A Clamping Mechanism Involved in SNARE-Dependent Exocytosis

See allHide authors and affiliations

Science  04 Aug 2006:
Vol. 313, Issue 5787, pp. 676-680
DOI: 10.1126/science.1129450


During neurotransmitter release at the synapse, influx of calcium ions stimulates the release of neurotransmitter. However, the mechanism by which synaptic vesicle fusion is coupled to calcium has been unclear, despite the identification of both the core fusion machinery [soluble N-ethylmaleimide–sensitive factor attachment protein receptor (SNARE)] and the principal calcium sensor (synaptotagmin). Here, we describe what may represent a basic principle of the coupling mechanism: a reversible clamping protein (complexin) that can freeze the SNAREpin, an assembled fusion-competent intermediate en route to fusion. When calcium binds to the calcium sensor synaptotagmin, the clamp would then be released. SNARE proteins, and key regulators like synaptotagmin and complexin, can be ectopically expressed on the cell surface. Cells expressing such “flipped” synaptic SNAREs fuse constitutively, but when we coexpressed complexin, fusion was blocked. Adding back calcium triggered fusion from this intermediate in the presence of synaptotagmin.

In the cell membrane, trafficking occurs constitutively. However, exocytosis—the fusion of vesicles containing stored secretory product—is tightly regulated by external signals and is often highly synchronized. Nevertheless, regulated exocytosis and constitutive vesicle trafficking rely on the same machinery—cognate vesicle (v-) and target membrane (t-) SNARE proteins. How, then, can fusion by a common mechanism be tightly triggered in one instance but occur spontaneously in another (1)?

It has been suggested that the fusion machinery is constitutively “on,” so that exocytosis—in which fusion is “off” in the absence of a signal for secretion—would require a protein “clamp” to block fusion in the basal state. Then, an additional protein “trigger” would be needed to reverse the clamp when a signal for secretion appears. Accumulation of fusion intermediates arrested at a discrete stage would create a synchronized and robust burst of secretion in response to urgent but transient physiological need. Isolated SNAREs can spontaneously fuse bilayers (2) on a time scale shorter than secretions can be stored—hours to days—but direct evidence for a clamp mechanism has been lacking.

Complexins are ∼20 kD proteins associated with the SNARE complexes mediating exocytosis at neuronal synapses (35) and may represent candidates to act as fusion clamps. The human genome encodes four complexins that are broadly coexpressed in neuronal and neuroendocrine cells, of which two are soluble proteins and two are membrane-anchored by C-terminal prenylation (6). Whereas complexin binds tightly to the assembled exocytic SNARE complex in a 1:1 fashion, it binds weakly or not at all to its individual v- and t-SNARE subunits (4, 79). Complexin binds in the groove of the four-helix bundle of the SNARE complex (10) in its membrane-proximal half. If complexin could bind analogously to the zippering SNAREpins that drive fusion, it could potentially clamp fusion by arresting the fusion mechanism (11), but it could also promote fusion by stabilizing a key intermediate (12).

Genetic and physiological studies are difficult to interpret in this regard. Increasing the concentration of complexin by presynaptic injection inhibits neurotransmission (13), but decreasing the concentration of complexin by knocking out two of the four complexins also blocks synchronous neurotransmitter release (14). Although complexins do not themselves appear to bind calcium (4), the phenotype of the double knockout can be bypassed simply by raising extracellular calcium (14).

Synaptotagmins I and II are the calcium sensors for synchronous synaptic transmission (15). Residing mainly in synaptic vesicles, they are members of a large family of neurally expressed proteins with two calcium-binding C2 domains (16). Knockout of mouse synaptotagmin I abolishes the fast synchronous component of calcium-mediated release (17). Mutations of synaptotagmin I reduce or eliminate the apparent cooperativity of calcium and alter the vesicle-release probability (18). Most important, defined mutations in synaptotagmin's calcium-binding site shift the calcium sensitivity of neurotransmission in a manner that reflects calcium binding by the isolated protein (19, 20).

The phenotypes of complexin and synaptotagmin deletions are virtually identical, suggesting that they cooperate in some fashion in calcium-triggered exocytosis (14, 17), whether directly or indirectly. Unfortunately, the complex nature of the phenotypes and the presence of multiple isoforms preclude further insights into molecular mechanism.

