Complexin Binding to Membranes and Acceptor t-SNAREs Explains Its Clamping Effect on Fusion

Abstract

Complexin-1 is a SNARE effector protein that decreases spontaneous neurotransmitter release and enhances evoked release. Complexin binds to the fully assembled four-helical neuronal SNARE core complex as revealed in competing molecular models derived from x-ray crystallography. Presently, it is unclear how complexin binding to the postfusion complex accounts for its effects upon spontaneous and evoked release in vivo. Using a combination of spectroscopic and imaging methods, we characterize in molecular detail how complexin binds to the 1:1 plasma membrane t-SNARE complex of syntaxin-1a and SNAP-25 while simultaneously binding the lipid bilayer at both its N- and C-terminal ends. These interactions are cooperative, and binding to the prefusion acceptor t-SNARE complex is stronger than to the postfusion core complex. This complexin interaction reduces the affinity of synaptobrevin-2 for the 1:1 complex, thereby retarding SNARE assembly and vesicle docking in vitro. The results provide the basis for molecular models that account for the observed clamping effect of complexin beginning with the acceptor t-SNARE complex and the subsequent activation of the clamped complex by Ca²⁺ and synaptotagmin.

Introduction

In regulated exocytosis, an increase in intracellular calcium triggers the fusion of secretory and synaptic vesicles with the plasma membrane, thereby releasing the contents of the vesicle. This fusion process is controlled by a highly specialized fusion machinery, where protein-soluble N-ethylmaleimide-sensitive factor attachment protein receptors (SNAREs) lie at the core of this machinery (1, 2, 3). SNAREs consist of conserved stretches of 60–70 amino acid residues called “SNARE motifs” that are involved in the formation of a highly stable four-helix coiled-coil structure termed the “SNARE complex” (4, 5).

Three SNARE proteins are present in regulated neuronal exocytosis: synaptobrevin-2 (Syb) on the vesicle membrane, and syntaxin-1a (Syx) and SNAP-25a (SNAP25) on the plasma membrane, where SNAP25 contributes two helices to the complex (6, 7, 8). Before fusion, the SNAREs are thought to be present in a partially assembled trans complex that is anchored through C-terminal transmembrane helical regions on Syb and Syx, in the secretory vesicle and plasma membranes, respectively (3). In this state, the vesicle and target membranes are in close proximity (9), and fusion occurs after the assembly of the SNAREs in an N- to C-terminal direction (10). This highly exergonic process drives vesicle and target membranes into contact and overcomes the energy barriers to fusion (11).

SNAREs are the minimal proteins required to reconstitute membrane fusion (12); however, a number of additional proteins are required to regulate SNARE assembly, produce efficient fusion, and provide a response to calcium (13, 14, 15, 16). Complexin-1 (Cpx) is a highly charged protein that is essential for Ca²⁺-mediated neurotransmission that appears to act by interacting with and regulating the SNAREs. Several studies have examined the role of Cpx in membrane fusion and yielded seemingly contradictory results (17, 18, 19, 20, 21, 22). Complexin-1 exerts both positive and negative effects on vesicle exocytosis, facilitating synchronous neurotransmission while inhibiting spontaneous fusion events. Complexin-1 seems to act at a postpriming step of the fusion process, as neurons extracted from cpx-1, cpx-2 double-knockout mice display a large reduction in Ca²⁺-triggered exocytosis, while still maintaining a wild-type number of primed vesicles (23). Additionally, different regions within Cpx appear to have different actions. The central α-helix (see Fig. 1) binds to SNARE complexes and appears to be essential but not sufficient for the protein to be fully functional. A more N-terminally localized accessory α-helix is critical for an inhibitory (clamping) effect of Cpx on spontaneous neurotransmission, whereas the unstructured amino-terminus facilitates evoked exocytosis (24). A long and flexible C-terminal region is required for clamping and priming but not for Ca²⁺-triggering of fusion events (25). In reconstituted systems, Cpx reportedly modulates the docking and fusion events between membranes containing t-SNAREs (Syx and SNAP25) and the v-SNARE (Syb). Depending on the study, Cpx has been reported to inhibit (26), facilitate (27, 28), or not affect docking at all (29), while suppressing spontaneous fusion and activating Ca²⁺-triggered fusion.

Figure 1.

Complexin interacts with the core SNARE complex. (a) Complexin-1 contains regions reported to bind membranes and SNAREs. The crystal structure of a fragment of Cpx bound to the core SNARE complex is shown (PDB: 1KIL) where SNAP-25a, syntaxin-1a, and synaptobrevin-2 helices are indicated in green, red, and blue, respectively. In this study, the spin-labeled side chain R1 was engineered into multiple sites in Cpx, covering the N- and C termini, and the central helical region (magenta spheres indicate the position of several labeled sites on Cpx in the Accessory and Central helices). (b) Shown here are X-band cw-EPR spectra for 13 spin-labeled Cpx mutants in solution (black trace) or incubated with the assembled core SNARE complex (red trace). Sites that interact with the SNAREs exhibit a broadening of the EPR spectrum and a decrease in normalized amplitude. All spectra are 100 Gauss scans with at least 10 averages taken. (c) Shown here is the ratio of the normalized intensities for the high-field resonance line of each mutant in the presence/absence of the core SNARE complex. Error bars represent the uncertainty in the normalized intensity ratio due to normalization and measurement of peak-to-peak amplitudes. To see this figure in color, go online.

Complexin-1 binds to the fully assembled postfusion SNARE core complex (30), and a crystal structure of a fragment of Cpx (residues 26–83) bound to this complex indicates that Cpx associates as a fifth α-helix (31) in an antiparallel fashion between Syx and Syb (see Fig. 1). The interaction involves the central helix of Cpx, leaving the more N-terminally localized accessory helix free while producing no significant conformational changes within the SNARE complex. In another crystal structure, the accessory helix bridges between two SNARE core complexes, leading to the proposal that cross linking of SNARE complexes accounts for the inhibitory activity of Cpx (32). Binding of Cpx to defined prefusion SNARE complexes composed of Syx and SNAP25 (i.e., without Syb) has been difficult to establish, although such an interaction could be inferred from attenuated Syb vesicle binding to t-SNARE vesicles that were preincubated with Cpx (33). Early work indicated that there were no interactions between Cpx and the soluble, individual SNAREs or to plasma membrane t-SNARE complexes composed of Syx and SNAP25 (34). In one study using single molecule methods, Cpx was reported to interact with and stabilize a 1:1 binary t-SNARE complex (35). However, the same group recently found that their SNARE complex preparations were heterogeneous, resulting in unspecific subconfigurations unless a stabilizing dNSyb peptide was used to assemble a well-defined SNARE acceptor complex (36). Similarly, a different group demonstrated Cpx binding to a t-SNARE complex by electron paramagnetic resonance (EPR) and proposed that high microMolar concentrations of complexin might act to inhibit vesicle docking through a weak interaction with the target membrane t-SNAREs (26). However, a more recent study by this group claimed that Cpx primes t-SNAREs by an unknown mechanism, and speculated that Cpx might play a role in converting an inactive 2:1 Syx/SNAP25 complex to an active complex (29).

Preparations of t-SNAREs are frequently not homogenous, and bulk assays may be complicated by signals from higher order oligomers of Syx (37) and 2:1 Syx/SNAP25 complexes that preferentially form and are known to be inefficient at fusion (10). This may explain why Cpx interactions with t-SNAREs have been difficult to characterize. As a result, the mechanism by which Cpx inhibits vesicle docking and fusion in the absence of calcium and stimulates fusion in its presence remains highly debated, and no consensus regarding the importance of the observed Cpx-SNARE complex interactions has emerged (38, 39).

