Calcium-dependent Oligomerization of Synaptotagmins I and II

Synaptotagmins constitute a large family of membrane proteins characterized by their distinct distributions and different biochemical features. Genetic evidence suggests that members of this protein family are likely to function as calcium sensors in calcium-regulated events in neurons, although the precise molecular mechanism remains ill defined. Here we demonstrate that different synaptotagmin isoforms (Syt I, II, and IV) are present in the same synaptic vesicle population from rat brain cortex. In addition, Syt I and II co-localize on the same small synaptic vesicle (SSV), and they heterodimerize in the presence of calcium with a concentration dependence resembling that of the starting phase of SSV exocytosis (EC50 = 6 ± 4 μm). The association between Syt I and Syt II was demonstrated by immunoprecipitation of the native proteins and the recombinant cytoplasmic domains and by using fluorescence resonance energy transfer (FRET). Although a subpopulation of SSV containing Syt I and IV can be isolated, these two isoforms do not show a calcium-dependent interaction. These results suggest that the self-association of synaptotagmins with different calcium binding features may create a variety of calcium sensors characterized by distinct calcium sensitivities. This combinatorial hypothesis predicts that the probability of a single SSV exocytic event is determined, in addition to the gating properties of the presynaptic calcium channels, by the repertoire and relative abundance of distinct synaptotagmin isoforms present on the SSV surface.

Synaptotagmins constitute a large family of membrane proteins characterized by their distinct distributions and different biochemical features. Genetic evidence suggests that members of this protein family are likely to function as calcium sensors in calcium-regulated events in neurons, although the precise molecular mechanism remains ill defined. Here we demonstrate that different synaptotagmin isoforms (Syt I, II, and IV) are present in the same synaptic vesicle population from rat brain cortex. In addition, Syt I and II co-localize on the same small synaptic vesicle (SSV), and they heterodimerize in the presence of calcium with a concentration dependence resembling that of the starting phase of SSV exocytosis (EC 50 ‫؍‬ 6 ؎ 4 M). The association between Syt I and Syt II was demonstrated by immunoprecipitation of the native proteins and the recombinant cytoplasmic domains and by using fluorescence resonance energy transfer (FRET). Although a subpopulation of SSV containing Syt I and IV can be isolated, these two isoforms do not show a calcium-dependent interaction. These results suggest that the self-association of synaptotagmins with different calcium binding features may create a variety of calcium sensors characterized by distinct calcium sensitivities. This combinatorial hypothesis predicts that the probability of a single SSV exocytic event is determined, in addition to the gating properties of the presynaptic calcium channels, by the repertoire and relative abundance of distinct synaptotagmin isoforms present on the SSV surface.
Neuronal communication depends upon the transduction of an electric nerve impulse for the release of neurotransmitters from their storage compartment, the small synaptic vesicles (SSV). 1 This tightly regulated process is triggered by the rapid increase of the intracellular calcium due to the opening of voltage-gated calcium channels at the active zones of the synaptic plasma membrane, which causes the activation of the SSV fusion machinery (1,2). Of the proteins involved in the physiological cycle of the SSV, the members of the synaptotagmin family represent the best candidates for the role of calcium sensors in neurotransmitter release and, more generally, in regulated exocytosis (3)(4)(5). This protein family comprises more than a dozen members having the same overall structure and a variable degree of homology (4,5). Synaptotagmins are integral membrane proteins with broad distributions in neuronal and non-neuronal tissues that vary between isoforms (4,5). In the central nervous system and neuroendocrine cells, synaptotagmins are localized on SSV and secretory granules (6). These proteins are characterized by a single membrane-spanning domain and by a large cytoplasmic portion containing two internal repeats that have homology with the C2 domain of protein kinase C (7,8). This domain is known to regulate the calciumdependent translocation of protein kinase C to membranes and is present in many proteins with different functions (5,8). In synaptotagmins, the two C2 domains are responsible for different interactions, both calcium-dependent and calcium-independent. The first C2 homology domain of synaptotagmin, termed C2A, whose structure has been recently solved (9 -12), binds calcium and acidic phospholipids (in particular, phosphatidylserine) in a ternary complex in a calcium-dependent manner (EC 50 ϭ 3-6 M Ca 2ϩ ) (13)(14)(15)(16). The C2A also binds syntaxin, a presynaptic t-SNARE (17) at relatively high calcium concentrations (EC 50 Ͼ 200 M Ca 2ϩ ) (11,18,19). The second C2 domain (C2B) of synaptotagmin I (Syt I) has recently been shown to interact with another class of acidic phospholipids, the phosphoinositides. The specificity of this interaction is strictly dependent on calcium as illustrated by the maximal binding of phosphatidylinositol 4,5-diphosphate at calcium concentrations between 20 and 100 M (20). The C2B domain also shows calcium-independent interactions. Among these, its binding to the soluble phosphoinositide analogues, inositol polyphosphates, may be important in modulating the interaction of synaptotagmin with phosphoinositides (21)(22)(23). Other interactions independent of calcium are with the clathrin adaptor AP-2 (24) and with ␤-SNAP, the neuronal isoform of a family of cytosolic proteins termed soluble N-ethylmaleimidesensitive factor (NSF) attachment proteins (SNAPs) (25). Synaptotagmin is also able to bind the other synaptic t-SNARE, SNAP-25. While the interaction with syntaxin is totally calciumdependent, SNAP-25 binding to synaptotagmin is only partially modulated by calcium (26 -28). Synaptotagmin also interacts with voltage-gated calcium channels (29 -38). These findings, together with functional studies (39 -41) and genetic evidence from Caenorhabditis elegans, Drosophila, and mouse (42)(43)(44)(45) strongly indicate that synaptotagmins are the main calcium sensor(s) for synchronous neurotransmitter release in the central nervous system.