The outside of a cell normally lacks SNAREs and cytoplasmically located regulatory proteins, which can be added back—either directly to the medium or by expression on the cell surface with artificially added signal peptides—to test directly for function in regulating fusion.

Here, we used fusion of intact cells mediated by flipped SNAREs–SNAREs ectopically expressed on the cell surface–to establish the functional capabilities of synaptotagmin and complexin in a biological context that is compositionally virgin (21).

To look for a direct modulatory role of complexin on fusion, we introduced bacterially expressed recombinant human complexin I into the flipped SNARE assay. Complexin inhibited the fusion of flipped-SNARE–expressing cells in a concentration-dependent manner (Fig. 1A), up to at least 56%. The complexin titration is not at a plateau, but we were unable to test higher concentrations. Thus, we sought to increase the local concentration of complexin I by “flipping” the protein and artificially anchoring it to the cell membrane by a glycerophosphatidylinositol (GPI) anchor (Fig. 1B). Under these conditions, residual fusion from cells expressing complexin-GPI (Cpx-I-GPI) was only 3% of the fusion observed in control cells (Fig. 1, C and D) and within the background level of fusion observed even in the presence of a competitive inhibitory VAMP2 fragment (CD-Vamp2) (Fig. 1D). Thus, SNARE-dependent fusion was completely abolished in Cpx-I-GPI–expressing cells. In contrast, a complexin point mutant with a critical VAMP2-binding residue mutated (R59H) (12, 22) was without effect (Fig. 1D). Cpx-I-GPI was an equally effective inhibitor when expressed on the v- or t-SNARE–expressing cells, suggesting that the 38 amino acid GPI linker provides the protein with sufficient conformational flexibility to behave essentially as a soluble protein.

Fig. 1.

Complexin inhibits SNARE-mediated fusion. Cell fusion (21) was performed between HeLa cells stably expressing flipped t-SNAREs (and cyan fluorescent protein in the nucleus) (t-cells) and HeLa cells stably expressing flipped VAMP2 (and red fluorescent protein in the cytoplasm) (v-cells) in Dulbecco's modified media containing 1.8 mM calcium, unless otherwise noted. Proteins were either transiently transfected together with a cotransfection marker [a variant of yellow fluorescent protein bearing a nuclear localization signal (YFP-nls)] or exogenously added as recombinant protein to transient transfectants [see (36) for details]. Experiments were quantified as percentage of YFP-nls transfected cells with blue nuclei. These axes differ from previous publications (21, 30), where fusion was measured as a percent of “contacting v- and t-cells.” The absence of a cytosolic or membrane stain for t-cells prevented such a measurement here, and therefore the fusion efficiency should be considered an underestimate. Data are mean ± SEM. (A) Recombinant complexin I was added at the indicated concentrations. (B) Domain structure of flipped regulatory proteins. The pre-prolactin signal sequence (SS) was fused to the N terminus of an AU1 epitope–tagged full-length human complexin I followed by a GPI anchoring motif. Flipped full-length synaptotagmin-I was generated by inverting the balance of charges around the transmembrane domain (TMD) (37). An N-terminal HA epitope was added, and cysteine was point-mutated to serine (C277S). (C) Cell fusion experiment between t-cells (cyan-nuclei, arrowheads), and v-cells (red cytoplasm) transiently transfected with YFP-nls (control) or YFP-nls plus Cpx-I-GPI. Fused cells have red cytoplasm and nuclei staining for both cyan and YFP (appears white; asterisks). Scale bar, 50μM. (D) Regulatory protein effects on cell fusion.

We also developed a flipped variant of synaptotagmin I (Fig. 1B and fig. S1). In contrast to complexin, neither flipped synaptotagmin (Syt-I) nor a soluble C2A-C2B fragment produced a significant change in SNARE-mediated fusion (Fig. 1D). In the following experiments, only the membrane-embedded Syt-I was used, but indistinguishable results were obtained with soluble synaptotagmin.

