View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector , Vol. 24, 363–376, October, 1999, Copyright 1999 by Cell Press Kinetics of Responses to Ca2؉ and Assembly with the Core SNARE Complex onto Membranes

Anson F. Davis,* Jihong Bai,* Dirk Fasshauer,† 1994), as a docking prior to (DiAn- Mark J. Wolowick,* Jessica L. Lewis,* tonio and Schwarz, 1994; Reist et al., 1998), and as a and Edwin R. Chapman*‡ for AP-2 in endocytosis (Zhang et al., 1994; *Department of Physiology Jorgensen et al., 1995). While these functions are not University of Wisconsin School of Medicine mutually exclusive, only a role for synaptotagmin I as Madison, Wisconsin 53706 the Ca2ϩ sensor for neuronal exocytosis would require † Department of Neurobiology microsecond “sensing” of Ca2ϩ concentrations capable Max Planck Institute for Biophysical Chemistry of triggering release (Ͼ20 ␮MCa2ϩ; Heidelberger et al., Am Fassberg 11 1994). D-37077 Go¨ ttingen Spontaneous membrane fusion persists in synapto- Germany tagmin I nulls, indicating that synaptotagmin plays a regulatory rather than a direct role in the catalysis of fusion. In contrast, recent studies indicate that the syn- aptic vesicle membrane fusion reaction requires the as- Summary sembly (Littleton et al., 1998; Chen et al., 1999) of the heterotrimeric SNARE (soluble N-ethylmaleimide-sensi- The synaptic vesicle protein synaptotagmin I binds tive fusion protein [NSF] attachment receptor) complex -Ca2؉ and is required for efficient re (So¨ llner et al., 1993). In synapses, this complex is com- lease. Here, we measure the response time of the posed of the target membrane SNAREs (t-SNAREs) syn- C2 domains of synaptotagmin to determine whether taxin soluble NSF attachment protein–25 (SNAP-25), and 2؉ synaptotagmin is fast enough to function as a Ca the vesicle membrane SNARE (v-SNARE) synapto- sensor for rapid exocytosis. We report that synapto- brevin. Syntaxin and SNAP-25 assemble into a binary 2؉ tagmin is “tuned” to sense Ca concentrations that complex (Chapman et al., 1994; Fasshauer et al., 1997). trigger neuronal exocytosis. The speed of response is Subsequent binding of synaptobrevin results in the for- unique to synaptotagmin I and readily satisfies the mation of a highly stable, SDS-resistant ternary SNARE kinetic constraints of synaptic vesicle membrane fu- complex (Hayashi et al., 1994; Fasshauer et al., 1997). 2؉ sion. We further demonstrate that Ca triggers pene- The -resistant core of the ternary complex has tration of synaptotagmin into membranes and simulta- been identified (Fasshauer et al., 1998) and its crystal neously drives assembly of synaptotagmin onto the structure, solved (Sutton et al., 1998). The core region base of the ternary SNARE (soluble N-ethylmaleimide- is composed almost entirely of four ␣ helices: syntaxin sensitive fusion protein [NSF] attachment receptor) and synaptobrevin contribute one helix each, while complex, near the transmembrane anchor of syntaxin. SNAP-25 contributes two helices. These helices assem- These data support a molecular model in which synap- ble into a parallel coiled-coil structure that extends close totagmin triggers exocytosis through its interactions to the membrane anchors of both syntaxin and synapto- with membranes and the SNARE complex. brevin, closely juxtaposing the vesicle and target mem- branes. The proximity of SNARE-tethered membranes and the free energy associated with ternary complex Introduction formation have been proposed to drive bilayer fusion (Hanson et al., 1997). Reconstitution experiments have Synaptic vesicle exocytosis is extremely fast; the maxi- provided direct evidence that assembled SNARE com- mal rate of Ca2ϩ-triggered fusion can be as rapid as plexes may indeed be sufficient to catalyze lipid mixing 1000–3000 sϪ1, and the lag time between increases in and membrane fusion in vitro, albeit with very slow kinet- 2ϩ [Ca ]i and fusion can be as little as 60–350 ␮s (Llinas et ics (Weber et al., 1998). This finding, coupled to the al., 1981; Heidelberger et al., 1994; Sabatini and Regehr, conservation of SNARE in all eukaryotic cells, 1996). These findings place definite kinetic constraints suggests that the SNARE complex may serve as a ubiq- on the Ca2ϩ-regulated rearrangements of the protein– uitous membrane fusion “machine.” protein and protein–lipid complexes that catalyze bi- If the SNARE complex serves as the core of the exocy- layer fusion. totic machinery, how is fusion regulated by Ca2ϩ– Genetic studies demonstrated that the Ca2ϩ-binding synaptotagmin? Synaptotagmin binds Ca2ϩ and effector synaptic vesicle protein, synaptotagmin I, is essential molecules via two homologous repeats designated as for efficient excitation–secretion coupling (Nonet et al., C2 domains. The membrane-proximal is des- 1993; Geppert et al., 1994; DiAntonio and Schwarz, 1994; ignated C2A, the membrane-distal C2 domain is desig- amino 130ف Littleton et al., 1994). However, interpretation of the ge- nated C2B. C2 domains are widespread netics data has not lead to a consensus as to the in vivo acid motifs that confer Ca2ϩ-regulated lipid-binding ac- function of synaptotagmin. Proposed functions include tivity to many of their parent molecules. In some cases, acting as the low-affinity Ca2ϩ sensor that regulates Ca2ϩ-C2 domain-mediated translocation of proteins to rapid exocytosis (Geppert et al., 1994; Littleton et al., membranes accelerates their catalytic function by colo- calizing them with their substrates (e.g., and lipid ). The C2 domains of synaptotagmin ‡ To whom correspondence should be addressed (e-mail: chapman@ may function in a similar manner by accelerating SNARE physiology.wisc.edu). complex–mediated membrane fusion in response to Neuron 364

Figure 1. Monitoring Ca2ϩ and Vesicle Binding to the C2A Domain of Synaptotagmin in Solution (A) Structure of the C2A domain of synaptotagmin, in which three bound Ca2ϩ ions are shown (modified from Shao et al., 1998, using MOLSCRIPT [Kraulis, 1991]). Two versions of C2A were generated, one with a Trp substitution at position 234, within Ca2ϩ–binding loop 3, and another with a Trp placed in position 173, within Ca2ϩ–binding loop 1. (B) Spectral properties of Trp substitution mutants. Protein (10 ␮M) in HBS was excited at 285 nm in the presence of 1 mM EGTA. Ca2ϩ was then added (1 mM free) followed by addition of vesicles (10 nM) composed of 25% PS/75% PC. Fluorescence of the Trp-234 reporter is slightly decreased upon addition of Ca2ϩ (9%); fluorescence of the Trp-173 reporter is significantly increased by Ca2ϩ (27%). Both reporters show large increases in fluorescence (51% and 103% for Trp-234 and Trp-173, respectively, compared with the intensity in EGTA) and blue- shifts in their emission spectra upon binding vesicles. Vesicles did not influence the spectra in the absence of Ca2ϩ. (C) C2A–Ca2ϩ–vesicle interactions are rapid and reversible. Samples were prepared and excited as in (B), and the emission intensity was monitored versus time at 340 nm. Additions were as in (B), and the Ca2ϩ effect was reversed by addition of 2 mM EGTA. binding Ca2ϩ. More specifically, Ca2ϩ triggers penetra- C2A domain of synaptotagmin in solution in real time. tion of the C2A domain into membranes (Chapman and To this end, we scanned the surface of C2A by placing Davis, 1998; Figure 5) and triggers the clustering of multi- tryptophan (Trp) reporters in different positions and ple isoforms of synaptotagmin via the C2B domain assaying for Ca2ϩ and/or vesicle-induced changes in (Chapman et al., 1998; Osborne et al., 1999). Genetic Trp fluorescence. In a previous study, we demonstrated studies have established that the C2B domain plays a that a Trp placed in position 234 (designated C2A- critical role in excitation–secretion coupling (Littleton et F234W), within Ca2ϩ binding–loop 3 (Sutton et al., 1995), al., 1994) and that C2B-mediated heterooligomerization reported penetration of the loop into lipid bilayers in may modulate the strength of synaptic transmission (Lit- response to Ca2ϩ.Ca2ϩ-triggered insertion resulted in a tleton et al., 1999). Finally, both C2 domains of synapto- marked increase in fluorescence intensity and a blue- tagmin mediate interactions with the t-SNAREs syntaxin shift in the emission spectrum (Chapman and Davis, (Chapman et al., 1995) and SNAP-25 (Schiavo et al., 1998). Here, we report a novel Trp reporter placed at 1997). These findings indicate that synaptotagmin may position 173 (designated M173W), within Ca2ϩ–binding trigger opening of the fusion pore via direct effects loop 1 (Sutton et al., 1995), that yields a strong increase on lipid bilayers, by driving conformational/structural in fluorescence upon binding Ca2ϩ alone and further changes in SNAREs or by altering the physical relation- increases in fluorescence upon binding vesicles in the ship between SNAREs and the lipid bilayer. Crucial tests presence of Ca2ϩ. The positions of these Trp reporters of these models are to determine whether the Ca2ϩ de- in the structure of C2A and their emission spectra are pendence and speed of Ca2ϩ-triggered synaptotagmin– shown in Figures 1A and 1B. The increase in fluores- membrane interactions and synaptotagmin oligomeriza- cence intensity upon addition of vesicles occurs only tion are consistent with the physiology of neuronal 2ϩ exocytosis. In this study, we directly measure the Ca2ϩ in the presence of Ca (data not shown). These data 2ϩ sensitivity and “response time” of synaptotagmin I. Fur- demonstrate that assembly of the C2A–Ca –vesicle 2ϩ thermore, we have investigated the interaction of the complex is ordered; C2A first binds Ca , followed by 2ϩ Ca2ϩ–synaptotagmin–vesicle complex with the core of binding of the C2A–Ca complex to vesicles. We note the assembled SNARE complex. These data are consoli- that F234W and M173W were made in a new “back- dated into a model for a novel Ca2ϩ–synaptotagmin– ground” C2A domain. In a previous study, we truncated membrane–SNARE complex, which may represent an C2A at residue 258 to avoid inclusion of the sole naturally intermediate structure in the Ca2ϩ-triggered fusion re- occurring Trp at position 259 (Chapman and Davis, action. 1998). However, this truncation increased the Ca2ϩ re- quirement for lipid binding (data not shown) and is there- Results fore not appropriate for kinetics studies. The C2A-Trp reporters generated for this study are full length C2A Ca2؉ and Vesicle Binding to the C2A Domain domains that harbor a second mutation, W259F, such of Synaptotagmin, Monitored that the Ca2ϩ loop Trp reporters constitute the sole Trp with Fluorescent Reporters residues. As discussed below, the new C2A-Trp report- Our first goal was to generate optical reporters that ers retain wild-type Ca2ϩ- and vesicle-binding proper- could be used to monitor Ca2ϩ and lipid binding to the ties. Manual mixing experiments are shown in Figure 1C Synaptotagmin Dynamics and Effector Interactions 365

