Direct, Ca2+-dependent Interaction between Tubulin and Synaptotagmin I

The synaptic vesicle protein synaptotagmin I probably plays important roles in the synaptic vesicle cycle. However, the mechanisms of its action remain unclear. In this study, we have searched for cytoplasmic proteins that interact with synaptotagmin I. We found that the cytoskeletal protein tubulin directly and stoichiometrically bound to recombinant synaptotagmin I. The binding depended on mm Ca2+, and 1 mol of tubulin dimer bound 2 mol of synaptotagmin I with half-maximal binding at 6.6 μm tubulin. The Ca2+ dependence mainly resulted from Ca2+ binding to the Ca2+ ligands of synaptotagmin I. The C-terminal region of β-tubulin and both C2 domains of synaptotagmin I were involved in the binding. The YVK motif in the C2 domains of synaptotagmin I was essential for tubulin binding. Tubulin and synaptotagmin I were co-precipitated from the synaptosome extract with monoclonal antibodies to tubulin and SNAP-25 (synaptosome-associated protein of 25 kDa), indicating the presence of tubulin/synaptotagmin I complex and tubulin binding to synaptotagmin I in SNARE (soluble N-ethylmaleimide-sensitive factor attachment protein receptor) complexes. Synaptotagmin I promoted tubulin polymerization and bundled microtubules in the presence of Ca2+. These results suggest that direct interaction between synaptotagmin I and tubulin provides a mechanism for attaching synaptic vesicles to microtubules in high Ca2+ concentrations.

Chemical synapses depend on fast release of neurotransmitter molecules from the presynaptic terminal by synaptic vesicle exocytosis (1). This release is triggered by instantaneous increase of intracellular Ca 2ϩ concentration around the release site (active zone), due to Ca 2ϩ influx through voltage-sensitive calcium channels opened by depolarization. After exocytosis, synaptic vesicles are rapidly endocytosed and regenerated by multiple pathways. Synaptic vesicle endocytosis is a process driven by concerted actions of many proteins and other substances, which is not yet fully understood (2,3). Thus synaptic vesicles undergo a complex, tightly regulated cycle. To understand the mechanisms of the synaptic plasticity underlying higher brain functions, it is essential to elucidate the molecular basis of this synaptic vesicle cycle.
Recent investigations have established a central role of SNAREs 1 in the release process of neurotransmitters; specific interactions between t-SNAREs (syntaxin and SNAP-25 from the presynaptic membrane) and a v-SNARE (VAMP/synaptobrevin from the synaptic vesicle membrane) are critical for synaptic vesicle exocytosis (4 -7). This trans-SNARE complex forms a parallel four-helix bundle (8 -10). Close apposition of the synaptic vesicle membrane and the presynaptic membrane by the tight trans-SNARE complex, with the energy released by its assembly, is proposed to drive fusion between the two membranes (11). The complex may serve as a minimal machinery for membrane fusion as supported by liposome fusion (12). However, many important aspects of the synaptic vesicle cycle still remain unclear. For instance, how does Ca 2ϩ trigger a cascade of interactions between the SNAREs? How can selective retrieval of the synaptic vesicle membrane from the plasma membrane be achieved? How are endocytosed synaptic vesicles distributed into different pools?
Synaptotagmin I is a Ca 2ϩ -binding, synaptic vesicle membrane protein that probably plays important roles in the synaptic vesicle cycle (13)(14)(15). The protein consists of a short N-terminal luminal domain, one transmembrane segment, and two C2 domains extending into the cytoplasm. In vitro, these C2 domains bind Ca 2ϩ (15), phospholipids (16 -18), syntaxin 1 (19 -20), and SNAP-25 (21,22). The protein forms complexes with SNAREs including VAMP through its Ca 2ϩ -dependent binding to syntaxin 1 and SNAP-25. The Ca 2ϩ affinity of its binding to syntaxin 1 and SNAP-25 apparently matches that of neurotransmitter release (23). Synaptotagmin mutants exhibit severe movement disorders in the nematode Caenorhabditis elegans (24) and a marked decrease of Ca 2ϩ -dependent neurotransmitter release in Drosophila (25,26). In synaptotagmin I-deficient mice, fast neurotransmitter release is profoundly impaired (27), suggesting an important role of the protein in the release mechanisms. A very recent genetic study has shown a quantitative relationship between the Ca 2ϩ -binding affinity of synaptotagmin I and the Ca 2ϩ sensitivity of transmitter release (28). Furthermore, the protein responds to Ca 2ϩ very rapidly to bind simultaneously to membrane and the ternary SNARE complex (29). Based on these findings, the protein has been widely assumed as a major Ca 2ϩ sensor in the release process. However, it remains to be elucidated how synaptotag-min I could transmit the Ca 2ϩ signal to the SNAREs, eventually bringing about the fusion of the synaptic vesicle membrane with the presynaptic membrane. Interestingly, overexpression of synaptotagmin I in PC12 cells modulated fusion pore kinetics, indicating its interaction with fusion pores (30).
