Ca 2 (cid:49) /Calmodulin Causes Rab3A to Dissociate from Synaptic Membranes*

The GTPase Rab3A has been postulated to cycle on and off synaptic membranes during the course of neurotransmission. Moreover, a Rab guanine nucleotide dissociation inhibitor has been shown to cause Rab3A to dissociate from synaptic membranes in vitro . We dem- onstrate here that Ca 2 (cid:49) /calmodulin also can cause Rab3A to dissociate from synaptic membranes in vitro . Like Rab guanine nucleotide dissociation inhibitor, it forms a 1:1 complex with Rab3A that requires both the lipidated C terminus of Rab3A and the presence of bound guanine nucleotide. In addition, a synthetic pep- tide corresponding to the Lys 62 -Arg 85 sequence of Rab3A can prevent the dissociating effect of each protein and disrupt complexes between each protein and Rab3A. However, Ca 2 (cid:49) /calmodulin’s effect differs from that of Rab guanine nucleotide dissociation inhibitor not only in being Ca 2 (cid:49) -dependent but also in having a less stringent requirement for GDP as opposed to GTP and in involving a less complete dissociation of Rab3A. The functional significance in vivo of Ca 2 (cid:49) /calmodulin’s effect remains to be determined; it may depend in part on the relative amounts of Ca 2 (cid:49) /calmodulin and Rab guanine nucleotide dissociation inhibitor that are available for binding to Rab3A in individual, activated nerve termini. The opening voltage-gated Ca 2 in active zones of brief, localized of g av , and the supernatants and membrane pellets were subjected to SDS-PAGE and Western blot analysis with anti- Rab3A antibodies. The Rab3A bands from supernatant ( S ) and pellet ( P ) fractions are compared for four different incubation conditions: lane 1 , 100 (cid:109) M CaCl 2 and 1 m M ATP, 1 m M MgCl 2 ; lane 2 , 100 (cid:109) M CaCl 2 and 60 (cid:109) M CaM; lane 3 , 100 (cid:109) M CaCl 2 , 60 (cid:109) M CaM, and 1 m M ATP, 1 m M MgCl 2 ; and lane 4 , 60 (cid:109) M CaM and 10 m M EGTA. Note that ATP appeared to enhance the Ca 2 (cid:49) /CaM-dependent dissociation of Rab3A from REM in this experiment but that ATP had no statistically signif- icant effect in three experiments.

The opening of voltage-gated Ca 2ϩ channels in active zones of nerve terminals causes a brief, localized influx of Ca 2ϩ followed by the secretion of neurotransmitters (1)(2)(3). The molecular basis of this effect is still unclear, but increased concentrations of intracellular Ca 2ϩ may act at several levels to trigger fast fusion of pre-docked synaptic vesicles with the synaptic plasma membrane, promote endocytosis of the vesicle membranes and subsequent vesicle reformation, and mobilize additional vesicles to release sites (1,4). Proteins that bind Ca 2ϩ probably mediate many of these actions, and a number of candidate proteins have been identified. They include rabphilin (5,6); the ␣-, ␤-II-, and ␥ isoforms of protein kinase C (7); and dynamin (8), all of which show Ca 2ϩ -dependent binding to acidic phosphoglycerides. They also include calmodulin (CaM) 1 (9), synaptotagmin (10 -13), and calcineurin (14), which bind Ca 2ϩ directly. CaM that contains bound Ca 2ϩ (Ca 2ϩ /CaM) can activate CaM kinase II and calcineurin (9), and both enzymes may play important regulatory roles (15)(16)(17)(18). Furthermore, Ca 2ϩ /CaM appears to be required for secretion in adrenal chromaffin cells (19 -21). However, how Ca 2ϩ -and Ca 2ϩ /CaM-dependent reactions are integrated to promote and optimize synaptic responses remains to be determined.
In the present investigation we examined the effects of Ca 2ϩ and CaM on the behavior of Rab3A, a low molecular mass, di-geranylgeranylated, guanine nucleotide-binding protein that is attached to neurotransmitter-containing synaptic vesicles (22,23). Previous investigators had shown that depolarization of rat brain synaptosomes causes a reduction in the contents of both Rab3A and a related guanine nucleotide-binding protein, Rab3C, in crude synaptic vesicles (Refs. 24 and 25 but see Ref. 26 for a conflicting view). Furthermore, action of a Rab guanine nucleotide dissociation inhibitor protein (Rab GDI) had been implicated because of its known ability to form a complex with Rab3A and cause it to dissociate from synaptic membranes in cell-free experiments (27)(28)(29). While exploring the possibility that increased concentrations of Ca 2ϩ might affect the Rab3A dissociation process, we discovered that Ca 2ϩ / CaM also can cause Rab3A to dissociate from synaptic membranes. Studies of the mechanism of this effect and its relation to that of Rab GDI are described below. 2

EXPERIMENTAL PROCEDURES
Materials-CaM was obtained from Calbiochem and freshly dissolved in 50 mM HEPES, pH 7.4, for each experiment. CaCl 2 , Suprapur grade, was from EM Science. BS 3 was obtained from Pierce. Rab3A peptides Lys 62 -Arg 85 , Ala 2 -Asn 18 , and Glu 177 -Asp 195 (Table I) and the Rab GDI peptide, Gly 21 -Ser 45 (GIMSVNGKKVLHMDRNPYYG-GESSS), were synthesized by the University of Washington Biopolymer Facility. The CaM kinase II peptide Leu 290 -Ala 309 (Table I) was from LC Laboratories. Stock solutions of peptides were prepared in Me 2 SO and then added to incubation mixtures at final Me 2 SO concentrations of Ͻ5%. GDP, GTP␥S, and unprenylated Rab3A were from Calbiochem. Rab GDI was purified from bovine brain as described (29), except that all buffers used after the ammonium sulfate precipitation step contained 10% glycerol, 0.25 mM phenylmethylsulfonyl fluoride, 2.5 g/ml each of aprotinin and leupeptin, and 1 g/ml pepstatin A. All other purchased chemicals were reagent grade from Sigma, and all procedures were performed at 4°C unless otherwise indicated.
