1. Structural Biology and Molecular Biophysics
  2. Neuroscience
Download icon

Functional synergy between the Munc13 C-terminal C1 and C2 domains

  1. Xiaoxia Liu
  2. Alpay Burak Seven
  3. Marcial Camacho
  4. Victoria Esser
  5. Junjie Xu
  6. Thorsten Trimbuch
  7. Bradley Quade
  8. Lijing Su
  9. Cong Ma
  10. Christian Rosenmund
  11. Josep Rizo  Is a corresponding author
  1. University of Texas Southwestern Medical Center, United States
  2. Charité-Universitätsmedizin Berlin, Germany
  3. Huazhong University of Science and Technology, China
Research Article
  • Cited 21
  • Views 1,957
  • Annotations
Cite as: eLife 2016;5:e13696 doi: 10.7554/eLife.13696

Abstract

Neurotransmitter release requires SNARE complexes to bring membranes together, NSF-SNAPs to recycle the SNAREs, Munc18-1 and Munc13s to orchestrate SNARE complex assembly, and Synaptotagmin-1 to trigger fast Ca2+-dependent membrane fusion. However, it is unclear whether Munc13s function upstream and/or downstream of SNARE complex assembly, and how the actions of their multiple domains are integrated. Reconstitution, liposome-clustering and electrophysiological experiments now reveal a functional synergy between the C1, C2B and C2C domains of Munc13-1, indicating that these domains help bridging the vesicle and plasma membranes to facilitate stimulation of SNARE complex assembly by the Munc13-1 MUN domain. Our reconstitution data also suggest that Munc18-1, Munc13-1, NSF, αSNAP and the SNAREs are critical to form a ‘primed’ state that does not fuse but is ready for fast fusion upon Ca2+ influx. Overall, our results support a model whereby the multiple domains of Munc13s cooperate to coordinate synaptic vesicle docking, priming and fusion.

https://doi.org/10.7554/eLife.13696.001

eLife digest

In the brain, neurons communicate with each other using small molecules called neurotransmitters. Electrical signals in one neuron trigger the release of the neurotransmitters, which then bind to receptor proteins on another neuron nearby. Neurotransmitters are packaged into small compartments called synaptic vesicles and are released from the neuron when these vesicles fuse with the membrane that surrounds the cell. Many proteins are involved in regulating this process to ensure that neurotransmitters are released at the right place and time.

A large protein called Munc13 plays an important role in the release of neurotransmitters. It contains many different regions, including a long domain called MUN and three additional domains called C1, C2B and C2C among others. However, it is not clear how all these domains work together to control neurotransmitter release. Here Liu, Seven et al. address this question using purified proteins inserted into membranes as well as experiments in neurons from mice. The experiments show that the C1, C2B and C2C domains all play key roles in neurotransmitter release. Together with the MUN domain, these three domains help to form bridges between synaptic vesicles and the membrane surrounding the neuron. These bridges could help other proteins involved in neurotransmitter release to form a group that induces vesicle fusion.

Liu, Seven et al.’s findings also suggest that Munc13 proteins cooperate with other proteins to form a 'primed' state in which a synaptic vesicle is ready to rapidly fuse with a neuron’s membrane when triggered to do so by an electrical signal. A future challenge is to find out how the proteins that form this primed state promote vesicle fusion.

https://doi.org/10.7554/eLife.13696.002

Introduction

The release of neurotransmitters by Ca2+-evoked synaptic vesicle exocytosis is a key event for communication between neurons and involves several steps, including vesicle docking at presynaptic active zones, a priming reaction(s) that leaves the vesicles ready for release, and fast Ca2+-triggered fusion of the vesicle and plasma membranes (Sudhof, 2013). Studies of the complex machinery that controls release have shown that eight proteins are particularly important and have established defined roles for them (Rizo and Xu, 2015; Jahn and Fasshauer, 2012; Brunger et al., 2015; Sudhof and Rothman, 2009): i) the soluble N-ethylmaleimide-sensitive factor attachment protein receptors (SNAREs) syntaxin-1, SNAP-25 and synaptobrevin bring the membranes together by forming a four-helix bundle called SNARE complex (Sollner et al., 1993; Poirier et al., 1998; Sutton et al., 1998), which is critical for membrane fusion (Hanson et al., 1997); ii) N-ethylmaleimide sensitive factor (NSF) and soluble NSF attachment proteins (SNAPs; no relation to SNAP-25) disassemble the SNARE complex (Sollner et al., 1993) to recycle the SNAREs (Mayer et al., 1996; Banerjee et al., 1996); iii) Munc18-1 and Munc13s orchestrate SNARE complex assembly, which involves initial binding of Munc18-1 to a self-inhibited ‘closed’ conformation of syntaxin-1 (Dulubova et al., 1999; Misura et al., 2000) and opening of syntaxin-1 by Munc13 (Richmond et al., 2001; Ma et al., 2011; Yang et al., 2015); iv) and Synaptotagmin-1 (Syt1) acts as the Ca2+ sensor that triggers fast release (Fernandez-Chacon et al., 2001), likely via interactions with both membranes (Arac et al., 2006) and with the SNARE complex (Brewer et al., 2015; Zhou et al., 2015).

Reconstitution experiments (Weber et al., 1998) have contributed to establishing some of these key concepts, providing a powerful tool to study the mechanism of synaptic vesicle fusion (Brunger et al., 2015). It is now clear that synaptobrevin-liposomes can fuse with syntaxin-1-SNAP-25 liposomes under some conditions but not others, and that fusion is stimulated to different degrees by Syt1, Munc18-1 or Munc13-4 (Weber et al., 1998; van den Bogaart et al., 2010; Tucker et al., 2004; Yu et al., 2013; Kyoung et al., 2011; Lee et al., 2010; Boswell et al., 2012; Parisotto et al., 2012), but such fusion is abolished by NSF and αSNAP because they disassemble syntaxin-1-SNAP-25 complexes (Weber et al., 2000; Ma et al., 2013). However, inclusion of Munc18-1 and a Munc13-1 fragment enable fusion at least in part because they protect against the disassembly activity of NSF-αSNAP while coordinating SNARE complex assembly (Ma et al., 2013). These results explained the essential nature of Munc18-1 and Munc13s for neurotransmitter release (Verhage et al., 2000; Richmond et al., 1999; Varoqueaux et al., 2002; Aravamudan et al., 1999) and correlated with earlier studies of the role of the HOPS tethering complex in yeast vacuolar fusion (Xu et al., 2010).

Despite these advances, fundamental questions remain about the mechanism of neurotransmitter release, in particular regarding the functions of Munc13s (Rizo and Xu, 2015). These large proteins of presynaptic active zones contain a variable N-terminal region that in some isoforms include a C2 domain (the C2A domain), and a highly conserved C-terminal region that includes (see Figure 1A for Munc13-1): i) a C1 domain involved in diacyglycerol (DAG)-phorbol ester-dependent augmentation of release (Rhee et al., 2002; Basu et al., 2007); a C2B domain that regulates release probability and modifies short term plasticity through its Ca2+- and phosphatidylinositolphosphate-binding activities (Shin et al., 2010); a MUN domain that is key for the crucial function of Munc13 in release (Basu et al., 2005) and mediates the activity of Munc13 in opening syntaxin-1 (Richmond et al., 2001; Ma et al., 2011; Yang et al., 2015); and a C2C domain that is also important for release (Madison et al., 2005; Stevens et al., 2005) and is not predicted to bind Ca2+ but may bind phospholipids because this is a common property of C2 domains (Rizo and Sudhof, 1998). The central function of Munc13s in neurotransmitter release was initially associated to an essential role in vesicle priming (Augustin et al., 1999), but later studies that used stringent definitions of vesicle docking (see discussion) uncovered a critical role for Munc13s in docking that was attributed to their activity in mediating SNARE complex assembly (Weimer et al., 2006; Hammarlund et al., 2007; Imig et al., 2014). However, it is unknown whether Munc13s participate in upstream interactions that might help bridging the two membranes to promote docking and priming. This possibility is attractive because the MUN domain is related to tethering factors involved in diverse forms of membrane traffic (Pei et al., 2009; Li et al., 2011) and is flanked by domains with demonstrated or potential lipid-binding properties. Moreover, while there is evidence that Munc13s modulate release probability and have a role beyond docking [e.g. (Rhee et al., 2002; Hammarlund et al., 2007; Shin et al., 2010)], it is unclear whether they form part of the primed complex after SNARE complex assembly, influencing membrane fusion downstream of priming.

Figure 1 with 1 supplement see all
Munc13-1 C1C2BMUN strongly stimulates lipid mixing between V- and T-liposomes.

(A) Domain diagram of Munc13-1. CaMb = calmodulin-binding sequence. (BC) Lipid mixing assays between V- and T-liposomes alone (T+V) or in the presence of different combinations of Munc13-1 C1C2BMUN, Syt1 C2AB fragment, Munc18-1 (M18), NSF-αSNAP (NSF/SNAP) and synaptobrevin cytoplasmic domain (Syb-cd). T-liposomes contained 1% DAG and 1% PIP2. (DE) Analogous lipid mixing assays performed in the presence of C1C2BMUN (D) or C1C2BMUN plus Munc18-1 and NSF-αSNAP (NSF/SNAP) (E) with T-liposomes containing 1% DAG and 1% PIP2, 1% PIP2 (-DAG), 1% DAG (-PIP2) or no DAG and PIP2 (-DAG-PIP2). Controls of T+V (D) or T+V in the presence of C1C2BMUN plus NSF-αSNAP (NSF/SNAP) (E), both with T-liposomes containing 1% DAG and 1% PIP2, are shown in orange. All experiments were started in the presence of 100 μM EGTA, and Ca2+ (600 μM) was added after 300 s.

https://doi.org/10.7554/eLife.13696.003

To shed light into these questions, we have used a combination of reconstitution and dynamic light scattering (DLS) assays together with electrophysiological experiments. Our results show that both the C1-C2B region and the C2C domain of Munc13-1 play important functions in release, and suggest that these domains help bridging synaptic vesicles to the plasma membrane, facilitating the activity of the Munc13-1 MUN domain in promoting SNARE complex assembly. Moreover, our results indicate that the neuronal SNAREs, Munc18-1, NSF, αSNAP and a Munc13-1 fragment including the C1, C2B, MUN and C2C domains (C1C2BMUNC2C) are sufficient to generate a ‘primed’ state that is ready to trigger fast membrane fusion upon addition of Ca2+, thus resembling the primed state of synaptic vesicles.

Results

Munc13-1 C1C2BMUN strongly stimulates SNARE-dependent lipid mixing

In experiments that followed our recent reconstitution study (Ma et al., 2013) and were directed at analyzing how different factors affect the efficiency of membrane fusion, we first analyzed lipid mixing between synaptobrevin-liposomes (V-liposomes) and syntaxin-1-SNAP-25 liposomes (T-liposomes) by monitoring de-quenching of the fluorescence of NBD-labeled lipids incorporated in the synaptobrevin-liposomes (Weber et al., 1998) (Figure 1). These experiments were initiated in the absence of Ca2+, and Ca2+ was added at 300 s to examine the Ca2+ -dependence of the results. The liposomes contained a synaptic-like lipid composition and, unless otherwise specified, the T-liposomes included DAG and PIP2, which activate the Munc13-1 C1 and C2B domains (Ma et al., 2013). We illustrate the reproducibility of some of the data in Supplementary Figures, showing the quantification of the data at 500 s.

In the NBD de-quenching assays we observed that lipid mixing between V- and T-liposomes is strongly stimulated by a Munc13-1 fragment spanning its C1, C2B and MUN domains (C1C2BMUN) (Figure 1B; red data), which contrasts with the smaller stimulation observed earlier [Figure S8 of (Ma et al., 2013)]. We note that we have been able to reproduce all other results from this previous study and we speculate that the sample of C1C2BMUN used for the experiments of Figure S8 of Ma et al. (2013), which were performed at the end of that study, might have been partially inactivated during storage in the freezer, as we have reproduced the results shown in Figure 1B in more than 20 subsequent reconstitution experiments employing at least five different preparations of C1C2BMUN. The enhancement of lipid mixing induced by C1C2BMUN was independent of Ca2+ and was SNARE-dependent, as it was strongly impaired by addition of the cytoplasmic region of synaptobrevin (Syb-cd) (Figure 1B and Figure 1—figure supplement 1A). The extent of lipid mixing between V- and T-liposomes in the presence of C1C2BMUN was comparable to that caused by a soluble fragment spanning the two C2 domains of Syt1 (C2AB fragment) in the presence of Ca2+; in contrast, Munc18-1 had only a small stimulatory effect on lipid mixing, and did not enhance the stimulation caused by C1C2BMUN (Figure 1B and Figure 1—figure supplement 1A).

As expected, addition of NSF-αSNAP abolished the strong stimulatory effect of C1C2BMUN but lipid mixing was highly efficient again when both C1C2BMUN and Munc18-1 were added in the presence of NSF-αSNAP (Figure 1C and Figure 1—figure supplement 1B), consistent with the notion that C1C2BMUN and Munc18-1 mediate SNARE complex assembly in an NSF-αSNAP-resistant manner (Ma et al., 2013). Note that these experiments did not include Syt1 C2AB and yet lipid mixing was Ca2+-dependent in the presence of C1C2BMUN, Munc18-1 and NSF-αSNAP. Moreover, removal of DAG, PIP2 or both from the T-liposomes caused increasingly stronger impairments of lipid mixing in these experiments but had much milder effects on the stimulation of lipid mixing caused by C1C2BMUN in the absence of NSF-αSNAP (Figure 1D,E and Figure 1—figure supplement 1C,D). These data show that the effect of Munc13-1 C1C2BMUN alone on lipid mixing arises from a property that is largely independent of Ca2+, DAG and PIP2, whereas the lipid mixing observed in the more complete reconstitutions including C1C2BMUN, Munc18-1 and NSF-αSNAP is stimulated by Ca2+, DAG and PIP2, thus exhibiting properties that are more similar to those of neurotransmitter release.

Munc13-1 C1C2BMUN clusters phosphatidylserine-containing liposomes

The ability of Syt1 C2AB to bind simultaneously to two membranes in a Ca2+-dependent manner (Arac et al., 2006) underlies at least in part its activity in stimulating SNARE-dependent lipid mixing (Tucker et al., 2004; Xue et al., 2008). Hence, we tested whether Munc13-1 C1C2BMUN is also able to bridge two membranes by monitoring the formation of liposome clusters by DLS. While C1C2BMUN did not cluster plain liposomes lacking phosphatidylserine (PS), dramatic increases in particle size observed by DLS revealed efficient clustering of PS-containing vesicles caused by C1C2BMUN (Figure 2A,B). Inclusion of synaptobrevin, DAG+PIP2 or syntaxin-1 in the PS-vesicles did not have major effects on the clustering induced by C1C2BMUN, as shown by the bar diagrams of Figures 2B–E and by the corresponding intensity autocorrelation curves (Figure 2F). The two types of representations provide different views of the DLS data; below we use one or the other depending on the aspect that we want to emphasize. We note that there is some degree of variability among the particle sizes observed, which makes it difficult to draw firm conclusions from the small differences observed in Figures 2B–E. Hence, these data show that PS is the main determinant for the vesicle clustering activity of C1C2BMUN, although we cannot rule out that synaptobrevin, syntaxin-1, DAG or PIP2 might affect clustering to a small degree. Similarly, Ca2+ did not have a major effect on clustering of PS-vesicles by C1C2BMUN, although it might increase clustering to a small extent (Figure 2—figure supplement 1).

Figure 2 with 1 supplement see all
Munc13-1 C1C2BMUN clusters PS-containing liposomes.

(AE) The particle size in samples containing phospholipid vesicles alone (gray bars) or after incubation with Munc13-1 C1C2BMUN for 5 min (red bars) in the absence of Ca2+ was measured by DLS. The liposomes had a standard lipid composition including no PS (A), PS (B), PS and synaptobrevin (C), PS+DAG+PIP2 (D) or PS and syntaxin-1 (E). (F) Intensity autocorrelation curves corresponding to the experiments shown in (AE) after incubation with Munc13-1 C1C2BMUN for 5 min.

https://doi.org/10.7554/eLife.13696.005

These results correlate with the observation that the strong stimulatory activity of C1C2BMUN in lipid mixing between V- and T-liposomes does not require Ca2+, DAG or PIP2 (Figures 1B,D), indicating that this activity arises from its ability to bind simultaneously to two PS-containing membranes in a Ca2+-independent manner, thus favoring SNARE complex assembly. This activity might contribute to the role of Munc13s in synaptic vesicle docking and is distinct from the function of Munc13-1 in mediating the transition from the syntaxin-1-Munc18-1 complex to the SNARE complex (Ma et al., 2011), but likely potentiates this function in the reconstitutions that include Munc18-1 and NSF-αSNAP by placing the MUN domain near the SNARE-Munc18-1 machinery.

