Report

SNARE Proteins: One to Fuse and Three to Keep the Nascent Fusion Pore Open

See allHide authors and affiliations

Science  16 Mar 2012:
Vol. 335, Issue 6074, pp. 1355-1359
DOI: 10.1126/science.1214984

Abstract

Neurotransmitters are released through nascent fusion pores, which ordinarily dilate after bilayer fusion, preventing consistent biochemical studies. We used lipid bilayer nanodiscs as fusion partners; their rigid protein framework prevents dilation and reveals properties of the fusion pore induced by SNARE (soluble N-ethylmaleimide–sensitive factor attachment protein receptor). We found that although only one SNARE per nanodisc is required for maximum rates of bilayer fusion, efficient release of content on the physiologically relevant time scale of synaptic transmission apparently requires three or more SNARE complexes (SNAREpins) and the native transmembrane domain of vesicle-associated membrane protein 2 (VAMP2). We suggest that several SNAREpins simultaneously zippering their SNARE transmembrane helices within the freshly fused bilayers provide a radial force that prevents the nascent pore from resealing during synchronous neurotransmitter release.

Efficient synaptic transmission requires the fast release of neurotransmitters from synaptic vesicles that fuse with the presynaptic plasma membrane upon entry of calcium ions (1). Membrane fusion necessarily implies a fusion pore that opens between the vesicle and its partner membrane at the instant of fusion. The conductance properties of such nascent fusion pores suggest that their typical diameters are in the range of ~2 nm, although considerable variability exists (25). Neurotransmitter is released from synaptic vesicles [diameter ~40 nm (6, 7)] by diffusion through the nascent pore in the first 100 to 200 μs, even before appreciable dilation of the pore occurs (4). The transient and variable nature of the fusion pore has severely limited biochemical and physical chemical studies.

We suggest that nanodiscs (811) provide an ideal model for such studies because the small amount of disc lipid will suffice to allow pores to open but not expand (Fig. 1 and Fig. 2A) beyond their nascent, physiologically relevant state for neurotransmitter release. Nanodiscs are synthetic lipoprotein particles that contain a small piece of circular lipid bilayer (up to ~17 nm in diameter) wrapped by two copies of membrane scaffold protein (MSP) derived from apolipoprotein A1. In the system we describe here, nanodiscs contain the synaptic vesicle SNARE (v-SNARE) VAMP2 and small unilamellar vesicles [diameter 30 to 60 nm (12)] contain the synaptic target membrane SNARE (t-SNARE) complex of syntaxin 1 and SNAP25. SNAREs are the core machinery for this and other cellular membrane fusion processes (1214). They assemble between bilayers as a four-helix bundle (15) that imparts sufficient force to cause bilayer fusion (16).

Fig. 1

(A) Cartoon showing the v-disc model. The nanodisc is a small piece of lipid bilayer wrapped by two MSPs (blue). VAMP2 (green) can insert into a nanodisc to form a v-disc (11). The lipid head groups are shown as gray spheres. (B) Elution profile of nanodisc or v-disc on Superdex 200 10/300 GL column. Embedment of VAMP2 results in the earlier elute volume of a v-disc (red major peak) relative to that of a VAMP2-free nanodisc (black peak). By gel filtration, the 6×His-SUMO tag (cleaved from VAMP2 by SUMO-protease, the red minor peak) can also be removed. (C) SDS–polyacrylamide gel electrophoresis (PAGE) gel stained with Coomassie Brilliant Blue showing the input and final nanodisc products after gel filtration. (D) V-disc samples were analyzed in an FEI Tecnai-12 electron microscope. Left panel: V-discs show regular “disc” shapes; VAMP2 protein can hardly be seen because of small protein size and flexible structure. Right panel: When soluble syntaxin 1A H3 domain and SNAP-25N/C domain were co-incubated with v-disc, they form SNARE complexes that can be seen as rodlike structures (red) protruding from two sides of the v-disc (green).