Preparation of Synaptosomes-Two different methods were used to prepare synaptosomes from cerebral cortex of nonhuman primates (Macaca nemestrina), obtained from the tissue distribution program of the Regional Primate Research Center at the University of Washington. In method 1, 50 g of cortex was sliced in ice-cold buffer A (320 mM sucrose; 2 mM EGTA; 0.1 mM phenylmethylsulfonyl fluoride; 1 g/ml each of leupeptin, aprotinin, and pepstatin A; and 2 g/ml p-aminobenzamidine). Then a gray matter-enriched portion, diluted with 10 volumes of buffer A, was homogenized successively with an Oster blender and a Potter-Elvehjem homogenizer. Finally, a crude synaptosomal pellet (9000 ϫ g av ) (where g av is g force at tube half-length) was prepared and purified on a Ficoll gradient using a modification of the methods of Fisher von Mollard et al. (24) and Barrie et al. (31). In method 2, synaptosomes to be used primarily for depolarization studies were prepared from 15 g of prefrontal and temporal cortex. The tissue was minced with a razor blade in ice-cold buffer A (with 5 mM HEPES, pH 7.4, 0.5 mM EGTA), rinsed twice with buffer, and homogenized in a total volume of 100 ml using a loose-fitting Potter-Elvehjem homogenizer (0.25 mm clearance; Ref. 32). Then a synaptosome-enriched fraction was prepared as described (24).
Preparation of Synaptosomal Lysates and Lysate Subfractions-For most studies synaptosomal lysates were prepared and successively centrifuged to yield a synaptosomal plasma membrane-enriched fraction (28,000 ϫ g av pellet), a crude synaptic vesicle fraction (176,000 ϫ g av pellet), and a corresponding high speed supernatant as described (33) except that EGTA (2 mM) was added to all buffers. The crude synaptic vesicle fraction, which contained Rab3A-enriched membranes (REM), was resuspended in 2-3 ml of buffer containing 300 mM glycine, 0.1 mM EGTA, 1 g/ml aprotinin, 1 g/ml leupeptin, and 5 mM HEPES, pH 7.4. Then aliquots were flash-frozen and stored at Ϫ70°C. Both freshly prepared and freeze-thawed REM gave similar results. The 176,000 ϫ g av supernatant was concentrated 10-fold with a PM10 membrane (Amicon), and aliquots were prepared and frozen as described above.
Rab Dissociation Assays and Quantitative Densitometry-REM (15 g of protein) were incubated for 30 min at 30°C with various additives in 50 l of buffer B (50 mM HEPES, pH 7.4, containing 0.1 mM CaCl 2 , 0.5 M MgCl 2 , 1 mM DTT, 2 g/ml aprotinin, and 2 g/ml leupeptin). For specific incubation conditions, refer to the figure legends. The reaction mixtures were centrifuged for 30 min at 100,000 ϫ g av in a Beckman TLA 45 rotor. The pellets were resuspended in 50 l of buffer B (but without Ca 2ϩ ) with brief sonication. For quantitation of Rab3A, aliquots of each pellet and supernatant fraction (5 and 7 l, respectively) were examined by SDS-PAGE and Western blot analysis with the use of anti-Rab3A antibodies (see below). Corresponding samples of untreated membranes (0.15, 0.3, 0.6, 1.2, 1.8, and 2.4 g of protein/sample) were used as standards. For quantitation of Rab1A and Rab5A, 10-and 12-l aliquots of pellet and supernatant fractions were used, respectively; the samples of untreated membranes used as standards were 0.5, 1.0, 1.5, 2.5, 3.0, 3.5, and 4.0 g of protein/sample. The Western blots were scanned with a Bio-Rad model GS-670 imaging densitometer, and the absorbance values obtained for the Rab protein bands were converted to g of equivalence of Rab proteins on the basis of the standard curves that were generated. The amount of each Rab protein dissociated from the REM was reported as a percentage of the total Rab protein recovered in the supernatant and membrane fractions. Unless indicated otherwise, data points shown in the figures represent single or duplicate measurements. In general, results of replicate measurements varied by no more than 2%. For Rab3A, the number of experiments represented by each figure is reported in the figure texts. For Rab1A and Rab5A, two dissociation experiments were performed and gave similar results.
