Munc13 C₂B-Domain – an Activity-Dependent Ca²⁺-Regulator of Synaptic Exocytosis

Abstract

Munc13 is a multidomain protein of presynaptic active zones that mediates the priming and plasticity of synaptic vesicle exocytosis, but the mechanisms involved remain unclear. Here, we use biophysical, biochemical, and electrophysiological approaches to demonstrate that the central C₂B-domain of Munc13 functions as a Ca²⁺-regulator of short-term synaptic plasticity. The crystal structure of the C₂B-domain revealed an unusual Ca²⁺-binding site with an amphipathic α-helix. This configuration confers onto the C₂B-domain unique Ca²⁺-dependent phospholipid-binding properties favoring phosphatidylinositolphosphates. A mutation that inactivated Ca²⁺-dependent phospholipid binding to the C₂B-domain did not alter neurotransmitter release evoked by isolated action potentials, but depressed release evoked by action potential trains. In contrast, a mutation that increased Ca²⁺-dependent phosphatidylinositolbisphosphate binding to the C₂B-domain enhanced release evoked by isolated action potentials and by action potential trains. Our data suggest that during repeated action potentials, Ca²⁺- and phosphatidylinositolphosphate-binding to the Munc13 C₂B-domain potentiate synaptic vesicle exocytosis, thereby offsetting synaptic depression induced by vesicle depletion.

INTRODUCTION

Synaptic transmission is initiated when Ca²⁺-influx during an action potential triggers neurotransmitter release¹. Synaptic transmission is not a constant point-to-point transfer of information from one neuron to the next, but changes as a function of use, rendering synapses elementary computational units of the brain². Many different types of use-dependent synaptic plasticity have been described, among which presynaptic short-term plasticity stands out because it is universally present at synapses, and can alter synaptic transmission more than 10-fold³. Short-term plasticity is of central importance for information processing by the brain; for example, it may underlie working memory formation in cortex⁴.

At first approximation, presynaptic short-term plasticity results from two opposing processes³. Repeated action potentials deplete the readily-releasable pool (RRP) of synaptic vesicles, thereby inducing synaptic depression. At the same time, Ca²⁺-influx during repeated action potentials causes accumulation of residual Ca²⁺, thereby inducing synaptic facilitation. As a consequence, a high release probability usually results in synaptic depression because the RRP becomes depleted, whereas a low release probability usually results in synaptic facilitation because vesicle depletion is delayed but accumulating residual Ca²⁺ increases Ca²⁺-triggering.

Considerable evidence, however, indicates that presynaptic plasticity is an active, regulated, and synapse-specific process that goes beyond a passive response dictated by the release probability and RRP size. For example, RIM1α and Munc13 are active zone proteins that interact with each other, and form a heterotrimeric complex with the synaptic vesicle protein Rab3^5–9. Mutations in each of these three proteins induce changes in short-term synaptic plasticity that cannot be accounted for by corresponding alterations in residual Ca²⁺, release probability, or RRP^8,10–14. These and other observations indicate that Ca²⁺ not only triggers release, but that during stimulus trains, the residual Ca²⁺ accumulating between action potentials regulates release by independent mechanisms.

At present, the major Ca²⁺-regulator in short-term synaptic plasticity is thought to be calmodulin¹⁵. Calmodulin regulates neurotransmitter release by multiple mechanisms, including a direct modulation of Ca²⁺-channels^16–18, activation of protein kinases¹⁹, regulation of synaptic vesicle priming via the cytoskeleton²⁰, and binding to Munc13-1 and -2 (13). Moreover, calmodulin acts in presynaptic long-term plasticity by activating adenylate cyclase during mossy-fiber LTP^21,22. However, the large variety of different types of presynaptic plasticity with distinct spatio-temporal profiles suggests that calmodulin is unlikely to account for all forms of short-term plasticity.

Munc13s and RIMs are essential for priming synaptic vesicles, and are additionally involved in short- and long-term synaptic plasticity^8,10. Three Munc13 isoforms (Munc13-1, -2, and -3, of which Munc13-2 is expressed in two principal isoforms called bMunc13-2 and ubMunc13-2) function in synaptic vesicle exocytosis^5,10,23,24. In addition, two ubiquitously expressed Munc13 isoforms (BAP-3 and Munc13-4 probably act in non-synaptic forms of exocytosis^25–27. Munc13s have variable N-terminal sequences, but contain similar central and C-terminal domains: a C₂B-domain, a large Munc13-homology region (the MUN domain¹⁴), and a Ca²⁺-independent C₂C-domain (Fig. 1A). It seems likely that the canonical Munc13-domains, i.e. their C₂B-, MUN- and C₂C-domains, mediate their shared functions, whereas their variable N-terminal domains modulate these functions. Consistent with this notion, the calmodulin-binding motif and the C₁-domain of Munc13-1 are involved in short-term synaptic plasticity^12,13, whereas its MUN domain mediates its priming function^14,28,29.

Figure 1. The Munc13-1 C₂B-domain is a Ca²⁺-binding module.

a. Domain organization of Munc13-1 and bMunc13-2. The binding activities of various domains are indicated above (CaM = calmodulin; DAG = diacylglycerol), and their presumed Ca²⁺-binding ability below the domains. An alignment of the Ca²⁺-binding loops from the synaptotagmin-1 C₂A-domain, the Munc13-1 C2B-domain, and the Munc13-2 C₂B-domain is shown below the domain organization (residues 162–241 from rat synaptotagmin-1 [acc. # X52772]; 695–779 from rat Munc13-1 [acc. # U24070]; and 619–703 from rat bMunc13-2 [acc. # AF159706]). In the alignment, conserved sequences are highlighted (black = Ca²⁺-binding residues; yellow = top loop; blue = β-strands; red = conserved charged sequence in loop 3 specific for the Munc13 C₂B-domains). The two C₂B-domain mutations analyzed (“DN” and “KW”) are described at the bottom.

b. Fluorescent emission spectra of WT and DN-mutant Munc13-2 C₂B-domains without and with 1 mM Ca²⁺ plus/minus EGTA, or with 10 mM Mg²⁺ (for data on Munc13-1 and for individual spectra, see Supplementary Fig. 1).

c. ¹H-¹⁵N HSQC spectra of the Munc13-1 C₂B-domain in the absence (black contours) and presence (red contours) of 0.5 mM Ca²⁺.

d, e. Ca²⁺-binding to the Munc13-1 C₂B-domain monitored with ¹H-¹⁵N HSQC spectra. The diagrams show expansions of superpositions of selected ¹H-¹⁵N HSQC spectra acquired during a titration of Ca²⁺ from 0 to 0.7 mM. The contours are color coded according to the Ca²⁺-concentration (indicated in μM next to the contours).

The C₂B-domains of all Munc13 isoforms, but not their C₂A- and C₂C-domains, contain the requisite Ca²⁺-binding residues of C₂-domains³⁰ (Fig. 1A), suggesting that Munc13's may universally bind Ca²⁺ via their C₂B-domains. However, previous attempts to demonstrate Ca²⁺-binding to the Munc13 C₂B-domain failed⁵. We now show that the rat Munc13 C₂B-domain binds Ca²⁺, and contains an unusual α-helix in its top Ca²⁺-binding loops. The Munc13 C₂B-domain exhibits Ca²⁺-dependent phospholipid binding with an unexpected PIP- and PIP₂-specificity that differs from that of other C₂-domains, and that mediates the Ca²⁺-dependent regulation of short-term synaptic plasticity by the Munc13 C₂B-domain. Thus, our data reveal that the Munc13 C₂B-domain functions as a Ca²⁺-regulator of short-term synaptic plasticity by interfacing with PIP and PIP₂.