The protein also interacts with the clathrin adaptor protein AP-2 with high affinity (31,32), suggesting its involvement in synaptic vesicle endocytosis. In synaptotagmin mutants of C. elegans (33) and Drosophila (34), synaptic vesicles in the nerve terminal were markedly decreased. A similar decrease of synaptic vesicles was observed in the squid nerve terminal injected with a polyclonal antibody to synaptotagmin I (35). Moreover, overexpression of synaptotagmin I or II in the neuromuscular junction of Xenopus led to a different distribution of synaptic vesicles without change in the total number of synaptic vesicles (36). These findings suggest that synaptotagmin I is involved not only in the endocytosis of synaptic vesicles but also in their distribution. The protein thus seems to regulate many steps of the synaptic vesicle cycle. However, its exact role in each step remains poorly understood. To understand the mechanisms of various functions of synaptotagmin I, we searched for cytosolic proteins that interact with synaptotagmin I. We have found that synaptotagmin I directly binds to tubulin in a Ca 2ϩdependent manner. Our findings suggest that this binding provides a mechanism for attaching synaptic vesicles to microtubules in high Ca 2ϩ concentrations. Preliminary accounts of this work have been published (37,38).

EXPERIMENTAL PROCEDURES
Materials-Taxol, subtilisin (type III), phenylmethylsulfonyl fluoride AEBSF, bovine thrombin, and benzamidine-agarose were purchased from Sigma. Glutathione-Sepharose and protein G-Sepharose fast flow were obtained from Amersham Biosciences.
Antibodies-Anti-␣-tubulin and ␤-tubulin mAbs (DM1A and TUB 2.1, respectively) were obtained from Sigma. Anti-SNARE mAbs have been reported previously (39). A polyclonal antibody against rat VAMP-2 was generated by immunizing BALB/c mice with recombinant, full-length VAMP-2 prepared by digestion of GST-VAMP-2 with thrombin. One hundred g (first injection) or 50 g (subsequent injections) of VAMP-2 in phosphate-buffered saline was emulsified with an equal volume of Freund's complete (first injection) or incomplete (subsequent injections) adjuvant and injected intraperitoneally at two-week intervals. One week after the fourth injection, mouse antisera were collected. A mAb (3F10) raised against the N-terminal region of synaptotagmin I (40) was a gift from M. Takahashi (Mitsubishi Kasei Institute for Life Sciences, Tokyo, Japan).
Subcellular Fractions-All procedures were carried out at 4°C in the presence of protease inhibitors (1 M pepstatin A/2 M leupeptin/0.3 mM phenylmethylsulfonyl fluoride). Fresh forebrains from three-week-old female rats were homogenized in 10 mM Hepes-NaOH (pH 7.4), and then solid NaCl was added to a final concentration of 150 mM. After incubation for 30 min, the homogenate was centrifuged at 200,000 ϫ g for 1 h, and the supernatant was used as the soluble fraction. The lysed P2 and crude synaptosome fractions were prepared from adult rat forebrains as follows. Forebrains were homogenized in 10 mM Hepes-NaOH (pH 7.4)/0.32 M sucrose and centrifuged at 700 ϫ g for 10 min. The supernatant was centrifuged again at 9000 ϫ g for 20 min, and the pellet (P2 fraction) was collected as crude synaptosome fraction. The P2 fraction was lysed by dilution into 9 volumes of 10 mM Hepes-NaOH (pH 7.4) and incubated for 30 min. The pellet obtained by centrifugation at 9000 ϫ g for 20 min was the lysed P2 fraction. The lysed P2 and crude synaptosome fractions were suspended in solution A (20 mM Hepes-NaOH (pH 7.4)/150 mM NaCl) at a protein concentration of 4 mg/ml, and equal volumes of solution A containing 2% (w/v) Triton X-100 were added. The mixtures were stirred for 30 min and then centrifuged at 100,000 ϫ g for 1 h. The resultant supernatants were diluted with equal volumes of solution A and used as the Triton X-100 extracts. These extracts were used immediately after preparation. Protein concentrations were determined using the Bio-Rad protein assay kit.