Phosphatidylinositol-specific phospholipase C (PI-PLC) cleaves between the glycerol backbone and the phosphate group of GPI. Addition of PI-PLC to Cpx-I-GPI–expressing cells released complexin from the cell surface (fig. S2, A and B). When PI-PLC was added to Cpx-I-GPI clamped fusion assays, there was a rapid recovery of fusion activity, accomplishing in 5 min what ordinarily takes several hours (Fig. 2A). The burst of fusion suggests that the SNAREs were at least partially assembled but could not complete fusion in the presence of Cpx-I-GPI. The burst plateaued at about 25% of overnight fusion, and the incomplete recovery could not be attributed to limiting concentrations of PI-PLC (fig. S2C). Instead, the relatively low recovery suggests that some complexin-SNARE clamped complexes remained associated even after GPI cleavage.

Fig. 2.

Recovery of the cell fusion activity by PI-PLC cleavage of Cpx-I-GPI. v- and t-cells were incubated overnight before addition of 1 U/ml PI-PLC (t = 0, where indicated) to release the complexin GPI anchor. v-cells were transiently transfected with (A) Cpx-I-GPI or (B) Syt-1 and Cpx-I-GPI. Control curves have no Cpx-I-GPI. The resultant cell fusion activity was determined as in Fig. 1. The average percentage recovery to control values (from five independent experiments) at the 10-min time point was 34.2% ± 2.8% for complexin alone and 68.8% ± 2.6% for the samples containing synaptotagmin. Total free calcium equals 1.8 mM.

When full-length membrane-embedded Syt-I was present on the v-cell membrane, recovery from the clamped state by addition of PI-PLC was enhanced by a factor of 2 (Fig. 2B) (in cell culture medium containing 1.8 mM calcium). However, in the absence of PI-PLC, Cpx-I-GPI continued to completely clamp the fusion reaction. Thus, synaptotagmin may facilitate the removal of complexin but cannot overcome the very high local concentration of complexin maintained by the GPI anchor.

In the ideal recapitulation of synaptic vesicle exocytosis, the v- and t-cells would be incubated together (plus complexin and synaptotagmin) in the presence of very low calcium. In this context, we could most easily assess the role of calcium. A limitation of our assay system is that the cells require calcium to remain stably associated with the underlying glass support but tolerate low levels of calcium for ∼40 min. Thus, we performed the following experiments with overnight incubations in the presence of calcium, followed by a transient EGTA [ethylene glycol bis-(2-aminoethyl ether)–N,N,N′N′-tetraacetic acid]–induced drop to 10 μM free calcium. Under these conditions, PI-PLC cleavage could be carried out in the absence of calcium and before activation of synaptotagmin.

There was no recovery of fusion following the PI-PLC cleavage when Syt-I was present but free calcium was kept low (10 μM) (Fig. 3A). Reduction to 100 nM free calcium gave indistinguishable results (fig. S3). This synaptotagmin-containing intermediate recovered full fusogenicity upon addition of calcium (Fig. 3A) over a calcium range (∼200 μM) that promotes synaptotagmin binding to SNARES (23) (Fig. 3B). Synaptotagmin binds the acidic lipid PIP2, but not PIP3, in a calcium-dependent manner (24). Likewise, recovery from the clamped state was stimulated by PIP2, but not PIP3 (fig. S4), and the half-maximal calcium concentration needed for stimulation is lower when the plasma membrane is enriched in PIP2 (but not PIP3) (Fig. 3B). Neither cleavage by PI-PLC (25) nor recovery from the complexin-block (in the absence of Syt-I) were calcium-sensitive (Fig. 3C). To eliminate the possibility that calcium-free Syt-I was simply interfering with PI-PLC cleavage of Cpx-I-GPI, we carried out order-of-addition experiments, with PI-PLC, a PI-PLC inhibitor, and calcium (Fig. 3D), confirming that both complexin and synaptotagmin were associated with the clamped complex in the absence of calcium.

Fig. 3.