Simultaneous evaluation of lipid and content mixing

As expected, Syt1 C2AB could not stimulate lipid mixing between V- and T-liposomes in the presence of NSF-αSNAP because NSF-αSNAP disassemble the syntaxin-1-SNAP-25 t-SNARE complex (Ma et al., 2013), but it was surprising that Syt1 C2AB did not have marked effects on the lipid mixing observed in the presence of Munc13-1 C1C2BMUN, Munc18-1 and NSF-αSNAP (data not shown; see also below). This observation prompted us to investigate to what extent the lipid mixing observed in these experiments reflects real membrane fusion. For this purpose, we used an assay that simultaneously measures lipid mixing from de-quenching of the fluorescence of Marina Blue-labeled lipids and content mixing from the development of FRET between PhycoE-Biotin trapped in the T-liposomes and Cy5-Streptavidin trapped in the V-liposomes (Zucchi and Zick, 2011). Addition of unlabeled streptavidin to the reaction ensures that the observed FRET arises only from content mixing. Using this approach, we again observed that efficient lipid mixing in the presence of NSF-αSNAP required C1C2BMUN and Munc18-1, as well as Ca2+, and that Syt1 C2AB did not markedly affect lipid mixing under these conditions (Figure 3A,C). However, content mixing in the presence of Munc13-1 C1C2BMUN, Munc18-1, NSF-αSNAP and Ca2+was inefficient in the absence of Syt1 C2AB and was strongly enhanced by Syt1 C2AB (Figure 3B,D). The difference between lipid and content mixing in the absence of Syt1 C2AB emphasizes the fact that lipid mixing may not necessarily reflect true membrane fusion, as described in previous studies [e.g. (Chan et al., 2009; Zick and Wickner, 2014; Kyoung et al., 2011; Diao et al., 2012; Lai et al., 2014)]. Overall, our results show that Syt1 C2AB selectively enhances content mixing but not lipid mixing under our conditions. These findings indicate that Syt1 plays a role in membrane fusion, in agreement with results from single vesicle assays using full-length Syt1 (Kyoung et al., 2011; Diao et al., 2012).

Figure 3 with 1 supplement see all
Syt1 is required for efficient content mixing but not lipid mixing in reconstitutions including Munc18-1, Munc13-1 C1C2BMUN and NSF-αSNAP.

Lipid mixing (A,C) between V- and T-liposomes was measured from the fluorescence de-quenching of Marina Blue-labeled lipids and content mixing (B,D) was monitored from the development of FRET between PhycoE-Biotin trapped in the T-liposomes and Cy5-Streptavidin trapped in the V-liposomes. The assays were performed in the presence of different combinations of Munc13-1 C1C2BMUN, Munc18-1 (M18) and NSF-αSNAP (NSF/SNAP), and in the absence (A,B) or presence (C,D) of Syt1 C2AB fragment. Experiments were started in the presence of 100 μM EGTA and 5 μM streptavidin, and Ca2+ (600 μM) was added after 300 s.

https://doi.org/10.7554/eLife.13696.007

Experiments performed without streptavidin revealed a small amount of leakiness in reactions containing C1C2BMUN, Munc18-1, NSF-αSNAP and Syt1 C2AB, but the leakiness occurred mostly in the beginning and likely arises because of the presence of a population of small, relatively unstable vesicles (Figure 3—figure supplement 1). Note that in these assays much of the Cy5 fluorescence increase caused by FRET from PhycoE (reflecting content mixing) should occur during the first round of fusion and that no further substantial increases are thus expected in subsequent rounds of fusion or upon detergent addition. Correspondingly, the maximum Cy5 fluorescence observed in our most efficient fusion reactions was similar to that observed upon detergent addition (e.g. Figure 3D, red curve; see Materials and Methods). In contrast, the lipid mixing signal expressed as percentage of maximum Marina Blue fluorescence is much smaller in the same reactions (e.g. Figure 3C, red curve) because fluorescence de-quenching is expected to continue in successive rounds of fusion and to undergo a further, large increase upon detergent addition due to additional probe dilution.

The Munc13-1 C2C domain strongly stimulates liposome fusion

The Ca2+-dependent membrane fusion observed in the presence of C1C2BMUN, Munc18-1, NSF-αSNAP and Syt1 C2AB is efficient but is much slower than that of neurotransmitter release, suggesting that our reconstitutions lack at least one key factor that contributes to the high speed of release in vivo. We hypothesized that the Munc13-1 C2C domain might be such a factor based on evidence suggesting that this domain plays an important role in release (Stevens et al., 2005; Madison et al., 2005). To test this hypothesis, we performed fusion assays between V- and T-liposomes, with or without Syt1 C2AB, in the presence of Munc18-1, NSF-αSNAP and fragments of Munc13-1 that contained the MUN domain alone or together with the C1C2B region, the C2C domain, or both (MUN, C1C2BMUN, MUNC2C and C1C2BMUNC2C, respectively). We found that the MUN and MUNC2C fragments did not support membrane fusion but C1C2BMUNC2C was much more efficient than C1C2BMUN in facilitating Ca2+-dependent fusion; in fact, Syt1 C2AB had no marked effect in the experiments performed with C1C2BMUNC2C (Figure 4 and Figure 4—figure supplement 1), likely because this fragment is already highly efficient in promoting fusion in the time scale of our measurements. Note that, in the absence of Ca2+, C1C2BMUNC2C was also more active than C1C2BMUN in promoting lipid mixing, but did not stimulate content mixing.

Figure 4 with 4 supplements see all
The Munc13-1 C2C domain strongly stimulates membrane fusion.

Lipid mixing (A,C) between V- and T-liposomes was measured from the fluorescence de-quenching of Marina Blue-labeled lipids and content mixing (B,D) was monitored from the development of FRET between PhycoE-Biotin trapped in the T-liposomes and Cy5-Streptavidin trapped in the V-liposomes. The assays were performed in the presence of Munc18-1, NSF-αSNAP and distinct Munc13-1 fragments as indicated, without (A,B) or with (C,D) Syt1 C2AB fragment. Experiments were started in the presence of 100 μM EGTA and 5 μM streptavidin, and Ca2+ (600 μM) was added after 300 s. (E) Intensity autocorrelation curves measured by DLS for isolated V- or T-liposomes, or at different time points as indicated in a fusion reaction performed as in (A,B) with C1C2BMUNC2C and 8-fold dilution of all proteins and liposomes. Lipid and content mixing curves for this reaction, as well as particle size distributions corresponding to several of these intensity autocorrelation curves, are shown in Figure 4—figure supplement 4.

https://doi.org/10.7554/eLife.13696.009

These results show that, indeed, the Munc13-1 C2C domain plays a key role in stimulating liposome fusion, but the lack of activity of the MUNC2C fragment shows that the region spanning the C1 and C2B domains is also important for such stimulation. Since these two domains bind DAG and PIP2, respectively, we tested the effects of removing DAG, PIP2 or both in the T-liposomes in the full reconstitutions including C1C2BMUNC2C and observed considerable impairments of fusion (Figure 4—figure supplement 2), similar to those observed in the lipid mixing assays of Figure 1E. We also attempted to examine the Ca2+-dependence of fusion in these full reconstitutions, which were normally performed with 100 μM EGTA to chelate any residual Ca2+ before addition of 600 μM Ca2+ (to make the concentration of free Ca2+ 500 μM). However, experiments where we added 100 or 120 μM Ca2+ at 300 s (Figure 4—figure supplement 3) yielded similar fusion efficiency to that observed when we added 600 μM Ca2+ (Figure 4D). Since the EGTA present should chelate most of the added 100 μM Ca2+, these results suggest that a small amount of residual Ca2+ (likely in the 1 μM range or below) is sufficient to trigger fusion in these experiments, but further research will be required to assess the Ca2+-dependence more accurately. Note that the sensitivity of the reaction to such low Ca2+ concentrations, compared to those that activate Syt1 (Fernandez-Chacon et al., 2001), can be attributed to the C2B domain present in C1C2BMUNC2C (Shin et al., 2010) (see below), and that residual Ca2+ might arise from the purified C1C2BMUNC2C fragment, which we did not treat with Ca2+ chelators to avoid removal of the Zn2+ ions bound to the C1 domain.

We also compared the effects of the Munc13-1 C1C2BMUN and C1C2BMUNC2C fragments on liposome fusion in the absence of NSF-αSNAP. Interestingly, C1C2BMUNC2C alone stimulated fusion between V- and T-liposomes strongly but in a Ca2+ independent manner (Figure 5A,B and Figure 5—figure supplement 1A,B), unlike the reactions that included Munc18-1 and NSF-αSNAP (Figure 4). The fusion efficiency caused by C1C2BMUNC2C alone was similar to that induced by Syt1 C2AB alone in the presence of Ca2+ and appeared to be somewhat increased by addition of Munc18-1, even though Munc18-1 alone did not stimulate fusion (Figure 5 and Figure 5—figure supplement 1). Syt1 C2AB did not enhance fusion further in the reactions containing Munc18-1 and C1C2BMUNC2C, presumably because fusion is already highly efficient. However, Syt1 C2AB did enhance fusion in the presence of Munc18-1 and C1C2BMUN fragment, which is less efficient than C1C2BMUNC2C (Figure 5 and Figure 5—figure supplement 1).

Figure 5 with 1 supplement see all
Munc13-1 C1C2BMUNC2C can induce Ca2+-independent fusion of V- and T-liposomes in the absence of NSF-αSNAP.

Lipid mixing (A,C) between V- and T-liposomes was measured from the fluorescence de-quenching of Marina Blue-labeled lipids and content mixing (B,D) was monitored from the development of FRET between PhycoE-Biotin trapped in the T-liposomes and Cy5-Streptavidin trapped in the V-liposomes. The assays were performed in the presence of different combinations of Munc18-1 (M18), Syt1 C2AB fragment and Munc13-1 C1C2BMUN or C1C2BMUNC2C as indicated. Experiments were started in the presence of 100 μM EGTA and 5 μM streptavidin, and Ca2+ (600 μM) was added after 300 s.

https://doi.org/10.7554/eLife.13696.014

Munc13-1 C1C2BMUNC2C favors bridging of V- to T-liposomes

Our results suggest that the Munc13-1 C2C domain contributes strongly to promote membrane fusion but in a Ca2+-independent manner, consistent with the fact that it does not contain a full set of the aspartate side chains that commonly form the Ca2+-binding sites of C2 domains (Rizo and Sudhof, 1998; Brose et al., 1995). To investigate the mechanism underlying these findings, we first performed clustering experiments with T- and V-liposomes under the same conditions of Figures 5A,B with addition of only C1C2BMUN or C1C2BMUNC2C at different concentrations, and monitored the particle size after 3 min by DLS. Although these measurements reflect not only liposome clustering but also fusion, clustering should dominate the formation of large particles. These data indicated that C1C2BMUNC2C is somewhat more efficient in liposome clustering than C1C2BMUN, exhibiting substantial clustering activity at 50–100 nM concentration (Figure 6—figure supplement 1). Liposome co-floatation assays also suggested that the presence of the C2C domain might increase the affinity of C1C2BMUNC2C for liposomes, but it is not sufficient for stable liposome binding in the context of MUNC2C (Figure 6—figure supplement 2). However, it was unclear whether these effects of the C2C domain on clustering and liposome affinity are sufficient to explain those observed on membrane fusion.

Interestingly, when we analyzed clustering of V- and T-liposomes separately, we found that C1C2BMUNC2C clustered V-liposomes only in the presence of Ca2+ whereas it clustered T-liposomes in the absence and presence of Ca2+ (Figure 6—figure supplement 3). These results contrast with those obtained with C1C2BMUN, which clusters V-liposomes even in the absence of Ca2+ (Figure 2), and suggest a delicate interplay between multiple lipid binding sites within these large protein fragments (see discussion). Since C1C2BMUNC2C clustered mixtures of T- and V-liposomes more efficiently than C1C2BMUN (Figure 6—figure supplement 1), these data suggested that C1C2BMUNC2C may preferentially bridge V-liposomes to T-liposomes. To test this notion more rigorously, we analyzed liposome clustering after 3 min at 20°C to minimize contributions of liposome fusion to the DLS data. Lipid mixing assays confirmed that very little fusion occurs under these conditions (Figure 6—figure supplement 4). At 20°C, C1C2BMUNC2C again was able to cluster T-liposomes but not V-liposomes in the absence of Ca2+, which required Ca2+ for clustering (Figures 6A–D). To test whether populations of clustered and non-clustered liposomes can be distinguished by DLS, we used plain vesicles that did not contain PS or proteins and that, correspondingly, were not clustered by C1C2BMUNC2C (Figure 6E). Indeed, a clearly bimodal distribution of clustered and non-clustered liposomes was observed by DLS when we added C1C2BMUNC2C to a 1:1 mixture of plain liposomes and T-liposomes (Figure 6F). Importantly, only clustered vesicles were detectable by DLS analysis of a 1:1 mixture of V- and T-liposomes in the presence of C1C2BMUNC2C and the absence of Ca2+ (Figure 6G,H). These data show that, whereas Ca2+-free C1C2BMUNC2C does not cluster V-liposomes (Figure 6A,B), it can bridge V-liposomes to T-liposomes.

Figure 6 with 4 supplements see all
The Munc13-1 C1C2BMUNC2C fragment bridges V-liposomes to T-liposomes.

(A,C,E,G) Intensity autocorrelation curves measured by DLS after 3 min incubations at 20°C on samples containing: (A) V-vesicles alone or in the presence of C1C2BMUNC2C and 100 μM EGTA or 500 μM Ca2+; (C) T-vesicles alone or in the presence of C1C2BMUNC2C and 100 μM EGTA or 500 μM Ca2+; (E) plain vesicles containing no PS alone or in the presence of C1C2BMUNC2C and 100 μM EGTA or 500 μM Ca2+, or a 1:1 mixture of plain vesicles and T-vesicles in the presence of C1C2BMUNC2C and 100 μM EGTA; (G) V-vesicles alone, T-vesicles alone, or 1:1 mixtures of V- and T-vesicles in the presence of C1C2BMUNC2C and 100 μM EGTA or 500 μM Ca2+. (B,D,F,H) Bar diagrams showing the particle size distribution in samples containing: (B) V-vesicles in the presence of C1C2BMUNC2C and 100 μM EGTA; (D) T-vesicles alone or in the presence of C1C2BMUNC2C and 100 μM EGTA; (F) a 1:1 mixture of plain vesicles and T-vesicles in the presence of C1C2BMUNC2C and 100 μM EGTA; (H) a 1:1 mixture of V- and T-vesicles in the presence of C1C2BMUNC2C and 100 μM EGTA. These bar diagrams correspond to the autocorrelation curves of selected samples among those shown in (A,C,E,G) and are intended to illustrate that mixtures of clustered and non-clustered vesicles can be readily distinguished (F), and that Ca2+-free C1C2BMUNC2C does not cluster isolated V-vesicles (B) but bridges V- to T-vesicles (H).

https://doi.org/10.7554/eLife.13696.016

Overall, this analysis suggests that the dramatic stimulation of membrane fusion caused by C1C2BMUNC2C compared to C1C2BMUN (Figures 4,5) arises at least in part because of this preferential bridging of V- to T-liposomes by C1C2BMUNC2C, and perhaps because such bridging is more efficient (Figure 6—figure supplement 1) and/or longer lasting. Hence, these results further suggest that Munc13s play a role in bridging synaptic vesicles to the plasma membrane, which may contribute to its function in docking and facilitate the activity of the MUN in opening syntaxin-1 (Figure 7), and indicate that this bridging activity involves a synergy between the C1, C2B and C2C domains. However, while the ability of Ca2+-free C1C2BMUNC2C to bridge V-liposomes to T-liposomes explains its stimulation of fusion between these liposomes in the absence of Ca2+ (Figure 5), it is unclear why the dramatic stimulation of fusion caused by C1C2BMUNC2C in the presence of Munc18-1 and NSF-αSNAP requires Ca2+ (Figure 4B).

Figure 7 with 1 supplement see all
Model of how bridging of synaptic vesicles to the plasma membrane by the highly conserved C-terminal region of Munc13s can create a cage-like environment and facilitate the activity of the MUN domain in promoting the transition from the syntaxin-1-Munc18-1 complex to the SNARE complex, thus favoring SNARE complex assembly.

Syntaxin-1 (Habc domain, orange; SNARE motif and N-terminus, yellow) is shown in a closed conformation bound to Munc18-1 (purple). Synaptobrevin is shown in red, SNAP-25 in green and the C-terminal region of Munc13-1 in brown. The model is inspired by the ability of C1C2BMUNC2C to bridge V- to T-liposomes (Figure 6) and assumes that the C1-C2B region binds to the plasma membrane while the C2C domain binds to the vesicle membrane. See text and the legend of Figure 7—figure supplement 1 for additional details.

https://doi.org/10.7554/eLife.13696.021

Attempts to test whether Ca2+ causes substantial additional clustering under these conditions were hindered by saturation of the DLS detector as the reaction progressed. We thus performed additional lipid and content mixing assays with all protein and liposome concentrations diluted 2-fold, 4-fold and 8-fold, and found that Ca2+ still stimulated fusion strongly even in the most diluted conditions, albeit at a somewhat slower rate (Figure 4—figure supplement 4A,B). Analysis of the fusion reaction with the 8-fold dilution by DLS revealed that efficient clustering already occurred after 3 min in the absence of Ca2+, and the particle size increased to a very little extent one minute after Ca2+ addition (Figure 4E; Figure 4—figure supplement 4C–F), although the size did increase considerably as the reaction progressed to completion. These results suggest that the action of C1C2BMUNC2C in these experiments is not limited to bridging V- to T-liposomes but also involves activities downstream of such bridging.

