Membrane Fusion Induced by Neuronal SNAREs Transits through Hemifusion*[boxs]

Synaptic transmission requires the controlled release of neurotransmitter from synaptic vesicles by membrane fusion with the presynaptic plasma membrane. SNAREs are the core constituents of the protein machinery responsible for synaptic membrane fusion. The mechanism by which SNAREs drive membrane fusion is thought to involve a hemifusion intermediate, a condition in which the outer leaflets of two bilayers are combined and the inner leaflets remain intact; however, hemifusion has been observed only as an end point rather than as an intermediate. Here, we examined the kinetics of membrane fusion of liposomes mediated by recombinant neuronal SNAREs using fluorescence assays that monitor both total lipid mixing and inner leaflet mixing. Our results demonstrate that hemifusion is dominant at the early stage of the fusion reaction. Over time, hemifusion transitioned to complete fusion, showing that hemifusion is a true intermediate. We also show that hemifusion intermediates can be trapped, likely as unproductive outcomes, by modulating the surface concentration of the SNARE proteins.

Synaptic transmission requires the controlled release of neurotransmitter from synaptic vesicles by membrane fusion with the presynaptic plasma membrane. SNAREs are the core constituents of the protein machinery responsible for synaptic membrane fusion. The mechanism by which SNAREs drive membrane fusion is thought to involve a hemifusion intermediate, a condition in which the outer leaflets of two bilayers are combined and the inner leaflets remain intact; however, hemifusion has been observed only as an end point rather than as an intermediate. Here, we examined the kinetics of membrane fusion of liposomes mediated by recombinant neuronal SNAREs using fluorescence assays that monitor both total lipid mixing and inner leaflet mixing. Our results demonstrate that hemifusion is dominant at the early stage of the fusion reaction. Over time, hemifusion transitioned to complete fusion, showing that hemifusion is a true intermediate. We also show that hemifusion intermediates can be trapped, likely as unproductive outcomes, by modulating the surface concentration of the SNARE proteins.
In the neuron, SNARE 1 assembly plays a critical role in promoting the fusion of the synaptic vesicles with the plasma membrane (1)(2)(3)(4)(5)(6)(7). Cognate SNAREs pair to form a coiled coil structure that bridges two membranes (8,9). The subsequent steps yielding one common phospholipid bilayer remain a matter of debate. It has been proposed that SNAREs involved in neurotransmitter release at synapses may promote membrane fusion by the formation of two juxtaposed transmembrane pores preassembled by the transmembrane domains of SNAREs in respective membranes (10). In sharp contrast, recent evidence for SNAREs involved in trafficking in yeast has indicated that hemifusion might be involved in the SNARE fusion pathway (11,12), analogous to the lipid-protein stalk model generally accepted for viral membrane fusion proteins (13)(14)(15)(16)(17). However, hemifusion has been observed only as an outcome rather than as an intermediate, raising some concerns as to whether hemifusion is an off-pathway product in SNAREmediated membrane fusion (16). Alternatively, the mechanism by which neuronal SNAREs induce membrane fusion might be entirely different from those for other systems including yeast SNAREs.
In this work, we examined the kinetics of membrane fusion of liposomes mediated by neuronal SNAREs syntaxin 1A, SNAP-25, and synaptobrevin using fluorescence assays (18) that monitored both total lipid mixing and inner leaflet mixing. Our results demonstrate that hemifusion is the main event at the early stage of the fusion reaction. Over time, hemifusion converts to the complete fusion, supporting strongly the theory that hemifusion is a true fusion intermediate.

MATERIALS AND METHODS
Protein Sample Preparation-Plasmid construction, protein expression, and purification for neuronal SNAREs were described elsewhere (19). Briefly, vesicle-associated (v-) SNARE synaptobrevin (amino acids 1-116) and a truncated version of target membrane (t-) SNARE syntaxin (amino acids 168 -288), for which the N-terminal ␣-helical Habc domain was deleted, were expressed as N-terminal glutathione S-transferase fusion proteins. Another t-SNARE, SNAP-25, was also expressed as a glutathione S-transferase fusion protein in which the four cysteines were replaced with alanines.