Antibodies-Rabbit polyclonal antibodies were raised against peptides corresponding to amino acids Ala 2 -Asn 18 of the cDNA-predicted sequence of bovine brain Rab3A (34) as described previously (35) except that a C-terminal cysteine was added when the peptide was synthesized to facilitate conjugation to keyhole limpet hemocyanin (Pierce). A 1/500 dilution of the affinity purified antibody was used for most anti-Rab3A Western analyses. However, for Western analyses in cross-linking experiments with Rab3A peptide Ala 2 -Asn 18 (see below), a 1/500 dilution of a Rab3A-specific antibody that had been raised against Rab3A peptide Lys 202 -Asp 217 (Santa Cruz Biotechnology) was used. The anti-Rab1A and anti-Rab5A antibodies (Santa Cruz Biotechnology) were each used at a 1/125 dilution. Antibodies were also prepared against peptides corresponding to amino acids Gly 21 -Ser 45 of the cDNA-predicted sequence of bovine brain Rab GDI (36). A 1/200 dilution of this affinity purified antibody was used for all anti-Rab GDI Western analyses.
Antibody specificity was determined by two-dimensional gel electrophoresis (37, 38) followed by Western analysis (35). In the case of Rab3A, 20 g of of purified synaptic vesicle protein (33) were analyzed, and only one immunoreactive band, representing Ͼ95% of the antibodyreactive material, was found. In the case of Rab GDI, 100 g of nonhu-man primate brain cytosol (9200 ϫ g av ϫ 20 min supernatant as described (33)) were similarly analyzed and found to contain two adjacent proteins (representing Ͼ90% of the antibody-reactive material) that most likely represented two charged isoforms of Rab GDI (39).
Sucrose Density Gradients-The apparent molecular mass of the Rab3A⅐CaM complex was determined by sucrose gradient centrifugation (27) modified as follows. Supernatants from Rab3A dissociation experiments with 90 g of REM, 100 M CaCl 2 , and 50 M CaM in 250 l of buffer B were overlaid onto a 5-ml, 5-25% continuous sucrose gradient containing 5 mM MgCl 2 , 1 mM DTT, 0.1 mM CaCl 2 , 50 mM HEPES, pH 7.4, and protease inhibitors as described for buffer B. After centrifugation for 7 h at 173,000 ϫ g av in a Beckman SW50.1 rotor, fractions were analyzed for Rab3A with an immuno-dot blot assay. The high speed supernatant fraction from synaptosomal lysates (1 mg/250 l) was similarly analyzed.
Cross-linking Studies-For cross-linking experiments, REM (15-25 g of protein) were washed twice by centrifugation in 50 mM HEPES, pH 7.4, containing 100 mM NaCl, 1 mM MgCl 2 , 1 mM DTT, and protease inhibitors as described for buffer B. The washed membranes were mixed with Ca 2ϩ /CaM in 25 l of buffer B and incubated for 30 min at 30°C as described in the figure legends. The incubation mixtures were centrifuged for 30 min at 100,000 ϫ g av ; the supernatants were treated for 30 min at 30°C with freshly prepared 1 mM BS 3 (40), and the reactions were quenched with Tris buffer and analyzed as described above (41). A similar BS 3 treatment procedure was used in all other cross-linking experiments.
Radiolabeling of CaM and Ca 2ϩ / 125 I-CaM-binding Experiments-CaM (225 g) in 50 l of 50 mM HEPES buffer, pH 7.6, was modified by reaction with 0.5 mCi of 125 I-labeled Bolton-Hunter reagent (2200 Ci/ mmol; NEN Life Science Products) according to the manufacturer's directions. The 125 I-labeled CaM was then separated from unreacted reagent with the use of a Bio-Gel P-6 DG (Bio-Rad) spin column as described (42). After centrifugation, a mixture of 50 l of 125 I-labeled CaM (10 nmol) and 50 l of REM (450 g) was incubated for 45 min at 37°C in the presence of 100 M CaCl 2 and 0.1 mM GDP. After incubation the mixture was centrifuged for 30 min at 100,000 ϫ g av , half of the supernatant was treated with BS 3 , and half was reserved as control. An aliquot of each half (50 l) was then immunoprecipitated with 50 l of anti-Rab3A IgG (1 g) for 24 h at 0°C. After immunoprecipitation, 40 l of immobilized protein A on Trisacryl beads (Pierce; 50% slurry), which had been pretreated with bovine serum albumin, were added, and the sample was mixed for 20 h at 4°C in a tube rotator. The protein A beads were then washed four times by a procedure that involved suspension in 50 mM HEPES, pH 7.4, containing 100 mM NaCl, 0.1 mM CaCl 2 , 0.1% Tween 20, 0.1 mM DTT, and 1 g/ml each of leupeptin and aprotinin, followed by centrifugation for 4 min at 350 ϫ g av . The washed beads were boiled for 4 min in 30 l of 2% SDS and prepared for SDS-PAGE analysis (described below).