To better characterize the interaction of Cpx with the prefusion acceptor t-SNARE complex, we took advantage of a new protocol that allows for the purification of monomeric Syx (37) and the assembly of a well-defined and highly active t-SNARE complex consisting of one copy of Syx and one copy of SNAP25 (40). This 1:1 complex rapidly binds Syb and promotes fusion at rates approaching that of an artificially stabilized t-SNARE complex (10). In this work, we use a range of biophysical methods to characterize the detailed molecular interactions of Cpx to this new acceptor complex, and we demonstrate that Cpx strongly associates with the complex while simultaneously interacting with the lipid bilayer at both its N- and C-terminal ends. The membrane interactions at both ends of Cpx are membrane-curvature dependent, confirming and extending previous work obtained in the absence of a t-SNARE complex (33, 41), and both t-SNARE and membrane interactions act cooperatively to promote Cpx association. Importantly, the affinity of Cpx for this complex is higher than for the ternary postfusion core SNARE complex, suggesting that binding to the 1:1 acceptor complex is at least as important in regulating SNARE assembly as binding to the 1:1:1 postfusion complex. We also find that binding of Cpx to the binary t-SNARE complex dramatically reduces the rate of lipid mixing between reconstituted Syb vesicles and t-SNARE vesicles, and that this reduction is the result of a reduced affinity of Syb for the t-SNARE complex. These data lead to a model for the inhibitory action of Cpx on spontaneous fusion events that places this Cpx/t-SNARE complex at the starting point for synchronous Ca²⁺ and synaptotagmin-triggered fusion events.

Materials and Methods

Protein mutagenesis, expression, and purification

Sequences encoding complexin-1 (full-length and fragments comprising residues 26–134 and 26–83), SNAP-25a (full-length wild-type and cysteine-less mutant), synaptobrevin-2 (residues 1–96 and residues 1–116), and syntaxin-1a (residues 183–288 and residues 191–253) were cloned into pET-28a vector. The sequence for the soluble fragment of syntaxin-1a (180–253) was cloned into pET-15b vector. All proteins were derived from Rattus norvegicus. Constructs carried a thrombin-cleavable amino-terminal His₆ tag to facilitate purification. Site-directed mutagenesis based on the PCR reaction with mutation-carrying primers (Integrated DNA Technologies, Coralville, IA) was used to introduce cysteine mutations into complexin-1. The superclamping complexin 1 mutant included mutations D27L, E34F, and R37A (42); the nonclamping mutant included mutations A30E, A31E, L41E, and A44E (32). Plasmids were amplified in TOP10 chemically competent cells (Invitrogen, Carlsbad, CA) and purified using a QIAprep Spin Miniprep Kit (Qiagen, Hilden, Germany) or a GeneJET PCR Purification Kit (Thermo Fisher Scientific, Waltham, MA). All sequences were confirmed by DNA sequencing (GENEWIZ, South Plainfield, NJ). All recombinant proteins were expressed in BL21 (DE3) cells (Agilent Technologies, Santa Clara, CA) and purified as described previously (10, 28, 43, 44).

Membrane-associated syntaxin-1a (183–288) does not form oligomers when purified in dodecylphosphocholine (DPC) and this preparation could be used to form an active 1:1 complex with SNAP-25a (37, 40). An active complex could also be formed using wild-type SNAP-25a where the four native cysteines are dodecylated through an alkylation reaction (40). The work presented here reports results using the dodecylated SNAP-25a complex with syntaxin 1a, but both complexes produced similar results in these measurements.

Assembly and purification of assembled SNARE complexes

SNARE proteins used for assembly of a soluble core SNARE complex were purified by affinity chromatography on a Profinity IMAC Ni-Charged Resin column (Bio-Rad. Hercules, CA). Cysteine-free SNAP-25a, syntaxin-1a (180–253), and synaptobrevin-2 (1–96) were mixed in a 1:1:1 molar ratio. SNAP-25a was first combined with half the amount of Syx and incubated for 1 h at 4°C. Subsequently, Syb was added to the solution, followed by the addition of the remaining amount of Syx. The mixture of all three proteins was placed at 4°C for the night. The complex was purified on a Mono Q 5/50 GL column using an ÄKTApurifier (GE Healthcare, Little Chalfont, UK).

A membrane-reconstituted cis-SNARE complex using syntaxin 1a (183–288) with full-length SNAP-25a and synaptobrevin-2 (1–116) was produced in a similar manner to that described previously (45). In this case, the complex was formed by mixing equal quantities of the three proteins in buffer (20 mM HEPES, pH 7.4, 150 mM NaCl, 1 mM EDTA) in the presence of 0.1% DPC. After incubating overnight at 4°C, mixtures were purified by Mono Q ion-exchange chromatography (GE Healthcare Life Sciences, Marlborough, MA) in the presence of 0.1% DPC. The complex was reconstituted as described below for the liposome SNARE reconstitution, except that the proteins were dialyzed from DPC.

A soluble binary complex (sBC, syntaxin-1a (191–253) and cysteine-less SNAP25) was assembled in DPC in a fashion similar to its membrane-associated counterpart (40). After overnight assembly, this complex was purified by Mono Q ion-exchange chromatography (GE Healthcare Life Sciences) in detergent-free buffers, followed by extensive dialysis against nuclear magnetic resonance (NMR) buffer (10 mM each HEPES, MES, and acetate, pH 6, 150 mM NaCl, 1 mM EDTA) to remove residual amounts of DPC. The complete removal of detergent was confirmed by the lack of signature acyl chain proton resonances in ¹H-NMR.

Preparation of liposomes

1-palmitoyl-2-oleoylphosphatidylcholine (POPC) and 1-palmitoyl-2-oleoyl-phosphatidylserine (POPS) (Avanti Polar Lipids) phospholipids were mixed in a 70:30 molar ratio. Lipid powder was dissolved in chloroform in a round-bottom flask. After the complete lipid dissolution, the solvent was removed under vacuum in a rotary evaporator for 4–5 h and the residual lipid film was thoroughly resuspended in a physiological buffer (139 mM KCl, 12 mM NaCl, 20 mM MOPS, pH = 7.4). The resultant multilamellar vesicles were freeze-thawed five times in liquid nitrogen and extruded repeatedly through either a 100- or 50-nm polycarbonate membrane (Whatman Nucleopore Track-Etched Membranes; Sigma-Aldrich, St. Louis, MO) to produce liposomes. Extrusion was performed on a LiposoFast Basic Extruder (Avestin, Ottawa, ON, Canada). To obtain small unilamellar vesicles of ∼25 nm in diameter, the resuspended lipid film was sonicated (Sonicator W-220F; formerly Heat Systems-Ultrasonics, now QSonica, Newtown, CT) in a pear-shaped flask that was kept on ice during the process. The concentration of vesicles was determined with a phosphate assay based on the modified Bartlett method (46, 47).

Liposome reconstitution of SNARE proteins

Proteoliposomes with lipid composition of 70:30 mol % POPC/POPS were prepared with plasma membrane SNARE proteins. Proteoliposomes with a lipid composition of 97:1.5:1.5 POPC/Rh-DOPE/7-nitrobenzofurazan (NBD)-DOPE were prepared for synaptobrevin-2 and used for lipid mixing and total internal reflection fluorescence (TIRF) binding. SNARE proteins were reconstituted using sodium cholate, as previously described (48, 49). The desired lipids were combined and organic solvents were evaporated under a stream of N₂ gas followed by vacuum desiccation for at least 1 h. The dried-down lipid films were dissolved with 25 mM sodium cholate in buffer (20 mM HEPES, 150 mM KCl, pH 7.4) followed by the addition of an appropriate volume of syntaxin-1a and dSNAP-25a in their respective detergents to reach a final lipid/protein ratio of 100 for EPR samples, 400 for samples used in lipid mixing and synaptobrevin-2 proteoliposomes in TIRF binding, and 3000 for plasma membrane SNAREs for forming planar-supported bilayers. After 1 h of equilibration at room temperature, the mixture was diluted below the critical micelle concentration by adding more buffer to the desired final volume. The sample was then dialyzed overnight against 1 L of buffer with one buffer change after ∼4 h.

