|Home | About | Journals | Submit | Contact Us | Français|
Unc13/Munc13s play a crucial function in neurotransmitter release through their MUN domain, which mediates the transition from the syntaxin-1/Munc18-1 complex to the SNARE complex. The MUN domain was suggested to be related to tethering factors, but no MUN-domain structure is available to experimentally validate this notion and address key unresolved questions about the interactions and minimal structural unit required for Unc13/Munc13 function. Here we identify an autonomously folded module within the MUN domain (MUN-CD) and show that its crystal structure is remarkably similar to several tethering factors. We also show that the activity in promoting the syntaxin-1/Munc18-1 to SNARE complex transition is strongly impaired in MUN-CD. These results show that MUN domains and tethering factors indeed belong to the same family and may have a common role in membrane trafficking, and suggest a model whereby the MUN-CD module is central for Munc13 function but full activity requires adjacent sequences.
Inracellular membrane fusion is mediated by members of conserved protein families that underlie a common mechanism of membrane fusion (Jahn and Scheller, 2006; Wickner and Schekman, 2008; Rizo and Rosenmund, 2008; Sudhof and Rothman, 2009). Particularly important for fusion are the SNAREs and Sec1/Munc18 (SM) proteins. Thus, SNAREs residing on two membranes assemble into SNARE complexes through their SNARE motifs, which brings the two membranes together and is key for membrane fusion (Sollner et al., 1993; Hanson et al., 1997; Poirier et al., 1998; Sutton et al., 1998). SM proteins are believed to play a central albeit currently unclear role in membrane fusion through interactions with SNARE complexes (Rizo et al., 2006; Toonen and Verhage, 2007; Sudhof and Rothman, 2009; Carr and Rizo, 2010). Munc18-1, the neuronal SM protein involved in neurotransmitter release, also binds to the SNARE syntaxin-1 folded into a closed conformation involving binding of the syntaxin-1 N-terminal Habc domain (Fernandez et al., 1998) to its SNARE motif (Dulubova et al., 1999). This interaction is not generally conserved in other systems (Carr et al., 1999; Dulubova et al., 2001) and gates entry of syntaxin-1 into SNARE complexes to control neurotransmitter release (Gerber et al., 2008). Although some models assume that SNARE complex assembly leads directly to fusion, strong evidence suggests that the priming step that leaves synaptic vesicles in a release-ready state involves syntaxin-1 opening and partial SNARE complex assembly (Rizo and Rosenmund, 2008; Walter et al., 2010).
Neurotransmitter release also depends on factors with specialized roles to meet its tight regulatory requirements, including invertebrate Unc13 and their vertebrate homologues, Munc13s, among others. These large (ca. 200 kDa) proteins from presynaptic active zones are crucial for priming synaptic vesicles to a release-ready state, and no release is observed in their absence (Augustin et al., 1999; Richmond et al., 1999; Aravamudan et al., 1999; Varoqueaux et al., 2002). This role can be ascribed to an autonomously folded C-terminal region called the MUN domain (Basu et al., 2005) (Figure 1A), although the minimal Unc13/Munc13 sequence required for priming is still unclear (Stevens et al., 2005; Madison et al., 2005). A syntaxin-1 mutant bearing a mutation that destabilizes its closed conformation (Dulubova et al., 1999) partially rescues neurotransmitter release in C. elegans unc13 nulls (Richmond et al., 2001), and the Munc13-1 MUN domain dramatically accelerates the transition from the closed syntaxin-1/Munc18-1 complex to the SNARE complex (Ma et al., 2011), showing that the Unc13/Munc13 priming function entails a direct role in opening syntaxin-1. However, conflicting reports have ascribed this role to binding of Unc13/Munc13s to the syntaxin-1 N-terminus (Betz et al., 1997; Stevens et al., 2005; Madison et al., 2005) or to its SNARE motif (Ma et al., 2011). In addition to their key priming function, Unc13/Munc13s govern varied presynaptic plasticity processes that underlie distinct forms of information processing in the brain (Brose et al., 2000), likely through intramolecular interactions of the MUN domain with other modules (Basu et al., 2005; Basu et al., 2007) and through interactions with other active zone proteins (Betz et al., 2001; Dulubova et al., 2005; Wang et al., 2009). Hence, the vital priming activity of the MUN domain probably connects a multitude of presynaptic signals to exocytosis.
MUN domains are also found in Unc31/CAPS, proteins with functions related to but distinct from those of Unc13/Munc13s (James et al., 2009; Liu et al., 2010), and in diverse proteins from plants and fungi (Koch et al., 2000; Basu et al., 2005). Notably, recent sequence analyses revealed remote homology between MUN domains and subunits of tethering complexes from diverse membrane compartments, such as the exocyst, GARP, Cog and Dsl1p complexes (Pei et al., 2009), and structural data is establishing similarities among these tethering factors (Sivaram et al., 2006; Tripathi et al., 2009). These findings bring up the intriguing possibility that all these proteins form an additional family with a general function in intracellular membrane fusion, in addition to having specialized roles in different membrane compartments. This notion could be fundamental for our overall understanding of the mechanisms of intracellular membrane traffic, but it is critical to obtain experimental support for the similarity of MUN domains to tethering factors, since the predicted sequence identities between them are below 10% (Pei et al., 2009) and no three-dimensional structure of a MUN domain is available.
