Munc13 supports fusogenicity of non-docked vesicles at synapses with disrupted active zones

  1. Chao Tan
  2. Giovanni de Nola
  3. Claire Qiao
  4. Cordelia Imig
  5. Richard T Born
  6. Nils Brose
  7. Pascal S Kaeser  Is a corresponding author
  1. Department of Neurobiology, Harvard Medical School, United States
  2. Department of Neuroscience, University of Copenhagen, Denmark
  3. Department of Molecular Neurobiology, Max Planck Institute for Multidisciplinary Sciences, Germany

Abstract

Active zones consist of protein scaffolds that are tightly attached to the presynaptic plasma membrane. They dock and prime synaptic vesicles, couple them to voltage-gated Ca2+ channels, and direct neurotransmitter release toward postsynaptic receptor domains. Simultaneous RIM + ELKS ablation disrupts these scaffolds, abolishes vesicle docking, and removes active zone-targeted Munc13, but some vesicles remain releasable. To assess whether this enduring vesicular fusogenicity is mediated by non-active zone-anchored Munc13 or is Munc13-independent, we ablated Munc13-1 and Munc13-2 in addition to RIM + ELKS in mouse hippocampal neurons. The hextuple knockout synapses lacked docked vesicles, but other ultrastructural features were near-normal despite the strong genetic manipulation. Removing Munc13 in addition to RIM + ELKS impaired action potential-evoked vesicle fusion more strongly than RIM + ELKS knockout by further decreasing the releasable vesicle pool. Hence, Munc13 can support some fusogenicity without RIM and ELKS, and presynaptic recruitment of Munc13, even without active zone anchoring, suffices to generate some fusion-competent vesicles.

Editor's evaluation

Tan and colleagues studied synaptic transmission, presynaptic protein levels, and synaptic ultra-structure in hippocampal cultures of mice lacking the key active-zone proteins RIM (1, 2), ELKS (1, 2), and Munc13 (1, 2). Compared to cultures lacking only RIM and ELKS, additional deletion of Munc13 results in a further decrease of synaptic Munc13-1 levels, a similar reduction of the number of docked synaptic vesicles, and a more pronounced decrease of the readily releasable vesicles. The results support the conclusion of the nonredundant role of Munc13 in synaptic vesicle priming. Overall, this study reinforces the notion that synapse formation is a remarkably resilient process that occurs even under strong perturbation of presynaptic function.

https://doi.org/10.7554/eLife.79077.sa0

Introduction

Neurotransmitter release is mediated by synaptic vesicle fusion at presynaptic active zones, and Munc13 proteins have a central role in this process (Brunger et al., 2018; Dittman, 2019; Wojcik and Brose, 2007). In addition to Munc13, active zone scaffolds contain RIM, ELKS, RIM-BP, Bassoon/Piccolo, and Liprin-α. Together, they form a protein machine that controls the speed and precision of synaptic transmission by docking and priming of synaptic vesicles, by organizing the coupling of these vesicles to sites of Ca2+ entry, and by targeting transmitter release toward postsynaptic receptor domains (Biederer et al., 2017; Südhof, 2012). Given the molecular complexity of the active zone and the essential roles of several of its protein components, understanding its assembly and mode of action has remained both a key aim and a challenge in cellular neuroscience.

Mouse knockout studies established major roles for Munc13 in vesicle priming at central nervous system synapses (Augustin et al., 1999; Varoqueaux et al., 2002). This process renders vesicles fusion-competent and adds them to the pool of readily releasable vesicles (RRP) that can rapidly undergo exocytosis upon action potential arrival. The functional RRP can be probed experimentally by applying stimuli that deplete it, for example, superfusion with hypertonic sucrose solution (Kaeser and Regehr, 2017; Rosenmund and Stevens, 1996). When Munc13 is deleted, release competence of vesicles is abolished or strongly decreased across all tested synapses in multiple model organisms, including mouse, Drosophila melanogaster, and Caenorhabditis elegans (Aravamudan et al., 1999; Augustin et al., 1999; Richmond et al., 1999; Varoqueaux et al., 2002). Munc13 ablation also results in an almost complete loss of docked vesicles, as defined by plasma membrane attachment in electron micrographs (Hammarlund et al., 2007; Imig et al., 2014; Siksou et al., 2009). These findings led to the model that vesicle docking and priming are morphological and functional states that correspond to release competence, a notion that is supported by similar correlations upon ablation of SNARE proteins (Chen et al., 2021; Hammarlund et al., 2007; Imig et al., 2014; Kaeser and Regehr, 2017). Due to their core function in vesicle priming, Munc13s mediate the resupply of fusion-competent vesicles as they are spent during synaptic activity and thereby control short-term plasticity and recovery from synaptic depression (Lipstein et al., 2013; Lipstein et al., 2021; Rosenmund et al., 2002). Furthermore, Munc13 nano-assemblies may account for secretory hotspots and recruit the SNARE protein syntaxin-1 (Reddy-Alla et al., 2017; Sakamoto et al., 2018).

Several recent studies took the approach to ablate combinations of active zone protein families, allowing analyses of active zone assembly and function upon elimination of redundant components (Acuna et al., 2016; Brockmann et al., 2020; Kushibiki et al., 2019; Oh et al., 2021; Tan et al., 2022; Wang et al., 2016; Zarebidaki et al., 2020). We simultaneously deleted RIM1, RIM2, ELKS1 and ELKS2, which resulted in a loss of RIM, ELKS, and Munc13, and in strong decreases of Bassoon, Piccolo, and RIM-BP levels at active zones (Tan et al., 2022; Wang et al., 2016; Wong et al., 2018). This disruption of active zone assembly led to a near-complete loss of synaptic vesicle docking as studied by electron microscopy and to an ~85% decrease in single action potential-evoked exocytosis as assessed electrophysiologically. However, some transmitter release persisted: up to ~35% of release evoked by hypertonic sucrose and ~50% of spontaneous release events remained under the tested experimental conditions, increasing extracellular Ca2+ to increase vesicular release probability p strongly enhanced single action potential-evoked release, and stimulus trains released vesicles surprisingly efficiently (Wang et al., 2016). Hence, some releasable vesicles persisted despite the loss of most docked vesicles. These findings supported alternative mechanistic models in which docking and priming are independent processes mediated by distinct molecular functions of Munc13 (Kaeser and Regehr, 2017). Further support for this model came from experiments with artificial retargeting of Munc13 to synaptic vesicles rather than to active zones, which increased vesicle fusogenicity but not docking in mutants that lack most RIM and ELKS sequences and docked vesicles (Tan et al., 2022). Together, these findings suggest that non-docked vesicles can contribute to the functional RRP after removal of RIM and ELKS. What remained enigmatic, particularly in view of the near-complete loss of Munc13 from the target membrane after RIM + ELKS ablation, was whether undocked vesicles engaged Munc13 for vesicle priming or used an alternative priming pathway.

In this study, we tested directly whether Munc13 is necessary to prime vesicles after the strong active zone disruption upon RIM + ELKS knockout. We generated mice to ablate Munc13-1 and Munc13-2 in addition to RIM and ELKS (that is RIM1, RIM2, ELKS1, and ELKS2) and compared phenotypes of hextuple knockout neurons with those from conditional RIM + ELKS knockouts only. We found that synapses formed despite this strong genetic manipulation and that overall their ultrastructure was largely normal except for a lack of docked vesicles. However, Munc13 ablation on top of RIM and ELKS knockout further impaired single action potential-evoked release and decreased the RRP at both excitatory and inhibitory synapses. Paired pulse ratios, used to monitor p, were not further affected by the additional removal of Munc13. Our data establish that Munc13 can functionally prime some vesicles in the absence of RIM and ELKS, indicate that Munc13 away from active zones is sufficient to confer vesicle fusogenicity, and support a growing body of evidence showing that synapse formation is overall resilient to severe perturbations of synaptic protein content and of synaptic activity. We propose that Munc13 recruitment to presynapses is rate-limiting to generate fusion competence of synaptic vesicles.

Results

Some synaptic Munc13 remains after RIM + ELKS knockout

With the overall goal to determine whether Munc13 mediates addition of vesicles to the functional RRP after RIM + ELKS knockout, we first confirmed key effects on synaptic transmission and Munc13 active zone levels in cultured neurons after ablation of RIM + ELKS (Figure 1) that we had described before (Tan et al., 2022; Wang et al., 2016). We cultured primary hippocampal neurons of mice with floxed (fl) alleles for RIM1, RIM2, ELKS1 and ELKS2 (Figure 1A). At 5 days in vitro (DIV5), a time point that is before the detection of functional synapses in these cultures (Held et al., 2020; Mozhayeva et al., 2002), the neurons were transduced with Cre-expressing lentiviruses or control lentiviruses (that express an inactive mutant of cre) to generate cKOR+E neurons or controlR+E neurons, respectively. We previously established that this induces strong defects in active zone assembly and neurotransmitter release, but the neurons form synapses at normal numbers, and postsynaptic receptor assemblies and functions are preserved (Tan et al., 2022; Wang et al., 2016). We first confirmed that excitatory and inhibitory synaptic transmission are strongly impaired in cKOR+E neurons (Figure 1B–F). Synaptic responses were induced by focal electrical stimulation, and whole-cell recordings served to monitor excitatory and inhibitory transmission via glutamate and GABAA receptors (GABAARs), respectively. Action potential-evoked excitatory transmission was monitored via NMDA receptors (NMDARs) to prevent the strong reverberant activity that is induced by stimulation of these neuronal networks when AMPA receptors (AMPARs) are not blocked.

Figure 1 with 3 supplements see all
Action potential-evoked neurotransmitter release and Munc13 active zone levels after ablation of RIM + ELKS.

(A) Strategy for deletion of RIM1, RIM2, ELKS1, and ELKS2 in cultured hippocampal neurons. Neurons of mice with floxed alleles for all four genes were infected with Cre-expressing lentiviruses (to generate cKOR+E neurons) or lentiviruses expressing a recombination-deficient version of Cre (to generate controlR+E neurons) as described (Tan et al., 2022; Wang et al., 2016). (B, C) Sample traces (B) and quantification (C) of excitatory postsynaptic currents (EPSCs) evoked by focal electrical stimulation, controlR+E 20 cells/3 cultures, cKOR+E 19/3. (D–F) Sample traces (D) and quantification of amplitudes (E) and 20–80% rise times (F) of inhibitory postsynaptic currents (IPSCs) evoked by focal electrical stimulation, 18/3 each. (G–K) Sample stimulated emission depletion (STED) microscopic images (G) and quantification (H–K) of side-view synapses in cultured hippocampal neurons stained for Munc13-1 (imaged in STED), PSD-95 (imaged in STED), and Synaptophysin (imaged in confocal). In (H, I), fluorescence levels at each position of line profiles (600 nm × 200 nm) positioned perpendicular to the center of the elongated PSD-95 structure and aligned to the PSD-95 peak are shown; in (J, K), peak values for each line profile are shown independent of peak position, 60 synapses/3 cultures each. Data are mean ± SEM; ***p<0.001 as determined by Welch’s t-tests (C, E, F), two-way ANOVA followed by Bonferroni’s multiple-comparisons post-hoc tests (H, I), or unpaired two-tailed Student’s t-tests (J, K). For assessment of Munc13-1 levels after Munc13 knockout using STED microscopy, Synaptophysin levels in cKOR+E synapses, and comparison of Munc13-1 levels by STED microscopy in cKOR+E and cKOM synapses, see Figure 1—figure supplement 1; for assessment of synaptic transmission after Munc13 knockout, see Figure 1—figure supplement 2; for assessment of Munc13-1 expression by confocal microscopy and Western blotting in cKOR+E and cKOM neurons, see Figure 1—figure supplement 3. Source data 1 contains numerical values of all means, errors, and p-values for this and all figures.

We next evaluated Munc13-1 positioning and levels at the active zone using stimulated emission depletion (STED) superresolution microscopy (Figure 1G–K). We stained for Synaptophysin to mark the synaptic vesicle cloud (imaged by confocal microscopy), PSD-95 to mark the postsynaptic density (PSD, imaged by STED), and Munc13-1 (imaged by STED). In these experiments, side-view synapses are defined as a synaptic vesicle cluster with an elongated PSD-95 structure aligned at one side of the vesicle cloud as described before (Emperador-Melero et al., 2021a; Held et al., 2020; Nyitrai et al., 2020; Tan et al., 2022; Wong et al., 2018). To evaluate Munc13-1 levels in the active zone area, we quantified PSD-95 and Munc13-1 fluorescence levels in 600 nm × 200 nm areas that were positioned perpendicular to the PSD through the center of the PSD-95 signal, and plotted their line profiles (Figure 1H and I) and peak levels (Figure 1J and K). Based on these analyses, Munc13-1 was largely lost from the active zone area of cKOR+E synapses.

For comparison, we analyzed Munc13-1 antibody signals at the active zone after ablation of Munc13. We used mice with floxed alleles for Munc13-1 and constitutive knockout alleles for Munc13-2 and Munc13-3 (Figure 1—figure supplement 1A; Augustin et al., 2001; Banerjee et al., 2022; Varoqueaux et al., 2002). In cultured neurons of these mice, Cre expression removes Munc13-1 (cKOM) without the potential for compensation by Munc13-2 or -3. Control experiments were performed on the same cultures but with lentiviral expression of an inactive Cre (controlM). Munc13-1 was ablated efficiently in cKOM neurons (Figure 1—figure supplement 1B–F), with the leftover signal not distinguishable from background levels that are typically observed in this approach (Held et al., 2020; Nyitrai et al., 2020; Wong et al., 2018). When we compared Munc13-1 levels in cKOR+E and cKOM synapses, the remaining signal was somewhat higher in cKOR+E synapses (Figure 1—figure supplement 1G). These higher levels did not arise from a peak at the position of the active zone area (around –70 to –20 nm) (Tan et al., 2022; Wong et al., 2018). Instead, Munc13-1 levels appeared higher more broadly, and the ratio of Munc13-1 at cKOR+E vs. cKOM synapses shifted upward throughout the presynaptic bouton (Figure 1—figure supplement 1G). In both types of neurons, Synaptophysin signals remained unchanged (Figure 1—figure supplement 1H and I). As observed before in autaptic cultures (Banerjee et al., 2022), there was some release left at these Munc13 knockout synapses (Figure 1—figure supplement 2). This is likely due to the very small amount of remaining Munc13-1 when using this conditional allele (Banerjee et al., 2022).