Fig. 2

(A) Schematics showing how the fusion pore can be envisioned. The diameter of the nanodisc is 16 nm. Lipids that naturally form flat surfaces will favor structures that have a zero net curvature (when neglecting the Gaussian curvature). Hence, in neck-like structures, the negative curvatures (shown here as perpendicular to the pore) are of the same order; likewise, for saddle-like structures, the positive (pore) curvatures are of the same order. A 4-nm pore would correspond to a 6-nm curvature for the exterior of the bilayer. This is approximately what is represented here. With no stress, a 1-nm pore can form. With a reasonable amount of stress (θ = 20°), a 4-nm pore diameter would result. (B) Lipid mixing is SNARE-specific. V-discs exchanged lipids with t-liposomes (blue). Discs without VAMP2 do not fuse with t-liposomes (red); cytoplasmic domain of VAMP2 (CDV), which titrates the free t-SNARE, also blocks the fusion (green). (C) Calcium release is SNARE-specific. Calcium (50 mM) is encapsulated into t-liposome; during the liposome-nanodisc fusion, the pore opens, calcium is released from the liposome to the exterior buffer, and the Mag-Fluo-4 signal is enhanced. An increasing Mag-Fluo-4 signal indicates that calcium is continuously released during the fusion (blue); limited calcium release is observed under nonfusogenic conditions (red and black). (D) Dithionite assay showed some NBD protection after 40 min of fusion (blue). To completely quench all NBD signal, detergent (De) was first added to disrupt the liposomes, followed by dithionite (Di) to obtain 100% quench (brown). With CDV to block the fusion, no NBD protection was observed after dithionite treatment (green).

After reconstitution (supplementary text), the nanodiscs containing VAMP2 (v-discs) were separated by gel filtration (Fig. 1, A to C). Each disc contained about 400 lipid molecules wrapped by two MSPs. With a starting VAMP2/MSP ratio of 6:2, we recovered ~7 VAMP2 copies per disc on average after removing VAMP2-free dics. Electron microscopy of v-discs confirmed an average diameter of 16 ± 2 nm (Fig. 1D). Not surprisingly, single VAMP2 proteins on these discs could not be readily distinguished because of their small size and flexible structure. However, addition of the soluble t-SNARE (a complex of syntaxin H3 cytosolic domain and SNAP25N/C helical domains) formed rodlike SNARE complexes that were seen to protrude from the nanodiscs (Fig. 1D). This confirms that VAMP2 on nanodiscs can form SNARE complexes.

We used a well-established lipid mixing assay (17) to test whether v-discs can fuse with t-vesicles. Nitro-2-1,3-benzoxadiazol-4-yl-phosphatidylethanolamine (NBD-PE) and rhodamine-PE were included in the v-discs (1.5 mol % each). This surface concentration of rhodamine effectively quenches the NBD fluorescence. However, when a nanodisc fuses with a liposome, NBD fluorescence will greatly increase because of the substantial (>50-fold) lipid dilution as the disc lipid mixes with the massive excess of vesicle lipid, as we observed (Fig. 2B). Little or no increase of NBD signal was observed in control experiments. The slow lipid mixing between nanodiscs and vesicles was limited by the rate of docking (initial SNARE assembly) and not by the rate of fusion (fig. S1), as is the case for vesicle-vesicle fusion systems (18). A similar fusion process was observed when the SNAREs were placed in the opposite topology, with v-SNAREs in the vesicle and t-SNAREs in the nanodisc (fig. S2).

To monitor efflux of content via fusion pores that necessarily form at least transiently during the fusion process, we encapsulated calcium (50 mM) in the liposomes, which were then incubated with v-discs in a medium containing a calcium-activated fluorophore, Mag-Fluo-4 (2 μM, Kd for calcium = 22 μM; Invitrogen). When pores open, calcium diffuses through the pores into the exterior buffer, inducing a fluorescence signal. The results (Fig. 2C) clearly show that calcium was released in a SNARE-specific manner. To ascertain that this efflux was due to diffusion through a pore, rather than transient lysis or leakage of the vesicle during fusion, we tested the rate of release of vesicle cargos of different sizes—specifically, the calcium chelator EDTA (Stokes radius, ~0.4 nm) and EGTA (Stokes radius, ~0.5 nm). EDTA release from liposomes was faster than EGTA release from liposomes by a factor of 2.3 (fig. S3). From these data, a pore size of ~2 nm can be calculated (supplementary text), similar in size to the nascent fusion pore size calculated from electrophysiological measurements (3).

