-7 in stochastic exocytosis. To further investigate this hypothesis, it would be interesting to knock down Syt-1 in the same genetic background used in this study. Syt-7 is subject to several post-translational modifications. It can undergo palmitoylation which affects Syt-7 targeting (Flannery et al., 2010) but also phosphorylation which regulates the secretion of insulin from pancreatic beta cells (Wu et al., 2015b). These post-translational modifications may affect both Syt-7 targeting and its function in neurotransmission. Although we found no significant effect for Syt-7 on spontaneous release, it does not preclude its implication in other cells types or systems in which Syt-7 post-translational modification could affect its capacity to regulate spontaneous release.
Previous investigations already demonstrated that Syt-7 is required for facilitation at several central synapses such as mossy fibers, corticothalamic projections, Schaffer collaterals and the perforant path (Jackman et al., 2016). Its role as Ca 2+ -sensor for asynchronous release in neurons was identified earlier and appeared visible only upon Syt-1 removal in most neuronal types (Bacaj et al., 2013) except in particular fast spiking synapses such as the calyx of Held, where it contributes to time-locked information transfer even in presence of Syt-2 (Luo and Südhof, 2017). Our results confirmed that in presence of Syt-1, Syt-7 removal only impacts short-term plasticity but not asynchronous release in hippocampal synapses.
Our study used an extracellular Ca 2+ concentration of 4 mM, which is slightly higher than physiological. It has been well established in our lab that hippocampal neurons are viable for many hours of recording without measurable effects on synaptic parameters, both in network and autaptic preparations (Bourgeois-Jaarsma et al., 2019;Díez-Arazola et al., 2020;Meijer et al., 2019). The relatively high Ca 2+ concentration favorably affects mEPSC amplitude and frequency, increasing the sensitivity of our analysis (Williams and Smith, 2018) (Bacaj et al., 2013;Groffen et al., 2010;Xu et al., 2009).   Our study focused on excitatory neurons. Recent findings revealed a different implication of the major Ca 2+ -sensors Syt-1 and Doc2 in excitatory versus inhibitory neurons (Courtney et al., 2018). In this study, both Syt-1 and Doc2 were confirmed to regulate minis, but their relative contributions depended on whether release was from excitatory or inhibitory neurons (Courtney et al., 2018). Certain mechanisms that trigger spontaneous release at central synapses are different at excitatory and inhibitory synapses (Williams and Smith, 2018). While Doc2c, rabphilin-3a and Syt-7 do not play a prominent role in spontaneous glutamate release, this conclusion cannot be extrapolated to inhibitory neurons without further investigation.
To conclude, the mechanism for Ca 2+ -dependent regulation of the residual spontaneous release in Doc2a,b DKO hippocampal neurons remains elusive. Other C 2 domain proteins could potentially contribute to residual mEPSCs including Syts, Munc-13 s, ferlins (Wu et al., 2015a) and copines (Liu et al., 2018). Expression levels of those proteins in hippocampal glutamatergic synapses can be taken in consideration to select the most likely effector. Syt-11, Copine6 (Cpne6) and Munc13-1 (Unc13a) isoforms are the most expressed isoforms among the proteins cited above (Saunders et al., 2018) and therefore the C 2 proteins most likely involved in spontaneous release regulation, if any.
The active zone proteins Munc13-1 to 4 contain multiple C 2 domains and function in priming and RRP generation. Munc13 is involved in mEPSC potentiation by phorbol esters (Lou et al., 2008). Munc13-1 binds Ca 2+ and interacts in a Ca 2+ -dependent manner with calmodulin, an important signalling partner for exocytosis modulation. Based on its expression level (Saunders et al., 2018), Munc13-1 is the best candidate among Munc13 isoforms. The Ca 2+ -sensing protein otoferlin is involved in spontaneous release at mouse inner hair cell ribbon synapses (Takago et al., 2018). Copine-6, a two C 2 -domain-containing protein, localizes to presynaptic terminals and binds to Syb-2 and other SNARE proteins in a Ca 2+ -dependent manner. It suppresses spontaneous neurotransmission once bound to Syb-2 (Liu et al., 2018). It is also possible that Ca 2+sensors responsible for the residual spontaneous release haven't been discovered yet. The wide variety and expression of C 2 domain proteins present at the active zone render the identification of the protein responsible for residual spontaneous release very challenging.
Another alternative mechanism for the occurrence of residual mEPSCs in absence of established Ca 2+ sensors is that residual spontaneous release isn't regulated by presynaptic Ca 2+ but instead depends on "sensorless" fusion only requiring SNARE complexes. Indeed, the three neuronal SNARE proteins, syntaxin, SNAP25, and Syb-2, are thought to constitute the minimal components sufficient to drive membrane fusion. In vitro studies (Weber et al., 1998) have shown that neuronal SNAREs can drive the fusion between synthetic liposomes. A more recent study replicated this finding in yeast vacuoles (Ko et al., 2014) where they showed that neuronal SNARE proteins can induce vacuole fusion in a Rab-and SM protein-independent manner.
Our investigation revealed that Doc2c and Syt-7 have no major contribution to the residual mEPSC frequency but rabphilin-3a seems to be implicated in spontaneous release in neuronal networks. This further reduces the collection of candidate proteins that might contribute to spontaneous release and points at rabphilin-3a for deeper investigation of quantal release regulation in inhibitory neurons.

Declaration of competing interest
All authors declare that they have no competing interests.