Because some Munc13-1 could be detected in cKOR+E neurons by STED microscopy (Figure 1G–K, Figure 1—figure supplement 1G) and Western blotting (Wang et al., 2016), we compared synaptic Munc13-1 levels in cKOR+E and cKOM neurons (Figure 1—figure supplement 3). There were somewhat higher Munc13-1 signals in confocal microscopic images in cKOR+E synapses compared to cKOM synapses (Figure 1—figure supplement 3A, B, D, and E). Similarly, a slight Munc13-1 band was detected in Western blots of cKOR+E neurons, but not of cKOM neurons (Figure 1—figure supplement 3C and F). The remaining Munc13-1 signal in immunostainings detected in cKOM neurons is likely mostly composed of antibody background in these experiments as quantifications were done without background subtraction and noise levels of ~25% are common (Wang et al., 2016). Altogether, our data indicate that some Munc13-1 might remain in nerve terminals of cKOR+E synapses, but the remaining Munc13 is not efficiently concentrated in the active zone area apposed to the PSD, supporting previous data on RIM-mediated recruitment of Munc13 to the active zone (Andrews-Zwilling et al., 2006).

Synapses form after deletion of Munc13 in addition to RIM and ELKS

With the overall goal to test whether Munc13-1 mediates the remaining release in cKOR+E neurons, we crossed the conditional knockout mice for RIM1, RIM2, ELKS1, and ELKS2 to conditional Munc13-1 and constitutive Munc13-2 knockout mice (Figure 2A). Cultured hippocampal neurons from these mice were infected with Cre-expressing or control lentiviruses at DIV5 to generate cKOR+E+M and controlR+E+M neurons, respectively, to remove Munc13 in addition to RIM + ELKS. We first used STED microscopy to analyze the localization and levels of RIM1 and Munc13-1 . RIM1 and Munc13-1 were effectively removed from active zones of cKOR+E+M neurons (Figure 2B–K, Figure 2—figure supplement 1A). PSD-95 levels as judged by this method were unaffected in most quantifications (Figure 2C, E, and J), but a very small change was detected in Figure 2H. Because PSD-95 is used for synapse selection (see ‘Materials and methods’), this isolated change likely reflects small differences in synapse selection in the two conditions in this specific experiment. The Munc13-1 signal that remained in STED experiments of cKOR+E neurons (Figure 1I and K, Figure 1—figure supplement 1G) further decreased in cKOR+E+M neurons (Figure 2—figure supplement 1B). In confocal images, there was a further decrease in synaptic Munc13-1 levels in cKOR+E+M neurons compared to cKOR+E neurons (Figure 2—figure supplement 1C–E), and Munc13-1 was below detection threshold in Western blotting (Figure 2—figure supplement 1F). Finally, Synaptophysin puncta number, size, and intensity, analyzed with a custom-written code to perform automatic two-dimensional segmentation for object detection (Emperador-Melero et al., 2021a; Held et al., 2020; Liu et al., 2018), were indistinguishable between controlR+E+M and cKOR+E+M neurons (Figure 2L–O). This indicates that synapse formation is intact in the absence of the tested active zone proteins. Similar results were obtained when we analyzed synapses in neurons infected with Cre-expressing lentiviruses at DIV2 (Figure 2—figure supplement 2) instead of DIV5.

Figure 2 with 2 supplements see all
Simultaneous deletion of RIM, ELKS, and Munc13 does not disrupt synapse formation.

(A) Strategy for simultaneous deletion of RIM1, RIM2, ELKS1, ELKS2, Munc13-1, and Munc13-2 in cultured hippocampal neurons (cKOR+E+M). Neurons were infected with Cre-expressing lentiviruses (to generate cKOR+E+M neurons) or viruses expressing a recombination-deficient version of Cre (to generate controlR+E+M neurons). (B–F) Sample images (B) and quantification (C–F) of side-view synapses stained for RIM1 (STED), PSD-95 (STED), and Synaptophysin (confocal), controlR+E+M 65 synapses/3 cultures, cKOR+E+M 66/3. (G–K) Same as (B–F), but for synapses stained for Munc13-1 instead of RIM1, 63/3 each. (L–O) Overview confocal images of anti-Synaptophysin staining (L) and quantification of Synaptophysin puncta density (M), intensity (N), and size (O); Synaptophysin objects were detected using automatic two-dimensional segmentation, confocal images of Synaptophysin staining are from the experiment shown in (B–F), 17 images/3 cultures each. Data are mean ± SEM; ***p<0.001 as determined by two-way ANOVA followed by Bonferroni’s multiple-comparisons post-hoc tests (C, D, H, I), unpaired two-tailed Student’s t-tests (E, J, M–O), or Welch’s t-tests (F, K). For Synaptophysin levels in cKOR+E+M synapses, comparison of Munc13-1 levels by STED microscopy in cKOR+E and cKOR+E+M neurons, and Munc13-1 expression in controlR+E+M and cKOR+E+M neurons assessed by confocal microscopy and Western blotting, see Figure 2—figure supplement 1; for microscopic assessment of synapse formation after lentiviral infection at 2 days in vitro (DIV2) instead of DIV5, see Figure 2—figure supplement 2.

We next analyzed synapse ultrastructure using high-pressure freezing followed by transmission electron microscopy in the cultured neurons (Figure 3) with established procedures (Tan et al., 2022; Wang et al., 2016). In these analyses, docked synaptic vesicles are defined as vesicles for which the electron density of the vesicular membrane touches that of the presynaptic plasma membrane, and less electron-dense space cannot be detected between the two membranes. Simultaneous deletion of RIM, ELKS, and Munc13 abolished vesicle docking (Figure 3A and B) similar to RIM + ELKS ablation (Tan et al., 2022; Wang et al., 2016). While bouton size was unchanged (Figure 3D), there was a small decrease in vesicle numbers and a mild increase in the length of the PSD in cKOR+E+M neurons (Figure 3C and E). This might be caused by homeostatic adaptations or by general roles of these proteins in synapse development, or be coincidental. Altogether, however, the morphological analyses, including STED, confocal and electron microscopy, establish that nerve terminals and synaptic appositions form and are overall ultrastructurally near-normal apart from a loss of docked vesicles despite the strong genetic manipulation with deletions of RIM1, RIM2, ELKS1, ELKS2, Munc13-1 and Munc13-2.

Synaptic ultrastructure after RIM + ELKS + Munc13 knockout.

(A–E) Sample images (A) and analyses (B–E) of synaptic morphology of high-pressure frozen neurons analyzed by electron microscopy; docked vesicles (B), total vesicles (C), bouton size (D), and postsynaptic density (PSD) length (E) per synapse profile are shown. Docked vesicles were defined as vesicles for which the electron density of the vesicular membrane touched that of the target membrane such that the two membranes were not separated by less electron-dense space, controlR+E+M 99 synapses/2 cultures, cKOR+E+M 100/2. Data are mean ± SEM; *p<0.05, ***p<0.001 as determined by Welch’s t-tests (B, C) or unpaired two-tailed Student’s t-test (D, E).

Deletion of Munc13 in addition to RIM and ELKS further impairs synaptic vesicle release

Using whole-cell recordings, we then assessed synaptic transmission in cKOR+E+M neurons and corresponding controls. We first measured spontaneous vesicle release by assessing miniature excitatory and inhibitory postsynaptic currents (mEPSCs and mIPSCs, respectively) in the presence of the sodium channel blocker tetrodotoxin. The frequencies of mEPSCs and mIPSCs were robustly decreased in cKOR+E+M neurons compared to controlR+E+M neurons, while their amplitudes remained unchanged (Figure 4A–F). In addition, there was a small increase in mEPSC rise times, similar to cKOR+E neurons (Tan et al., 2022), while mEPSC decay times and mIPSC kinetics were unchanged (Figure 4—figure supplement 1). Hence, vesicle release is impaired, but postsynaptic neurotransmitter detection is in essence intact in cKOR+E+M neurons.

Figure 4 with 2 supplements see all
Neurotransmitter release is strongly impaired after RIM + ELKS + Munc13 ablation.

(A–C) Sample traces (A) and quantification of miniature excitatory postsynaptic current (mEPSC) frequencies (B) and amplitudes (C) in controlR+E+M and cKOR+E+M neurons, 27 cells/3 cultures each. (D–F) Sample traces (D) and quantification of miniature inhibitory postsynaptic current (mIPSC) frequencies (E) and amplitudes (F), 22/3 each. (G, H) Sample traces (G) and quantification (H) of EPSCs evoked by focal electrical stimulation, 20/3 each. (I–K) Sample traces (I) and quantification of amplitudes (J) and 20–80% rise times (K) of IPSCs evoked by focal electrical stimulation, controlR+E+M 28/3, cKOR+E+M 31/3. (L) Comparison of EPSCs normalized to their own controls for cKOR+E (absolute data from Figure 1C) and cKOR+E+M (from H) neurons, cKOR+E 19/3, cKOR+E+M 20/3. (M) Comparison of IPSCs normalized to their own controls for cKOR+E (absolute data from Figure 1E) and cKOR+E+M (from J) neurons, cKOR+E 18/3, cKOR+E+M 31/3. (N) Observed ratio and distribution of T* values for NMDAR-EPSCs after 100,000 rounds of hierarchical bootstrap, T* = [mean(cKOR+E)/mean(controlR+E)]/[mean(cKOR+E+M)/mean(controlR+E+M)]; 95% confidence intervals (Cl) and the probability of the null hypothesis PH0 are indicated, n as in (L). (O) As (N), but for IPSCs, n as in (M). Data are mean ± SEM unless noted otherwise; **p<0.01, ***p<0.001 as determined by Welch’s t-tests (K, L) or Mann–Whitney tests (B, C, E, F, H, J, M). For mEPSC and mIPSC kinetics in cKOR+E+M neurons, see Figure 4—figure supplement 1. For a workflow of the hierarchical bootstrap, see Figure 4—figure supplement 2.

We then measured single action potential-evoked release at cKOR+E+M synapses (Figure 4G–K). The cKOR+E+M neurons had very strong reductions in evoked release, both at excitatory (Figure 4G and H) and inhibitory (Figure 4I–K) synapses compared to controlR+E+M neurons. To directly test whether the additional Munc13 ablation further reduced release compared to cKOR+E (Figure 1B–E), we normalized the PSC amplitudes measured in cKOR+E and cKOR+E+M neurons to their own controls, which are genetically identical within each mouse line except for the expression of Cre recombinase. We then compared the normalized data using standard statistical tests (Figure 4L and M). These analyses revealed that for both excitatory and inhibitory synapses, Munc13 knockout in addition to RIM + ELKS knockout decreased the remaining PSCs by ~40–50% compared to RIM + ELKS knockout only (Figure 4L and M). The design of the experiments imparts structure to the data: repeated measurements (sweeps) from one cell are more similar to each other than to measurements from another cell, and the cells from one culture batch might be more similar to each other than to those from other batches. To account for this structure in our experiments, we performed a hierarchical bootstrap (Saravanan et al., 2020) following the workflow in Figure 4—figure supplement 2. Through hierarchical resampling, we calculated 100,000 bootstrap replicates (T* values) of our test statistic T (the control-normalized PSC ratio of cKOR+E to cKOR+E+M), thus allowing us to estimate the sampling distribution of T, to calculate 95% confidence intervals, and to calculate the probability PH0 of the null hypothesis (T* ≤ 1) given the data. For both datasets, these distributions were robustly above 1 with correspondingly low probabilities for the null hypothesis (Figure 4N and O). Thus, using both standard statistical tests and a hierarchical bootstrap, we found strong evidence for a robust, additional decrease in action potential-evoked release after ablation of Munc13. The remaining release in cKOR+E+M neurons might be due to the low amount of exon 21/22-deficient Munc13-1 that persists after conditional Munc13-1 knockout with this allele (Banerjee et al., 2022; Figure 1—figure supplement 2), to Munc13-1 that persists beyond 11 days of Cre expression, to very low levels of Munc13-3 that escaped our detection in previous studies as Munc13-3 was not deleted in the hextuple knockout mice, or to an alternative release pathway that does not depend on RIM, ELKS, and Munc13. Altogether, however, the data establish that the remaining neurotransmitter release after RIM and ELKS knockout depends at least partially on the presence of Munc13-1 and Munc13-2.

Munc13 contributes to a remaining functional RRP after active zone disruption

Given the further reduction of synaptic transmission when Munc13 is ablated in cKOR+E neurons, we analyzed vesicle priming and release probability in cKOR+E+M neurons. The goal was to determine which release properties are controlled by Munc13 through comparison of these parameters with cKOR+E neurons. We assessed the functional RRP at both excitatory and inhibitory synapses through the application of hypertonic sucrose, a method that has been broadly used to evaluate correlations between vesicle docking and priming (Imig et al., 2014; Rosenmund and Stevens, 1996; Wang et al., 2016; Zarebidaki et al., 2020). We detected robust reductions of the vesicle pool assessed by this method for both AMPAR and GABAAR-mediated responses in cKOR+E neurons (Figure 5A–D), but the functional RRP was not fully eliminated. Deletion of Munc13 on top of RIM and ELKS revealed an additional decrease, with an almost complete loss of releasable vesicles at excitatory cKOR+E+M synapses and an >80% reduction at inhibitory cKOR+E+M synapses (Figure 5E–H). Comparison of these two genotypes by a standard statistical approach showed that cKOR+E+M neurons had a significantly smaller RRP size than cKOR+E neurons in both synapse types (Figure 5I and J). The hierarchical bootstrap analyses revealed probabilities for the null hypotheses to be true of <0.001 for excitatory synapses and of 0.084 for inhibitory synapses, respectively (Figure 5K and L). We conclude that the fusion competence of vesicles that remains after active zone disruption by RIM + ELKS knockout is mediated at least in part by Munc13. Because Munc13 is not active zone-anchored and docked vesicles are barely detectable in cKOR+E neurons, these vesicles are undocked, but likely associated with Munc13 at some distance away from the active zone.

The remaining functional RRP in RIM + ELKS-deficient synapses depends at least in part on Munc13.

(A, B) Sample traces (A) and quantification (B) of excitatory postsynaptic currents (EPSCs) triggered by hypertonic sucrose in controlR+E and cKOR+E neurons, the first 10 s of the EPSC were quantified to estimate the RRP, controlR+E 18 cells/3 cultures, cKOR+E 17/3. (C, D) As (A, B), but for inhibitory postsynaptic currents (IPSCs), 18/3 each. (E–H) As for (A–D), but for controlR+E+M and cKOR+E+M neurons, (F) 23/3 each, (H) 21/3 each. (I) Comparison of EPSCs triggered by hypertonic sucrose normalized to their own controls for cKOR+E (absolute data from B) and cKOR+E+M (from F), cKOR+E 17/3, cKOR+E+M 23/3. (J) Comparison of IPSCs triggered by hypertonic sucrose normalized to their own controls for cKOR+E (absolute data from D) and cKOR+E+M (from H), cKOR+E 18/3, cKOR+E+M 21/3. (K) Observed ratio and distribution of T* values for sucrose EPSC charge after 100,000 rounds of hierarchical bootstrap, n as in (I). (L) As (K), but for sucrose IPSC charge, n as in (J). Data are mean ± SEM unless noted otherwise; **p<0.01, ***p<0.001 as determined by Mann–Whitney tests (B, F, H, I, J) or Welch’s t-tests (D).