Incorporation of membrane-anchored Syt1 in the reconstitution assays

The use of the soluble Syt1 C2AB fragment in our assays allows analysis of the effects of including or omitting the Syt1 C2 domains on fusion, but in vivo Syt1 is anchored on synaptic vesicles. To study how membrane-anchoring of Syt1 affects fusion in our assays, we used liposomes containing synaptobrevin and a Syt1 fragment spanning its transmembrane (TM) and cytoplasmic regions (VSyt1-liposomes). These liposomes contained a smaller percentage of PS (6.8%) that resembles that of synaptic vesicles and prevents inhibition of fusion due to binding of Syt1 to the membrane where it is anchored (Stein et al., 2007). The VSyt1-liposomes fused efficiently with T-liposomes in a Ca2+-independent manner, as described previously (Stein et al., 2007), and the Munc13-1 C1C2BMUN fragment slightly increased the fusion efficiency, but Munc18-1 had no marked effect (Figure 8A,B). However, as expected, fusion was abolished by NSF-αSNAP and, in their presence, fusion required Munc18-1, C1C2BMUN and Ca2+ (Figure 8C,D). These latter results are similar to those obtained in the experiments where the soluble Syt1 C2AB was added instead of incorporating Syt1 into the V-liposomes (Figure 3C,D).

Ca2+-independent membrane fusion between Syt1-containing V-liposomes and T-liposomes becomes Ca2+-dependent in the presence of Munc18-1, Munc13-1 C1C2BMUN and NSF-αSNAP.

Lipid mixing (A,C) between V-liposomes containing Syt1 (VSyt1) and T-liposomes was measured from the fluorescence de-quenching of Marina Blue-labeled lipids and content mixing (B,D) was monitored from the development of FRET between PhycoE-Biotin trapped in the T-liposomes and Cy5-Streptavidin trapped in the V-liposomes. The assays were performed in the presence of different combinations of Munc18-1 (M18), Munc13-1 C1C2BMUN and NSF-αSNAP as indicated. Experiments were started in the presence of 100 μM EGTA and 5 μM streptavidin, and Ca2+ (600 μM) was added after 300 s.

https://doi.org/10.7554/eLife.13696.023

Fusion assays between VSyt1- and T-liposomes in the presence of NSF-αSNAP, Munc18-1 and different Munc13-1 fragments (Figure 9A,B and Figure 9—figure supplement 1) also yielded similar results to those obtained with soluble Syt1 C2AB (Figure 4C,D), as C1C2BMUNC2C was much more active than C1C2BMUN while MUN and MUNC2C remained inactive. Content mixing between VSyt1- and T-liposomes in the presence of NSF-αSNAP, Munc18-1 and C1C2BMUNC2C again required Ca2+, even though there was lipid mixing before adding Ca2+, and was very fast upon Ca2+ addition. Moreover, absence of either Munc18-1 or Munc13-1 C1C2BMUNC2C completely abolished fusion (Figure 9C,D), in correlation with the absolute requirement of both proteins for neurotransmitter release in vivo.

Figure 9 with 3 supplements see all
Fast, Ca2+-dependent membrane fusion between VSyt1- and T-liposomes in the presence of Munc18-1, Munc13-1 C1C2BMUNC2C and NSF-αSNAP, which depends on Ca2+ binding to the Munc13-1 C2B domain.

Lipid mixing (A,C,E) between V-liposomes containing Syt1 (VSyt1) and T-liposomes was measured from the fluorescence de-quenching of Marina Blue-labeled lipids and content mixing (B,D,F) was monitored from the development of FRET between PhycoE-Biotin trapped in the T-liposomes and Cy5-Streptavidin trapped in the V-liposomes. In (A,B), the assays were performed in the presence of Munc18-1, NSF-αSNAP and distinct Munc13-1 fragments as indicated. In (C,D), experiments were performed in the presence of NSF-αSNAP with or without addition of Munc18-1 and/or Munc13-1 C1C2BMUNC2C. In (E,F), assays were performed in the presence of Munc18-1, NSF-αSNAP and WT or D705N,D711N mutant Munc13-1 C1C2BMUNC2C. All experiments were started in the presence of 100 μM EGTA and 5 μM streptavidin, and Ca2+ (600 μM) was added after 300 s.

https://doi.org/10.7554/eLife.13696.024

Functional importance of Ca2+ binding to the Munc13-1 C2B domain

The finding that the data obtained with NSF-αSNAP, Munc18-1 and Munc13-1 C1C2BMUNC2C do not depend strongly on Syt1 but exhibit a drastic Ca2+ dependence indicates that such Ca2+ dependence arises from the Munc13-1 C2B domain, which contains the only known Ca2+-binding sites in these proteins (Shin et al., 2010). To test this idea, we analyzed fusion between VSyt1- and T-liposomes in the presence of NSF-αSNAP, Munc18-1 and a mutant Munc13-1 C1C2BMUNC2C fragment where two of the aspartate Ca2+ ligands were mutated to asparagine (D705N,D711N) to disrupt Ca2+ binding. We found that content mixing was strongly impaired by the D706N,D711N mutation while Ca2+-independent lipid mixing was not affected (Figure 9E,F and Figure 9—figure supplement 2), indicating that Ca2+ binding to the Munc13-1 C2B domain is critical for fusion under these conditions. To examine how these results are related to the membrane bridging activity of C1C2BMUNC2C, we analyzed the particle size in fusion reactions where all protein and liposome concentrations were diluted eight-fold, which still allows a strong Ca2+-induced stimulation of content mixing for WT C1C2BMUNC2C (as observed in the experiments performed with the soluble Syt1 C2AB fragment; Figure 4—figure supplement 4A,B) but not for the D706N,D711N mutant (Figure 9—figure supplement 3A–D). DLS analysis of the 8-fold diluted fusion reactions including WT C1C2BMUNC2C showed that much of the liposome clustering had already occurred after 3 min in the absence of Ca2+, while little changes were observed 1 min after adding Ca2+ and a moderate increase in particle size occurred as the reaction progressed to completion (Figure 9—figure supplement 3E). In analogous reactions with the C1C2BMUNC2C D706N,D711N mutant, efficient clustering occurred after 3 min and did not increase further afterwards (Figure 9—figure supplement 3F).

These results suggest that, while Ca2+  binding to the C2B domain of WT C1C2BMUNC2C might contribute to more efficient bridging of V- to T-liposomes, it is unlikely that an effect on such bridging alone can explain the dramatic enhancement of content mixing induced by Ca2+. Note also that the finding that the D706N,D711N mutation disrupts content mixing but not Ca2+-independent lipid mixing (Figure 9E) correlates with the observation that efficient content mixing in the presence of WT C1C2BMUNC2C requires Ca2+, while there is substantial lipid mixing without Ca2+ (Figures 9A,C,E). Indeed, quantification at 300 s, before Ca2+ addition, showed that the fluorescence increase reflecting lipid mixing was 28.9% of that observed 200 s after Ca2+ addition while content mixing was minimal at 300 s (fluorescence 3.3% of that observed 200 s after adding Ca2+) (Figure 9—figure supplement 2). This difference is exacerbated by the fact that the maximal fluorescence associated with content mixing is expected to correspond to only one round of fusion, whereas additional rounds of fusion can contribute to the maximal fluorescence in the lipid mixing signal. Hence, these results suggest that C1C2BMUNC2C, NSF-αSNAP and Munc18-1 enable formation of a ‘primed state’ that is ready for membrane fusion and includes assembled trans-SNARE complexes, as lipid mixing can occur, but requires Ca2+ binding to the Munc13-1 C2B domain for fast full fusion that might be further accelerated by Ca2+ binding to Syt1.

The Munc13-1 C2B and C2C domains are critical for neurotransmitter release

Rescue experiments in which Munc13-1 fragments were overexpressed using the Semliki Forest virus in neurons from Munc13-1/2 double KO mice indicated that the MUN domain is sufficient to rescue neurotransmitter release (Basu et al., 2005), but other functional studies in chromaffin cells and C. elegans indicated that the C2C domain is also critical to rescue Munc13 function (Stevens et al., 2005; Madison et al., 2005). To clarify the functional importance of different domains for Munc13-1, we performed additional rescue experiments in autaptic neuronal cultures from Munc13-1/2 double KO mice (Varoqueaux et al., 2002) but expressing Munc13-1 fragments with a lentiviral expression vector, which does not overexpress proteins at such high levels as the Semliki Forest virus. Analysis of excitatory postsynaptic currents (EPSCs) revealed that a Munc13-1 fragment encompassing the C1C2BMUNC2C region rescued close to 50% of the EPSC amplitude compared to rescue with WT Munc13-1, whereas Munc13-1 fragments spanning the MUN, MUNC2C or C1C2BMUN sequences led to only very small levels of rescue (Figure 10A,B). Similar results were obtained when we analyzed the readily-releasable pool using hypertonic sucrose (Figure 10C,D). These results need to be examined with caution because distinct expression levels were observed for the different fragments (Figure 10—figure supplement 1). However, all Munc13-1 fragments were expressed at higher levels than WT, and hence their levels should be sufficient to rescue release if the fragments are functional. Indeed, since there is no release in Munc13-1/2 double KO neurons (Varoqueaux et al., 2002), the very small amounts of release observed for the rescues with MUN, MUNC2C and C1C2BMUN fragments imply that these fragments can perform Munc13-1 function to some degree, albeit with very low efficiency, and vast protein production may have compensated for functional deficiency in the previous rescues with Semliki Forest virus overexpression of the MUN domain (Basu et al., 2005). Importantly, the robust rescue observed with the C1C2BMUNC2C fragment compared to C1C2BMUN (Figure 10) shows that the Munc13-1 C2C domain indeed plays an important role in neurotransmitter release, in clear correlation with our reconstitution data. Moreover, the C1C2B region is also critical for Munc13-1 function, as the MUNC2C fragment is much less active than C1C2BMUNC2C in both the rescue experiments (Figure 10) and in our fusion assays (Figures 4 and 9).

Figure 10 with 1 supplement see all
The Munc13-1 C1, C2B and C2C domains are critical for neurotransmitter release.

(A) Representatives traces of single AP-evoked EPSCs from Munc13-1/2 DKO hippocampal neurons rescued with Munc13-1 full length, or C-terminal Munc13-1 fragments, in response to 2 ms somatic depolarization. Depolarization artifacts and action potentials were blanked. (B) Normalized summary plot of EPSC peak amplitudes from Munc13-1/2 DKO hippocampal neurons rescued with Munc13-1 full length or C-terminal fragments. Data were collected during 4 consecutive days of recording. Data were normalized to the mean value of the control group (Munc13-1 full length). Error bars represent SEM. Normalized data were pooled from two independent cultures. Values that differ significantly from controls are indicated (*p<0.05; ***p<0.001) by Non parametric Kruskal-Wallis test with a post hoc Dunn's Multiple comparison test. (C) Representative traces of RRP sizes induced by 5 s hypertonic sucrose solution application, from Munc13-1/2 DKO hippocampal neurons rescued with Munc13-1 full length and C-terminal fragments. (D) Normalized summary plot of RRP charge. For the rescue experiments, approximately equal numbers of green positive Munc13-1/2 DKO neuron rescues with Munc13-1 full length or C-terminal fragments were collected the same day. But due to the fact that neurons that lacked both Munc13-1 and Munc13-2 proteins show no evoked excitatory postsynaptic currents (EPSCs), and no response with sucrose stimulation, EPSC and RRP that show no responses were not quantified in the plots. The following numbers of EPSC or RRP responses were observed out of the total green positive neurons for each condition: FL, 53/55; C1C2BMUNC2C, 45/54; C1C2BMUN, 6/50; MUN, 6/50; MUNC2C, 4/51. EPSC means ± SEM (nA) excluding 0: FL, 1.963 ± 0.2511; C1C2BMUNC2C, 0.83120 ± 0.1427; C1C2BMUN, 0.03298 ± 0.008110; MUN, 0.05984 ± 0.0009355; MUNC2C, 0.05892 ± 0.0009125. EPSC charge means ± SEM (pC) excluding 0: FL, 226.3 ± 42.04; C1C2BMUNC2C, 139.5 ± 23.90; C1C2BMUN, 5.090 ± 1.190; MUN, 1.847 ± 1.092; MUNC2C 8.259 ± 1.092.

https://doi.org/10.7554/eLife.13696.028

Discussion

Great advances have been made to characterize the central components of the neurotransmitter release machinery, but fundamental questions remain about how these components work together to trigger Ca2+-dependent membrane fusion. Strong evidence indicates that Munc18-1 and Munc13s orchestrate SNARE complex formation in an NSF-SNAP-resistant manner (Ma et al., 2013), and that the participation of Munc13s in SNARE complex formation underlies their key functions in vesicle docking and priming (Weimer et al., 2006; Hammarlund et al., 2007; Imig et al., 2014). However, it was unclear whether Munc13s have additional roles upstream and/or downstream of SNARE complex assembly. Moreover, it was unknown how the functions of the different domains that form the highly conserved C-terminal region of Munc13s are integrated. Our results now show that both the C1C2B-region preceding the MUN domain and the C2C domain at the C-terminus are critical for neurotransmitter release, and suggest a strong functional synergy between these domains that arises because they help bridging the synaptic vesicle and plasma membranes, facilitating the activity of the MUN domain in mediating opening of syntaxin-1 (Figure 7). Our results also indicate that the neuronal SNAREs, Munc18-1, Munc13, NSF and α-SNAP are crucial to generate a ‘primed state’ that includes Munc13 as an integral component and is ready to release but needs Ca2+ to trigger fast membrane fusion.

Our reconstitutions recapitulate many key features of synaptic vesicle fusion, providing an ideal system to investigate the mechanism of release. The total abrogation of neurotransmitter release observed in the absence of Munc18-1 and Munc13s (Verhage et al., 2000; Varoqueaux et al., 2002) established that these two proteins are the most central factors of the membrane fusion apparatus together with the SNAREs; accordingly, membrane fusion depends strictly on Munc18-1 and Munc13-1 C1C2BMUNC2C in our reconstitutions, as shown in a compelling fashion by Figure 9D. Moreover, fusion exhibits a tight dependence on Ca2+ (Figure 9D) and removal of the C1-C2B region or the C2C domain of Munc13-1 markedly impairs fusion in our reconstitutions (Figures 4B,D and 9B) as well as neurotransmitter release in neurons (Figure 10). The dependence of fusion on DAG and PIP2 (Figure 4—figure supplement 2) correlates with the notion that these factors enhance neurotransmitter release in part via their respective interactions with the Munc13 C1 and C2B domains (Rhee et al., 2002; Shin et al., 2010). NSF and αSNAP are key for the strict requirement of Munc18-1, Munc13-1 and Ca2+ for fusion (Figures 4,5), and for the dependence of lipid mixing on DAG and PIP2 (Figure 1D,E), because they disassemble syntaxin-1-SNAP-25 heterodimers (Weber et al., 2000), ensuring that vesicle docking, priming and fusion proceed through the Munc18-1-Munc13-dependening pathway (Ma et al., 2013).

Although our reconstitutions are incomplete (see below), these multiple correlations with physiological data imply that the mechanisms of action of the proteins included are likely to be related at least in part to those operating in vivo. The finding that the Mun13-1 C1C2BMUNC2C fragment can bridge V-liposomes to T-liposomes (Figure 6) is particularly revealing because it suggests a natural model for how the different domains that form the conserved C-terminal region of Munc13s cooperate to mediate synaptic vesicle docking and priming (Figure 7), providing an explanation for why this fragment rescues neurotransmitter release and stimulates liposome fusion much more efficiently than shorter fragments (Figures 4, 9 and 10). In this model, we assume that NSF-αSNAP disassemble the syntaxin-1-SNAP-25 complex in the T-liposomes and Munc18-1 binds to the released syntaxin-1 folded into a closed conformation. Bridging of the two membranes through respective interactions with the C1-C2B region and the C2C domain at opposite ends of the highly elongated MUN domain creates a ‘cage-like’ environment to facilitate SNARE complex assembly, placing the MUN domain in an ideal position to exert its activity in accelerating the transition from the syntaxin-1-Munc18-1 complex to the SNARE complex (Figure 7). This model is consistent with studies that revealed an important function for Munc13s in docking using stringent vesicle-plasma membrane distance criteria [direct contact or <5 nm; (Weimer et al., 2006; Hammarlund et al., 2007; Imig et al., 2014)], and that suggested that docking is equivalent to priming, reflecting partial SNARE complex formation (note that this notion may not be valid for more relaxed definitions of docking). Our model postulates that Munc13s function in docking and priming not only because they promote SNARE complex assembly but also because they participate in upstream interactions that provide a bridge between the two membranes.