Ca++-dependent release of the complexin block. v- and t-cells were incubated overnight before addition of 1 U/ml PI-PLC and 1.8 mM EGTA (t = 0, where indicated) to release the complexin GPI anchor and reduce free calcium in the media to 10 μM. At t = 5 min, free calcium was raised to 200 μM (unless otherwise indicated). Control assays are without Cpx-I-GPI. Cells were fixed and data quantified as in Fig. 1. (A) Calcium dependence of synaptotagmin I mediated recovery. v-cells were transiently transfected with Cpx-I-GPI and Syt-I. (B) Calcium dependence of recovery of cell fusion. Cell fusion experiments as in (A), with v-cells expressing Cpx-I-GPI and Syt-I. Before starting the recovery, the cells were incubated for 3 hours in the presence of either buffer, 80 μM PIP2 or 80 μM PIP3. EGTA was introduced (t = 0) at 2 mM to give a free calcium concentration of 500 nM. After the 5-min EGTA/PI-PLC incubation, calcium was added to give the indicated free concentrations. Fusion recovery at 10 min is plotted as percent of the control fusion. (C) Recovery in the absence of Syt-I is not calcium responsive. v-cells were transfected with Cpx-I-GPI. (D) The calcium-sensitive complex includes soluble complexin. As in (A), except a PI-PLC inhibitor (U73122) was included before calcium addition (green curve) or coincident with PI-PLC addition (blue curve). (E) Syt-I does not stabilize clamp when expressed in the t-cells. Experiment is the same as 3A except Syt-I is transiently transfected in t-cells. v-cells were transfected with Cpx-I-GPI. (F) Effect of synaptotagmin C2A and C2B mutants on the fusion recovery. v-cells were transiently transfected with Cpx-I-GPI plus wild-type Syt-I (squares), C2A mutant (D178N) (triangles), C2B mutant (D303N,D309N) (crosses), or the double C2A/C2B mutant (D178N,D303N,D309N) (circles). Control experiments as squares but not transfected with Cpx-I-GPI (diamonds). Data for all panels are the mean ± SEM of two independent experiments.

When synaptotagmin was expressed on the t-cell (fig. S1B), neither the stabilization of the clamp nor the calcium-dependent enhancement of fusion recovery was maintained (Fig. 3E). Thus, flipped synaptotagmin maintains a topological restriction that is consistent with the principal synaptic vesicle localization of synaptotagmin and reminiscent of a topological restriction observed in liposomes (26). The calcium responsiveness was reduced when a conserved calcium-chelating aspartic acid residue was mutated in the Syt-I C2A domain and virtually eliminated when residues in the C2B domain were mutated (Fig. 3F), which parallels the consequences of mutating conserved aspartic acids in vivo (15).

The fast recovery kinetics implies the existence of a trans-SNARE clamped complex. Accessibility experiments with molecules that associate with different regions on the unstructured/uncomplexed individual SNAREs should reveal the extent of complex assembly. First, fusion appeared to be clamped after the initiation of SNAREpin assembly, because fusion proceeded after anchor cleavage even in the presence of the inhibitory v-SNARE cytoplasmic domain (fig. S5), which would bind and titrate out uncomplexed free t-SNAREs. Second, the clamped SNAREpin appeared not to be fully assembled, as assessed by neurotoxin treatment. Botulinum toxin B (BoNT/B) and Tetanus toxin each bind to free VAMP2 and cleave after the same amino acid (27), although Tetanus binds at a membrane distal position and BoNT/B binds more membrane proximal (28). When added to v cells prior to mixing with t-cells, each completely inhibited cell-cell fusion (Fig. 4A). When added to a cell-cell assay after accumulation of the clamped state, however, the toxins showed distinct effects. The assay remained completely resistant to the membrane-distally targeted tetanus but was partially susceptible to BoNT/B (Fig. 4B). Before nerve terminal activity, similar toxin accessibility differences are found in crayfish neuromuscular junctions (29). Fusion was clamped before a hemifusion intermediate, as judged by the lack of mixing of the ganglioside GM1 between cells (fig. S6), using an assay previously published (30). Thus, it is likely that the clamped trans-SNARE complex includes both calcium-free synaptotagmin and soluble complexin, a multi-component complex that may resemble the prestimulation configuration of presynaptic SNAREs.

Fig. 4.