We finally used paired pulse ratios to monitor vesicular release probability p in controlR+E+M and cKOR+E+M neurons and repeated the measurements in cKOR+E neurons for direct comparison (Figure 6). Paired pulse ratios are inversely correlated with p (Zucker and Regehr, 2002) and strongly increased at short interstimulus intervals after knockout of RIM and ELKS at excitatory and inhibitory synapses (Figure 6A–D; Tan et al., 2022; Wang et al., 2016). In cKOR+E+M neurons, p was also strongly decreased, illustrated by robust increases in EPSC and IPSC paired pulse ratios (Figure 6E–H), to an extent that is overall very similar to cKOR+E neurons. Comparison of genotypes through standard statistical methods or hierarchical bootstrap supported the conclusion that Munc13 knockout in addition to RIM and ELKS did not further decrease p, as effects in cKOR+E synapses and in cKOR+E+M synapses were indistinguishable (Figure 6I–L). Consistent with this comparison, spontaneous mEPSC and mIPSC frequencies (Figure 6—figure supplement 1A–H) and depression of IPSCs during stimulus trains (Figure 6—figure supplement 1I–L) were similar in the two genotypes. Hence, while the remaining Munc13 at cKOR+E synapses is sufficient to maintain a small functional RRP, it does not enhance vesicular release probability of these vesicles.

Figure 6 with 1 supplement see all
Vesicular release probability is not further impaired by combined RIM + ELKS + Munc13 knockout.

(A, B) Sample traces (A) and quantification (B) of excitatory postsynaptic current (EPSC) paired pulse ratios in controlR+E and cKOR+E neurons, controlR+E 15 cells/3 cultures, cKOR+E, 16/3. (C, D) As (A, B), but for inhibitory postsynaptic currents (IPSCs) (sample traces of 2500 ms intervals are not shown in C for simplicity), 17/3 each. (E–H) As for (A–D), but for controlR+E+M and cKOR+E+M neurons, (F) 19/3 each, (H) 19/3 each. (I) Comparison of EPSC paired pulse ratios across interstimulus intervals normalized to their own controls for cKOR+E (absolute data from B) and cKOR+E+M (from F), cKOR+E 16/3, cKOR+E+M 19/3. (J) Comparison of IPSC paired pulse ratios across interstimulus intervals normalized to their own controls for cKOR+E (absolute data from D) and cKOR+E+M (from H), cKOR+E 17/3, cKOR+E+M 19/3. (K) Observed ratio and distribution of T* values for EPSC paired pulse ratios at 50 ms interstimulus intervals after 100,000 rounds of hierarchical bootstrap, n as in (I). (L) As (K), but for IPSC paired pulse ratios at 20 ms interstimulus intervals, n as in (J). Data are mean ± SEM unless noted otherwise; ***p<0.001 as determined by two-way ANOVA followed by Bonferroni’s multiple-comparisons post-hoc tests (B, D, F, H) or by Mann–Whitney tests (I, J). For miniature EPSCs (mEPSCs) and miniature IPSCs (mIPSCs) in cKOR+E neurons, and for IPSCs evoked by stimulus trains in cKOR+E and cKOR+E+M neurons, see Figure 6—figure supplement 1.

Discussion

We found previously that the functional RRP is not fully disrupted after ablating vesicle docking by simultaneous knockout of RIM and ELKS (Wang et al., 2016). Here, we show that fusion of these remaining RRP vesicles depends at least in part on Munc13. Even though Munc13-1 is not active zone-anchored after RIM + ELKS ablation, knocking out Munc13 in addition decreased the remaining pool of releasable vesicles at excitatory and inhibitory hippocampal synapses. We conclude that Munc13 can render some vesicles fusogenic in the absence of RIM and ELKS. Our work adds a very strong compound knockout mutation to disrupt synaptic function, with removal of six important active zone proteins, to a growing body of literature that is based on the compound mutant approach. Beyond its relevance for mechanisms of neurotransmitter release, our work further supports that synapse formation is remarkably resilient to even massive perturbations of presynaptic function and protein composition.

Disrupting active zones by removing redundancy

Knockout studies on active zone gene families have defined multiple roles for these proteins in the neurotransmitter release process. While some functions and mechanisms, for example, vesicle priming, strongly depend on single proteins (Aravamudan et al., 1999; Augustin et al., 1999; Richmond et al., 1999; Varoqueaux et al., 2002), other functions are redundant between members within a protein family or across protein families, and hence much more difficult to study mechanistically. This is particularly true for the scaffolding mechanisms that hold the active zone together and connect it with the plasma membrane and the vesicle cluster. These mechanisms are not well defined, and models ranging from self-assembly of complexes with defined stoichiometries to phase separation via multivalent low-affinity interactions have been proposed (Chen et al., 2020; Emperador-Melero and Kaeser, 2020; Südhof, 2012). Recent studies with compound mutants to remove combinations of active zone protein families started to identify the required active zone scaffolds (Acuna et al., 2016; Kushibiki et al., 2019; Oh et al., 2021; Wang et al., 2016). We found previously that simultaneous deletion of RIM and ELKS in hippocampal neurons strongly disrupts active zone protein assemblies with subsequent loss of Munc13, RIM-BP, Piccolo, and Bassoon (Tan et al., 2022; Wang et al., 2016; Wong et al., 2018). This loss of active zone material causes a near-complete disruption of vesicle docking and strongly impairs action potential-triggered neurotransmitter release. Unexpectedly, however, some release persists either due to the presence of remaining release-competent but non-docked vesicles or due to the rapid generation of release-competent vesicles in response to stimulation. The present study establishes that the transmitter release that remains after RIM and ELKS deletion relies at least in part on Munc13 as ablating Munc13 on top of RIM and ELKS decreased the remaining functional RRP further. The remaining release in cKOR+E+M neurons might be Munc13-independent, which is currently not possible to establish. Perhaps more likely, it is mediated by the small amount of remaining Munc13-1 protein in the neurons derived from the conditional mutant mice used here (Banerjee et al., 2022), which is not seen with the constitutive Munc13-1 knockout mouse line (Augustin et al., 1999; Man et al., 2015). We further found that synapse structure per se is resilient to this major genetic and functional perturbation as synapses formed at normal densities and showed overall near-normal ultrastructure despite the knocking out of six important active zone genes in the cultured neurons. Our work adds to a growing body of data demonstrating that neurotransmitter release, presynaptic Ca2+ entry, and active zone scaffolding are all dispensable for the formation of prominent types of central nervous system synapses (Held et al., 2020; Sando et al., 2017; Sigler et al., 2017; Verhage et al., 2000).

One way to interpret our data is to compare them with properties of fusion in other secretory pathways. Synaptic vesicle exocytosis after RIM and ELKS knockout has resemblance with secretion from chromaffin cells (Neher, 2018; Wang et al., 2016). In these cells, the release-ready pool of vesicles depends on Munc13, has relatively slower kinetics compared to synapses, but appears to not rely on the sequences of Munc13 that interact with RIM or on active zone scaffolds more generally (Betz et al., 2001; Man et al., 2015; Neher, 2018; Stevens et al., 2005). The similarities of the remaining fusion after disrupting active zone assembly by knockout of RIM and ELKS are striking in that the release kinetics appear slowed down (Wang et al., 2016), the remaining fusion depends at least in part on Munc13, and docking and the functional RRP are not fully correlated. Some of these similarities are also reminiscent of recent studies in C. elegans, where vesicle priming does not rely on the interaction of unc-13/Munc13 with unc-10/RIM, and possibly unc-10/RIM itself (Liu et al., 2019). Altogether, these studies and our work indicate that disrupting the active zone through RIM + ELKS knockout renders synaptic vesicle exocytosis similar to chromaffin cell secretion. Notably, there are specific vertebrate synapses with similar properties. For example, at hippocampal mossy fiber synapses, strong depolarizations lead to RRP estimates that are larger than the number of docked vesicles (Maus et al., 2020), supporting the model that RRP vesicles can be rapidly generated and released during such stimuli, perhaps similar to the remaining RRP at cKOR+E synapses.

Can Munc13 prime vesicles in the absence of RIM?

Our data reveal that the transmitter release that remains after active zone disruption upon RIM and ELKS deletion depends at least in part on Munc13. Thus, Munc13 might render some vesicles fusion-competent in the absence of RIM and when Munc13 is not anchored at the active zone. This mechanism, however, is inefficient as it only maintains a fraction of the functional RRP of a wildtype synapse. An alternative or complementary model is that fusion competence is rapidly generated and immediately followed by exocytosis during pool-depleting stimuli. We recently established that the functional RRP after RIM + ELKS knockout can be further boosted by re-expressing RIM zinc finger domains without enhancing vesicle docking (Tan et al., 2022). When the RIM zinc finger domain was expressed in RIM + ELKS knockout neurons, it co-localized with the vesicle cluster, it was not concentrated at the active zone, and it recruited Munc13 to the vesicle cluster. Hence, Munc13 can enhance vesicle fusogenicity through association with non-docked vesicles, at least when this association is generated artificially. Here, we show that the remaining fusion in RIM + ELKS knockouts depends at least in part on Munc13 (Figures 4 and 5), but that Munc13-1 is barely detectable at active zones, while some Munc13 is present at synapses. These data indicate that endogenous Munc13 can be near vesicles and enhance their fusogenicity even if Munc13 is not active zone-anchored. Hence, Munc13 on non-docked vesicles might mediate the remaining fusion through generation of a pool of vesicles that can be rapidly primed and released upon stimulation. Altogether, a model arises that the rate-limiting step for generation of a functional RRP is the presynaptic recruitment of Munc13, even if Munc13 is not fully anchored and activated at the active zone. Upon stimulation, roles of Munc13 in SNARE complex assembly and fusion can be executed quickly.

Previous work established that RIM recruits Munc13 to active zones and activates it. This mechanism operates through binding of the RIM zinc finger to Munc13 C2A domains, which is necessary for rendering Munc13 monomeric and active in fusion (Andrews-Zwilling et al., 2006; Betz et al., 2001; Brockmann et al., 2020; Camacho et al., 2017; Deng et al., 2011; Dulubova et al., 2005; Lu et al., 2006). In the experiments presented here, the neurotransmitter release remaining in the absence of RIM and ELKS depends at least partially on Munc13. This indicates that not all priming requires the RIM-mediated activation mechanism of Munc13, consistent with previous observations that many but not all RRP vesicles are lost after RIM knockout (Han et al., 2011; Han et al., 2015; Kaeser et al., 2011; Kaeser et al., 2012). An alternative mechanism could operate via bMunc13-2 (which lacks the C2A-domain that binds to RIM) and ELKS (Kawabe et al., 2017). While ELKS provides a Munc13-recruitment and priming mechanism that is independent of RIM, most ELKS is also removed in the RIM + ELKS knockout neurons (Kaeser et al., 2009; Liu et al., 2014; Wang et al., 2016), and only a very small subset of synapses relies on bMunc13-2 (Kawabe et al., 2017). Hence, this mechanism might be insufficient to explain the remaining release from RIM + ELKS knockout neurons. An alternative and perhaps more likely explanation is that not all C2A-domain-containing Munc13 requires RIM. Monomeric, active Munc13 is in equilibrium with dimeric, inactive Munc13, and RIM shifts the equilibrium to the active form. Even in the absence of RIM, some Munc13 will be monomeric and available for assembling SNARE complexes, accounting for the vesicular exocytosis that remains in RIM + ELKS knockout neurons. Finally, this mechanism may not be restricted to docked vesicles, but vesicles associated with Munc13 may be amenable to release, explaining why some release persists in RIM + ELKS mutants despite the loss of active zone-anchored Munc13 and of a strong reduction in docked vesicles.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Genetic reagent (Mus musculus)Rims1tm3Sud/J
(RIM1αβfl/fl)
Kaeser et al., 2008RRID:IMSR_JAX:015832
Genetic reagent (M. musculus)Rims2tm1.1Sud/J
(RIM2αβγfl/fl)
Kaeser et al., 2011RRID:IMSR_JAX:015833
Genetic reagent (M. musculus)Erc1tm2.1Sud/J
(ELKS1αfl/fl)
Liu et al., 2014RRID:IMSR_JAX:015830
Genetic reagent (M. musculus)Erc2tm1.2Sud/J
(ELKS2αfl/fl)
Kaeser et al., 2009RRID:IMSR_JAX:015831
Genetic reagent (M. musculus)Unc13atm1.1Bros
(Munc13-1fl/fl)
Banerjee et al., 2022MGI:7276178
Genetic reagent (M. musculus)Unc13btm1Rmnd
(Munc13-2-/-)
Varoqueaux et al., 2002RRID:MGI:2449706
Genetic reagent (M. musculus)Unc13ctm1Bros
(Munc13-3-/-)
Augustin et al., 2001RRID:MGI:2449467
Cell line (Homo sapiens)HEK293T cellsATCCCat# CRL-3216; RRID:CVCL_0063
Recombinant DNA reagentpFSW EGFP CreLiu et al., 2014pHN131014; lab plasmid code (LPC): p009
Recombinant DNA reagentpFSW EGFP ΔCreLiu et al., 2014pHN131015; LPC: p010
AntibodyAnti-RIM (rabbit polyclonal)SySyCat# 140003;
RRID:AB_887774; lab antibody code (LAC): A58
Immunofluorescence (IF) (1:500)
AntibodyAnti-PSD-95 (mouse monoclonal)NeuroMabCat# 73-028;
RRID:AB_10698024; LAC: A149
IF (1:500)
AntibodyAnti- Synaptophysin (guinea pig polyclonal)SySyCat# 101004;
RRID:AB_1210382; LAC: A106
IF (1:500)
AntibodyAnti-Munc13-1 (rabbit polyclonal)SySyCat# 126103;
RRID:AB_887733; LAC: A72
IF (1:500);
Western blot (WB) (1:1000)
AntibodyAnti-Synapsin-1 (mouse monoclonal)SySyCat# 106001;
RRID:AB_2617071; LAC: A57
WB (1:4000)
AntibodyAnti-rabbit Alexa Fluor 488 (goat polyclonal)Thermo FisherCat# A-11034; RRID:AB_2576217; LAC: S5IF (1:200)
AntibodyAnti-mouse IgG2a Alexa Fluor 555 (goat polyclonal)Thermo FisherCat# A-21137; RRID:AB_2535776; LAC: S20IF (1:200)
AntibodyAnti-guinea pig Alexa Fluor 633 (goat polyclonal)Thermo FisherCat# A-21105; RRID:AB_2535757; LAC: S34IF (1:500)
AntibodyAnti-mouse peroxidase-conjugated (goat polyclonal)MP BiologicalsCat# 0855550;
RRID:AB_2334540;
LAC: S52
WB (1:10000)
AntibodyAnti-rabbit peroxidase-conjugated (goat polyclonal)MP BiologicalsCat# 0855676;
RRID:AB_2334589;
LAC: S53
WB (1:10000)
Software, algorithmMATLABMathWorksRRID:SCR_001622; https://www.mathworks.com/products/matlab.html
Software, algorithmPrismGraphPadRRID:SCR_002798; https://www.graphpad.com/scientific-software/prism
Software, algorithmFiji/ImageJNIHRRID:SCR_002285; https://imagej.net/software/fiji/downloads
Software, algorithmpClampMolecular DevicesRRID:SCR_011323; https://www.moleculardevices.com/products/software/pclamp.html
Software, algorithmMATLAB code for object recognition and analysis of 2D imagesThis paperhttps://doi.org/10.5281/zenodo.6388196This code was adapted for 2D images from a previously generated code (Liu et al., 2018; Liu et al., 2022) and is freely accessible at zenodo.org
Software, algorithmMATLAB code for hierarchical bootstrapThis paperhttps://github.com/kaeserlab/Hierarchical_Bootstrap_Analysis_RB; Born, 2022This code was generated for this paper and is freely accessible at github.com

