Fusion of Cells by Flipped SNAREs

See allHide authors and affiliations

Science  13 Jun 2003:
Vol. 300, Issue 5626, pp. 1745-1749
DOI: 10.1126/science.1084909


The SNARE (soluble N-ethylmaleimide–sensitive factor attachment protein receptor) hypothesis suggests that pairs of proteins known as vesicle (v-) SNAREs and target membrane (t-) SNAREs interact specifically to control and mediate intracellular membrane fusion events. Here, cells expressing the interacting domains of v- and t-SNAREs on the cell surface were found to fuse spontaneously, demonstrating that SNAREs are sufficient to fuse biological membranes.

The fusion of intracellular membranes is mediated by cognate SNARE (soluble N-ethylmaleimide–sensitive factor attachment protein receptor) proteins (1) that assemble between lipid bilayers as SNAREpins (2). SNAREpins consist of a bundle of four helices (3). For fusion to occur, three of these (contributed by the target membrane or t-SNARE), must emanate from one of the membrane partners. The fourth helix, contributed by the cognate vesicle or v-SNARE, must be rooted in the opposing membrane (4). Energy made available from protein folding—the spontaneous assembly of the SNARE complex—is coupled by the SNAREpins to do work on the lipid bilayers, promoting fusion (5). SNARE-dependent fusion is exquisitely specific, so much so that the pattern of membrane fusion in the cell is predicted from the pattern of fusion of artificial liposomes by isolated SNAREs with an accuracy exceeding 99% (4, 68).

Can SNAREpins also fuse entire cells? Viral-encoded fusion proteins use helical hairpin bundles that are analogous to SNAREpins (9). SNARE proteins normally face the cytoplasm, within which their helical domains can pair to link membranes for fusion. Thus, to ascertain whether SNAREs can fuse cells, their orientation must be “flipped,” and cognate cells must be engineered that express either the v- or the t-SNAREs.

For this purpose (Fig. 1A), we inserted a signal sequence that specifies translocation across the endoplasmic reticulum (ER) ahead of the coding sequence of each of the three subunits of a plasma membrane exocytic SNARE complex: VAMP2, functionally defined as the v-SNARE, and Syntaxin 1A (Syntaxin) and SNAP-25, which together compose the t-SNARE (10). SNAP-25 contributes two helices to the bundle and Syntaxin contributes one. The membrane-anchoring sequence at the C-terminus of Syntaxin and VAMP is expected to terminate translocation initiated by the signal sequence and to anchor the flipped SNAREs to the plasma membrane after expression and intracellular transport. SNAP-25 is normally anchored to the cytoplasmic side of the plasma membrane by palmitoylation of a Cys-rich loop. To prevent disulfide bonds that could interfere with folding in the ER lumen (and which would not form in the reducing environment of the cytosol), we used a fully functional SNAP-25 mutant with all four Cys residues substituted by Ser to generate flipped SNAP-25 (5). Translocated SNAP-25 is expected to assemble with flipped Syntaxin on the lumenal side of the ER when they are coexpressed to form a flipped t-SNARE. The extracellular domains of the flipped v- and t-SNAREs do not contain any Cys residues.

Fig. 1.