RESULTS

The Munc13 C₂B-domain is a Ca²⁺-binding module

The C₂B-domains of all Munc13 isoforms contain the canonical Ca²⁺-binding sites of C₂-domains³⁰ (Fig. 1a), but exhibit only limited sequence homology to other C₂-domains, hindering prediction of domain boundaries. Thus, we first examined the minimum sequence necessary to obtain an autonomously folded Munc13 C₂B-domain, and identified a C₂B-domain fragment that was soluble and monomeric (residues 675–820 in Munc13-1).

To determine whether the C₂B-domain binds Ca²⁺, we recorded fluorescence spectra of purified C₂B-domains from Munc13-1 and Munc13-2 in the absence or presence of Ca²⁺ or Mg²⁺ (Fig. 1b and Supplementary Fig. 1). For Munc13-2, we also examined a mutant C₂B-domain in which two canonical aspartates (D⁶²⁹ and D⁶³⁵) in the presumptive Ca²⁺-binding sites were replaced by asparagines (referred to as the DN-mutant, Fig. 1a). Application of Ca²⁺ but not Mg²⁺ enhanced the intrinsic fluorescence of the wild-type (WT) C₂B-domains, but had no effect on DN-mutant C₂B-domain, suggesting that the Munc13 C₂B-domains specifically bind Ca²⁺ (Fig. 1b).

We next acquired ¹H-¹⁵N heteronuclear single quantum correlation (HSQC) spectra of the Munc13-1 C₂B-domain in the absence and presence of Ca²⁺ (Fig. 1c, black and red contours, respectively). Ca²⁺ induced extensive cross-peak changes as expected for a Ca²⁺-binding module. During Ca²⁺-titrations, some cross-peaks exhibit progressive Ca²⁺-induced shifts (e.g. Fig. 1d, upper left corner). Other cross-peaks disappeared during the titration, or shifted to different parts of the spectrum (Figs. 1d and 1e), suggesting that the exchange between Ca²⁺-free and Ca²⁺-bound states is slow on the NMR time scale. The curved, progressive Ca²⁺-induced shifts of some cross-peaks, and the differential shifts between cross-peaks (Figs. 1d and 1e), show that at least two Ca²⁺-ions bind to the C₂B-domain. At a C₂B-domain concentration of 120 μM, no major cross-peak shifts occurred beyond 250–300 μM Ca²⁺, demonstrating that Ca²⁺-binding was saturable. The cross-peak movements can be fitted to a binding model with two Ca²⁺-binding sites and a K_D of <100 μM.

Crystal structure of Ca²⁺-free and Ca²⁺-bound C₂B-domain

To study how Ca²⁺ binds to the Munc13 C₂B-domain, we crystallized the Ca²⁺-free and Ca²⁺-bound Munc13-1 C₂B-domain. The Ca²⁺-free and Ca²⁺-bound C₂B-domain cystals exhibited distinct space groups (C222₁ and P4₃2₁2, respectively). Using diffraction data and the crystal structure of the synaptotagmin-1 C₂A-domain³¹ as a search model for molecular replacement, we determined the structure of the Ca²⁺-free Munc13-1 C₂B-domain to a resolution of 1.90 Å (Table 1). The resulting model was then employed for molecular replacement together with diffraction data to determine the structure of the Ca²⁺-bound C₂B-domain to a resolution of 1.37 Å (Fig. 2).

Table 1.

Data collection and refinement statistics for the Munc13-1 C₂B domain

	Munc13-1 C2B-domain Ca2+-free	Munc13-1 C2B-domain Ca2+-bound
Data collection
Space group	C2221	P43212
Cell dimensions
a, b, c (Å)	42.57, 101.14, 68.04	55.94, 55.94, 89.98
α, β, γ (°)	90.0, 90.0, 90.0	90.0, 90.0, 90.0
Resolution (Å)*	26.43-1.90 (1.93-1.90)	30.00-1.37 (1.39-1.37)
Rmerge (%)	5.0 (37.9)	5.5 (46.4)
I/σI	33.1 (2.4)	40.8 (3.1)
Completeness (%)	86.5 (47.5)	99.8 (96.4)
Redundancy	7.5 (3.3)	12.6 (6.1)
Refinement
Resolution (Å)	26.00 – 1.90 (1.94-1.90)	30.00 – 1.37 (1.40-1.37)
No. reflections (unique/Rfree)	9,905/501	30,262/1,546
Rwork / Rfree	0.217 (0.263) / 0.282 (0.318)	0.170 (0.208) / 0.193 (0.231)
No. atoms	1,019	1,352
Protein	935	1,175
Ligand/ion	19/2	12/4
Water	63	161
B-factors
Protein	43.9	17.3
Ligand/ion	45.7/49.0	14.7/15.5
Water	45.9	18.4
R.m.s. deviations
Bond lengths (Å)	0.017	0.018
Bond angles (°)	1.579	1.813

For each structure, data was collected from one crystal.

^*Values in parentheses are for highest-resolution shell.

Figure 2. Three-dimensional structures of the Ca²⁺-free and Ca²⁺-bound Munc13-1 C₂B-domain.

a. Ribbon diagram of the crystal structure of the Ca²⁺-bound Munc13-1 C₂B-domain (blue = β-strands; orange = α-helices). Bound Ca²⁺-ions are shown as yellow spheres; β-strands are numbered from 1 to 8. The top loops are labeled loop 1 – loop 4; N and C indicate N- and C-termini, respectively. See Supplementary Fig. 2 for analysis of crystal contacts.

b. Backbone superposition of the Ca²⁺-free (orange) and Ca²⁺-bound (blue) Munc13-1 C₂B-domains.

c. Backbone superposition of the crystal structures of the Ca²⁺-bound Munc13-1 C₂B-domain (blue) and the Ca²⁺-free synaptotagmin-1 C₂A-domain (red; PDB accession code 1rsy). N- and C-termini of both domains are indicated with letters of the corresponding color.

d. 2Fo−Fc electron density map contoured at 1σ of the Ca²⁺-binding region of the Munc13-1 C₂B-domain superimposed with a stick model of the protein. Ca²⁺-ions and water molecules are represented by yellow spheres and red stars, respectively. In this and the following panels, protein atoms are color coded: green, carbon; blue, nitrogen; red, oxygen.

e. Ribbon-and-stick diagram summarizing the Ca²⁺-binding mode of the Munc13-1 C₂B-domain. The water molecules are not shown for simplicity. All other Ca²⁺-ligands are shown as stick models and labeled; K704CO and E758CO denote the backbone carbonyl group of the corresponding residues. Ca²⁺-ions are labeled Ca1 and Ca2.

f. Ribbon-and-stick diagram of the Munc13-1 C₂B-domain illustrating the amphipathic character of the α-helix of loop 3. The side chains of the Ca²⁺ ligands and of all residues in loop 3 are shown as stick models.

