Calcium-dependent Binding of Sorcin to the N-terminal Domain of Synexin (Annexin VII)*

The annexins are characterized by their ability to bind phospholipid membranes in a Ca2+-dependent manner. Sequence variability between the N-terminal domains of the family members may contribute to the specific cellular function of each annexin. To identify proteins that interact with the N-terminal domain of synexin (annexin VII), a fusion protein was constructed composed of glutathioneS-transferase fused to amino acids 1–145 of human synexin. Affinity chromatography using this construct identified sorcin as a Ca2+-dependent synexin-binding protein. Overlay assays confirmed the interaction. The glutathioneS-transferase construct associates with recombinant sorcin over the range of pCa2+ = 4.7–3.1 with no binding observed at pCa2+ = 5.4. Overlay assays using deletion constructs of the synexin N-terminal domain mapped the sorcin binding site to the N-terminal 31 amino acids of the synexin protein. Additionally, synexin forms a complex with sorcin and recruits this protein to chromaffin granule membranes in a Ca2+-dependent manner. Sorcin is able to inhibit synexin-mediated chromaffin granule aggregation in a manner saturable with increasing sorcin concentrations, but does not influence the Ca2+ sensitivity of synexin-mediated granule aggregation. Therefore, the interaction between sorcin and synexin may serve to regulate the functions of these proteins on membrane surfaces in a Ca2+-dependent manner.

The annexins are a family of homologous proteins that are characterized by their ability to bind phospholipid membranes in a Ca 2ϩ -dependent manner (1). Synexin (annexin VII) was the first to be isolated (2, 3) on the basis of its ability to cause adrenal medullary chromaffin granule aggregation in vitro. Structurally, the annexins are composed of a conserved Cterminal core region containing four or eight repeats of a 70amino acid sequence that is 40 -60% conserved, and a unique N-terminal domain. The conserved core region is thought to be responsible for the annexin Ca 2ϩ and lipid binding properties (4). The sequences of the N-terminal domains of this family are highly variable, leading to the hypothesis that the differences in the N-terminal domains may contribute to the specific cellular function of each annexin (9,10,45,46).
Although members of the annexin family have been implicated in playing diverse roles in cellular processes that include the anti-inflammatory response (5), ion channel formation (6), vesicular trafficking (7), and exocytosis (8), the physiological function has not been definitively determined for any annexin. However, it has been proposed that the annexins may perform their in vivo roles by interacting with other proteins. In fact, several members of the annexin family have been shown to bind members of the S100 subfamily of EF-hand proteins. Annexin I binds S100C (9,10,45), annexin II binds p11 (11), and annexin XI binds calcyclin (12) with each interaction characterized by the S100 protein binding to the N-terminal domain of the respective annexin. The annexin I-S100C and annexin XI-calcyclin interactions are Ca 2ϩ -dependent, while the annexin II-p11 interaction is Ca 2ϩ -independent.
Most annexins have short N-terminal domains of less than 50 residues. However, synexin (annexin VII) has an N terminus composed of 167 amino acids. This domain is largely uncharged and is rich in proline, glycine, tyrosine, and glutamine residues (13). The function of this domain is particularly enigmatic, although annexin XI also has an N-terminal domain of similar size and amino acid composition (41), and this domain has been shown to interact with calcyclin (12) and has been proposed to function as a nuclear localization signal in a cell type-specific manner (38).
The expression of mammalian annexins in yeast secretory mutants has yielded results that suggest that there may be a synexin binding protein that interacts with the N-terminal domain of the annexin. When human synexin is expressed in the late sec mutants sec2, sec4, and sec15, their growth is inhibited and their content of secretory vesicles is increased (14). Coexpression of the N-terminal domain with full-length synexin inhibits this synthetic lethality. 1 It is possible that the effect of the N-terminal domain is due to competition with an interacting protein at a specific binding site.
Since annexins, in general, must be extracted in EGTAcontaining buffers to remove them from membranes, it is possible that proteins that interact with annexins in a Ca 2ϩ -dependent manner are lost during isolation of the annexins. Therefore, we have focused our attention on identifying proteins that bind annexins in a Ca 2ϩ -dependent manner. To identify intracellular mammalian proteins that interact with the N-terminal domain of synexin, a fusion protein consisting of glutathione S-transferase and the N-terminal domain of synexin was constructed. Utilizing this fusion protein in affinity chromatography and overlay techniques, we have identified sorcin, an intracellular EF-hand, Ca 2ϩ -binding protein, as a synexin-binding protein.