Obtaining structural information at atomic resolution on the MUN domain of Unc13/Munc13s is also crucial to rationalize multiple data on these proteins that have accumulated over the years. This includes the effects of disease-causing mutations in the Munc13-4 MUN domain (Feldmann et al., 2003), and mutagenesis results that have yielded a confusing picture regarding which is the minimal Unc13/Munc13 sequence required for the vesicle priming function and which interactions with syntaxin-1 are important for this function (see above). To help address these questions and to provide a structural framework to understand Unc13/Munc13 function, we have devoted very extensive efforts to crystallize fragments spanning the Munc13 MUN domain sequences. We have found that the C-terminal half of the Munc13-1 MUN domain constitutes an autonomously folded module (referred to as MUN-CD), and we have solved its crystal structure at 2.7 Å resolution. The structure of MUN-CD is remarkably similar to those of several tethering complex subunits. These results demonstrate that Unc13/Munc13s are indeed related to tethering complexes, supporting the notion that all these proteins belong to a common family. Using NMR spectroscopy, we also show that the MUN-CD module exhibits a much weaker affinity for the syntaxin-1 SNARE motif than the MUN domain, and also a much weaker activity in accelerating the transition from the syntaxin-1/Munc18-1 complex to the SNARE complex. These data, together with examination of the MUN-CD structure, provide clear explanations for the functional results obtained with different Unc13/Munc13 mutants and support the conclusion that the priming function of Unc13/Munc13s is mediated by binding to the syntaxin-1 SNARE motif.
Unc13/Munc13s contain a variable N-terminal region that may include a C2 domain, and a conserved C-terminal region that includes two C2 domains and in most cases a C1 domain (illustrated for rat Munc13-1 in Figure 1A). Definition of structural or functional units within the region between the C2B and C2C domains has been challenging because of its large size and the abundance of helical sequences, which hinders computational comparisons when homology is low. Initial analyses identified two homology regions that were called Munc13-homology-domains (MHD1 and MHD2) and correspond to residues 1106-1249 and 1358-1525 of rat Munc13-1 (Koch et al., 2000), but attempts to express up to 25 Munc13-1 fragments based on these results did not yield any properly folded, well-behaved protein (Basu et al., 2005). Additional computational analyses revealed that the sequence homology in this family extends to most of the region between the C2B and C2C domains; this region was shown to form an autonomously folded unit, which was termed the MUN domain, and to be sufficient to rescue neurotransmitter release in Munc13-1/2 double knockout mice (Basu et al., 2005). However, some degree of vesicle priming was also supported by Unc13 and Munc13-1 fragments that had an N-terminally truncated MUN domain and included part of or the full C2C domain [Unc13 fragment corresponding to residues 891-1688 of rat Munc13-1 (Madison et al., 2005) and rat Munc13-1 fragment spanning residues 1100-1735 (Stevens et al., 2005); see Figure 1A]. These findings emphasized the urgent need to obtain three-dimensional structural information to understand the relationship between structural and functional units in these proteins.
With this goal, we attempted to crystallize multiple fragments spanning the MUN domain of Munc13-1 (Figure 1B), aided in part by NMR experiments to check that the expressed fragments are properly folded and well behaved [see (Chen et al., 2002; Lu et al., 2006)]. We started with the initial fragment identified as the Munc13-1 MUN domain [residues 859-1531; (Basu et al., 2005)], which yields highly-quality NMR spectra [Figure S1A]. Crystals were obtained under diverse conditions, but the best crystals diffracted only to a dmin of 6-7 Å, and changing the N-terminus of the fragment did not help to get better crystals (Figure 1B). Interestingly, we observed that the protein in the crystals was fully cleaved at residue Phe1444, presumably by a residual protease, and mutating this residue to Ala or Gln hindered crystallization. This residue is located in a predicted long loop that exhibits poor sequence conservation and alternative splicing (Brose et al., 1995). We attempted to co-express fragments corresponding to the proteolytic products, and we also made several MUN domain fragments with different deletions within this loop, which in most cases improved the solubility and expression of the fragment, but no crystals were obtained with these proteins (Figure 1B). We also attempted to crystallize similar fragments of Munc13-2 and Munc13-3, but the former were largely insoluble and the latter lead to poorly diffracting crystals.