Recently, another protein-protein interaction involving the C2B domain of synaptotagmin has been described. This interaction involves the calcium-dependent self-association of the C2B domain of Syt I and consequent synaptotagmin dimerization (46,47). This result has suggested a model in which, following calcium rise at the nerve terminal, synaptotagmin first dimerizes and then associates with syntaxin, creating a protein complex involved in the calcium-dependent step(s) leading to SSV fusion with the active zones of the synaptic plasma membrane (47).
In this paper, we present evidence for a broader distribution of some synaptotagmin isoforms than expected on the basis of the reported literature. We analyzed SSV preparations isolated from adult rat brain cortex for the presence of synaptotagmins I, II, and IV (Syt I, Syt II, and Syt IV) and found that the same population of SSV contains an overlapping distribution of these isoforms. In addition, both Syt I and II are present on the same SSV, and they heterodimerize efficiently with a concentration dependence for calcium resembling that of the starting phase of SSV exocytosis. This process was confirmed in vitro with the soluble cytoplasmic domains of Syt I and II using both immunoprecipitation and fluorescence resonance energy transfer (FRET) techniques. In contrast, the distribution of Syt IV on SSV only partially overlaps with that of Syt I and II, and the ability of Syt IV to associate with Syt I is low. Taken together, these results suggest that the self-association of synaptotagmins with different calcium binding features possesses a certain degree of specificity and may create a variety of calcium sensors characterized by distinct calcium sensitivities.

EXPERIMENTAL PROCEDURES
Antibody Production and Purification-Peptides MVSASHPEAL-AAPVTTVATC (corresponding to residues 1-19 of rat Syt I with an additional cysteine at the C terminus), CMRNIFKRNQEPIVAPAT (residues 1-17 of rat Syt II), and CMAPITTSRVEFDEIPT (residues 1-16 of rat Syt IV; both peptides with an additional cysteine at the N terminus) were conjugated to maleimide-activated keyhole limpet hemocyanin (Pierce) following the manufacturer's specifications. Polyclonal isoformspecific antibodies were raised in New Zealand White rabbits by intramuscular injection of 1 mg of the corresponding peptide together with Freund's adjuvant. The antibodies were affinity-purified by using Sulfolink resin (3 ml; Pierce) previously coupled with 3 mg of the specific peptide and eluted with 100 mM glycine, pH 2.5, 0.1% bovine serum albumin. This method yielded a satisfactory purification of anti-Syt I and -Syt II antibodies, but not of anti-Syt IV, that either proved sensitive to both acidic and basic elution protocols or did not elute from the peptide column (48). Anti-Syt-IV antibodies were partially purified by ammonium sulfate precipitation and protein A-Sepharose affinity chromatography. The specificity of the purified antibodies was evaluated by neutralization with the generating peptides and by comparison with the preimmune serum.
Synaptic Vesicle Purification and Immunoisolation-SSV from rat brain cortex were prepared following the method of Huttner et al. (49) with minor variations. Briefly, the membrane-enriched fraction (LP2) obtained by centrifugation of the lysate supernatant was resuspended as described and then loaded into a linear continuous sucrose gradient prepared from equal amounts of 50 and 800 mM sucrose (Life Technologies, Inc.) by using a model 106 Gradient Master (BioComp Instrument Inc., Fredericton, Canada). Samples were spun at 26,000 rpm (90,000 ϫ g av ) in a Beckman SW28 rotor for 4 h. The fraction corresponding to a sucrose concentration range between 200 and 400 mM and visible as a translucent band was loaded into a glycerol-coated controlled pore glass beads column (25 ϫ 1000 mm; CPG, Lincoln Park, NJ) equilibrated in 4 mM HEPES-NaOH, pH 7.4, 300 mM glycine, 0.04% NaN 3 . The column was eluted at 0.7 ml/min with the same buffer, and 8-ml fractions were collected. For the data presented in Fig. 1B, proteins corresponding to each fraction (400 l) were recovered by precipitation with trichloroacetic acid by using 0.5 g of sodium deoxycholate as carrier and then analyzed by SDS-PAGE. For immunoprecipitation and immunoisolation experiments, fractions corresponding to the second peak of absorbance at 280 nm were pooled and centrifuged at 45,200 rpm (198,000 ϫ g av ) in a 50.2 Ti rotor for 2 h. The SSV pellet was resuspended in 1 ml of 4 mM HEPES-NaOH, pH 7.4, 300 mM glycine, 0.04% NaN 3 , 100 M phenylmethylsulfonyl fluoride, 2 g/ml pepstatin, and the suspension was homogenized by forcing it eight times through a 25-gauge needle. Protein concentration was routinely determined with Bio-Rad protein assay reagent by using purified immunoglobulins as a reference.