Preparation of planar-supported bilayers containing 1:1 Syx (183–288):d-SNAP25 SNARE complexes

Planar-supported bilayers with reconstituted plasma membrane SNAREs were prepared by the Langmuir-Blodgett/vesicle fusion technique, as described in previous studies (50, 51, 52). Quartz slides were cleaned by boiling in Contrad detergent (Decon Labs, King of Prussia, PA) for 10 min, using hot bath-sonication while still in detergent for 20 min, and rinsing thoroughly with deionized water. Immediately before use the slides were cleaned using 3:1 sulfuric acid/hydrogen peroxide and then rinsed thoroughly with Milli-Q water (Millipore, Billerica, MD). The first leaflet of the bilayer was prepared by Langmuir-Blodgett transfer directly onto the quartz slide using a Nima 611 Langmuir-Blodgett trough (Nima, Coventry, UK) by applying the POPC lipid from a chloroform solution. After allowing the solvent to evaporate for 10 min, the monolayer was compressed at a rate of 10 cm²/min to reach a surface pressure of 32 mN/m. After equilibration for 5–10 min, a clean quartz slide was rapidly (200 mm/min) dipped into the trough and slowly (5 mm/min) withdrawn, while a computer maintained a constant surface pressure and monitored the transfer of lipids with headgroups down onto the hydrophilic substrate. Proteoliposomes containing binary SNARE complex (77 mM total lipid in 1.3 mL) were added and incubated at room temperature for 2 h to introduce the protein complex and form the second leaflet of the supported bilayer. Excess unfused proteoliposomes were then removed by perfusion with 5 mL of reaction buffer (20 mM HEPES, 150 mM KCl) containing 100 μM EDTA.

Protein spin labeling and continuous wave-EPR measurements

All complexin-1 variants were labeled by addition of 10-fold molar excess of the thiol-specific spin label, (1-oxy-2,2,5,5-tetramethylpyrrolinyl-3-methyl)methanethiosulfonate and incubated overnight at 4°C. Excessive spin label was removed by desalting in physiological buffer (139 mM KCl, 12 mM NaCl, 20 mM MOPS, pH = 7.4) using the HiPrep 26/10 column (GE Healthcare Life Sciences). For sample preparation, protein or lipid aliquots were mixed in the specified molar ratios. Continuous wave (cw)-EPR measurements were performed on 6 μL of sample loaded with Hamilton syringe into the 0.60 × 0.84 mm (inner diameter × outer diameter) borosilicate glass capillaries (Vitrocom, Mountain Lakes, NJ). Spectra were recorded on an EMX X-band EPR Spectrometer (Bruker, Billerica, MA) at 2 mW incident microwave power with modulation amplitude of 1 G and frequency of 100 kHz. The magnetic field was swept through 100 G, and up to 30 scans were performed to increase the signal/noise ratio. Spectra were then processed using LabVIEW programs provided by Christian Altenbach (University of California, Los Angeles, CA), normalized, and plotted in OriginPro 7.5 (OriginLab, Northampton, MA).

Binding affinities determined by EPR spectroscopy

The binding affinity of spin-labeled Cpx mutants to membranes or SNARE-containing proteoliposomes was determined by EPR spectroscopy essentially as described previously (53). In this case, the normalized amplitudes of the EPR spectra, rather than the absolute amplitudes, were used to generate a plot of the fraction of membrane or SNARE-bound Cpx (fb) as a function of accessible lipid concentration. The data were fit to the following expression:

$$fb=K\left[\text{L}\right]/\left(1+K\left[\text{L}\right]\right),$$

where [L] is the accessible lipid concentration, and K is the reciprocal molar partition coefficient.

Power saturation EPR

Power saturation EPR experiments were conducted at room temperature on an EMX X-band EPR Spectrometer (Bruker). A quantity of 75 μM complexin-1 was incubated with sonicated PCPS small unilamellar vesicles and the sample was loaded into a gas-permeable TPX capillary. The microwave power was varied from 0.25 to 100 mW and 30 scans of the central peak were averaged for each power step. The applied magnetic field was swept through 10 G with a modulation amplitude of 1 G and frequency of 100 kHz. The power saturation was conducted on spin-labeled complexin-1 in the presence of air (20% O₂), N₂, or N₂ with 10 mM NiEDDA (53). In each of these conditions and for every power step, the amplitude of the central peak was measured and the P_1/2 value was extracted using a LabVIEW program provided by Christian Altenbach (University of California, Los Angeles). The values of ΔP_1/2 (O₂) or ΔP_1/2 (NiEDDA) were then determined from the difference in P_1/2 values in the presence and absence of either O₂ or NiEDDA, respectively. A depth parameter, Φ, related to the local concentrations of O₂ and NiEDDA, which vary as a function of depth in the lipid bilayer (54), was determined from the values of ΔP_1/2. Based on a defined calibration curve, the obtained values of depth parameter were applied in the hyperbolic tangent function that describes behavior of Φ relative to distance from the membrane surface (55).

Ensemble lipid mixing assay

Proteoliposomes were prepared using lipid/protein ratios of 400 for both plasma membrane t-SNARE and synaptobrevin-2 proteoliposomes. Experiments were performed at 37°C using a FluoroMax-3 Spectrofluorometer from Horiba Scientific (Kyoto, Japan) to measure NBD dequenching with an excitation wavelength of 460 nm, emission wavelength of 532 nm, and time resolution of 2 s. Proteoliposomes were mixed at a 1:1 ratio using a concentration of 20 μM lipid for both SNARE proteoliposome samples. After each subsequent fusion reaction, 0.1% Triton-X was added to observe the total amount of fluorescence, which was used to correct differences between the total amounts of probes from different proteoliposome reconstitutions. Curves were normalized to the lipid mixing condition without complexin-1.

Fluorescence anisotropy measurements

Native cysteine C105 of complexin-1 was fluorophore-labeled with Alexa546. Experiments were performed on a FluoroMax-3 Spectrofluorometer (Horiba Scientific). Fluorescence anisotropy of Cpx was measured in a 1.2 mL volume containing 50 nM Cpx before and after addition of proteoliposomes with reconstituted proteins, as indicated in text at lipid/protein of 400. Protein-free vesicles were used as a control sample, where the lipid concentration is the same as with the proteoliposomes added.

NMR spectroscopy

TROSY-versions of three-dimensional (3D) backbone experiments (56) [HNCA, HN(CA)CB, HNCO, HN(CA)CO] were performed on ²H,¹³C,¹⁵N-labeled full-length Cpx in buffer (10 mM each HEPES, MES, and Acetate, pH 6, 150 mM NaCl, 1 mM EDTA) or in DPC (the same buffer with the addition of 150 mM DPC), and ¹³C,¹⁵N-labeled Cpx (26–83) in buffer (10 mM each HEPES, MES, and acetate, pH 6, 150 mM NaCl, 1 mM EDTA). Data were collected on an Avance 800 MHz Spectrometer (Bruker) at 25°C. ¹⁵N-labeled Cpx samples were employed in the study with nanodisc and vesicle interactions. ¹⁵N-labeled Cpx (26–83) samples were employed in interaction studies with the soluble binary complex. Protein-free nanodiscs were assembled with POPC and three different membrane scaffold proteins (MSP1D1, MSP1D1-ΔH5, and MSP1D1-ΔH4ΔH5), and subsequently purified according to published protocols (57). Protein-free nanodiscs were added to CpxI samples in NMR buffer to make final NMR samples with protein/MSP/lipid ratios of 1:2:120. All three different MSP samples resulted in very similar NMR spectra, and the data from the MSP1D1 nanodisc sample were used in Fig. 3 a. Small POPC/POPG (90:10) vesicles were formed by freeze-thaw-sonicate cycles, and added to a Cpx NMR sample in buffer to make a final sample with a protein/lipid ratio of 1:50. These two-dimensional spectra were collected on either the 800- or a 600-MHz Avance Spectrometer (Bruker). All spectra were processed and analyzed with NMRPipe (58) and SPARKY (59).

Figure 3.