The observation of a remote sequence homology between MUN domains and some tethering factors (Pei et al., 2009) and the available crystal structures of several of these factors (Dong et al., 2005; Sivaram et al., 2006; Wu et al., 2005; Tripathi et al., 2009; Ren et al., 2009) led us to hypothesize that, similar to these factors, the Munc13-1 MUN domain may contain four subdomains (named A-D; Figure 1A). Since some of the structures of tethering factors were obtained with fragments containing only the A-B or C-D subdomains, we prepared new fragments spanning the predicted A-B and C-D subdomains of rat Munc13-1. While we were unable to express A-B fragments (residues 859-1149 or similar) in soluble form, we obtained high yields of soluble fragments corresponding to the C-D subdomains (residues 1148-1516 or 1148-1531) (Figure 1B). These soluble proteins still did not yield crystals, but removal of the entire long loop in the Munc13-1(1148-1516) fragment (resulting in the 1148-1407,EF,1464-1516 fragment) did lead to needle clusters (Figure S1C). A similar fragment where we kept 11 residues from the loop (1148-1407,EF,1453-1516 fragment) was highly soluble, exhibited high quality NMR spectra (Figure S1B) and yielded crystals that after optimization diffracted to a dmin of 2.75 Å resolution (Figure S1D). Unfortunately, the diffraction data that we obtained with crystals of seleno-methionine labeled protein revealed multiple lattices, crystal twinning, data anisotropy, low anomalous signal and severe radiation damage, which prevented us from deriving a structure from these data.
Finally, we were able to obtain improved crystals (Figure S1E) that exhibited P42 crystallographic symmetry and diffracted to a dmin of 2.65 Å resolution with a similar fragment extended at the C-terminus (1148-1407,EF,1453-1531 fragment) (Figure 1B). The diffraction data obtained with this fragment still exhibited anisotropy and crystallographic pseudosymmetry, but were no longer subject to severe radiation damage, twinning or multiple lattices. A structure solution was achieved by exploiting the strong crystallographic pseudosymmetry. Data obtained from the seleno-methionine labeled protein collected at the anomalous absorption edge for selenium was indexed in the pseudosymmetric I4 cell and used for heavy atom location, anomalous phasing and automated model building. Final refinement of the model versus native data processed in the crystallographic P42 cell yielded two molecules of MUN-CD and 99 waters. Data collection, phasing and refinement statistics are listed in Table 1. Because this fragment spans an autonomously folded module and contains the predicted C-D subdomains, we refer to this fragment as the MUN-CD module.
The contacts between the two molecules that form the asymmetric unit of the MUN-CD crystals (Figure 2A) are rather limited (380 Å2 buried surface area). Each molecule engages in an additional contact with a crystallographic symmetry mate (700 Å2 buried surface area) that is primarily mediated by residues 1400-1407,E of the truncated loop. MUN-CD is monomeric in solution based on NMR and dynamic light scattering data, indicating that these contacts are induced by crystallization and are unlikely to be physiologically relevant. MUN-CD has an elongated structure (ca. 90 Å from end to end) containing two lobes that correspond to the predicted C and D subdomains (Figure 2B). Both subdomains form five-helix bundles, but there are a total of only nine helices (numbered α1 to α9 in Figure 2B) because a very long, continuous helix (α5) spans the two subdomains. Consecutive helices pack against each other in an antiparallel fashion throughout the MUN-CD module, while alternating helices (α1 vs. α3, α2 vs α4, etc.) are packed in a parallel arrangement. Some of the loops connecting the helices did not yield interpretable electron density, suggesting that they are disordered. Part of the loop connecting helices α7 and α8 that contains the engineered deletion is disordered, but some residues of this loop are observable and participate in crystal contacts (see above), thus providing an explanation as to why alterations in this region impact the quality of the crystals.
There are intimate contacts between subdomains C and D, and the root mean square (r.m.s.) deviation between the two molecules of the asymmetric unit is 0.5 Å for 294 common Cα atoms, indicating that the relative orientation between the two subdomains is rigid. The long axes of the two subdomains cross at an angle of approximately 130°, leading to a boomerang-like shape with a shallow concave surface at the bottom in the orientation shown in Figure 2C. Accordingly, helix α5 bends and changes direction gradually as it crosses from subdomain C to subdomain D (Figure 2B). Helix α6 exhibits a sharp bend, likely to accommodate the bending of helix α5, and helix α9 also has a sharp bend with a non-helical turn connecting the two sides of the helix (see * in Figure 2B). Each one of these helix bends contains one proline residue in the corner, which likely causes the change in direction.
The surface of the MUN-CD module is considerably charged, with a predominance of negative charges; however, there are a few positive patches that might be involved in membrane binding [see (Guan et al., 2008)], particularly at one side of the concave bottom surface (Figure 2C). No large, exposed areas with a hydrophobic character can be observed on the surface of the MUN-CD module, with the exception of a hydrophobic patch at the N-terminus that is most likely involved in contacts with the B subdomain in the full-length protein (Figure S2A).