Recombinant proteins were expressed in Escherichia coli Rosetta (DE3) pLysS (Novagene). Glutathione S-transferase fusion proteins were purified by affinity chromatography using glutathione-agarose beads (Sigma). The protein was cleaved by thrombin in cleavage buffer (50 mM Tris-HCl, 150 mM NaCl, pH 8.0). We added 1% OG for syntaxin and synaptobrevin). Purified proteins were examined with 15% SDS-PAGE, and the purity was at least 90% for all of the proteins (data not shown).
Membrane Reconstitution-The mixture of POPC (1-palmitoyl-2-dioleoyl-sn-glycero-3-phosphatidylcholine) and DOPS (1,2-dioleoyl-sn-glycero-3-phosphatidylserine) (molar ratio of 65:35) in chloroform was dried in a vacuum and was resuspended in a buffer (50 mM Tris-HCl, 150 mM NaCl, pH 8.0) to make the total lipid concentration about 50 mM. Protein-free large unilamellar liposomes (ϳ100 nm in diameter) were prepared by extrusion through polycarbonate filters (Avanti Polar Lipids). Syntaxin (480 l, 21 M) and SNAP-25 (630 l, 16 M) were mixed at room temperature for about 60 min to allow the formation of t-SNAREs. The preformed t-SNAREs were concentrated down to 90 l using a 5-kDa cutoff centrifugal filter (Millipore) and were mixed with 10 l of liposomes for about 15 min at room temperature, resulting in a 50:1 lipid/protein molar ratio. The fluorescent liposomes containing POPC, DOPS, NBD-PS (1,2-dioleoyl-sn-glycero-3-phosphoserine-N-(7nitro-2-1,3-benzoxadiazol-4-yl)), and rhodamine-PE (1,2-dioleoyl-snglycero-3-phosphoethanolamine-N-(lissamine rhodamine B sulfonyl)) in the molar ratio of 62:35:1.5:1.5 were prepared following the procedure described above, and the final lipid concentration was ϳ10 mM. Synaptobrevin (80 l, 50 M) was mixed with 20 l of fluorescent liposomes for about 15 min at room temperature. The liposome/protein mixture was diluted two times, which makes the concentration of OG below the critical micelle concentration. After dialyzing against 2 liters of dialysis buffer (25 mM HEPES, 100 mM KCl, 5% (w/v) glycerin, pH 7.4) at 4°C * This work was supported by grants from the National Institutes of Health (to Y.-K. S.) and the Welch Foundation (to J. A. M.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbs.org) contains a supplemental figure and supplemental data.
The reconstitution efficiency was determined using SDS-PAGE and visualized by Coomassie Blue staining. The amount of protein in liposomes was estimated by comparing the band in the gel with that of the same protein of known concentration. The reconstitution efficiency was more than 90% for both t-and v-SNAREs (Fig. 1B). The orientation of the SNAREs in the liposomes was examined with the trypsin digestion experiments (19). The SNAREs-reconstituted vesicles were treated with trypsin (0.5 mg/ml) under room temperature for 1 h. Nearly all SNARE proteins were digested by trypsin, indicating that SNARE molecules are oriented inside out, exposing the soluble domain to the solution phase (Fig. 1B).
Total Lipid Mixing Fluorescence Assay-To measure the lipid mixing, v-SNARE liposomes were mixed with t-SNARE liposomes in the ratio of 1:9. The final solution contained ϳ1 mM lipids with a total volume of 100 l. Fluorescence was measured at excitation and emission wavelengths of 465 and 530 nm, respectively. Fluorescence changes were recorded with a Varian Cary Eclipse model fluorescence spectrophotometer using a quartz cell of 100 l with a 2-mm path length. The maximum fluorescence intensity (MFI) was obtained by adding 0.2% n-dodecylmaltoside. All of the lipid mixing experiments were carried out at 35°C. For each lipid/protein (L/P) ratio, the experiments were performed at least three times with newly prepared samples. To make sure that the percentage of MFI was independent of the probe concentrations, we measured the total lipid mixing at 1.5 and 0.7 mol % NBD-PS while keeping the rhodamine-PE concentration at 1.5 mol %. We found that the percentages of MFI were identical for both NBD-PS concentrations (supplemental information).