When vesicles fuse to target membranes, the fusion pore eventually expands and the vesicle is incorporated into the target membrane. With nanodiscs, however, the pore cannot appreciably expand beyond its nascent diameter of ~2 nm, so the only means available to reduce membrane stress (resulting from the extreme curvature inherent in a small pore) is for the pore to eventually reseal. To confirm the prediction that nanodisc-vesicle pores reseal, we introduced dithionite (5 mM) into samples 40 min after beginning the fusion assay. Dithionite quenches all externally accessible NBD (19), including NBD on both faces of unreacted nanodiscs and NBD-PE that had diffused into the outer leaflets of liposomes via hemifusion or full fusion. Dithionite is also small enough (Stokes radius, 0.2 to 0.3 nm) to readily diffuse through any 2-nm fusion pores that may remain open and quench the NBD signal on the inner leaflets in the case that the pore remains open, but will not gain access to the interior if the pore has (as predicted) closed off. We observed that some of the NBD dye remained protected against dithionite, and only after the fusion reaction (Fig. 2D and fig. S4). No NBD protection was found for the negative controls. Thus, there are some fusion events that correspond to full fusion between the nanodisc and the liposome, in which a pore must have opened and then resealed. Reverse experiments where NBD-PE was initially only on the t-liposomes confirmed that essentially no pores remained open after full fusion (fig. S5). The dithionate data indicate that ~ 50% of all fusion events entailed full fusion with subsequent resealing of the pore (see supplementary text). The ~50% balance of SNARE complex–dependent events can be accounted for by events resembling hemifusion in which no pore opens and only the outer leaflets are shared between the liposome and nanodisc. In all cases, the nanodisc remained attached to the liposomes after the pore resealed (fig. S6, A to D), consistent with the idea that after full fusion, the pore reseals to a hemifusion-like state in which a stalk of lipid bilayer permanently connects the outer leaflet of the vesicle to what had been the SNARE complex–containing leaflet of the nanodisc (fig. S6F). After resealing, VAMP2 is fully resistant to toxin cleavage, which suggests that it is in cis-SNARE complexes (fig. S6, E and F).

The number of VAMP2 copies per disc can be controlled by adjusting the input VAMP2/MSP ratio. With increasing VAMP2/MSP ratios during assembly of nanodiscs, the v-disc products eluted progressively earlier on gel filtration columns (Superdex 200), consistent with increasing size and more VAMP2 being inserted into each disc (Fig. 3A). To test how the number of SNAREpins affects fusion, we purified seven sets of nanodiscs containing, respectively, an average of 1.2 (ND1), 2.2 (ND2), 3.15 (ND3), 4.3 (ND4), 5.5 (ND5), 7.4 (ND7), and 9.3 (ND9) copies of VAMP2 after VAMP2-free nanodiscs were removed by affinity purification (Fig. 3B and fig. S7). VAMP2 and MSP were mixed in these preparations in different proportions, and the VAMP2/MSP ratios in the final isolated nanodiscs used in the fusion experiments were established by three independent methods: Coomassie Blue–based protein determinations (Fig. 3B), quantitative Western blotting (fig. S7), and counting of the discrete photobleaching steps of single nanodisc particles containing fluorescently labeled VAMP2 (fig. S13). The three methods agreed very closely [within ±3% (SD/mean × 100) for ND1 and ND2 in particular; see table S2].