Expression and Purification of GST Fusion Proteins-cDNAs for full-length and cytoplasmic regions of wild-type or mutant rat synaptotagmin I were prepared by PCR and cloned into the pGEX-KG vector to express as GST fusion proteins. Substituted mutations of constituents of the Ca 2ϩ ligand (D230N/D232N and D363N/D365N), the polylysine motifs (K(189 -192)A and K(324 -327)A) or the (SDP) YVK motifs (Y180A/V181A/K182A and Y311A/V312A/K313A) were prepared using the overlapping primer method (41). Plasmids for GST fusion proteins of the cytoplasmic regions of rat syntaxin 1A and VAMP-2 and fulllength mouse SNAP-25 were similarly prepared. All constructs were verified by DNA sequencing. These constructs were transfected to the Escherichia coli strain BL21. GST fusion proteins were purified using glutathione-Sepharose. For some experiments, GST-synaptotagmin I bound to glutathione-Sepharose was cleaved with thrombin (1 unit/mg GST fusion protein). Thrombin in the released proteins was removed with benzamidine-agarose. All recombinant proteins were used within 3 days after purification.
Affinity Chromatography on GST Fusion Proteins-GST alone, GSTentire cytoplasmic portion of synaptotagmin I (referred to as GSTsynaptotagmin I), or GST-synaptotagmin I fragments (all 3 M) immobilized on glutathione-Sepharose beads (100 l) were incubated with the rat brain-soluble fraction (1 mg of protein/ml) or purified tubulin (1 mg/ml) in 0.5% Triton X-100/HNa buffer (10 mM Hepes-NaOH (pH 7.4)/150 mM NaCl/1 M pepstatin A/2 M leupeptin/0.3 mM phenylmethylsulfonyl fluoride) containing 3 mM CaCl 2 or 1 mM EGTA for 2 h at 4°C. The beads were then washed three times with 1 ml of 0.1% Triton X-100/HNa buffer containing 3 mM CaCl 2 or 1 mM EGTA. The bound materials were eluted with the sample buffer (42) and subjected to SDS-PAGE.
Affinity Chromatography on Tubulin-Sepharose-Purified tubulin (see below) was coupled to CNBr-activated Sepharose 4B (Amersham Biosciences) (1.1 mg of tubulin/ml gel). Tubulin-Sepharose beads were incubated with the Triton X-100 extract of the lysed P2 fraction (1 mg of protein/ml in 0.5% Triton X-100/HNa buffer/3 mM CaCl 2 ) for 2 h at 4°C. The beads were washed three times with 10 volumes of 0.5% Triton X-100/HNa buffer/3 mM CaCl 2 and then with 10 volumes of 0.1% Triton X-100/HNa buffer/3 mM CaCl 2 . The bound materials were eluted with the sample buffer, fractionated by SDS-PAGE, and detected by immunoblotting.
Purification and Subtilisin Treatment of Tubulin-Whole microtubule proteins (CS3) were prepared from porcine brains by temperaturedependent cycles of assembly-disassembly following the procedures of Shelanski et al. (43). Pure tubulin was isolated from CS3 by chromatography on phosphocellulose (Whatman P-11) (44), concentrated by ultrafiltration to 10 mg/ml, and stored at Ϫ80°C until use. Digestion of tubulin with subtilisin was carried out by modifying the method of Rodionov et al. (45). Briefly, taxol-stabilized microtubules (10 mg/ml tubulin treated with 20 M taxol) were digested with subtilisin (10 g/ml) for 15 min or overnight at 37°C. The cleavage reactions were terminated by addition of 2 mM AEBSF. Subtilisin-treated microtubules were sedimented through a cushion of 10% (w/v) sucrose in RB buffer (100 mM Mes-KOH (pH 6.8)/0.5 mM MgCl 2 /1 mM EGTA) containing 1 mM GTP, 20 M taxol, and 2 mM AEBSF at 100,000 ϫ g for 50 min at 30°C. The pellets were washed and resuspended in HNa buffer containing 2 mM AEBSF. The subtilisin-digested tubulins were subjected to SDS-PAGE by the modified Laemmli's method (46).
Blot Overlay Assay-Purified tubulin was resolved by SDS-PAGE on 12.5% gel and transferred to nitrocellulose membrane. The membrane was blocked overnight at 4°C with 3% skim milk in TBST buffer (0.05% Tween 20/10 mM Tris-HCl (pH 8.0)/150 mM NaCl) and then incubated for 6 h with the Triton X-100 extract of the lysed P2 fraction (final protein concentration of 1 mg/ml in 0.5% Triton X-100/HNa buffer/3 mM CaCl 2 ). After washing five times with TBST/3 mM CaCl 2 , synaptotagmin I bound to tubulins was detected by a mAb to the protein.