Inner Leaflet Mixing Assay-The inner leaflet mixing assay was modified from the method developed by Meers et al. (20). The method is based on the fact that sodium dithionite reacts more rapidly with NBDs in the outer leaflet than those in the inner leaflet. By controlling the time and amount of dithionite, the reaction can be limited to the outer leaflet. Small aliquots (ϳ0.7 l) of 100 mM sodium dithionite in 50 mM Tris buffer, pH 10, were added to the v-SNARE liposomes (100 l, 0.2 mM lipid) until a desired reduction of NBD was achieved. The reaction was monitored at room temperature by scanning the fluorescence signal for 15 min from 500 to 700 nm with the excitation at 460 nm. Typically, in 10 min the reduction was complete, and no more change of the spectrum was observed. The liposomes without NBDs in the outer leaflets were subjected to the lipid mixing assay described above. To make sure that the percentage of MFI was independent of the extent of the NBD reduction, the inner leaflet mixing assay was performed at the 55 and 65% reduced conditions. We found that the percentages of MFI were identical for both conditions (supplemental information).

RESULTS AND DISCUSSION
We investigated the fusion of liposomes induced by neuronal SNAREs with a well characterized fluorescence lipid mixing assay. Recombinant t-SNARE complexes containing the H3 "core" domain of syntaxin 1A as the t-SNARE heavy chain and SNAP-25 as the t-SNARE light chain were reconstituted into the liposomes containing POPC (65 mol)/DOPS (35 mol) (Fig.  1A). Additionally, the v-SNARE synaptobrevin was also reconstituted into a separate population of the same POPC/DOPS liposomes containing fluorescent lipids, NBD-PS, and rhodamine-PE (1.5 mol % each). For both t-and v-SNAREs, the initial lipid/protein ratio was set at 50:1 (Fig. 1B). When the t-SNARE liposomes were mixed with the v-SNARE liposomes at 35°C, an increase of the fluorescence intensity was observed, indicating that the fusion occurred ( Fig. 2A, red trace). Although both v-and t-SNAREs have roughly equal surface density, the absolute amount of t-SNARE liposomes in the cuvette was about 10 times greater than the v-SNARE lipo-somes. These amounts of protein and lipid were used to better ensure first order kinetics. Following an initial rapid rise, the fluorescence signal approached a plateau of roughly 40% of maximum ( Fig. 2A). The half-time of the fusion reaction was ϳ9 Ϯ 1 min (mean Ϯ S.D.) (ϳ540 Ϯ 60 s), consistent with the previous reported half-time of 10 min with a similarly truncated syntaxin 1A (21).
Because the fluorescent lipids were distributed equally in the inner leaflet and the outer leaflet, the observed total lipid mixing should be the sum of outer leaflet mixing and inner leaflet mixing. To selectively measure inner leaflet mixing separately, we treated the v-SNARE liposomes with sodium dithionite. Under controlled conditions, sodium dithionite reduces NBD attached to the lipid head group in the outer leaflet to a nonfluorescent derivative while leaving NBD in the inner leaflet largely unaffected. When we mixed the dithionitetreated v-SNARE liposomes with the t-SNARE liposomes, inner leaflet mixing was observed ( Fig. 2A, blue trace). The extent of the NBD reduction did not affect the kinetics of inner leaflet mixing (supplemental information). Interestingly, the halftime of inner leaflet mixing was ϳ20 Ϯ 2 min (ϳ1,200 Ϯ 120 s), which was about twice the half-time of total lipid mixing. The kinetic difference in the half-times of the two processes suggests that outer leaflet mixing and inner leaflet mixing were not simultaneous but sequential in time. These results suggest that outer leaflet mixing likely occurred faster than inner leaflet mixing.
Because we collected the time traces of total lipid mixing and inner leaflet mixing separately, it was straightforward to calculate the percentage of hemifusion (defined as 2(P T Ϫ P I )/ [2(P T Ϫ P I ) ϩ P I ] ϫ 100, where P T is the percentage of maximum for total lipid mixing and P I is the percentage of maximum for inner leaflet mixing ( Fig. 2A)) as a function of time. As expected, at the beginning of the fusion reaction, the fluorescence change was mainly due to outer leaflet mixing (Fig. 2B), indicating that hemifusion was the dominant event. As time progressed, however, the percentage of hemifusion decreased dramatically. Hemifusion was about 90% at 1 min and was extrapolated to be nearly 100% at the start of the reaction. The percentage steadily declined to roughly 20% at 40 It should be noted that these estimates are from the ensemble of ϳ7 ϫ 10 11 liposomes in the reaction and that each individual event is likely to be very fast (18,21). These results provide strong kinetic evidence for the conversion of hemifusion to complete fusion over time and therefore for the sequential mechanism in which hemifusion is an on-pathway intermediate (Fig. 2C).