Preparation of Rab3A-depleted Synaptic Membranes and Transfer of Rab3A to These Membranes-To generate Rab3A-depleted membranes, REM (30 g) were incubated for 30 min at 30°C with 1.6 M Rab GDI in 50 l of buffer B (but without Ca 2ϩ ), pelleted by centrifugation for 30 min at 100,000 ϫ g av , and suspended in 5 l of buffer B (but without Ca 2ϩ ). To study the transfer of Rab3A to these membranes, they were mixed with medium containing Rab3A⅐Ca 2ϩ /CaM complex (prepared by incubating REM with buffer containing 75 M CaM and 100 M CaCl 2 ) and incubated for 30 min at 30°C in the presence or absence of one of the peptides listed in Table I. Then the incubation mixtures were subfractionated by centrifugation and analyzed as described above.
Depolarization Assays-The ability of synaptosomes to secrete glutamate was measured after KCl-induced depolarization (24) or treatment with 4-aminopyridine ϩ 4␤-phorbol dibutyrate (31). A modification of the method previously described (24) was used. Briefly, control and depolarized samples were incubated for 10 min at 37°C and then placed in ice water for 2 min and centrifuged for 2 min at 13,000 ϫ g av . The amount of NADPH that had been produced was determined by measuring the absorbance of the supernatant at 360 nm on a Beckman DU 640 spectrophotometer using 390 nm as the reference wavelength.
Other Methods-SDS-PAGE was performed as described by Laemmli (43) but with 14% gels. In some cases, EGTA was added to the samples just before they were boiled. Proteins were transferred from SDS-PAGE gels to Immobilon P membranes for 30 min at 80 V for Western analysis of Rab3A or transferred for 60 min to identify cross-linked products of higher molecular mass. Immunoblots were performed as described (35). Protein concentrations were determined with the Bradford method (44) (Bio-Rad) or, for SDS-containing samples, the micro-BCA method (Pierce). Free calcium ion concentrations were varied in the presence of 1 mM EGTA, on the basis of established binding constants (45).

Ca 2ϩ /CaM Causes Rab3A to Dissociate from Membranes-
The initial aim of this investigation was to determine whether Ca 2ϩ and CaM influence the dissociation of Rab3A from synaptic membranes. To examine this possibility, we isolated synaptosomes from samples of macaque cerebral cortex, lysed the synaptosomes in hypotonic medium, and prepared Rab3A-enriched membranes (REM) from the lysates by ultracentrifugation. Then we suspended the REM in medium containing Ca 2ϩ and/or various other additives, incubated the mixtures for 30 min at 30°C, separated the membranes from the medium by centrifugation, and separately measured the amounts of Rab3A recovered in the membrane and supernatant fractions. The results of these experiments demonstrated that medium containing both Ca 2ϩ and CaM, i.e. a Ca 2ϩ /CaM complex, caused Rab3A to dissociate from the membranes but that medium containing either 100 M Ca 2ϩ or 60 M CaM alone did not ( Fig.  1). The dissociation of Rab3A occurred in the absence of added ATP, and half-maximal effects were observed when the concentrations of Ca 2ϩ and CaM were about 0.5 and 20 M, respectively. Maximal dissociation of Rab3A (approximately 65%) was obtained when the concentrations of Ca 2ϩ and CaM were about 10 and 65 M, respectively (data not shown).
Mechanism of the Effects of Ca 2ϩ /CaM-To examine the mechanism of the Rab3A-dissociating effect of Ca 2ϩ /CaM, we first incubated REM in the presence of medium that contained Ca 2ϩ /CaM and then recovered the medium and subfractionated it by sucrose gradient ultracentrifugation. Upon measuring the content of Rab3A in the subfractions, we detected a peak of material that had an apparent molecular mass of about 40 kDa ( Fig. 2A). This peak could be distinguished easily from the peak of Rab3A-containing material detected in sucrose gradient ultracentrifugation experiments with a high speed supernatant fraction from a synaptosomal lysate ( Fig. 2A). The peak from the lysate supernatant had a considerably larger apparent molecular mass and probably corresponded to a complex of Rab3A and Rab GDI (28).
In subsequent studies we incubated REM with Ca 2ϩ /CaM, recovered the medium, and added the cross-linking agent, BS 3 , to it. Then we analyzed the cross-linked material by SDS-PAGE and Western blotting with an antibody to Rab3A. In agreement with the sucrose gradient experiments, the results revealed the presence of Rab3A-containing material that had an apparent molecular mass of 43 kDa (Fig. 2B, lane 2). The combined results of these experiments suggested that Ca 2ϩ / CaM could form a 1:1 molar complex with Rab3A.
To obtain additional evidence concerning this possibility, we incubated REM with Ca 2ϩ /CaM that contained 125 I-labeled CaM, recovered the medium, and added BS 3 to it. Then we immunoprecipitated Rab3A-containing material, analyzed it by SDS-PAGE, and identified 125 I-CaM-containing bands by Ca 2ϩ /CaM is known to form complexes with many different proteins, and generally similar mechanisms appear to be involved. When Ca 2ϩ binds to CaM, it induces a conformational change in CaM that exposes binding sites for both hydrophobic and basic amino acids (46 -48). To obtain further evidence that the 1:1 complex of Rab3A and CaM involves Ca 2ϩ /CaM, we incubated REM with Ca 2ϩ and CaM, recovered the medium, and incubated aliquots of it in the presence of different concentrations of EGTA. Then we treated the incubation mixtures with BS 3 and analyzed the cross-linked products by SDS-PAGE. The results revealed that incubation of the medium in the presence of 1 mM EGTA decreased the electrophoretic mobility of the Rab3A⅐CaM complex (apparent molecular mass 43 3 50 kDa) (Fig. 3, compare lanes 2 and 3), whereas incubation with 10 mM EGTA greatly reduced the amount of the complex that could be detected (Fig. 3, compare lanes 2 and 4). The binding of Ca 2ϩ to CaM is known to increase its electrophoretic mobility (49; see also Fig. 4, lanes 8 and 9 and Fig. 7B, lanes 1  and 2). Therefore, both results provided evidence for the formation of a Rab3A⅐Ca 2ϩ /CaM complex.