Materials
The following reagents were obtained from commercial sources: pGEX-KG expression vector (American Type Culture Collection, Rockville, MD) (15), glutathione and anti-glutathione S-transferase (anti-GST) 2 polyclonal antiserum (Sigma), glutathione-Sepharose and CNBr-* This work was supported by National Institutes of Health Grant GM53266. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Methods
Construction of Glutathione S-Transferase-Synexin N-terminal Domain Expression Plasmid (pGEX-KG-syntail)-The synexin N-terminal domain from residues 1-145 was amplified by PCR from human synexin cDNA that had been previously subcloned into a yeast expression vector (14). The upstream primer incorporated a NcoI site at the initiation codon and consequently changed the second amino acid, following the initiating methionine, from a serine to an alanine. The downstream primer incorporated a HindIII site and a termination codon adjacent to the codon for amino acid residue 145. The PCR product was ligated into the yeast expression vector YEpC 2 1 (14), from which it was subsequently excised with NcoI and HindIII and then ligated into the complementary sites of pGEX-KG. The resulting construct (pGEX-KG-syntail) creates a fusion protein between GST and amino acids 1-145 of human synexin where amino acid 2 of synexin has been changed from serine to alanine to facilitate subcloning into the NcoI site. The identity of the construct was verified by sequencing using the Sanger method (16) on an Applied Biosystems Prism 377 automated DNA sequencer.
Isolation of GST-Synexin N-terminal Domain Fusion Protein-Escherichia coli XL-1 Blue cells were transformed with pGEX-KG-syntail, and an ampicillin-resistant colony was used to inoculate 100 ml of LB medium having a concentration of 100 g/ml ampicillin. This preculture was grown to stationary phase overnight at 37°C. One liter of LBampicillin was inoculated the following morning with the preculture and was grown for 1 h at 37°C. Protein expression was then induced by the addition of isopropyl-1-thio-␤-D-galactopyranoside to a final concentration of 100 M and incubation at 37°C was continued for 5 h. The cells were collected by centrifugation (5000 ϫ g, 10 min), resuspended in ice-cold phosphate-buffered saline (150 mM NaCl, 8 mM Na 2 HPO 4 , 2 mM NaH 2 PO 4 , pH 7.3), and lysed by sonication in the presence of 1% Triton X-100. Insoluble material was removed by centrifugation (10000 ϫ g, 10 min), and the fusion protein in the resulting supernatant was isolated by binding to GSH-Sepharose. Following several washes with phosphate-buffered saline, the GST fusion protein was eluted with 10 mM reduced glutathione and was dialyzed overnight at 4°C against 50 mM Tris-HCl, pH 8.0. The protein was then stored frozen at Ϫ20°C until use. SDS-PAGE of the dialyzed protein followed by Coomassie Blue staining of the gel revealed one major protein band at 41 kDa.
Preparation of Postmicrosomal Supernatants and Affinity Chromatography-Postmicrosomal supernatants were prepared from adrenal medullary tissue as described previously (17). The affinity matrix was prepared as follows; 1.5 mg of the GST-synexin N-terminal domain construct or unfused GST protein prepared as described above was suspended in 20 ml of coupling buffer (500 mM NaCl, 100 mM NaHCO 3 ) with 1 mM phenylmethanesulfonyl fluoride and 1 mM DL-dithiothreitol added and mixed with a slurry of CNBr-activated Sepharose 4B prepared from 5 g of freeze-dried powder according to the manufacturer's (Pharmacia) recommendations. Coupling was allowed to proceed overnight at 4°C on a rocking table. The gel was then washed once with coupling buffer and incubated for 2 h at room temperature with 250 ml of 1 M ethanolamine chloride, pH 8.0.
Affinity chromatography was performed using an adaptation of the method that was described previously (17). Briefly, the coupled Sepharose was washed with 50 ml of column buffer (240 mM sucrose, 150 mM NaCl, 2 mM EGTA, 1 mM MgCl 2 , and 25 mM HEPES-NaOH adjusted to pH 7.3 at 37°C) and packed into a water-jacketed column (1.6 cm ϫ 7.5 cm), giving a bed volume of approximately 15 ml. The column temperature was maintained at 37°C during chromatography. All column buffers were pumped with an LKB Microperpex peristaltic pump operated at maximum speed (2 ml/min). The packed column was washed with 50 ml of column buffer, to which 4 mM CaCl 2 had been added such that the final free Ca 2ϩ concentration equaled 2 mM. Solutions of NaCl, HEPES, MgCl 2 , EGTA, and CaCl 2 were added to 40 ml of the postmicrosomal supernatant so that the final concentrations equaled those of the Ca 2ϩ -containing buffer. The postmicrosomal supernatant, containing 200 -300 mg of protein, was then applied to the column, and the column was subsequently washed with 50 ml of column buffer contain-ing 500 mM NaCl and 2 mM free Ca 2ϩ . Ca 2ϩ -dependent binding proteins were eluted from the column with 50 ml of column buffer (containing 2 mM EGTA) to which no CaCl 2 had been added.