Immunoprecipitation-Protein G-Sepharose fast flow (250 l of resin) was incubated with a mAb to ␣-tubulin (DM1A) or mAb to SNAP-25 (6H4, see Ref. 39) for 2 h and then washed three times with 0.2 M borate buffer (pH 9.0). Solid dimethyl pimelimidate was added to a final concentration of 20 mM, and the mixture was incubated for 45 min at room temperature. Then the Sepharose beads were washed twice and incubated for 2 h in 0.2 M ethanolamine-HCl (pH 8.0). The mAb-coupled Sepharose beads were rinsed in phosphate-buffered saline and stored at 4°C. Tubulin and SNAP-25 were immunoprecipitated from 2 ml of the Triton X-100 extract (1 mg of protein/ml) of crude synaptosomes by incubating with 50 l of mAb-coupled protein G-Sepharose for 2 h at 4°C. The resin was washed three times with 10 volumes of HNa buffer/0.1% Triton X-100. The proteins precipitated with the resin were eluted with the sample buffer, fractionated by SDS-PAGE, and detected by immunoblotting using the ECL Western blotting detection reagents (Amersham Biosciences).
Measurement of Tubulin Polymerization-CS3 (2 mg/ml) or a mixture of thrombin-released synaptotagmin I (1.5 M) and CS3 was incubated in the presence or absence of 1 mM CaCl 2 for 30 min at 4°C and transferred to cuvettes. The cuvettes were warmed to 37°C, and 1 mM GTP was added. Microtubule formation was assayed by measuring turbidity (absorbance at 340 nm) at 37°C. Aliquots of polymerization reactions taken at 30 min at 37°C were analyzed by electron microscopy.
Co-sedimentation of Synaptotagmin I and Tubulin-A mixture of CS3 and thrombin-released synaptotagmin I was incubated for 30 min at 30°C in the presence of 1 mM GTP and 1 mM CaCl 2 and then centrifuged at 100,000 ϫ g for 50 min at 30°C through a cushion of 10% sucrose in RB buffer without EGTA. The pellet was homogenized on ice in RB buffer/1 mM GTP/1 mM CaCl 2 and then centrifuged at 100,000 ϫ g for 50 min at 4°C. Coomassie Blue staining after SDS-PAGE detected proteins in each supernatant and pellet.
Electron Microscopy-Samples were applied to collodion-coated grids (400 mesh) for 30 s and then fixed with 1% (w/v) glutaraldehyde for 5 s. After rinses in distilled water, samples were negatively stained with 3% (w/v) uranyl acetate solution. The grids were examined with the Hitachi H-7100 electron microscope.

RESULTS
Binding of Tubulin to Synaptotagmin I-To identify soluble proteins that specifically interact with synaptotagmin I, the soluble fraction from the rat brain was incubated with GST fusion protein of the cytoplasmic portion of synaptotagmin I immobilized on glutathione-Sepharose. After thorough washing, the bound materials were eluted with SDS and subjected to SDS-PAGE. When the soluble fraction was incubated with GST-synaptotagmin I in the presence of EGTA, no major binding components were detected. (data not shown). However, in the presence of 3 mM CaCl 2 , only a protein of 55 kDa was found in the bound material (Fig. 1a). This component was not retained by immobilized GST alone (Fig. 1b), GST fusion proteins of the cytoplasmic portion of syntaxin 1A or VAMP-2, or fulllength SNAP-25 (data not shown), indicating the specificity of the binding. The 55-kDa protein was recognized by mAbs spe-cific to ␣or ␤-tubulin (Fig. 1b). Based on the molecular mass value and the reaction with the mAbs, the 55-kDa band was identified as a mixture of ␣and ␤-tubulin.
Ca 2ϩ Dependence of the Binding-To characterize tubulin binding to synaptotagmin I in more detail, we used purified tubulin instead of the brain-soluble fraction. We determined tubulin binding as the molar ratio between synaptotagmin I and tubulin dimer (␣␤) (usual form of tubulin) by NIHImage analysis of Coomassie Blue-stained bands. Concentration dependence of tubulin binding to synaptotagmin I was determined by incubating immobilized GST-synaptotagmin I with increasing concentrations of purified tubulin in the presence or absence of CaCl 2 (Fig. 2a). Purified tubulin efficiently bound to GST-synaptotagmin I in the presence of CaCl 2 . Thus the interaction between GST-synaptotagmin I and tubulin was direct. The binding was saturated at 1.5 mg/ml tubulin with halfmaximal binding at ϳ6.6 M (0.72 mg/ml). At the saturation, the molar ratio between tubulin dimer and synaptotagmin I was ϳ0.5, indicating that one tubulin dimer binds two synaptotagmin I molecules. In the absence of CaCl 2 , the maximal binding was about one-fifth of that in CaCl 2 . Fig. 2b illustrates the Ca 2ϩ dependence of tubulin (2 mg/ml) binding to synaptotagmin I. Tubulin binding was saturated at 2.0 mM CaCl 2 with half-maximal binding at ϳ0.9 mM CaCl 2 . Ca 2ϩ -dependent binding of tubulin to synaptotagmin I appeared to consist of at least two components with different Ca 2ϩ affinities (see the curve in Fig. 2b), like syntaxin 1A binding to synaptotagmin I (19).