In agreement with the previous predictions (Pei et al., 2009), the structures in the Protein Data Bank (PDB) with closest similarity to MUN-CD found using DALI (Holm and Sander, 1993) correspond mostly to subunits of vesicle tethering complexes. MUN-CD also displays similarities with myosins that tether vesicles to actin filaments and had been previously found to resemble the tethering factors (Tripathi et al., 2009). The highest structural similarity with MUN-CD found is for the exocyst complex subunit Sec6p (PDB code 2FJI), with a Z score of 18.9 and an r.m.s. deviation of 2.7 Å for 261 equivalent α carbons (Figure 3A). This represents a remarkable similarity considering that the sequences of MUN-CD and Sec6p share less than 10% sequence identity based on the resulting structure-based sequence alignment (Figure 4).
Sec6p was indeed predicted to be the tethering factor most closely related to MUN domains, and a homology model of MUN-CD built based on this finding (Pei et al., 2009) also exhibits a clear similarity with the crystal structure of the MUN-CD described here, although with some important differences (Figure S3A). A superposition of the model with the crystal structure using DaliLite (Holm and Park, 2000) yielded a 3.0 Å r.m.s. deviation for 263 equivalent α carbons. The approximate ends and register of five of the helices (α1, α2, α4, α7 and α8) were correctly predicted by the model, but helices α3 and α5 were shifted in register by 4 and 7 residues, respectively. These shifts reveal discrepancies with the sequence alignment obtained previously (Pei et al., 2009) and together with other structure-based adjustments lead to the revised alignment of Figure 4. Note also that the computational model did not contain the bends in helices α5, α6 and α9 of the MUN-CD structure (Figure S3A). Naturally, these structural differences directly reflect distinctions between the structure of MUN-CD and that of Sec6p (Figure 3A), which was used to build the model and does not have these helical bends. The proline residues in the corners of the helical bends of the MUN-CD structure (Pro1323, Pro1363 and Pro1502) are not conserved in Sec6p or other tethering factors and exhibit different degrees of conservation in Unc13/Munc13s (Figure 4), suggesting that the helical bends are unique features of a subset of Unc13/Munc13 MUN-CD modules.
Varied degrees of similarity between MUN-CD and other tethering factors were found with DALI (illustrated in Figure 3). The Tip20p subunit of the Dsl1p complex (PDB code 3FHN) also exhibits a high similarity to MUN-CD (Figure 3B), with a Z score of 17.0 and a 3.3 Å r.m.s. deviation for 253 equivalent α carbons. Clear similarities with somewhat worse r.m.s. deviations were observed for other tethering factors, including the Sec15p subunit of the exocyst complex (PDB code 2A2F; Z score 14.3; 5.6 Å r.m.s. deviation for 247 equivalent α carbons; Figure 3C) and the Dsl1p subunit of the Dsl1p complex (PDB code 3K8P; Z score 11.9; 6.2 Å r.m.s. deviation for 231 equivalent α carbons; Figure 3D), while the Exo70p and Exo84p exocyst subunits are more distantly related. These higher r.m.s. deviations reflect in part some defined structural distinctions, but they also arise from different relative orientations of the C and D subdomains.
Helix α5, which connects the two subdomains, is the most variable feature of all these structures. It is thus not surprising that the correct register and length of this helix is difficult to predict from sequence analyses. Note that the seven-residue shift in the register of helix α5 in the computational model of MUN-CD (Pei et al., 2009) with respect to the crystal structure still preserves the orientation of the hydrophobic side chains toward the core of the structure, and that the hydrophobicity patterns are conserved features in all these structures (Figure 4) that are key to establish their remote sequence homologies by sequence analyses. In the sequence alignment of Figure 4 we have also highlighted ten residues that were found to be very highly conserved in MUN domains (Pei et al., 2009) (white with red background), residues that are conserved in MUN domains but not in tethering factors (orange), and residues that appear to be conserved in Unc13/Munc13 MUN domains but not in other MUN domains (blue). In general, the highly conserved residues are mostly buried and are probably key to maintain correct folding; W1165 is exposed in MUN-CD but is expected to pack against the B subdomain in the full-length protein, while D1204 is partially exposed (Figure S3B) and might participate in a key interaction of the MUN domain. Residues that are selectively conserved in MUN domains or in Unc13/Munc13 MUN domains are evenly distributed over the MUN-CD structure (Figure S3C).
While the Munc13-1 MUN domain robustly rescues vesicle priming in Munc13-1/2 double knockout neurons, a shorter fragment spanning residues 1045-1531 (Figure 1A) rescues only 1% of the readily-releasable pool (Basu et al., 2005). These findings show that truncation of the predicted A and B subdomains strongly impairs MUN domain function. Since the MUN domain is thought to underlie priming by accelerating the transition from the closed syntaxin-1/Munc18-1 complex to the SNARE complex, and this activity depends on interactions of the MUN domain with the syntaxin-1 SNARE motif and probably with Munc18-1 (Ma et al., 2011) (see model of Figure 5), we examined whether these properties are impaired in MUN-CD using the same methodology.