Flipped SNAREs are expressed on the cell surface. (A) The domain structure of flipped SNAREs. The pre-prolactin signal sequence (SS) was fused to the N-terminus of VAMP2, the Syntaxin H3 domain, and SNAP-25. The putative N-glycosylation sites were eliminated (×) by point mutations: T27A in flipped VAMP2 and T79A and N188A in flipped SNAP-25. A Myc tag (red) was engineered between the N-terminus of the Syntaxin H3 domain and the signal sequence. The four cysteines in SNAP-25 were replaced with serines (C → S). TMD, transmembrane domain. The coiled-coil segments represent SNARE motifs. (B and C) Flipped SNAREs are glycosylated. Whole-cell lysates of COS cells transfected with flipped VAMP2, flipped VAMP2/T27A, empty pcDNA3.1(+) vector (Mock), both flipped Syntaxin and flipped SNAP-25, or both flipped Syntaxin and flipped SNAP-25/T79A,N188A were immunoblotted with (B) antibody to VAMP2 or (C) antibody to SNAP-25. Treatment with tunicamycin (5 μg/ml, overnight) or mutation of the glycosylation site(s) removed the higher molecular weight bands (arrows). (D) Expression of flipped v-SNAREs on the cell surface. Unpermeabilized COS cells expressing flipped GFP-VAMP2/T27A (GFP fluorescence, left panel) were stained with antibody to GFP (red, right panel). (E) Soluble t-SNAREs bind to flipped v-SNAREs. COS cells were transfected with an IRES plasmid that encoded flipped VAMP2/T27A and enhanced GFP (EGFP), then incubated with a soluble t-SNARE complex (5 μM) for 1 hour at 37°C. Surface-bound t-SNARE was detected with an antibody to Syntaxin (red, merged with EGFP fluorescence). No binding of the t-SNARE to the control cells (expressing EGFP only, right panel) was observed. (F) Expression of flipped t-SNAREs on the cell surface. COS cells cotransfected with flipped Syntaxin and flipped SNAP-25/T79A,N188A were stained with antibody to Myc (Syntaxin, green) and antibody to SNAP-25 (red). Top row: Intact cells. Bottom row: Permeabilized cells. Scale bars, 10 μm.

SNAREs are unfolded or incompletely folded except when they are in the four-helix bundle. In particular, v-SNAREs like VAMP2, with their single coil motif, are intrinsically unfolded (11), and the t-SNARE (Syntaxin and SNAP-25) exists in a partially folded state in which the membrane-distal portion of the t-SNARE forms a stable three-stranded helical bundle, but the membrane-proximal portion is unzipped (1214). Because the quality control system operative in the lumen of the ER serves to retain incompletely folded proteins (15), only a small fraction of the flipped SNAREs would be expected to reach the cell surface. In particular, the most unfolded SNARE, VAMP2, would be expected to be least efficiently transported. However, with a high enough level of expression, at least some of each SNARE should reach the cell surface.

In preliminary experiments, cell fusion only occurred when the cells that expressed the flipped v- and t-SNAREs were treated with tunicamycin, an antibiotic that inhibits Asn-linked glycosylation (16). This modification occurs in the lumen of the ER but not in the cytosol. SNAREs, which are not physiologically N-glycosylated, could become artifactually glycosylated when they are flipped. There are Asn-linked glycosylation sites (Asn-X-Ser/Thr) predicted within VAMP2 (Asn25), SNAP-25 (Asn77 and Asn188) and Syntaxin (Asn107 and Asn135).

Flipped VAMP2 (Fig. 1B) and flipped SNAP-25 (Fig. 1C) were indeed Asn-glycosylated when expressed in COS cells. The artifactual glycosylation was prevented by glycosylation-site mutations that generated flipped VAMP2/T27A and flipped SNAP-25/T79AN188A, which were used in all further experiments. Two artifactual glycosylation sites are predicted in Syntaxin, both within its N-terminal regulatory domain, Habc (17, 18), which folds back on the helical bundling domain to negatively regulate fusion by blocking access to the v-SNARE (19). Artifactual glycosylation, as well as substantial autoinhibition, were simultaneously eliminated by the removal of this domain in flipped Syntaxin.

COS cells transfected with flipped SNAREs expressed these proteins on the cell surface. Cells expressing flipped VAMP2/T27A tagged at the N-terminus with green fluorescent protein (GFP-VAMP2/T27A) showed a pattern of fluorescence that indicates some VAMP2 reached the cell surface, although the vast majority stayed inside (Fig. 1D), as would be expected from quality control considerations. Immunostaining of these unpermeabilized cells with an antibody to GFP confirmed the surface localization of a portion of the flipped VAMP2 (Fig. 1D).