Domain of Synaptotagmin I Involved in Tubulin Binding-To determine the domain of synaptotagmin I involved in tubulin binding, binding of purified tubulin to GST fusion proteins of various portions of the cytoplasmic domain of synaptotagmin I was examined (Figs. 3, a and b). Neither the region between the transmembrane segment and the C2A domain (fragment I) nor the short linker domain between the C2A and C2B domains (fragment III) exhibited significant binding. Binding of C2A and C2B domains was respectively about onehalf of that of the entire cytoplasmic domain. C2A domain plus C2B domain without the linker bound tubulin almost as efficiently as the entire cytoplasmic domain (data not shown). Therefore, both C2A and C2B domains are involved in tubulin binding.
Based on these findings, we inferred that the sequences common to both C2 domains of synaptotagmin I might be important for tubulin binding. We focused on the polylysine (KKKK) (47) and YVK (also called SDPYVK) (16,17) motifs conserved between the two C2 domains. In a previous study, replacement of the polylysine motif with alanine did not impair syntaxin/synaptotagmin I binding but inhibited neurotransmitter release from PC12 cells (48). The YVK motif has been shown to be involved in its Ca 2ϩ -dependent binding to syntaxin 1 and phospholipids (16,17). We replaced these two motifs in both C2 domains with alanine and examined tubulin binding of these mutants (Fig. 3c). Although the polylysine motif mutant exhibited normal binding, binding was negligible in the YVK motif mutant (Fig. 3, d and e). Thus the YVK motif was essential for tubulin binding. Tubulin binding in the absence of CaCl 2 was also greatly decreased.
Four aspartic acid residues in the C2A (Asp-230 and Asp-232) (15) and C2B domains (Asp-363 and Asp-365) (49) are important constituents of the Ca 2ϩ -binding sites (Ca 2ϩ ligands) of synaptotagmin I. Replacement of all these residues with alanine leads to a loss of Ca 2ϩ sensitivity of the bindings of SNAREs (syntaxin 1A and SNAP-25) and phospholipids (50). To examine whether tubulin binding to synaptotagmin I depends on Ca 2ϩ sensitivity of synaptotagmin I as reported for these substances, we prepared the same mutant. Ca 2ϩ -depend- ent tubulin binding of the mutant synaptotagmin I was markedly decreased (38% of wild-type). In contrast, Ca 2ϩ -independent binding did not significantly change (Fig. 3, d and e). Thus Ca 2ϩ dependence of tubulin binding to synaptotagmin I mainly results from Ca 2ϩ binding to the Ca 2ϩ ligands of synaptotagmin I.
Tubulin Binds to Synaptotagmin I in SNARE Complexes-The above results have shown that tubulin, like some SNAREs, Ca 2ϩ -dependently binds to synaptotagmin I. This raises the possibility that these SNAREs and tubulin compete for synaptotagmin I. To check the possibility we immunoprecipitated the tubulin/synaptotagmin I complex from the Triton X-100 extract of crude synaptosomes with mAbs to tubulin and the t-SNARE SNAP-25. An anti-synaptotagmin I mAb precipitated tubulin and SNAREs as well as synaptotagmin I, showing that the synaptotagmin I/tubulin complex exists (data not shown). However, this does not tell whether tubulin and SNAREs compete for synaptotagmin I as synaptotagmin I could bind SNARE complexes and tubulin, separately. We therefore used an anti-SNAP-25 mAb as SNAP-25 does not directly bind tubulin (see above). The immunoprecipitate obtained with a mAb to tubulin contained synaptotagmin I, syntaxin 1, and SNAP-25 in addition to tubulin (Fig. 4a). Conversely, an anti-SNAP-25 mAb precipitated tubulin and synaptotagmin I together with the SNAREs (syntaxin 1, SNAP-25, and VAMP-2) (Fig. 4b). Practically all of the syntaxin 1 immunoprecipitated with a mAb to SNAP-25 existed as the ternary complex as shown by SDS resistance at 37°C. As already mentioned, GST-syntaxin 1A, -SNAP-25, or -VAMP-2 did not bind to tubulin. A previous study has reported a very weak binding between tubulin and syntaxin 1A (51). However, as the molar ratio (tubulin to syn-taxin 1A) of the binding is about 0.02, it is very unlikely that syntaxin 1A/tubulin binding significantly contributed to the immunoprecipitates. Thus immunoprecipitation of synaptotagmin I with a mAb to tubulin probably indicates the presence of tubulin/synaptotagmin I binding in vivo. The immunoprecipitation experiments also show that SNAREs and tubulin do not compete with each other for synaptotagmin I. This conclusion was supported by the finding that purified tubulin-conjugated Sepharose bound SNARE/synaptotagmin I complexes present in the Triton X-100 extract of the lysed P2 fraction (Fig. 4c).