For SSV immunoisolation purposes, M48 monoclonal antibodies against Syt I (50) were purified with DEAE blue resin (Bio-Rad) (51) and then dialyzed extensively against distilled water. The purified IgG fraction (0.4 mg) was coupled to Eupergit C1Z methacrylate microbeads (98.6 mg; 1-m diameter; Röhm Pharma, Darmstadt), previously washed with distilled water, by incubation at room temperature for 1 h (52). The reaction was stopped by centrifugation, followed by incubation with 1 M glycine, pH 6.5, for 12 h at room temperature to block the remaining reactive sites. Beads were subjected to six alternate pH washes (100 mM Tris-HCl, pH 8.0, 150 mM NaCl and 100 mM sodium acetate, pH 4.5, 150 mM NaCl) to remove loosely bound material. The beads were stored in 20 mM Tris-HCl, pH 7.2, 150 mM NaCl, 0.02% NaN 3 . In selected experiments, protein G-Sepharose fast flow beads (Amersham Pharmacia Biotech) previously conjugated with M48 monoclonal antibodies (50) were used. Immunoisolation was performed by using the purified SSV fraction obtained from the glycerol-coated controlled pore glass beads column as starting material. 10 l of 50% M48-conjugated and control beads were washed in glycine buffer (4 mM HEPES-NaOH, pH 7.4, 300 mM glycine, 0.04% NaN 3 ) and preincubated with 1 mg of asolectin (Sigma) to block any nonspecific lipid binding sites. They were washed again in glycine buffer and incubated with 10 g of SSV in the presence of 0.2 mg/ml ovalbumin, 5% glycerol at 4°C for 2 h. The reaction was stopped by centrifugation. The beads were washed three times in glycine buffer and resuspended in SDS-containing sample buffer, while the supernatant proteins were precipitated with trichloroacetic acid in the presence of 0.05% (w/v) sodium deoxycholate as carrier and resuspended as above. Proteins were analyzed by SDS-PAGE and Western blotting using anti-Syt II, anti-Syt IV, and anti-VAMP 2 (WAKO, Richmond, VA) antibodies. Immunoreactive bands were visualized with a rabbit anti-mouse IgG or a sheep antirabbit peroxidase conjugated IgG followed by ECL detection (Amersham Pharmacia Biotech).
Immunoprecipitation of Native Synaptotagmins-50 g of the column-purified pooled cortical SSV fraction and 100 g of an impure vesicular fraction (LP2) from rat cerebellum were solubilized in 25 mM HEPES-KOH, pH 7.6, 100 mM KCl, 1% glycerol containing 100 M phenylmethylsulfonyl fluoride, 2 g/ml pepstatin, and 4% octyl-␤-Dglucosopyranoside (OG) for 30 min at 4°C. Detergent-free buffer was added to give a final OG concentration of 1.2%, and the solutions were centrifuged to remove insoluble material. The solubilized proteins were incubated for 2 h at 4°C with protein G-agarose beads (Boehringer Mannheim) previously conjugated with M48 monoclonal antibodies (50) or anti-Syt I or anti-Syt II antibodies or with the protein G beads alone. The reactions were stopped by centrifugation. The proteins in the supernatant were precipitated with trichloroacetic acid and resuspended in 20 l of SDS-containing sample buffer, while the beads were washed three times in incubation buffer containing 0.8% OG and then prepared for SDS-PAGE. 8 l of each sample was analyzed by SDS-PAGE and Western blotting, using anti-Syt I, anti-Syt II or anti-Syt IV antibodies.
For the experiments investigating the calcium dependence, similar procedures were used, except that Ca 2ϩ /EGTA (final concentration 2 mM EGTA) buffers were added to the incubation buffer to yield the free calcium concentrations indicated. 1 mM MgCl 2 was also added, and the immunoprecipitation step was carried out in 2% OG. Immunoblots were detected either with ECL or by incubation with iodinated anti-rabbit secondary antibodies (17.6 Ci/g, 1 Ci/ml, Amersham Pharmacia Biotech). In the case of Western blotting detected with the ECL detection system, Syt II and Syt IV were quantified with the NIH Image (version 1.61) software by comparing the signal present in the immunoprecipitate with that obtained from an antigen standard curve prepared by loading increasing amounts of either Syt II or Syt IV (0.25, 0.5, 0.75, and 1 times the starting material) (53). Quantitation of Syt II with iodinated anti-rabbit secondary was performed by using a Molecular Dynamics PhosphorImager. Syt recovery in the immunoprecipitate was calculated as the percentage of total Syt II present in the sample. Both the sum of the immunoreactivity present in the supernatant (S) and in the pellet (P) and the total input of the sample (T) were used as denominators to calculate the percentage of Syt recovery, and the two values were compared. In all cases, the maximal S.D. between the measurements obtained with the two methods was less than 10, with an average value of 6. Both values were used to determine the EC 50 of the calcium dependence of Syt I/Syt II co-immunoprecipitation, and the variability between the two methods is taken into account in the error bars in Fig. 4. To compare different experiments, the percentage of Syt II present in the Syt I immunoprecipitate was expressed as a percentage of the maximal recovery (Fig. 4).