The Ca²⁺-free and Ca²⁺-bound C₂B-domains contained a typical C₂-domain β-sandwich fold with two four-stranded β-sheets and a type II C₂-domain topology (Figs. 2a and 2b)³². Ca²⁺ did not cause major changes in the β-sandwich, as described for synaptotagmin-1 C₂-domains^31,33,34, but induced large changes in the top loops. In the Ca²⁺-free C₂B-domain, substantial parts of loops 1, 3 and 4 exhibited little electron density, indicating that the Ca²⁺-free loops are disordered (Fig. 2b). In the Ca²⁺-bound C₂B-domain, however, we observed well-defined electron densities for all four top loops (Figs. 2a and 2b), for two bound Ca²⁺-ions, and for multiple water molecules (partly illustrated in Fig. 2d). The Ca²⁺-binding region is likely stabilized by Ca²⁺ binding, similar to the synaptotagmin-1 C₂-domains^33,34. In addition, the top loops in the Ca²⁺-bound C₂B-domain structure were involved in extensive crystal contacts that may have contributed to the stabilization of the top loops.

Structural comparisons using DALI³⁵ revealed that several structures of other type II C₂-domains in the Protein Data Bank (PDB) exhibited close similarity to the Munc13-1 C₂B-domain, including structures of the C₂-domains from PKCζ, PLC-δ1 and cPLA₂ (PDB accession codes 1gmi, 1djx and 1rlw, respectively). However, the lowest rms deviation from the Ca²⁺-bound Munc13-1 C₂B-domain (1.1 Å for 97 equivalent Cα carbons) was observed for the synaptotagmin-1 C₂A-domain structure (PDB accession codes 1rsy and 1byn), despite the fact that it exhibits a type I topology. Backbone superpositions confirmed the similarity of the synaptotagmin-1 C₂A-domain and the Munc13-1 C₂B-domain (Fig. 2c). Strikingly, the major difference between the Munc13-1 C₂B-domain and all other C₂-domains is an extended loop 3 of the Munc13 C2B-domain that contains a unique protruding α-helix, and is absent from other C₂-domains (Figs. 2a,f).

Ca²⁺-binding mode of the Munc13 C₂B-domain

C₂-domains commonly bind two or three Ca²⁺-ions through five conserved aspartate side chains from loop 1 (aspartates D1 and D2) and loop 3 (aspartates D3–D5)^30,36. The crystal structure of the Ca²⁺-bound Munc13-1 C₂B-domain revealed two bound Ca²⁺-ions (Figs. 2), consistent with the NMR data (Figs. 1d and 1e). The two Ca²⁺-ions are coordinated by the five canonical aspartate residues, accounting for the block of Ca²⁺-binding by the DN-mutation (Fig. 1b). In addition, the Ca²⁺-ions were coordinated by two backbone carbonyl oxygens and two water molecules (Figs. 2d and 2e). Loop 3 of the Munc13-1 C₂B-domain contains additional acidic residues that did not participate in Ca²⁺-binding but are oriented towards the Ca²⁺-binding sites, and may increase its Ca²⁺-affinity (Fig. 2f). Moreover, exposed basic and hydrophobic residues confer an amphipathic character onto the unique α-helix of loop 3 in the Munc13 C₂B-domain. This amphipathic character could be increased by extending the α-helix towards the C-terminus to include R⁷⁶⁹ to R⁷⁷², and the helical structure may have been partially distorted by crystal contacts (Supplementary Fig. 2). Hence, a longer α-helix spanning residues 762–772 may be formed by loop 3 of the Munc13 C₂B-domain in solution and/or upon binding of the C₂B-domain to phospholipid bilayers (see below).

The Munc13 C₂B-domain is a Ca²⁺/phospholipid-binding module

We next examined whether Munc13 C₂B-domains bind to phospholipids in a Ca²⁺-dependent manner. First, we measured fluorescence resonance-energy transfer (FRET) from the Munc13-1 C₂B-domain to dansyl-labeled liposomes containing PIP and PIP₂. Ca²⁺ increased FRET only when both protein and liposomes were present, and this increase was reversed by EGTA (Fig. 3a and Supplementary Fig. 3). FRET may be mediated by the conserved tryptophan residue near loop 3 of the Munc13 C₂B-domain (Fig. 1a), as this residue is in the phospholipid-interacting sequence of C₂-domains.

Figure 3. Ca²⁺-dependent binding of the Munc13 C₂B-domain to PIP/PIP₂-containing liposomes.

a. FRET assays of Ca²⁺-dependent binding of the Munc13 C₂B-domain to dansyl-labeled `synaptic' liposomes containing 0.5% PIP and 0.1% PIP₂ (0.03 mg/ml; total volume = 0.6 ml). Fluorescence spectra (excitation = 282 nm) were monitored in solutions containing either only the C₂B-domain, liposomes, or both as indicated on the right³⁴. Spectra were first recorded in Ca²⁺-free buffer (black traces, covered by overlying green, red, or blue traces), then after addition of 2 mM Mg²⁺ (blue traces, under the overlying green or red traces), then after addition of 0.2 mM Ca²⁺ (red traces), then again after further addition of 1 mM EGTA (green trace, done only for the samples containing both liposomes and C₂B-domain protein). Data show a representative experiment repeated multiple times; see Supplementary Fig. 3 for individual spectra.

b, c. Centrifugation assays of Ca²⁺-dependent Munc13 C₂B-domain binding to `synaptic' liposomes containing 0.5% PIP and 0.1% PIP₂ (b), or 0.25% PIP and 0.05% PIP₂ (c). GST-fused Munc13 C₂B-domains and the synaptotagmin-1 C₂A/B-domain fragment (used as an internal control) were bound to liposomes at the indicated free Ca²⁺-concentrations clamped with Ca²⁺/EGTA buffer containing 2 mM Mg². Co-pelleted Munc13 and synaptotagmin-1 C₂-domains were analyzed by SDS-PAGE and Coomassie Blue staining, and quantified by scanning (top panels = representative experiments; bottom panels = summary graphs (means ± SEMs [n=3]); data were normalized to binding at the highest Ca²⁺-concentration; quantitations for synaptotagmin-1 for panel b are shown in Supplementary Fig. 4).

Next, we employed a centrifugation assay with liposomes containing a synaptic phospholipid composition with 0.5% PIP and 0.1% PIP₂ (ref. 37). Besides the wild-type Munc13-1 and 13-2 C₂B-domains and the DN-mutant Munc13-2 C₂B-domain, we examined an additional Munc13-2 C₂B-domain mutant called the KW-mutant, in which lysine⁶³⁰ in loop 1 is exchanged for tryptophan (Fig. 1a). The KW-mutant was designed to render the Munc13 C₂B-domain more similar to the synaptotagmin-1 C₂A-domain. Synaptotagmin-1 contains at this position a hydrophobic methionine (Fig. 1a) that inserts into the phospholipid bilayer in a Ca²⁺-dependent manner³⁸, and enhances the Ca²⁺-dependent phospholipid binding affinity of the synaptotagmin-1 C₂A-domain^39–41. Tryptophan was used instead of methionine to maximize the membrane penetration of the Munc13 C₂B-domain, as previously shown for the synaptotagmin-1 C₂A-domain^40,41.

The Munc13-1 and 13-2 C₂B-domains bound poorly to synaptic liposomes in the absence of Ca²⁺, but strongly in the presence of Ca²⁺ (Fig. 3b). Quantitation of Coomassie-blue stained SDS-gels revealed that the Munc13-1 and 13-2 C₂B-domains exhibited similar apparent Ca²⁺-affinities (Munc13-1: EC₅₀ = 5.5±0.9 μM Ca²⁺ [n=3]; Munc13-2: EC₅₀ = 5.3±0.8 μM Ca [n=3]; means ± SEMs) that were indistinguishable from that of the synaptotagmin-1 C₂A/B-domain fragment (EC₅₀ = 6.4 ± 0.5 μM Ca²⁺ [n=4; means ± SEMs]; Supplementary Fig. 4). The DN-mutation blocked all Ca²⁺-dependent phospholipid binding (EC₅₀ >1 mM Ca²⁺), whereas the KW-mutation did not alter the extent or apparent Ca²⁺-affinity of Ca²⁺-dependent phospholipid binding under these conditions (EC₅₀ = 5.4±0.9 μM Ca²⁺ [n=3; means ± SEM]; Fig. 3b).