Mouse lines

Request a detailed protocol

The quadruple homozygote floxed mice for Rims1 (to remove RIM1α and RIM1β, RRID:IMSR_JAX:015832, Kaeser et al., 2008), Rims2 (to remove RIM2α, RIM2β, and RIM2γ, RRID:IMSR_JAX:015833, Kaeser et al., 2011), Erc1 (to remove ELKS1α, RRID:IMSR_JAX:015830, Liu et al., 2014), and Erc2 (to remove ELKS2α, RRID:IMSR_JAX:015831, Kaeser et al., 2009) were previously described (Wang et al., 2016). Exon 6 (E6) or 26 (E26) were flanked by loxP sites in the Rims1 or Rims2 floxed mice, respectively. Exons 2 (E2) and 3 (E3) were flanked by loxP sites in the Erc1 floxed mice, and exon 3 (E3) was flanked by loxP sites in the Erc2 floxed mice. Floxed Unc13a mice (Munc13-1, Exon 21 [E21] flanked by loxP sites, Unc13atm1.1Bros, MGI:7276178, Banerjee et al., 2022) were crossed to constitutive knockout mice for Unc13b (Munc13-2, Unc13btm1Rmnd, RRID_MGI:2449706, Varoqueaux et al., 2002) and Unc13c (Munc13-3, Unc13ctm1Bros, RRID_MGI:2449467, Augustin et al., 2001) to produce Munc13 triple homozygote mice. Mice for simultaneous ablation of RIM1, RIM2, ELKS1, ELKS2, Munc13-1, and Munc13-2 were generated by crossing the corresponding conditional (RIM1αβ, RIM2αβγ, ELKS1α, ELKS2α, and Munc13-1) and constitutive (Munc13-2) knockout alleles to homozygosity. All animal experiments were approved by the Harvard University Animal Care and Use Committee (protocol number IS00000049).

Cell lines, primary neuronal culture, and lentiviral infection

Request a detailed protocol

Primary mouse hippocampal cultures were generated from newborn pups within 24 hr after birth as described (Held et al., 2020; Tan et al., 2022; Wang et al., 2016); cells from mice of both sexes were mixed. Mice were anesthetized by hypothermia and the hippocampus was dissected out. Cells were dissociated and plated onto glass coverslips in tissue culture medium composed of Minimum Essential Medium (MEM) with 10% Fetal Select bovine serum (Atlas Biologicals FS-0500-AD), 2 mM l-glutamine, and 25 µg/mL insulin, 0.1 mg/mL transferrin, 0.5% glucose, and 0.02% NaHCO3. Cultures were maintained in a 37°C tissue culture incubator, and after ~24 hr the plating medium was exchanged with growth medium composed of MEM with 5% Fetal Select bovine serum, 2% B-27 supplement (Thermo Fisher 17504044), 0.5 mM l-glutamine, 0.1 mg/mL transferrin, 0.5% glucose, and 0.02% NaHCO3. At DIV3, depending on growth, 50 or 75% of the medium were exchanged with growth medium supplemented with 4 µM cytosine β-d-arabinofuranoside (AraC). Cultured neurons were transduced with lentiviruses produced in HEK293T cells (CRL-3216, RRID:CVCL_0063, immortalized human cell line of female origin, purchased mycoplasma-free) by Ca2+ phosphate transfection; viral transduction was at DIV5 unless noted otherwise. These lentiviruses expressed EGFP-tagged Cre recombinase (to generate cKO neurons) or a truncated, enzymatically inactive EGFP-tagged Cre protein (ΔCre, to generate control neurons). Expression in lentiviral constructs was driven by the human Synapsin-1 promoter to restrict expression to neurons (Liu et al., 2014; Wang et al., 2016). Analyses were performed at DIV16–19.

Electrophysiology

Request a detailed protocol

Electrophysiological recordings in cultured hippocampal neurons were performed as described (Held et al., 2020; Tan et al., 2022; Wang et al., 2016) at DIV16–19. The extracellular solution contained (in mM) 140 NaCl, 5 KCl, 2 MgCl2, 1.5 CaCl2, 10 glucose, 10 HEPES-NaOH (pH 7.4, ~300 mOsm). To avoid network activity induced by AMPAR activation, NMDAR-mediated excitatory postsynaptic currents (NMDAR-EPSCs) were measured to assess action potential-triggered excitatory transmission. For NMDAR-EPSCs, 6-Cyano-7-nitroquinoxaline-2,3-dione (CNQX, 20 µM) and picrotoxin (PTX, 50 µM) were present in the extracellular solution. Electrically evoked inhibitory postsynaptic currents (IPSCs) were recorded in the presence of d-amino-5-phosphonopentanoic acid (D-APV, 50 µM) and CNQX (20 µM) in the extracellular solution. Recordings were performed at room temperature (20–24°C). Action potentials were elicited with a bipolar focal stimulation electrode fabricated from nichrome wire. Paired pulse ratios were calculated as the amplitude of the second PSC divided by the amplitude of the first at each interval from the average of three to four sweeps per cell and interval. The baseline value for the second PSC was taken immediately after the second stimulus artifact. For analysis of action potential trains (50 stimuli at 10 Hz), the baseline value of each IPSC within the train was taken immediately after the corresponding stimulus artifact. For AMPAR-mediated transmission (mEPSC and sucrose-induced EPSCs), TTX (1 µM), PTX (50 µM), and D-APV (50 µM) were added to the extracellular solution. For mIPSC and sucrose-induced IPSC recordings, TTX (1 µM), CNQX (20 µM), and D-APV (50 µM) were added to the extracellular solution. The RRP was estimated by application of 500 mM sucrose in extracellular solution applied via a microinjector syringe pump for 10 s at a rate of 10 µL/min through a tip with an inner diameter of 250 µm. mEPSCs and mIPSCs were identified with a template search followed by manual confirmation by an experimenter, and their frequencies and amplitudes were assessed during a 100 s recording time window after reaching a stable baseline (>3 min after break-in). Glass pipettes were pulled at 2–5 MΩ and filled with intracellular solutions containing (in mM) for EPSC recordings: 120 Cs-methanesulfonate, 2 MgCl2, 10 EGTA, 4 Na2-ATP, 1 Na-GTP, 4 QX314-Cl, 10 HEPES-CsOH (pH 7.4, ~300 mOsm); and for IPSC recordings: 40 CsCl, 90 K-gluconate, 1.8 NaCl, 1.7 MgCl2, 3.5 KCl, 0.05 EGTA, 2 Mg-ATP, 0.4 Na2-GTP, 10 phosphocreatine, 4 QX314-Cl, 10 HEPES-CsOH (pH 7.2, ~300 mOsm). Cells were held at +40 mV for NMDAR-EPSC recordings and at –70 mV for evoked IPSC, mEPSC, mIPSC, sucrose EPSC, and sucrose IPSC recordings. Access resistance was monitored and compensated to 3–5 MΩ, and cells were discarded if the uncompensated access exceeded 15 MΩ. Data were acquired at 5 kHz and lowpass filtered at 2 kHz with an Axon 700B Multiclamp amplifier and digitized with a Digidata 1440A digitizer. Data acquisition and analyses were done using pClamp10. For electrophysiological experiments, the experimenter was blind to the genotype throughout data acquisition and analyses.

STED and confocal imaging

Request a detailed protocol

Light microscopic analyses were in essence performed as previously described (Emperador-Melero et al., 2021b; Emperador-Melero et al., 2021a; Held et al., 2020; Tan et al., 2022; Wong et al., 2018). Neurons were cultured on 0.17-mm-thick 12-mm-diameter coverslips. At DIV16–18, cultured neurons were washed two times with warm PBS and fixed in 4% PFA in PBS for 10 min. After fixation, coverslips were rinsed twice in PBS, then permeabilized in PBS + 0.1% Triton X-100 + 3% BSA (TBP) for 1 hr. Primary antibodies were diluted in TBP and stained for 24–48 hr at 4°C. The following primary antibodies were used: rabbit anti-RIM1 (1:500, RRID:AB_887774, A58), rabbit anti-Munc13-1 (1:500, RRID:AB_887733, A72), guinea pig anti-Synaptophysin (1:500, RRID:AB_1210382, A106), mouse anti-PSD-95 (1:500, RRID:AB_10698024, A149). After primary antibody staining, coverslips were rinsed twice and washed 3–4 times for 5 min in TBP. Alexa Fluor 488 (anti-rabbit, RRID:AB_2576217, S5), 555 (anti-mouse IgG2a, RRID:AB_1500824, S20), and 633 (anti-guinea pig, RRID:AB_2535757, S34) conjugated antibodies were used as secondary antibodies at 1:200 (Alexa Fluor 488 and 555) or 1:500 (Alexa Fluor 633) dilution in TBP, incubated for 24–48 hr at 4°C followed by rinsing two times and washing 3–4 times 5 min in TBP. Stained coverslips were post-fixed for 10 min with 4% PFA in PBS, rinsed two times in PBS + 50 mM glycine, and once in deionized water, air-dried, and mounted on glass slides. STED images were acquired with a Leica SP8 Confocal/STED 3X microscope with an oil immersion 1.44 numerical aperture 100x objective and gated detectors as described before (Emperador-Melero et al., 2021a; Held et al., 2020; Tan et al., 2022; Wong et al., 2018). Images of 46.51 × 46.51 µm2 areas were scanned at a pixel density of 4096 × 4096 (11.4 nm/pixel). Alexa Fluor 633, Alexa Fluor 555, and Alexa Fluor 488 were excited with 633 nm, 555 nm, and 488 nm using a white light laser at 2–5% of 1.5 mW laser power. The Alexa Fluor 633 channel was acquired first in confocal mode using 2× frame averaging. Subsequently, Alexa Fluor 555 and Alexa Fluor 488 channels were acquired in both confocal and STED modes. Alexa Fluor 555 and 488 channels in STED mode were depleted with 660 nm (50% of max power, 30% axial depletion) and 592 nm (80% of max power, 30% axial depletion) depletion lasers, respectively. Line accumulation (2–10×) and frame averaging (2×) were applied during STED scanning. Identical settings were applied to all samples within an experiment. Synapses within STED images were selected in side-view, defined as synapses that contained a synaptic vesicle cluster labeled with Synaptophysin and associated with an elongated PSD-95 structure along the edge of the vesicle cluster as described (Emperador-Melero et al., 2021b; Emperador-Melero et al., 2021a; Held et al., 2020; de Jong et al., 2018; Nyitrai et al., 2020; Tan et al., 2022; Wong et al., 2018). For intensity profile analyses, side-view synapses were selected using only the PSD-95 signal and the vesicle signal for all experiments by an experimenter blind to the protein of interest. A region of interest (ROI) was manually drawn around the PSD-95 signal and fit with an ellipse to determine the center position and orientation. A ~1200-nm-long, 200-nm-wide rectangle was then positioned perpendicular to and across the center of the elongated PSD-95 structure. Intensity profiles from –400 nm (presynaptic) to +200 nm (postsynaptic) relative to the center of the PSD-95 signal were obtained for all three channels within this ROI. To align individual profiles, the PSD-95 signal only was smoothened using a moving average of 5 pixels, and the smoothened signal was used to define the peak position of PSD-95. All three channels (vesicle marker, test protein, and smoothened PSD-95) were then aligned to the PSD-95 peak position and averaged across images for line profile plots. Peak values for each line profile were determined independent of peak position and used to generate the plots of peak levels. For Figure 1—figure supplement 3A, B, D, and E and Figure 2—figure supplement 1C and D, Munc13-1 levels were analyzed using Synaptophysin to define ROIs in the confocal images with ImageJ. For Figure 2L–O and Figure 2—figure supplement 2A–E, ROI selection was performed using an adapted custom-written code to perform automatic two-dimensional segmentation (Emperador-Melero et al., 2021a; Held et al., 2020; Liu et al., 2018; Liu et al., 2022); the code was deposited to Zenodo at https://doi.org/10.5281/zenodo.6388196. After Synaptophysin object detection, the density, intensity, and area of these objects were quantified (Figure 2L–O, Figure 2—figure supplement 2A–D). In Figure 2—figure supplement 2E, the Synaptophysin objects (confocal) that exceeded the overlap threshold of 0% with PSD-95 objects (STED) were included in the quantification. Analyses were performed on raw images without background subtraction, and adjustments were done identically across experimental conditions. Representative images were brightness- and contrast-adjusted to facilitate inspection, and these adjustments were made identically for images within an experiment. The experimenter was blind to the condition/genotype for image acquisition and analyses.