Figure 2. Ca2ϩ Sensitivity of Wild-type and C2A Reporter Constructs (A) C2A-Trp reporter mutations do not affect the Ca2ϩ dependence of vesicle binding. Wild-type (triangles), M173W (circles), and F234W (squares) C2A domains were immobi- lized as GST fusion proteins on glutathione- Sepharose beads at low-protein density (0.3 ␮g/␮l) and assayed for binding to 3H-labeled liposomes (25% PS/75% PC) as a function of 2ϩ [Ca ]free. Data were normalized, plotted, and fit with GraphPad Prism 2.0 software. In all 2ϩ cases, the [Ca ]1/2 values (21, 21, 16 ␮M) and Hill coefficients (2.4, 1.8, and 2.8) were similar for wild-type, F234W, and M173W C2A do- mains, respectively. (B) C2A-Trp reporter mutations do not affect the extent of vesicle binding. Wild-type (WT), M173W, or F234W mutant C2A domains (15 ␮g) were immobilized as GST fusion proteins and assayed for vesicle binding (1.75 ␮g total 3H-labeled lipids) in 2 mM EGTA (Ϫ)or1mM Ca2ϩ (ϩ) in 150 ␮l HBS, as described (Chap- man and Davis, 1998). (C) Ca2ϩ dependence of C2A-F234W-vesicle 2ϩ interactions measured in solution. The emission spectra shown in Figure 1B were corrected, integrated, and plotted as a function of [Ca ]free 2ϩ 2ϩ ([Ca ]1/2 ϭ 74 ␮MCa , Hill coefficient ϭ 1.9). In all experiments, error bars represent the standard deviations from a minimum of three independent determinations. and demonstrate the utility of using the Trp reporters and results in greater Hill coefficients (data not shown), to monitor assembly and disassembly reactions in solu- presumably via the polyvalent nature of the binding reac- tion. Interactions with Ca2ϩ and vesicles were reversible tion. This effect was abrogated by reducing the density and occurred on the subsecond timescale. Therefore, of the fusion protein to 0.3 ␮g fusion protein/␮l beads. these interactions were resolved in time by a stopped- As shown in Figures 2A and 2B, the Ca2ϩ dependence flow rapid mixing approach, as described below. and the extent of vesicle binding to immobilized wild-type 2ϩ and Trp mutant C2A domains were similar ([Ca ]1/2 ϭ Ca2؉ Sensitivity of the C2A Domain of Synaptotagmin 16–21 ␮M with Hill coefficients of 2–3), demonstrating The threshold for secretion is Ͼ20 ␮MCa2ϩ, and half- that these reporters are valid for use in kinetics studies. maximal rates of transmitter release occur at Ca2ϩ con- To more accurately measure the Ca2ϩ dependence -␮M (Heidelberger et al., 1994). In and cooperativity of C2A–Ca2ϩ–vesicle interactions, in 200ف centrations of contrast, some groups have reported that the [Ca2ϩ] for dependent of the immobilization and washing protocols 2ϩ half-maximal binding ([Ca ]1/2) of vesicles to C2A is 4–6 required in the bead-binding experiments, we cleaved ␮M (Davletov and Su¨ dhof, 1993; Li et al., 1995). These C2A-F234W from the GST fusion moiety using thrombin findings suggest that the Ca2ϩ dependence on C2A– and used the Trp reporter to monitor membrane binding 2ϩ membrane interactions is too low to function directly in solution as a function of [Ca ]free (Figure 2C). Under in triggering exocytosis. We addressed this issue by these conditions, the cooperativity of the Ca2ϩ dose– determining precisely the Ca2ϩ dependence for the inter- response was identical to that observed in bead-binding action of the C2A domain of synaptotagmin with vesicles assays using low densities of immobilized fusion protein containing anionic lipids. These studies were also di- (Hill coefficient of 1.9). Thus, two or more Ca2ϩ-binding rected at determining whether the reporter mutations sites regulate C2A–membrane interactions. However, 2ϩ affect the equilibrium-binding properties of C2A-F234W when measured in solution, the [Ca ]1/2 for C2A-F234W and C2A-M173W in subsequent kinetics studies. Be- binding to vesicles was 74 ␮M. These data indicate that 2ϩ 2ϩ 2ϩ cause the [Ca ] for half-maximal binding ([Ca ]1/2)to the Ca requirement for binding to vesicles containing membranes is dependent on the acidic phospholipid 25% PS is greater (21–74 ␮MCa2ϩ) than previously esti- content (Brose et al., 1992), our experiments employed mated (4–6 ␮MCa2ϩ; Davletov and Su¨ dhof, 1993) and vesicles that contained 25% phosphatidyl serine (PS) corresponds to or exceeds the Ca2ϩ threshold for neu- and 75% phosphatidyl choline (PC) to mimic the acidic ronal exocytosis (Ͼ20 ␮MCa2ϩ; Heidelberger et al., lipid content of both synaptic vesicles and synapto- 1994). These data are consistent with the proposed somal plasma membranes (Breckenridge et al., 1972, function of synaptotagmin as a low-affinity Ca2ϩ sensor. 1973). Ca2ϩ-dependent vesicle binding was first assayed –with glutathione-Sepharose-immobilized glutathione Kinetics of C2A–Ca2؉ and C2A–Ca2؉ S-transferase-C2A (GST–C2A) fusion proteins and 3[H]- Vesicle Interactions labeled vesicles. In the course of these experiments, Because the Trp reporter mutations did not affect the we discovered that the density of the GST–C2A fusion Ca2ϩ-binding properties of C2A, these constructs were protein on the beads influences the Ca2ϩ dependence of used for kinetics experiments. We first carried out the interaction. High-protein density (e.g., 12 ␮g fusion steady-state titrations of C2A-M173W with Ca2ϩ in the protein/␮l beads) shifts the Ca2ϩ dependence to the left absence of vesicles (Figure 3A). Half-maximal binding Neuron 366

Figure 3. Steady-State and Kinetic Measure- ments of Ca2ϩ Binding to the C2A Domain of Synaptotagmin (A) The Ca2ϩ-induced change in C2A-M173W fluorescence was exploited to measure the dissociation constant for this binding site(s). C2A-M173W (5 ␮M) was excited at 285 nm, and spectra were collected from 300–400 nm as a function of [Ca2ϩ]. Spectra were inte- grated, corrected, normalized, and plotted 2ϩ (closed circles) versus [Ca ]free. Data were fit

with GraphPad Prism 2.0 software. The Kd was 61 ␮M, consistent with NMR studies of Ca2ϩ–binding site 1 (Shao et al., 1996); the Hill coefficient was 1.1. For comparison, the amplitudes of the kinetic traces in (B) are also 2ϩ plotted versus [Ca ]free (open circles), demonstrating the agreement between the steady-state and kinetics experiments. (B) Kinetics of Ca2ϩ binding to C2A-M173W. This series of stopped-flow experiments was carried out at 4ЊC using 5 ␮M protein. The sample was excited at 285 nm and emitted light collected with a 335 nm cutoff filter. Amplitudes of response were normalized and plotted versus time. At the lowest [Ca2ϩ] tested, equilibrium was reached within the dead time of the instrument (1.2 ms), providing a lower limit for the .MϪ1sϪ1 108ف second-order rate constant of