Complexin binds to membranes at both its N- and C-terminal regions. (a) Shown here are NMR intensity ratios of Cpx bound to POPC nanodiscs (cyan bars) or POPC/POPG (90:10) bilayers (red bars). (b) Shown here are EPR spectra in the absence (black trace) and presence (red trace) of POPC/POPS (70:30) liposomes and (c) intensity ratios for the high field resonance of spin-labeled Cpx mutants with/without POPC/POPS (70:30) liposomes. Error bars represent the uncertainty in the normalized intensity ratio due to normalization and measurement of peak-to-peak amplitudes. The data indicate the N- and C termini of Cpx are involved in direct membrane binding because the largest intensity drops in both the NMR data (a) and EPR data (b) are seen for the N- and C-terminal regions. To see this figure in color, go online.

TIRF microscopy

All experiments were carried out on an Axiovert 35 Fluorescence Microscope (Carl Zeiss, Thornwood, NY), equipped with a 40× water immersion objective (N.A. = 0.95; Carl Zeiss) and prism-based TIRF illumination. The light source was an OBIS 532 LS Laser (Coherent, Santa Clara, CA). Fluorescence was observed through a 610-nm band-pass filter (D610/60; Chroma Technology, Brattleboro, VT) by an EMCCD camera (DU-860E; Andor Technology, Belfast, Northern Ireland). The EMCCD camera was cooled to −70°C, and the gain was typically set to an electron gain factor of ∼100. The prism-quartz interface was lubricated with glycerol to allow easy translocation of the sample cell on the microscope stage. The beam was totally internally reflected at an angle of 72° from the surface normal, resulting in an evanescent wave that decays exponentially with a characteristic penetration depth of ∼100 nm. An elliptical area of 250 × 65 μm was illuminated. The laser intensity, shutter, and camera were controlled by a homemade program written in LabVIEW (National Instruments, Austin, TX).

TIRF complexin-1 binding assay

Alexa647-labeled complexin-1 (at native C105) diluted with reaction buffer to indicated concentration was flowed onto the planar-supported bilayer in increasing concentrations. Binding was measured by increases in fluorescence in the TIRF field by taking images every 15 s. After binding was saturated, a higher concentration of complexin-1 was added and saturation monitored until saturation occurred again from 0.001 up to 10 μM complexin-1.

TIRF synaptobrevin-2 proteoliposome docking

Synaptobrevin-2 proteoliposomes with a lipid composition of 99:1 POPC/Rh-DOPE were diluted to a final concentration of 5 μM lipid in 3 mL of reaction buffer. Before Syb-proteoliposomes were injected into the chamber, the planar-supported bilayer with reconstituted t-SNARE complexes was preincubated for 15 min with the indicated concentration of complexin-1. Images were taken every 30 s, and the total amount of fluorescence was used to determine the number of proteoliposomes bound per μm² at each time point.

Results

Complexin binds to the soluble SNARE core complex

Site-directed spin labeling is well suited to examine the molecular interactions of Cpx to both soluble and membrane-associated SNAREs, and the spin-labeled side chains of R1 (Fig. 1 a) were introduced one at a time into full-length Cpx by the reaction of the thiol-specific 1-oxyl-2,2,5,5-tetramethylpyrroline-3-methyl spin label to single site-directed cysteines.

As a starting point, we first probed the well-known interaction of Cpx to an assembled four-helical soluble SNARE core complex (1:1:1 syntaxin-1 (180–253):SNAP25:synaptobervin-2 (1–96)) by EPR spectroscopy using 13 different spin-labeled sites on Cpx (Fig. 1 b). The addition of the soluble SNARE core complex resulted in a decrease in normalized amplitude of the Cpx spectra, where positions in or near the central helix yielded the most dramatic lineshape changes (e.g., E47R1, K54R1, and D68R1). Fig. 1 c plots these changes as a ratio of the normalized EPR intensity in the presence and absence of the four-helix SNARE complex. Changes in EPR lineshape may occur due to a reduction in the overall tumbling rate of Cpx resulting from SNARE association; however, a comparison of the spectra in Fig. 1 b with those for Cpx in sucrose (see Fig. S1) indicates that slowing the tumbling rate alone cannot account for the lineshape changes at some sites. At several sites (E47R1, K54R1, and D68R1), the changes in normalized intensity reflect a reduction in local backbone dynamics that is due to the induction of stable secondary structure and direct interactions with SNAREs. There is no indication in these spectra of tertiary contact of the label with the SNARE complex, consistent with the sites being outward facing in the Cpx/SNARE complex structure (PDB: 1KIL (31)).

We also tested the binding of these Cpx mutants to individual soluble SNAREs lacking their transmembrane segments. Upon the addition of spin-labeled Cpx to any of the three soluble individual SNAREs—Syb (1–96), Syx (180–253), or SNAP25—no change in lineshape was detected (see Fig. S2), demonstrating that at the concentrations of Cpx and SNAREs used here, no significant interaction takes place, which is consistent with previous findings (34).

The Cpx/SNARE complex used for EPR (Fig. 1) should resemble the previously published structure (PDB: 1KIL (31)), and a comparison of the spectra in Fig. 1 b with this structure indicates that the resolved helical region of Cpx in the crystal structure (sites 32–72) varies significantly in its backbone dynamics. Lineshapes from sites E36R1 and E39R1 are consistent with helical structures (60, 61), but indicate that the N-terminal end of this accessory helix is highly dynamic and is either fraying or undergoing significant rocking motions. This is consistent with the published structure (31), where the resolved N-terminal end of Cpx (residues 32–41) is not making efficient contact with the SNAREs. Near the C-terminal side of the central helix, D68R1 is also more dynamic than labels from the central region of this helix. These data indicate that sites from the C-terminal side of the accessory helix and N-terminal region of the central helix are largely responsible for the binding of Cpx to the SNARE complex. EPR spectra from the N-terminal (E2R1, T13R1, and K18R1) and C-terminal (R78R1, Q83R1, E108R1, and T119R1) ends of full-length Cpx indicate that these segments remain unstructured upon interaction with the soluble four-helical core SNARE complex.

Complexin binds to a soluble binary acceptor t-SNARE complex

To determine whether Cpx binds to a soluble binary acceptor t-SNARE complex, we used a fragment of syntaxin (residues 191–253) lacking the transmembrane domain. This was purified and subsequently assembled with SNAP25 to form a binary complex as described in Kreutzberger et al. (40) and further purified as described in Materials and Methods. We incubated this soluble binary complex with a Cpx construct comprising the accessory and central helices, i.e., residues 26–83, and measured their interaction by NMR. All resonances of free Cpx could be assigned by heteronuclear experiments (see Materials and Methods) and yielded chemical shifts that were consistent with previous assignments (34) (Fig. 2 a). After mixing with the sBC in a 1:1 ratio, many resonances of Cpx became too broad to be observed (Fig. 2, b and c), particularly those toward the C-terminal end of the central helix. This indicates that these residues bind to a rather large complex, possibly undergoing intermediate exchange on an approximately millisecond timescale. Significant signal attenuation was also observed for residues on the N-terminal side of central helix, whereas residues in the accessory helix were progressively less attenuated from the C-terminal toward the N-terminal end, indicating a weaker interaction with the binary SNARE complex. In contrast, the extreme ends of the construct exhibit no signal attenuation, and are not likely to be interacting with the binary complex. When we changed the ratio of Cpx (26–83):sBC from 1:1 to 2:1, 5:1, and 10:1, the respective attenuations decreased, but retained the same pattern with the intensities of the central helix residues, showing higher degrees of attenuation compared to those of the accessory helix residues. The fact that significant attenuation persists even at these high molar ratios supports the notion of intermediate molecular exchange on the NMR timescale. The data also support the notion that the binding of Cpx to the binary complex is initiated from the C-terminal half of the central helix and gradually extends toward both ends of Cpx. Because we have not performed more detailed NMR exchange experiments on this system, we cannot distinguish between thermodynamic and kinetic factors that might both contribute to the observed gradual binding pattern.

Figure 2.