High-pressure freezing and electron microscopy

Request a detailed protocol

Neurons were cultured on sapphire coverslips (6 mm diameter) coated with matrigel. At DIV16–18, cultured neurons were frozen using a Leica EM ICE high-pressure freezer in extracellular solution containing (in mM) 140 NaCl, 5 KCl, 2 CaCl2, 2 MgCl2, 10 HEPES-NaOH (pH 7.4), 10 glucose (~300 mOsm), with PTX (50 µM), CNQX (20 µM), and D-AP5 (50 µM) added. After freezing, samples were first freeze-substituted (AFS2, Leica) in anhydrous acetone containing 1% glutaraldehyde, 1% osmium tetroxide, and 1% water. The process of freeze substitution was as follows: –90°C for 5 hr, 5°C per hr to –20°C, –20°C for 12 hr, and 10°C per hr to 20°C. Following freeze substitution, samples were Epon-infiltrated and baked for 48 hr at 60°C followed by 80°C overnight before sectioning at 50 nm. For ultrathin sectioning, the sapphire coverslip was removed from the resin block by plunging the sample first in liquid nitrogen and followed by warm water several times until the sapphire was completely detached. The resin block containing the neurons was then divided into four pieces, and one piece was mounted for sectioning. Ultrathin sectioning was performed on a Leica EM UC7 ultramicrotome, and the 50 nm sections were collected on a nickel slot grid (2 × 1 mm) with a carbon-coated formvar support film. The samples were counterstained by incubating the grids with 2% lead acetate solution for 10 s, followed by rinsing with distilled water. Images were taken with a transmission electron microscope (JEOL 1200 EX at 80 kV accelerating voltage) and processed with ImageJ. The total number of vesicles, the number of docked vesicles per synapse profile, the area of the presynaptic bouton, and the length of the PSD were analyzed in each section. Docked vesicles were defined as vesicles for which the electron densities of the vesicular membrane and the presynaptic plasma membrane merged such that the two membranes were not separated by less electron-dense space. Bouton size was calculated from the measured perimeter of each synapse. Experiments and analyses were performed by an experimenter blind to the genotype.

Western blotting

Request a detailed protocol

At DIV15–19, cultured neurons were harvested in 20 µL 1× SDS buffer per coverslip and run on standard SDS-PAGE gels followed by transfer to nitrocellulose membranes. Membranes were blocked in filtered 10% nonfat milk/5% goat serum for 1 hr at room temperature and incubated with primary antibodies (rabbit anti-Munc13-1, 1:1000, RRID:AB_887733, A72; mouse anti-Synapsin-1, 1:4000, RRID:AB_2617071, A57) in 5% nonfat milk/2.5% goat serum overnight at 4°C, and HRP-conjugated secondary antibodies (1:10,000, anti-mouse, RRID:AB_2334540; anti-rabbit, RRID:AB_2334589) were used. Western blotting was repeated 3–8 times per genotype from selected cultures used for electrophysiology, immunostaining, and electron microscopy. For illustration in figures, images were adjusted for brightness and contrast to facilitate visual inspection, and the same adjustments were used for the entire scan.

Statistics

Standard statistical tests were performed using GraphPad Prism 9; hierarchical bootstrap was performed using MATLAB. Data are displayed as mean ± SEM unless noted otherwise, and significance of standard tests is presented as *p<0.05, **p<0.01, and ***p<0.001. Sample sizes were determined based on previous studies, and no statistical methods were used to predetermine sample size. No outliers were excluded. Parametric tests were used for normally distributed data (assessed by Shapiro–Wilk tests) or when sample size was n ≥ 30. Unpaired two-tailed Student’s t-tests were used for datasets with equal variance, or Welch’s unequal variances t-tests for datasets with unequal variance. For non-normally distributed data, Mann–Whitney tests or Kruskal–Wallis tests followed by Dunn’s multiple-comparisons post-hoc tests were used. For paired pulse ratios, two-way ANOVA tests with Bonferroni’s post-hoc tests were used. For STED side-view analyses, two-way ANOVA tests with Bonferroni’s post-hoc tests were used on a 200-nm-window centered around the active zone peak. The hierarchical bootstrap analyses (Saravanan et al., 2020) were performed using a custom-written MATLAB code (https://github.com/kaeserlab/Hierarchical_Bootstrap_Analysis_RB; Born, 2022). This method is similar to other hierarchical statistical approaches, such as repeated-measures ANOVA and linear mixed-effects models; however, it has the added advantage of not making any distributional assumptions about the underlying data and allows for unequal variance among the groups. To test whether knockout of Munc13 caused additional decrease for electrically or sucrose-evoked PSCs, or an increase in paired pulse ratios, the test statistic, T, was defined as:

T=[mean(cKOR+E)/mean(controlR+E)]/[mean(cKOR+E+M)/mean(controlR+E+M)]

The null value for T is 1, and the alternative hypothesis is T > 1 for PSCs and T < 1 for paired pulse ratios. The sampling distribution of T was estimated by resampling with replacement from the raw data while preserving the hierarchical relationships created by the design of the experiment. For each bootstrap iteration, group identity (controlR+E, cKOR+E, controlR+E+M, or cKOR+E+M) was preserved. Resampling was done at three nested levels: batch of culture, cell, and sweep (except for sucrose-evoked release for which only one sweep was recorded). After each bootstrap iteration for the four groups, a bootstrap replicate of T called T* was calculated using the above formula. The procedure was repeated 100,000 times, producing an estimate of the sampling distribution of T, which was plotted in the frequency histograms. Based on the distribution of the 100,000 T* values, we calculated 95% confidence intervals using the percentile method (Efron and Tibshirani, 1994). In addition, we calculated the probability of the null hypothesis (PH0) given our data for action potential- and sucrose-evoked PSCs as

PH0=#{T1}/100,000,

and for paired pulse ratios as:

PH0=#{T1}/100,000.

Note that this is not the traditional p-value calculated with standard statistical tests, which is the probability of obtaining a result as extreme or more extreme assuming the null hypothesis to be true. Rather, the metric computed above is a more direct measure of a given hypothesis being true – in this case, the null hypothesis – given the measured data points (Saravanan et al., 2020). For all datasets, sample sizes and the specific tests used are stated in the figure legends.

Materials, data, and code availability

Request a detailed protocol

Plasmids used for this study will be shared upon request. Mouse lines will be shared upon request within the limits of the respective material transfer agreements. Analyses codes have been deposited to Zenodo and GitHub and are publicly available as listed in the Key Resources Table. All data generated or analyzed in this study, including individual data points, are included in the figures. Source data files for Figure 1—figure supplement 3, Figure 2—figure supplement 1, and Figure 2—figure supplement 2 are provided, and a source data table that contains all means, errors, statistical tests, and p-values is also included. Plasmids and mice should be requested from the corresponding author (kaeser@hms.harvard.edu).

Data availability

All data generated or analyzed in this study, including individual data points, are included in the figures. Source data files for Figure 1—figure supplement 3, Figure 2—figure supplement 1 and Figure 2—figure supplement 2 are provided, and a source data table that contains all means, errors, statistical tests and p-values is also included.

References

    1. Saravanan V
    2. Berman GJ
    3. Sober SJ
    (2020)
    Application of the hierarchical bootstrap to multi-level data in neuroscience
    Neurons, Behavior, Data Analysis and Theory 3:https://nbdt.scholasticahq.com/article/13927-application-of-the-hierarchical-bootstrap-to-multi-level-data-in-neuroscience.

Decision letter

  1. Axel T Brunger
    Reviewing Editor; Stanford University School of Medicine, Howard Hughes Medical Institute, United States
  2. Lu Chen
    Senior Editor; Stanford University, United States

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Munc13 supports vesicle fusogenicity after disrupting active zone scaffolds and synaptic vesicle docking" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Lu Chen as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions

1. It is unclear why Cre mediated removal of the presynaptic scaffold proteins was initiated at DIV 5, at a time point when synapse formation is already well underway. This likely contributed to the hypomorphic phenotype. Please comment.

2. The authors conclude that postsynaptic response is intact. On the other hand, they find kinetic changes in EPSC and IPSC (Figure 1 and Figure 4) that deserve to be discussed. In addition, they do not show any kinetic analysis of mEPSC and mIPSCs; this type of analysis is straightforward to be obtained and it might be useful to assess in part if there are or not changes at the postsynaptic level.

3. The major conclusion of the study is that the remaining release after ablation of RIM+ELKS (R+E) is mainly Munc13 dependent. First, the relative changes in mean EPSC/IPSC amplitude or sucrose-dependent RRP size between R+E and RIM+ELKS+Munc13 (R+E+M) KO cultures are comparably small, at most around 15% (Figures 4, 5). This is expected given the strong reduction in Munc13-1 levels in R+E KO (Figure 1; Wang et al., 2016). The relative differences between the experimental groups are smaller than the differences between different control data sets. Moreover, previous data sets obtained from R+E KO cultures under similar conditions by the same group suggest a similar relative decrease in E/IPSC amplitude or RRP size in R+E KO (Wang et al., 2016; Tan et al., 2022) compared to R+E+M KOs (both by ~90%). Given the parameter distributions (e.g., control EPSC amplitudes range between 0.5 and >1.5 nA and display clusters that likely reflect different cultures, Figure 4H), how confident can the authors robustly resolve average relative changes by ~10%? To evaluate whether the remaining release in R+E KOs is indeed Munc13-dependent, it would be helpful to report effect sizes and to provide a post-hoc power calculation. Irrespectively of the outcome of a power analysis, it may be advisable to increase sample sizes for some of the major experiments.

4. Although EPSC/IPSC amplitude, RRP size, and mEPSC/mIPSC frequency are strongly reduced in R+E+M hextuple KO cultures, these cultures still display significant evoked and spontaneous synaptic transmission. For instance, ~20% of the IPSC RRP (Figure 5H) or ~40% of mEPSCs (Figure 4B) remain in R+E+M KO neurons. The authors attribute this to incomplete loss of Munc13 or Munc13-independent release (Discussion). In line with the first hypothesis and previous reports, their confocal data of Munc13-1,-2,-3 KO cultures indicate a significant anti-Munc13-1 signal (~25% of control; Figure 1 —figure supplement 2B). Thus, it remains unclear whether the remaining release in R+E KO neurons is Munc13 dependent. An alternative hypothesis is that a significant fraction, if not the majority, of the remaining release is Munc13 independent. This would be equally, if not more interesting. Thus, unless the authors directly demonstrate that the release remaining in R+E KO neurons indeed requires Munc13, all respective statements in the manuscript should be revised accordingly (e.g., "Our data reveal that the transmitter release that remains after active zone disruption upon RIM and ELKS deletion depends on Munc13., l. 324)". Indeed, an alternative interpretation of the current data could be quite similar to that of Wang et al. 2016: Fusion competent vesicles persist upon ablation of RIM-1/2, ELKS-1/2, and Munc13-1/2.

5. The rise time of NMDAR EPSCs was previously shown to be strongly attenuated in R+E KO cultures (Wang et al., 2016). How does the EPSC rise time in R+E+M KO cultures relate to the one in R+E KO neurons?

6. The paired-pulse ratio data does not suggest changes in release probability (pr) between R+E and R+E+M conditions (Figure 6). Could the authors provide an additional, independent pr (and RRP?) estimate, e.g., based on (IPSC?) trains? Moreover, it would be interesting to plot the relative decrease in mini frequency for R+E KO and R+E+M KO and discuss how a potential change would relate to pr.

7. Regarding the quantification of STED fluorescence intensity data: Can the authors exclude that crosstalk between both channels causes the remaining Munc13-1 fluorescence in R+E+M cultures (Figure 1J, 2I)? How do the changes in fluorescence intensity at STED resolution compare to the corresponding changes at confocal resolution? What is the justification for restricting the analysis to side-view synapses? Finally, it would be helpful to plot relative changes in fluorescence intensity for R+E and R+E+M, similar to Figure 1 —figure supplement 1F.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Munc13 supports vesicle fusogenicity after disrupting active zone scaffolds and synaptic vesicle docking" for further consideration by eLife. Your revised article has been evaluated by Lu Chen (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

Reviewer #2 (Recommendations for the authors):

The authors have significantly improved the manuscript with new experimental results and further substantial discussion of key aspects of the study. They have satisfactorily attended to my initial concerns.

Reviewer #3 (Recommendations for the authors):

In general, most points have been addressed by new experiments and text revisions. However, some of the concerns were apparently misunderstood by the authors. We would appreciate if the following remaining points could be clarified:

1. Regarding the major finding and effect sizes: Release is strongly compromised in both, RE KO (IPSC amplitude by ~82% of control) and REM KO cultures (~89% of control). EPSCs or IPSC amplitudes recorded in RE cultures were reduced by 90% and 81% (Wang et al., 2016), or by ~90% and ~85% (Tan et al., 2022) compared to time-matched controls in previous papers by the same group. Hence, the question arises whether the relative release defect of REM compared to RE cultures observed in the present study would persist in subsequent data sets.

Furthermore, the authors base their interpretation and effect size estimation on data that are normalized to the mean of the respective control group. Crucially, although the relative difference seems very high (~60%, comparing 11% with 18%), it strongly depends on the magnitude of the release defect. Conversely, the authors could have compared the relative reduction by 80% vs. 90% instead of the remaining fraction, which would give a relative difference of only ~13%. Thus, power calculations based on the relative effects of data normalized to the control mean are not informative in this context. Moreover, basing the sample sizes of the present study on previous studies with effect sizes of 80-90% appears inappropriate as well. To assess if release is indeed more strongly reduced in REM cultures, the authors could perform a two-way ANOVA and test for an interaction effect after not only normalizing the RE and REM datasets to the respective control means but also the control groups to their respective mean values. This analysis would reveal if there were indeed a significant difference between RE and REM. Otherwise, new data has to be collected to support this central finding of the paper.

2. Based on their data, the authors put forward a model in which the remaining release in RE KOs "depends on" Munc13. As most readers will unlikely know/look up the definition of "depend on", we strongly suggest using the term "partially depends on" (or the likes) throughout the text to clearly emphasize that they cannot differentiate between Munc13-dependent and independent mechanisms.

3. In their response, the authors claim that the relationship between confocal and STED fluorescence intensity measurements has been characterized extensively in previous papers (for example in Figures 1 and S1 of (Wong et al., 2018) and Figure S2 of (Tan et al., 2022)). However, data showing a clear correlation between confocal and STED fluorescence intensities were not presented in these references. Although we do not doubt the major conclusions based on the STED data, the authors should either cite another paper or provide a corresponding correlation in the present manuscript.

https://doi.org/10.7554/eLife.79077.sa1

Author response

Essential revisions

1. It is unclear why Cre mediated removal of the presynaptic scaffold proteins was initiated at DIV 5, at a time point when synapse formation is already well underway. This likely contributed to the hypomorphic phenotype. Please comment.