Fig. 3

(A) Elution profile of v-discs with different numbers of VAMP2 copies per disc. With more VAMP2 incorporation, v-discs eluted in smaller elution volumes on a Superdex 200 column. (B) Determination of the number of VAMP2 copies per nanodisc. The v-disc samples obtained by gel filtration were analyzed by SDS-PAGE staining with Coomassie Brilliant Blue. The number of VAMP2 copies per disc was determined by the VAMP2/MSP ratio according to the quantification of the corresponding protein bands. Each v-disc population has about 1.2 (ND1), 2.2 (ND2), 3.15 (ND3), 4.3 (ND4), 5.5 (ND5), 7.4 (ND7), and 9.3 (ND9) copies of VAMP2 per disc on average. (C) Lipid mixing assays demonstrate that discs with varying VAMP2 copy numbers are equally efficient in fusing with calcium-encapsulated t-liposomes. (D) Calcium release assay shows different kinetics when v-discs with different copy numbers of VAMP2 fuse with t-liposomes. Calcium release is gradually increased with higher copy numbers of VAMP2 inserted into the nanodiscs. The error bars are SEM. (E) The endpoint values after a 40-min fusion reaction are presented for both lipid mixing (blue) and calcium release (green). Lipid mixing, normalized by the average endpoint value, does not vary significantly with the number of VAMP2 copies. Calcium release, normalized by the value for ND9, varies as a sigmoid with an inflection point at ~5 VAMP2 copies per nanodisc.

Remarkably, all seven samples drove lipid mixing at the same rate (Fig. 3C), which implies that a single v-SNARE per nanodisc yields the maximum rate of membrane fusion. Furthermore, the dithionite protection assay shows that for any VAMP2 copy number, the percentage of lipid-mixing events corresponding to full fusion remains unchanged at 45 ± 7% (SD; fig. S8). Together, these two results clearly establish that a single SNAREpin is sufficient to drive complete membrane fusion.

The pores forming in our nanodisc-liposome system are transient, eventually resealing of their own accord (fig. S5). Because our lipid-mixing experiments result in full membrane fusion (i.e., full exchange of lipid), we can set a lower limit for how long these pores must have remained open. In order for all of the fluorescent phospholipid to equilibrate between the nanodisc and the liposome, the pore must remain open for ~10 μs (it takes ~10 μs for a phospholipid with a diffusion coefficient of 5 μm2/s to cover 50 nm2, half of a nanodisc embedded bilayer, and reach the fusion pore). In contrast, simple considerations and in vivo observations of neurotransmitter release from synaptic vesicles suggest that a much longer time (~100 μs) is required for full efflux of the content with similar size as neurotransmitter from ~40-nm liposomes or synaptic vesicles (4, 2022).

Do the fusion pores between nanodiscs and vesicles remain open long enough for such cargo to efflux? The answer (Fig. 3D) is that they do when several or more SNARE complexes can form, but not when only a single SNAREpin is available. In contrast to the rate of lipid mixing, maximum cargo efflux decreased precipitously from 12% (ND9 and ND7) to less than 2% (ND1 and ND2) as the number of VAMP2s per disc decreased from ~9 to ~1. Thus, even though the rate and frequency of full membrane fusion events do not depend on the number of VAMP2 molecules per disc, the efficiency of cargo release is highly sensitive to SNAREpin number, increasing markedly as the number of SNAREpins increases above two per disc.

At very low SNARE numbers (i.e., ND1 or ND2), the pore opens only long enough to exchange lipid, and thus only a small fraction of the content is released. The amount of release increases starting with ND3 discs and reaches a maximum with ND7 discs, which reconstituted with about 7 VAMP2 proteins total or about 3.5 per nanodisc face (table S2); this finding suggests that maximum efflux requires 3 or 4 VAMP2 proteins. The simplest model is that a limited content release occurs when one SNAREpin is engaged, whereas a sudden increase in content release occurs above a threshold when enough SNAREpins are involved (fig. S12C). In the nanodisc system, that critical number is achieved when any one side of the nanodisc has at least the minimum necessary number of SNAREs. Indeed, if we assume that the VAMP2 distributes randomly between the two sides of each disc (fig. S12B), the calcium release across the whole of the titration fits well to such a “cooperative” model and describes the threshold number of SNAREpins for efficient content release as ~3 (fig. S12C), which is consistent with in vivo observations.