Complexin binds to soluble binary t-SNARE complex. (a) Shown here is the ¹⁵N-¹H HSQC spectrum of Cpx (26–83) in buffer. Assignments of backbone amides are denoted by one-letter amino acid abbreviations followed by their sequence numbers. (b) Shown here is the ¹⁵N-¹H HSQC spectrum of Cpx (26–83) mixed with an equimolar concentration of sBC, consisting of Syx(191–253) and SNAP25 (green), overlayed onto the HSQC spectrum of free Cpx (26–83) (red). (c) Shown here are intensity ratios of backbone amide resonances versus residue numbers at different molar ratios of Cpx (26–83)/sBC, with ratios of 10:1, 5:1, and 1:1 represented in red, blue, and green bars, respectively. The intensity ratios are the intensities of the Cpx (26–83)/sBC samples over the intensities of Cpx (26–83) alone and were calibrated using N- and C-terminal residues to account for differences in sample concentrations. Only peaks with S/N higher than 20 were analyzed, and error bars were propagated from S/N levels of spectra and calibrated with redundant measurements. To see this figure in color, go online.

Complexin binds to membranes via both its N- and C-terminal domains and this binding is highly dependent on membrane curvature

Complexin-1 has been reported to interact with lipid bilayers in its C-terminal region at two specific sites: a C-terminal motif, and a nearby amphipathic helix (41, 62). Binding of the N-terminal domain to membranes has also been reported (33). Here, we used both NMR and cw-EPR to examine these membrane interactions of full-length Cpx and determine whether any structural changes occurred within Cpx upon membrane binding.

Numerous backbone amide peaks undergo changes in chemical shift when phospholipid-mimicking DPC micelles are added to Cpx in solution. Based on resonance assignments obtained from respective NMR 3D backbone experiments, residues in the N- and C-terminal regions display large chemical shift differences, a few residues at the C-terminal end of central helix show moderate changes, and residues assigned to the accessory and central helix show almost no change in chemical shift (Fig. S3). To observe changes in a more lipidlike environment, we have also measured HSQC spectra of Cpx in the presence of nanodiscs and liposomes. Shown in Fig. 3 a are the intensity ratios from HSQC spectra for full-length ¹⁵N-labeled Cpx in the presence and absence of vesicles or nanodiscs. Upon lipid binding, there is an increase in the rotational correlation time for segments directly interacting with the bilayer, and NMR lineshapes in these regions broaden due to reduced averaging of chemical shift anisotropy and dipolar interactions. As seen in Fig. 3 a, the reductions in NMR peak intensity are not uniform, but occur primarily for residues within either the N- and C-terminal domains of Cpx, whereas the more centrally localized sites show small or no levels of attenuation. The first 12 and the last 21 residues show the largest decreases, indicating that these sites are directly involved in simultaneous membrane binding.

To further test this result, EPR spectra (Fig. 3 b) were recorded for 10 spin-labeled mutants covering the Cpx sequence in the presence and absence of POPC/POPS (70:30 mol %) small unilamellar vesicles (SUVs). The normalized intensity ratios for the high-field resonance peaks of Cpx in the absence/presence of liposomes are shown in Fig. 3 c. Upon membrane association, the EPR lineshapes broaden, due to a combination of changes in backbone dynamics and a change in the selection of R1 rotamers that occurs upon insertion of the spin label into the membrane (63, 64). The data indicate that both the N- and C-terminal regions of Cpx contact the membrane interface, consistent with the NMR result. CpxE2R1, T13R1, K18R1, S115R1, and T119R1 yield highly averaged EPR lineshapes in the absence of membrane, but upon addition of SUVs, all of these sites exhibit a significant increase in spectral linewidth and a drop in signal amplitude. Smaller changes are observed for CpxD68R1 and perhaps CpxK54R1, indicating that they may also interact with the lipid interface. Other sites in Cpx yield no or small changes in EPR lineshapes in the presence of lipid bilayers.

To determine the position of Cpx relative to the membrane interface, we power-saturated EPR spectra from sites E2R1, D68R1, and T119R1 and obtained membrane-depth parameters (see Materials and Methods). The membrane-depth parameters (Φ) and position of the spin label relative to the lipid phosphate plane are shown in Table 1. Values obtained for all three Cpx mutants (E2R1, D68R1, and T119R1) place the labels at a depth of 1.6–2.2 Å from the phosphate plane. The labels do not reach the acyl chain region of the bilayer, and the data demonstrate that Cpx is peripherally associated with the membrane interface.

Table 1.

Depth Parameters, Φ, and Distances from the Phospholipid Phosphate Plane Determined for R1-labeled Cpx by Power Saturation of the EPR Spectrum

Position	Φ	Distance (Å)
E2R1	−1.1 ± 0.2	1.7 ± 0.9
D68R1	−1.1 ± 0.1	1.6 ± 0.5
T119R1	−0.96 ± 0.14	2.2 ± 0.6

Positive values of the membrane depth (Å) indicate the nitroxide R1 side chain is positioned on the hydrocarbon side of the phosphate plane within the lipid bilayer. The errors in the depth are calculated based on standard deviation of Φ.

It has been previously reported that the binding of the C terminus of Cpx to lipid vesicles depends on membrane curvature (33, 41). To compare the curvature-dependent binding at both ends of full-length Cpx, we determined the equilibrium binding affinities of Cpx to liposomes of different sizes, ≈100, ≈50, and ≈25 nm in diameter (see Fig. S4 for size distribution histograms) by EPR spectroscopy using both CpxE2R1 and T119R1. As seen in Fig. 4, b and c, labels from opposite ends of Cpx show a strong and similar dependence upon curvature, exhibiting significantly stronger binding to the smallest and most curved vesicles. The C-terminal site, Cpx119R1 (Fig. 4 c; Table 2), displays an 80-fold difference in affinity between the largest and smallest vesicles, corresponding to a free energy difference of ∼2.6 kcal/mol. The N-terminus shows a similar affinity to the smallest vesicles and a 40-fold weaker affinity toward the largest vesicles (Fig. 4 b; Table 2). The result for the C terminus is generally consistent with previous findings (33, 41) and the data for the N-terminus indicates that it has a similar curvature-dependent binding. We also tested to see if the N-terminal and C-terminal interactions of Cpx were cooperative, by measuring the affinity at site T119R1 when the N-terminus was truncated (Cpx (26–134)). As seen in Fig. 4 d, when the N-terminus is truncated the membrane affinity of the C-terminal end is not significantly changed. Thus, both ends of Cpx bind to membranes in a curvature-dependent manner and act independently of each other. This may be a result of long disordered segments being present between the N- and C-terminal regions of Cpx.

Figure 4.

Complexin senses membrane curvature. (a) EPR spectra of E2R1 and T119R1 show dramatic changes in lineshape upon membrane binding, and these changes were used to determine Cpx membrane affinity. Titrations of (b) E2R1 and (c) T119R1 with unilamellar vesicles of varied size (≈100, ≈50, and ≈25 nm in diameter) yield the fraction of membrane bound Cpx as a function of accessible lipid. These data were fit (solid lines) to a simple binding isotherm to yield the membrane binding partition coefficient shown in Table 2 (see Materials and Methods). In (d), the binding of Cpx119R1 in full-length Cpx was compared with that for an N-terminally truncated version Cpx (26–134) to small 25-nm-diameter vesicles, and indicate that the N- and C-terminal regions act independently. All data points are averages of three titrations. Error bars represent deviations from the best fit. To see this figure in color, go online.

Table 2.

Partition Coefficients and Binding Affinities for the Binding of CpxE2R1 and CpxT119R1 to Vesicles of Different Size

	E2R1 Partition Coefficient [M−1]	E2R1 Lipid Binding Affinity [mM]	T119R1 Partition Coefficient [M−1]	T119R1 Lipid Binding Affinity [mM]
92 ± 23 nm LUVs	64 ± 6	16 ± 1	30 ± 2	34 ± 2
58 ± 13 nm SUVs	260 ± 31	3.9 ± 0.4	72 ± 15	14 ± 2.5
26 ± 8 nm SUVs	2400 ± 180	0.42 ± 0.03	2300 ± 210	0.43 ± 0.04

Phase partition coefficients and the respective binding affinities were determined for complexin-1 incubated with liposomes. The lipid binding affinities are equal to K⁻¹ in milliMolar, where K is the reciprocal molar partition coefficient. The lipid binding affinity, K⁻¹, represents the concentration of accessible lipid (in milliMolar) that produces half-maximal membrane binding of Cpx.