We thank the reviewers and editors for giving us the opportunity to comment, and we have performed a new experiment to address this important point. As a matter of background, in the culture system we use here, functional synapses are not detected at this time point. Spontaneous release is typically detectable only at DIV8-DIV9 (Mozhayeva et al., 2002), and evoked release builds up after DIV10. We have established before that with DIV5 cre infection to remove CaV2 channels (Held et al., 2020), for example, release never builds up to robust levels (see Figure S4A and S4B, in (Held et al., 2020)). Thus, we think that DIV5 infection is appropriate in general. This point has been added on lines 119-120 of the manuscript.

In a previous study, we have shown that even when the Munc13-1 allele used here is knocked out throughout development by constitutive deletion in the germline, there continues to be some release in cultured autaptic neurons (see Figure S3 in (Banerjee et al., 2022)). We now include data that this is very similar in the Munc13 conditional knockout mice in the cultured neurons with DIV5 cre lentivirus infection (cKOM neurons, new Figure 1 —figure supplement 2). We think that together, these observations make it highly likely that the persistence of some release in cKOR+E+M neurons is not due to DIV5 cre infection, but to the allele that we use. This allele is described in detail in (Banerjee et al., 2022) and discussed in the present manuscript on lines 161-164.

We felt that the most important point was to establish that the persistence of synapse formation in the hextuple knockout neurons was not due to late infection. Therefore, we pursued an additional experiment in which Cre expression was initiated by lentiviral infection at DIV2 (Figure 2 —figure supplement 2). The density of synapses identified by Synaptophysin and PSD-95 was unaffected, indicating the synapse formation proceeds when RIM, ELKS and Munc13 are deleted with early cre infection.

2. The authors conclude that postsynaptic response is intact. On the other hand, they find kinetic changes in EPSC and IPSC (Figure 1 and Figure 4) that deserve to be discussed. In addition, they do not show any kinetic analysis of mEPSC and mIPSCs; this type of analysis is straightforward to be obtained and it might be useful to assess in part if there are or not changes at the postsynaptic level.

We thank the reviewers for bringing this up and have added analyses of kinetics of mEPSCs and mIPSCs (Figure 4 —figure supplement 1). There was a small increase of mEPSC rise times, similar to what we showed before in cKOR+E neurons (Tan et al., 2022), while mEPSC decay times and mIPSC kinetics were unchanged. Overall, we think that together with the unchanged mPSC amplitudes, this argues that receptors are fundamentally there and detect neurotransmitter. It is currently not possible to fully exclude any postsynaptic effects, but we think in general they cannot be large effects dominating the phenotypes. We added this point on lines 223-226 and used wording that does not exclude that small effects may be present.

Kinetic changes in evoked IPSCs are indeed present in cKOR+E synapses (as shown before (Wang et al., 2016)) and cKOR+E+M synapses, and are presented in Figures 1 and 4. We think that they may reflect asynchrony of release, but drawing strong conclusions is not possible in the studied synapses, and we hope that the reviewers and editors agree that presenting the specific data point of increased IPSC rise times in a descriptive way is appropriate.

3. The major conclusion of the study is that the remaining release after ablation of RIM+ELKS (R+E) is mainly Munc13 dependent. First, the relative changes in mean EPSC/IPSC amplitude or sucrose-dependent RRP size between R+E and RIM+ELKS+Munc13 (R+E+M) KO cultures are comparably small, at most around 15% (Figures 4, 5). This is expected given the strong reduction in Munc13-1 levels in R+E KO (Figure 1; Wang et al., 2016). The relative differences between the experimental groups are smaller than the differences between different control data sets. Moreover, previous data sets obtained from R+E KO cultures under similar conditions by the same group suggest a similar relative decrease in E/IPSC amplitude or RRP size in R+E KO (Wang et al., 2016; Tan et al., 2022) compared to R+E+M KOs (both by ~90%). Given the parameter distributions (e.g., control EPSC amplitudes range between 0.5 and >1.5 nA and display clusters that likely reflect different cultures, Figure 4H), how confident can the authors robustly resolve average relative changes by ~10%? To evaluate whether the remaining release in R+E KOs is indeed Munc13-dependent, it would be helpful to report effect sizes and to provide a post-hoc power calculation. Irrespectively of the outcome of a power analysis, it may be advisable to increase sample sizes for some of the major experiments.

We thank the reviewers and editors for raising this point and for allowing us to comment on it. We have performed power analyses as requested and describe it below. We think that there are two important components to the point that is raised and we have made a concerted effort in the revised manuscript to be exceedingly clear on them.

Comparisons of controls across experiments. We think that these comparisons are complicated by a number of factors. The experiments are often months or years apart. There are genetic background differences between the mouse lines. In the specific case of the work presented here, there are genotype differences between the controls (the controlR+E+M neurons are Munc13-2 constitutive knockout neurons). Therefore, appropriate controls have to be performed for each experiment, which is now clearly described in the manuscript on lines 120-122, 147-152, 182-186. In the case of the presented experiments, this means that recordings are always done interleaved on the same day and with neurons that are identical in genotype except for the application of cre-expressing virus vs. Δcre-virus (a recombination-deficient version of cre). Hence, any cross-comparison of effects requires normalization to these controls, rather than direct comparison of absolute values, and controls should not be directly compared.

Effect magnitudes of control-normalized data. The only way to cross-compare is to normalize the mutant data to their controls that are matched in terms of genotype and interleafed on the day of recording. By doing so, in Figures 4 and 5, we compare means with 40-70% differences and assess significance based on this comparison. In these figures, the data are shown on a scale to 1.0 (100%) to reveal clearly that the data are normalized, but comparisons and statistics are done on effects sized >40%, not on 10-15% effect sizes. We are confident about these effects based on these numbers, and also based on the observation that they are present in EPSCs, IPSCs, and EPSC-RRP and IPSC-RRP measurements, which cross-validates the observations to some extent.

As instructed, we have also done post-hoc power analyses on these effects. Based on this analysis, the confidence to conclude that we detected an effect correctly in Figures 4L, 4M, 5I and 5J is 90.1%, 86.4%, 99.7% and 65.3% respectively. A different way to use post-hoc power analyses is to ask how many n’s we need minimally to detect the observed effect sizes with 90% confidence given the control data and its variability. We calculated this as well and found that we would need 16 cells in Figure 4L (actual experiment: 19 and 20 cells), 13 cells in Figure 4M (actual: 18 and 31 cells), 7 cells in Figure 5I (actual: 17 and 23 cells), and 9 cells in Figure 5J (actual: 18 and 21 cells). Altogether, these post-hoc calculations support that the number of observations we used in these experiments are sufficient to detect the observed effects with reasonable confidence.

Finally, we think that it is inappropriate to add n after an experiment is completed as we set n before we started the experiment. The only possible approach would be to repeat the entire experiment from the beginning independently with a larger n, which is not feasible for a revision (and we think not necessary given the points above). In summary, we feel that the effects are robust and established, and we hope that the reviewers and editors agree.

4. Although EPSC/IPSC amplitude, RRP size, and mEPSC/mIPSC frequency are strongly reduced in R+E+M hextuple KO cultures, these cultures still display significant evoked and spontaneous synaptic transmission. For instance, ~20% of the IPSC RRP (Figure 5H) or ~40% of mEPSCs (Figure 4B) remain in R+E+M KO neurons. The authors attribute this to incomplete loss of Munc13 or Munc13-independent release (Discussion). In line with the first hypothesis and previous reports, their confocal data of Munc13-1,-2,-3 KO cultures indicate a significant anti-Munc13-1 signal (~25% of control; Figure 1 —figure supplement 2B). Thus, it remains unclear whether the remaining release in R+E KO neurons is Munc13 dependent. An alternative hypothesis is that a significant fraction, if not the majority, of the remaining release is Munc13 independent. This would be equally, if not more interesting. Thus, unless the authors directly demonstrate that the release remaining in R+E KO neurons indeed requires Munc13, all respective statements in the manuscript should be revised accordingly (e.g., "Our data reveal that the transmitter release that remains after active zone disruption upon RIM and ELKS deletion depends on Munc13., l. 324)". Indeed, an alternative interpretation of the current data could be quite similar to that of Wang et al. 2016: Fusion competent vesicles persist upon ablation of RIM-1/2, ELKS-1/2, and Munc13-1/2.

We agree with the overall assessment, and we have made a concerted effort in the manuscript to be clear in phrasing. In respect to the cited statement above: we use the word “depends on” to express that release is further impaired in the RIM+ELKS+Munc13 knockouts, and we do not understand it equal to “is required for”. We think that this use is correct. For example, it is difficult to disagree with the following statement that we provide to illustrate our use of “to depend on”: “While our everyday lives depend a lot on email by now, email is not required for life”. Indeed, the Merriam-Webster definition of “depend on” is “to need (someone or something) for support”. We think that in that sense, the use of “depends on” is appropriate; our data reveal that the remaining release in cKOR+E neurons depends on Munc13 because it is reduced by ~50% in cKOR+E+M neurons. We have scrutinized the revised manuscript for the correct wording around this topic, and have changed “depends on” to “depends on at least partially” (or similar) in many cases, for example on lines 242, 261, 285, 315 and 349. The wording we use in captions is: “Munc13 contributes to a remaining functional RRP after active zone disruption” (results); “The remaining functional RRP in RIM+ELKS-deficient synapses depends on Munc13.” (Figure 5 caption).

The key question is of course why there is release left in cKOR+E+M. We think that there are two alternative, fundamentally different explanations for this point:

Incomplete loss of Munc13-1. A previous study has established in cultured autaptic neurons after germline deletion of Munc13-1 that excitatory synaptic transmission was impaired but not abolished in the allele we used here (Banerjee et al., 2022), different from previous Munc13-1 null alleles (Augustin et al., 1999). In a new experiment (Figure 1 —figure supplement 2), we confirm that this is also the case in dissociated mixed cultures and with application of cre-lentiviruses instead of germline recombination. The previous study (Banerjee et al., 2022) has described that a very small amount of spliced-over Munc13-1 remains after cre-recombination (estimated to be <5%). In our study, it was required to use the conditional allele because a hextuple knockout experiment would be excruciatingly complicated to do with a lethal allele. This is described on lines 161-164 of the manuscript.

Munc13-1 independent release. The alternative possibility is that some release remains because it is fully Munc13-independent. While this is an exciting possibility, it is impossible to make this conclusion in our experiments because there is a small amount of Munc13-1 left (point above) (Banerjee et al., 2022). We have included it as a possibility on lines 240-241 and 317-320 of the revised manuscript.

5. The rise time of NMDAR EPSCs was previously shown to be strongly attenuated in R+E KO cultures (Wang et al., 2016). How does the EPSC rise time in R+E+M KO cultures relate to the one in R+E KO neurons?

We thank the reviewers for pointing this out. The 20-80% rise time of NMDAR-EPSCs is significantly increased in cKOR+E+M synapses compared with that in controlR+E+M synapses (controlR+E+M: 7.6 ± 0.4 ms, cKOR+E+M: 9.1 ± 0.5 ms, p = 0.037, 20 cells/3 cultures each). This effect is qualitatively similar to cKOR+E synapses. The effect magnitude is somewhat smaller here than in Ref. (Wang et al., 2016). Possible explanations are: (i) Munc13-2 is knocked out in controlR+E+M synapses but not in controlR+E synapses; (ii) different extracellular calcium concentrations are used here (1.5 mM) vs. before (2 mM) (Wang et al., 2016); (iii) other technical factors, for example cre infection time points or other differences between the current experiments and the ones that were done ~8 years ago (exact composition of some of the culture media ingredients). Finally, we note that the kinetics of these NMDAR responses are very slow (~5-10 ms rise kinetics) compared to the sub-millisecond kinetics of the fusion reaction. Hence, it is unclear how much of the slowdown reflects changes in release kinetics. Because of this uncertainty, we would prefer to not include the NMDAR kinetic analyses in the paper although they are slowed down and consistent with previous results. We believe that no interpretation should be made. We also think that this is different for IPSC kinetics, which are included in the manuscript because they are much faster and hence likely provide insight into release kinetics. We hope that the reviewers agree given our explanations.

6. The paired-pulse ratio data does not suggest changes in release probability (pr) between R+E and R+E+M conditions (Figure 6). Could the authors provide an additional, independent pr (and RRP?) estimate, e.g., based on (IPSC?) trains? Moreover, it would be interesting to plot the relative decrease in mini frequency for R+E KO and R+E+M KO and discuss how a potential change would relate to pr.

We thank the reviewers for these suggestions and have performed new experiments for both points.

Comparison of mini frequencies. For comparison of spontaneous synaptic transmission in cKOR+E+M synapses and cKOR+E synapses, we performed new recordings of mEPSCs and mIPSCs in controlR+E and cKOR+E synapses (Figure 6 —figure supplements 1A, 1B, 1E, 1F; note that (Wang et al., 2016) used a different extracellular calcium concentration). At the same time, we performed new recordings of mEPSCs and mIPSCs in controlR+E+M synapses and cKOR+E+M synapses such that the recordings were done during the same time frame and not years apart (Figures 4A-4F). This analysis reveals that the mini frequencies in cKOR+E+M synapses and cKOR+E synapses are similar (Figure 6 —figure supplements 1D and 1H). We also have previous datasets that were done simultaneously ~2 years ago (Figure 4A), allowing for an independent comparison of mEPSC frequencies in cKOR+E+M and cKOR+E synapses (the cKOR+E results were published in (Tan et al., 2022)). In this earlier dataset as well, mEPSC frequencies were similar in the two genotypes.

Comparison of responses to stimulus trains. We now provide data recorded during train stimulation, and we find that the reduced depression during these stimulus trains is very similar in cKOR+E synapses and cKOR+E+M synapses (Figure 6 —figure supplements 1I-1L).

We think that overall, these data support the PPR data in Figure 6 to establish that release probability is not further affected by additional knockout of Munc13.

One often performed analysis is an RRP-determination based on back-extrapolation of trains, a method that might be pursued based on the reviewer comment above. We performed the analyses (Author response image 1), and overall, the outcomes look reasonable and consistent with all of our data. We note however, that assumptions that these methods require are not met by our data. Most importantly, vesicular release probability p has to be high in order for this analysis to be valid (Kaeser and Regehr, 2017; Neher, 2015; Thanawala and Regehr, 2016). In our mutants, p is severely affected after multi-protein knockout, and it cannot be rendered high enough, even by increasing extracellular calcium (Wang et al., 2016). A second important point is that the analysis should only be done on a homogenous population of synapses, which is not the case in our cultured hippocampal neurons. Hence, although the overall outcomes look reasonable, we strongly feel that the analyses should not be added to the paper because key assumptions are not met. We hope that the reviewers agree that just showing the IPSCs during stimulus trains (Figure 6 —figure supplements 1I-1L), but not the back-extrapolation (Author response image 1), is the most appropriate way to present these data.