Unusual phospholipid specificity of the Munc13 C₂B-domain

We next examined whether decreasing the PIP- and PIP₂-concentrations alters Ca²⁺-dependent phospholipid binding by wild-type or mutant Munc13 C₂B-domains. An only two-fold decrease of the PIP- and PIP₂-concentrations (to 0.25% PIP and 0.05% PIP₂) abolished Ca²⁺-dependent liposome binding of the wild-type Munc13 C₂B-domain, but not of the synaptotagmin-1 C₂-domains (Fig. 3c). Strikingly, the KW-mutation converted the Munc13 C₂B-domain into a synaptotagmin-like domain, with full Ca²⁺-induced binding to the liposomes containing reduced PIP- and PIP₂-concentrations (Munc13-2 KW: EC₅₀ = 18.8±2.6 μM Ca²⁺ [n=4]; Syt-1: EC₅₀ = 17.2±2.4 μM Ca²⁺ [n=3]; means ± SEMs; Fig. 3c). Thus, the Munc13 C₂B-domain is more sensitive to the PIP- and PIP₂-concentrations than the synaptotagmin-1 C₂-domains, but a single amino-acid substitution renders the lipid-binding properties of the Munc13 C₂B-domain similar to those of synaptotagmin-1.

Assays of Ca²⁺-dependent binding of wild-type Munc13-2 C₂B-domain to liposomes containing increasing concentrations of PIP or PIP₂ revealed that both phosphatidylinositolphospholipids equally promoted Ca²⁺-dependent binding (Fig. 4 and Supplementary Fig. 5). This behavior was unexpected because Ca²⁺-dependent binding of the synaptotagmin-1 C₂-domains to liposomes exhibits a strong preference for PIP₂ due to its higher negative charge^42,43. Again, the KW-mutant Munc13 C₂B-domain preferentially bound to the PIP₂-containing liposomes, similar to the synaptotagmin-1 C₂-domains (Fig. 4).

Figure 4. PIP- and PIP₂-dependence of Ca²⁺-induced liposome binding to Munc13 C₂B-domains.

a, b Quantitation of Ca²⁺-dependent Munc13 C₂B-domain binding to `synaptic' liposomes as a function of the PIP- (a) or PIP₂-concentration (b). Binding assays were carried out using the centrifugation assay (Figs. 3b and 3c) in the absence (open symbols) or presence of 0.1 mM Ca²⁺ (filled symbols) as a function of the concentration of PIP (a) or PIP₂ (b) in the liposomes. The top panels display representative experiments, and the bottom panels summary graphs (means ± SEMs [n=3]; data were normalized to binding at the highest free Ca²⁺ concentration). Wild-type and KW-mutant C₂B-domains are not significantly different for the PIP titration (a), but are significantly different for the PIP₂ titration (b; p=0.0016 using a 2-way ANOVA test; see Supplementary Fig. 5 for direct comparison of the binding of the WT C₂B-domain to PIP- or PIP₂-containing liposomes).

In synaptotagmin-1, C₂-domains not only bind to phospholipids, but also to SNARE proteins^43–46. However, pull-down experiments with solubilized brain proteins uncovered Ca²⁺-dependent binding of the Munc13 C₂B-domains only to tubulin (which as an abundant protein binds non-specifically to many proteins), but not to SNARE proteins, suggesting that the Munc13 C₂B-domain does not interact with SNARE proteins (Supplementary Fig. 6).

Role of Munc13-2 C₂B-domain in neurotransmitter release

In order to assess the functional importance of Ca²⁺-binding to the Munc13 C₂B-domain, we analyzed synaptic transmission in autapses formed by hippocampal neurons that were cultured on micro-islands of glia cells. The neurons were isolated from mice that lack Munc13-1 and 13-2, and were rescued by viral expression of the `ubMunc13-2' variant of Munc13-2, used because of its pronounced effects on short-term synaptic plasticity^24,47.

We first analyzed synaptic transmission induced by isolated action potentials. Wild-type, DN-mutant, and KW-mutant Munc13-2 rescued the loss of synaptic transmission induced by deletion of Munc13-1 and 13-2 (Fig. 5a). Rescue with WT and DN-mutant Munc13-2 caused no major change in EPSC amplitudes, whereas rescue with KW-mutant Munc13-2 increased the EPSC amplitudes almost two-fold (Fig. 5b). To test whether this change is due to a difference in the size of the RRP, we measured the RRP by application of hypertonic sucrose⁴⁸, but detected no significant change (Fig. 5c). We also determined the vesicular release probability (P_vr) for each neuron expressing WT or mutant Munc13-2 by calculating the ratio of integrated EPSC and RRP charges. The DN-mutation did not alter the vesicular release probability, whereas the KW-mutation nearly doubled it (Fig. 5d).

Figure 5. Effect of Munc13-2 C₂B-domain mutations on release induced by isolated action potentials.

All experiments in this figure and Fig. 6 were performed in hippocampal autaptic neurons cultured from Munc13-1/-2 double KO mice. Neurons were infected with recombinant Semliki Forest Virus expressing WT, DN-mutant, or KW-mutant Munc13-2, and excitatory postsynaptic currents (EPSCs) were recorded in whole-cell mode.

a, Representative EPSCs evoked by isolated action potentials (left) or 0.5 M sucrose (right) in neurons expressing WT (black), DN-mutant (blue), or KW-mutant Munc13-2 (red).

b–d, Mean EPSC amplitudes (b), RRP size (c, measured as the response to 0.5 M sucrose, integrating the transient current component for 4 s); and vesicular release probability (d, calculated as the ratio of the charge of evoked responses to that of the RRP). Data shown are means ± SEMs (n= (WT: n=58; DN: n=57; KW: n=79; ***, p<0.001 by paired t-test).

e, Representative EPSCs evoked by isolated action potentials in neurons expressing WT (black), DN-mutant (blue), or KW-mutant Munc13-2 (red) at three different Ca²⁺-concentrations as indicated.

f, Mean ratio of the EPSC amplitudes monitored at low vs. high Ca²⁺ in neurons expressing WT (black), DN-mutant (blue), or KW-mutant Munc13-2 (red; WT, n=16; DN, n=14; KW, n=16; *, p<0.05; see Supplementary Table 2 for a numerical listing of all electrophysiologically results.

To confirm that the KW- but not the DN-mutation of the C₂B-domain of Munc13 alters the release probability during islated action potentials, we monitored the relative EPSC amplitudes of synapses expressing WT, DN-, or KW-mutant Munc13-2 at low (1 mM) or high (12 mM) extracellular Ca²⁺-concentrations (Fig. 5e). Consistent with an unchanged basal release probability, WT and DN-mutant Munc13-2 exhibited the same relative Ca²⁺-dependent changes in ESPC amplitudes. In contrast, KW-mutant Munc13-2 displayed a relative increase in EPSC amplitude at the low ambient Ca²⁺-concentration, confirming the hypothesis (Fig. 5f).