1H-13C HMQC spectra of MUN-CD 1H,13C-labeled at the Ile, Leu and Val methyl groups in a perdeuterated background (2H-ILV-13CH3-MUN-CD) (39 kDa) exhibited progressive decreases in cross-peak intensities upon addition of Munc18-1 (67 kDa), SNARE complex formed (55 kDa) or both (Figures 6A,B). These data show that MUN-CD binds to Munc18-1 very weakly and to the SNARE complex somewhat tighter but still weakly, and that there is some cooperativity in these interactions. These results are similar to those obtained with the MUN domain, but the SNARE complex induced stronger decreases in cross-peak intensities for the MUN domain (Ma et al., 2011), showing that binding to the SNARE complex is impaired in MUN-CD. Moreover, MUN-CD caused perturbations (broadening and slight shifts) on the same 1H-15N HSQC cross-peaks of the syntaxin-1 SNARE motif that were perturbed by the MUN domain (Figure 6C), but these perturbations were considerably smaller than those induced by the MUN domain (Ma et al., 2011), suggesting that MUN-CD binds more weakly to the syntaxin-1 SNARE motif. This conclusion was confirmed through titrations of 15N-syntaxin-1 SNARE motif with MUN-CD monitored by 1D 1H-15N HSQC spectra (Arac et al., 2003; Dulubova et al., 2005), which did not reach saturation even at 60 μM MUN-CD (Figure 6D). We estimate that the Kd is larger than 300 μM [compared to 46 μM for the MUN domain (Ma et al., 2011)].
To monitor the transition from the syntaxin-1/Munc18-1 complex to the SNARE complex, we used 1H-13C HMQC spectra of perdeuterated cytoplasmic region of syntaxin-1 where the Ile methyl groups were 1H,13C-labeled [2H-I-13CH3-syntaxin-1], which allow to easily distinguish the cross-peaks from the two complexes (illustrated for the Ile203 cross-peak in Figure 7A, left panel) (Ma et al., 2011). Consecutive 1H-13C HMQC spectra acquired on a sample of 2H-I-13CH3-syntaxin-1/Munc18-1 complex upon addition of the synaptobrevin and SNAP-25 SNARE motifs in the presence of MUN-CD module showed that the cross-peaks from the 2H-I-13CH3-syntaxin-1/Munc18-1 complex gradually decrease, and the cross-peaks from the SNARE complex emerge slowly over time [Figure 7A, middle and right panels]. The time dependence of the intensity of the Ile203 cross-peak from the SNARE complex yielded a rate constant of 0.38 ± 0.05 hr−1 for SNARE complex formation (Figure 7B), whereas the kinetics of disappearance of the Ile203 cross-peak from the syntaxin-1/Munc18-1 complex yielded a rate constant of 0.33 ± 0.13 hr−1. For comparison, the rate constants measured in the absence of MUN-CD were smaller than 0.021-0.16 hr−1, and in the presence of MUN domain they were 4.0-5.1 hr−1 (Ma et al., 2011). Hence, these data indicate that the MUN-CD module can slightly accelerate the transition from the syntaxin-1/Munc18-1 complex to the SNARE complex but this activity is much weaker than that of the MUN domain, in correlation with the findings that truncating the N-terminus of the MUN domain to residue 1045 strongly impairs its priming activity in vivo (Basu et al., 2005), and that MUN-CD has a much weaker affinity for the syntaxin-1 SNARE motif than the MUN domain. Overall, these results support the conclusion that binding to the SNARE motif is critical for Unc13/Munc13 MUN domain function, and suggest a refined model whereby the MUN-CD module is at the heart of this function but needs the assistance of neighboring sequences to efficiently execute this function (Figure 5).
Unc13/Munc13s perform a crucial function in synaptic vesicle priming through their MUN domain. While it is now clear that the MUN domain mediates the transition from the Munc18-1/syntaxin-1 complex to the SNARE complex, key questions remained about the minimal structural-functional units in Unc13/Munc13s and about which interactions underlie this function. Moreover, it seems likely that the Unc13/Munc13 MUN domain performs an additional, general function(s) that may be shared by all MUN-domain containing proteins, and the finding of remote homology between MUN domains and tethering factors involved in traffic at diverse membrane compartments (Pei et al., 2009) suggested that all these proteins form a family with a common role in membrane traffic. However, given the fundamental nature of this notion to understand intracellular membrane traffic and the very low sequence identities between MUN domains and tethering factors, it was crucial to obtain experimental evidence supporting the connection between these proteins. Our structure of the Munc13-1 MUN-CD module now reveals a striking similarity to the structures of Sec6p and other tethering factors, providing definitive prove that these proteins are related and indeed belong to the same family. Furthermore, the structure of the MUN-CD module, together with our biophysical studies of MUN-CD interactions and function, yield a framework to rationalize ample functional data that are currently available for Unc13/Munc13s, and suggest the refined model of Figure 5.