Cell surface biotinylation and quantitative immunoblotting studies (20) revealed that 4.5 ± 1.8% (n = 3 experiments) of the flipped v-SNARE reached the cell surface, corresponding to 1.9 ± 0.6 × 106 (n = 3 experiments) copies of VAMP2 on the surface of each transfected cell. For a cell of 2000 μm2 surface area, this number of flipped v-SNAREs results in a surface density of ∼1000 per μm2, which is somewhat less than the surface concentration of VAMP2 in synaptic vesicles, ∼4000 per μm2, as estimated on a basis of ∼20 copies per vesicle of ∼40 nm diameter (21, 22).

The surface-localized flipped VAMP2/T27A could bind t-SNAREs (Fig. 1E). Here, COS cells transfected with flipped VAMP2/T27A were incubated with soluble cognate t-SNAREs, consisting of the bacterially expressed cytoplasmic domain of Syntaxin complexed with SNAP-25. Soluble t-SNAREs only bound to cells that expressed flipped VAMP2/T27A, and did not bind to cells that had been mock-transfected (Fig. 1E).

When COS cells were cotransfected with constructs that encode flipped SNAP-25/T79A,N188A and flipped Syntaxin both proteins were expressed on the cell surface (Fig. 1F). As expected from quality control, most of the flipped t-SNAREs remained within the cells, presumably within the ER (Fig. 1F), but 12.4 ± 3.7% (n = 3 experiments) of total flipped SNAP-25/T79A,N188A and 20.1 ± 6.4% (n = 3 experiments) of total flipped Syntaxin proteins were located on the cell surface [as determined by surface biotinylation (20)]. This corresponds to 5.6 ± 2.4 × 106 (n = 3 experiments) copies of flipped SNAP-25 on the cell surface of each transfected cell, which is similar to the number of viral fusion proteins expressed on the cell surface needed to trigger cell fusion (23).

To ascertain whether cells expressing flipped cognate v- and t-SNAREs fuse, we developed a color-mixing assay (Fig. 2A). The cytoplasm of cells that express flipped v-SNAREs (v-cells) was marked by a red fluorescent protein (RFP-nes) that contained a nuclear export signal. The nuclei of cells that express flipped t-SNAREs (t-cells) were marked by a cyan-colored variant (CFP-nls) of GFP that contained a nuclear localization (import) signal. In this way, fused cells could be distinguished easily because their red cytoplasm surrounded two or more blue nuclei: The CFP-nls from the t-cells and the CFP-nls newly synthesized in the fused cells will equilibrate among the nuclei present in a common cytoplasm after fusion.

Fig. 2.

Cell-cell fusion mediated by flipped SNAREs. (A) Flipped v- and t-SNAREs were expressed on the surface of two separate cell populations. The cytoplasm of COS cells that express flipped v-SNAREs was labeled by RFP-nes. The nuclei of the cells that express flipped t-SNAREs were labeled by CFP-nls. SNARE-dependent cell-cell fusion generated fused cells with a red cytoplasm that surrounds two or more cyan nuclei. Cells expressing flipped v-SNAREs were detached from culture dishes, then seeded on a coverslip that already contained cells that expressed flipped t-SNAREs. Cells were fixed after 6 hours at 37°C. Representative confocal images are shown. (B) In the fused cells, multiple cyan nuclei (arrowheads) shared a common red cytoplasm. Neighboring unfused cells had either red cytoplasm or cyan nuclei. Arrows indicate dark nuclei that likely result from the most recent cell fusion events in which CFP-nls has not reached equilibrium. (C) Controls. The cytoplasmic domain of VAMP2 (20 μM) completely inhibited cell fusion mediated by flipped SNAREs (+ CD-VAMP). Cell fusion was not observed when either the v-SNARE (– Flipped VAMP) or one of the t-SNAREs (– Flipped SNAP-25) was omitted from transfection. Scale bars, 10 μm. Graph: Cell fusion was not observed with the noncognate v-SNARE flipped Bet1; with a functionally inactivated flipped SNAP-25/T79A,N188A mutant (SNAP-25/I178A,I181A); or with a 26–amino acid C-terminal deletion of flipped SNAP-25/T79A,N188A (SNAP-25Δ26). A 9 –amino acid deletion in flipped SNAP-25 (SNAP-25Δ9) greatly diminished cell fusion. Cell fusion was quantitated as in Fig. 4.