Expression and Immunoprecipitation of Recombinant Synaptotagmins-Recombinant glutathione S-transferase (GST)-synaptotagmin fusion proteins were prepared by inserting the DNA corresponding to residues 95-421 of rat Syt I and 103-422 of rat Syt II (GenBank TM accession numbers X52772 and M64488, respectively) into the EcoRI/ NcoI sites of the expression vector pGEX-KG (54). The proteins correspond to the published sequences (55,56), except for the substitution of Glu 188 for Asp, Gly 374 for Asp, and Ile 393 for Met in rat Syt I. These variations, generated by single base changes, may be ascribed to the different DNA sources used for the cloning of rat Syt I (Rattus norvegicus here versus Rattus rattus used in Ref. 56). For immunoprecipitation purposes, a tagged version of the rat Syt II GST fusion protein was prepared by inserting the sequence YPYDVPDYA (corresponding to the hemagglutinin epitope (HA)) immediately downstream of the thrombin cleavage site. GST fusion proteins were purified on glutathione-agarose beads (Sigma), and the cytoplasmic domains of Syt I and Syt II were released by thrombin cleavage (54). Proteins were purified by ion exchange chromatography on a Mono-Q matrix (Amersham Pharmacia Biotech), dialyzed against 20 mM HEPES-KOH, pH 7.6, 150 mM KCl, 10% glycerol, 0.1 mM dithiothreitol, and, after freezing in liquid nitrogen, stored at Ϫ80°C.
For immunoprecipitation experiments, Syt I and HA-tagged Syt II were thawed and immediately diluted in 25 mM HEPES-KOH, pH 7.5, KCl 100 mM, glycerol 1%, dithiothreitol 0.1 mM, OG 0.8% (final concentrations: Syt II 44 nM, Syt I 435 nM). The protein solution was precleared by centrifugation to eliminate any aggregated material and then mixed with the appropriate Ca 2ϩ /EGTA buffers to yield the free calcium concentrations indicated. In selected samples, MgCl 2 was added to reach a free Mg 2ϩ concentration of 0.5 mM. Samples were incubated for 1 h at 4°C and then immunoprecipitated by adding an excess of monoclonal anti-HA antibody 12CA5 (57) prebound to protein G-Sepharose fast flow (Amersham Pharmacia Biotech) for 1 h at 4°C. Beads were recovered by gentle centrifugation and washed extensively with incubation buffer containing the appropriate free calcium concentration and 0.5 mM MgCl 2 where indicated. Proteins bound to beads were then separated by SDS-PAGE and analyzed either by densitometric analysis of the Coomassie Blue-stained gels or by Western blotting. In both cases, Syt I recovery was quantified with the NIH Image software by comparing the signal present in the immunoprecipitate with that obtained from a standard curve prepared by loading increasing amounts of recombinant Syt I (50, 100, 150, 200, and 250 ng in the case of Western blotting and 0.2, 0.4, 0.6, 0.8, and 1 g for Coomassie staining) (53). Data were expressed as a percentage of the maximal Syt I present in the Syt II immunoprecipitate.
Similar results were obtained in experiments performed at 25°C and without detergent in the incubation buffer. As previously noticed (28), unspecific binding of Syt I to beads is totally abolished in the presence of nonionic detergent. For this reason, OG was included in the incubation buffer.
FRET Measurements-The cytoplasmic domain of Syt I and II were dialyzed extensively against 20 mM bicine, pH 8.5, 100 mM KCl and then labeled with N-succinimidyl-CY3 and N-succinimidyl-CY5 (Amersham Pharmacia Biotech) for 30 min at room temperature using a protein/dye ratio of 1:17 (58). The reaction was blocked by the addition of 100 mM glycine and incubation for 5 min at 4°C. Excess dye was removed on a PD10 gel filtration column (Amersham Pharmacia Biotech) pre-equilibrated in 20 mM HEPES-NaOH, pH 7.6, 100 mM KCl, and the yield of labeling was determined spectrophotometrically by using extinction coefficients of 31,270 M Ϫ1 cm Ϫ1 for Syt I and 28,830 M Ϫ1 cm Ϫ1 for Syt II, both at 280 nm, and extinction coefficients of 150,000 M Ϫ1 cm Ϫ1 at 554 nm for CY3 and 250,000 M Ϫ1 cm Ϫ1 at 650 nm for CY5.
Spectrophotometric experiments were performed by adding increasing amounts of calcium to a 50-l cuvette (Hellma, Jena) containing 0.1 M CY3-Syt II in 20 mM HEPES-NaOH, pH 7.6, 100 mM KCl, 2 mM EGTA, and either 0.5 M CY5-Syt I or unlabeled Syt I. In selected samples, 0.5 mM free Mg 2ϩ was also added to monitor the effect of divalent cations different from Ca 2ϩ on the equilibrium. Fluorescence emission spectra were recorded in a 710 PTI spectrofluorimeter (Photon Technology International, South Brunswick, NJ) with an excitation wavelength of 540 nm. Excitation and emission slit widths were set to 4 nm. The average CY3 fluorescence in the range of 560 -590 nm was then normalized for the maximum fluorescence emission intensity at 570 nm. This correction allowed the comparison of different experiments despite variations in the initial CY3-Syt II concentration. Data were expressed as FRET efficiency (E F ), where E F ϭ 1 Ϫ R FЈ , and R FЈ corresponds to the CY3 fluorescence in the presence or in the absence of FRET acceptor (R FЈ ϭ FЈ CY3-Syt II/CY5-Syt I /FЈ CY3-Syt II/Syt I ).