The extensive efforts that were required to solve the structure of the MUN-CD module, which spanned 13 years of research in our laboratory and involved preparation of about 50 different protein fragments [Figure 1 and (Basu et al., 2005)], emphasize the difficulties that can arise in structural studies of large helical regions where it is difficult to identify a clear structural unit from sequence analyses. Computational biology (Koch et al., 2000; Basu et al., 2005; Pei et al., 2009) has played a key role in guiding our efforts, and the continued improvements in sequence analysis methods that led to the identification of remote homology between MUN domains and tethering factors (Pei et al., 2009) were particularly critical to design the fragment that eventually allowed us to solve the structure of the MUN-CD module. Clearly, these methods still have difficulties in predicting specific structural features that are not conserved in the family (e.g. the bends in helices α5, α6 and α9 of the MUN-CD module), or correctly defining the alignment of the helices when there is substantially variability within the family (e.g. helices α3 and α5). However, the similarity of the MUN-CD structure described here to those of Sec6p and other tethering factors provides a vivid illustration of the power of current computational methods to identify remote homologies and predict overall protein folds based on these homologies, even when sequence identities are very low (< 10%).
Although the MUN-CD structure resembles those of several subunits of the exocyst complex, it is most closely related to Sec6p. It is noteworthy that the available structure of Sec6p covers only the CD subdomains (Sivaram et al., 2006) and that N-terminal fragments corresponding to the AB subdomain could not be expressed in soluble form (Sivaram et al., 2005), in analogy to the behavior of Munc13-1 fragments (Figure 1). These finding suggest that, functionally, Sec6p might also be more closely related to Unc13/Munc13s than other exocyst subunits. However, Sec6p binds to the SNAP-25 homologue Sec9p and this interaction inhibits SNARE complex formation (Sivaram et al., 2005), whereas the Munc13-1 MUN domain accelerates SNARE complex formation by binding to syntaxin-1 (Ma et al., 2011). On the other hand, activities in promoting SNARE complex formation have been described for CAPS (James et al., 2009) and the Dsl1p complex (Ren et al., 2009) and we note that a pattern of divergence was observed in studies of SM protein/SNARE interactions (Carr et al., 1999; Dulubova et al., 1999; Dulubova et al., 2002), and yet subsequent studies are gradually revealing more convergence than suggested by the initial studies (Dulubova et al., 2003; Dulubova et al., 2007; Shen et al., 2007; Furgason et al., 2009). Hence, while it is now clear that MUN domains and tethering factors belong to a common family, much research will be needed to unravel which functions are conserved and which are specific in the members of the family. Based on the above observations, the known role of tethering complexes in bridging membranes, and data supporting roles for CAPS and Unc13/Munc13s in vesicle docking (Hammarlund et al., 2007; James et al., 2009; Weimer et al., 2006; Siksou et al., 2009), we speculate that membrane docking-tethering and facilitation of SNARE complex assembly are general roles of this protein family, even though the underlying mechanisms may sometimes vary. Note also that this hypothesis does not rule out the possibility of an additional, direct role for these proteins in membrane fusion [see (Rizo and Rosenmund, 2008; James et al., 2009)].
Multiple mutagenesis data on Unc13/Munc13s that have accumulated over the years can now be rationalized based on the structure of the MUN-CD module described here. Thus, the structure shows that a prevalent mutation in Munc13-4 that causes a grave human disease (Feldmann et al., 2003) results in deletion of four residues (V608-A611) in helix α1 of subdomain C, which should lead to protein misfolding and thus loss of function. Moreover, the MUN-CD structure suggests that mutational studies claiming a correlation between disruption of Unc13/Munc13-1 function and binding to the syntaxin-1 N-terminus (Madison et al., 2005; Stevens et al., 2005) need to be reinterpreted. Thus, the mutations used in these studies (corresponding to F1234A/K1236A, Q1190R/L1279P/D1655E and I1364F in Munc13-1) are also likely to cause misfolding, as they involve replacement of one side chain from the hydrophobic core of subdomain C or D, and in the case of L1279P a proline is introduced in the middle of helix α4 (Figure S2B). Note that the mutated hydrophobic residues are buried in the structure of MUN-CD (therefore cannot participate in binding to syntaxin-1), and that the syntaxin-1 fragments used in these studies only contained part of the Habc domain (hence cannot fold). Moreover, we have been unable to detect the interaction with the syntaxin-1 N-terminus using well-folded recombinant proteins (Basu et al., 2005; Ma et al., 2011). Hence, there is currently no convincing evidence for the relevance of the interaction of Unc13/Munc13s with the syntaxin-1 N-terminus, although we cannot rule out that this interaction occurs in vivo and perhaps involves a post-translational modification.