Author response image 1
(a, b) Cumulative IPSC amplitude plots in stimulus trains (original data from Figure 6 —figure supplements 1I-1L, n’s as in those figures). (c, d) Back-extrapolation to time zero yields the IPSC-amplitudes at Y-intercept to estimate the recovery-corrected pool size for each cell, back-extrapolation was based on the last ten (4150) responses. (e) Comparison of IPSC at Y-intercept normalized to their own controls in cKOR+E (absolute data from c) and cKOR+E+M (from d) neurons.

7. Regarding the quantification of STED fluorescence intensity data: Can the authors exclude that crosstalk between both channels causes the remaining Munc13-1 fluorescence in R+E+M cultures (Figure 1J, 2I)? How do the changes in fluorescence intensity at STED resolution compare to the corresponding changes at confocal resolution? What is the justification for restricting the analysis to side-view synapses? Finally, it would be helpful to plot relative changes in fluorescence intensity for R+E and R+E+M, similar to Figure 1 —figure supplement 1F.

We have performed two new experiments as outlined below. As a matter of background: we have used the method of side-view analyses extensively in the past five years and have published it in many papers (Emperador-Melero et al., 2021a, 2021b; Held et al., 2020; de Jong et al., 2018; Nyitrai et al., 2020; Tan et al., 2022; Wong et al., 2018). We also provide revised methods in the manuscript that refer to these papers and describe the methods in great detail. We note that we have also previously provided detailed supplemental figures on the justification and process of side-view synapse selection (Figure S3 in (Held et al., 2020), Figure S2 in (Nyitrai et al., 2020), Figure S6 in (Emperador-Melero et al., 2021b) and Figure S1 in (Tan et al., 2022)) and on the comparison between confocal and STED microscopy in this specific system (Figure S1 in (Wong et al., 2018), Figure S2 in (Tan et al., 2022)).

Crosstalk between channels. Based on our extensive experience with the use of this method for the characterization of knockouts, we are confident that there is no detectable crosstalk. To exclude the crosstalk for our specific conditions, we performed a new experiment in which we included a negative control without primary antibodies for the protein of interest, Munc13-1, with the same combination of primaries (Author response image 2). In the “no primary antibody” condition, no detectable signal in the relevant STED channel was detected, indicating that crosstalk is not the reason for background in our experimental setup.

Author response image 2
(a-c) Sample STED images (a) and quantification (b, c) of side-view synapses stained for Munc13-1 (imaged in STED), PSD-95 (imaged in STED), and Synaptophysin (imaged in confocal). Munc13-1 primary antibody was not added during staining process as negative control (“no primary antibody”). Peak position and levels were analyzed in line profiles (600 nm x 200 nm) positioned perpendicular to the center of elongated PSD-95 structure and aligned to the PSD-95 peak, 20 synapses/1 cultures each.

Justification for side-view synapse selection. Side-view synapse selection is necessary because the analysis of active zone localization depends on assessing distances between the PSD and the active zone. This distance cannot be measured in synapses in other orientations, for example top-view, for two reasons. First, z-resolution in STED is worse than x-y resolution. Second, and critically, STED microscopy bleaches significantly in the area of signal acquisition. Hence, we only take single sections and do not reconstruct individual synapses in 3D. In a top-view synapse, the active zone and the PSD are overlayed in a single section and it is not possible to measure their distance. Hence, side-view selection is required. Because it is an experimenter-based selection process, this is done only based on the Synaptophysin and PSD-95 signals; at the time of side-view selection the experimenter is blind to the protein of interest. This is described in detail on lines 688-699.

Relationship of fluorescence intensities in confocal and STED imaging. We have characterized this in previous papers extensively, for example in Figures 1 and S1 of (Wong et al., 2018) and Figure S2 of (Tan et al., 2022). Hence, if there are changes in overall levels of a protein in a synapse, STED and confocal intensities typically correlate well. To directly compare Munc13-1 confocal (Figure 2 —figure supplements 1C and 1D) and STED levels (Figures 2G-2K) in cKOR+E+M synapses, STED and confocal images stained for Synaptophysin, PSD-95 and Munc13-1 in controlR+E+M and cKOR+E+M neurons were taken at the same time in new experiments. The Munc13-1 signals were removed efficiently at the active zone area in STED (Figures 2G-2K) and from synapses in confocal (Figure 2 —figure supplements 1C and 1D), and the reductions correlated well with one another.

Relative changes in fluorescence intensity for cKOR+E and cKOR+E+M synapses. We now provide this comparison in Figure 2 —figure supplement 1B. These analyses show that Munc13-1 levels in cKOR+E synapses were higher than that in cKOR+E+M synapses with a broad shoulder towards the inside of the nerve terminal, similar to what we observed in confocal microscopy (Figure 2 —figure supplement 1E).

We hope that these points, together with our previous extensive work on establishing this STED microscopy workflow, sufficiently answer the reviewers’ comments.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

Reviewer #3 (Recommendations for the authors):

In general, most points have been addressed by new experiments and text revisions. However, some of the concerns were apparently misunderstood by the authors. We would appreciate if the following remaining points could be clarified:

We thank the reviewer for giving us the opportunity to clarify these remaining points.

1. Regarding the major finding and effect sizes: Release is strongly compromised in both, RE KO (IPSC amplitude by ~82% of control) and REM KO cultures (~89% of control). EPSCs or IPSC amplitudes recorded in RE cultures were reduced by 90% and 81% (Wang et al., 2016), or by ~90% and ~85% (Tan et al., 2022) compared to time-matched controls in previous papers by the same group. Hence, the question arises whether the relative release defect of REM compared to RE cultures observed in the present study would persist in subsequent data sets.

Furthermore, the authors base their interpretation and effect size estimation on data that are normalized to the mean of the respective control group. Crucially, although the relative difference seems very high (~60%, comparing 11% with 18%), it strongly depends on the magnitude of the release defect. Conversely, the authors could have compared the relative reduction by 80% vs. 90% instead of the remaining fraction, which would give a relative difference of only ~13%. Thus, power calculations based on the relative effects of data normalized to the control mean are not informative in this context. Moreover, basing the sample sizes of the present study on previous studies with effect sizes of 80-90% appears inappropriate as well. To assess if release is indeed more strongly reduced in REM cultures, the authors could perform a two-way ANOVA and test for an interaction effect after not only normalizing the RE and REM datasets to the respective control means but also the control groups to their respective mean values. This analysis would reveal if there were indeed a significant difference between RE and REM. Otherwise, new data has to be collected to support this central finding of the paper.

After discussion of this point with the reviewing and senior editors for obtaining guidance, we have consulted with three biostatistics experts. As outlined below, we have added new statistical analyses. We thank the reviewer for insisting on this point, as we think that the newly added analyses, described below (and presented in the paper in Figures 4N, 4O, 5K, 5L, 6K, 6L, Figure 4 —figure supplement 2, Materials and methods) are a significant addition to this paper and hopefully for the field for better accounting of the data structure that is inherent to the types of analyses we perform.

Consultation. We consulted with the following experts. (1) Dr. Clement Ma is a Biostatistician and Assistant Professor at the Dalla Lana School of Public Health University of Toronto, he is also Co-Director of the Biostatistics Core Services at the Centre for Addiction and Mental Health. (2) The Harvard Biostatistics Student Consulting Center is a service for Harvard Researchers to consult about statistical testing. (3) Dr. Richard Born is a faculty member in the Neurobiology Department at Harvard and an expert in the quantitative analyses of neuroscience data. Dr. Born teaches all Harvard neuroscience graduate students in statistics, and this is a program with a heavy emphasis on quantitative methods. Graduates regularly go on to have careers in highly quantitative fields such as artificial intelligence research and computational neuroscience. The overall outcomes of these consultations were as follows:

a) The use of conventional statistical testing after control-normalization is appropriate as long as we only reach conclusions for the comparison of the two mutant groups, but not for control groups. We think that this is exactly how we used these comparisons. The controls are different (genetic background, constitutive knockout of Munc13-2) and should not be directly compared unless an experiment is designed to do so.

b) One general limitation in our analyses is that we compare means of individual measurements, and the statistical methods assume that the measurements are independent. However, this is not necessarily the case 1, We record multiple sweeps from each cell and multiple cells from each batch of culture, and we then average per cell. To account for this nested structure, we adapted an analysis method that does not suffer from this limitation and offers a fully independent test of the hypothesis that additional knockout of Munc13 impairs release over the deficit observed in RIM+ELKS knockout. The null hypothesis is that no further deficit is observed. We used a hierarchical bootstrap to test the probability of the null hypothesis. This method is similar to other hierarchical statistical approaches, such as repeated measures ANOVA and linear mixed effects models, however it has the added advantage of not making any distributional assumptions about the underlying data and it allows for unequal variance among the groups. In addition, it is much more intuitive and straightforward to implement even when there are multiple nested levels to the data – see ref. 1 for details and a comparison with other methods. These new analyses directly address the overarching concern that the detected reduction through the use of normalization and conventional statistics across four data sets may be explained by factors other than genotype.

Description of the hierarchical bootstrap. Two gene families involved in synaptic release (RIM and ELKS) are removed, which reduces synaptic transmission by about 80%. Now, it is addressed whether knockout of a third gene family, Munc13, further reduces synaptic transmission over this ~80% reduction. However, since the conditional RIM+ELKS knockout is in a different mouse strain from the conditional RIM+ELKS+Munc13 knockout, the measured parameters of synaptic transmission in each conditional knockout strain must be compared to their own controls. In each strain, the controls and test conditions are genetically identical except for the absence/presence of Cre recombinase.

We define: strain 1: R+E (short for RIM+ELKS)

strain 2: R+E+M (short for RIM+ELKS+Munc13)

To perform an experiment in a given strain, the hippocampi of several newborn mice are dissected out, and the cells are dissociated and pooled together in one primary culture. This culture is then divided into two pools subjected to different conditions: one is treated with a lentivirus containing Cre (= cKO); the other is treated with a lentivirus containing a catalytically inactive version of Cre (= control). From each culture, multiple cells are tested. Each cell is patch-clamped and synaptic transmission is tested by measuring the size of the postsynaptic current (EPSC or IPSC) evoked by an action potential or by the application of hypertonic sucrose. This measurement (sweep, technical replicate) is repeated 5-6 times for each cell for action potential-evoked release, and once per cell for sucrose-evoked release. The culture procedure is typically repeated in at least three batches of culture.

We define:

condition 1: cKO

condition 2: control

Thus, each measurement is uniquely identified by five numbers, which are variables (columns) in the tabulated data:

strain, R+E and R+E+M

condition, cKO and control

batch, always 3 batches per experiment

cell, may vary for different batches, 6 to 12 cells per batch

sweep, 5 or 6 per cell (action potential-evoked) or 1 per cell (sucrose-evoked)

Four experimental groups are defined:

cKOR+E

controlR+E

cKOR+E+M

controlR+E+M

The scientific question is whether knocking out Munc13 in addition to RIM+ELKS (cKOR+E+M) causes a greater relative decrease (vs. controlR+E+M) in synaptic transmission as compared to cKOR+E. So, we define our test statistic, T, as:

T = [mean(cKOR+E) / mean(controlR+E)] / [mean(cKOR+E+M) / mean(controlR+E+M)] (1)

The null value for this statistic is: T = 1, corresponding to no further reduction,

and the alternate hypothesis is: T > 1, corresponding to a reduction.

In the hierarchical bootstrap 1, we estimate the sampling distribution of T by resampling with replacement from the raw data, while preserving the hierarchical relationships created by the design of the experiment. For each bootstrap iteration, group identity (combination of strain and condition) is preserved. Resampling is done at three nested levels: batch, cell and sweep. In the framework of a linear mixed effects model, group (= strain x condition) would be a fixed effect and batch, cell and sweep would be random effects. We developed a MATLAB code to perform the hierarchical bootstrap (https://github.com/kaeserlab/Hierarchical_Bootstrap_Analysis_RB).

For each of the four experimental groups, we perform the following hierarchical resampling. Starting with cKOR+E, we first resample with replacement from the batches of cells in this experiment. In all cases, there were three batches per experiment, so we use the MATLAB command 'unidrnd(3,3,1)' to draw 3 random samples from the uniform discrete distribution from 1 to 3. We might, for example, draw B2, B2, B1 for this iteration (Figure 4—figure supplement 2c).

We then start with batch 2, determine the number of cells sampled in this batch, and randomly resample from the cells, replicating the same number of cells but containing a different combination from that of the actual experiment. If there were 7 cells tested in batch 2, we might draw the following bootstrap sample: C1, C3, C7, C6, C7, C5, C6, in which cells 6 and 7 are included twice and cells 2 4 not at all. Then we proceed through this list of cells, each time randomly resampling (always with replacement) from the set of technical replicates for that cell and appending these resampled measurements to the bootstrap sample for this group. After doing this for each of the 7 cells, we go back and repeat the entire process, first for batch 2 (again), but selecting a different sample of the 7 cells, and then for batch 1. At the end we have a bootstrap sample for this experimental group that is exactly the same size as our original data set, but it contains a different subset of measurements and, importantly, it preserves the nested, hierarchical structure of our experiment. The above process is then repeated for the remaining three groups (controlR+E, cKOR+E+M, controlR+E+M), then T* is calculated using the formula (1) (the '*' denotes a bootstrap replicate of our test statistic). This entire procedure is repeated 100,000 times, producing an estimate of the sampling distribution of T. Based on the distribution of the 100,000 T* values, we calculated 95% confidence intervals using the percentile method 2. To do this, we sorted the 100,000 values of T* in ascending order, then defined the lower bound as the 2,500th value and the upper bound as the 97,500th value. In addition, we calculated the probability of the null hypothesis, PH0, given our data as:

PH0 = #{T* ≤ 1} / 100,000

Note that this is not a traditional p-value calculated with standard statistical tests, which is the probability of obtaining a result as extreme or more extreme assuming the null hypothesis to be true, but rather a measure of the probability of a given hypothesis being true given our data 1.

Outcome. Using this method, we found low probabilities for the null hypotheses to be true: 0.042 for action potential-evoked EPSCs (Figure 4N), 0.009 for action potential-evoked IPSCs (Figure 4O), < 0.001 for sucrose-evoked EPSCs (Figure 5K), and 0.084 for sucrose-evoked IPSCs (Figure 5L). This confirms our conclusions reached with conventional statistics that synaptic transmission and the pool of releasable vesicles are further reduced by additional removal of Munc13.