In increasing release, KW-mutant Munc13-2 could act either as a Ca²⁺-sensor for triggering release analogous to synaptotagmin, or as an auxiliary Ca²⁺-regulator of Ca²⁺-triggering by synaptotagmin. To differentiate between these two possibilities, we tested whether wild-type or KW-mutant Munc13-2 confer Ca²⁺-triggered neurotransmiter release onto synapses from synaptotagmin-1 KO mice that lack almost all such release⁴⁹. However, neither WT nor KW-mutant Munc13-2 rescued the loss of Ca²⁺-induced synchronous release in synaptotagmin-deficient synapses, suggesting that Munc13-2 functions as an auxiliary Ca²⁺-regulator in release (Supplementary Fig. 7).

The Munc13-2 C₂B-domain in short-term plasticity

We next monitored synaptic responses induced by 2.5, 10, and 40 Hz stimulus trains in synapses expressing WT or mutant Munc13-2. Plots of normalized responses revealed that as expected, synapses expressing WT Munc13-2 exhibited strong facilitation at 2.5 Hz stimulation, and less facilitation at 10 Hz stimulation (Figs. 6a–6d; and Supplementary Fig. 8). In contrast, synapses expressing DN- or KW-mutant Munc13-2 both displayed no facilitation, but transient depression during the 2.5 Hz stimulation, and persistent depression during the 10 Hz stimulation (Figs. 6a and 6c).

Figure 6. Ca²⁺-binding to the Munc13 C₂B-domain regulates release during high-frequency action potential trains.

a, b. Mean normalized (left panels) and absolute (right panels) EPSC amplitudes in response to a 2.5 Hz (a) or 10 Hz (b) action potential train in Munc13-deficient neurons expressing WT (black), DN-mutant (blue), or KW-mutant Munc13-2 (red; means ± SEMs). In the normalized plots (left panels), p<0.001 for WT compared to DN- and KW-mutant Munc13-2; in the absolute responses (right panels), the initial responses are significantly larger (p<0.01) for the KW-mutant Munc13-2 compared to the WT or DN-mutant protein, whereas the final responses are significantly smaller (p<0.001) for the DN-mutant compared to the WT and KW-mutant Munc13-2 (2.5 Hz, WT n=18; KW n=21; DN; n=16; 10 Hz, WT n=50; DN n=41; KW n=64).

c., d. Normalized (c) and absolute EPSC amplitudes (d) in response to a low-frequency stimulus train (0.2 Hz) that is interrupted by a 5 sec 10 Hz stimulus train to induce augmentation (gray area)⁴⁷. Munc13-deficient neurons expressing WT (black), DN-mutant (blue), or KW-mutant Munc13-2 (red) were analyzed (for normalized responses, degree of augmentation is significantly higher (p<0.001) for WT compared to DN- and KW-mutant Munc13-2; for absolute responses, all three Munc13 forms differ significantly from each other at the p<0.001 level (WT n=50; DN n=41; KW n=64)).

e. Relative potentiation by PDBu (1 μM) of EPSC amplitudes evoked at 0.2 Hz in Munc13-deficient neurons expressing WT (black), DN-mutant (blue), or KW-mutant Munc13-2 (red). The relative PDBu potentiation was significantly lower (p<0.001) in synapses expressing KW-mutant Munc13-2 than in synapses expressing WT or DN-mutant Munc13-2 (WT, n=30; DN, n=31; KW, n=43).

f. Plot of the degree of PDBu potentiation as a function of the initial vesicular release probability (P_vr) in individual neurons. Each individual data point represents a Munc13-deficient neuron expressing WT (black), DN-mutant (blue), or KW-mutant Munc13-2 (red). The solid symbols represent the mean values for each group.

Strikingly, plots of absolute synaptic responses indicated that the analysis of normalized responses is misleading. Specifically, synapses expressing KW-mutant Munc13-2 started off with a much higher absolute EPSC value than synapses expressing WT Munc13-2, and exhibited continuously larger EPSCs, whereas synapses expressing DN-mutant Munc13-2 stated off with an unchanged EPSC value, but experienced more severe synaptic depression during the stimulus trains (right panels, Figs. 6b and 6d). Thus, in synapses containing KW-mutant Munc13-2, the initially increased release probability leads to a faster depletion of the RRP and apparent depression. In contrast, in synapses expressing DN-mutant Munc13-2, the initial release probability is normal, and depression develops because the accumulating Ca²⁺ that normally augments release by binding to the Munc13 C₂B-domain can no longer bind to the domain.

To confirm these conclusions, we analyzed a second, related form of short-term synaptic plasticity: augmentation of synaptic responses observed after a short high-frequency stimulus train. We measured synaptic responses before and after application of a 5 s, 10 Hz stimulus train and again analyzed normalized and absolute EPSC amplitudes (Figs. 6e and 6f). Plots of normalized responses showed that augmentation was largest in synapses expressing wild-type Munc13-2 (2.3 ± 0.1 fold, n=77), but was impaired in synapses expressing DN- or KW-mutant Munc13-2 (1.3 ± 0.1, n=71, p<0.001; 1.5 ± 0.1, n=93, p<0.001). Plots of absolute responses, however, showed that DN-mutant expressing synapses exhibited a true loss of augmentation, whereas the apparent loss of augmentation in KW-mutant expressing synapses was not present since the synapses start from an enhanced `plateau' (Fig. 6f). Interestingly, the loss of augmentation in synapses expressing DN-mutant Munc13-2 applied only to the very initial phase; later in the stimulus train, responses recovered, consistent with the notion that multiple Ca²⁺-regulators mediate augmentation^{13,16–20,50}.

Our data suggest that the main mechanism by which Ca²⁺-binding to the Munc13 C₂B-domain mediates synaptic augmentation involves a change in vesicle release probability, as the gain-of-function KW mutant has a two-fold higher initial release probability. To test further whether additional effects on vesicle repriming could explain the phenotype, we measured the recovery of the RRP after vesicle depletion by a high-frequency stimulus train (40 Hz for 2.5 s). Although there was a trend towards a slower EPSC recovery in synapses expressing DN- or KW-mutant Munc13-2, this effect was not statistically significant (Supplementary Fig. 8).

Synaptic but not non-synaptic Munc13 isoforms contain adjacent C₁- and C₂-domains similar to PKC, where they cooperate with each other⁵¹. At a synapse, phorbol esters increase the presynaptic vesicle release probability without changing the RRP size^47,52, at least in part by activating Munc13 (refs. 52 and 53). To test whether the Munc13 C₁- and C₂-domains also cooperate, we analyzed the effect of phorbol esters (1 μM PDBu applied for 1 min). We found that the relative potentiation by PDBu was significantly decreased in synapses expressing KW-mutant Munc13-2, whereas synapses expressing WT or DN-mutant Munc13-2 displayed similar degrees of potentiation (Figs. 6g). Plots of the relative potentiation of release by PDBu against the vesicular release probability for individual synapses revealed an inverse correlation (Fig 6h), indicating that the increased vesicular release probability caused by the KW-mutation occludes the PDBu potentiation.

DISCUSSION

Munc13's are essential components of the synaptic release machinery that prime synaptic vesicles for exocytosis, and regulate short-term plasticity of synaptic exocytosis^{10,13,47,53,54}, but their mechanisms of action remain unclear. Here, we show that the central C₂B-domain of Munc13's exhibits unusual Ca²⁺- and PIP/PIP₂-dependent phospholipid binding properties. These properties structurally correlate with a unique accessory α-helix of the Munc13 C₂B-domain that is part of its Ca²⁺-binding site. The unusual properties of the Munc13 C₂B-domain enable Munc13's to mediate Ca²⁺-dependent augmentation of synaptic vesicle exocytosis during high-frequency trains of action potentials. As described below, we believe that this augmentation is likely based on the Ca²⁺-dependent binding of the Munc13 C₂B-domain to the plasma membrane, which in turn is enabled by increased synthesis of PIP and PIP₂ induced by accumulating Ca²⁺ during a high-frequency action potential train.