2ϩ 10 Ϫ1 Ϫ1 occurred at 61 ␮MCa (Hill coefficient of 1.1). Nuclear kon (0.8 ϫ 10 M s ; Equation 1). The value for kon magnetic resonance (NMR) studies indicated the pres- closely approaches the collisional limit calculated for ence of two binding sites with dissociation constants this reaction (1.9 ϫ 1010 MϪ1sϪ1 for vesicles with a diame- of 60 and 400 ␮MCa2ϩ (Shao et al., 1996). It is likely ter of 100 nm; Lu et al., 1995; Equations 2 and 3). We that Trp-173, in Ca2ϩ–binding loop 1, is reporting binding note that identical rates are observed when fluores- of Ca2ϩ to the “high”-affinity site. The fluorescence cence resonance energy transfer (FRET) measurements changes that result from Ca2ϩ binding were resolved in are used to monitor C2A-F234W-Ca2ϩ interactions with time by a stopped-flow rapid mixing approach (Figure dansyl-labeled vesicles (discussed in Figure 5). FRET 3B). Even at the lowest [Ca2ϩ] tested (30 ␮M), binding monitors the binding of C2A-F234W to the membrane was complete in the dead time of the instrument (1.2 surface via proximity of the Trp–dansyl donor–acceptor ms). From these data, the lower limit for the Ca2ϩ on- pair, providing a means to measure binding independent MϪ1sϪ1. These data suggest of Trp insertion. Binding of C2A-F234W to vesicles and 108ف rate is estimated to be that Ca2ϩ binds C2A with kinetics that approach or ex- insertion of the Trp residue are virtually simultaneous .MϪ1sϪ1; Falke et al., 1994; reactions and are governed solely by diffusion 108ف) ceed the collisional limit Equations 2 and 3). These findings are consistent with The kinetics data described above indicate that C2A structural studies, indicating that Ca2ϩ does not drive binds to membranes very tightly (Kd ϭ 40 nM at 100 ␮M 2ϩ large conformational changes in C2A (Shao et al., 1998), Ca and vesicles composed of 25% PS/75% PC). Thus, 2ϩ making it possible for C2A to rapidly bind Ca2ϩ. From PS and Ca cooperate in the stabilization of this com- 2ϩ the Ca2ϩ on-rate and equilibrium constant (Figure 3), it plex, as predicted by the physical proximity of the Ca - 2ϩ and lipid-binding sites within C2A (Chapman and Davis, is apparent that the Ca off-rate (koff-Ca2ϩ) exceeds 6 ϫ 103 sϪ1. 1998). We next determined the rate at which Ca2ϩ triggers We next focused on the ability of the C2A-F234W- Ca2ϩ-vesicle complex to respond to decreases in Ca2ϩ binding and penetration of C2A-F234W into lipid bi- concentration. For these experiments, C2A-F234W was layers. In stopped-flow rapid mixing experiments, we assembled onto vesicles (25% PS/75% PC) with 100 observed that the rate at which C2A-F234W-Ca2ϩ bound ␮MCa2ϩ. Complexes were subsequently disassembled to vesicles was identical regardless of whether Ca2ϩ was by rapid mixing with excess Ca2ϩ chelator. As shown in premixed with C2A-F234W or vesicles, or both (data not 2ϩ Figure 4C, the disassembly kinetics were extremely fast shown). These data demonstrate that Ca binding is Ϫ1 2ϩ 2ϩ (kdiss ϭ 700 s ), indicating that Ca is rapidly exchanged not rate limiting for C2A-F234W-Ca -vesicle assembly. in the protein–membrane complex. These data demon- Thus, a simple bimolecular model can be applied to strate that synaptotagmin can rapidly respond to de- 2ϩ study complex formation between C2A-F234W-Ca 2ϩ creases as well as increases in [Ca ]i. As with koff (de- and vesicles. An example of Ca2ϩ-triggered membrane scribed above), kdiss can be reduced by increasing the association of C2A-F234W is shown in Figure 4A. C2A- PS content of the vesicles or by assembling complexes 2ϩ F234W was mixed with Ca (100 ␮M) and vesicles (11 at higher [Ca2ϩ] (data not shown). Again, the effect of nM, 25% PS/75% PC), resulting in a rapid increase in increasing [Ca2ϩ] persists at concentrations beyond sat- Trp fluorescence. We observed that higher concentra- uration of the protein and is likely to be due to direct 2ϩ tions of Ca reduce the observed rate constant, kobs, effects of Ca2ϩ on the vesicles. At lower, subsaturating 2ϩ by reducing koff (data not shown) and by “trapping” C2A- [Ca ], Trp-quenching experiments (described in Figure F234W via Ca2ϩ-induced aggregation of vesicles. 5A) indicate that the [Ca2ϩ] determines the depth of The second-order rate constant for C2A–Ca2ϩ–vesicle penetration of the Trp reporter (data not shown), per- assembly with respect to [vesicle] kon was determined haps by occupying multiple binding sites. by plotting kobs as a function of the [vesicle] (Figure 4B). The kinetics data described above are summarized Ϫ1 The Y intercept yields koff (240 s ), and the slope yields in the model shown in Figure 4D. Synaptotagmin Dynamics and Effector Interactions 367

Figure 4. Assembly and Disassembly Kinetics of C2A–Ca2ϩ–Vesicle Complexes The fluorescence change exhibited by C2A-F234W upon Ca2ϩ-triggered binding to vesicles was used to monitor the kinetics of C2A–Ca2ϩ–vesicle interactions. C2A-F234W was excited at 285 nm, and emitted light was collected with a 326.1 nm band-pass filter. All kinetics data are shown referenced to an arbitrary time ϭ 0. The 2 ms plateau phase at the beginning of each trace corresponds to the signal, which is collected during the “flow” and prior to the “stop”; the dead time was 1.2 ms. (A) Binding of C2A-F234W-Ca2ϩ to vesicles in real time. Vesicles (25% PS/75% PC; 11 nM final concentration) were premixed with Ca2ϩ (100 ␮M final concentration) and then rapidly mixed with C2A-F234W (5 ␮M final concentration). As a control, mixing experiments were also carried out using 2 mM EGTA instead of Ca2ϩ (lower trace), providing a true minimum reference point. Data (1000 points) were collected for 100 ms Ϫ1 and are plotted with a best-fit single exponential function (kobs ϭ 312 s ). The first 10 ms are shown on an expanded timescale in the inset. 2ϩ (B) Determination of kon and koff for the C2A-F234W-vesicle complex in the presence of Ca . Stopped-flow experiments were carried out as in (A), and kobs was determined by fitting the data with single exponential functions, and plotted versus [vesicle]. The Y intercept yields koff for 2ϩ Ϫ1 10 Ϫ1 Ϫ1 the C2A-F234W-vesicle complex in the presence of Ca (240 s ), and the slope yields kon (0.8 ϫ 10 M s ) for binding of C2A-F234W- Ca2ϩ to vesicles (Equation 1). Error bars represent standard deviations from four independent experiments. (C) Disassembly kinetics of the C2A-F234W-Ca2ϩ-vesicle complex upon chelation of Ca2ϩ. C2A-F234W-Ca2ϩ-vesicle complexes were assembled at 100 ␮MCa2ϩ with vesicles composed of 25% PS/75% PC. Disassembly reactions were carried out by rapidly mixing these complexes with 5 mM EGTA (final). As a control, samples were mixed with buffer lacking EGTA, thus providing a maximum signal reference. Disassembly data were collected for 100 ms; the plot includes a best-fit single exponential function. The first 10 ms are shown on an expanded timescale in the inset. 2ϩ 8 Ϫ1 Ϫ1 3 Ϫ1 (D) Model for C2A–Ca –vesicle dynamics. Summary of the rate constants: kon-Ca2ϩ Ն 10 M s ;koff-Ca2ϩ Ն 6 ϫ 10 s (determined from 10 Ϫ1 Ϫ1 Ϫ1 2ϩ kon-Ca2ϩ and a Kd of 61 ␮M; Figure 3A; Shao et al., 1996); kon ≈ 10 M s ;koff ϭ 240 s (in the presence of 100 ␮MCa with vesicles composed Ϫ1 2ϩ of 25% PS/75% PC); kdiss ϭ 700 s (for complexes assembled at 100 ␮MCa with vesicles composed of 25% PS/75% PC).

Mechanism of Ca2؉-Triggered C2A–Vesicle Assembly instrument, we were able to compare the insertion rates Trp-173 exhibits an increase in fluorescence intensity of Trp-234 and Trp-173 by simply monitoring Trp fluores- and a blue-shifted emission spectrum upon binding to cence. As shown in Figure 5B, the loop 3 reporter, Trp- vesicles, indicating that, like Trp-234, Trp-173 pene- 234, penetrates bilayers more rapidly than does the loop trates into lipid bilayers. A recent study demonstrated 1 reporter, Trp-173. This difference was further charac- that all three flexible loops of cytosolic terized, using FRET between the Trp reporter and dansy-

A2 (cPLA2) make direct contacts with membranes (Nalef- lated lipids in the vesicle surface. As shown in Figures ski and Falke, 1998). To directly address this question, 5C and 5D, FRET measurements precisely coincided we utilized membrane-embedded fluorescence quench- with changes in Trp fluorescence, demonstrating that ers (Chapman and Davis, 1998). Vesicles were prepared binding of the Ca2ϩ loops to the vesicle surface is virtu- with 25% PS and 75% brominated PC. If a Trp reporter ally simultaneous with insertion of the Trp reporter. pierces the bilayer, it will contact the bromine group and become efficiently quenched. As shown in Figure 5A, Comparison of Synaptotagmin I C2A Domain both Trp-234 and Trp-173 were quenched by the mem- Dynamics with Other C2 Domains brane-embedded bromine groups, demonstrating that In the final series of C2A kinetics experiments, we sought both Ca2ϩ-binding loops penetrate into the lipid bilayer. to determine whether this speed of response is unique We next sought to determine whether the loops pene- to synaptotagmin I. To address this question, we com- trate simultaneously or sequentially. Because Ca2ϩ pared the dynamics of C2A-F234W, derived from synap- binds C2A within the dead time of the stopped-flow totagmin I, with the C2A domain of synaptotagmin III Neuron 368

Figure 5. Sequential Insertion of Loops 1 and 3 of C2A into Lipid Bilayers (A) Ca2ϩ binding–loops 1 and 3 of C2A pene- trate into lipid bilayers. Upper panel corre- sponds to C2A-F234W, lower panel corre- sponds to C2A-M173W. Protein (10 ␮M) was excited at 285 nM in the presence of 1 mM Ca2ϩ and either 25% PS/75% PC, 25% PS/ 75% 6–7-dibromo-PC, or 11–12-dibromo-PC. The number refers to the position of bromine quenchers on the sn-2 acyl chain of PC. F234W and M173W were quenched 36% and 29% by 11–12-dibromo-PC, and 52% and 56% by 6–7-dibromo-PC, respectively. (B) Kinetics of C2A-F234W and C2A-M173W penetration into lipid bilayers. Measurements were made as described in Figure 4A using 5 ␮M protein, 100 ␮MCa2ϩ, and 10 nM vesicles. Insertion is sequential, Trp-234 inserts before Trp-173. (C and D) Comparison of Trp insertion rates and FRET measurements of C2A-F234W and C2A-M173W. Trp insertion was monitored as described in (B) above. FRET was monitored by exciting the Trp residue at 285 nm and measuring the emission of a dansyl-phospha- tidyl ethanolamine (PE) acceptor in the vesi- cles (composed of 25% PS/70% PC/5% dan- syl-PE) via a 523 nm band-pass filter. Conditions were 5 ␮M protein, 100 mM Ca2ϩ, and 10 nM vesicles. Trp insertion and binding measured via FRET are virtually simultaneous.