A function of Munc13s in docking in the traditional sense of bridging two membranes (sometimes referred to as tethering when the intermembrane distances are longer) seems natural given the architecture of their conserved C-terminal region, with an elongated MUN domain that is flanked by domains with demonstrated or putative membrane-binding properties and that is related to tethering factors involved in traffic at diverse compartments (Li et al., 2011). Indeed, these tethering factors likely facilitate SNARE complex formation by similar mechanisms to that proposed here for Munc13-1 (Yu and Hughson, 2010). The presence of C1 domain and C2 domains adjacent to the MUN domain in Munc13s provides opportunities for modulation by factors that regulate release such as DAG, PIP2 and Ca2+. In addition, Munc13-1 contains an N-terminal C2A domain that contributes to vesicle docking via interaction with αRIMs [(Dulubova et al., 2005); M. Camacho and C. Rosenmund, unpublished results], which underlies the finding that rescue of release by C1C2BMUNC2C is incomplete (Figure 10) and further supports the notion of an overall role for Munc13s in docking. We also note that reconstitution studies had previously shown that Munc13-4 promotes docking of V- to T-membranes in a Ca2+-dependent manner (Boswell et al., 2012), but it is unclear to what extent this activity is related to that describe here for Munc13-1 because the C-terminal C2 domain of Munc13-4 binds Ca2+, whereas the Munc13-1 C1C2BMUNC2C is not predicted to bind Ca2+ (Rizo and Sudhof, 1998).

Synaptic vesicle docking and priming occur before Ca2+ influx and hence the underlying interactions are not expected to require Ca2+. Correspondingly, C1C2BMUNC2C can bridge V- to T-membranes in the absence of Ca2+ (Figure 6). The interactions that cause such bridging are still unclear and extensive studies will be required to characterize them because the distinct vesicle clustering properties of C1C2BMUN and C1C2BMUNC2C (Figure 2, Figure 2—figure supplement 1, Figure 6) suggest that these large protein fragments contain multiple membrane binding sites that can cooperate in cis to interact with a single membrane or in trans to bind to two membranes (Figure 7—figure supplement 1). We have been unable to express the C2C domain alone and we have not detected specific interactions of this domain with the SNAREs in preliminary NMR experiments using the MUNC2C fragment. Although this fragment does not bind tightly to liposomes (Figure 6—figure supplement 2), it seems likely that the C2C domain contains a lipid-binding site(s) with moderate affinity, as this feature would explain the stronger overall liposome clustering activity of C1C2BMUNC2C compared to C1C2BMUN (Figure 6—figure supplement 1) and phospholipid binding is a characteristic property of C2 domains (Rizo and Sudhof, 1998). We speculate that Ca2+-free C1C2BMUNC2C may bridge T- and V-liposomes because the C1-C2B region binds to the DAG-PIP2-containing T-liposomes in a configuration that favors binding of the C2C domain in trans to another membrane, i.e. a V-liposome (Figure 7; Figure 7—figure supplement 1C). Cooperation between multiple C2C domains could readily strengthen the V-liposome binding and hence the bridging activity.

Ca2+ had generally small effects on vesicle clustering (Figure 2—figure supplement 1, Figure 6C,G, Figure 4E, Figure 9—figure supplement 3E) except for a strong stimulation of V-liposome docking by C1C2BMUNC2C (Figure 6A) that is unlikely to have physiological relevance. Hence, an effect on docking cannot explain the dramatic effects of Ca2+ in the reconstitutions that include C1C2BMUNC2C, Munc18-1 and NSF-αSNAP (Figures 4B,D,9B,D). Note that the ability of C1C2BMUNC2C to stimulate both lipid and content mixing of V- and T-liposomes is largely independent of Ca2+ in the absence of Munc18-1 and NSF-αSNAP (Figure 5A,B), as expected because Ca2+-free C1C2BMUNC2C clusters V- to T-liposomes (Figure 6G,H). In the presence of C1C2BMUNC2C, Munc18-1 and NSF-αSNAP, there is substantial lipid mixing but practically no content mixing in the absence of Ca2+, and both lipid and content mixing occur very fast upon Ca2+ addition (Figures 4 and 9; Figure 9—figure supplement 2). These observations indicate that partial SNARE complex formation already occurs in the absence of Ca2+, resulting in the formation of a ‘primed state’ that is ready for fusion but only fuses (and fast) upon Ca2+ addition [note in this context that the lipid mixing observed before adding Ca2+ can occur through lipid transfer when the SNARE complex brings membranes transiently into proximity without the need for fusion or hemifusion (Zick and Wickner, 2014)]. Formation of this primed state requires Munc18-1, Munc13-1, NSF, αSNAP and the three SNAREs, but not Syt1 (Figure 4A,B). Such requirements correlate with the findings that Munc18-1 and Munc13s are essential for vesicle priming (Verhage et al., 2000; Varoqueaux et al., 2002) whereas Syt1 is not (Geppert et al., 1994; Bacaj et al., 2015), supporting the notion that the primed state formed in our reconstitutions resembles the primed state of synaptic vesicles.

The nature of the primed protein complex underlying this state is unclear, but it is likely to include C1C2BMUNC2C and Munc18-1 since the MUN domain binds to membrane-anchored SNARE complexes (Guan et al., 2008) and Munc18-1 also binds to SNARE complexes (Dulubova et al., 2007; Shen et al., 2007). This proposal is consistent with data suggesting a role for Sec1p (the yeast Munc18-1) after SNARE complex assembly (Grote et al., 2000) and with the observation that a constitutively open syntaxin-1 mutant fully rescues the docking defect observed in Unc13 nulls in C. elegans but rescues neurotransmitter release only partially, which also suggested a role for Unc13 downstream of SNARE complex formation (Hammarlund et al., 2007). It is also possible that α-SNAP forms part of this primed complex (Zick et al., 2015). Regardless of its composition, our data suggest that the primed state is metastable and requires Ca2+ binding to the Munc13-1 C2B domain for efficient content mixing (Figure 9F) either because the Ca2+ -bound Munc13-1 C2B domain contributes directly to facilitate membrane fusion or because Ca2+ binding releases an inhibitory interaction existing in the primed state. The finding that a mutation in a Ca2+ -binding loop of the Munc13-2 C2B domain increases release probability (Shin et al., 2010) is consistent with both possibilities and supports the notion that Munc13s form intrinsic part of the primed state of synaptic vesicles.

Clearly, our reconstitutions raise many questions and are incomplete, as they do not incorporate other important proteins that control release such as CAPS, complexins, RIMs or Rab3s (Rizo and Sudhof, 2012). While Syt1 C2AB stimulates content mixing in reconstitutions with C1C2BMUN (Figure 3), the effects of Syt1 become masked with C1C2BMUNC2C (Figure 4), likely because this Munc13-1 fragment promotes fusion with high efficiency and no further acceleration can be observed in our experiments. Thus, effects of Syt1 may only be observable at faster time scales, as Syt1 is not essential for release but is key to trigger release with high speed upon Ca2+ influx (Sudhof, 2013). Approaches that allow faster measurements [e.g. single vesicle assays (Lee et al., 2010; Kyoung et al., 2011); Diao et al., 2012; Lai et al., 2014] will be required to test this prediction. An additional advantage of these assays is that they allow distinction of the docking event from lipid and content mixing. Note also that the strong effect of the D705N,D711N mutation in the Munc13-1 C2B domain in our reconstitutions (Figure 9E,F) contrasts with the mild effects of an analogous mutation in the C2B domain of the related isoform Munc13-2 on evoked release (Shin et al., 2010). However, it is plausible that these mild effects arise because this isoform has some functional differences with Munc13-1 (Rosenmund et al., 2002) or because there is some functional redundancy with another protein not included in our reconstitutions (e.g. CAPS), and the increased release probability caused by the mutation in the Munc13-2 C2B domain Ca2+-binding loops (Shin et al., 2010) suggests a role as a Ca2+ sensor that might cooperate with Syt1. Mutating the Munc13-2 Ca2+-binding sites did have marked effects on release during action-potential trains; hence, our reconstitution results, which were obtained in the presence of DAG and PIP2, may be related to hyper-activated states present during these trains.

Further research will be required to distinguish between these possibilities, but we would like to emphasize that our reconstitutions recapitulate many key features of neurotransmitter release. Hence, unexpected findings from our reconstitutions that appear to contradict physiological data may actually uncover novel mechanistic aspects of release that were not observed in physiological experiments because of functional redundancy or compensatory effects by factors not included in the reconstitutions.

Materials and methods

Recombinant proteins

Expression and purification of full length rat syxtaxin-1A, full length human SNAP-25A (with its four cysteines mutated to serines), full-length rat Synaptobrevin, full-length rat Munc18-1, the rat synaptotagamin-1 C2AB fragment (residues 131–421), full length Cricetulus griseus NSF V155M mutant, full-length Bos taurus αSNAP, and rat Munc13-1 fragments spanning the MUN and MUNC2C regions (residues 859–1531 and 859–1735, respectively, both with residues 1408–1452 from a flexible loop replaced by the sequence EF), were described previously (Ma et al., 2011; 2013; Chen et al., 2002; 2006; Dulubova et al., 1999; Xu et al., 2013).

A construct encoding TM and cytoplasmic regions of Syt1 (residues 57–421 with the following cysteine mutations: C74S, C75A, C77S, C79I, C82L) with a C-terminal His-tag within a Pet28a vector (Mahal et al., 2002) was a kind gift from Thomas Sollner. The protein was expressed in Escherichia coli BL21(DE3) cells in Terrific Broth media at 16°C for 18 hr with 0.4 mM isopropyl β-D-1-thiogalactopyranoside. Cells were re-suspended in a buffer containing 50 mM Hepes pH 7.4 and 600 mM KCl with a protease inhibitor mixture, and lysed using an Avestin EmulsiFlex-C5 homogenizer. The soluble fraction of the cell lysate was collected after centrifugation at 48,000 × g for 30 min; 1% β-OG was very slowly added to this soluble fraction and incubated on an orbital shaker for 4 hr at 4°C. This mixture was centrifuged at 48,000 × g for 30 min, and the soluble fraction was incubated with Ni-NTA resin (Qiagen; Valencia, CA) at 4°C for 2 hr. The resin was washed with a buffer containing 50 mM Hepes pH 7.4, 600 mM KCl, 10 mM Imidazole and 1% β-OG. Nucleic acid contaminants were cleared with benzonase treatment of the resin using 40 units of benzonase per milliliter of solution. The Syt1 fragment was eluted using the washing buffer supplemented with 250 mM imidazole and further purified with size-exclusion chromatography on a Superdex 200 16/60 column using 20 mM Hepes pH 7.4 containing 600 mM KCl and 1% β-OG as the buffer.

To express rat Munc13-1 fragments encoding the C1C2BMUN and C1C2BMUNC2C regions (residues 529–1531 and 529–1735, respectively, both with residues 1408–1452 from a flexible loop replaced by the sequence EF), the corresponding DNA sequences originating from full-length rat Munc13-1 (Basu et al., 2005) were cloned into the pFastBac vector (the vector was modified by adding a GST tag and a TEV cleavage site in front of the EcoRI cloning site). The construct was used to generate a baculovirus using the Bac-to-Bac system (Invitrogen). Insect cells (sf9) were infected with the baculovirus, harvested about 72–96 hr post-infection, and re-suspended in lysis buffer (50 mM Tris pH8.0, 250 mM NaCl, 1 mM TCEP). Cells were lysed and centrifuged at 18,000 rpm for 45 min, and the clear supernatant was incubated with GST agarose at room temperature for 2 hr. The beads were washed with: i) lysis buffer; ii) lysis buffer containing 1% TX-100; iii) lysis buffer containing 1M NaCl; and iv) lysis buffer. The protein was treated with TEV protease on the GST agarose at 22°C for 2 hr. The protein was further purified by ion exchange chromatography and gel filtration, and was concentrated to 1–4 mg/ml for storage in 10 mM Tris buffer (pH 8.0) containing 10% glycerol, 5 mM TCEP and 250 mM NaCl. The C1C2BMUNC2C D705N,D711N mutant was generated by site-directed mutagenesis, and purified as the WT fragment.

Liposome co-floatation assays

Lipids mixture containing 37.5% POPC, 18% POPE, 20% DOPS, 2% PIP2, 2% DAG, 20% Cholesterol and 0.5% Rhodamine-PE were dried in glass tubes with nitrogen gas and kept under vacuum overnight. Lipid films were re-suspended in buffer (25 mM HEPES, pH 7.4, 150 mM KCl, 10% glycerol (v/v)) and vortexed for 5 min. The re-suspended lipid films were frozen and thawed for five times, then extruded through a 80 nm polycarbonate filter with an Avanti extruder for at least 19 times and the size of the liposomes was analyzed by DLS. Liposome solutions containing 2 mM lipids were incubated with 1 µM Munc13-1 fragments at room temperature for 1 hr. The liposomes and bound proteins were isolated by a co-floatation assay on a Histodenz density gradient (40%:35%:30%) as described previously (Guan et al., 2008). Samples from the top of the gradient were taken and analyzed by SDS-PAGE and Coomassie blue staining.

Lipid mixing assays monitored by NBD-fluorescence de-quenching

These lipid mixing assays were performed as described in (Ma et al., 2013) with some modifications. Donor liposomes with synaptobrevin (V) contained 40% POPC, 20% DOPS, 17% POPE, 20% Cholesterol, 1.5% NBD PE, and 1.5% Liss Rhod PE. Donor liposomes with both synaptobrevin and Synaptotagmin-1 (VSyt1) contained 40% POPC, 6.8% DOPS, 30.2% POPE, 20% Cholesterol, 1.5% NBD PE, and 1.5% Liss Rhod PE. Acceptor liposomes with syntaxin-1-SNAP-25 (T) contained 38% POPC, 18% DOPS, 20% POPE, 20% Cholesterol, 2% PIP2 and 2% DAG. Lipid mixtures were dried in glass tubes with nitrogen gas and kept under vacuum overnight. Lipid films were re-suspended and dissolved in buffer (25 mM HEPES, pH 7.4, 150 mM KCl, 0.5 mM TCEP, 10% glycerol (v/v)) with 1% β-OG. Purified SNARE proteins containing 1% β-OG were added to liposomes to make Syx:SNAP25:Lipid ratios = 1:5:800 for T-liposomes, Syb:Lipid = 1:500 for V-liposomes and Syt1:Syb:Lipid = 1:2:1000 for VSyt1 liposome. The mixtures were incubated at room temperature for 30 min and dialyzed against the reaction buffer (25 mM HEPES, pH 7.4, 150 mM KCl, 0.5 mM TCEP, 10% glycerol (v/v)) with 1 g/L Biobeads SM2 (Bio-Rad; Hercules, CA) 3 times.

For lipid mixing assays, donor liposomes (0.125 mM lipids) were mixed with acceptor liposomes (0.25 mM lipids) with various additions of the other proteins in a total volume of 200 μl. For experiments with NSF-αSNAP, acceptor liposomes were first incubated with 0.8 μM NSF, 2 μM αSNAP, 2.5 mM MgCl2, 2 mM ATP, 0.1 mM EGTA and 1 μM Munc18-1 at 37°C for 25 min, and then mixed with donor liposomes and 0.5 μM Munc-13 fragments, 1 μM C2AB fragment, 1 µM excess SNAP-25, and 0.1 mM EGTA (this amount of EGTA is critical to prevent effects from residual Ca2+ co-purified with the Munc13-1 fragments and is low enough to preserve the Zn2+-binding sites of the C1 domain). To measure the effects of Ca2+, 0.6 mM Ca2+ was added after 300 s of the start of the reaction. The NBD-PE fluorescence probe was excited at 460 nm and the emission signal from NBD was monitored at 538 nm with a PTI Spectrofluorometer (Edison, NJ). All experiments were performed at 37°C. At the end of each reaction, 1% w/v β-OG was added to solubilize the liposomes, and all the data were normalized to the maximum fluorescence signal achieved after addition of β-OG. All the experiments were repeated at least three times with a given preparation and the results were verified in multiple experiments performed with different preparations. For quantification, we calculated the average fluorescence at 500 s, expressed as percentage of the maximum fluorescence, and the corresponding standard deviation.