We analyzed outer and inner leaflet mixing separately on the basis of the sequential mechanism (Fig. 2C). For this purpose, net outer leaflet mixing was obtained by subtracting the kinetics of inner leaflet mixing from that of total lipid mixing (Fig.  2E). The time trace of outer leaflet mixing fitted well to a simple exponential function representing the first order kinetics with the first order rate constant k 1 ϭ 1.5 ϫ 10 Ϫ3 s Ϫ1 (Fig.  2E). The kinetics of inner leaflet mixing was analyzed with the first order kinetics theory for the sequential mechanism depicted in Fig. 2C (22). The theory fitted the data very well with first order rate constant k 2 ϭ 3.7 ϫ 10 Ϫ3 s Ϫ1 and the backward rate constant k Ϫ1 ϭ 6.3 ϫ 10 Ϫ3 s Ϫ1 . It is quite interesting to find that k Ϫ1 is almost twice as big as k 2 . Thus, once hemifusion is formed it is twice as likely to go back to the two separate liposomes than to advance toward the complete fusion.
One might argue that these results are equally consistent with a parallel mechanism as well (Fig. 2D). In this alternative mech-anism, hemifusion is an off-pathway product in equilibrium with the unfused liposomes. However, the data argue against the parallel mechanism. At the beginning of the reaction, the fusion events were almost exclusively hemifusion (Fig. 2B), which is a clear indicator for a sequential mechanism. Thus, the results favor the sequential mechanism and establish that SNARE-mediated membrane fusion transitions through hemifusion.
Work with viral fusion proteins as well as SNARE proteins has suggested that the surface protein density of the fusion proteins may be an important parameter that determines the outcome of the fusion events (12,23). To gain further insight into the role of the protein surface density, we reduced the input L/P ratios to 100:1 and 200:1, from the original 50:1. Liposomes generated with a 100:1 L/P ratio yielded results that were qualitatively similar to those for the previous 50:1 L/P ratio (Fig. 3, A and B), although the overall fusion efficiency was lower and the conversion from hemifusion to complete fusion was slower. With this surface density, ϳ45% of full fusion occurred during the 75-min reaction. However, at the L/P ratio of 200:1 we did not observe the time-dependent shift from hemifusion to complete fusion (Fig. 3, C and D). This result suggests that the majority (Ͼ60%) of hemifusion intermediate remained without transitioning to complete fusion. Thus, the surface density of SNARE proteins is indeed a determining factor for the outcome of SNARE-induced membrane fusion.
FIG. 2. Fluorescence assay for total lipid mixing and inner leaflet mixing. A, fluorescence changes for total lipid mixing (red traces) and inner leaflet mixing (blue traces), normalized with respect to the MFI, are shown for the L/P ratio of 50:1. MFI was obtained by adding 0.2% n-dodecylmaltoside (sudden increase at the end). The black trace is a control run with the t-SNARE liposomes reconstituted with syntaxin 1A only (without SNAP-25). Inset, residual fluorescence changes for total lipid and inner leaflet mixing recorded at the longer period of time (8,500 -9,100 s); the former was 43%, and the latter was 40%. B, the percentage of hemifusion versus time. The percentage of hemifusion was calculated using the equation 2(P T Ϫ P I )/[2(P T Ϫ P I ) ϩ P I ] ϫ 100, where P T is the percentage of maximum for total lipid mixing and P I is the percentage of maximum for inner leaflet mixing. C, a schematic diagram for the sequential mechanism in which hemifusion is an on-pathway intermediate. k 1 , first order constant for time trace of outer leaflet mixing; k Ϫ1 , backward rate constant; k 2 , first order constant for time trace of inner leaflet mixing. D, a schematic diagram for the parallel mechanism in which hemifusion is an off-pathway product. E, analysis of outer and inner leaflet mixing based on the sequential mechanism in C and the first order kinetics. The solid lines are the best fits to the first order kinetics. The data were fitted with the program DYNAFIT (22).
In summary, we have shown that membrane fusion induced by neuronal SNAREs transitions from hemifusion to complete fusion in a kinetically resolvable manner, establishing that hemifusion is a true intermediate along the SNARE-induced membrane fusion pathway. We also have shown that, under low surface protein density, hemifusion can be trapped as an outcome of SNARE-induced membrane fusion.