RESULTS AND DISCUSSION
Analyzing the roles of individual synaptotagmins is complicated by the large number of isoforms present in mammals, by their overlapping distributions, and by the difficulties in obtaining isoform-specific tools. In the past, isoform-specific antibodies have been extremely useful for the analysis of the subcellular distribution of several other synaptic proteins, including amphiphysin, syntaxin, SNAP-25, and VAMP/synaptobrevin (59 -63). For the generation of isoform-specific antibodies, it is crucial to identify stretches of amino acids with relatively low homology. The comparison of the 13 synaptotagmin sequences presently available indicates that the regions of high similarity are restricted to the two C2 domains and the C terminus, while the hydrophilic neck region and the intravesicular N-terminal segment show only limited homology (4,5). This last segment is particularly interesting because, as shown for Syt I, it is unlikely that an antibody directed against it will perturb the physiological function of the protein (64). Furthermore, antibodies against the N terminus can be used to monitor cycles of SSV exoendocytosis in neurons (64 -66). We therefore immunized rabbits with peptides corresponding to the N terminus of Syt I, II, and IV. After affinity purification on the corresponding immobilized peptide column, synaptic fractions were probed with the antibodies for specific Syt immunoreactivity. As shown in Fig. 1A, each antibody recognized a single major band in both cortical synaptic vesicles and a membrane fraction from cerebellum, and this reactivity was abolished by preincubation with an excess of the correspondent generating peptide. In the case of the anti-Syt IV, the immunoreactivity was reduced significantly but not completely abolished (not shown). As complementary proof of specificity, preimmune and immune anti-Syt IV sera were tested against the same cortical and cerebellar membrane fractions. A single band with an apparent molecular mass of 62 kDa was detected only with the immune serum (Fig. 1A).
Fractions derived from the glycerol-coated size exclusion chromatography, the last step of the SSV purification, were analyzed by SDS-PAGE and probed with the affinity-purified isoform-specific antibodies (Fig. 1B). As expected, the immunoreactivity corresponding to Syt I was localized with the second large peak of the chromatographic profile, previously referred to as CPG-II (49), which corresponds to a homogeneous vesicle population with an average diameter of 39 Ϯ 3 nm, determined by electron microscopy (not shown). Syt II and Syt IV are also present in the cortical SSV preparation (Fig. 1B, lower panel). The distribution of Syt II exactly overlaps the one observed for Syt I, while Syt IV immunoreactivity is slightly shifted. The presence of Syt II in a population of rat cortical SSV is particularly surprising, based on the relatively low abundance of Syt II at the mRNA level in the rat brain cortex and its specialized localization in cerebellum and spinal cord, suggesting a mutually exclusive Syt I and Syt II distribution (56). The presence of multiple Syt isoforms in a homogeneous SSV preparation suggests two possible scenarios. The first implies the existence of SSV containing a single synaptotagmin isoform and characterized, in terms of calcium sensitivity, by the features of the unique synaptotagmin present on its surface. The entire neuronal SSV pool would then consist of distinct synaptotagminspecific SSV subpopulations that could be mobilized by different calcium levels. This possibility is compatible with the recent finding that at least two populations of SSV can be distinguished biochemically in rat brain cortex (67). The second possibility envisages the presence of multiple synaptotagmin isoforms on the surface of the same vesicle. In this case, the SSV calcium dependence would thus be determined by the repertoire and relative abundance of the different isoforms on its surface. To discriminate between these two possibilities, SSV immunoisolation experiments were carried out. This strategy, successfully used in the past to isolate glutamate-containing vesicles from brain homogenates, was chosen because it does not perturb the integrity of the SSV by avoiding the presence of detergents (52). In this condition, the N-terminal directed Syt I antibody presents its limits due to the inaccessibility of the lumen of the SSV. We therefore used a well established monoclonal antibody against synaptotagmin, termed M48 (50), with no appreciable immunoreactivity against Syt II (not shown). After incubation of the immobilized M48 beads with the purified CPG-II fraction, beads were analyzed by Western blotting with anti-Syt II or anti-Syt IV-specific polyclonal antibodies and, as a control, with antibodies against the v-SNARE VAMP-synaptobrevin, a protein specifically localized on SSV (61,68). As shown in Fig. 2, Syt I-specific beads immunopurified SSV containing the majority of the Syt II immunoreactivity; more than 65% of Syt II immunoreactivity was found in the pellet, as determined by quantitative Western blotting and scanning analysis. In contrast, less than half of the Syt IV is associated with the Syt I-positive SSV (Յ45%), thus indicating the presence of this isoform in at least two populations of SSV, one of which lacks Syt I. This immunoisolation result has two important consequences. First, it confirms the finding that SSV populations with different biochemical features (i.e. specific membrane protein composition) can be identified (67) and, second, that a single SSV is characterized by a specific panel of synaptotagmin isoforms present on its surface. No conclusions can be drawn at this point re-FIG. 2. Immunoisolation of SSV. Highly purified SSV from rat brain cortex were immunoisolated in the absence of detergents by using anti-Syt I monoclonal antibody conjugated to protein G-Sepharose beads (Syt I IgG beads) or mock beads (empty beads). The resulting immunoisolated SSV fractions were then analyzed by SDS-PAGE and probed for