A binary t-SNARE complex promotes simultaneous membrane and t-SNARE binding of complexin

To determine whether Cpx is capable of simultaneous interactions with both membranes and membrane-embedded SNAREs, spin-labeled full-length Cpx was added to membrane-reconstituted Syx (183–288), a dodecylated version of SNAP-25a (dSNAP25), or a reconstituted 1:1 Syx:dSNAP25 t-SNARE complex. The dSNAP25 was produced by chemical alkylation of the four native cysteines closely resembling the palmitoylated form of the protein, which is predominant in neurons or SNAP25-expressing insect cells (65, 66). This 1:1 syx:dSNAP25 acceptor t-SNARE complex is functionally similar to an artificial 1:1:1 Syx (183–288)/SNAP25:Syb (49–96) acceptor t-SNARE complex (ΔN complex) that produces fast and efficient fusion events (40), but it does not require the short Syb peptide needed to stabilize Syx and SNAP25 in a 1:1 state (10, 48).

Shown in Fig. 5 are EPR spectra from six R1-labeled sites within Cpx that interact with SNAREs or membranes (Figs. 1, 2, and 3). In Fig. 5 a, 30 μM Cpx is added to protein-free liposomes prepared by dialysis at a lipid concentration of 4 mM. For these larger liposomes, 4 mM is not a lipid concentration high enough to bind a significant fraction of Cpx. As expected, no significant interaction is detected at any Cpx site. In Fig. 5 b, these same Cpx mutants are added to the binary t-SNARE complex (Syx/dSNAP25) reconstituted into membranes at this same lipid concentration (4 mM) and a protein concentration of 40 μM. In contrast to Fig. 5 a, significant lineshape broadening is seen both at the N- and C termini (positions 2, 115, and 119), which interact with membranes, and within the central helical region (positions 47, 54, and 68), which interact with the core SNARE complex (Fig. 1) (31, 32) or the soluble 1:1 t-SNARE complex (Fig. 2). The relative normalized EPR intensity changes seen upon the addition of pure lipid membranes or the membrane-reconstituted acceptor t-SNARE complex are shown in Fig. 5 c. The observed lineshape changes at the N- and C termini seen in Fig. 5 b resemble those that take place at sites on membranes (Fig. 3). The lineshapes for sites near the central helix of Cpx (sites 46, 54, and 68) resemble those seen with the SNARE core complex (Fig. 1) and are consistent with the R1 label at exposed helical sites (60). No significant changes in lineshape occur when labeled Cpx is added to membranes containing only reconstituted Syx; however, lineshape changes are observed when Cpx is added to membranes containing only dSNAP25 at the same protein/lipid ratio as in the experiments with the binary t-SNARE complex. These changes yield a different pattern of contact than that seen with the binary t-SNARE complex, and were not further investigated (see Fig. S5).

Figure 5.

Membranes with the reconstituted acceptor t-SNARE complex enhance the membrane binding of Cpx. Shown here are EPR spectra from different spin-labeled Cpx mutants in solution (black traces) or incubated with either (a) liposomes at 4 mM lipid (70:30 POPC/POPS) (cyan trace) or (b) proteoliposomes with 4 mM lipid and t-SNARE complex (70:30 POPC/POPS) at a protein/lipid ratio of (1:100) (red trace). (c) Shown here are normalized intensity ratios for spin-labeled Cpx in the presence/absence of liposomes (4 mM lipid) (cyan bars) or in the presence/absence of proteoliposomes with t-SNAREs (red bars). The ratios are determined from normalized intensities measured from the high-field nitroxide resonance line (M_I = −1). The error range represents the uncertainty in the determination of the ratio. Protein-free liposomes were prepared by dialysis in an identical manner to that for the proteoliposomes. To see this figure in color, go online.

Fig. 5 demonstrates that the presence of the binary acceptor t-SNARE complex promotes Cpx binding under conditions where lipid interactions alone are insufficient to bind Cpx. As a result, simultaneous membrane-SNARE interactions are contributing to the free energy of Cpx binding, and the EPR spectra data provide strong evidence that Cpx is simultaneously contacting both membranes and the 1:1 t-SNARE complex.

To estimate the contribution that the SNAREs make to the membrane binding of Cpx, we used site T119R1 to measure Cpx membrane affinity by titrating increasing lipid concentrations into a sample having a fixed concentration of labeled Cpx. As seen in Fig. 6 a and Table 3, the affinity of Cpx to membranes containing Syx is not significantly different from lipid alone; however, the presence of the 1:1 acceptor t-SNARE complex dramatically increases the affinity of Cpx. The partition coefficient is increased by 19-fold, indicating that the interaction of Cpx with the membrane-associated t-SNARE complex increases the free energy of membrane association by ∼1.8 kcal/mole. Cpx also binds the assembled membrane-reconstituted 1:1:1 SNARE core complex, but with a threefold weaker affinity than the membrane-associated binary t-SNARE complex.

Figure 6.

Complexin strongly binds to proteoliposomes with reconstituted acceptor t-SNARE complexes. (a) Shown here are binding curves for Cpx, using the spin-labeled mutant CpxT119R1, produced by recording EPR spectra with increasing concentrations of proteoliposomes that were either protein-free (black trace) or reconstituted with Syx (red trace), dSNAP25 (blue trace), the Syx:dSNAP25 t-SNARE complex (green trace), or the assembled cis-SNARE complex (orange trace). All proteins were present at a protein/lipid ratio of 1:100. Points are averages of three titrations, and errors represent deviations from the best fit. (b) Fluorescence anisotropy measurements are consistent with EPR, where fluorophore-labeled Cpx was mixed with either protein-free liposomes, dSNAP25, Syx, or t-SNARE-containing proteoliposomes composed of 70:30 POPC/POPS at a lipid/protein ratio of 400:1. The effect of the interaction was quantified and plotted as a change in fluorescence anisotropy relative to free Cpx. Anisotropy changes are the average of four experiments, and the error bars represent standard errors. (c) The binding of Cpx to planar-supported bilayers was measured with TIRF microscopy. Increasing concentrations of labeled Cpx were titrated into the planar bilayer, and increases in fluorescence in the TIRF field were monitored. The fluorescence as a function of Cpx concentration is shown for protein-free planar bilayers (70:30 POPC/POPS) (red trace) and for planar bilayers (70:30 POPC/POPS) reconstituted with syntaxin-1/dSNAP25 (lipid/protein of 3000) (black traces). For each condition, values are averages from three separate bilayer preparations. Error bars are standard errors. To see this figure in color, go online.

Table 3.

Partition Coefficients and Binding Affinities for the Binding of Cpx to Membranes in the Presence and Absence of Reconstituted SNAREs

CpxT119R1 with:	Partition Coefficient M−1]	Lipid Binding Affinity [mM]
Liposomesa (POPC/POPS)	330 ± 30	3.0 ± 0.3
Syx proteoliposomes	290 ± 30	3.5 ± 0.3
dSNAP25 proteoliposomes	1460 ± 190	0.69 ± 0.08
Syx/dSNAP25 proteoliposomes	6300 ± 900	0.16 ± 0.02
Syx/SNAP25/Syb2 proteoliposomes	2100 ± 240	0.50 ± 0.06

Phase partition coefficients and the respective binding affinities were determined for CpxT119R1 incubated with the SNARE proteoliposomes. The lipid binding affinities are equal to K⁻¹ in milliMolar, where K is the reciprocal molar partition coefficient (see Materials and Methods). The lipid binding affinity, K⁻¹, represents the concentration of accessible lipid (in milliMolar) that produces half-maximal membrane and SNARE binding of Cpx. The protein/lipid ratio of SNARE containing proteoliposomes is 1:100.