Domain of Tubulin Involved in Synaptotagmin I Binding-We next examined whether one or both of the tubulin dimer subunits were involved in synaptotagmin I binding by the blot overlay method (Fig. 5a). Purified tubulin transferred to nitrocellulose membrane was incubated with the Triton X-100 extract of the lysed P2 fraction, and bound synaptotagmin I was detected by a mAb. ␤-Tubulin but not ␣-tubulin bound synaptotagmin I. To identify the synaptotagmin I binding domain, purified tubulin was subjected to limited subtilisin digestion. Short-time digestion of tubulin dimer with subtilisin is known to remove the C-terminal region of ␤-tubulin (␣ s ␤), and prolonged digestion leads to the cleavage of ␣-tubulin as well (␣ s ␤ s ). Tubulins digested for a short time and overnight were incubated with immobilized GST-synaptotagmin I, and the bound tubulin was measured (Fig. 5b). The binding of synaptotagmin I was lost by removing the C-terminal region of ␤-tubulin. Taken together, these results indicate that synaptotagmin I binds to the C-terminal region of ␤-tubulin.  4. Immunoprecipitation of synaptotagmin I/tubulin complex from brain synaptosomes. a, precipitation with a mAb to tubulin. b, precipitation with a mAb to SNAP-25. The Triton X-100 extract of synaptosomes was immunoprecipitated, and the precipitated proteins were detected by immunoblots. a, synaptotagmin I (syt), syntaxin 1 (syn 1), and SNAP-25 were precipitated besides tubulin with a mAb to tubulin. VAMP-2 was undetectable, probably because of the small amount precipitated. None of these proteins were detected in the precipitate with normal mouse IgG (nor. mouse IgG). b, the precipitate obtained with a mAb to SNAP-25 was incubated in the sample buffer for 5 min at 95 or 37°C, subjected to SDS-PAGE, and then blotted. The blots were incubated with a mixture of antibodies to synaptotagmin I, tubulin, syntaxin 1, SNAP-25, and VAMP-2 (left lane) or with a mAb to syntaxin 1 (right lane). Note that practically all of the syntaxin 1 precipitated existed as the ternary complex (indicated by an asterisk). c, binding of SNARE/synaptotagmin I complex to purified tubulin-conjugated Sepharose. SNAREs and synaptotagmin I were detected by immunoblotting. The Triton X-100 extract of the lysed P2 fraction (lane 1) and unbound fraction (lane 2) were probed with a mixture of mAbs to synaptotagmin I, syntaxin 1, and SNAP-25. Lanes 3-5, bound fraction was probed with mAbs to synaptotagmin I, syntaxin 1, and SNAP-25, respectively. Note the significant decrease of synaptotagmin I in the unbound fraction. in the molecular motor or in the regulation of tubulin assembly. The binding of MAPs to tubulin C termini promotes stabilization and bundling of microtubules (52). To see whether this is also the case for synaptotagmin I we examined the effect of synaptotagmin I on tubulin polymerization by measuring turbidity (absorbance at 340 nm). Tubulin polymerization was induced by addition of GTP at 30°C. Fig. 6a illustrates a typical example of such experiments. In the presence of Ca 2ϩ , turbidity of tubulin alone changed little, indicating very little polymerization, in agreement with the well known inhibition of tubulin polymerization by Ca 2ϩ (53). Its turbidity steadily increased in the absence of Ca 2ϩ . The co-presence of wild-type synaptotagmin I increased turbidity more than tubulin alone. Importantly, unlike tubulin alone, Ca 2ϩ further increased turbidity. However, when the YVK/AAA mutant of synaptotagmin I was used, enhancement of tubulin polymerization by Ca 2ϩ was not observed; Ca 2ϩ strongly inhibited polymerization as in the case of tubulin alone. This finding is consistent with the poor binding of the YVK/AAA mutant to tubulin (see Fig. 3). Formation of microtubules was analyzed by electron microscopy (Fig. 6b). Compared with normal microtubules in the absence of Ca 2ϩ , those in the presence of Ca 2ϩ were rare, short, and incompletely assembled. Microtubules tended to be bundled in the presence of both synaptotagmin I and Ca 2ϩ . These results show that synaptotagmin I facilitated tubulin polymerization and microtubule bundling in the presence of Ca 2ϩ .