the presence of Syt II (anti-Syt II), Syt IV (anti-Syt IV), and VAMP/synaptobrevin (anti-VAMP 2). Lane T, total input of purified SSV; P, proteins associated with the immunoprecipitate; S, supernatant fraction. Both Syt II and VAMP 2 are specifically enriched in the Syt I-containing SSV, while Syt IV distributes equally between the pellet and supernatant, thus indicating the presence of this isoform in a subpopulation of SSV lacking Syt I. FIG. 1. A, specificity of the isoform-specific anti-synaptotagmin N-terminal antibodies. Affinity-purified anti-Syt I and II antibodies were incubated either with dimethyl sulfoxide (Յ4% (v/v) final concentration) (Ϫ) or with the generating peptide corresponding to the N terminus of the specific isoform (0.5 mg dissolved in dimethyl sulfoxide) (ϩ) and then used to probe nitrocellulose membranes containing 15 g of purified SSV from cortex (CTX) and 50 g of a crude vesicle fraction from cerebellum (CBM). Alternatively, the same brain fractions were used to test preimmune and immune anti-Syt IV serum. B, Syt I, II, and IV colocalize in an SSV preparation from rat brain cortex. The elution profile of a glycerol-coated controlled pore glass beads column (monitored at 280 nm) loaded with an impure fraction of rat cortical SSV is shown in the upper panel. Proteins from the resulting fractions were separated by SDS-PAGE, transferred onto nitrocellulose, and then analyzed with anti-Syt I, II, and IV-specific antibodies. The immunoreactivity for Syt I and Syt II co-localizes and overlaps with the second peak of the elution profile (CPG-II), which corresponds to pure SSV. Syt IV immunoreactivity also peaks in the CPG-II area, but it is slightly shifted compared with the signal for Syt I and II.
garding the presence of other synaptic markers on Syt IVpositive vesicles. Future experiments characterizing the subcellular distribution of Syt IV are necessary to conclusively define the vesicular compartment containing this synaptotagmin isoform.
Recently, Syt I was demonstrated to homodimerize via pairing of its C2B domain (46,47). The presence of multiple synaptotagmin isoforms on the surface of the synaptic membrane raises the possibility that they could interact by forming heterocomplexes. To test this hypothesis, immunoprecipitation experiments were performed on different membrane preparations (Fig. 3). N-terminal Syt I antibodies conjugated to protein G-Agarose beads were applied to membrane extracts obtained by solubilizing purified cortical SSV with 4% OG. After incubation, the immunoprecipitate was extensively washed and then analyzed as described in Fig. 2. The same experiments were performed in parallel on a crude vesicular fraction from cerebellum, a tissue enriched in Syt II. In both cases, co-immunoprecipitation between Syt I and II was observed (Fig. 3,  upper panel). Experiments were also performed using immobilized anti-Syt II antibodies, and the same conclusion holds true (data not shown), thus suggesting that Syt I and II have an intrinsic ability to interact. In contrast, Syt IV exhibits very low binding to Syt I in membrane extracts from both cortex and cerebellum (5-10%; Fig. 3, lower panel). Even at calcium concentrations supporting maximal interaction between Syt I and Syt II (Ͼ100 M calcium; see below), no significant calcium-dependent interaction between Syt I and IV can be detected. Under these conditions, only 5-12% of the total Syt IV is associated with Syt I (not shown). An interpretation of this result is that the pairing between different isoforms is characterized by a certain level of specificity and distinct calcium dependences, determined by the structure and binding properties of their C2B domains. This situation would be similar to that of the C2A domain of synaptotagmins, known to interact differently with phosphatidylserine and syntaxin in distinct isoforms (19). Interestingly, the C2A domain of Syt IV does not show any calcium-dependent binding of at least one of these ligands (69,70), and it is therefore classified as a member of the calcium-independent synaptotagmin subfamily (4). The existence of different classes of C2B domains is suggested by recent observations that distinct C2B domains have different binding activities for inositol 1,3,4,5-tetrakisphosphate (23) and phos-phoinositides. 2 Future experiments investigating the heterodimerization abilities of different isoforms are needed to further clarify this point.
Recently, findings from several laboratories have indicated that the C2B domain-dependent homodimerization of Syt I is calcium-dependent (46,47). To determine if this feature is shared by the interaction between Syt I and Syt II, solubilized proteins from purified cortical brain SSV were immunoprecipitated with immobilized anti-Syt I antibodies in the presence of different calcium concentrations. Calcium potently promotes the interaction between the two isoforms and, at equilibrium, more than 45% of Syt II is engaged in the binding with Syt I, as shown by Western blotting with anti-Syt II antibodies (Fig. 4,  upper panel). The concentration of calcium triggering the halfmaximal effect (EC 50 ) is 6 Ϯ 4 M, while maximal association between Syt I and II is reached at a concentration Ն100 M Ca 2ϩ . Together, these results suggest that, at calcium concentrations experienced by an SSV during exocytosis, a large fraction of Syt II may be in a complex with Syt I. This finding, together with the observation that multiple synaptic isoforms 2 C. Thomas and G. Schiavo, unpublished results. Purified SSV from rat brain cortex (CTX) and a crude vesicular fraction from rat cerebellum (CBM) were solubilized with OG and then immunoprecipitated in the absence of externally added calcium by using immobilized anti-Syt I antibodies. The resulting pellets (P) and supernatants (S) were then probed with anti-Syt II (upper panel) or anti-Syt IV antibodies (lower panel). Association between Syt I and Syt II is detectable in both membrane fractions, while Syt I and Syt IV show only a minimal interaction (5-12% of the total).