(Mizuta et al., 1994). Synaptotagmin III is the putative synaptotagmin oligomerization, using a stopped-flow Ca2ϩ sensor hypothesized to trigger slow secretion from rapid mixing approach. As shown in Figure 6C, rapid large dense-core vesicles, and slow, asynchronous re- mixing of synaptotagmin with 1 mM Ca2ϩ resulted in lease from synaptic vesicles (Geppert et al., 1994; Li et rapid clustering. The majority of the signal was obtained al., 1995). For these experiments, a sole Trp reporter in the dead time of the instrument (1.2 ms) at only 8 ␮M was placed in Ca2ϩ–binding loop 3 of the C2A domain [protein]. Thus, Ca2ϩ triggers synaptotagmin oligomer- of synaptotagmin III as described for C2A-F234W, from ization on the submillisecond timescale. The density synaptotagmin I (Figure 1). We also conducted similar of synaptotagmin on the surface of synaptic vesicles experiments using a Trp reporter version of cPLA2. Fi- indicates that clustering would occur on an even more nally, we exploited the Trp residues that naturally occur rapid timescale in vivo and is clearly rapid enough to at the lipid-binding interface of the C2 domain of protein couple Ca2ϩ to the regulation of the fusion pore. We note C␤ (PKC␤; Sutton and Sprang, 1998). Assembly that at least three distinct kinetic phases are apparent in (Figure 6A) and disassembly (Figure 6B) reactions were the kinetic trace. These phases are likely due to the carried out with each of the four C2 domains. Clearly, formation of larger synaptotagmin aggregates; the in- the C2A domain of synaptotagmin associates and disso- creasing size of these aggregates reduces the rate of ciates from membranes more rapidly than either of the oligomerization. other C2 domains. These data indicate that the C2A It should be noted that the Ca2ϩ dependence for sy- domain of synaptotagmin I is specialized for speed of naptotagmin oligomerization varies, depending on the response. method used to measure the interaction (Chapman et al., 1996; Sugita et al., 1996; Osborne et al., 1999). How- Kinetics of Ca2؉-Triggered C2B-Mediated ever, a recent solution-based binding assay revealed 2ϩ Synaptotagmin Oligomerization a [Ca ]1/2 of 140 ␮M for the heterooligomerization of In the final series of kinetics experiments, we examined synaptotagmin I and II, consistent with the Ca2ϩ require- the speed at which Ca2ϩ triggers C2B-mediated cluster- ment for secretion (Heidelberger et al., 1994). ing of synaptotagmin. The C2B domain plays an impor- tant role in excitation–secretion coupling; mutations in Simultaneous Binding of Synaptotagmin this domain strongly reduce the Ca2ϩ sensitivity of the to Membranes and to the Four Helix secretory machinery (Littleton et al., 1994), and C2B- Core of the SNARE Complex mediated heterooligomerization of synaptotagmin I and The data described above demonstrate that the Ca2ϩ IV may modulate release probability (Littleton et al., sensitivity and speed of response of the C2 domains of 1999). We speculate that C2B-mediated clustering may synaptotagmin are consistent with physiological mea- regulate the assembly, opening, or dilation of a fusion surements of neuronal exocytosis. While synaptotagmin pore (Fernandez-Chacon and Alvarez de Toledo, 1995). may function to regulate fusion, mounting evidence indi- Light scattering was used to monitor the kinetics of cates that the SNARE complex may directly mediate Synaptotagmin Dynamics and Effector Interactions 369

Figure 6. Comparison of the Kinetics of the C2A Domain of Synaptotagmin I with Other C2 Domains and Kinetics of C2B Domain– Mediated Oligomerization (A and B) Association (A) and disassembly (B) kinetics of the C2A-F234W (derived from synaptotagmin I; syt I), synaptotagmin III (syt III; Mizuta et al., 1994; Fukuda et al., 1995),

PKC␤ (Shao et al., 1996), and cPLA2 (Nalefski et al., 1997; Perisic et al., 1998) were exam- ined by monitoring Trp fluorescence as a function of time as described in Figure 4. Complexes were assembled with 1 mM [Ca2ϩ] (final) and disassembled with 5 mM EGTA. Vesicles (11 nM) were composed of 25% PS/ 75% PC. Because of the differences in the kinetics of each C2 domain, data were col- lected over a logarithmic timescale. (C) Kinetics of Ca2ϩ-induced C2B-mediated synaptotagmin oligomerization. For these ex- periments, the cytoplasmic domain of recom- binant oligomerization-competent synapto- tagmin (8 ␮M) was rapidly mixed with 1 mM Ca2ϩ (upper trace) or 2 mM EGTA (lower trace) in HBS. Clustering of synaptotagmin was monitored by illuminating at 335 nm and col- lecting scattered light via a 335 nm cutoff filter. The light-scatter signal is shown on an expanded time frame in the inset. Mg2ϩ did not trigger oligomerization and served as a negative control (data not shown). lipid bilayer fusion (Littleton et al., 1998; Weber et al., We then sought to determine whether synaptotagmin 1998). Thus, if Ca2ϩ-triggered synaptotagmin–membrane can bind to the mini complex and to membrane vesicles complex formation serves as a coupling step in exo- composed of PS/PC at the same time. The immunopre- cytosis, it is likely that this complex would interact with cipitation shown in Figure 7B was repeated to include SNAREs to regulate their function. Synaptotagmin binds samples lacking mini complex or synaptotagmin, and the isolated t-SNAREs syntaxin and SNAP-25 in a Ca2ϩ- the immunoprecipitates were assayed for Ca2ϩ-trig- promoted manner (Chapman et al., 1995; Schiavo et gered PS/PC vesicle-binding activity. As shown in Fig- al., 1997; data not shown). However, given the speed ure 7C, synaptotagmin that coimmunoprecipitated with of secretion, it is unlikely that the targets of Ca2ϩ– the mini complex was capable of simultaneously binding synaptotagmin action are isolated t-SNAREs, but rather, vesicles. The ability of synaptotagmin to simultaneously are t-SNAREs in preformed binary complexes or par- penetrate membranes and bind the core of the SNARE tially/fully assembled ternary SNARE complexes (Weber complex indicates that the Ca2ϩ- and lipid-binding loops et al., 1998). Therefore, in the next series of experiments, of synaptotagmin are not available to form direct con- 2ϩ we sought to determine whether Ca –synaptotagmin, tacts with SNAREs. Therefore, a distinct “face” of the 2ϩ as well as the Ca –synaptotagmin–vesicle complex, C2 domains of synaptotagmin engages the core of the can interact with the fully assembled SNARE complex. SNARE complex (discussed below). To date, the struc- Previous domain mapping studies demonstrated that tural elements that mediate synaptotagmin–H3 interac- synaptotagmin binds to the H3 domain of syntaxin tions have not been reported. (Chapman et al., 1995; Kee and Scheller, 1996). There- The mini complex contains two helical strands of 2ϩ fore, we determined whether Ca can trigger binding SNAP-25. A previous report indicated that the synapto- of synaptotagmin to the H3-containing minimal core of tagmin-SNAP-25 interaction is largely Ca2ϩ independent the SNARE complex (referred to as the mini complex; (Schiavo et al., 1997). In our experiments, the Ca2ϩ- Fasshauer et al., 1998; Sutton et al., 1998). This complex dependent interaction between synaptotagmin and is solely composed of the four helix bundle and lacks t-SNAREs is lost a few days following purification and all other domains. A gel of the mini complex is shown may account for this apparent discrepancy. in Figure 7A. This complex is resistant to SDS but can be dissociated by boiling. The assembled complex was 2 –incubated with recombinant synaptotagmin in the pres- Structure of the Ca ؉–Synaptotagmin ence and absence of Ca2ϩ, and immunoprecipitated with Membrane–SNARE Complex anti-synaptobrevin antibodies. Robust Ca2ϩ-triggered There are disparate reports regarding the structural ele- binding of synaptotagmin to the mini complex was ob- ments that mediate synaptotagmin–syntaxin interac- 2ϩ ␮M tions. In the next series of experiments, we sought to 100ف) served (Figure 7B). The [Ca ]1/2 for binding Ca2ϩ; data not shown) is in agreement with the Ca2ϩ determine whether synaptotagmin binds both the H3 dependence for secretion. These experiments establish (Chapman et al., 1995; Kee and Scheller, 1996) and Habc that the core of the SNARE complex serves as a target domains (Shao et al., 1997, 1998; Fernandez et al., 1998) for synaptotagmin–Ca2ϩ action. of syntaxin. We also sought to identify the structural Neuron 370

Figure 7. Ca2ϩ Triggers Simultaneous Bind- ing of Synaptotagmin to Membranes and to the Four Helix Core of the SNARE Complex (A) The “mini” SNARE complex is SDS resis- tant. Mini SNARE complex (1.5 ␮g) was disso- ciated into its component parts (residues 1–96 of synaptobrevin, 1–80 and 120–206 of SNAP-25, and 180–262 of syntaxin) by boiling in SDS sample buffer. (B) The ability of synaptotagmin (2.5 ␮M) to bind to the mini complex (3 ␮M) in 2 mM EGTA (ϪCa2ϩ) or 1 mM Ca2ϩ (ϩCa2ϩ) was assayed by coimmunoprecipitation with anti-synapto- brevin antibodies as described in the Experi- mental Procedures. Note: in all immunopre- cipitation experiments, a nonoligomerizing version of synaptotagmin I was employed (see Experimental Procedures for details). (C) Synaptotagmin–Ca2ϩ–mini complexes were assembled (at 4 ␮M synaptotagmin and 4 ␮M mini complex) and purified by immuno- precipitation as described in (B). In parallel, controls lacking either SNARE complexes or synaptotagmin were also prepared. The im- munoprecipitated complexes were then as- sayed for radiolabeled liposome-binding ac- tivity as described in Figure 2B. Error bars represent the standard deviation from tripli- cate determinations.