To investigate the role of hydrophobic interactions in forming the Rab3A⅐Ca 2ϩ /CaM complex, we used two different approaches. First, we incubated REM with Ca 2ϩ /CaM, recovered the medium, and incubated aliquots of it in the presence or absence of the neutral detergent, Triton X-100. Then we added BS 3 to the incubation mixtures and analyzed the cross-linked products by SDS-PAGE. Treatment with 0.1% Triton X-100 completely disrupted the complex (data not shown). In the second approach we incubated unmodified, recombinant Rab3A with Ca 2ϩ /CaM, then added BS 3 , and analyzed the cross-linked material. The fact that no complex of recombinant Rab3A with Ca 2ϩ /CaM could be detected (Fig. 4, compare lanes 5 and 8) provided evidence that the lipidated C terminus of Rab3A is required for binding to Ca 2ϩ /CaM.
Among the many proteins that form complexes with Ca 2ϩ / CaM is CaM kinase II (50). Furthermore, a basic-and hydrophobic amino acid-containing binding site for Ca 2ϩ /CaM on this kinase, Leu 290 -Ala 309 (Table I), has been identified (50,51). Rab3A also contains a sequence that is enriched in basic and hydrophobic amino acids, Lys 62 -Arg 85 . To investigate the possibility that the Rab3A Lys 62 -Arg 85 sequence might include a binding site for Ca 2ϩ /CaM, we synthesized a peptide that cor-responded to it (Table I) and then compared the effects of this peptide with those of other synthetic peptides in the following experiments.
First, we incubated REM with Ca 2ϩ /CaM in the presence of each peptide and then measured the amount of Rab3A that dissociated to the medium. The results revealed that a 100 M concentration of the Rab3A Lys 62 -Arg 85 peptide or of a peptide corresponding to the CaM kinase II Leu 290 -Ala 309 sequence blocked the Rab3A-dissociating effect of Ca 2ϩ /CaM (Fig. 5A); half-maximal values were observed at concentrations of 42 and 18 M, respectively. In contrast, 100 M concentrations of peptides that respectively corresponded to regions near the Rab3A N terminus or unmodified C terminus (Rab3A Ala 2 -Asn 18 or Glu 177 -Asp 195 ; Table I) had no effect.
Second, after incubating REM with Ca 2ϩ /CaM, we recovered the medium and incubated aliquots of it with the Rab3A Lys 62 -Arg 85 peptide, the CaM kinase II Leu 290 -Ala 309 peptide, or the Rab3A Ala 2 -Asn 18 peptide, then added BS 3 to each incubation mixture, and analyzed the content of cross-linked Rab3A⅐Ca 2ϩ / CaM complex. Both the Rab3A Lys 62 -Arg 85 peptide and the CaM kinase II Leu 290 -Ala 309 peptide reduced the amount of complex that could be detected, but the Rab3A Ala 2 -Asn 18 peptide had no effect (Fig. 6A).
Third, we incubated the Rab3A Lys 62 -Arg 85 , Ala 2 -Asn 18 -, or Glu 177 -Asp 195 peptides with CaM in the presence of Ca 2ϩ or EGTA, then added BS 3 , and analyzed the products by SDS-PAGE. The results showed that the Lys 62 -Arg 85 peptide could form a Ca 2ϩ -dependent complex with CaM but the other Rab3A peptides could not (Fig. 7). Taken together, the results of these experiments provided strong evidence that Ca 2ϩ /CaM promotes the dissociation of Rab3A from synaptic membranes by binding to amino acids within the Rab3A Lys 62 -Arg 85 sequence.
Transfer of Rab3A from Rab3A⅐Ca 2ϩ /CaM to Membranes-Having shown that the Rab3A Lys 62 -Arg 85 and CaM kinase II Leu 290 -Ala 309 peptides could separately disrupt Rab3A⅐Ca 2ϩ / CaM complexes, we examined the possibility that disruption of the complexes might promote the transfer of Rab3A to membranes. We did this by incubating REM with Ca 2ϩ /CaM to generate a Rab3A⅐Ca 2ϩ /CaM complex, recovering the medium, and incubating aliquots of it with one or the other of the peptides in the presence of Rab3A-depleted synaptic membranes. After the incubations we separated the membranes from the medium by centrifugation and measured the contents of Rab3A in the pellet and supernatant fractions. The results demonstrated that each peptide could cause Rab3A to translocate from the medium to the membranes, whereas a control peptide had no effect (Fig. 8).