FIG. 4. Calcium dependence of Syt I and Syt II co-immunoprecipitation.
Purified SSV from rat brain cortex were solubilized with OG and then immunoprecipitated with immobilized anti-Syt I antibody in the absence (EGTA) or in the presence of EGTA-calcium buffers with a concentration of free Ca 2ϩ ranging from 1 M to 1 mM and then analyzed by Western blotting with an anti-Syt II-specific antibody (upper panel). Syt II associated with the Syt I immunoprecipitate was quantified with the NIH Image software by comparing the Syt II signal present in the immunoprecipitate with that obtained from a standard curve (0.25, 0.5, 0.75, and 1 times the starting material). The fraction of Syt II associated with the pellets in different conditions was expressed as a percentage of the total Syt II in the sample. Both the sum of the immunoreactivity present in the supernatant (S) and in the pellet (P) and the total input of the sample (T) were used as denominators to calculate the percentage of synaptotagmin recovery. The two values were compared, and both were used to determine the EC 50 of the calcium dependence of Syt I/Syt II co-immunoprecipitation. The variability between the two calculation methods is included in the error bars in the lower panel. To compare different experiments (n ϭ 5), data were normalized, and mean values were plotted as a function of calcium concentration (lower panel). The EC 50 of the equilibrium is 6 Ϯ 4 M Ca 2ϩ with maximal Syt I-Syt II association reached at concentration Ն 100 M Ca 2ϩ . are present on the same SSV, raises the possibility that the self-association of synaptotagmins with different calcium binding features might create a variety of calcium sensors characterized by distinct calcium sensitivities.
The heterodimerization process was confirmed in vitro with recombinant proteins by using two different approaches. The first involved the co-immunoprecipitation of the cytoplasmic domain of Syt I (residues 95-421) with the homologous portion of Syt II (residues 103-422). In order to minimize steric hindrance during immunoprecipitation, Syt II was tagged at the amino terminus with a 9-residue version of the hemagglutinin epitope (HA-Syt II). The two purified isoforms, showing a slightly different mobility in SDS-PAGE (Fig. 5, inset; lanes A  and B), were mixed, and Ca 2ϩ /EGTA buffers were added to obtain the desired free calcium concentration. HA-Syt II was then quantitatively recovered with immobilized HA-specific antibodies, and the pellet was analyzed for Syt I. As shown in the inset of Fig. 5, the presence of Syt I in the immunoprecipitate is dependent on the addition of HA-Syt II to the incubation buffer (compare lanes C and D). The process is dependent on calcium and shows an EC 50 of 5 Ϯ 3 M (n ϭ 4), a value very similar to the one observed with the native proteins. The addition of magnesium to the equilibrium, previously reported to alter the homodimerization process of Syt I (46), only slightly shifts the EC 50 of the phenomenon toward higher calcium concentrations (9 Ϯ 5 M; n ϭ 3), but it is very efficient in reducing the amount of calcium-independent binding between Syt I and Syt II (Fig. 5). These results demonstrate that the recombinant cytoplasmic domain of Syt I and II are competent for calcium-dependent heterodimerization and can mimic the behavior of the native proteins, in contrast to a previous report investigating the dimerization of Syt I, where at least one native protein was required for efficient and calcium-dependent dimer formation (47). The reason for this discrepancy is not clear, but it could be ascribed to differences in the conditions used for synaptotagmin expression and for immunoprecipitation, together with a different protein tagging approach.
The second strategy used to test the in vitro interaction between Syt I and II is based on a FRET approach (58,71). For this purpose, the recombinant cytoplasmic portions of Syt I and II were labeled with either the fluorescent dye CY3 or with CY5. For both dyes, the degree of modification was optimized to obtain an incorporation of 1 mol of dye/mol of protein. Experimentally, we used modified synaptotagmins with a ratio of dye to protein ranging between 0.5 and 1.2.
To test if Syt I and Syt II interact functionally in solution, CY3-modified Syt II (fluorescent donor) was mixed with CY5modified Syt I (fluorescent acceptor) in the absence of calcium and then in the presence of increasing calcium concentrations (Fig. 6A). To exclude quenching effects due to the direct interaction of the added protein with the donor dye CY3, the same experiment was performed with CY3-Syt II alone (Fig. 6D) or by adding recombinant Syt I without CY5 modification (Fig.  6C). In both of these samples, the addition of calcium caused a decrease in the fluorescence emission of Cy3-Syt II, the extent of which is less pronounced than that seen in the presence of the Cy5-Syt I acceptor. The emission peak at 664 nm present in the samples of Fig. 6, A and B, is partly due to direct excitation of CY5-Syt I (Fig. 6B) and, in the presence of the donor Cy3-Syt II, to FRET. Its decrease (instead of increase) with increasing calcium concentrations could be attributed to internal quenching of this emission peak, together with a decrease of the intensity of the donor spectra, which constitutes its base line. A calcium-dependent internal quenching is also visible in the solution of pure CY5-Syt I with increasing calcium (Fig. 6B). The FRET efficiency (E F ) of the dimerization was therefore calculated using the emission peak of the donor, where an efficient control for the FRET-independent quenching of the fluorophore is available (Fig. 6C).