elements of synaptotagmin that mediate syntaxin bind- bind to a region of the H3 domain that extends to, and ing. The goal of these studies is to arrive at a structural potentially includes a portion of, the transmembrane model for the novel Ca2ϩ–synaptotagmin–membrane– domain of syntaxin. The effect of Ca2ϩ to potentiate SNARE complex. ␣/␤-SNAP binding to full-length syntaxin is currently un- Two independent approaches were employed to de- der investigation. termine whether synaptotagmin binds syntaxin via both The inability of the Habc-containing GST–1–265, GST– the Habc and H3 domains. In the first approach, domains 193, and GST–1–177 fusion proteins to bind native syn- of syntaxin 1A were immobilized as GST fusion proteins aptotagmin could be due to steric hindrance from the and assayed for their abilities to bind native synaptotag- GST moiety. We therefore generated soluble 1–177 via min derived from synaptosomal detergent extracts. As thrombin cleavage of the corresponding GST fusion pro- positive controls, we assayed for ␣/␤-SNAP, SNAP-25, tein and assayed the ability of the untagged fragment and synaptobrevin binding to immobilized syntaxins. to bind immobilized GST–C2A derived from synaptotag- As a negative control, we assayed for synaptophysin mins I and IV. The C2A domain of synaptotagmin IV binding. contains a naturally occurring aspartate-to-serine muta- Synaptotagmin and / -SNAP compete for binding to ␣ ␤ tion that disrupts Ca2ϩ-binding activity (von Poser et al., syntaxin (So¨ llner et al., 1993; data not shown), indicating 1997) and serves as a negative control. As shown in that they bind to a common site. Indeed, synaptotagmin Figure 8B, the 1–177 fragment failed to form detectable and ␣/␤-SNAP bound to all syntaxin constructs that complexes with GST-C2A-I at concentrations 20-fold contained an intact H3 domain and transmembrane do- greater than those used for the full-length syntaxin-posi- main (1–288, 160–288, 180–288, 194–288), but neither protein bound to fragment 230–288. Thus, the N-terminal tive control. The 1–177 fragment also failed to coimmu- border of the synaptotagmin/␣/␤-SNAP-binding domain noprecipitate with the full-length cytoplasmic domain lies between 194–230. All C-terminal truncation mutants of synaptotagmin (data not shown). Furthermore, the tested (1–265, 1–193, 1–177) failed to bind significant isolated soluble H3 domain bound to GST-C2A-I to the levels of synaptotagmin and ␣/␤-SNAP, despite the same extent as full-length syntaxin. We conclude that presence of an intact Habc domain. To further map the all detectable synaptotagmin-binding activity lies within synaptotagmin-binding site, we prepared a deletion mu- the H3/transmembrane domain of syntaxin (Chapman tant lacking the membrane-proximal 15 amino acids of et al., 1995; Kee and Scheller, 1996). These findings the H3 domain. This deletion (⌬250–265) strongly inhib- are consistent with a model in which synaptotagmin ited synaptotagmin and ␣/␤-SNAP binding, further sup- regulates exocytosis via interactions with the “base” porting a model in which synaptotagmin and ␣/␤-SNAP of the SNARE complex, near the membrane anchor of Synaptotagmin Dynamics and Effector Interactions 371

Figure 8. The H3 Domain of Syntaxin Contains the Sole Synaptotagmin-Binding Site (A) Binding of native synaptic proteins to syntaxin truncation and deletion mutants. GST–syntaxin fusion proteins (20 ␮g) were immobilized with glutathione-Sepharose and incubated with synaptosomal Triton X-100 extracts (1 mg protein at 1 mg/ml) in the presence of 2 mM EGTA (Ϫ)or1mMCa2ϩ (ϩ)for2hrat4ЊC. Beads were washed and subjected to SDS–PAGE and immunoblot analysis to detect synaptotagmin and, as controls, SNAP-25, synaptobrevin II, ␣/␤-SNAP, and synaptophysin. Immunoreactive bands were visualized with enhanced chemilumi- nescence. Twenty-five percent of the bound material and 6 ␮g extract (total) were analyzed. (B) Binding of soluble syntaxin fragments to the immobilized C2A domain of I and IV. Full-length (0.5 ␮M), 1–177 (10 ␮M), and 180–288 (0.5 ␮M) fragments of syntaxin were prepared as described in the Experimental Procedures and incubated with 20 ␮g GST, GST-C2A-I, or GST-C2A-IV in 150 ␮l TBS plus 0.5% Triton X-100 in either 2 mM EGTA (ϪCa2ϩ) or 1 mM Ca2ϩ (ϩCa2ϩ)for2hrat4ЊC. Pellets (30%) were loaded onto the gel; “total” corresponds to 100 ng of full-length syntaxin and 50 ng of the 1–177 and 180–288 fragments. (C) Schematic representation of the regions of syntaxin that interact with other proteins. Helical regions (designated by “H”) are shaded. Ha, Hb, and Hc correspond to helical domains that assemble into a three-stranded helical bundle (Fernandez et al., 1998), and H3 corresponds to a region that forms a helix within the ternary SNARE complex (Sutton et al., 1998). The closed rectangle corresponds to the transmembrane domain. Binding domains for other binding proteins are indicated with bars (Betz et al., 1997, and references therein; Fujita et al., 1998; Naren et al., 1998). syntaxin, where C2A can simultaneously interact with Finally, we neutralized two acidic Ca2ϩ ligands, aspar- membranes (Figure 7C). tates 230 and 232, within the C2A domain of synaptotag- In the final series of experiments, we sought to deter- min by substitution with asparagine residues. These mu- mine whether synaptotagmin binds syntaxin solely via tations inhibit the ability of C2A to bind Ca2ϩ and syntaxin the C2A domain (Li et al., 1995) or whether both C2A (Li et al., 1995) and disrupted Ca2ϩ-triggered synapto- and C2B participate in binding (Chapman et al., 1996). tagmin binding to syntaxin in our assay system, even We note that relatively low levels of syntaxin bound to though the C2B domain remained intact (Figure 1B). We conclude that C2A contains Ca2ϩ ligands that are ;%10ف) the immobilized C2A domain of synaptotagmin Figure 8B). In contrast, stable and stoichiometric synap- essential for synaptotagmin–syntaxin interactions, while totagmin–Ca2ϩ–syntaxin complexes are formed by incu- C2B possesses a second contact site that contributes bating the cytoplasmic domain of synaptotagmin I to the stability of the complex. The syntaxin-binding (amino acids 96–421) with full-length syntaxin at micro- domain of synaptotagmin is shown schematically in Fig- molar concentrations (Chapman et al., 1995). These ure 9C. complexes can be immunoprecipitated with anti-syn- taxin antibodies and visualized by staining with Coo- massie blue. This assay was used to identify the shortest Discussion sequence of synaptotagmin that forms high-affinity complexes with syntaxin. C-terminal synaptotagmin We have determined the Ca2ϩ sensitivity and speed of truncations to residue 96–381, 96–359, or 96–337 were response of synaptotagmin I, the putative Ca2ϩ sensor tolerated (Figure 9A). However, all shorter truncations, for rapid neuronal exocytosis. We observed that the C2A to residues 96–315, 96–293, and 96–265, resulted in fail- domain of synaptotagmin is “tuned” to respond to Ca2ϩ ure to form stoichiometric complexes. These data indi- concentrations (21–74 ␮MCa2ϩ; Figures 2A and 2C) that cate that for high-affinity binding to syntaxin, the C2A trigger synaptic vesicle membrane fusion (threshold, domain of synaptotagmin (residues 96–265) is not suffi- Ͼ20 ␮MCa2ϩ; half-maximal rates at 194 ␮MCa2ϩ; Heidel- cient and must be extended more than halfway into berger et al., 1994). This Ca2ϩ sensitivity is in marked

C2B domain of the molecule, to residue 337 (Figure 8C). contrast to the C2 domain of cPLA2, which exhibits a Neuron 372

Figure 9. Both C2 Domains of Synaptotag- min Are Required for High-Affinity Syntaxin- Binding Activity (A) Binding of synaptotagmin C-terminal trun-

cation mutants to syntaxin. Full-length His6- syntaxin (syx; 0.5 ␮M) was incubated with recombinant fragments of synaptotagmin (stg; 2.5 ␮M) in 170 ␮l TBS plus 0.5% Triton X-100 in the presence of 2 mM EGTA (-Ca2ϩ) or1mMCa2ϩ (ϩCa2ϩ). As a control for non- specific precipitation of synaptotagmin, sam- ples were prepared lacking syntaxin (Ϫsyx). The sequence of each synaptotagmin frag- ment is indicated on the left. Synaptotagmin binding was assayed by coimmunoprecipita- tion with syntaxin as described in the Ex- perimental Procedures. Thirty percent of the immunoprecipitate (IP) and 0.5 ␮g of synap- totagmin were subjected to SDS–PAGE and stained with Coomassie blue. Abbreviation: L-chain, light chain from the anti-syntaxin monoclonal antibody. (B) Ca2ϩ binding to C2A is essential for pro- moting synaptotagmin–syntaxin interactions. Aspartate residues 230 and 232 were neutral- ized by substitution with asparagines, and the ability of the mutant cytoplasmic domain of synaptotagmin to bind syntaxin analyzed by coimmunoprecipitation as described in (A). The D230,232N mutant failed to bind syntaxin. (C) Schematic diagram indicating the primary sequence of synaptotagmin that is required to form high-affinity complexes with syntaxin. The closed rectangle corresponds to the transmembrane domain; the C2 domains are shaded. Residues 96–139 are not required for C2A–syntaxin interactions (Li et al., 1995; data not shown); therefore, the syntaxin-binding domain is shown extending from residues 140–337.