Additional assessment of data structure. It seems a priori plausible that the design of the experiments would impart structure to our data. That is, it is likely that repeated measurements (sweeps) from one cell are more similar to each other than to measurements from another cell. And we might also expect that the cells from one batch are more similar to each other than to those from other batches, despite our best efforts to standardize the treatment of different batches. But is this in fact the case? Might it still be legitimate to treat all of the measurements from a given experimental group as independent?

To address this question, we performed two additional analyses. First, within each experimental group (i.e. eliminating the fixed effects of strain and condition), we performed 2-way ANOVA with batch and cell as factors. An example distribution of data points for the NMDAR-EPSC is shown in Figure 4—figure supplement 2b. For the two experiments (action potential-evoked EPSC and IPSC) in which there were repeated measurements for single cells, we ran the ANOVA for each of the four groups, yielding 8 tests, and found that cell was a significant factor in all 8 of them (at p < 0.001) and batch was a significant factor in 7 of 8 (also at p < 0.001). Including all four experiments (the above two experiments plus the sucrose-evoked EPSCs and IPSCs), we found batch to be a significant factor less frequently, but there was still evidence for higher between-batch variance than within-batch variance in 50% of the 16 cases (p < 0.05). This is not surprising, because each batch represented a mixture of neurons from several animals and were thus likely to be more homogeneous. Nevertheless, the ANOVA provides strong evidence that it is right to consider structure in the data to begin with and hence strengthen the point that a method that accounts for such structure, like the hierarchical bootstrap, should be used for analyses.

The second analysis was to repeat the bootstrap in a non-hierarchical way and compare the distributions of T* and the resulting confidence intervals with those from the hierarchical bootstrap. To do this, we simply pooled the measurements from all of the batches and cells from a given experimental group, and resampled with replacement from this pool. This instantiates the assumption that all measurements were independent. As expected, these distributions were much narrower and more bell-shaped than the corresponding distributions from the hierarchical bootstrap (Author response image 3), with the resulting 95% confidence intervals being, on average, less than half the width of their hierarchical counterparts. Importantly, extreme values of T* were much less likely to be represented in the non-hierarchical distribution, and the significance level of the results was overestimated. While this analysis does not directly address the issue of structure in the way that the ANOVA does, it provides a vivid picture of the consequences of not analyzing the data in the appropriate way, and further justifies the use of hierarchical bootstrap as a conservative method to judge the effects assessed here.

Author response image 3
Comparison of hierarchical vs. non-hierarchical bootstrap for the analyses of NMDAR-EPSCs.

Assessment of paired pulse ratios with hierarchical bootstrap. Using the same methodology, we also analyzed the paired pulse ratios presented in Figure 6. Using conventional statistics, these ratios were found to be similar between groups (cKOR+E and cKOR+E+M). With the hierarchical bootstrap, we found the probability of the null hypothesis (T ≥ 1) given our data was 0.69 (NMDAR-EPSC paired pulse ratio, Figure 6K), and 0.61 (IPSC paired pulse ratio, Figure 6L). Thus, not only did we fail to reject the null hypothesis by conventional statistics, but we found strong evidence in support of it, confirming that vesicular release probability is not further affected by removal of Munc13 in cKOR+E synapses.

Power calculations. The power calculations (previous response, page 4) were done based on all data including the control conditions. Even though the data were normalized for the calculation, the control values were included in the calculation and the outcomes are hence independent of whether they are based on normalized or absolute data, all groups were part of the calculation. They are accurate regardless of normalization. We apologize that we did not explain this clearly in the previous response. Finally, the hierarchical bootstrap analyses presented here further support the point that sampling was big enough, because our conclusions are supported with an independent analyses method that accounts for the other variables in the experiments.

Changes to the manuscript. The following changes to the manuscript were made presenting these new analyses:

– inclusion of outcomes of hierarchical bootstrap in Figures 4N, 4O, 5K, 5L, 6K, 6L

– discussion of outcomes in corresponding Results section (page 11, top)

– inclusion of a new figure to describe the analysis procedure, Figure 4 —figure supplement 2

– generation and uploading of a new MATLAB code to perform these analyses, https://github.com/kaeserlab/Hierarchical_Bootstrap_Analysis_RB

– detailed description of the methodology of the bootstrap analyses in the “Materials and methods section”, sub-section “Statistics”

We thank the reviewer for insisting on finding a better way to analyze the effects across mouse lines. We feel that we have adapted a method that independently confirms the major conclusions. We will definitely use this method in the future, and we hope that others will as well.

2. Based on their data, the authors put forward a model in which the remaining release in RE KOs "depends on" Munc13. As most readers will unlikely know/look up the definition of "depend on", we strongly suggest using the term "partially depends on" (or the likes) throughout the text to clearly emphasize that they cannot differentiate between Munc13-dependent and independent mechanisms.

We now use "partially depends on" (or similar) throughout the text.

3. In their response, the authors claim that the relationship between confocal and STED fluorescence intensity measurements has been characterized extensively in previous papers (for example in Figures 1 and S1 of (Wong et al., 2018) and Figure S2 of (Tan et al., 2022)). However, data showing a clear correlation between confocal and STED fluorescence intensities were not presented in these references. Although we do not doubt the major conclusions based on the STED data, the authors should either cite another paper or provide a corresponding correlation in the present manuscript.

We thank the reviewer for asserting that there is no doubt about the major conclusion. The best comparison for this point is provided in Figures S2A, S2G, and S2H of reference 3. This comparison is the most appropriate because RIM1 expression levels are directly controlled using a stronger (RIM1αhigh) or weaker (RIM1αlow) promotor. We think that this experiment is the only experiment that directly controls expression levels for direct comparison of Author response image 4 to plot the STED data similar to the confocal data, and identical to the way we plot peak values in the current manuscript (for example in Figures 1J, 1K, 2E, 2F, 2J and 2K). This analysis reveals that levels assessed with STED microscopy are highly similar to levels assessed with confocal microscopy.

Author response image 4
(a, b) Comparison of RIM1 levels between confocal and STED images. In a, RIM1 levels in indicated conditions are from confocal images (from Figure S2A of 3). In b, RIM1 levels in indicated conditions are from STED images (data from Figure S2H of 3).

We do not have a similarly valid comparison in the current manuscript with an experiment in which we actively titrate expression levels. In fact, there are no experiments with exogenous expression in the current manuscript at all. There are, however, experiments for which a strong correlation is expected. The best current experiment in this respect is the analysis of Munc13 antibody stainings in controlM and cKOM neurons. These stainings are analyzed both with STED microscopy (Figure 1 —figure supplement 1F) and with confocal microscopy (Figure 1 —figure supplement 3E). The signal reductions are similar to one another with 28% remaining in confocal microscopy and 21% remaining in STED microscopy (we note that there is a small amount of Munc13-1 left in these mutants, ref. 4). We now state this in the supplemental figure legend of Figure 1 —figure supplement 3E.

References

1. Saravanan, V., Berman, G. J. and Sober, S. J. Application of the hierarchical bootstrap to multi-level data in neuroscience. Neurons, Behav. data Anal. theory 3, (2020).

2. Efron, B. and Tibshirani, R. J. An Introduction to the Bootstrap. (Chapman and Hall/CRC, 1994). doi:10.1201/9780429246593

3. Tan, C., Wang, S. S. H., de Nola, G. and Kaeser, P. S. Rebuilding essential active zone functions within a synapse. Neuron 110, 1498-1515.e8 (2022).

4. Banerjee, A. et al. Molecular and functional architecture of striatal dopamine release sites. Neuron 110, 248-265.e9 (2022).

https://doi.org/10.7554/eLife.79077.sa2

Article and author information

Author details

  1. Chao Tan

    Department of Neurobiology, Harvard Medical School, Boston, United States
    Contribution
    Conceptualization, Resources, Formal analysis, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3787-0336
  2. Giovanni de Nola

    Department of Neurobiology, Harvard Medical School, Boston, United States
    Contribution
    Formal analysis, Investigation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  3. Claire Qiao

    Department of Neurobiology, Harvard Medical School, Boston, United States
    Contribution
    Resources, Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2084-2478
  4. Cordelia Imig

    1. Department of Neuroscience, University of Copenhagen, Copenhagen, Denmark
    2. Department of Molecular Neurobiology, Max Planck Institute for Multidisciplinary Sciences, Goettingen, Germany
    Contribution
    Resources, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7351-8706
  5. Richard T Born

    Department of Neurobiology, Harvard Medical School, Boston, United States
    Contribution
    Formal analysis, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4360-427X
  6. Nils Brose

    Department of Molecular Neurobiology, Max Planck Institute for Multidisciplinary Sciences, Goettingen, Germany
    Contribution
    Conceptualization, Resources, Writing - review and editing
    Competing interests
    Reviewing editor, eLife
  7. Pascal S Kaeser

    Department of Neurobiology, Harvard Medical School, Boston, United States
    Contribution
    Conceptualization, Formal analysis, Supervision, Funding acquisition, Visualization, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    kaeser@hms.harvard.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-1558-1958

Funding

National Institute of Mental Health (MH113349)

  • Pascal S Kaeser

National Institute of Neurological Disorders and Stroke (NS083898)

  • Pascal S Kaeser

Harvard Medical School

  • Pascal S Kaeser

Max Planck Institute for Multidisciplinary Sciences (open access funding)

  • Cordelia Imig
  • Nils Brose

German Research Foundation (EXC 2067/1-390729940)

  • Nils Brose

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank J Wang and V Charles for technical support, all members of the Kaeser laboratory for insightful discussions and feedback, M Verhage and J Broeke for a MATLAB macro to analyze electron microscopic images, C Ma for advice on data analyses, the Harvard Biostatistics Student Consulting Center for feedback on statistical testing methods, and C Liu for adapting a 3D image analysis code for 2D images. CQ is currently a graduate student at Peking University. This work was supported by grants from the NIH (R01MH113349 and R01NS083898 to PSK) and the German Research Foundation (EXC 2067/1-390729940 to NB). We acknowledge the Neurobiology Imaging Facility (supported by a P30 Core Center Grant P30NS072030) and the Electron Microscopy Facility at Harvard Medical School.

Ethics

This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All animal experiments were approved by the Harvard University Animal Care and Use Committee (protocol number IS00000049).

Senior Editor

  1. Lu Chen, Stanford University, United States

Reviewing Editor

  1. Axel T Brunger, Stanford University School of Medicine, Howard Hughes Medical Institute, United States

Version history

  1. Received: March 29, 2022
  2. Preprint posted: April 1, 2022 (view preprint)
  3. Accepted: November 17, 2022
  4. Accepted Manuscript published: November 18, 2022 (version 1)
  5. Version of Record published: January 6, 2023 (version 2)

Copyright

© 2022, Tan et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,326
    Page views
  • 275
    Downloads
  • 2
    Citations

Article citation count generated by polling the highest count across the following sources: PubMed Central, Crossref, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Chao Tan
  2. Giovanni de Nola
  3. Claire Qiao
  4. Cordelia Imig
  5. Richard T Born
  6. Nils Brose
  7. Pascal S Kaeser
(2022)
Munc13 supports fusogenicity of non-docked vesicles at synapses with disrupted active zones
eLife 11:e79077.
https://doi.org/10.7554/eLife.79077

Further reading

    1. Cell Biology
    2. Microbiology and Infectious Disease
    Juan Xiang, Chaoyang Fan ... Pei Xu
    Research Article Updated

    The relative positions of viral DNA genomes to the host intranuclear environment play critical roles in determining virus fate. Recent advances in the application of chromosome conformation capture-based sequencing analysis (3 C technologies) have revealed valuable aspects of the spatiotemporal interplay of viral genomes with host chromosomes. However, to elucidate the causal relationship between the subnuclear localization of viral genomes and the pathogenic outcome of an infection, manipulative tools are needed. Rapid repositioning of viral DNAs to specific subnuclear compartments amid infection is a powerful approach to synchronize and interrogate this dynamically changing process in space and time. Herein, we report an inducible CRISPR-based two-component platform that relocates extrachromosomal DNA pieces (5 kb to 170 kb) to the nuclear periphery in minutes (CRISPR-nuPin). Based on this strategy, investigations of herpes simplex virus 1 (HSV-1), a prototypical member of the human herpesvirus family, revealed unprecedently reported insights into the early intranuclear life of the pathogen: (I) Viral genomes tethered to the nuclear periphery upon entry, compared with those freely infecting the nucleus, were wrapped around histones with increased suppressive modifications and subjected to stronger transcriptional silencing and prominent growth inhibition. (II) Relocating HSV-1 genomes at 1 hr post infection significantly promoted the transcription of viral genes, termed an ‘Escaping’ effect. (III) Early accumulation of ICP0 was a sufficient but not necessary condition for ‘Escaping’. (IV) Subnuclear localization was only critical during early infection. Importantly, the CRISPR-nuPin tactic, in principle, is applicable to many other DNA viruses.

    1. Cell Biology
    Enrico Radaelli, Charles-Antoine Assenmacher ... Marco Spinazzi
    Research Article Updated

    Impaired spermatogenesis and male infertility are common manifestations associated with mitochondrial diseases, yet the underlying mechanisms linking these conditions remain elusive. In this study, we demonstrate that mice deficient for the mitochondrial intra-membrane rhomboid protease PARL, a recently reported model of the mitochondrial encephalopathy Leigh syndrome, develop early testicular atrophy caused by a complete arrest of spermatogenesis during meiotic prophase I, followed by degeneration and death of arrested spermatocytes. This process is independent of neurodegeneration. Interestingly, genetic modifications of PINK1, PGAM5, and TTC19 – three major substrates of PARL with important roles in mitochondrial homeostasis – fail to reproduce or modify this severe phenotype, indicating that the spermatogenic arrest arises from distinct molecular pathways. We further observed severe abnormalities in mitochondrial ultrastructure in PARL-deficient spermatocytes, along with prominent electron transfer chain defects, disrupted coenzyme Q (CoQ) biosynthesis, and metabolic rewiring. These mitochondrial defects are associated with a germ cell-specific decrease in GPX4 expression leading arrested spermatocytes to ferroptosis – a regulated cell death modality characterized by uncontrolled lipid peroxidation. Our results suggest that mitochondrial defects induced by PARL depletion act as an initiating trigger for ferroptosis in primary spermatocytes through simultaneous effects on GPX4 and CoQ – two major inhibitors of ferroptosis. These findings shed new light on the potential role of ferroptosis in the pathogenesis of mitochondrial diseases and male infertility warranting further investigation.