Effect of neurotoxins on cell fusion recovery. (A) BoNT/B and TeNT inhibit the cell fusion reaction. v-cells were incubated for 3 hours with buffer (Mock), BoNT/B, or TeNT, then washed and mixed with t-cells overnight. (B) Recovery from the clamped state is partially sensitive to BoNT/B. Cell fusion was performed as in Fig. 3A, using v/Cpx-1-GPI/Syt-1-cells. Before starting the recovery, the cells were incubated for 3 hours in the presence of buffer (control) (squares), BoNT/B (triangles), or TeNT (diamonds). Cells were washed, then incubated with EGTA and PI-PLC (at t = 0) (to give baseline free calcium of 10 μM). At t = 5 min, Ca++ (200 μM free final) was added to the reaction. The extent of fusion was quantified as in Fig. 1.

Functional reconstitution has revealed a fundamental property of complexin: its ability to operate as a reversible molecular clamp. Clamping is enhanced by membrane anchoring of complexin, either by a GPI anchor (here), or, for some complexins, by prenylation (6). Clamping was observed whether complexin was anchored to the v- or t-SNARE–expressing cells, consistent with the behavior of a soluble protein, and was abrogated in a point mutant that also prevents overexpressed complexin from inhibiting exocytosis (22). Fusion was clamped after the initiation of SNAREpin assembly but before the SNAREs were fully zippered.

Synaptotagmin is a thoroughly studied protein whose role as the sensor for synchronous transmitter release is well established (15, 31). We suggest that another fundamental property of synaptotagmin may be the capacity to couple the calcium signal to SNAREs in a mechanism that requires complexin. Coupling of fusion to calcium is only reconstituted when synaptotagmin is added or expressed together with complexin. Synaptotagmin could couple from the v-side but not the t-side of the reaction, reflecting its primary location to synaptic vesicles in vivo. Mutation of residues in synaptotagmin's calcium-binding sites that are critical for calcium coupling in vivo also prevented reconstituted coupling, and coupled fusion was enhanced by phosphoinositide.

The calcium sensitivity (∼200 uM) lies within the expected range of calcium microdomains (32) and physiologically is sufficient for some presynaptic terminals (31) but not others. Reintroduction of just one synaptotagmin-specific lipid (PIP2) increased our sensitivity to ∼40 μM. Replenishment of the full spectrum of acidic lipids, including especially phosphatidylserine, may ultimately bring the assay into the low micromolar range. The lack of phosphatidylserine may also bear on a second partial discrepancy with other studies. Provision of synaptotagmin without complexin neither inhibited nor significantly stimulated fusion, even in the presence of calcium, showing that synaptotagmin is neither inherently a clamp nor a fusion protein. This characteristic differs from liposome reconstitutions, where Ca-synaptotagmin alone stimulates SNARE assembly and SNARE-mediated proteoliposome fusion (33, 34). However, each of these activities is strongly or entirely dependent on the presence of phosphatidylserine in the liposomes.

By their nature, reductionist systems that are compositionally defined are artificial, and in vivo studies will ultimately be required to confirm the physiological relevance of the inherent properties we have uncovered. To this end, the recent observation of increased spontaneous fusion events in cortical neuronal synapses derived from synaptotagmin knockout mice (35) is consistent with a negative (i.e., clamp-associated) role for synaptotagmin. To date, only autapses have been studied in complexin knockout mice (14), in which exocytosis was dramatically reduced, perhaps suggesting an additional positive role of complexin in stabilizing or activating the fusion machinery. The existing genetics are also confounded by the large number of potentially compensatory synaptotagmin and complexin isoforms.

Fusion triggered from clamped flipped SNAREs takes about 5 min in the reconstituted system as compared with milliseconds or less in synapses. It is possible that this difference reflects the kinetics of enzymatic cleavage of GPI-complexin (i.e., GPI-complexin bound to SNAREs is a poor substrate) or another property of GPI-complexin (perhaps arresting fusion at an earlier stage or perhaps the absence of other important factors from our defined system). However, we think a likely explanation may be that msec synaptic transmission requires only the transient opening of a fusion pore to allow egress of transmitter, whereas our assay measures macromolecular content mixing between cells, which will always be much slower than the initial fusion pore.

Non-neuronal exocytosis, needed for the diverse forms of intercellular communication and physiological response in the body, may also rely on a similar clamping principle—although using different clamps—and similar triggering mechanisms to govern the common and constitutive engine for membrane fusion, SNARE proteins.

Supporting Online Material

Materials and Methods

Figs. S1 to S6

Table S1


References and Notes

View Abstract

Navigate This Article