2ϩ submicromolar [Ca ]1/2 for lipid binding (Clark et al., to rapidly respond to both decreases as well as in- 1991). creases in [Ca2ϩ]. Kinetics measurements revealed that C2A is also The speed of Ca2ϩ-triggered exocytosis suggests that tuned to respond rapidly to changes in Ca2ϩ concentra- the fusion complex is preassembled prior to the arrival tion. The C2A kinetics data are summarized in the model of the Ca2ϩ signal. We therefore examined the interaction shown in Figure 4D. C2A first binds Ca2ϩ with diffusion- of Ca2ϩ–synaptotagmin with the fully assembled four helix 8 Ϫ1 Ϫ1 2ϩ limited kinetics (kon-Ca2ϩ Ն 10 M s ); the Ca -bound bundle that forms the core of the SNARE complex. These loops of C2A then penetrate into membranes at the experiments revealed that the Ca2ϩ–synaptotagmin– 10 Ϫ1 Ϫ1 collisional limit (kon ≈ 10 M s ). Thus, C2A responds vesicle complex assembles onto the base of the SNARE to Ca2ϩ as rapidly as possible. How does this speed complex, within the membrane-proximal H3 domain of compare with electrophysiological measurements of the syntaxin, adjacent to the lipid bilayer. A significant level 2ϩ kinetics of secretion? From kon-Ca2ϩ, it is apparent that a of binding was observed in the absence of Ca , which 2ϩ rise in [Ca ]i to 200 ␮M would “load” C2A in Ͻ50 ␮s. may serve to poise the synaptotagmin–SNARE complex This value lies within the minimal lag time for secretion for rapid responses to Ca2ϩ influx. Assembly of the Ca2ϩ– (Llinas et al., 1981; Heidelberger et al., 1994; Sabatini synaptotagmin–vesicle complex onto the base of the and Regehr, 1996). Furthermore, the collision-limited in- SNARE complex is an important finding, given current teraction of C2A–Ca2ϩ with membranes, coupled to the models of SNARE-mediated membrane fusion. Assem- close proximity of C2A and membranes in vivo, indicates bly of the H3 domain of syntaxin into the four helix that, to regulate exocytosis, this interaction would also bundle is an essential step in exocytosis (Littleton et al., occur on an extremely rapid timescale (microsecond to 1998) and may in fact mediate membrane fusion (Weber submicrosecond), as described below. et al., 1998; Chen et al., 1999). Certainly, assembly of Our kinetics data revealed that the dissociation con- the four helix bundle would result in very close juxtaposi- stant for the C2A domain–vesicle complex is 40 nM, tioning of the vesicle and target membranes, providing kcal to the fusion reaction an ideal platform for interactions with an exocytotic Ca2ϩ 10ف potentially providing (Equation 4). This high affinity, however, does not come sensor. These structural data also provide biochemical at the expense of speed; the complex releases Ca2ϩ support for the findings of Neher and colleagues, who and disassembles, upon addition of Ca2ϩ chelator, with recently reported that cleavage of SNAP-25 by botuli- extraordinarily rapid kinetics (Figure 6B). This speed of num neurotoxin A slows the exocytotic burst at high response is unique to the C2A domain of synaptotagmin [Ca2ϩ] and suggested that this effect could be due to I as compared with other C2 domains (Figure 6). For the diminished ability of the Ca2ϩ sensor to interact with example, addition of Ca2ϩ chelators results in synapto- the C-terminal base of the SNARE complex (Xu et al., tagmin–vesicle disassembly rates that are three orders 1998). 2ϩ of magnitude faster than those measured for cPLA2 (see Since the Ca -binding loops of C2A embed into the also Nalefski et al., 1997). These data demonstrate that lipid bilayer, a distinct region of synaptotagmin must the C2A domain of synaptotagmin is highly specialized interact with the core of the SNARE complex. This region Synaptotagmin Dynamics and Effector Interactions 373

has not been identified but is likely to undergo a small Indeed, a recent study indicates that assembly of pro- conformational change, upon binding of Ca2ϩ to the ductive ternary SNARE complexes occurs between rises Ca2ϩ- and membrane-binding loops of C2A, that in- in Ca2ϩ and fusion (Chen et al., 1999). creases synaptotagmin’s affinity for t-SNAREs. This The function of rapid Ca2ϩ-triggered synaptotagmin conformational change may not be apparent in the ab- clustering remains to be established. As noted above, sence of effector molecules (Shao et al., 1998). Prelimi- mutations in C2B strongly affect the Ca2ϩ dependence nary NMR data indicate that the C-terminal fragment of of secretion (Littleton et al., 1994), and C2B-mediated SNAP-25 (residues 120–206) interacts with a groove on heterooligomerization may modulate synaptic transmis- the “side” of the C2A domain of synaptotagmin (Y. K. sion (Littleton et al., 1999). Thus, C2B appears to play Chae, E. R. C., and J. L. Markley, unpublished data). an important role in excitation–secretion coupling. Here, However, because NMR studies make use of very high- we demonstrate that Ca2ϩ-triggered C2B domain– protein concentrations, interpretation of these data mediated synaptotagmin oligomerization occurs on the must be made with caution. For example, in bead-bind- submillisecond timescale, even at relatively low-protein ing assays, we did not detect binding of either of the concentrations (Figure 6C). Thus, both C2 domains of isolated “strands” of SNAP-25 with C2A (data not synaptotagmin are ideally suited to regulate synaptic shown). What is striking, however, is that the groove vesicle exo- and/or endocytosis in response to rapid 2ϩ that is on the side of C2A (and which is also predicted changes in [Ca ]i. It is tempting to speculate that rapid to be present in C2B; Sutton et al., 1995) matches pre- synaptotagmin clustering drives assembly of the SNARE cisely the diameter of the four helix bundle of the SNARE complex into a “fusion pore” and/or regulates the Ca2ϩ- complex (Sutton et al., 1998) onto which synaptotagmin triggered opening or the Ca2ϩ-dependent expansion of assembles in response to Ca2ϩ (Figure 7). These findings the pore (Fernandez-Chacon and Alvarez de Toledo, prompt a preliminary model in which synaptotagmin as- 1995). In this light, it will be of interest to determine sembles onto the core of the SNARE complex by “wrap- whether synaptotagmin oligomerization can cluster ping” around the four helix bundle. This mode of binding SNARE complexes or alter SNARE dynamics. would allow for sequential penetration of loops 3 and 1 of the C2 domains of synaptotagmin into lipid bilayers Experimental Procedures (e.g., PS [Chapman and Davis, 1998; Figure 5A] or PIP2 [Schiavo et al., 1996]). Recombinant Proteins The simultaneous interaction of synaptotagmin with cDNA encoding rat synaptotagmin I (Perin et al., 1990), rat synapto- membranes and with the membrane-proximal region tagmin III (Mizuta et al., 1994), and rat syntaxin 1A (Bennett et al., of the SNARE complex prompts molecular models 1992) were kindly provided by T. C. Su¨ dhof (Dallas, TX), Susumu Seino (Chiba, Japan), and R. Scheller (Stanford, CA), respectively. 2ϩ by which Ca –synaptotagmin could trigger exocytotic cDNA encoding a second form of rat synaptotagmin I was provided membrane fusion. For example, the rapid and high-affin- by G. Schiavo (London, UK; Osborne et al., 1999). An expression 2ϩ ity Ca -triggered penetration of synaptotagmin into vector (minipRSET) to produce a His6-tagged Trp reporter version membranes could increase membrane tension by pull- of the C2 domain of cPLA2 (W71F and Y96W) was kindly provided ing the SNARE complex toward the bilayer or by rotating by Roger Williams (Cambridge, UK); a pGEX expression vector to the four helix bundle to pull the vesicle and target mem- produce the C2 domain of C␤ was also kindly pro- vided by T. C. Su¨ dhof (Dallas, TX; Shao et al., 1996), and a pGEX branes toward one another. Alternatively, synaptotag- expression vector to produce the C2A domain of mouse synaptotag- min–membrane interactions could function in parallel min IV was provided by M. Fukuda (Ibaraki, Japan; Fukuda et al., with the SNARE complex to directly mediate lipid re- 1995). arrangements, at the base of the SNARE complex, that The cytoplasmic (residues 96–421) and C2A (residues 96–265) serve to accelerate SNARE-mediated membrane fusion. domains of synaptotagmin I were prepared as described (Chapman It is unclear whether the SNARE complex can fully et al., 1995, 1996). A series of truncation (96–381, 96–359, 96–337, 96–315, 96–293) and point mutants (D230,232N) were generated by “zipper” together prior to fusion. Scale molecular mod- PCR, subcloned into pGEX-2T, and expressed, purified, and cleaved els (Sutton et al., 1998) reveal how close together com- from the GST fusion moiety with thrombin, as described (Chapman plete assembly would bring the vesicle and target mem- et al., 1996). Synaptotagmin refers to the recombinant cytoplasmic branes and suggest that final “zippering” may not occur domain of synaptotagmin I, unless otherwise indicated. A series of until during or after fusion. These observations prompt syntaxin truncation (1–265, 1–193, 1–177; 160–288, 180–288, 194– an additional model in which the Ca2ϩ-triggered associa- 288, 230–288) and a deletion mutant (⌬250–265) were prepared in the same manner as GST fusion proteins. Soluble syntaxin fragment tion of synaptotagmin with partially assembled SNARE 1–177 was generated by thrombin cleavage of GST–1–177. Full- complexes could drive final zippering together of the length (1–288) and the 180–288 syntaxin fragment were also gener- base of the complex to accelerate membrane fusion ated as N-tagged His6 fusion proteins by subcloning in pTrcHis (In 2ϩ (Weber et al., 1998) in response to Ca influx. This Vitrogen). His6-tagged proteins were purified as described (Chap- could occur by “ordering” the t-SNAREs into helices man et al., 1996). A preliminary version of C2A-F234W was described previously to facilitate rapid assembly of the C-terminal region of (Chapman and Davis, 1998) and for this study was modified to synaptobrevin into the complex. This model is sup- extend from residue 96–265 to encompass the entire C2A domain of ported by the findings that synaptotagmin binds equally synaptotagmin I. This construct harbors a second mutation, W259F, well to isolated and assembled t-SNAREs. If the synap- such that the Trp reporter is the sole Trp residue. C2A-M173W was totagmin-binding site is the same in both cases, the prepared in the same manner in the W259F background. cPLA2 t-SNAREs in the synaptotagmin-Ca2ϩ-t-SNARE complex (W71F and Y96W) was purified according to Perisic et al. (1998), with minor modifications. The C2A domain of synaptotagmin III must adopt a conformation resembling the assembled (amino acids 291–423) was generated with the Trp reporter muta- ternary SNARE complex. It follows that synaptobrevin tions (F390W, W417F) via PCR and subcloned into pGEX-4T as could rapidly zipper together with ordered t-SNAREs. described (Chapman and Davis, 1998). GST fusion proteins were Neuron 374