Simultaneous lipid mixing and content mixing assays

Assays that simultaneously measured lipid mixing from de-quenching of the fluorescence of Marina Blue-labeled lipids and content mixing from the development of FRET between PhycoE-Biotin trapped in the T-liposomes and Cy5-Streptavidin trapped in the V-liposomes were performed as described (Zucchi and Zick, 2011; Zick and Wickner, 2014) with some modifications. V-liposomes with synaptobrevin contained 39% POPC, 19% DOPS, 19% POPE, 20% Cholesterol, 1.5% NBD PE, and 1.5% Marine Blue PE. VSyt1-liposomes with both synaptobrevin and Synaptotagmin-1 contained 40% POPC, 6.8% DOPS, 30.2% POPE, 20% Cholesterol, 1.5% NBD PE, and 1.5% Marine Blue PE. T-liposomes with syntaxin-1-SNAP-25 contained 38% POPC, 18% DOPS, 20% POPE, 20% Cholesterol, 2% PIP2 and 2% DAG. Lipid mixtures were dried in glass tubes with nitrogen gas and under vacuum overnight. Lipid films were re-suspended and hydrated in buffer (25 mM HEPES, pH 7.4, 150 mM KCl, 0.5 mM TCEP, 10% glycerol (v/v)) with 1% β-OG by vortex and sonication. Purified SNARE proteins and fluorescence labeled proteins were added to lipid mixtures to make Syx:SNAP25:Lipid = 1:5:800 and PhycoE-Biotin 4 µM for T-liposomes; Syb:Lipids = 1:500 and Cy5-Streptavidin 8 µM for V-liposomea; and Syt1:Syb:Lipids = 1:2:1000 and Cy5-Streptavidin 8 µM for VSyt1 liposomes. The mixtures were incubated at room temperature for 30 min and dialyzed against the reaction buffer (25 mM HEPES, pH 7.4, 150 mM KCl, 0.5 mM TCEP, 10% glycerol (v/v)) with 1g/L Biobeads SM2 (Bio-Rad) 3 times at 4°C. The proteoliposomes were purified by floatation on a three-layer histodenz gradient (35%, 25%, and 0%) and harvesting from the topmost interface.

To simultaneously measure lipid mixing and content mixing, T-liposomes (0.25 mM lipids) were mixed with V-liposomes (0.125 mM lipids) with various additions in a total volume of 200 μl under analogous conditions as those described above for lipid mixing assays, including 100 μM EGTA. All experiments were performed at 30°C, and 0.6 mM Ca2+ was added at 300 s. The fluorescence signals from Marine Blue (excitation at 370 nm, emission at 465 nm) and Cy5 (excitation at 565 nm, emission at 670 nm) were recorded to monitor lipid and content mixing, respectively. At the end of each reaction, 1% w/v β-OG was added to solubilize the liposomes, and the lipid mixing data were normalized to the maximum fluorescence signal achieved after addition of β-OG. To measure content mixing without interference from vesicle leakiness, most experiments were performed in the presence of 5 μM streptavidin. Control experiments without streptavidin were performed to measure the maximum Cy5 fluorescence attainable upon detergent addition. However, there was a large variability in the maximum Cy5 fluorescence values observed, perhaps because of binding of the dye to the detergent. Since the average maximum values observed in detergent were similar to the maximum Cy5 fluorescence observed at the end of the most efficient fusion reactions (those including the Munc13-1 C1C2BMUNC2C fragment, e.g. Figure 9B, blue trace) and the latter was more reproducible, in practice we used averaged maximum values of these reactions for normalization. All the experiments were repeated at least three times with a given preparation and the results were verified in multiple experiments performed with different preparations. For quantification, we calculated the average fluorescence at 500 s, expressed as percentage of the maximum fluorescence, and the corresponding standard deviation.

Measuring the clustering activity of Munc13-1 fragments by DLS

The clustering activity of Munc13-1 fragments was measured by DLS using a DynaPro instrument (Wyatt Technology; Santa Barbara, CA) basically as described (Arac et al., 2006). For experiments with liposomes and Munc13-1 fragments, the conditions and liposome compositions were similar to those of the liposomes used for lipid mixing assays unless specifically indicated. Thus, 0.5 μM Munc13-1 fragments were mixed with T-liposomes (0.25 mM lipids) and/or V-liposomes (0.125 mM lipids) and incubated in a buffer containing 25 mM Hepes (pH 7.4), 150 mM KCl, 10% (v/v) glycerol, 0.5 mM TCEP, 2.5 mM MgCl2, 0.1 mM EGTA at 25°C. For titrations with Munc13-1 fragments (Figure 6—figure supplement 1), different concentrations of the fragments were mixed with T-liposomes (0.25 mM lipids) and V-liposomes (0.125 mM lipids) and incubated in the same buffer at 30°C. For experiments to monitor particle size in parallel with lipid and content mixing, lipid and content mixing assays were performed at 30°C under the standard conditions described above but with all protein and liposome concentrations diluted 8-fold, and identical samples were analyzed by DLS as a function of time at 30°C.

Lentiviral constructs

The cDNAs of Munc13-1 full length and C-terminal fragments (C1C2BMUNC2C, aa529-1693; C1C2BMUN, aa529-1508; MUN domain, aa859-1508; and MUNC2C, aa859-1693) were generated from rat Munc13-1 (Basu et al., 2005) by PCR amplification. The reverse primer harbors a 3xFLAG sequence (Sigma-Aldrich) to allow expression analysis. The corresponding PCR products were fused to a P2A linker (Kim et al., 2011) after a nuclear localized GFP sequence into the lentiviral shuttle vector, which allows a bicistronic expression of NLS-GFP and the Munc13-1-Flag protein/fragment under the control of a human SYNAPSIN1 promoter. Concentrated lentiviral particles were prepared as described (Lois et al., 2002).

Autaptic hippocampal neuronal cultures and lentiviral infection

Animal welfare committees of Charité Medical University and the Berlin state government Agency for Health and Social Services approved all protocols for animal maintenance and experiments (license no. T 0220/09). Astrocytes were plated at a density of 5000 cells/cm2 onto microdots coated coverslips to allow them to grow onto the growth permissive substrate. Hippocampi were dissected from embryonic day 18.5 Munc13 1/2 DKO mouse and enzymatically treated with 25 units ml-1 of papain for 45 min at 37°C. After enzyme digestion, hippocampi were mechanically dissociated and the neuron suspension was plated onto the astrocytes microislands at a final density of 300 cells cm-2. Neurons were incubated at 37°C and 5% CO2 for 13–16 days to mature before starting the experiments; 24 hr after plating, neurons were infected with lentiviral rescue constructs per 35 mm diameter well.

Electrophysiology

Synaptic function was assayed by whole-cell voltage clamp. Synaptic currents were monitored using a Multiclamp 700B amplifier (Axon instrument). The series resistance was compensated by 70% and only cells with series resistances <10 MΩ were analyzed. Data were acquired using Clampex 10 software (Axon instrument) at 10 kHz and filtered using a low-pass Bessel filter at 3 kHz. Recordings were done at room temperature in autaptic hippocampal Munc13- 1/2 DKO neurons at 13–16 days in vitro (DIV). Borosilicate glass pipettes with a resistance between 2–3.5 MΩ were used. Internal pipette recording solution contained the following (in mM): 136 KCl, 17.8 HEPES, 1 EGTA, 4.6 MgCl2, 4 Na2ATP, 0.3 Na2GTP, 12 creatine phosphate, and 50 Uml-1 phosphocreatine kinase; 300 mOsm; pH 7.4. Neurons were continuously perfused with standard extracellular solution including the following (in mM): 140 NaCl, 2.4 KCl, 10 HEPES, 10 glucose, 2 CaCl2, 4 MgCl2; 300 mOsm; pH 7.4. Action potential-evoked EPSCs were triggered by 2 ms somatic depolarization from −70 to 0 mV. To determine the size of the readily-releasable pool (RRP), hypertonic solution, 500 mM sucrose added to standard extracellular solution, was applied directly onto the isolated neuron for 5s using a fast application system. A transient inward current component that lasts for 2–3 s represents the RRP charge (Rosenmund and Stevens, 1996).

Western blot

Hippocampal neurons from E18.5 Munc13-1/2 DKO at a density of 10.000 / cm2 were plated into 6 well plates containing monolayer cultures of astrocytes. Neurons were lysed after 15 DIV at 4°C with 50 mM Tris·HCl, pH 7.9, 150 mM NaCl, 5 mM EDTA, 1% Triton X-100, 1% sodium deoxycholate, 250 μM phenylmethylsulfonyl fluoride, 1% Nonidet P-40, and protease inhibitor cocktail-complete mini (Roche Diagnostics, Berlin, Germany). Lysates were mixed with Laemmli Buffer containing 0.3 mM DTT, and boiled 10 min at 95°C. 30 µg of protein lysates were used for the SDS-PAGE electrophoresis. After separation by SDS-PAGE proteins were transferred to a polyvinyl difluoride (PVDF) membrane. Membranes were blocked with 5% skim milk in TBST, and incubated at 4°C over night with primary antibodies: anti-Flag M2 (F1804; Sigma-Aldrich), and anti-Living Colors GFP (632375; Clontech; Mountain View, CA). Secondary antibodies were horseradish peroxidase-conjugated (Jackson ImmunoResearch; West Grove, PA). The immunoreactive proteins were detected by ECL Plus Western Blotting Detection Reagents (GE Healthcare Biosciences; Pittsburgh, PA) in a Fusion FX7 detection system (Vilber Lourmat, Eberhardzell, Germany).

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
    Munc18-1 binds directly to the neuronal SNARE complex
    1. I Dulubova
    2. M Khvotchev
    3. S Liu
    4. I Huryeva
    5. TC Südhof
    6. J Rizo
    (2007)
    Proceedings of the National Academy of Sciences of the United States of America 104:2697–2702.
    https://doi.org/10.1073/pnas.0611318104
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
    In vitro system capable of differentiating fast Ca2+-triggered content mixing from lipid exchange for mechanistic studies of neurotransmitter release
    1. M Kyoung
    2. A Srivastava
    3. Y Zhang
    4. J Diao
    5. M Vrljic
    6. P Grob
    7. E Nogales
    8. S Chu
    9. AT Brunger
    (2011)
    Proceedings of the National Academy of Sciences of the United States of America, 108, 10.1073/pnas.1107900108.
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
    Sec17 can trigger fusion of trans-SNARE paired membranes without Sec18
    1. M Zick
    2. A Orr
    3. ML Schwartz
    4. AJ Merz
    5. WT Wickner
    (2015)
    Proceedings of the National Academy of Sciences of the United States of America 112:E2290–E2297.
    https://doi.org/10.1073/pnas.1506409112
  74. 74
  75. 75

Decision letter

  1. Axel T Brunger
    Reviewing Editor; Stanford University, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your work entitled "Synergy between Munc13 and Synaptotagmin C2 domains in Synaptic Vesicle Fusion" for consideration by eLife. Your article has been evaluated by Randy Schekman as Senior Editor and three peer reviewers, one of whom is a member of our Board of Reviewing Editors.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission. Major revisions are needed before a final decision can be made.

Summary:

In this work the function of Munc13 is studied using (1) reconstituted bulk liposome content and lipid mixing assays and (2) electrophysiology experiments in neuronal cultures with rescue using Munc13 DKO hippocampal neurons. The results suggest that the Munc13 C2 domains (especially C2B and C2C) play key roles in priming as demonstrated by Figure 9 (reconstitution experiments) and 10 (neuronal culture experiments). It is also suggested that the Munc13 C2 domains act synergistically together with those of synaptotagmin upon Ca2+-triggering. The study extends a previous one by the authors (Ma et al., 2013) in fundamental ways.

Since fusion reactions were monitored in bulk, the effects on fusion efficiency cannot be separated from effects on vesicle docking, posing questions on the proposed role on intrinsic fusion mechanisms (points 2, 3, 6 and 9). Moreover, the absence of complexin is a limitation in their reconstituted experiments and may have influenced the effect of Munc13 in their in vitro experiments. While the experiments presented in this work expand on previous work on the role of Munc13 on synaptic vesicle priming and plasticity, the suggested role as acting in "synergy" with the synaptotagmin C2 domains is less clear and, ideally, should be further tested by additional experiments as outlined below (points 1, 2, and 3). The role of Munc13 in fusion per se is interesting, but the available data are still a bit too inconclusive to make such a strong statement – even in the title. For example, the DD/NN mutation in C2B impairs liposome fusion, but corresponding mutations do not affect evoked EPSCs. It might be prudent to discuss these data a bit more carefully.

The manuscript is rather cumbersome to understand and there are rather different results obtained with the soluble Syt2 C2AB domain vs. full-length Syt1. Thus the work would benefit from elimination of all results with the soluble Syt1 C2 AB domain (points 5, 8, 9), also considering that full-length Syt1 is more biologically relevant.

Essential revisions:

1) Figure 9 suggests that the Ca2+ binding region of the Munc13 C2B is important for enhancing Ca2+ triggered content/lipid mixing in a bulk assay with SNAREs, full-length Syt1, αSNAP, NSF, and C1 C2BMUNC2C (the figure shows that the DN mutant of the Munc13 C2B domain has a reduced Ca2+ triggered lipid/content mixing effect). This result is at odds with the results by Shin et al. (2010) using rescue experiments with Munc13-2 in cultured hippocampal neuronal cultures derived from Munc13-1/2 DKO mice. The experiments of Shin et al. revealed no effect of the DN mutant on isolated EPSCs, and on the RRP. However, Shin et al. observed an effect on short term plasticity that was explained by an induced increase in PIP2 concentration upon high frequency stimulation. This discrepancy between Figure 9 and the EPSC/RRP results by Shin et al. needs to be discussed, and follow up experiments may be necessary (see next point).

2) As is apparent from both the lipid and content mixing traces in Figure 9, spontaneous fusion reactions occur at a substantial rate before Ca2+ is injected, i.e., the system has not reached a steady state level at the time point of Ca2+ injection. Since in their bulk assay, t- and v-vesicles may continuously undergo docking, this produces opportunities for both lipid and content mixing to occur as a consequence of an increasing number of docked vesicles. Upon injection of Ca2+, the observed burst of both content and lipid mixing signals could be a consequence of an increased probably of vesicle docking upon Ca2+ binding to Munc13-C2B rather than an increase of intrinsic fusion probability of already docked vesicle pairs. Thus, it would important to perform clustering experiments plus/minus Ca2+ with wildtype C1 C2BMUNC2C and the DN mutant under the same conditions as Figure 9 in order to assess the potential effect of the Ca2+ bound C2B domain of Munc13 on vesicle clustering.

3) Figure 6 suggests that C1 C2BMUNC2C greatly enhances both content and lipid mixing in the absence of Ca2+ together with SNAREs, compared to just SNAREs alone. However, this result could also be a consequence of enhanced docking by C1 C2BMUNC2C in the absence of Ca2+, rather than an effect on intrinsic spontaneous fusion itself. To assess this question, clustering experiments should be attempted with exactly the same conditions and constructs to assess if C1C2MUNC2C enhances vesicle clustering.

4) Figure 7 suggests that injection of Ca2+ does not induce any increase in lipid/content mixing in presence of SNAREs and full length Syt1 (orange curve, i.e., no Munc18, Munc13, αSNAP/NSF). This is at odds with previous reconstitution experiments that showed an increase of both properties with SNAREs and Syt1 alone upon Ca2+ injection (Malsam et al., EMBO J 2012; Lai et al., eLife 2014). An explanation is needed.

5) The effect of Munc13 in spontaneous (Ca2+ independent) and Ca2+-triggered fusion are very different when Syt1-C2AB fragment or reconstituted full-length Syt1 is used (compare Figures 5 and 7), indicating that Syt1-C2AB behaves quite differently from full-length Syt1 in their reconstituted system. The use of reconstituted full-length Syt1 is more physiological. For the benefit of a clearer presentation of the results, all experiments with Syt1 C2AB could be eliminated from the paper.

6) Figures 6 and 7 show the effect of C1C2BMUN on lipid mixing in the absence or presence of NSF/SNAP and Munc18-1, by the different effects of calcium in this context. The authors suggest that different mechanisms are at play, but there may be other possibilities. The calcium independent lipid mixing effects of C1C2BMUN are thought to reflect support of SNARE complex assembly by C1C2BMUN. In the presence of NSF/SNAP and Munc18-1 there are at least two explanations: (1) NSF/SNAP reverts the C1C2BMUN-mediated SNARE complex assembly in the absence of calcium, while in the presence of calcium more C1C2BMUN is recruited to the liposomes and can then override the antagonistic effect of NSF/SNAP. (2) The membrane recruited C1C2BMUN even protects assembled SNARE complexes from NSF/SNAP.

7) Surprisingly, Syt1 does not appear to be essential to control either the speed or the calcium-sensitivity of the reaction when the longest forms of Munc13 are included in their assay (Figure 5). Thus, in this simplified system, the calcium-switch is encoded entirely within the C2 domains of Munc13 at a calcium sensitivity somewhere below 100 μM. Establishing this limit with some precision would help ascertain how the activity of this system compares to the resting state in the synapse. The authors also speculate that Syt1-dependence may become apparent if they had faster time resolution in their assay. Given the central role of this calcium-sensor in synaptic biology, establishing whether it has any activity here is essential. However, as pointed out in (5) above, the results with the soluble Syt1 C2AB domain are rather different from the results with full-length Syt1. Thus, it might be prudent to eliminate all results with the Syt1 C2AB domain and focus on full-length reconstituted Syt1 throughout this work.