Conversely, our recent studies have provided very strong evidence for a more natural mechanism of opening syntaxin-1 (Ma et al., 2011) that is further supported and refined by the results presented here (Figure 5). The key aspect of this model is that binding of the MUN domain to the syntaxin-1 SNARE motif, and likely to Munc18-1, helps to ‘extract’ the SNARE motif from the closed conformation, thus allowing syntaptobrevin and SNAP-25 binding to form the SNARE complex. This overall mechanism may prevent formation of syntaxin-1 oligomers and 2:1 syntaxin-1/SNAP-25 heterodimers, which hinders SNARE complex assembly (Rizo and Rosenmund, 2008). Note also that the MUN domain binds to 1:1 syntaxin-1/SNAP-25 heterodimers (Guan et al., 2008; Weninger et al., 2008) and hence may favor the correct syntaxin-1/SNAP-25 stoichiometry. Our data now show that the MUN-CD module binds weakly to Munc18-1 and the SNARE complex, with some degree of cooperativity between these interactions, and interacts also with the isolated syntaxin-1 SNARE motif (Figure 6). This behavior is very similar to that of the MUN domain (Ma et al., 2011), but the interactions with the SNARE complex and particularly with the syntaxin-1 SNARE motif are strongly impaired. As a consequence, the activity of MUN-CD in accelerating the transition from the syntaxin-1/Munc18-1 complex to the SNARE complex is severely impaired (Figure 7), in correlation with the observation that the MUN domain robustly rescues release while truncation of the N-terminal A,B subdomains strongly disrupts the rescue activity (Basu et al., 2005). Note however that fragments that are partially truncated at the MUN domain N-terminus but include the C2C domain can partially support Unc13 and Munc13 function (Stevens et al., 2005; Madison et al., 2005). Altogether, these results suggest that, although the MUN-CD module is a minimal structural unit that is at the heart of Unc13/Munc13 function, it can barely support this function by itself and requires the assistance of the adjacent MUN A,B subdomains or the C2C domain.
Based on our data, it seems very likely that the MUN-AB region cooperates with MUN-CD in binding to the syntaxin-1 SNARE motif to confer full activity to the MUN domain (Figure 5). Without the help of the MUN-AB region, the MUN-CD module still has some residual activity that could be enhanced by the C2C domain through other interactions that contribute to Unc13/Munc13 function, e.g. via the putative interaction with the syntaxin-1 N-terminus (but see above). Since the most common function of C2 domains is Ca2+-dependent membrane binding (Rizo and Sudhof, 1998), and the C2C domain is not predicted to bind Ca2+ but is basic, an alternative possibility is that the C2C domain binds constitutively to the plasma membrane (Figure 5), which is expected to enhance binding to the syntaxin-1 SNARE motif. Evidently, membrane interactions likely play multiple roles in modulating Unc13/Munc13 function, since the C1 domain and C2B domain preceding the MUN domain bind to diacylglycerol and phosphoinositides, respectively (Betz et al., 1998; Shin et al., 2010), and even the MUN domain exhibits weak membrane interactions that cooperate with binding to the SNARE complex (Guan et al., 2008). Note also that, while Figure 5 depicts interactions with the plasma membrane, some interactions could also be established with the vesicle membrane directly or indirectly (e.g. through Rab proteins) to tether the vesicles. Clearly, further research will be required to test all these possibilities and to unravel how the various interactions of Unc13/Munc13s with membranes and proteins control vesicle priming and presynaptic plasticity. The structure of the MUN-CD module described here provides a key piece to solve this fascinating puzzle.
Constructs for bacterial expression of full-length rat Munc18-1, fragments of rat synaptobrevin-2 (residues 29-93), human SNAP25 (residues 11-82 and 141-203), the cytoplasmic domain (residues 2-253) of rat syntaxin-1A or its SNARE motif (residues 191-253) were described previously (Dulubova et al., 1999; Dulubova et al., 2007; Xu et al., 2010; Chen et al., 2002). The vectors to express the various Munc13-1 fragments were prepared using standard recombinant DNA techniques, starting with fragments spanning full-length rat Munc13-1 or Munc13-1(859-1531) lacking residues 1415-1437, which are alternatively spliced (Brose et al., 1995). All proteins were expressed as GST fusions, isolated by affinity chromatography, and purified by gel filtration and/or ion exchange chromatography as described (Dulubova et al., 1999; Dulubova et al., 2007; Basu et al., 2005; Chen et al., 2002). Isotopic labeling was performed using well-established procedures (Tugarinov et al., 2004). Munc18-1–syntaxin-1(2-253) complexes and pre-assembled SNARE complexes formed with purified synaptobrevin(29-93), syntaxin-1(2-253), SNAP-25(11-82) and SNAP-25(141-203) were prepared as described (Ma et al., 2011).