Sequencing of p22-p22 was separated from the other proteins found in the peak fractions of the GST-synexin N-terminal domain fusion protein column by SDS-PAGE. The 22-kDa protein band was first excised from the SDS-polyacrylamide gel and then the protein was electroeluted from the polyacrylamide slice. The electroeluted protein was precipitated with chloroform and methanol to remove SDS and Coomassie Blue. Fifty microliters of 8 M urea was added to the precipitated protein and was then diluted to 200 l to give 2 M urea and 50 mM Tris, pH 8.0. Three hundred nanograms of lysyl peptidase purified from Achromobacter lyticus (Waco Bioproducts, Richmond, VA) was added, and the sample was incubated for 15 h at 37°C. The resulting peptides were separated using reverse-phase high performance liquid chromatography on a C18 column (Buffer A, 0.1% trifluoroacetic acid in water; Buffer B, 0.09% trifluoroacetic acid in acetonitrile). The amino acid sequences of two peptides were determined by Edman degradation on an Applied Biosystems model 470A gas phase sequencer with a model 120A phenylthiohydantoin analyzer.
GST-Synexin N-terminal Domain Overlay Assays-Four micrograms of the proteins found in the peak fractions of the GST-synexin Nterminal domain fusion protein column that contained sorcin were separated by one-dimensional gel electrophoresis. These separated proteins were then transferred to nitrocellulose, and the nitrocellulose filter was blocked overnight in blocking buffer (150 mM NaCl, 8 mM Na 2 H PO 4 , 2 mM NaH 2 PO 4 (pH 7.3), 2 mM CaCl 2 , and 5% nonfat dry milk) at 4°C. Individual strips were then incubated for 1 h in buffer B (20 mM Tris-HCl, 150 mM NaCl, 0.1% Tween 20, 2 mM CaCl 2 , and 5% bovine serum albumin) with either 5 g/ml unfused GST protein or 5 g/ml GST-synexin N-terminal domain construct. All wash steps consisted of three 10-min incubations with blocking buffer. The strips were washed and then incubated for 1 h in blocking buffer with rabbit polyclonal antiserum raised against GST (1:1500 dilution). Following another wash step, the strips were incubated for 1 h in blocking buffer with a goat anti-rabbit HRP-conjugated antibody (1:1500 dilution). Following a final wash, binding was detected using a chemiluminescent substrate system (Pierce).
Subcloning and Expression of Recombinant Human Sorcin-Sorcin cDNA was amplified by PCR from human heart cDNA (CLONTECH) using primers that incorporated NcoI and SacI restriction sites at the 5Ј and 3Ј ends of the DNA, respectively. The amplified product was subcloned into pGEX-KG and used to transform E. coli XL-1 Blue. The identity of the construct was verified by DNA sequencing. Bacterial growth, protein induction, and isolation of the fusion protein were performed as described above. However, following the final wash with phosphate-buffered saline, the GSH-Sepharose beads were washed twice with cleavage buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 2.5 mM CaCl 2 , 0.1% ␤-mercaptoethanol). To cleave sorcin from the fusion protein, thrombin (Boehringer Mannheim, 0.5 unit/0.5 ml of slurry) was added to the bead slurry and incubated for 40 min at room temperature on a rotator. Released protein was recovered by washing 5-6 times with wash buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 10 mM EGTA). Recombinant protein was dialyzed overnight in 10 mM Tris, pH 7.4 with 2 mM EGTA and was further purified by application to an FPLC Poros Q anion exchange column (PerSeptive Biosystems, Cambridge, MA) equilibrated with 10 mM Tris, pH 7.4, at 4°C and elution with a linear gradient of 0 -1 M NaCl in 10 mM Tris. Sorcin eluted between 0.2 and 0.25 M NaCl. SDS-PAGE of the column fractions followed by Coomassie Blue staining of the gel revealed one protein band at approximately 23 kDa. Polyclonal antiserum was raised in rabbits against recombinant sorcin according to methods described previously (37).
Measurement of the Ca 2ϩ Dependence of the in Vitro Interaction of Sorcin with the GST-syntail Fusion Protein-Eight micrograms of the synexin-(1-145) GST fusion protein bound to GSH-Sepharose were incubated on ice for 20 min with 4 g of recombinant sorcin in binding buffer (40 mM Hepes, pH 7.0, 30 mM KCl, 240 mM sucrose, 0.1% Triton X-100) to which Ca 2ϩ -EGTA buffer was added to give a final concentration of 2.5 mM EGTA and the desired Ca 2ϩ concentration. The final Ca 2ϩ concentration of each buffer was verified with a calcium-selective electrode (Radiometer). After binding, the beads were rapidly washed with 1 ml of the appropriate Ca 2ϩ binding buffer and were resuspended in Laemmli sample buffer. The proteins were then separated on a 10% SDS-polyacrylamide gel and were visualized by Coomassie Blue staining.