Properties of the Munc13 C₂B-domain

Structurally, the Munc13 C₂B-domain is composed of a standard C₂-domain β-sandwich in which aspartate residues in the `top' loops coordinate two Ca²⁺-ions (Figs. 1 and 2). A distinctive feature of the Munc13 C₂B-domain is Ca²⁺-binding loop 3, which includes an extended sequence that folds into an amphipathic α-helix (Fig. 2). Biochemically, the Munc13 C₂B-domain binds to phospholipids in a Ca²⁺-dependent manner similar to other C₂-domains, but differs from other known C₂-domains, such as those from synaptotagmin, in that Ca²⁺-dependent phospholipid binding requires relatively high concentrations of PIP or PIP₂ (Figs. 3 and 4). This unusual biochemical property is likely mediated, at least in part, by the unique α-helix formed by Ca²⁺-binding loop 3, which contains a highly positively charged region that may act similarly to the PIP₂-dependent amphipathic α-helix observed in epsin⁵⁵.

Our data show that in addition to the amphipathic α-helix, the conserved, positively charged K630 residue in Ca²⁺-binding loop 1 is a major determinant of the unusual phospholipid binding properties of the Munc13 C₂B-domain. In the Ca²⁺-binding loops of synaptotagmin-1 C₂-domains, the residues analogous to K630 of Munc13 are hydrophobic (M173 and V304; Fig. 1a). During Ca²⁺-dependent phospholipid binding of synaptotagmin-1 C₂-domains, M173 and V304 insert into the phospholipid bilayer, and contribute to the relatively non-specific but tight Ca²⁺-dependent phospholipid-binding of these C₂-domains^38–41. Moreover, exchanging M173 and V304 in the synaptotagmin-1 C₂-domains for tryptophan further enhances their Ca²⁺-dependent phospholipid binding, indicating that tryptophan increases Ca²⁺-dependent phospholipid binding mediated by hydrophobic residues^40,41,56. These observations suggested to us that K630 in the Munc13 C₂B-domain may contribute to its unique specificity for high PIP/PIP₂-concentrations. To test this hypothesis, we substituted K630 of the Munc13 C₂B-domain for tryptophan, resulting in the KW-mutation. The KW-mutation rendered the Munc13 C₂B-domain responsive to PIP₂ at a concentration at which WT Munc13 is inert but synaptotagmin-1 is active, and thus confers synaptotagmin-like properties onto the Munc13 C₂B-domain (Fig. 4). As a result, the KW-mutation constitutes a gain-of-function mutation that enables Ca²⁺-dependent binding of the Munc13 C₂B-domain to phospholipid membranes containing lower PIP/PIP₂-concentrations than the Munc13 C₂B-domain would normally bind to.

The C₂B-domain regulates synaptic plasticity

Strikingly, abolishing Ca²⁺-binding to the Munc13-2 C₂B-domain with the DN-mutation did not alter vesicle exocytosis triggered by isolated action potentials (Fig. 5), but impaired facilitation of synaptic vesicle exocytosis induced by repeated action potentials (Fig. 6). Thus, the Munc13 C₂B-domain acts as a Ca²⁺-regulator of short-term synaptic plasticity, consistent with the notion that synaptic facilitation during high-frequency stimulus trains is not passively caused by residual Ca²⁺, but actively induced by Ca²⁺-binding to the Munc13 C₂B-domain (and other Ca²⁺-binding proteins). The KW-mutation, in contrast, increased the amount of Ca²⁺-triggered release during single and repeated action potentials (Figs. 5 and 6), without itself acting as a Ca²⁺-sensor for release (Supplementary Fig. 7).

Viewed together, our data suggests that during isolated action potentials, the lower PIP/PIP₂-content at rest prevents Ca²⁺-dependent binding of the WT Munc13 C₂B-domain to the membrane. During stimulus trains, a Ca²⁺-dependent phosphatidylinositol kinase may increase the presynaptic PIP/PIP₂-content, allowing residual Ca²⁺ to activate Munc13 and thereby to boost release. Consistent with this hypothesis, Ca²⁺ stimulates PIP and PIP₂ synthesis in neuroendocrine cells⁵⁷, and depolarization of neurons activates presynaptic PIP kinase Iγ by dephosphorylation, thereby increasing the plasma membrane PIP/PIP₂-content^58–60. Our hypothesis explains why the DN-mutation has no effect on Munc13 function during isolated action potentials, but interferes with Munc13 function during repeated action potentials. The hypothesis also accounts for the gain-of-function effect of the KW-mutation, since the PIP/PIP₂-content at rest is proposed to be too low to allow Ca²⁺-activation of the WT Munc13 C₂B-domain, but may suffice for Ca²⁺-activation of the KW-mutant C₂B-domain.

Mechanism of the Munc13 C₂B-domain action

In boosting release during a stimulus train, Ca²⁺-binding to the Munc13 C₂B-domain likely increases vesicle priming by enhancing the priming function of Munc13. The KW-mutation may enable the Munc13 C₂B-domain to perform the same activity even at rest, possibly because the KW-mutant C₂B-domain binds to the plasma membrane even without the increase in PIP and PIP₂ concentrations that is thought to occur during repeated action potentials^57–60.

However, two of our findings with KW-mutant Munc13 appear to argue against the hypothesis that the Munc13 C₂B-domain boosts the priming function of Munc13:

1. the KW-mutation selectively increased the Ca²⁺-sensitivity of release induced by isolated action potentials without increasing the size of the RRP (Fig. 5)

2. the KW-mutation did not significantly alter the rate by which evoked release recovered after depletion of the RRP, although there was a trend towards acceleration of priming (Supplementary Fig. 8)

Despite these findings, we believe that the priming hypothesis is correct for the following reasons: Priming likely involves a partial, if not complete assembly of SNARE-complexes between vesicles and the plasma membrane⁶¹. The number of assembled SNARE complexes per primed vesicle may determine (among others) the apparent Ca²⁺-affinity of synaptic vesicle fusion⁶². Thus, Ca²⁺-binding to the Munc13 C₂B-domain may increase the ability of Munc13 to catalyze SNARE-complex assembly of docked vesicles during priming, which would result in the appearance of an increased Ca²⁺-sensitivity of release for the KW-mutation. This hypothesis is consistent with the fact that KW-mutant Munc13 does not act as a Ca²⁺-sensor for exocytosis itself (Supplementary Fig. 7). Although plausible, the priming hypothesis cannot be tested directly until its underlying tenet – namely that Munc13 mediates vesicle priming by catalyzing SNARE-complex assembly – has been confirmed⁶³.