Rat Munc13-1 MUN-CD (1148-1407,EF,1453-1531 fragment) dissolved in 10 mM Tris (pH 8.0), 150 mM NaCl, 10% (v/v) glycerol and 5 mM TCEP was concentrated to 7.5-8.5 mg/mL for crystallization using the hanging drop vapor diffusion method. Drops in a ratio of 1 μl protein to 1 μl well solution were equilibrated against 200 μl 0.1 M MES (pH 5.8-6.2), 18-25% (v/v) PEG 400 at 20 °C. Crystals appeared overnight and grew to ~15 × 30 × 100 μm within two weeks. Selenomethionine-derivatized (SeMet) crystals were obtained under similar conditions. For cryoprotection, crystals were transferred in steps into reservoir solutions containing increasing concentrations of PEG 400, to a final concentration of 35% (v/v), and flash-cooled in liquid nitrogen. MUN-CD crystals exhibited the symmetry of space group P42 with unit cell parameters of a = 160.8 Å, c = 42.2 Å and contained two molecules of MUN-CD per asymmetric unit. Native and SeMet MUN-CD crystals displayed strong anisotropy and strong I4 pseudosymmetry. Native MUN-CD crystals diffracted to a dmin of 2.65 Å when exposed to synchrotron radiation. Data were indexed, integrated and scaled using the HKL-3000 program package (Minor et al., 2006). Data collection statistics are in Table 1.
Phases for the MUN-CD fragment were obtained from a two-wavelength anomalous dispersion experiment using selenomethionyl-substituted protein with data to a dmin of 2.70 Å. Attempts to locate sufficient heavy atom sites to phase the structure were unsuccessful using data processed and scaled in the space P42. Initial phasing and model building calculations were carried out using data processed and scaled in the pseudosymmetric I4 space group, which contains one monomer of MUN-CD; seven of nine possible selenium sites were located using the program SHELXD (Schneider and Sheldrick, 2002). Initial phases were calculated and refined with the program MLPHARE (Otwinowski, 1991), resulting in an overall figure-of-merit of 0.15 for data between 38.0 and 2.70 Å. Phases were further improved by cycles of alternating density modification with the program Parrot (Zhang et al., 1997) and model building with the program Buccaneer (Cowtan, 2006). The resulting model contained 75% of all MUN-CD residues, and two copies of this model were placed in the SeMet protein P42 unit cell via the program Phaser (McCoy et al., 2007). Additional rounds of density modification and model building were performed as described, using only the SeMet peak dataset. The resulting model contained 84.5% of all MUN-CD residues in each monomer. Two copies of the MUN-CD monomer were placed in the native protein P42 unit cell via the program Phaser. Additional residues were manually modeled in the program coot (Emsley and Cowtan, 2004). Refinement was performed using the program Phenix (Adams et al., 2002) with a random 5% of all data set aside for an Rfree calculation. The current model contains 86.8% of all residues in two MUN-CD monomers (labeled A and B, respectively); included are residues 1156-1340, 1352-1407, E (Glu), F (Phe), 1453-1458, 1469-1514 of chain A, residues 1156-1340, 1352-1407, E (Glu), F (Phe), 1453-1457, 1468-1516 of chain B. The working R-factor is 0.244, and the free R-factor is 0.302. A Ramachandran plot generated with Molprobity (Davis et al., 2007) indicated that 89.7% of all protein residues are in the favored regions. The coordinates have been deposited in the PDB (accession code 3SWH).
1H-13C HMQC spectra were acquired at 25 °C on a Varian INOVA800 spectrometer equipped with a cold probe, using samples dissolved in 20 mM Tris (pH 8.0), 150 mM NaCl, 2 mM TCEP, using D2O as the solvent. 1H-15N HSQC spectra were acquired at 15 °C on Varian INOVA500 or 600 spectrometers with samples dissolved in 20 mM HEPES (pH 7.4), 125 mM NaCl, 2 mM TCEP, using H2O/D2O 95:5(v/v) as the solvent. Analysis and quantification of the data were performed as described (Ma et al., 2011).
The crystal structure of a C-terminal module of Munc13-1 (MUN-CD) is described MUN-CD is at the heart of Munc13 function but full activity requires adjacent modules The structure of MUN-CD is very similar to those of diverse tethering factors MUN-domain containing proteins and tethering factors form a protein family
We thank Nan Shen for initial work on crystallization of the MUN domain, Yilun Sun for expert technical assistance, Bingke Yu, Jimin Pei and Nick Grishin for fruitful discussions, and Zbyszek Otwinowsky for advice on data reduction. The structure shown in this report is derived from work performed on beamlines 19-BM and 19-ID at Argonne National Laboratory, Structural Biology Center at the Advanced Photon Source. Argonne is operated by UChicago Argonne, LLC, for the U.S. Department of Energy, Office of Biological and Environmental Research under contract DE-AC02-06CH11357. This work was supported by a postdoctoral fellowship from the American Heart Association (to Y.X.), by Welch foundation grant I-1304, and by NIH grant NS37200 (to J.R.).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.