purified and cleaved from the GST moiety as described (Chapman coimmunoprecipitation experiments. Immunoprecipitation of re- and Davis, 1998). All constructs were confirmed by DNA sequencing. combinant SNAREs and SNARE complexes was carried out as de- “Mini complex” refers to the minimal core of the synaptic SNARE scribed (Chapman et al., 1995). Briefly, recombinant SNAREs or complex (Fasshauer et al., 1998; Sutton et al., 1998) and is com- SNARE complexes were incubated with the indicated concentration posed of residues 180–262 of syntaxin 1A, 1–80 and 120–206 of of recombinant synaptotagmin in Tris-buffered saline (TBS; 20 mM SNAP-25A, and 1–96 of synaptobrevin II. Components were ex- Tris [pH 7.4], 150 mM NaCl) plus 0.5% Triton X-100 in the presence pressed, purified, and assembled as described (Fasshauer et al., of 2 mM EGTA or 1 mM Ca2ϩ for 2 hr. Syntaxin and synaptobrevin 1998). were immunoprecipitated by incubating the samples with HPC-1 (5 ␮l) or 69.1 (1.5 ␮l) ascites, respectively, for 2 hr, and 12 ␮l protein G-Sepharose fast-flow (Pharmacia) for 1 hr. The immunoprecipitates Bead-Binding Assays were washed three times and analyzed by SDS–PAGE and staining Binding of native proteins to immobilized recombinant proteins was with Coomassie blue. As a control for nonspecific precipitation of carried out as described (Chapman et al., 1995). Triton X-100 deter- synaptotagmin, samples were also prepared lacking SNAREs. In gent extracts (1 ml at 1 mg/ml) were incubated with the indicated each case, the immunoprecipitating antibodies did not bind syn- immobilized fusion protein (10–30 ␮g, as indicated) in the indicated aptotagmin, and, under the conditions of the binding assays, syn- volume in the presence of 2 mM EGTA or 1 mM Ca2ϩ for 2 hr. Beads aptotagmin did not precipitate in the absence of SNAREs. Thus, were washed three times in binding buffer, boiled in SDS sample for the experiments shown in the Figures, (ϪSNARE) samples also buffer, and analyzed by SDS–PAGE and immunoblotting. Immunore- lacked immunoprecipitating antibodies. active bands were visualized with HRP-conjugated secondary anti- bodies and enhanced chemiluminescence. Liposome-Binding Assays PS and PC were from Avanti Polar Lipids, 3[H]-PC was from Phar- Fluorescence Measurements and Miscellaneous Procedures macia/Amersham. 3[H]-labeled 25% PS/75% PC liposome-binding Steady-state fluorescence measurements were made using a PTI assays were carried out as described (Chapman and Davis, 1998), QMϪ1 fluorometer. Stopped-flow rapid mixing experiments were with modifications. Samples were incubated in Micro Bio-Spin chro- carried out with an Applied Photophysics SX.18MV stopped-flow matography columns (BioRad) for 15 min at room temperature. Col- spectrometer. For kinetics experiments using liposomes, fluores- umns were loaded onto a manifold and rapidly washed three times cence was monitored with a 326.1 nm band-pass filter to reduce with binding buffer plus 2 mM EGTA or 1 mM Ca2ϩ, and binding scattered light. In the absence of liposomes, Trp fluorescence was was quantified by liquid scintillation counting. collected with a 335 nm cutoff filter. For FRET measurements, dansyl fluorescence was collected with a 523 nm band-pass filter. All fluo- Calculations rescence experiments were carried out using HEPES-buffered saline The on- (k ) and off- (k ) rates of C2A–Ca2ϩ interactions with vesi- (HBS; 50 mM HEPES [pH 7.4], 100 mM NaCl) at room temperature, on off cles were calculated, assuming pseudo first-order kinetics, ac- unless otherwise indicated. cording to Equation 1: k ϭ [vesicle]k ϩ k . Collision constants The rate of vesicle–protein association reactions is dependent on obs on off were calculated according to Equation 2: k ϭ 4␲N (D ϩ D )(r ϩ the diameter of the vesicles (Lu et al., 1995). Therefore, uniform 100 coll A v p v r )/1000, where N is Avagadro’s number, D and D are the diffusion nm unilamellar vesicles were prepared as described (Chapman and p A v p coefficients of the vesicle and protein, and r and r are the radii of Davis, 1998) and used in all kinetics experiments. [Vesicle] was v p the vesicle and protein. Diffusion coefficients were calculated with confirmed by measuring phosphate content (Chapman and Davis, the Stokes-Einstein equation (Equation 3): D ϭ kT/6␲␩r, where k is 1998). Boltzman’s constant, T is temperature, ␩ is viscosity, and r is the For Ca2ϩ titration experiments, [Ca2ϩ] was determined with a free radius. Free energy was calculated according to Equation 4: ⌬G ϭ Microelectrode MI-600 Ca2ϩ electrode and an MI-402 microrefer- ϪRTlnk /k , where R is the gas constant, and T is temperature. ence electrode (Bedford, NH), and World Precision Instruments (Sar- on off asota, FL) Ca2ϩ standards (pCa2ϩ range of 1–8). Ca2ϩ concentrations below 100 ␮M were buffered with 2 mM EGTA. Acknowledgments

We thank Radhika Desai, Valentin Robu, and Young Kee Chae for Light Scattering assistance with experiments; Reinhard Jahn and Silke Engers for There are two reported synaptotagmin I sequences: one with an generous gifts of antibodies; Meyer Jackson, Jeff Walker, and Ro- aspartate at position 374 (D374; Perin et al., 1990; confirmed by berto Coronado for stimulating discussions; and Troy Littleton for J. T. L. and E. R. C.), and another with a glycine at this position discussions and critical comments on this manuscript. We also (G374; Osborne et al., 1999). With the exception of the D374 version thank Richard Scheller, Roger Williams, Susumu Seino, Mitsunori of synaptotagmin I, all synaptotagmin isoforms, from all species, Fukuda, Giampietro Schiavo, and Thomas Su¨ dhof for providing 2ϩ contain a glycine at this position. This glycine is essential for Ca - cDNA and/or expression constructs, and David Gaston for molecular triggered oligomerization; the G374 but not the D374 version of modeling. This study was supported by grants from the National 2ϩ synaptotagmin clusters in response to Ca (R. D. and E. R. C., Institutes of Health (GM 56827–01), the American Heart Association unpublished data). We note that a copy of the D374 version can serve (9750326N), and the Milwaukee Foundation. as an acceptor for Ca2ϩ-triggered binding of G374 synaptotagmin (Chapman et al., 1996; Sugita et al., 1996). The basis for this se- Received May 4, 1999; revised August 10, 1999. quence variability is under investigation. For light-scattering experiments, oligomerization-competent (i.e., G374) recombinant synaptotagmin I (Osborne et al., 1999) was pre- References pared in HBS with 0.1 mM EGTA to circumvent Ca2ϩ-induced aggre- gation and rapidly mixed with 1 mM Ca2ϩ (final concentration) in the Bennett, M.K., Calakos, N., and Scheller, R.H. (1992). Syntaxin: a stopped-flow instrument. Samples were illuminated with 335 nm synaptic protein implicated in docking of synaptic vesicles at pre- light, and scattered light was collected through a WG335 filter. synaptic active zones. Science 257, 255–259. Betz, A., Okamoto, M., Benseler, F., and Brose, N. (1997). Direct interaction of the rat unc-13 homologue munc13–1 with the N termi- Immunoprecipitation and Monoclonal Antibodies nus of syntaxin. J. Biol. Chem. 272, 2520–2526. Mouse monoclonal antibodies directed against synaptotagmin I (604.4 and 41.1), syntaxin (HPC-1), synaptobrevin (69.1), synapto- Breckenridge, W.C., Morgan, I.G., Zanetta, J.P., and Vincendon, G. physin (7.2), SNAP-25 (71.2), and ␣/␤-SNAP (77.1) were kindly pro- (1972). The lipid composition of adult rat brain synaptotsomal vided by S. Engers and R. Jahn (Gottingen, Germany). Monoclonal plasma membranes. Biochim. Biophys. Acta 266, 695–707. antibody 77.1 recognizes ␣- and ␤-SNAP equally well. Breckenridge, W.C., Morgan, I.G., Zanetta, J.P., and Vincendon, The nonoligomerizing D374 form of synaptotagmin I (described G. (1973). Adult rat brain synaptic vesicles. II. Lipid composition. in detail under “Light Scattering,” above) was employed in all SNARE Biochim. Biophys. Acta 320, 681–686. Synaptotagmin Dynamics and Effector Interactions 375