^aThese protein-free liposomes are prepared by dialysis from cholate, as are the proteoliposomes; as a result, the binding affinities of CpxT119R1 are not directly comparable to those for the extruded liposomes in Table 2.

To confirm these results using a different approach, we measured the fluorescence anisotropy of Alexa 546-labeled Cpx in the presence of liposomes without or with Syx, dSNAP25, or the Syx/dSNAP25 acceptor t-SNARE complex. As seen in Fig. 6 b, there is little anisotropy change in the presence of the protein-free liposomes, slight increases in the presence of Syx, larger changes with dSNAP25 and the membrane-reconstituted SNARE core complex, and the largest changes with the acceptor t-SNARE complex. This roughly follows the order of membrane affinities measured by EPR. It is also important to note that there is a lower affinity of Cpx to the artificial ΔN(Syb) acceptor complex (10), and almost no binding to a Syx/SNAP25 complex assembled from CHAPS. The lower affinity to the ΔN complex suggests that Syb (49–96) competes with the Cpx binding site on the t-SNAREs, and the complex assembled in CHAPS has a 2:1 Syx/SNAP25 stoichiometry (40).

We also employed TIRF microscopy to estimate the binding of Cpx to planar-supported bilayers in the presence and absence of SNAREs, where t-SNARE complexes were incorporated into supported bilayers using a procedure described previously (see Materials and Methods). Fig. 6 c shows the binding of fluorescently labeled Cpx as a function of concentration to membranes with and without t-SNAREs. Complexin-1 displays only a weak interaction with the protein-free membrane, whereas the presence of the binary acceptor t-SNARE complex causes a dramatic increase in this interaction. As seen in Fig. 6 c, ∼2 μM Cpx is sufficient to saturate the planar-supported bilayer as higher Cpx concentrations do not significantly increase the measured fluorescence. This result is also consistent with the EPR result, showing the SNARE and membrane interactions act cooperatively to promote Cpx binding.

Acceptor t-SNARE complex and membrane-bound complexin inhibits membrane fusion by blocking synaptobrevin insertion into the complex and SNARE-dependent vesicle docking

As reported previously, the reconstituted Syx:dSNAP25 acceptor t-SNARE complex used here promotes fast membrane fusion with synaptobrevin-containing vesicles (40). Here, we tested the effect of Cpx on fusion in this system. We first employed a lipid mixing assay for fusion where vesicles containing Syb or the binary t-SNARE complex are mixed at a 1:1 ratio in the presence of Cpx. When lipid mixing occurs, rhodamine and NBD in the v-SNARE membrane are diluted and the NBD fluorescence increases. As seen in Fig. 7 a, Cpx has a potent inhibitory effect on lipid mixing in the absence of Ca²⁺. With no Cpx added, proteoliposomes fuse efficiently, whereas introduction of 8 μM Cpx blocks fusion almost entirely (Fig. 7 b), reaching only 5% of the lipid mixing observed in the absence of Cpx. These data demonstrate that Cpx has an inhibitory effect on lipid mixing in the absence of Ca²⁺ in this system.

Figure 7.

Complexin inhibits membrane fusion by reducing synaptobrevin affinity for t-SNAREs. (a) Shown here is lipid mixing between Syb and binary acceptor t-SNARE complex proteoliposomes (t-SNARE membrane composed of 70:30 POPC/POPS with a lipid/protein ratio of 400; v-SNARE membrane composed of 97:1.5:1.5 POPC/Rh-DOPE/NBD-DOPE, lipid/protein ratio of 400). Observed fluorescence is a function of time, and presented as normalized signal intensity. After each fusion reaction, 0.1% Triton X-100 was added to determine total fluorescence and normalize the data between different preparations. Traces correspond to lipid mixing in the absence of Cpx (black) or in the presence of 1 (red), 2 (blue), 4 (violet), or 8 μM Cpx (green trace), as indicated. All traces are normalized to 0 μM complexin. (b) Saturation of the lipid mixing taken at 500 s as a function of the concentration of added Cpx. The inhibitory effect of Cpx is concentration dependent, and at 8 μM, Cpx has reduced fusion by ∼20-fold. Error bars are standard errors from four measurements. (c) Shown here is docking of synaptobrevin proteoliposomes to SNARE acceptor complexes reconstituted into planar-supported bilayers (Syb proteoliposomes composed of 99:1 POPC/Rh-DOPE) as a function of time. The docking was measured in the absence of Cpx (black) and in the presence of 0.5 μM (red), 1 μM (cyan), and 2 μM Cpx (green), as indicated. (d) The number of bound Syb proteoliposomes at 1200 s as a function of concentration of Cpx. Values are averages from measurements on three bilayers, and errors are standard errors. (e) Shown here is the binding of a fluorescently labeled soluble Syb (1–96) peptide after a 20 min incubation to a planar-supported bilayer (70:30 POPC/POPS) containing the t-SNARE complex (SyxH3/dSNAP25) at a lipid/protein ratio of 3000. The binding of Syb was measured without Cpx (black), and at Cpx concentrations of 0.5 μM (red), 1 μM (blue), and 2 μM (green). Values are averages of three measurements and errors are standard errors. (f) The K_d for Syb binding calculated from the data in (e) is shown as a function of Cpx concentration for wild-type Cpx (black trace) as well as the superclamping (SC, blue trace) and nonclamping mutants (NC, purple trace). Error bars represent error in the fit. The mutations of SC are D27L, E34F, and R37A (42), and the mutations of NC are A30E, A31E, L41E, and A44E (32). To see this figure in color, go online.

The fusion rates observed in this lipid mixing assay may be due to the effects of Cpx upon membrane docking or upon fusion itself. To distinguish between docking and fusion, we added Cpx to planar-supported bilayers containing reconstituted 1:1 acceptor t-SNARE complexes without the addition of Ca²⁺. Fluorophore-labeled Syb proteoliposomes were added and the number of bound vesicles per μm² area of the supported bilayer was determined at each time point using TIRF microscopy. Fig. 7 c shows the cumulated docking events that occur as a function of time. Syb proteoliposomes bind to Syx/SNAP25 acceptor t-SNARE complexes reconstituted into the planar bilayers reaching a density of ∼6 vesicles/μm² in a period of 1200 s. As seen in Fig. 7, c and d, Cpx inhibits vesicle docking, where the inhibition saturates at ∼2 μM Cpx. It is important to note that although Cpx inhibits docking, it does not completely prevent docking even at a concentration of 2 μM. These data indicate that the reduction in docked v-SNARE vesicles in the presence of Cpx is due to a reduction in affinity of Syb to the acceptor t-SNARE complex. To directly test this result, we measured the binding of a soluble fragment of Syb (residues 1–96, with residue 55 mutated to a Cys and labeled with Alexa546) to the supported bilayers containing acceptor t-SNARE complexes. As seen in Fig. 7 e, the level of bound Syb increases as the concentration of labeled Syb is increased, but increasing concentrations of Cpx inhibit this binding. These data were fit to a simple 1:1 binding isotherm and the results are plotted in Fig. 7 f. As seen in Fig. 7 f, at the highest concentration used (2 μM) the effect of Cpx has saturated and the affinity of Syb for the reconstituted t-SNARE complex has been reduced by a factor of >10.

To determine whether this effect of Cpx on the association of Syb and the binary t-SNARE complex correlates with physiological results, we tested the binding of Syb using this approach in the presence of both superclamping and nonclamping mutants of Cpx. These mutants either slightly depress (the superclamping) or dramatically increase (the nonclamping) spontaneous fusion events in cultured cortical neurons (67). As seen in Fig. 7 f, the superclamping mutant produces a response close to wild-type Cpx, and may be just slightly more effective at reducing the affinity of Syb to the acceptor t-SNARE complex. In contrast, the nonclamping mutant is much less effective at reducing the affinity of Syb to this complex. These results correlate well with the physiological data and indicate that Cpx acts by reducing the affinity of Syb for the acceptor t-SNARE complex.