Relation between the Effects of Ca 2ϩ /CaM and Those of Rab GDI-Rab GDI can form a 1:1 complex with digeranylgeranylated Rab3A and cause it to dissociate from synaptic membranes (22,24). To examine the relation between this effect and that of Ca 2ϩ /CaM, we first sought to determine whether the two proteins bind to similar sites on Rab3A. In one set of experiments we incubated REM with Ca 2ϩ /CaM to generate a soluble Rab3A⅐Ca 2ϩ /CaM complex or used the high speed supernatant from a synaptosomal lysate as a source of Rab3A-Rab GDI. Then we incubated the Rab3A⅐Ca 2ϩ /CaM complex for 30 min at 30°C with Rab GDI or incubated the synaptosomal lysate supernatant under similar conditions with Ca 2ϩ /CaM.
Following the incubations, we treated the incubation mixtures with BS 3 and analyzed the Rab3A-containing material (Fig. 9). The results of these experiments showed that Ca 2ϩ /CaM and Rab GDI could compete with each other for binding to Rab3A. In a second set of experiments we incubated REM with Rab GDI in the presence of the Rab3A Lys 62 -Arg 85 peptide, the CaM kinase II Leu 290 -Ala 309 peptide, the Rab3A Ala 2 -Asn 18 peptide, or the Rab3A Glu 177 -Asp 195 peptide. Then we recovered the medium and measured the content of Rab3A in the supernatant by immunoblotting. The Rab3A Lys 62 -Arg 85 and CaM kinase II Leu 290 -Ala 309 peptides separately blocked the Rab GDI-dependent dissociation of Rab3A from REM (Fig. 5B); half-maximal values were observed at concentrations of 46 and 41 M, respectively (not shown). In contrast, the Rab3A Ala 2 -

Rab3A Dissociation from Membranes
Asn 18 and Glu 177 -Asp 195 peptides were inactive (Fig. 5B). In a third set of experiments we incubated REM with Rab GDI, isolated the medium, and incubated it for 30 min at 30°C in the presence or absence of the Rab3A Lys 62 -Arg 85 peptide or the Rab3A Ala 2 -Asn 18 peptide. Then we added BS 3 to the incubation mixtures and analyzed the Rab3A-containing cross-linked material (Fig. 6B). The Rab3A Lys 62 -Arg 85 peptide reduced the amount of cross-linked Rab3A⅐GDI complex recovered but the Rab3A Ala 2 -Asn 18 peptide did not. The similarity between these results and those obtained in the corresponding experiments with Ca 2ϩ /CaM (compare Fig. 5, A with B, and Fig. 6, A with B) provided strong evidence that Rab GDI and Ca 2ϩ /CaM interact with similar binding sites within the Rab3A Lys 62 -Arg 85 sequence.
The Rab GDI-induced dissociation of Rab3A from membranes is known to be under the control of guanine nucleotides (27). To determine whether guanine nucleotides also control the Ca 2ϩ /CaM-induced dissociation of Rab3A, we preincubated REM for 1 h at 37°C in the absence of added guanine nucleotides, treated the REM with Ca 2ϩ /CaM or Rab GDI, then measured the amount of Rab3A that dissociated from the REM to the medium (Fig. 10). Only a small response to Ca 2ϩ /CaM or Rab GDI could be detected (compare preincubated samples with nonpreincubated controls). In contrast, Rab3A dissociated from REM that had been preincubated in the presence of GDP or GTP␥S before being treated with Ca 2ϩ /CaM or Rab GDI, and GDP was more effective than GTP␥S. Notice that Rab GDI had a more stringent requirement for GDP than Ca 2ϩ /CaM did.
The mechanism of these guanine nucleotide-dependent effects remains to be determined. However, the conformation of Ras-GDP is known to differ from that of Ras-GTP (52), and guanine nucleotides are presumed to have similar effects on Rab proteins and other Ras-related proteins. Furthermore, a series of experiments with BS 3 -treated REM provided direct evidence that guanine nucleotides alter the conformation of Rab3A (Fig. 11). First, analyses of untreated control REM by SDS-PAGE and Western blotting revealed a single major band of Rab3A that had an apparent molecular mass of about 28 kDa, but analyses of BS 3 -treated control REM revealed two major, Rab3A-containing bands that had respective apparent molecular masses of about 28 and 23 kDa (Fig. 11, compare  lanes 1 and 3). Second, REM that had been preincubated for 1 h at 37°C in the absence of guanine nucleotides before being treated with BS 3 contained an increased amount of the 28-kDa band but a reduced amount of the 23-kDa band (Fig. 11, compare lanes 3 and 4). Third, REM that had been preincubated in the presence of GDP or GTP␥S before being treated with BS 3 showed a distribution of 23-and 28-kDa bands which resembled that in BS 3 -treated control membranes (Fig. 11, compare  lanes 3, 5, and 6). These results suggest that treatment with BS 3 can stabilize a guanine nucleotide-dependent conformation of Rab3A that has an increased electrophoretic mobility. Thus, it seems reasonable to postulate that the Rab3A of control REM or REM that have been incubated in the presence of added GDP or GTP␥S contains bound guanine nucleotides but that these nucleotides dissociate from Rab3A when REM are incubated in the absence of added GDP or GTP␥S or when Rab3A is analyzed by SDS-PAGE (see also Ref. 53). Furthermore, it seems likely that Rab3A that contains bound GDP or GTP has a more compact conformation than guanine nucleotide-free Rab3A does and that this compact conformation can be stabilized by BS 3 -dependent, intramolecular cross-linking reactions.