How does increased phospholipid binding induced by Ca²⁺-binding to the Munc13 C₂B-domain potentiate Ca²⁺-triggered release, be it via priming or otherwise? As a component of the biochemically insoluble active zone, Munc13 already is normally close to the plasma membrane. Thus, Ca²⁺-dependent C₂B-domain binding to the plasma membrane would not re-localize Munc13, but rather pull on the adjacent plasma membrane and stretch it. Such an activity may, analogous to what has been proposed for the mechanism of action of synaptotagmin⁶⁴, promote exocytosis by decreasing the energy requirement for Ca²⁺-triggered fusion-pore opening. DAG-binding to the C1-domain of Munc13 – which also induces Munc13 plasma membrane binding – may potentiate release by an analogous, but Ca²⁺-independent mechanism. An alternative hypothesis is that the Munc13 C₁- and C₂B-domains are normally inhibitory, and that DAG- and Ca²⁺-binding reverse this inhibition⁵⁴. This second hypothesis would require that the C₁- and the C₂B-domain have additional unknown biochemical interactions beyond lipid binding which are altered by the various mutations, a possibility that remains to be explored.

Munc13 as a computational unit for synaptic transmission

The activity of synaptic Munc13 isoforms is regulated via three distinct, adjacent signaling motifs: the previously described calmodulin-binding sequence and C₁-domain^12,13, and the C₂B-domain we characterize here (Fig. 7). All three motifs are directly or indirectly activated by Ca²⁺, and have profound roles in controlling neurotransmitter release during short-term plasticity, but differ from each other in their mechanisms of activation and action. The C₂B-domain is directly Ca²⁺-activated by Ca²⁺-influx during action potentials, but is presumably only stimulated after accumulating residual Ca²⁺ has induced the synthesis of PIP/PIP₂. In contrast, the calmodulin-binding sequence is indirectly activated by binding of accumulating residual Ca²⁺ to calmodulin, which likely acts on many synaptic targets simultaneously. The C₁-domain is activated indirectly via Ca²⁺-dependent induction of phospholipase C. Thus, the Ca²⁺-concentration dependence and time course of activation of the three regulatory motifs likely differ, leading to a common readout (synaptic potentiation) that results from the integration of multiple signals acting differentially on the three signaling motifs. Consistent with this model, the KW-mutation of Munc13 occludes the effect of phorbol esters on release triggered by isolated action potentials, whereas the DN-mutation (which has no effect on release triggered by isolated action potentials) has no effect on the phorbol ester potentiation under those conditions (Fig. 6). Moreover, at least mutations in the calmodulin-binding motif and the C₂B-domain act additively during short-term synaptic plasticity (Supplementary Fig. 9).

Figure 7. Model for the Ca²⁺-regulation of short-term plasticity by Munc13.

The top diagrams depict the domain structures of the three sub-families of Munc13 proteins, the two classes of long Munc13s expressed primarily in brain, and the class of short Munc13s expressed primarily in peripheral organs. Top arrows illustrate a possible regulation of the N-terminal RIM-binding sequences and the C-terminal MUN-domain of Munc13s by ligand-binding to the central C₁- and C₂-domains. The central regulatory domains of Munc13's are illustrated below the domain diagrams: the calmodulin-binding sequence found in Munc13-1 and bMunc13-2, the DAG-binding C₁-domain found in all variants of Munc13-1, -2, and -3 but not the ubiquitous Munc13 isoforms, and the Ca²⁺-binding C₂B-domain that is universally present in all neuronal and ubiquitous Munc13 isoforms. Note that in addition to binding to the C₂B-domain, Ca²⁺ also serves to stimulate the production of DAG from PIP₂ on the one hand, and the synthesis of DAG on the other hand.

In summary, our data establish that the Munc13 C₂B-domain operates as a Ca²⁺-regulator of short-term synaptic plasticity. Apart from the importance of the Munc13 C₂B-domain Ca²⁺-binding properties for synaptic exocytosis, our results also suggest that this domain may act as a Ca²⁺-regulator of exocytosis for non-synaptic Munc13 isoforms, which likely function in other forms of exocytosis (such as Munc13-4 in lymphocyte exocytosis²⁷).

METHODS

Plasmids

Four different types of plasmids encoding rat Munc13-1 and 13-2 were used for this study: i. pGEX-KT derived plasmids for bacterial expression. ii. pFastBac™1 (Invitrogen life technologies) derived baculovirus expression plasmids into which we inserted the GST-coding region. The resulting plasmid was used for cloning pFastBac-GST-Munc13-1 C₂B (residues 675–820; numbering based on U24070); pFastBac-GST-Munc13-2 C₂B (residue 599–744; numbering based on AF159706); pFastBac-GST-Munc13-2 C₂B-DN (D629N and D635N); and pFastBac-GST-Munc13-2 C₂B-KW (K630W); iii. Semliki Forest Virus expression plasmids (pSFV ubMunc13-2 WT, pSFVubMunc13-2 C₂B-D629N,D635N, pSFVub Munc13-2K630W) for neuronal cultures. iv. Lentiviral expression plasmid pFUW-Munc13-2 K603W that encodes ubMunc13-2 with the KW mutation.

Production of recombinant proteins

1. Bacterial expression

GST fusion proteins were expressed at 25 °C in E. coli BL21, isolated by affinity chromatography on glutathione-Sepharose followed by on-resin cleavage with thrombin, and cleaved proteins were further purified by ion-exchange and gel-filtration chromatography on MonoS and S75 columns (Amersham). Uniform ¹⁵N-labeling was achieved by growing the bacteria in ¹⁵NH₄Cl as the sole nitrogen source.

2. Bacuolovirus expression

Munc13 C₂B-domain proteins were generated using Bac-to-Bac® Baculovirus Expression System (Invitrogen Life Technologies). Recombinant GST-Munc13 C₂B proteins were generated by infecting 400 ml of Sf9 cells (~2×10⁶ cells/ml) in a 2 liter flask after inoculating 20 ml of baculovirus (~10⁷ pfu/ml) for 3 days at 28 °C. Recombinant GST-fusion proteins were purified using glutathione-Sepharose™ 4B beads (Amersham Biosciences).

Antibodies were either described previously^2,3 or obtained from Synaptic Systems GmbH (Göttingen, Germany), except when stated otherwise. SDS-PAGE and immunoblotting were performed as described³.

Fluorescence spectroscopy

Fluorescence spectra were recorded at 290 nm excitation in an LS55 luminescence spectrometer (PerkinElmer Life Sciences) with 0.3 μM purified Munc13 C₂B-domains in 0.5 ml of 20 mM HEPES-NaOH pH 7.2 and 0.1 M NaCl, with the indicated additions. All buffers were passed through the AG MP-50 resin (Bio-Rad) to eliminate contaminating Ca²⁺.

NMR spectroscopy was carried out at 25 °C on Varian INOVA500 or INOVA600 spectrometers with samples containing ~120 μM Munc13-1 C₂B-domain (residues 675–820) in 20 mM MES-NaOH pH 6.2, 0.1 M NaCl, and 0.5 mM TCEP. Ca²⁺-titrations were monitored by ¹H-¹⁵N HSQC experiments as described³⁰. All NMR data were processed with NMRPipe⁶⁵ and analyzed with NMRView⁶⁶.