The role of the SNARE transmembrane domains (TMDs) in fusion has been unclear. Membrane anchorage of the assembling cytosolic domains of VAMP2 and syntaxin is needed, and when this is provided by membrane-spanning lipids, fusion still occurs (17). In absolute contrast, point mutations in the syntaxin 1 TMD reduce the amplitude of the foot signal in electrophysiology (23), and deletion in the VAMP2 TMD significantly reduces neurotransmitter release (24). This implies a role for the TMDs either in the opening of the nascent physiological fusion pore, or in maintaining it open for the ~100 μs needed for transmitter efflux, or both. Because fusion pores must open (at least transiently) when lipid anchors mediate fusion, the simplest possibility is that the TMDs somehow keep the fusion pore from resealing when transmitter is exiting and the pore has not begun to appreciably expand. In this connection, it is noteworthy that the VAMP2 and syntaxin TMDs extend as a two-helix bundle through the entire span of the membrane (25). This raises the possibility that force resulting from the terminal zippering of SNARE TMDs within the bilayer could provide a source of energy to tip the balance against resealing in the nascent fusion pore.

To test this hypothesis, we used three chimeric VAMP2 proteins in which the VAMP2 TMD was replaced by (i) a dioleoyl phospholipid that spans only one monolayer (C18), (ii) a long C45 isoprenoid that can span the lipid bilayer (C45), or (iii) a non-SNARE TMD from platelet-derived growth factor receptor (PDGFR) (Fig. 4A). As with the wild-type VAMP2, seven samples of PDGFR–VAMP2-discs (characterized in fig. S9) were prepared: PND9, PND7, PND5, PND4, PND3, PND2, and PND1, respectively containing an average of 9.0, 7.2, 5.7, 4.4, 3.75, 2.7, and 1.2 copies of PDGFR-VAMP2. Lipid anchors (C18 and C45, both with an average of ~5 copies per nanodisc) were less efficient for fusion than the VAMP2 TMD by a factor of 5 to 10 (Fig. 4B), whereas the PDGFR TMD fused at the same rate as the native VAMP2 TMD at all numbers of VAMP2s per nanodisc (Fig. 4C). With C18 VAMP2, the fluorescence signal collapsed back to the background level after dithionite injection (fig. S10), which suggests that only the outer leaflets are shared (i.e., hemifusion). By contrast, C45 and PDGFR TMDs achieved full fusion with essentially the same ~50% efficiency as the native VAMP2 TMD. This confirms that a membrane-spanning domain (either lipid or protein) is required to achieve full fusion and also demonstrates that fusion pores open when only the cytosolic domains of the SNARE complex have zippered. Thus, bilayer fusion and the concomitant opening of the nascent pore do not require assembly of the syntaxin and VAMP2 TMDs into the bilayer-spanning helical bundle.

Fig. 4

(A) Schematics showing the wild type and various chimeric forms of VAMP2. The structures of C18 and C45 (solanesyl ester) and the protein sequences of VAMP2 and PDGFR TMDs are shown (abbreviations for amino acids: A, Ala; C, Cys; F, Phe; G, Gly; I, Ile; L, Leu; M, Met; S, Ser; T, Thr; V, Val; W, Trp; Y, Tyr; asterisks denote identical residues, and colons denote conserved substitutions). C18 spans only one leaflet of the lipid bilayer, whereas C45 and protein TMDs cross the lipid bilayer. (B) Lipid mixing assay of t-liposome with v-disc prepared with wild-type VAMP2 (ND5) or CDV with C18 or C45. The lipid anchor, either C18 or C45, shows compromised fusion efficiency relative to wild-type VAMP2 with native TMD. (C) Lipid mixing assay shows that the TMDs of VAMP2 and PDGFR have similar fusion kinetics, which suggests that these two TMDs are equivalent in lipid mixing. (D) Calcium release assay shows substantially reduced release when protein TMD is replaced by lipid anchor, either C18 or C45. (E) Calcium release assay reveals that VAMP2 TMD is more efficient than PDGFR TMD for content release. The error bars are SEM.

To establish the lifetime of the open pore in these artificial fusion events, we used the calcium release assay. By contrast to lipid mixing, none of the nonnative VAMP2 TMDs efficiently released cargo (Fig. 4, D and E). These experiments show that the native VAMP2 TMD is specifically required for efficient release of vesicle content after the pore opens, which it allows by virtue of lowering the rate of resealing of the nascent fusion pore.