DISCUSSION
The interaction of Ca 2ϩ /CaM with Rab3A resembles that of Rab GDI with Rab3A in several respects. Both proteins form soluble 1:1 complexes with Rab3A and cause it to dissociate from synaptic membranes. Formation of each of the complexes requires both the lipidated C terminus of Rab3A and the presence of guanine nucleotides. Both Ca 2ϩ /CaM and Rab GDI evidently bind to sites within the Rab3A Lys 62 -Arg 85 sequence.
The interactions of Ca 2ϩ /CaM and Rab GDI with Rab3A also differ in several respects. Importantly, the interaction of Ca 2ϩ / CaM with Rab3A depends on Ca 2ϩ . In addition, half-maximal effects of Ca 2ϩ /CaM on the dissociation of Rab3A from REM or the dissociation of Rab3A from Rab3A-Rab GDI occur at concentrations of Ca 2ϩ /CaM (ϳ20 M) that are much higher than the concentrations of Rab GDI required to dissociate Rab3A from REM or Rab3A-Ca 2ϩ /CaM (Ͻ0.5 M; data not shown). The Ca 2ϩ /CaM-dependent dissociation of Rab3A from REM is less extensive than the Rab GDI-dependent dissociation of Rab3A. And the Rab3A-dissociating effect of Ca 2ϩ /CaM has a less stringent requirement for GDP than does that of Rab GDI.
The precise mechanism of the Rab3A-dissociating effect of Ca 2ϩ /CaM remains to be determined, but it is of interest that the Rab3A Lys 62 -Arg 85 peptide contains a cluster of basic amino acids toward its N terminus, while its hydrophobic amino acids are more evenly distributed (Table I). Furthermore, a helical wheel projection of the peptide's sequence suggested that the clustered, basic amino acids may be located on one side of an amphipathic helix (not shown). The CaM kinase II Leu 290 -Ala 309 peptide has similar characteristics (Table I  and Ref. 47); and a recent crystallographic study of its interaction with Ca 2ϩ /CaM has shown that the latter can "wrap around" the peptide to make close contact with its basic and hydrophobic amino acids (48). Ca 2ϩ /CaM may conceivably in-teract with the Rab3A peptide in the same way. However, Ca 2ϩ /CaM may interact differently with native Rab3A because its binding to the protein appears to require the presence of the modified C terminus. Molecular modeling studies of Rab3A might provide some insight into this issue.
Modeling studies of the GDP-bound form of human Rab5A have suggested that the Rab5A Gln 60 -His 83 sequence, QTV-CLDDTTVKFEIWDTAGQEGYH, which is homologous to the Rab3A Lys 62 -Arg 85 sequence, may be partially exposed on the protein's surface (54). The cluster of hydrophilic amino acids toward the N terminus of the Rab5A Gln 60 -His 83 sequence occupies an exposed position adjacent to loop 3 of the Rab5A molecule, but the hydrophobic amino acids of the sequence are generally much less exposed and interact with other amino acids in the protein. If modeling studies of Rab3A suggest that the amino acids of the Lys 62 -Arg 85 sequence (Table I) occupy similar positions related to the protein's surface, the possibility that the clustered basic amino acids in the sequence may be available for binding to Ca 2ϩ /CaM would be worth examining.
Experimental tests of the role of individual basic amino acids in the sequence might be done by site-directed mutagenesis. A similar approach has been used to examine the regulatory role of amino acids in the Rab3A Asp 77 -Glu 82 sequence, which corresponds to the G2 guanine nucleotide-binding region. A Gln 81 3 Leu mutation altered the k off (GDP) and k off (GTP) of Rab3A and greatly reduced the ability of Rab3A to respond to Rab3A guanine nucleotide releasing factor (55). In addition, a recent study of a Rab6-v-Ha-Ras chimera showed that the Rab6 Arg 60 -Trp 67 sequence (-RTVRLQLW-), which is homologous to the Rab3A Lys 69 -Trp 76 sequence (-KRIKLQIW-), includes binding sites for Rab GDI and Rab geranylgeranyl transferase (56).
It might also be of interest to examine the effects of sitedirected mutagenesis within the corresponding regions of several Rab proteins. Thus, the first portion of the Rab3A Lys 62 -Arg 85 sequence contains five clustered basic amino acids, but the first portions of the corresponding sequences of Rab1A and Rab5A contain one and three basic amino acids, respectively. Furthermore, these differences may correlate with differences in the dissociating effects of Ca 2ϩ /CaM on the three Rab proteins. In a representative experiment, we incubated REM (from synaptosomes prepared using method 2; see "Experimental Procedures") for 30 min at 30°C in the presence of 1 mM GDP and 80 M Ca 2ϩ /CaM. Measurements by quantitative densitometry revealed that this caused the dissociation of 55% of the membrane-bound Rab3A but only 10% of the Rab1A and 20% of the Rab5A (data not shown). Therefore, mutation experiments designed to alter the number and/or distribution of basic amino FIG. 11. Evidence that Rab3A undergoes a conformational change in the absence of guanine nucleotides. REM were preincubated for 60 min at 37°C in the absence of added guanine nucleotides or in the presence of added GDP (1 mM) or GTP␥S (1 mM). The samples were then treated with BS 3 and analyzed as described in Figs. 1 and 2. Note that the cross-linking of Rab3A by BS 3 under these conditions is probably incomplete and that the identity of the cross-linked Rab3Acontaining bands of Ն46 kDa is unknown. acids within the Lys 62 -Arg 85 -like regions of these proteins might be informative.