Binding of Synaptotagmin I to Microtubules-The results described above indicate that synaptotagmin I, like MAPs, can also bind to microtubules. We determined whether synaptotagmin I/tubulin binding was present even after tubulin polymerization. For this purpose tubulin was subjected to polymerization-depolymerization cycles by heat/cold treatment in the presence of synaptotagmin I (Fig. 7). After polymerization, most synaptotagmin I was recovered in the pellet (microtubules). After depolymerization, synaptotagmin I was partially recovered in the supernatant. Thus synaptotagmin I could bind both free tubulin dimer and tubulins in microtubules.

DISCUSSION
The present study has demonstrated direct, Ca 2ϩ -dependent binding of synaptotagmin I to the cytoskeletal protein tubulin. To our knowledge, this is the first demonstration of direct interaction between the two proteins. The existence of synaptotagmin I/tubulin complex in vivo was supported by its immunoprecipitation from the detergent extract of crude synaptosomes. The stoichiometry of the binding was two synaptotagmin I molecules per one tubulin dimer (composed of one ␣-tubulin and one ␤-tubulin). Our results indicate that the C-terminal region of ␤-tubulin is responsible for synaptotagmin I binding. Thus the C terminus of ␤-tubulin may be able to hold two synaptotagmin I molecules. Alternatively, the C terminus may preferentially bind a homodimer of synaptotagmin I as synaptotagmin I oligomerizes (mainly dimers) in the presence of Ca 2ϩ (54,55).
Our data show that the Ca 2ϩ dependence of the binding mainly derives from Ca 2ϩ binding to the Ca 2ϩ ligands in both C2 domains of synaptotagmin I, similar to the binding of syntaxin 1A, SNAP-25, and phospholipids to synaptotagmin I (50). However, the binding of tubulin needs much higher Ca 2ϩ concentrations. At present, we cannot exclude the possibility that some factor(s) in the cell greatly increase the Ca 2ϩ sensitivity of the interaction between tubulin and synaptotagmin I. The reason for the requirement for high Ca 2ϩ concentrations for the interaction between tubulin and synaptotagmin I is unclear. Tubulin/synaptotagmin I interaction may require all the Ca 2ϩbinding sites in the C2 domains of synaptotagmin I to be filled with Ca 2ϩ ; this would need Ca 2ϩ concentrations of Ͼ1 mM (15). Binding of phospholipids increases the apparent affinity for Ca 2ϩ 1000-fold. In addition, tubulin/synaptotagmin I interaction may require Ca 2ϩ binding to all the low affinity Ca 2ϩbinding sites of tubulin, which needs mM level of Ca 2ϩ (56). Unlike the high affinity sites (see below), the low affinity Ca 2ϩbinding sites in tubulin are not removed by deprivation of the C-terminal regions, indicating that these sites are distant from the C-terminal regions (57). Binding of Ca 2ϩ to these low af- finity sites may allosterically regulate the C-terminal region of ␤-tubulin responsible for synaptotagmin I binding.
Our findings indicate that tubulin interacts with both C2A and C2B domains of synaptotagmin I. Like the binding of syntaxin 1A and phospholipids, the YVK motif in both C2 domains is involved in tubulin binding. This motif, highly conserved among C2 domain-containing proteins, exists as a part of the ␤-sheet of C2 domain and is thought to attract negative charges of binding substances such as syntaxin 1A and phospholipids (16,17). Consistent with this notion, the C-terminal region of ␤-tubulin is very rich in acidic residues. Their negative charges are probably important for its binding to the YVK motif. Our data, together with a previous report (50), show that tubulin, syntaxin 1A, and SNAP-25 bind to similar domains of the synaptotagmin I molecule. Nevertheless, tubulin does not compete with either syntaxin 1A or SNAP-25 for synaptotagmin I as indicated by the immunoprecipitation experiments. Presumably, the binding sites of these three proteins only partially overlap.
The C-terminal portions of both ␣and ␤-tubulins are exposed to the surface of microtubules (58). The C terminus of ␤-tubulin binds MAPs and tau more strongly than that of ␣-tubulin (52) and regulates vinblastine-induced tubulin polymerization (59). Thus it probably regulates assembly and disassembly of tubulins. Usually Ca 2ϩ inhibits tubulin polymerization directly by high affinity binding to the C-terminal regions of ␣and ␤-tubulin (56). However, the present study has shown that synaptotagmin I also binds to the same region of ␤-tubulin and promotes tubulin polymerization in the presence of Ca 2ϩ . It has previously been shown that tubulin dimer deprived of the C-terminal region of ␤-tubulin (␣␤ s ) by digestion with subtilisin is able to polymerize in mM Ca 2ϩ (60). The removal of the inhibitory effect of Ca 2ϩ on polymerization is due to the loss of high affinity Ca 2ϩ binding to the C-terminal region (57). Binding of synaptotagmin I to the C terminus of ␤-tubulin may block high affinity Ca 2ϩ binding to the C terminus, allowing tubulin polymerization in high Ca 2ϩ concentrations.