For v-cells, a plasmid encoding flipped VAMP2/T27A was cotransfected with a plasmid that encoded RFP-nes. For t-cells, an internal ribosomal entry site (IRES) plasmid was used that encoded both flipped SNAP-25/T79A,N188A and CFP-nls on the same transcript, and this was cotransfected with a plasmid that encoded flipped Syntaxin. Typically, about 30 to 40% of the COS cells expressed either v- or t-SNAREs as a result of transfection. The v-cells were harvested without proteolysis (with an EDTA buffer) and overlaid on a coverslip that contained an ∼80% confluent monolayer of t-cells. Enough v-cells were added to saturate the coverslip as a monolayer.

Cell fusion became apparent as the v-cells began to spread and contact t-cells, which typically required about an hour and continued thereafter (Fig. 2B, Fig. 3). Many examples of multinuclear cells were apparent (Fig. 2B), indicating that multiple fusions took place. Fusion was not observed (Fig. 2C) when the cytoplasmic domain of VAMP2 was added to titrate the flipped t-SNAREs on the surface of t-cells, or when flipped VAMP2/T27A or flipped SNAP-25/T79A,N188A was omitted from transfection.

Fig. 3.

The reconstructed three-dimensional structure of fused cells. (A) 0.5-μm confocal optical (Z-) sections were taken from a fused cell. The reconstructed three-dimensional image can be viewed from the top (left panel) and from the side (right panel). Multiple cyan nuclei can be identified within a single red cytoplasm. (B) Consecutive Z-sections through the fused cell. Scale bars, 10 μm.

In additional controls (Fig. 2C), fusion was not detectable with a noncognate v-SNARE (flipped yeast Bet1), despite the fact that flipped Bet1 was expressed on the cell surface. As another test, we employed a functionally inactivated SNAP-25 mutant (I178A,I181A) that assembles into the SNARE complex but with insufficient stability to drive exocytosis in neuroendocrine cells (24). Although this protein was expressed on the cell surface to a similar extent as SNAP-25/T79A,N188A, cell fusion did not occur (Fig. 2C). Further, a C-terminal–deleted SNAP-25 mutant (Δ26), which represents the product of cleavage by botulinum neurotoxin E (BoNT/E), was expressed on the cell surface but was also inactive in cell fusion. Deletion of the nine C-terminal residues (Δ9), representing the product of cleavage by botulinum neurotoxin A (BoNT/A), markedly reduced but did not completely prevent cell fusion, which is consistent with the complete blockage of neuroendocrine secretion by BoNT/E but not BoNT/A (25). In a typical experiment, 0.9 × 106 copies of flipped Bet1, 1.4 × 106 copies of flipped GFP-VAMP2/T27A, 3.2 × 106 copies of flipped SNAP-25/T79A,N188A, 4.2 × 106 copies of flipped SNAP-25/I178A,I181A, 1.1 × 106 copies of flipped SNAP-25Δ26 and 3.1 × 106 copies of flipped SNAP-25Δ9 were expressed on the surface of each transfected cell.

SNARE-dependent cell fusion was efficient: 37% of the v- and t-cells that were visibly in contact fused (Fig. 4A). Furthermore, ∼40% of the fused cells had three or more nuclei (e.g., at the 24-hour time points), apparently the product of multiple fusion events. Thus, the measured efficiency of SNARE-dependent cell fusion should be regarded as a conservative underestimate. Fusion of cells was accelerated by a peptide (Vc-peptide) derived from the C-terminal membrane-proximal half of VAMP2 (12). This peptide targets an intrinsic switch in the t-SNARE that controls the rate at which the helical bundle can zip up to drive bilayer fusion. The peptide binds reversibly to its cognate region in the t-SNARE and prestructures this region into a coil ready to accept the corresponding region of the v-SNARE. Vc-peptide accelerated cell fusion as it does SNARE-dependent fusion of liposomes (12), implying that zippering is a rate-limiting step in cell fusion.

Fig. 4.