X-ray crystallography

Purified Munc13-1 C₂B-domain (residues 675–820; in 20 mM MES-NaOH pH 6.2, 0.1 M NaCl and 0.5 mM TCEP) was concentrated to 20 mg/ml, and crystallized in 30% (w/v) PEG-MME 2000 and 0.1 M Bis-Tris-Propane pH 6.8 at 20 °C without Ca²⁺ or with 0.01–0.1 M CaCl₂ using the hanging-drop vapor-diffusion method. Ca²⁺-free Munc13-1 C₂B-domain crystals appeared in two days, grew to a final size of 0.08 × 0.05 × 0.15 mm within a week, and were transferred into a solution of 35% (w/v) PEG-MME 2000, 0.1 M Bis-Tris-Propane pH 6.8 and 5% (v/v) glycerol, and flash-cooled in liquid propane. Ca²⁺-bound Munc13-1 C₂B-domain crystals appeared within one week as needle-clusters and gradually transformed into a diamond-like shape over 3–4 weeks, with a final size of ~0.1 mm, after which they were transferred into a solution of 32% PEG-MME 2000, 0.1 M Bis-Tris-Propane pH 6.8, 0.1 M NaCl, 10% glycerol, 1 mM CaCl₂ and flash-cooled in liquid propane. Diffraction data were collected at the Structural Biology Center beamlines 19BM and 19ID of the Advanced Photon Source at 100K. The diffraction of these crystals was highly anisotropic, leading to a gradual decrease in the completeness at resolutions above 2.2 Å. Crystals of the Ca²⁺-free Munc13-1 C₂B-domain diffracted to a d_min of ~1.89 Å (space group C222₁; unit cell parameters a = 43 Å, b = 101 Å, c = 68 Å; 1 molecule per asymmetric unit). Crystals of the Ca²⁺-bound Munc13-1 C₂B-domain diffracted to a Bragg spacing (d_min) of ~1.37 Å (space group P4₃2₁2; unit cell parameters a = 57 Å, b = 57 Å, c = 90 Å, 1 molecule per asymmetric unit). Data were processed in the HKL2000 program suite⁶⁷. The structure of the Ca²⁺-free Munc13-1 C₂B-domain was determined via molecular replacement with the program AMoRe⁶⁸ using the synaptotagmin-1 C₂A-domain (PDB code 1rsy) as search model. The Ca²⁺-bound Munc13-1 C₂B-domain structure was determined via molecular replacement with the program Phaser⁶⁹ using the final structure of Ca²⁺-free Munc13-1 C₂B-domain as search model. Models were completed using the program Arp/Warp⁷⁰, followed by manual adjustments with the program O⁷¹, and refinements with the program Refmac⁷² of the CCP4 package⁷³, with a random subset of all data set aside for the calculation of free R factors. After complete refinement, solvent molecules were added where chemically reasonable. The final model for the Ca²⁺-bound Munc13-1 C₂B-domain contains residues 673–675 and 687–819, two Ca²⁺-ions, two Cl⁻-ions, one glycerol molecule, and 161 water molecules (final R = 17.0; R_free = 19.3; overall B-factor = 16.7). The final model for the Ca²⁺-free Munc13-1 C₂B-domain contains residues 678–705, 708–763, 773–801 and 807–819, one molecule of Bis-Tris-Propane, 2 Cl⁻-ions, and 63 water molecules (final R = 21.7; R_free = = 28.2; overall B-factor = 41.0). For data collection and refinement statistics, see Table 1.

Phospholipid binding assays

1. Centrifugation assay

Phospholipids and cholesterol (obtained from Avanti) were dissolved in chloroform:methanol (1:1; cholesterol and PIPs) or chloroform (PS, PI, PC, and PE), mixed in a `synaptic' composition (41% PC, 32% PE, 12% PS, 5% PI, and 10% cholesterol with or without additional PIP and PIP₂; phospholipid composition approximates that of synaptic vesicles^13–15), and dried under nitrogen. Lipids were resuspended by vortexing 175 mg lipid mixture for 20 min in 3.5 ml HEPES buffer (50 mM HEPES-NaOH pH 6.8, 0.1 M NaCl, and 4 mM EGTA) containing 0.5 M sucrose. Lipids were sonicated for 5 min in a bath sonicator (model G112SP1G; Laboratory Supply Co. Inc.), 14 ml HEPES buffer without sucrose were added, and liposomes were centrifuged at 100,000g for 30 min to separate heavy liposomes from free phospholipids. Liposomes were washed once, and repelleted (20,800g for 10 min). Binding assays were performed in 1 ml with 10 μg recombinant GST-fusion proteins (~0.125 μM) and 100 μg liposomes in HEPES buffer containing 2 mM MgCl₂ and various Ca²⁺-concentrations that result in the indicated free [Ca²⁺] as calculated with EqCal software (Biosoft, Ferguson, MI). Binding reactions were incubated 10 min at 30 °C with 800 rpm shaking, pelleted by centrifugation (20,800g for 10 min), and washed three times with 1 ml of the corresponding buffers. Final liposome pellets were dissolved in chloroform:methanol (1:2, v/v). Precipitated proteins were recovered by centrifugation (20,800g for 15 min), resuspended in 30 μl of 2× SDS sample buffer, and analyzed by SDS-PAGE and Coomassie Blue staining^2,16. Bound proteins were quantified analysis of scanned Coomassie-stained gels with the Image Quant program (version 5.2, Molecular Dynamics).

2. FRET assays

were performed essentially as described¹⁷ at room temperature in 0.5 ml of 20 ml HEPES-NaOH pH 7.2, 0.1 M NaCl, with 1 μM of Munc13-2 C₂B-domain protein and 30 μg/ml liposomes containing 41% PC, 22% PE, 10% dansyl-PE, 12% PS, 5% PI, 10% cholesterol, 0.5% PIP, and 0.1% PIP₂. Emission spectra (excitation: 282 nm) were first recorded without metal ions, then after addition of 2 mM MgCl₂, then after addition of 0.2 mM CaCl₂, and then after further addition of 1 mM EGTA.

GST pulldowns were performed as described². One unstripped rat brain (~1.5 g/brain; Pel-Freez Biologicals, Rogers, Arkansas) was homogenized with a tissue homogenizer (Thomas Scientific, Philadelphia, Pennsylvania) in 30 ml HEPES buffer containing 1 mM DTT, 1 mM PMSF, 5 μg/ml leupeptin, and 2 μg/ml aprotinin. 1% Triton X-100 was added, proteins were extracted for 1 hr at 4 °C with rotation, insoluble proteins were removed by centrifugation (100,000g for 1 hr), and the supernatant was used for pulldowns using 30 μg GST-Munc13 C₂B-domain proteins attached to glutathione-Sepharose, and 0.5 ml rat brain lysate. Pulldown reactions (1 ml volume) in HEPES buffer containing 2 mM MgCl₂ and 0.5% Triton X-100 were incubated 1.5 hr at 4 °C with rotation, washed six times with corresponding buffer, and bound proteins were analyzed by SDS-PAGE and immunoblotting.

Electrophysiology

Microdot neuronal cultures were prepared and electrophopysiological analyses were performed at 22–25 °C as described^18–20,74. Currents were recorded using an Axopatch 200B amplifier (Molecular Devices). Series resistance was within 10 MΩ and was electronically compensated at least 70%. Data were analyzed using AXOGRAPH software (version 4.9, Molcular Devices). For analyses of Syt-1 KO mice, primary cortical neurons from Syt-1 KO and littermate wild-type control mice were cultured in Modified Eagle Medium (MEM, Gibco) supplemented with B27 (Gibco), glucose, transferrin, fetal bovine serum, and Ara-C (Sigma)⁷⁴. To monitor synaptic responses, whole-cell patch-clamp recordings were made with neurons at 14–16 days in vitro. Synaptic responses were triggered by a 1 ms current pulse (900 μA) through a local extracellular electrode (FHC, Inc.), and recorded in whole-cell voltage-clamp mode using a Multiclamp 700B amplifier (Axon Instruments, Inc.)⁷⁴. Data were analyzed using Clampfit 9.02 (Axon Instruments, Inc) or Igor 4.0 (Wavemetrics).

Statistical analyses were performed using paired Student's t-test or ANOVA.