In the nanodisc system, the extent of content release can only be determined by the total amount of time the pore is open before permanently resealing. Should the pore reseal within ~100 μs (as calculated from diffusion constants and vesicle and pore diameter), only a commensurate fraction of the cargo will exit even though the lipids in the disc and vesicle bilayers will have fully mixed. These considerations are fundamental for understanding the different requirements for the number of SNAREs and their TMDs for vesicle fusion and content release, as revealed by our nanodisc experiments. Specifically, our results suggest that although a single SNARE complex suffices to open a fusion pore between nanodisc and vesicle, this pore is too short-lived to allow much transmitter to exit before the pore closes. Only when there are several or more SNAREpins assembling at the same pore does it remain open long enough for effective transmitter release. Even if there are enough SNAREs, it appears that the pore is short-lived unless the TMDs of VAMP2 and syntaxin can zipper in the bilayer to keep the pore open in one way or another, perhaps by pushing outward radially as their TMDs zipper within the bilayer. A single zippering SNAREpin could not do this, which explains why it would necessarily be ineffective. But it is easy to see that three or more SNAREpins pushing away from each other radially could channel the energy of trans-bilayer zippering to keep the nascent pore open. In this speculative model, the restraining force against resealing may only last for as long as it takes the TMDs to zipper. That length of time is likely to be much more than 100 μs, based on the maximum speed for folding of a two-helix coil (2628) and considering the higher viscosity of hydrocarbon relative to water (29); this would be more than enough time to allow complete neurotransmitter release.

A previous study showing that a single SNARE complex could mediate vesicle-vesicle fusion also measured content mixing (30). Interestingly, normalizing the published data for SNARE-free vesicles (see supplementary text and table S1) reveals that here, too, content mixing is greatly reduced relative to bilayer fusion with one SNARE complex per vesicle. Vesicle-vesicle fusion is inherently a poor model of neurotransmitter release through a nascent fusion pore because content mixing occurs not only through a nascent pore (as in nanodiscs and at the synapse) but also subsequently as the vesicles more slowly complete their fusion. Yet despite these limitations, indications of the mechanism we have uncovered can still be found.

These findings provide a simple mechanistic basis for understanding published data that had seemed to be contradictory. Titrations of dominant-interfering SNARE mutants in permeabilized or intact neurosecretory cells suggest a minimum requirement for three SNAREpins to open a fusion pore sufficient for neurotransmitter efflux (31, 32). A quantitative analysis of titrations of botulinum toxin A in relation to cleavage of SNAP25 has been interpreted to mean that a minimum of 10 to 15 SNAP25 molecules are required (33), but this analysis assumes that all SNAP25 is present in SNARE complexes and must be considered an upper limit only. By contrast, an elegant single-particle, single-molecule analysis combined with vesicle fusion reactions clearly established that a single SNAREpin was present in many fused vesicles (30). We can now see that all of these results emerge from a single underlying mechanism in which the dynamics of the nascent fusion pore are determined by the number of SNAREpins involved. Synaptic vesicles have ~70 copies of the v-SNARE VAMP2 (6) and the active zone is rich in t-SNAREs (34), ensuring that multiple SNAREpins are always available to keep the pore open and let transmitter out as rapidly as possible.

Supporting Online Material

www.sciencemag.org/cgi/content/full/335/6074/1355/DC1

Materials and Methods

SOM Text

Figs. S1 to S13

Tables S1 and S2

References (3539)

References and Notes

  1. Acknowledgments: Supported by Agence Nationale de la Recherche grant 09-Blanc-0129 (F.P.), NIH grants (J.E.R.), and a Partner University Funds exchange grant between the Yale and Ecole Normale Supérieure laboratories. We thank J. Shen (University of Colorado, Boulder) for providing the PDGFR-VAMP2 expressing vector; R. Beck, E. Rhoades, E. Karatekin, and A. Nath for many helpful discussions; and W. Xu and J. Coleman for their kind help.
View Abstract

Navigate This Article