The time course of cell fusion by flipped SNAREs. (A) To quantitate cell fusion and to clearly identify those v- and t-cells that were in contact but did not fuse, a second cell fusion readout system was used, in which EGFP (v-cells) and red fluorescent protein DsRed2 (t-cells) labeled the cytosol. At different time points, we determined the total number of fused cells (F, orange cytosol) in 50 random fields. In parallel, we also determined the total numbers of unfused v-cells (V, green cytosol) and unfused t-cells (T, red cytosol) that were in contact with each other or with fused cells. The efficiency of fusion (%) was calculated as follows: 2F/(V + T + 2F) × 100. Because this quantitation method does not determine the number of nuclei in a fused cell, it will not detect multiple fusion events, and consequently will underestimate the number of fusion events. The Vc-peptide (60 μM) promoted SNARE-dependent cell fusion, whereas the cytoplasmic domain of VAMP2 (CD-VAMP2, 20 μM) completely inhibited it. t-cells expressing flipped Syntaxin but not flipped SNAP-25/T79A,N188A did not fuse with v-cells in the presence or absence of the Vc-peptide. The data shown represent three independent experiments. (B) Live-cell imaging of SNARE-dependent cell fusion. Selected frames are shown from a movie (movie S1) recording the fusion process in Fig. 2. In this example, t = 0 corresponds to 9 min after the addition of Vc-peptide.

In spite of the fact that the overall fusion of a cell population takes hours (Fig. 4A), time-lapse video recording revealed that the fusion of individual cells took place much more rapidly, typically over 1 to 2 min (Fig. 4B and movie S1). This is consistent with the fact that we rarely observe apparent intermediates in the fusion process. The redistribution of CFP-nls after fusion (from the initially labeled nuclei from t-cells into the initially unlabelled nuclei from v-cells) took much longer (>60 min), which reflects the kinetics of nuclear trafficking processes.

The efficiency of fusion by flipped SNAREs correlates with the level of expression of flipped SNAREs on the cell surface (26). This is well documented for viral-mediated cell fusion (23), which requires multiple fusion proteins to act cooperatively. SNARE-dependent fusion is also cooperative (27) and strongly concentration-dependent in liposome systems (2). The liposomes fused by reconstituted SNARE proteins, though in the size range of many transport vesicles, are much smaller than most organelles. The fact that flipped SNAREs can fuse entire cells, which are by definition larger than any organelle, establishes that the size of the membrane is unimportant for SNARE-dependent fusion.

The well-accepted concept of viral-encoded fusion proteins (i.e., that a specific subunit of the viral envelope comprises the active principle that fuses the virus with the cell) is rooted in three lines of evidence. First, mutation of the gene that encodes the putative fusion protein blocks fusion. Second, the isolated putative fusion protein mediates fusion of artificial lipid bilayers under physiologically relevant conditions. Third, expression of the putative fusion protein on the surface of cells triggers cell fusion. Mutations in SNARE genes have long been known to prevent fusion in vivo, and isolated cognate SNAREs have previously been shown to efficiently fuse artificial bilayers. With the current results, the evidence that SNAREs provide the core, general principle of intracellular membrane fusion now rises to the same level of proof as for viral fusion proteins.

Certain other cellular membrane proteins have been posited to play this role instead of SNAREs or to requisitely cooperate with SNAREs in doing so (28, 29). The level of evidence in support of this does not rise to the level of proof now available for viral fusion proteins and SNAREs. Moreover, in the present study these proteins would of course not have been flipped to the outside of cells along with the cognate SNAREs, and therefore they cannot be the part of the cell's core machinery—SNAREpins—that is demonstrably endowed with the capacity to efficiently merge the lipid bilayers of cellular membranes. Nonetheless, such proteins may still potentially have a regulatory role in certain specialized fusion events within the cell.

Cell fusion occurs physiologically. It is vital for fertilization (sperm-egg), development (forming syncytial tissues such as muscle), and physiology (bone resorption). Although progress is being made, the responsible fusion proteins are still unknown (30). However, it is now easy to imagine how DNA rearrangements that linked signal sequences to SNAREs could have triggered an evolutionary process that led to the creation of proteins to mediate physiologic cell fusion.

Supporting Online Material

Materials and Methods


Movie S1

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article