8) When shorter versions of Munc13 are tested where the C2C domain is missing, the authors now see Syt1-dependent effects on membrane fusion. In particular, they make the important point that lipid-mixing proceeds "efficiently" without Syt1, while contents-mixing is strongly enhanced by Syt1 C2AB. However, in Figure 4, it is not obvious that the effect of Syt1 is more pronounced for content vs lipid-mixing. Although the change in the curves for content-mixing appear more dramatic than for lipid-mixing, this seems to be due to the scale of the axes and not necessarily to the rates. The initial rate changes about 3 fold for contents mixing (comparing B and D). A similar increase in lipid mixing would be hard to see at their resolution, but looks plausible when comparing A and C. A similar burst in lipid-mixing appears to be present in Figure 5C. Without meaningful resolution over this range, it is challenging to draw the conclusion "the finding that Syt1 C2AB selectively enhances content mixing but not lipid mixing provides strong evidence that Syt1 plays a direct role in membrane fusion…". The same point is echoed strongly in the Discussion. At a minimum, it would be helpful if the authors provided initial rates from each experiment, rather than endpoints. Otherwise, the discussion surrounding whether Syt1 is uniquely designed to support full-fusion should probably be more guarded. However, all these results were generated with the soluble Syt1 C2AB, so again, it might be better to eliminate all results with the Syt1 C2AB domain from this work.

9) In the Introduction, the authors write "…the mechanism by which Munc13s play an additional role in vesicle docking… is unclear." The papers cited here do actually propose a mechanism for the role of vesicle docking by Munc13s, i.e. docking/membrane attachment, priming, and SNARE complex assembly may be manifestations of the same process, that is, that Munc13s mediate vesicle priming by promoting SNARE complex assembly (i.e., by opening the Munc18-syntaxin complex and catalyzing trans SNARE complex formation). The data provided in these papers are quite compelling. The three papers should be cited separately, including the corresponding conclusions, and then explain why the authors think that these papers are not helpful or believable in defining (part of) the mechanism by which Munc13 mediates vesicle docking. Somewhat in the same context, can one use the liposome-clustering activity of proteins and protein fragments as a basis to extrapolate towards corresponding roles of the respective proteins in synaptic vesicle docking in neurons? For example, in the case of Syt1, such a conclusion can probably not be drawn – loss of Syt1 does not really affect synaptic vesicle docking much.

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

Thank you for resubmitting your work entitled "Functional Synergy between the Munc13 C-terminal C1 and C2 domains" for further consideration at eLife. Your revised article has been favorably evaluated by Randy Schekman (Senior editor) and three reviewers, one of whom is a member of our Board of Reviewing Editors.

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

The reviewers would like to thank the authors for addressing the concerns raised previously. Based on their new findings, the authors have changed the conclusions from their studies. Rather than suggesting a synergism between the synaptotagmin and Munc13 C2 domains, they now suggest synergism between the C2B and C2C domains of Munc13-1 and that Munc13-1 may bridge V- and T-liposomes in a calcium independent manner (Figure 6—figure supplement 5). The authors' new data correlating the activation of SNARE fusion with (at least in part) a role in the specific clustering/docking of V- and T-liposomes by Munc13 significantly clarifies the primary message of the paper and is a result that will be of general interest.

Comments:

1) The manuscript remains very long. Considering the revised conclusions, the authors are encouraged to take a fresh look and examine if all the data are really essential for the key conclusions.

2) There are some supplementary figures that are actually quite important (e.g., Figure 4—figure supplement 4, and Figure 6—figure supplement 5). Please restrict supplementary figures to repeat experiments, raw data, and non-essential experiments.

3) A general comment regarding all figures that compare lipid and content mixing results (Figures 4, 5, 7, 8, 9 and numerous supplemental figures): both types of experiments (lipid mixing and content mixing, respectively) are normalized differently: for lipid mixing 1% BOG is added and the resulting fluorescence intensity value is used for normalization; whereas for content mixing, it is the maximum value for each group of conditions in the particular panel. So, 100% content mixing could actually correspond to substantially fewer vesicles undergoing content mixing than vesicles undergoing lipid mixing. A suggestion would be to change all figure panel titles from "lipid mixing" to "relative lipid mixing" and "content mixing" to "relative content mixing", and to change all the y-axis labels from "Fluorescence (% of max)" to "Fluorescence (arbitrary units)". Another comment: ensemble experiments cannot determine what fraction of vesicles that undergo lipid mixing also undergo content mixing.

4) Figure 4—figure supplement 4C. We appreciate that the authors have performed DLS measurements to address the concern that the large increase in content mixing upon calcium addition could be due to a calcium-dependent docking/clustering effect. However the presence of large vesicles might dominate in the intensity autocorrelation functions shown. Percent Intensity plots (similar to those shown in Figures 2 and 6) should also be provided for Figure 4—figure supplement 4C.

5) Figure 4—figure supplement 4C was only done for the soluble Syt1 C2AB fragment. If at all possible, please also perform the DLS experiments also with full-length Syt1. The reason is that there is much less lipid mixing in these experiments prior to calcium addition than for the experiments with the soluble Syt1 C2AB fragment (compare Figures 7 and 8).

6) Figure 9: It is perhaps possible that there could be compensating factors in neurons that do not show a large phenotype for the DN mutant of Munc13. Nevertheless, if the author's model of calcium-independent bridging of T and V-vesicles by C1C2BMUNC2C is correct (Figure 6—figure supplement 5), why would calcium binding to the C2B domain of Munc13 have such a large effect? Please discuss.

7) Figure 6—figure supplement 5: While the preference for the C1C2B domain for the T-vesicle membrane can be explained by the presence of DAG and PIP2 in that membrane, what is the mechanism that C2C would only interact with the V-vesicle membrane?

8) Figures 2 and 6: why does C1C2BMUN cluster V-vesicles, whereas C1C2BMUNC2C does not? This question is also related to the Discussion section, paragraph five. Please provide an explanation why the C2C domain would favor trans-interactions with V-vesicles.

9) Subsection “Simultaneous evaluation of lipid and content mixing”, paragraph two": "A problem…". Actually, synaptic vesicles are acidic, so the property of sulforhodamine to produce an acidic interior of the proteoliposomes is not necessarily a problem. Moreover, the lack of leakiness under such conditions is desirable. In any case, it is good that the authors confirmed the content mixing results with the different content mixing assay by Zucchi and Zick. Thus, there is really no "problem" with the sulforhodamine content mixing assay.

10) Figures 1D and E: the effect of the different lipid compositions (PIP2 and DAG) appears to be much more pronounced in the presence of NSF/SNAP/Munc18, although C1C2BMUN is present in both experiments.

11) Discussion, paragraph seven: in addition to Lee and Kyoung, please also cite the improvements of this assay by Diao et al. (eLife 2012); Lai et al. (eLife 2014). Another comment: the single vesicle content mixing assay by Kyoung et al. (2011); Lai et al. (2014) allows discrimination between effects that are due to docking vs. due to fusion probabilities in addition to the improved time resolution offered by such single vesicle assays.

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

Thank you for resubmitting your work entitled "Functional Synergy between the Munc13 C-terminal C1 and C2 domains" for further consideration at eLife. Your revised article has been favorably evaluated by Randy Schekman (Senior editor) and a Reviewing Editor.

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

We thank the authors for their thoughtful considerations of the concerns raised previously. However, the Reviewing Editor has two remaining concerns:

1) Related to Point 9, the authors write:

"A potential concern with the use of sulforhodamine B fluorescence de-quenching to monitor content mixing in these bulk reconstitution experiments is that de-quenching can also arise from liposome leakiness. We attempted to perform experiments where both liposome populations are loaded with sulforhodamine B to assess how much of the de-quenching arises from leakiness (Yu et al., 2013). However, we were unable to observe lipid mixing in these control experiments. Since the solutions of high sulforhodamine B concentrations trapped in the liposomes are highly acidic (pH about 2), which may underlie the lack of lipid mixing, we attempted to increase the pH of these solutions to 5.5 or 7.4 before trapping them into the liposomes, but at both pH values the liposomes were strongly leaky."

Their observation of lack of lipid mixing when both vesicle populations are filled with sulforhodamine B is strange, considering that the authors observed efficient lipid mixing when only one class of vesicles is filled (Figure 3—figure supplement 1). Moreover, Kyoung et al., 2011 and Diao et al. 2012 observed both lipid and content mixing using their sulforhodamine B content mixing method, with only an extremely small fraction of leakage (0.01% of the docked vesicles). Could the problems be related to differences in reconstitution methods, rather than an issue with the sulforhodamine method per se?

This Reviewing Editor also would like to point out that the pH of the interior of the synaptobrevin/synaptotagmin vesicles ("donor" vesicles) is likely more neutral rather than pH 2 when using the published reconstitution protocol (Kyoung et al. Nat Protoc 8(1):1-16, 2013): the initial protein/lipid/detergent solution uses a detergent concentration that is at the CMC (OG ~ 30 mM), so vesicles do not yet form at this stage. This solution is then injected on a Cl-4B column that has been pre-equilibrated with vesicle buffer (pH 7.4, without sulforhodamine B), followed by elution with vesicle buffer (pH 7.4, again without sulforhodamine B) (Kyoung et al. Nat Protoc 8(1):1-16, 2013). Vesicles are forming on the column in a time course of several minutes as the detergent concentration decreases. One would expect that the pH of the interior of the column rapidly reaches 7.4.

In any case, as the authors point out, the alternative PhycoE-Biotin/Cy5-Streptavidin content mixing assay produces similar results (Figure 3—figure supplement 3). However, as written, the second paragraph of subsection “Simultaneous evaluation of lipid and content mixing” may come across to the casual reader as if there is something wrong with the sulforhodamine B content mixing method. Since this is a rather technical issue that distracts from the main messages of this paper, and that may be related to differences in reconstitution methods, the authors may wish to consider deleting this entire paragraph. In the future, perhaps the authors may wish to investigate a variety of content mixing assay jointly with other groups who have developed these assays for a more technical publication.

2) Point 11, related to the sentence starting "Approaches that allow faster measurements [e.g. single vesicle assays (Lee et al., 2010; Kyoung et al., 2011); will be required to test this prediction. " It is not just the faster time resolution that is important for future studies. The single vesicle-vesicle lipid/content mixing assay by Kyoung et al. 2011, Diao et al., 2012 et al. (but not the assay by Lee et al., 2010) can distinguish between effects due to docking, lipid mixing and content mixing on a 100 msec time scale for each individual vesicle pair. In addition to the improved time resolution, the discrimination between effects due to docking and/or fusion will be important for future studies. The authors may wish to consider making the discussion of the relevant publications clearer.

https://doi.org/10.7554/eLife.13696.030

Author response

Summary: In this work the function of Munc13 is studied using (1) reconstituted bulk liposome content and lipid mixing assays and (2) electrophysiology experiments in neuronal cultures with rescue using Munc13 DKO hippocampal neurons. The results suggest that the Munc13 C2 domains (especially C2B and C2C) play key roles in priming as demonstrated by Figure 9 (reconstitution experiments) and 10 (neuronal culture experiments). It is also suggested that the Munc13 C2 domains act synergistically together with those of synaptotagmin upon Ca2+-triggering. The study extends a previous one by the authors (Ma et al., 2013) in fundamental ways. Since fusion reactions were monitored in bulk, the effects on fusion efficiency cannot be separated from effects on vesicle docking, posing questions on the proposed role on intrinsic fusion mechanisms (points 2, 3, 6 and 9). Moreover, the absence of complexin is a limitation in their reconstituted experiments and may have influenced the effect of Munc13 in their in vitro experiments. While the experiments presented in this work expand on previous work on the role of Munc13 on synaptic vesicle priming and plasticity, the suggested role as acting in "synergy" with the synaptotagmin C2 domains is less clear and, ideally, should be further tested by additional experiments as outlined below (points 1, 2, and 3). The role of Munc13 in fusion per se is interesting, but the available data are still a bit too inconclusive to make such a strong statement – even in the title. For example, the DD/NN mutation in C2B impairs liposome fusion, but corresponding mutations do not affect evoked EPSCs. It might be prudent to discuss these data a bit more carefully.

The manuscript is rather cumbersome to understand and there are rather different results obtained with the soluble Syt2 C2AB domain vs. full-length Syt1. Thus the work would benefit from elimination of all results with the soluble Syt1 C2AB domain (points 5, 8, 9), also considering that full-length Syt1 is more biologically relevant.

Please note that we have included a new figure 6 and several new supplementary figures, and we have renumbered some of the figures: the previous Figure 3 is now Figure 3—figure supplement 1; and the previous figures 4, 5 and 6 are now figures 3, 4 and 5, respectively.

A) We appreciate the assessment that our paper extends the previous Ma et al. 2013 paper in fundamental ways. We also feel indebted to the reviewers because their thoughtful and detailed comments have helped to make the paper of higher quality and more fundamental (see in particular point B below).

B) We agree that it was premature to draw the strong conclusions on a role of the Munc13-1 C2 domains in fusion that we had described in the original manuscript, and we found particularly valuable the concerns from the reviewers regarding the possibility that at least some of the effects observed in the fusion assays arise from effects on clustering. While in the original manuscript we did conclude that the tendency of the Munc13-1 fragments to cluster vesicles contribute to their activity in the fusion assays, we did not make this a major point of the paper because the clustering data did not appear to explain the much stronger stimulation of fusion by C1C2BMUNC2C compared to C1C2BMUN nor the dramatic stimulation of fusion by Ca2+ in the reconstitutions that included Munc18-1, NSF-αSNAP and Munc13-1 fragments. We had performed a detailed analysis of the factors that determine the clustering ability of C1C2BMUN (shown in Figure 2) and had not seen any strong effects of Ca2+ on clustering (shown in new Figure 2—figure supplement 1). We also felt that it was unlikely that the strong effects of the C2C domain in the fusion assays could arise from the slightly higher tendency of C1C2BMUNC2C to cluster mixtures of V- and T-liposomes, compared to C1C2BMUN (previous Figure 2—figure supplement 1 that is now replaced with new data that are shown in Figure 6—figure supplement 1 and was performed under conditions analogous to those of the fusion assays).

We thought that the C2C domain was simply adding an additional lipid-binding site to the C1C2 BMUN sequence but there was an important property that we had not anticipated. Because of the overall concern of the reviewers on the clustering issue, we investigated the clustering properties of C1C2BMUNC2C in more detail and found that C1C2BMUNC2C was unable to cluster V-liposomes in the absence of Ca2+ (new Figure 6A and new Figure 6—figure supplement 3), in contrast to C1C2BMUN (Figure 2C). Importantly however, C1C2BMUNC2C does cluster all liposomes in mixtures of V-liposomes and T-liposomes (new Figure 6G,H), showing that C1C2BMUNC2C favors bridging of V-liposomes to T-liposomes. These results and the overall data presented in the paper have led us to emphasize in the revised manuscript the notion that Munc13-1 may bridge synaptic vesicles to the plasma membrane to contribute to docking and facilitate SNARE complex formation. We believe that this notion makes the paper stronger and with a clearer message that makes a lot of sense because the highly conserved C-terminal region of Munc13s is formed by an elongated MUN domain related to tethering factors that function in diverse membrane compartments, and by adjacent C1 and C2 domains with demonstrated or potential membrane-binding properties.

Because of these new data, and in response to the additional reviewer concerns described in the summary above, in the revised paper we have changed the title and we have drastically toned-down conclusions about a role for the Munc13 C2 domains in membrane fusion and about synergy with the synaptotagmin C2 domains. However, because it seems unlikely that clustering alone can explain the strong stimulation of fusion by Ca2+ (see point 2), we still mention that our data suggest the formation of a primed state that needs Ca2+ to trigger fast fusion and we make the suggestion that Munc13 may function in docking, priming and fusion. We hope that the reviewers will agree that our overall data support this suggestion even though do not fully prove it.

C) With regard to some of the other general concerns, we are well aware that our reconstitutions have their limitations, which we point out in the Discussion. However, we would like to emphasize that reconstitution experiments are not physiological by definition. The hope is that by including enough components we can get results that recapitulate at least to a certain extent what happens in vivo. Arguments commonly used to reach such conclusion include: i) there are clear correlations between the reconstitution data and physiological results; and ii) the biochemical properties underlying the reconstitution results make overall sense in the context of what is known about this system from research by a variety of approaches. Thus, many of the papers in the field that used reconstitutions have provided important insights even though they had striking contradictions with physiological data, most notably the fact that lipid and/or content mixing did not require Munc18-1 and Munc13s. Our reconstitutions are the only ones so far that can account for the dramatic disruption of release observed in the absence of Munc18-1 or Munc13s in vivo (Figure 8D is particularly striking in this regard), they include the (arguably) eight most central components of the release machinery, and there are many correlations between the results obtained with these reconstitutions and physiological data (see Discussion).

For these reasons, we believe that our reconstitutions provide a great framework to try to understand the mechanism of release and that biochemical properties that explain our data (e.g. the V- to T-liposome clustering activity of C1C2BMUNC2C) are likely to underlie at least in part the mechanism of action of the same proteins in vivo. Similarly, if some of our reconstitution data appear to contradict physiological results (e.g. the effects of the DD/NN mutation in the Munc13-1 C2B domain), it is probable that there is an explanation due to functional redundancy or compensatory effects in vivo. In any case, we agree with the reviewers that we should be prudent when drawing conclusions because it is very difficult to have definitive proof for molecular mechanisms. Hence, we have toned down any strong conclusions we made in the manuscript and tried to use soft terms such as ‘suggest’ or ‘indicate’. We have also tried to improve the discussion of the results obtained with the DD/NN mutation but without extending it to avoid further lengthening the manuscript.