The average CY3 fluorescence quenching in the range of 560 -590 nm was then normalized for the maximum emission wavelength ( ϭ 570 nm), and the data were expressed as the FRET efficiency versus the calcium concentration. E F was calculated as E F ϭ 1 Ϫ R FЈ , where R FЈ is the ratio between the fluorescence measured in the presence of FRET acceptor and in its absence (58). As shown in Fig. 6 (A, C, and E) the addition of acceptor at calcium concentrations Յ 1 nM (2 mM EGTA) causes a significant quenching (E F ϭ 0.05) that could be attributed to the calcium-independent Syt I/Syt II association, a feature already observed in the immunoprecipitation experiments performed with both the native and recombinant synaptotagmins (see Figs. 4 and 5). The addition of calcium induces an increase in the FRET efficiency (Fig. 6E). This parameter is directly linked to the CY5-dependent quenching of CY3 due to FRET and is corrected for the quenching effect due to the direct interaction of the added protein with the donor CY3. The EC 50 of the calcium dependence of the Syt I/Syt II interaction corresponds to 140 Ϯ 80 M (n ϭ 5). Analogous results were obtained in experiments performed using the opposite labeling strategy (CY3-modified Syt I and CY5-modified Syt II, not shown). Millimolar magnesium only slightly influenced the interaction between Syt I and Syt II (not shown). The reason for the discrep- The recombinant cytoplasmic portions of Syt I and Syt II, the latter tagged at the N terminus with HA, were mixed, and after incubation in the presence of different calcium concentrations, HA-Syt II was immunoprecipitated with HA-specific antibodies. Syt I associated with the beads was analyzed either by densitometric analysis of the Coomassie Blue-stained gels or by Western blotting with anti-Syt I-specific antibodies. Recovery was quantified by comparing the signal present in the pellets with that obtained from a standard curve of recombinant Syt I (from 50 to 250 ng for Western blotting and from 0.2 to 1 g for Coomassie staining). Data were expressed as a percentage of the maximal Syt I present in the Syt II immunoprecipitate. The EC 50 for the calcium dependence of the equilibrium is similar in both the absence (closed circles) or the presence (empty circles) of 0.5 mM free magnesium. The addition of magnesium reduces significantly the amount of calcium-independent binding between Syt I and Syt II. ancy between the EC 50 values of the heterodimerization determined by immunoprecipitation and by FRET is not surprising considering the differences between the two experimental systems. In fact, in contrast to the immunoprecipitation experiments, FRET analysis is performed at equilibrium with both ligands present in a homogeneous phase during the entire measurement. Future experiments monitoring the real time heterodimerization of the full-length Syt I and Syt II in living cells via FRET and life-time measurements will be helpful in addressing this issue.
How could this data on synaptotagmin heterodimerization be integrated into the general mechanism of docking and fusion of SSV at the active zones of a nerve terminal? At the present time, no unitary view exists concerning the molecular details of the tethering, docking, priming, and fusion of a SSV with the presynaptic membrane and the role(s) of synaptotagmins in these mechanisms (72). A series of compelling results indicates that the members of the synaptotagmin family represent the best candidates for the role of calcium sensors in regulated exocytosis (3)(4)(5). At the molecular level, this function can be correlated with the calcium-dependent interactions with syntaxin, phosphatidylserine, and phosphatidylinositol 4,5diphosphate (11, 13-15, 18 -20). In addition, due to the ability of the C2B domain to bind different ligands at resting calcium concentrations (i.e. phosphatidylinositol 3,4,5-triphosphate and SNAP-25) (20, 26 -28), synaptotagmins may also be involved in the tethering and docking of SSV to the active zones of the synaptic membrane. The formation of homo-and heterodimers of synaptotagmins can be integrated into this network of interactions as an additional element in the regulation of the calcium-dependent phase of exocytosis. In fact, the calciumdependent association of synaptotagmins with different calcium binding features could create a variety of calcium sensors characterized by distinct calcium sensitivities. This combinatorial hypothesis predicts that the probability of a single SSV exocytic event is determined by the repertoire of synaptotag- mins present on the SSV surface and the gating properties of the calcium channels at the synapse. Future experiments will assess the specific targeting and association of synaptotagmin isoforms characterized by distinct calcium sensitivities in living cells and monitor the efficiency and calcium sensitivity of neurotransmitter release under these conditions. This experimental system will also offer the possibility to investigate the physiological modulator(s) of several synaptotagmin isoforms, like Syt IV, VI, VIII, IX, and XI, that appear not to be modulated by calcium (4) and are likely to be involved in some aspects of membrane trafficking outside the nervous system.