Brose, N., Petrenko, A.G., Su¨ dhof, T.C., and Jahn, R. (1992). Synap- Calcium dependence of the rate of exocytosis in a synaptic terminal. totagmin: a calcium sensor on the synaptic vesicle surface. Science Nature 371, 513–515. 256, 1021–1025. Jorgensen, E.M., Hartwieg, E., Schuske, K., Nonet, M.L., Jin, Y., Chapman, E.R., and Davis, A.F. (1998). Direct interaction of a Ca2ϩ and Horvitz, H.R. (1995). Defective recycling of synaptic vesicles binding loop of synaptotagmin with lipid bilayers. J. Biol. Chem. in synaptotagmin mutants of Caenorhabditis elegans. Nature 378, 273, 13995–14001. 196–199. Chapman, E.R., An, S., Barton, N.J., and Jahn, R. (1994). SNAP-25: Kee, Y., and Scheller, R.H. (1996). Localization of synaptotagmin- a t-SNARE which binds to both syntaxin and synaptobrevin via binding domains on syntaxin. J. Neurosci. 16, 1975–1981. domains that may form coiled coils. J. Biol. Chem. 269, 27427– Kraulis, P.J. (1991). MOLSCRIPT: a program to produce both de- 27432. tailed and schematic plots of protein structures. J. Appl. Crystal. Chapman, E.R., Hanson, P.I., An, S., and Jahn, R. (1995). Ca2ϩ regu- 24, 946–950. lates the interaction between synaptotagmin and syntaxin 1. J. Biol. Li, C., Ullrich, B., Zhang, J.Z., Anderson, R.G.W., Brose, N., and Chem. 270, 23667–23671. Su¨ dhof, T.C. (1995). Ca2ϩ-dependent and independent activities of Chapman, E.R., An, S., Edwardson, J.M., and Jahn, R. (1996). A neural and non-neural synaptotagmins. Nature 375, 594–599. novel function for the second C2 domain of synaptotagmin: Ca2ϩ- Li, C., Davletov, B.A., and Su¨ dhof, T.C. (1995). Distinct Ca2ϩ and triggered dimerization. J. Biol. Chem. 271, 5844–5849. Sr2ϩ binding properties of synaptotagmins. J. Biol. Chem. 270, Chapman, E.R., Desai, R., Davis, A.F., and Tornhel, C. (1998). Delin- 24898–24902. eation of the oligomerization, AP-2- and synprint-binding region of Littleton, J.T., Stern, M., Perin, M., and Bellen, H.J. (1994). Calcium the C2B-domain of synaptotagmin. J. Biol. Chem. 273, 32966– dependence of neurotransmitter release and rate of spontaneous 32972. vesicle fusions are altered in Drosophila synaptotagmin mutants. Chen, Y.A., Scales, S.J., Patel, S.M., Doung, Y.C., and Scheller, R.H. Proc. Natl. Acad. Sci. USA 91, 10888–10892. (1999). SNARE complex formation is triggered by Ca2ϩ and drives Littleton, J.T., Chapman, E.R., Kreber, R., Garment, M.B., Carlson, membrane fusion. Cell 97, 165–174. S.D., and Ganetzky, B. (1998). Paralytic mutations demonstrate that Clark, J.D., Lin, L.L., Kriz, R.W., Ramesha, C.S., Sultzman, L.A., assembly and disassembly of the SNARE-complex are essential Lin, A.Y., Milona, N., and Knopf, J.L. (1991). A novel arachodonic steps in neurotransmitter release. Neuron 21, 401–413. 2ϩ acid–selective cytosolic PLA2 contains a Ca -dependent transloca- Littleton, J.T., Sereno, T.L., Rubin, G.M., Ganetzky, B., and Chap- tion domain with homology to PKC and GAP. Cell 65, 1043–1051. man, E.R. (1999). Synaptic function modulated by changes in the Davletov, B.A., and Su¨ dhof, T.C. (1993). A single C2 domain from ratio of synaptotagmin I and IV. Nature, 400, 757–760. synaptotagmin I is sufficient for high affinity Ca2ϩ/phospholipid bind- Llinas, R., Steinberg, I.Z., and Walton, K. (1981). Relationship be- ing. J. Biol. Chem. 268, 26386–26390. tween presynaptic calcium current and postsynaptic potential in DiAntonio, A., and Schwarz, T.L. (1994). The effects on synaptic squid giant synapse. Biophys. J. 33, 323–352. physiology of synaptotagmin mutations in Drosophila. Neuron 12, Lu, Y., Bazzi, M.D., and Nelsestuen, G.L. (1995). Kinetics of 909. VI, calcium and phospholipid association and dissociation. Bio- chemistry 34, 10777–10785. Falke, J., Drake, S.K., Hazard, A.L., and Peersen, O.B. (1994). Molec- ular tuning of ion binding to calcium signaling proteins. Q. Rev. Mizuta, M., Inagaki, N., Nemoto, Y., Matsukura, S., Takahashi, M., Biophys. 27, 219–290. and Seino, S. (1994). Synaptotagmin III is a novel isoform of rat synaptotagmin expressed in endocrine and neuronal cells. J. Biol. Fasshauer, D., Otto, H., Eliason, W.K., Jahn, R., and Bru¨ nger, A.T. Chem. 269, 11675–11678. (1997). Structural changes are associated with soluble N-ethylmalei- mide-sensitive fusion protein attachment receptor complex forma- Nalefski, E.A., and Falke, J.J. (1998). Location of the membrane- 2ϩ tion. J. Biol. Chem. 272, 28036–28041. docking face on the Ca -activated C2-domain of cytosolic phos- pholipase A2. Biochemistry 37, 17642–17650. Fasshauer, D., Eliason, W.K., Brunger, A.T., and Jahn, R. (1998). 2ϩ Identification of a minimal core of the synaptic SNARE complex Nalefski, E.A., Slazas, M.M., and Falke, J.J. (1997). Ca -signaling sufficient for reversible assembly and disassembly. Biochemistry cycle of a membrane-docking C2 domain. Biochemistry 36, 12011– 12018. 37, 10345–10353. Naren, A.P., Quick, M.W., Collawn, J.F., Nelson, D.J., and Kirk, K.L. Fernandez, I., Ubach, J., Dulubova, I., Zhang, X., Su¨ dhof, T.C., and (1998). Syntaxin 1A inhibits CFTR chloride channels by means of Rizo, J. (1998). Three-dimensional structure of an evolutionarily con- domain-specific protein–protein interactions. Proc. Natl. Acad. Sci. served N-terminal domain of syntaxin 1A. Cell 94, 841–849. USA 95, 10972–10977. Fernandez-Chacon, R., and Alvarez de Toledo, G. (1995). Cytosolic Nonet, M.L., Grundahl, K., Meyer, B.J., and Rand, J.B. (1993). Synap- calcium facilitates release of secretory products after exocytotic tic function is impaired but not eliminated in C. elegans mutants vesicle fusion. FEBS Lett. 363, 221–225. lacking synaptotagmin. Cell 73, 1291–1305. Fujita, Y., Shirataki, H., Sakisaka, T., Asakura, T., Ohya, T., Kotani, Osborne, S.L., Herreros, J., Bastiaens, P.I.H., and Schiavo, G. (1999). H., Yokoyama, S., Nishioka, H., Matsuura, Y., Mizoguchi, A., et al. Calcium-dependent oligomerization of synaptotagmins I and II. J. (1998). Tomosyn: a syntaxin-1-binding protein that forms a novel Biol. Chem. 274, 59–66. complex in the neurotransmitter release process. Neuron 20, Perin, M., Fried, V.A., Mignery, G.A., Jahn, R., and Su¨ dhof, T.C. 905–915. (1990). Phospholipid binding by a synaptic vesicle protein homolo- Fukuda, M., Kojima, T., Aruga, J., Niinobe, M., and Mikoshiba, K. gous to the regulatory region of protein kinase C. Nature 345, (1995). Functional diversity of C2 domains of synaptotagmin family. 260–263. J. Biol. Chem. 270, 26523–26527. Perisic, O., Fong, S., Lynch, D.E., Bycroft, M., and Williams, R.L. Geppert, M., Goda, Y., Hammer, R.E., Li, C., Rosahl, T.W., Stevens, (1998). Crystal structure of a calcium-phospholipid binding domain 2ϩ C.F., and Su¨ dhof, T.C. (1994). Synaptotagmin I: a major Ca sensor from cytosolic phospholipase A2. J. Biol. Chem. 273, 1596–1604. for transmitter release at a central synapse. Cell 79, 717–727. Reist, N.E., Buchanan, J., Li, J., DiAntonio, A., Buxton, E., and Hanson, P.I., Roth, R., Morisaki, H., Jahn, R., and Heuser, J.E. (1997). Schwarz, T.L. (1998). Morphologically docked synaptic vesicles are Structure and conformational changes in NSF and its membrane reduced in synaptotagmin mutants of Drosophila. J. Neurosci. 18, receptor complexes visualized by quick-freeze/deep-etch electron 7662–7673. microscopy. Cell 90, 523–535. Sabatini, B.L., and Regehr, W.G. (1996). Timing of Hayashi, T., McMahon, H., Yamasaki, S., Binz, T., Hata, Y., Su¨ dhof, at fast synapses in the mammalian brain. Nature 384, 170–172. T.C., and Niemann, H. (1994). Synaptic vesicle membrane fusion Schiavo, G., Gu, Q.M., Prestwich, G.D., So¨ llner, T.H., and Rothman, complex: action of clostridial neurotoxins on assembly. EMBO J. J.E. (1996). Calcium-dependent switching of the specificity of phos- 13, 5051–5061. phoinositide binding to synaptotagmin. Proc. Natl. Acad. Sci. USA Heidelberger, R., Heinemann, C., Neher, E., and Matthews, G. (1994). 93, 13327–13332. Neuron 376

Schiavo, G., Stenbeck, G., Rothman, J.E., and So¨ llner, T.H. (1997). Binding of the synaptic vesicle v-SNARE, synaptotagmin, to the plasma membrane t-SNARE, SNAP-25, can explain docked vesicles at neurotoxin-treated synapses. Proc. Natl. Acad. Sci. USA 94, 997– 1001. Shao, X., Davletov, B.A., Sutton, R.B., Su¨ dhof, T.C., and Rizo, J. (1996). Bipartite Ca2ϩ binding motif in C2 domains of synaptotagmin and protein kinase C. Science 273, 248–251. Shao, X., Li, C., Fernandez, I., Zhang, X., Su¨ dhof, T.C., and Rizo, J. (1997). Synaptotagmin–syntaxin interaction: the C2 domain as a Ca2ϩ-dependent electrostatic switch. Neuron 18, 133–142. Shao, X., Fernandez, I., Su¨ dhof, T.C., and Rizo, J. (1998). Solution structures of the Ca2ϩ-free and Ca2ϩ-bound C2A domain of synapto- tagmin: does Ca2ϩ induce a conformational change? Biochemistry 37, 16106–16115. So¨ llner, T., Bennett, M.K., Whiteheart, S.W., Scheller, R.H., and Roth- man, J.E. (1993). A protein assembly–disassembly pathway in vitro that may correspond to sequential steps of synaptic vesicle docking, activation, and fusion. Cell 75, 409–418. Sugita, S., Hata, Y., and Su¨ dhof, T.C. (1996). Distinct Ca2ϩ- dependent properties of the first and second C2-domains of synap- totagmin I. J. Biol. Chem. 271, 1262–1265. Sutton, B., and Sprang, S.R. (1998). Structure of the protein kinase C–beta phospholipid-binding C2 domain complexed with Ca2ϩ. Structure 6, 1395–1405. Sutton, R.B., Davletov, B.A., Berghuis, A.M., Su¨ dhof, T.C., and

Sprang, S.R. (1995). Structure of the first C2 domain of synaptotag- min I: a novel Ca2ϩ/phospholipid-binding fold. Cell 80, 929–938. Sutton, R.B., Fasshauer, D., Jahn, R., and Brunger, A.T. (1998). Crys- tal structure of a SNARE complex involved in synaptic exocytosis at 2.4 A˚ resolution. Nature 395, 347–353. von Poser, C., Ichtchenko, K., Shao, X., Rizo, J., and Su¨ dhof, T.C. (1997). The evolutionary pressure to inactivate a subclass of synap- totagmins with an amino acid substitution that abolishes Ca2ϩ bind- ing. J. Biol. Chem. 272, 14314–14319. Weber, T., Zemelman, B.V., McNew, J.A., Westermann, B., Gmachl, M., Parlati, F., So¨ llner, T.H., and Rothman, J.E. (1998). SNAREpins: minimal machinery for membrane fusion. Cell 92, 759–772. Xu, T., Binz, T., Niemann, H., and Neher, E. (1998). Multiple kinetic components of exocytosis distinguished by neurotoxin sensitivity. Nat. Neurosci. 1, 192–200. Zhang, J.Z., Davletov, B.A., Su¨ dhof, T.C., and Anderson, R.G.W. (1994). Synaptotagmin I is a high affinity receptor for clathrin AP-2: implications for membrane recycling. Cell 78, 751–760.