D) With regard to the absence of complexin in our experiments, this protein is believed to inhibit release in the absence of Ca2+ and to stimulate release upon Ca2+ influx. Hence, it would not be expected to have much effect in our optimal reconstitutions because there is little fusion before adding Ca2+ and fast fusion upon Ca2+ addition in the time scale of our experiments (e.g. Figure 8D). Indeed, in unpublished experiments we have found that complexin-1 has no measurable effects in our fusion assays, but we prefer not to include these data because we believe that the roles of complexin will be better studied with methods that have faster time scales than our bulk assays (as has already been shown in reconstitution studies of complexin function by the labs of Axel Brunger and Yeon-Kyun Shin). At the end of the Discussion we do point out that our reconstitutions are incomplete and miss several important factors (not only complexin).

E) We have tried to improve the presentation of the results, adding substantial new data without lengthening the manuscript too much. We are reluctant to remove the results obtained with the soluble Syt1 C2AB fragment because these results are important for how the story is developed (e.g. the switch to a different method of monitoring lipid and content mixing) and because, based on the abundant data already available in the literature with this fragment, it is natural that many researchers in the field would wonder what happens if we add this fragment. Note that experiments using full-length Syt1 do have the advantage of being more physiological, as pointed out in the review, but do not allow a direct comparison of the results with and without the Syt1 C2 domains with the same liposome preparations. For instance, there is a clear increase in content mixing when Syt1 C2AB is added in reconstitutions with C1C2BMUN (Figure 4), but it would be more difficult to reach this conclusion comparing results obtained with different V-liposome preparations that include or lack full-length Syt1. This result is important because it shows an effect of the Syt1 C2 domains in this system and it is likely that this effect cannot be observed when we use C1C2BMUNC2C because the system is too efficient (see Discussion). Finally, we would like to emphasize that, while it is true that some results obtained with C2AB and full-length Syt1 are different (mostly in the absence of Ca2+, Munc18-1 and NSF-αSNAP) the data obtained in the presence of Munc18-1, NSF-αSNAP and different Munc13-1 fragments are quite similar for C2AB and full-length Syt1 (compare Figure 4C,D with Figure 8A,B).

Essential revisions:

1) Figure 9 suggests that the Ca2+ binding region of the Munc13 C2B is important for enhancing Ca2+ triggered content/lipid mixing in a bulk assay with SNAREs, full-length Syt1, αSNAP, NSF, and C1C2BMUNC2C (the figure shows that the DN mutant of the Munc13 C2B domain has a reduced Ca2+ triggered lipid/content mixing effect). This result is at odds with the results by Shin et al. (2010) using rescue experiments with Munc13-2 in cultured hippocampal neuronal cultures derived from Munc13-1/2 DKO mice. The experiments of Shin et al. revealed no effect of the DN mutant on isolated EPSCs, and on the RRP. However, Shin et al. observed an effect on short term plasticity that was explained by an induced increase in PIP2 concentration upon high frequency stimulation. This discrepancy between Figure 9 and the EPSC/RRP results by Shin et al. needs to be discussed, and follow up experiments may be necessary (see next point).

This discrepancy was discussed in the original manuscript. We have tried to improve the discussion of this issue, as mentioned in point C of our response to the Summary.

2) As is apparent from both the lipid and content mixing traces in Figure 9, spontaneous fusion reactions occur at a substantial rate before Ca2+ is injected, i.e., the system has not reached a steady state level at the time point of Ca2+ injection. Since in their bulk assay, t- and v-vesicles may continuously undergo docking, this produces opportunities for both lipid and content mixing to occur as a consequence of an increasing number of docked vesicles. Upon injection of Ca2+, the observed burst of both content and lipid mixing signals could be a consequence of an increased probably of vesicle docking upon Ca2+ binding to Munc13-C2B rather than an increase of intrinsic fusion probability of already docked vesicle pairs. Thus, it would important to perform clustering experiments plus/minus Ca2+ with wildtype C1C2BMUNC2C and the DN mutant under the same conditions as Figure 9 in order to assess the potential effect of the Ca2+ bound C2B domain of Munc13 on vesicle clustering.

We have performed the requested clustering assays but with reactions diluted 8-fold to prevent saturation of the DLS detector (Figure 9—figure supplement 2 for experiments with VSyt1 vesicles; Figure 4—figure supplement 4 for experiments with V-vesicles and not Syt1). The immediate effects of Ca2+ addition on the DLS data are minimal and correlate with the observation that, in general, Ca2+ does not have strong effects on clustering ability (except for clustering of V-vesicles by C1C2BMUNC2C, which is unlikely to have any physiological relevance). Hence, it is unlikely that an effect on clustering underlies the drastic stimulation of content mixing induced by Ca2+.

We would like to clarify that there is substantial lipid mixing before Ca2+ addition, but very little content mixing. In Figure 9—figure supplement 1, we now present quantification of lipid mixing at 300 s, before adding Ca2+, in addition to the quantification at 500 s (200 s after Ca2+ addition) that was included previously. In subsection “Functional importance of Ca2+ binding to the Munc13-1 C2B domain” we describe that ‘the fluorescence increase reflecting lipid mixing was 28.9% of that observed 200 s after Ca2+ addition while content mixing was minimal at 300 s (fluorescence 3.3% of that observed 200 s after adding Ca2+)’. We also would like to emphasize that this difference is exacerbated by the fact that the maximal fluorescence associated with content mixing is expected to correspond to only one round of fusion, whereas additional rounds of fusion could contribute to the maximal fluorescence in the lipid mixing signal, which could easily correspond to more than 0.5 rounds of fusion if lipid mixing indeed arose from real fusion (which we do not believe). Hence, there is a dramatic difference in the amounts of lipid mixing and content mixing that occur before Ca2+ addition. Note also that the clear contrast between lipid and content mixing was also observed in the experiments without Syt1 or with Syt1 C2AB, in the presence of Munc18-1, NSF-αSNAP and C1C2BMUNC2C (Figure 4, blue curves). Moreover, the lipid and content mixing curves are much more similar to each other in experiments performed with T- and V-liposomes plus C1C2BMUNC2C (without Munc18-1 and NSF-αSNAP; see new Figure 5A,B, blue curves), showing how content mixing occurs concomitantly with lipid mixing under these conditions but not when NSF-αSNAP and Munc18-1 are present. All of these observations and the fact that there is already extensive clustering before Ca2+ addition provide strong evidence that the drastic effect of Ca2+ in the presence of Munc18-1, NSF-αSNAP and C1C2BMUNC2C does not arise because of increased clustering but rather from a role of the Munc13-1 C2B domain in fusion and/or in the release of some inhibitory interaction. We have de-emphasized the proposal of a role for Munc13-1 in fusion throughout the Abstract and the Discussion, but we hope that the reviewers will agree that we can suggest these two possibilities based on the available data.

3) Figure 6 suggests that C1C2BMUNC2C greatly enhances both content and lipid mixing in the absence of Ca2+ together with SNAREs, compared to just SNAREs alone. However, this result could also be a consequence of enhanced docking by C1C2BMUNC2C in the absence of Ca2+, rather than an effect on intrinsic spontaneous fusion itself. To assess this question, clustering experiments should be attempted with exactly the same conditions and constructs to assess if C1C2MUNC2C enhances vesicle clustering.

We agree with the interpretation proposed by the reviewers, which is supported by the new clustering data that we provide (Figure 6 and several figure 6 supplements).

4) Figure 7 suggests that injection of Ca2+ does not induce any increase in lipid/content mixing in presence of SNAREs and full length Syt1 (orange curve, i.e., no Munc18, Munc13, αSNAP/NSF). This is at odds with previous reconstitution experiments that showed an increase of both properties with SNAREs and Syt1 alone upon Ca2+ injection (Malsam et al., EMBO J 2012; Lai et al., eLife 2014). An explanation is needed.

Comparisons of our data with those obtained with single-vesicle assays such as those of Lai et al. 2014 are difficult because many differences have been observed previously between these assays and bulk assays (for instance in studies of complexins). There is no contradiction with the data of Malsam et al. as can be seen from Figure 1A of this paper. Note that small sudden jumps at the moment of Ca2+ addition are difficult to interpret because they may arise from temporarily high local Ca2+ concentrations when the Ca2+ is added, and that the curve after the jump continues as a natural extension of the curve before Ca2+ addition.

5) The effect of Munc13 in spontaneous (Ca2+ independent) and Ca2+-triggered fusion are very different when Syt1-C2AB fragment or reconstituted full-length Syt1 is used (compare Figures 5 and 7), indicating that Syt1-C2AB behaves quite differently from full-length Syt1 in their reconstituted system. The use of reconstituted full-length Syt1 is more physiological. For the benefit of a clearer presentation of the results, all experiments with Syt1 C2AB could be eliminated from the paper.

See point E of the response to the Summary.

6) Figures 6 and 7 show the effect of C1C2BMUN on lipid mixing in the absence or presence of NSF/SNAP and Munc18-1, by the different effects of calcium in this context. The authors suggest that different mechanisms are at play, but there may be other possibilities. The calcium independent lipid mixing effects of C1C2BMUN are thought to reflect support of SNARE complex assembly by C1C2BMUN. In the presence of NSF/SNAP and Munc18-1 there are at least two explanations: (1) NSF/SNAP reverts the C1C2BMUN-mediated SNARE complex assembly in the absence of calcium, while in the presence of calcium more C1C2BMUN is recruited to the liposomes and can then override the antagonistic effect of NSF/SNAP. (2) The membrane recruited C1C2BMUN even protects assembled SNARE complexes from NSF/SNAP.

We have considerably changed the interpretation of these experiments. We hope that the reviewers will agree with the current interpretation.

7) Surprisingly, Syt1 does not appear to be essential to control either the speed or the calcium-sensitivity of the reaction when the longest forms of Munc13 are included in their assay (Figure 5). Thus, in this simplified system, the calcium-switch is encoded entirely within the C2domains of Munc13 at a calcium sensitivity somewhere below 100 μM. Establishing this limit with some precision would help ascertain how the activity of this system compares to the resting state in the synapse. The authors also speculate that Syt1-dependence may become apparent if they had faster time resolution in their assay. Given the central role of this calcium-sensor in synaptic biology, establishing whether it has any activity here is essential. However, as pointed out in (5) above, the results with the soluble Syt1 C2AB domain are rather different from the results with full-length Syt1. Thus, it might be prudent to eliminate all results with the Syt1 C2AB domain and focus on full-length reconstituted Syt1 throughout this work.

It is true that the Ca2+ switch does come largely from the Munc13-1 C2B domain in these experiments (Figure 9). However, Syt1 C2AB does have an effect on content mixing in experiments with C1C2BMUN (Figure 4) and there also appears to be an effect of full-length Syt1 (compare Figures 4B and 8B), although this comparison is difficult to make because of the use of different proteoliposome populations. In the experiments with C1C2BMUNC2C, we cannot really expect much of an effect by Syt1 because the stimulation of fusion by Ca2+ is so fast in the time scale of our measurements. This is why we believe that techniques with faster time scales and that can distinguish docking from fusion, such as single vesicle assays, are more adequate to dissect the relative contributions of Munc13-1 and Syt1 to fusion.

We would like to clarify that the experiments of Figure 4—figure supplement 3 show that fusion is already maximized with 100 μM Ca2+ and 100 μM EGTA. Hence, the Ca2+ sensitivity is somewhere around 1 μM or below (not 100 μM or below). In the text we explain the difficulty of assessing this sensitivity more accurately because of the danger of removing the Zn2+ ions from the C1 domain. See also see point E of the response to the Summary with regard to the issue of removing data obtained with Syt1 C2AB.

8) When shorter versions of Munc13 are tested where the C2C domain is missing, the authors now see Syt1-dependent effects on membrane fusion. In particular, they make the important point that lipid-mixing proceeds "efficiently" without Syt1, while contents-mixing is strongly enhanced by Syt1 C2AB. However, in Figure 4, it is not obvious that the effect of Syt1 is more pronounced for content vs lipid-mixing. Although the change in the curves for content-mixing appear more dramatic than for lipid-mixing, this seems to be due to the scale of the axes and not necessarily to the rates. The initial rate changes about 3 fold for contents mixing (comparing B and D). A similar increase in lipid mixing would be hard to see at their resolution, but looks plausible when comparing A and C. A similar burst in lipid-mixing appears to be present in Figure 5C. Without meaningful resolution over this range, it is challenging to draw the conclusion on pg 12 "the finding that Syt1 C2AB selectively enhances content mixing but not lipid mixing provides strong evidence that Syt1 plays a direct role in membrane fusion…". The same point is echoed strongly in the Discussion. At a minimum, it would be helpful if the authors provided initial rates from each experiment, rather than endpoints. Otherwise, the discussion surrounding whether Syt1 is uniquely designed to support full-fusion should probably be more guarded. However, all these results were generated with the soluble Syt1 C2AB, so again, it might be better to eliminate all results with the Syt1 C2AB domain from this work.

As pointed out in the response to point 4, we cannot draw conclusions from the small jumps observed upon Ca2+ addition. There are many ways of quantifying the lipid and content mixing assays, and we chose to quantify the data at 500 s because this represents a compromise that reflects in part the initial speed and in part the completion of the reaction at longer times. The quantification provided in Figure 4—figure supplement 1 strongly supports the conclusion that there is indeed an enhancement of content mixing by C2AB in the experiments performed with C1C2BMUN, while there is no clear enhancement of lipid mixing. We have reproduced these results in multiple additional experiments with other preparations, and the same conclusions are reached with the sulfhorhodamine assay (Figure 3—figure supplement 1; previously Figure 3). Nevertheless, we have drastically cut the discussion of these data because this is not a main point of this paper. We still conclude that C2 AB helps with membrane fusion in these experiments and cite other papers that have previously reached similar conclusions regarding the stimulation of fusion by Syt1.

9) In the Introduction, the authors write "…the mechanism by which Munc13s play an additional role in vesicle docking… is unclear." The papers cited here do actually propose a mechanism for the role of vesicle docking by Munc13s, i.e. docking/membrane attachment, priming, and SNARE complex assembly may be manifestations of the same process, that is, that Munc13s mediate vesicle priming by promoting SNARE complex assembly (i.e., by opening the Munc18-syntaxin complex and catalyzing trans SNARE complex formation). The data provided in these papers are quite compelling. The three papers should be cited separately, including the corresponding conclusions, and then explain why the authors think that these papers are not helpful or believable in defining (part of) the mechanism by which Munc13 mediates vesicle docking. Somewhat in the same context, can one use the liposome-clustering activity of proteins and protein fragments as a basis to extrapolate towards corresponding roles of the respective proteins in synaptic vesicle docking in neurons? For example, in the case of Syt1, such a conclusion can probably not be drawn – loss of Syt1 does not really affect synaptic vesicle docking much.

We have considerably revised the Introduction and the Discussion with regard to everything related to docking, as we now emphasize the role of Munc13s in docking much more than in the original manuscript. We have put our results in the context of different definitions of docking used in the literature and explain that our data are fully consistent with the interpretation of these three papers. In the text we indicate that these papers use a stringent definition of docking that then becomes equivalent to priming and SNARE complex assembly. However, we also propose that Munc13s are involved in upstream interactions with the vesicle and plasma membranes that help to bridge the membranes and facilitates the activity in promoting SNARE complex formation. With regard to the question about extrapolation, this is of course is a very important question and we believe that such extrapolations are more likely to be valid if they involve clear correlations between reconstitution and physiological data, and are consistent with other data in the field about the functions of the proteins under study (see point C of the response to the Summary).

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

1) The manuscript remains very long. Considering the revised conclusions, the authors are encouraged to take a fresh look and examine if all the data are really essential for the key conclusions.

We have considered this issue very carefully but we are reluctant to remove any of the data presented because, although some of the data may not be essential for the main conclusions, they still make important points that would be lost if the data are not presented here. We tried to reduce a little bit the length of the text while introducing changes to address the remaining concerns. While we agree that the manuscript is long, there are a lot of data that help building the overall story.

2) There are some supplementary figures that are actually quite important (e.g., Figure 4—figure supplement 4, and Figure 6—figure supplement 5). Please restrict supplementary figures to repeat experiments, raw data, and non-essential experiments.

The most critical panel of the previous Figure 4—figure supplement 4, panel C, has been moved to Figure 4 as panel E. The previous Figure 6—figure supplement 5 is now Figure 7 in the revised paper. As a consequence of this change, the previous Figure 7 is now Figure 8. To avoid increasing the overall number of figures, we have merged the previous Figures 8 and 9 into a single new Figure 9. For consistency with this change, the figure supplements corresponding to the old Figures 8 and 9 have been renumbered and are now Figure 9—figure supplements 13.