Discussion

Complexin-1 inhibits spontaneous release while at the same time synchronizing exocytosis when intracellular calcium levels rise. Some characteristics of the in vivo function of Cpx have successfully been reconstituted (28, 42, 68, 69); however, biochemical studies of Cpx and SNARE proteins have not revealed a clear mechanism by which Cpx inhibits spontaneous fusion. In this work, we employed EPR and NMR spectroscopy, fluorescence anisotropy, and TIRF microscopy to characterize the interaction among Cpx, SNAREs, and the lipid bilayer. We demonstrate that Cpx not only binds to the assembled four helical coiled-coiled complex of Syx/SNAP25/Syb, but also binds a binary (1:1) t-SNARE complex of Syx/SNAP25 with high affinity while simultaneously interacting with the lipid bilayer. These protein and membrane interactions of Cpx are cooperative. The t-SNARE complex enhances the membrane affinity of Cpx (Fig. 5); and membrane-associated SNAREs are also more likely to interact with Cpx than their soluble counterparts. For example, Cpx shows an affinity for dSNAP25 (Fig. 6, a and b, and S5) that is not observed for SNAP25 in solution (Fig. S2). As shown previously (40), this binary t-SNARE complex undergoes rapid fusion with Syb vesicles. Using this system, we find that Cpx inhibits Ca²⁺-independent fusion by inhibiting vesicle docking, and it does so by reducing the affinity of Syb for the t-SNAREs. This reduction in Syb affinity likely accounts for the decrease in spontaneous fusion events observed in the absence of calcium in vivo.

As indicated above, previous reports have been controversial and inconsistent regarding the binding of Cpx to the t-SNAREs. Part of the difficulty in characterizing this interaction may be due to the difficulty in producing t-SNARE complexes that are homogeneous and efficient at fusion. Syntaxin-1a and SNAP25 efficiently form a 2:1 complex when reconstituted into membranes (44, 70). This off-pathway complex is slow to disassemble, and this disassembly accounts for the slow and inefficient fusion rates seen in some reconstituted systems. An artificially stabilized acceptor complex may be produced using the C-terminal end of Syb (49–96 or ΔN synaptobrevin), and this complex facilitates rapid and efficient fusion (10). However, as seen in Fig. 6 b, the binding of Cpx to this ΔN complex is weak, as is the binding of Cpx to a 2:1 (Syx/SNAP25) complex produced by a common reconstitution procedure using CHAPS. Binding to the fully assembled ternary cis-SNARE complex is stronger, but we observe the most efficient binding to the activated binary t-SNARE complex.

As seen in Figs. 3 and 4, we observe efficient membrane binding at both the N- and C-terminal ends of Cpx by EPR and NMR spectroscopy; moreover, both the N- and C-terminal interactions are dependent upon the curvature of the membrane, where the two interactions are independent of each other. These findings are generally consistent with recent reports. For example, curvature-dependent membrane binding at the C terminus has been reported for complexin from C. elegans (41) and mammalian complexin (33), and there is evidence that a fragment of Cpx containing its N-terminus participates in membrane binding (69). In previous work, the strong curvature dependence of Cpx binding was taken as evidence for an interaction of Cpx with the vesicle membrane rather than an interaction with the plasma membrane (41). Whether Cpx interacts with the vesicle membrane or the plasma membrane in vivo has not been established. From the dependence on curvature seen here (Fig. 4), we estimate that either end of Cpx will exhibit a roughly 10-fold higher affinity toward the synaptic vesicle membrane; however, given the strong t-SNARE interactions of Cpx, the N- and C-terminal domains may associate with the plasma membrane simply because of their proximity to this membrane. Of course, one can imagine that Cpx binds to the acceptor t-SNARE complex and the vesicle membrane when the two are in close proximity, as they will be when Syb inserts into the t-SNARE complex. Once the SNAREs assemble, Cpx might then also bind to the ternary complex. Regardless of the sequence of events, both membrane and t-SNARE interactions of Cpx will contribute to its overall association, and the removal of any region that contributes to either t-SNARE or membrane binding may reduce the effectiveness of Cpx. This cooperativity likely explains why removal of the Cpx C terminus modulates Cpx activity (71).

A number of models have been proposed to account for the inhibitory action of Cpx on membrane fusion. In one, Cpx binds to the SNAREs to yield a state that is clamped and not yet able to fully assemble and drive fusion (72). In this state, the Ca²⁺ sensor synaptotagmin-1 (Syt) triggers synchronous release by binding the SNAREs and displacing Cpx. However, work that is more recent shows that Cpx and Syt may both simultaneously bind SNAREs, leading to an alternate model where Syt binding rearranges the conformation of Cpx when bound to the SNAREs (73). Complexin-1 has also been proposed to arrest SNARE assembly by cross linking partially assembled SNAREs through its accessory helix (32). This model has been the subject of some debate, and whether the accessory helix interacts across SNAREs has not been resolved (38, 39). There is also a report that Cpx can exist in two configurations—one where it interacts exclusively with the ternary postfusion complex and a second where Cpx bridges the ternary complex with a binary complex (36). How such an interaction might regulate SNARE assembly is not clear. The data shown here (Fig. 3) indicate that both the N- and C termini of Cpx simultaneously interact with membranes; as a result, interactions of the accessory helix across SNAREs may not be structurally feasible. There has also been a proposal that electrostatic repulsion between vesicle and target membranes by the Cpx accessory helix accounts for the inhibitory activity of Cpx (38). However, such a model requires that the repulsive interactions be sufficiently large to inhibit SNARE assembly, which requires that the physical dimensions over which the repulsive interactions act be less than the Debye length, or <1 nm under physiological buffer conditions. All these models are built on the premise of a rapid and precisely timed recruitment of Cpx to the ternary SNARE complex where Syb is in various stages of assembly (depending on the model) with the target SNAREs. Our data clearly demonstrate that Cpx will associate with the binary target SNARE complex in the absence of vesicle SNAREs and potentially before the initiation of v- and t-SNARE assembly, providing a plausible starting point for the inhibitory action of Cpx.

In summary, the data shown here clearly demonstrate that Cpx inhibits the docking of v-SNARE vesicles and weakens the affinity of Syb for an acceptor t-SNARE complex, and it does so by simultaneous membrane and t-SNARE binding. These observations support a model, similar to one previously proposed (26), in which Cpx acts to lower the frequency of spontaneous fusion events in neuronal exocytosis by binding to the acceptor t-SNARE complex and delaying assembly of the ternary SNARE complex. Although many details regarding the state of the SNAREs and the role of Syt are still unknown, it is likely that Syx and SNAP25 are in, at some stage, a 1:1 complex with the regulatory protein Munc18 and perhaps Munc13 (74). As shown in a recent study, the binding of Munc18 to Syx does not preclude the formation of 1:1 interaction between Syx and SNAP25 (75), and Syb may also be templated to these regulatory proteins in a manner similar to that proposed for the v-SNARE in the yeast vacuole system (76). There is evidence that the Ca²⁺-triggering step does not involve a direct binding of Syt to the SNAREs. The interaction of Syt with the ternary SNARE complex is both heterogeneous and weaker than its interactions with membranes containing PI(4,5)P₂; moreover, the SNARE interaction of Syt does not appear to be strongly Ca²⁺ dependent (77). Because Syt binds strongly to membranes in the presence of Ca²⁺ and has the capacity to bridge vesicle and target membranes in the presence of Ca²⁺ (78), we imagine a model where Syt facilitates synchronous release by driving vesicle and target membranes closer together. By closing this gap, Syt would place Syb into close proximity to the acceptor t-SNARE complex, which would enhance the local concentration of Syb at the active t-SNARE complex and overcome the inhibitory action of Cpx.

Author Contributions

R.Z., A.K., B.L., V.K., L.K.T., and D.S.C. wrote the paper and designed the experiments. R.Z., A.K., B.L., and V.K. carried out the work.

Editor: Kalina Hristova.

Supporting Citations

References (79, 80) appear in the Supporting Material.