Interestingly, the GTPase Rad, which was recently shown to bind Ca 2ϩ /CaM by a GDP-dependent mechanism, contains an unprenylated C-terminal sequence, Lys 279 -Lys 308 , that is enriched in basic amino acids (57). Moreover, both selective truncation experiments involving this sequence and experiments with synthetic peptides have provided evidence that this sequence contains the binding site for Ca 2ϩ /CaM. The functional significance of the binding of Ca 2ϩ /CaM to Rad has yet to be determined.
The myristoylated, alanine-rich, protein kinase C substrate, MARCKS, also can form a complex with Ca 2ϩ /CaM (58,59). MARCKS contains a region, enriched in basic amino acids, that promotes its binding to vesicles containing acidic phosphoglycerides. In vitro experiments have shown that MARCKS dissociates from such vesicles when serine residues in this region are phosphorylated by protein kinase C (60,61). Alternatively, Ca 2ϩ /CaM may cause MARCKS to dissociate from membranes (60). Ca 2ϩ /CaM apparently binds to the unmodified region because its affinity for MARCKS is greatly reduced by the same protein kinase C-dependent phosphorylation reactions (58).
How important is the Rab3A-dissociating effect of Ca 2ϩ /CaM likely to be during neurotransmission in vivo? The answer to this question may depend on the amounts of Ca 2ϩ /CaM and Rab GDI that are available for binding Rab3A in activated nerve terminals. The amount of Ca 2ϩ /CaM available for binding is likely to be a complex function not only of the total, local concentration of Ca 2ϩ /CaM, but also of the rates and affinities of Ca 2ϩ /CaM binding to other nerve-terminal proteins including CaM kinase II (see particularly Ref. 62). The amount of Rab GDI available for binding to Rab3A presumably depends in part on the concentration of unbound Rab GDI in the nerve terminal, i.e. that fraction of the total, soluble Rab GDI that contains no bound Rab proteins. The amounts of available Ca 2ϩ /CaM and Rab GDI remain to be determined. However, the concentration of Ca 2ϩ surrounding activated Ca 2ϩ channels in nerve terminals may be as high as 100 M (63), the content of total CaM in brain may be ϳ500 mg/kg (ϳ30 M; Ref. 64), and the concentration of Rab GDI in brain may be ϳ270 mg/kg (ϳ5 M; Ref. 65).
It is noteworthy that some regions of the brain contain a membrane-associated protein called neuromodulin, GAP43, or B50 (66 -69), that binds CaM in vitro in the absence of Ca 2ϩ . Importantly, neuromodulin releases CaM in the presence of high concentrations of Ca 2ϩ or when one of its serine residues, Ser 41 , is phosphorylated by protein kinase C (70,71). Furthermore, neuromodulin has been shown to be attached to the inner surface of the plasma membrane in rat frontal cortex nerve terminals (72), has been implicated in the control of neurotransmission (73), and may be phosphorylated by protein kinase C in activated synaptosomes (74). Therefore, it is possible that activation of nerve terminals may be followed by release of bound CaM from neuromodulin and that the released CaM may increase the local amount of CaM available for Ca 2ϩ -dependent binding to nerve terminal proteins.
Unpublished experiments in our laboratory have shown that neuromodulin is present in macaque frontal cortex synaptosomes. In addition, these synaptosomes (prepared by method 2, see "Experimental Procedures") show glutamate responses to K ϩ -induced depolarization or treatment with 4-aminopyridine ϩ phorbol ester, comparable to those observed by others for rat and guinea pig synaptosomes (24,75). Therefore, the macaque synaptosomes might provide a useful model for examining the functional significance of the Rab3A-dissociating effect of Ca 2ϩ / CaM. Experiments designed to examine the effects of depolarization on the distribution of Rab3A are in progress.
In summary, we have shown that Ca 2ϩ /CaM can cause Rab3A to dissociate from synaptic membranes and studied the mechanism of this effect. In addition, we have identified similarities and differences between the effects of Ca 2ϩ /CaM and those of Rab GDI. These results raise a number of questions. For example, how important is the Rab3A-dissociating effect of Ca 2ϩ /CaM in vivo? If important in vivo, what is the significance of the differences between this effect and that of Rab GDI? Does the effect of Ca 2ϩ /CaM complement that of Rab GDI in some unknown way? Do complementary Rab3A-dissociating effects of Ca 2ϩ /CaM and Rab GDI increase the efficiency of synaptic vesicle recycling? These questions may suggest directions for future research.