3) A general comment regarding all figures that compare lipid and content mixing results (Figures 4, 5, 7, 8, 9 and numerous supplemental figures): both types of experiments (lipid mixing and content mixing, respectively) are normalized differently: for lipid mixing 1% BOG is added and the resulting fluorescence intensity value is used for normalization; whereas for content mixing, it is the maximum value for each group of conditions in the particular panel. So, 100% content mixing could actually correspond to substantially fewer vesicles undergoing content mixing than vesicles undergoing lipid mixing. A suggestion would be to change all figure panel titles from "lipid mixing" to "relative lipid mixing" and "content mixing" to "relative content mixing", and to change all the y-axis labels from "Fluorescence (% of max)" to "Fluorescence (arbitrary units)". Another comment: ensemble experiments cannot determine what fraction of vesicles that undergo lipid mixing also undergo content mixing.

The terms ‘lipid mixing’ and ‘content mixing’ in the figures are just meant to be used as titles for easy identification of the data by the reader. We believe that adding the term relative may be more confusing than clarifying. The key issue that we agree needs to be made clearer is why the fluorescence signal reflecting lipid mixing is much smaller than that reflecting content mixing when both are expressed as percentages of the maximum fluorescence. We have clarified this issue in the text with the following paragraph:

‘Note that in these assays much of the Cy5 fluorescence increase caused by FRET from PhycoE (reflecting content mixing) should occur during the first round of fusion and that no further substantial increases are thus expected in subsequent rounds of fusion or upon detergent addition. Correspondingly, the maximum Cy5 fluorescence observed in our most efficient fusion reactions was similar to that observed upon detergent addition (e.g. Figure 3D, red curve; see Materials and methods). In contrast, the lipid mixing signal expressed as percentage of maximum Marina Blue fluorescence is much smaller in the same reactions (e.g. Figure 3C, red curve) because fluorescence de-quenching is expected to continue in successive rounds of fusion and to undergo a further, large increase upon detergent addition due to additional probe dilution.’

4) Figure 4—figure supplement 4C. We appreciate that the authors have performed DLS measurements to address the concern that the large increase in content mixing upon calcium addition could be due to a calcium-dependent docking/clustering effect. However the presence of large vesicles might dominate in the intensity autocorrelation functions shown. Percent Intensity plots (similar to those shown in Figures 2 and 6) should also be provided for Figure 4—figure supplement 4C.

We include percent intensity plots for the most critical DLS data in panels C-F of the revised Figure 4—figure supplement 4.

5) Figure 4—figure supplement 4C was only done for the soluble Syt1 C2AB fragment. If at all possible, please also perform the DLS experiments also with full-length Syt1. The reason is that there is much less lipid mixing in these experiments prior to calcium addition than for the experiments with the soluble Syt1 C2AB fragment (compare Figures 7 and 8).

The suggested experiments with full-length Syt1 were already described in the previous manuscript and shown in the old Figure 9—figure supplement 2, which is now Figure 9—figure supplement 3 in the revised manuscript (because of the changes summarized in point 2).

6) Figure 9: It is perhaps possible that there could be compensating factors in neurons that do not show a large phenotype for the DN mutant of Munc13. Nevertheless, if the author's model of calcium-independent bridging of T and V-vesicles by C1C2BMUNC2C is correct (Figure 6—supplement figure 5), why would calcium binding to the C2B domain of Munc13 have such a large effect? Please discuss.

We briefly discussed in the previous manuscript that

‘…Our data suggest that the primed state is metastable and requires Ca2+ binding to the Munc13-1 C2B domain for efficient content mixing (Figure 9F) either because the Ca2+-bound Munc13-1 C2B domain contributes directly to facilitate membrane fusion or because Ca2+ binding releases an inhibitory interaction existing in the primed state. The finding that a mutation in a Ca2+-binding loop of the Munc13-2 C2B domain increases release probability (Shin et al., 2010) is consistent with both possibilities and supports the notion that Munc13s form intrinsic part of the primed state of synaptic vesicles.’

We have left this same paragraph in the revised manuscript without additional discussion to avoid further lengthening of the text.

7) Figure 6—figure supplement 5: While the preference for the C1C2B domain for the T-vesicle membrane can be explained by the presence of DAG and PIP2 in that membrane, what is the mechanism that C2C would only interact with the V-vesicle membrane?

The model proposed in the new Figure 7, which was previously Figure 6—figure supplement 5, assumes that the C1 and C2 B domains bind to the plasma membrane and such binding places Munc13 in an orientation that favors binding of the C2C domain in trans to another apposed membrane (the synaptic vesicle membrane in this case). The model is discussed in pages 23-24, where we make it clear that the mechanism of liposome clustering by these large Munc13 fragments remains to be established, as there are multiple sites in these fragments that could potentially interact with membranes. Hence, extensive studies will be needed to define this mechanism clearly.

8) Figures 2 and 6: why does C1C2BMUN cluster V-vesicles, whereas C1C2BMUNC2C does not? This question is also related to the Discussion section, paragraph five. Please provide an explanation why the C2C domain would favor trans-interactions with V-vesicles.

In the previous manuscript, we had provided an explanation for the DLS data but not a model figure because there are multiple potential explanations and we preferred not to favor too much one over another. In response to this concern, we have added a new Figure 7—figure supplement 1 in the revised manuscript that describes plausible models to explain the DLS data. In the figure legend we make it clear that there are other potential models, as we do not want to mislead the reader into thinking that the proposed models are well established.

9) Subsection “Simultaneous evaluation of lipid and content mixing”, paragraph two": "A problem…". Actually, synaptic vesicles are acidic, so the property of sulforhodamine to produce an acidic interior of the proteoliposomes is not necessarily a problem. Moreover, the lack of leakiness under such conditions is desirable. In any case, it is good that the authors confirmed the content mixing results with the different content mixing assay by Zucchi and Zick. Thus, there is really no "problem" with the sulforhodamine content mixing assay.

We agree that the acid interior of the liposomes is not necessarily a problem, but the pH in the concentrated sulforhodamine solution that becomes encapsulated into the liposomes is very acidic (pH 2). We devoted extensive efforts to do fusion assays with sulforhodamine in both vesicle populations and were never able to observe efficient lipid mixing. In any case, we have replaced the word ‘problem’ with ‘potential concern’ in that sentence to address this issue.

10) Figures 1D and E: the effect of the different lipid compositions (PIP2 and DAG) appears to be much more pronounced in the presence of NSF/SNAP/Munc18, although C1C2BMUN is present in both experiments.

Yes, this is an important point that we emphasized in the previous manuscript and still emphasize in the revised manuscript in the sentence (Results section):

‘These data show that the effect of Munc13-1 C1C2BMUN alone on lipid mixing arises from a property that is largely independent of Ca2+, DAG and PIP2, whereas the lipid mixing observed in the more complete reconstitutions including C1C2 BMUN, Munc18-1 and NSF-αSNAP is stimulated by Ca2+, DAG and PIP2, thus exhibiting properties that are more similar to those of neurotransmitter release.’

11) Discussion, paragraph seven: in addition to Lee and Kyoung, please also cite the improvements of this assay by Diao et al. (eLife 2012); Lai et al. (eLife 2014). Another comment: the single vesicle content mixing assay by Kyoung et al. (2011); Lai et al. (2014) allows discrimination between effects that are due to docking vs. due to fusion probabilities in addition to the improved time resolution offered by such single vesicle assays.

We have included the additional requested references.

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

We thank the authors for their thoughtful considerations of the concerns raised previously. However, the Reviewing Editor has two remaining concerns:

1) Related to Point 9, the authors write:

"A potential concern with the use of sulforhodamine B fluorescence de-quenching to monitor content mixing in these bulk reconstitution experiments is that de-quenching can also arise from liposome leakiness. We attempted to perform experiments where both liposome populations are loaded with sulforhodamine B to assess how much of the de-quenching arises from leakiness (Yu et al., 2013). However, we were unable to observe lipid mixing in these control experiments. Since the solutions of high sulforhodamine B concentrations trapped in the liposomes are highly acidic (pH about 2), which may underlie the lack of lipid mixing, we attempted to increase the pH of these solutions to 5.5 or 7.4 before trapping them into the liposomes, but at both pH values the liposomes were strongly leaky."

Their observation of lack of lipid mixing when both vesicle populations are filled with sulforhodamine B is strange, considering that the authors observed efficient lipid mixing when only one class of vesicles is filled (Figure 3—figure supplement 1). Moreover, Kyoung et al., 2011 and Diao et al. 2012 observed both lipid and content mixing using their sulforhodamine B content mixing method, with only an extremely small fraction of leakage (0.01% of the docked vesicles). Could the problems be related to differences in reconstitution methods, rather than an issue with the sulforhodamine method per se?

This Reviewing Editor also would like to point out that the pH of the interior of the synaptobrevin/synaptotagmin vesicles ("donor" vesicles) is likely more neutral rather than pH 2 when using the published reconstitution protocol (Kyoung et al. Nat Protoc 8(1):1-16, 2013): the initial protein/lipid/detergent solution uses a detergent concentration that is at the CMC (OG ~ 30 mM), so vesicles do not yet form at this stage. This solution is then injected on a Cl-4B column that has been pre-equilibrated with vesicle buffer (pH 7.4, without sulforhodamine B), followed by elution with vesicle buffer (pH 7.4, again without sulforhodamine B) (Kyoung et al. Nat Protoc 8(1):1-16, 2013). Vesicles are forming on the column in a time course of several minutes as the detergent concentration decreases. One would expect that the pH of the interior of the column rapidly reaches 7.4.

In any case, as the authors point out, the alternative PhycoE-Biotin/Cy5-Streptavidin content mixing assay produces similar results (Figure 3—figure supplement 3). However, as written, the second paragraph of subsection “Simultaneous evaluation of lipid and content mixing” may come across to the casual reader as if there is something wrong with the sulforhodamine B content mixing method. Since this is a rather technical issue that distracts from the main messages of this paper, and that may be related to differences in reconstitution methods, the authors may wish to consider deleting this entire paragraph. In the future, perhaps the authors may wish to investigate a variety of content mixing assay jointly with other groups who have developed these assays for a more technical publication.

We would like to clarify that in the first revision we insisted in keeping these data because the first round of review questioned the conclusion that synaptotagmin C2AB was required for efficient content mixing but not lipid mixing in the reconstitutions including C1C2BMUN, and the sulforhodamine assays provided further support for this conclusion. In the second revision we did not remove the data obtained with the sulforhodamine method because the argument made in the second round of review (stating that there is no problem) was not compelling; we did have a big problem with the method when we included sulforhodamine on both vesicles. With regard to the concerns raised in this third round of review, it is unclear whether the buffer can handle the strongly acidic pH caused by the very high concentrations of sulforhodamine that are being used. However, we agree that the problems that we described for the sulforhodamine method could represent a distraction. Hence, we have removed these data in the third revised manuscript.

2) Point 11, related to the sentence starting "Approaches that allow faster measurements [e.g. single vesicle assays (Lee et al., 2010; Kyoung et al., 2011); will be required to test this prediction. " It is not just the faster time resolution that is important for future studies. The single vesicle-vesicle lipid/content mixing assay by Kyoung et al. 2011, Diao et al., 2012 et al. (but not the assay by Lee et al., 2010) can distinguish between effects due to docking, lipid mixing and content mixing on a 100 msec time scale for each individual vesicle pair. In addition to the improved time resolution, the discrimination between effects due to docking and/or fusion will be important for future studies. The authors may wish to consider making the discussion of the relevant publications clearer.

After the aforementioned sentence, we have now added the following: ‘An additional advantage of these assays is that they allow distinction of the docking event from lipid and content mixing.’

https://doi.org/10.7554/eLife.13696.031

Article and author information

Author details

  1. Xiaoxia Liu

    1. Department of Biophysics, University of Texas Southwestern Medical Center, Dallas, United States
    2. Department of Biochemistry, University of Texas Southwestern Medical Center, Dallas, United States
    3. Department of Pharmacology, University of Texas Southwestern Medical Center, Dallas, United States
    Contribution
    XL, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Contributed equally with
    Alpay Burak Seven
    Competing interests
    No competing interests declared.
  2. Alpay Burak Seven

    1. Department of Biophysics, University of Texas Southwestern Medical Center, Dallas, United States
    2. Department of Biochemistry, University of Texas Southwestern Medical Center, Dallas, United States
    3. Department of Pharmacology, University of Texas Southwestern Medical Center, Dallas, United States
    Contribution
    ABS, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Contributed equally with
    Xiaoxia Liu
    Competing interests
    No competing interests declared.
  3. Marcial Camacho

    Department of Neurophysiology, NeuroCure Cluster of Excellence, Charité-Universitätsmedizin Berlin, Berlin, Germany
    Contribution
    MC, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    No competing interests declared.
  4. Victoria Esser

    1. Department of Biophysics, University of Texas Southwestern Medical Center, Dallas, United States
    2. Department of Biochemistry, University of Texas Southwestern Medical Center, Dallas, United States
    3. Department of Pharmacology, University of Texas Southwestern Medical Center, Dallas, United States
    Contribution
    VE, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    No competing interests declared.
  5. Junjie Xu

    1. Department of Biophysics, University of Texas Southwestern Medical Center, Dallas, United States
    2. Department of Biochemistry, University of Texas Southwestern Medical Center, Dallas, United States
    3. Department of Pharmacology, University of Texas Southwestern Medical Center, Dallas, United States
    Contribution
    JX, Acquisition of data, Analysis and interpretation of data
    Competing interests
    No competing interests declared.
  6. Thorsten Trimbuch

    Department of Neurophysiology, NeuroCure Cluster of Excellence, Charité-Universitätsmedizin Berlin, Berlin, Germany
    Contribution
    TT, Acquisition of data, Analysis and interpretation of data
    Competing interests
    No competing interests declared.
  7. Bradley Quade

    1. Department of Biophysics, University of Texas Southwestern Medical Center, Dallas, United States
    2. Department of Biochemistry, University of Texas Southwestern Medical Center, Dallas, United States
    3. Department of Pharmacology, University of Texas Southwestern Medical Center, Dallas, United States
    Contribution
    BQ, Acquisition of data, Analysis and interpretation of data
    Competing interests
    No competing interests declared.
  8. Lijing Su

    1. Department of Biophysics, University of Texas Southwestern Medical Center, Dallas, United States
    2. Department of Biochemistry, University of Texas Southwestern Medical Center, Dallas, United States
    3. Department of Pharmacology, University of Texas Southwestern Medical Center, Dallas, United States
    Contribution
    LS, Conception and design
    Competing interests
    No competing interests declared.
  9. Cong Ma

    1. Key Laboratory of Molecular Biophysics of the Ministry of Education, Huazhong University of Science and Technology, Wuhan, China
    2. College of Life Science and Technology, Huazhong University of Science and Technology, Wuhan, China
    Contribution
    CM, Conception and design, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    No competing interests declared.
    ORCID icon 0000-0002-7814-0500
  10. Christian Rosenmund

    Department of Neurophysiology, NeuroCure Cluster of Excellence, Charité-Universitätsmedizin Berlin, Berlin, Germany
    Contribution
    CR, Conception and design, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    CR: Reviewing editor, eLife
  11. Josep Rizo

    1. Department of Biophysics, University of Texas Southwestern Medical Center, Dallas, United States
    2. Department of Biochemistry, University of Texas Southwestern Medical Center, Dallas, United States
    3. Department of Pharmacology, University of Texas Southwestern Medical Center, Dallas, United States
    Contribution
    JR, Conception and design, Analysis and interpretation of data, Drafting or revising the article
    For correspondence
    Jose.Rizo-Rey@UTSouthwestern.edu
    Competing interests
    No competing interests declared.
    ORCID icon 0000-0003-1773-8311

Funding

German Research Council (SFB958)

  • Christian Rosenmund

European Research Council (SynVglut)

  • Christian Rosenmund

Welch Foundation (I-1304)

  • Josep Rizo

National Institutes of Health (NS037200)

  • Josep Rizo

National Institutes of Health (NS040944)

  • Josep Rizo

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

Acknowledgements

We thank Shae Padrick for fruitful discussions and the Charite viral core facility for provision of lentiviral constructs. This research was supported by a grant from the Welch Foundation (I-1304) (to JR), by NIH grants NS037200 and NS040944 (to JR), by German Research Council grant SFB958 (to CR) and by the ERC grant SynVglut (to CR).

Ethics

Animal experimentation: Animal welfare committees of Charité Medical University and the Berlin state government Agency for Health and Social Services approved all protocols for animal maintenance and experiments (license no. T 0220/09).

Reviewing Editor

  1. Axel T Brunger, Reviewing Editor, Stanford University, United States

Publication history

  1. Received: December 9, 2015
  2. Accepted: May 22, 2016
  3. Accepted Manuscript published: May 23, 2016 (version 1)
  4. Version of Record published: June 29, 2016 (version 2)

Copyright

© 2016, Liu 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

  • 1,957
    Page views
  • 573
    Downloads
  • 21
    Citations

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

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)

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

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

Further reading

    1. Structural Biology and Molecular Biophysics
    Johan E Larsson et al.
    Research Article
    1. Structural Biology and Molecular Biophysics
    David Jon Wasilko et al.
    Research Article