Microtubules have rarely been observed in the nerve terminal (61). It is possible that microtubules growing along the axon cease to extend into the nerve terminal due to inhibition by Ca 2ϩ provided by influx through voltage-sensitive calcium channels. However, our findings suggest the possibility that microtubules may be formed and maintained in the nerve terminal by synaptotagmin I binding to tubulin. In fact, microtubules have been observed near the active zone by electron microscopy of the rat brain and the frog neuromuscular junction under certain conditions of fixation (62,63) and immunohistochemically (64). These previous studies have reported synaptic vesicle association with microtubules in the nerve terminal. Microtubules that would be formed and stabilized by synaptotagmin I/tubulin binding might be short-lived due to rapid decline of Ca 2ϩ concentration in the nerve terminal. Consistent with this possibility, some structure containing tubulin was transiently formed near the presynaptic membrane after depolarization of the nerve terminal at the neuromuscular junction of Drosophila larva (65). The Ca 2ϩ -dependent synaptotagmin I/tubulin binding found in this study may underlie such transient structure and be involved in transient retention of synaptic vesicles near the presynaptic membrane. Further studies are required to determine whether microtubules or some other tubulin-related structures are actually formed in the nerve terminal after Ca 2ϩ influx on depolarization. Because synaptotagmin I is restricted to synaptic vesicles, our findings suggest that direct synaptotagmin I/tubulin binding provide a mechanism for retaining synaptic vesicles on microtubules. Previous models postulated that the synaptic vesicle pool in the nerve terminal is formed by attachment of synaptic vesicles to actin fibers through the binding of synapsin (synaptic vesicle protein) to actin fibers (66). Recent studies have shown the presence of two distinct pools of synaptic vesicles in the nerve terminal: readily releasable pool (exo/endo cycling pool) and reserve pool (67,68). The former is close to the active zone, and the latter is distant from the presynaptic membrane. In synapsin I-deficient mice, the number of synaptic vesicles distant (150 -500 nm) from the active zone was decreased, while neither the total number of synaptic vesicles nor the number of synaptic vesicles close (0 -150 nm) to the active zone was affected (69,70). Consistent with this finding, injection of anti-synapsin antibodies into the lampry reticulospinal axon dramatically decreased the number of synaptic vesicles more than 300 nm away from the active zone (67). These studies suggest the involvement of synapsin in the reserve pool. Cytochalasin D, an inhibitor of actin polymerization and depolymerization, eliminated only the reserve pool (68). Therefore, the interaction between synapsin and actin filaments is important for the reserve pool but not for the readily releasable pool. What makes up the readily releasable pool? Our data suggest that synaptotagmin I/tubulin binding retains synaptic vesicles on microtubules even in high Ca 2ϩ concentrations. Because of the localization of voltage-sensitive calcium channels directly involved in neurotransmitter release in close proximity to the active zone (71)(72)(73), the Ca 2ϩ concentration around the active zone is expected to become much higher than elsewhere in the nerve terminal when the channels open. Thus synaptotagmin I/tubulin binding may be involved in the recovery of synaptic vesicles into the readily releasable pool after their exocytosis.
Analyses of Drosophila mutants of Stoned proteins that probably mediate removal of synaptotagmin I from the plasma membrane have revealed a striking decrease in the size of the exo/endo cycling pool and that synaptotagmin I overexpression restored normal levels of endocytotic recycling in these mutants (74). Furthermore, synaptotagmin I overexpression in wild-type flies enhanced synaptic vesicle endocytosis. These findings indicate a direct, essential role of synaptotagmin I in the regulation of synaptic vesicle pool and are consistent with the above possibility that synaptotagmin I is involved in the regulation of the readily releasable pool. As mentioned above, synaptic vesicle attachment to microtubules mediated by synaptotagmin I/tubulin binding in high Ca 2ϩ would be transient due to rapid removal of Ca 2ϩ within the nerve terminal. Synaptic vesicles may be transferred quickly from microtubules to some other structure, which could hold them stably near the presynaptic membrane.
In this study, we have focused on synaptotagmin I, a major form of synaptotagmin in the brain. As recent studies have indicated that different synaptotagmins have different localizations and functions, it will be interesting to examine whether other forms of synaptotagmin (II-XII) also exhibit similar tubulin binding.
In conclusion, we have demonstrated direct, Ca 2ϩ -dependent, stoichiometric interaction between the synaptic vesicle protein synaptotagmin I and the cytoskeletal protein tubulin. This interaction may be involved in the regulation of synaptic vesicle distribution within the nerve terminal.