Subcloning and Expression of Recombinant Human Synexin-Fulllength human synexin cDNA was amplified by PCR off of the vector Cao3, which is a derivative of YEpDB60-synexin (14) in which the leu2 polymerase chain reaction; HRP, horseradish peroxidase; PAGE, polyacrylamide gel electrophoresis; MES, 4-morpholineethanesulfonic acid. gene is disrupted. PCR was performed with primers that incorporated XbaI and XhoI restriction sites at the 5Ј and 3Ј ends of the DNA, respectively. The amplified product was subcloned into pGEX-KG, which was used to transform E. coli XL-1 Blue. The identity of the construct was verified by DNA sequencing. Bacterial growth, protein induction, isolation of the protein, and cleavage with thrombin were performed as described for recombinant sorcin. However, the synexin wash buffer consisted of 50 mM MES, pH 6.0, 150 mM NaCl, 10 mM EGTA. Recombinant synexin was dialyzed overnight in 25 mM MES, pH 6.0 with 1 mM EGTA. The protein was further purified by application to an FPLC Poros S cation exchange column (PerSeptive Biosystems, Cambridge, MA) equilibrated with 25 mM MES, pH 6.0, and 500 M Ca 2ϩ at 4°C and elution with a linear gradient of 0 -1 M KCl in 25 mM MES. To promote binding of synexin to the Poros S resin, the Ca 2ϩ concentration of the protein solution was brought to 2 mM before loading onto the column. Recombinant synexin eluted between 0.25 to 0.3 M KCl. SDS-PAGE of the column fractions followed by Coomassie Blue staining of the gel revealed one protein band at approximately 48 kDa. Polyclonal antiserum was raised in rabbits against recombinant synexin according to methods described previously (37).
Construction of Deletions and Sorcin Overlay Assays-The human cDNAs for the synexin N-terminal region deletions were amplified by PCR off of the vector pGEX-KG-synexin. PCR was performed with primers that incorporated XbaI and XhoI restriction sites at the 5Ј and 3Ј ends of the DNA, respectively. The amplified products were subcloned into pGEX-KG and were transformed into E. coli XL-1 Blue. The identities of all of the constructs were verified by DNA sequencing. Three constructs were made that deleted C-terminal portions of the first 145 amino acids that comprise the synexin N-terminal domain. Bacterial growth, protein induction, and isolation of these fusion proteins were performed as described above. GSH-Sepharose beads with 4 g of fusion protein bound were resuspended in Laemmli sample buffer. The protein was run on a 10% SDS-polyacrylamide gel and was then transferred to nitrocellulose. The nitrocellulose filter was blocked overnight in overlay assay buffer (20 mM Tris-HCl, pH 7.4, 300 mM NaCl, 2 mM CaCl 2 , and 5% nonfat dry milk) at 4°C. The immobilized proteins were then incubated for 1 h in overlay assay buffer with 2 g/ml recombinant human sorcin. All wash steps consisted of three 10-min incubations with overlay assay buffer. The nitrocellulose filters were washed and then incubated for 1 h in overlay assay buffer with rabbit polyclonal antiserum raised against sorcin (1:10,000 dilution). Following another wash step, the immobilized proteins were incubated for 1 h in overlay assay buffer with a goat anti-rabbit HRP-conjugated antibody (1:10,000 dilution). Following a final wash, sorcin binding was detected using a chemiluminescent substrate system (Pierce). A control was performed as described above, but using only the sorcin antiserum with no protein overlay to determine background antiserum binding levels. Fusion proteins were visualized by performing Western blots with GST antiserum (Sigma).

Sorcin Overlays of Human Synexin Expressed in Yeast and Sorcin
Overlays of Adrenal Cytosol-Recombinant human synexin was expressed in yeast without a fusion tag and was purified by binding bovine brain lipid vesicles in the presence of Ca 2ϩ as described previously (14). Postmicrosomal supernatants were prepared from adrenal medullary tissue as described previously (17). Five micrograms of purified synexin or 65 g of postmicrosomal supernatant were run on a 10% SDS-polyacrylamide gel and were transferred to nitrocellulose. Sorcin overlay assays were performed exactly as described above for the sorcin overlays of the deletion constructs. Controls were performed using the sorcin antiserum with no protein overlay to determine background antiserum binding levels.
Construction and Expression of the Glutathione S-Transferase-Annexin XI N-terminal Domain Fusion Protein-A cDNA encoding human annexin XI was obtained as an I.M.A.G.E. Consortium Clone (ID26147, American Type Culture Collection, Rockville, MD) (40). The cDNA encoding the first 121 N-terminal amino acids of annexin XI was amplified by PCR using primers that incorporated XbaI and XhoI restriction sites at the 5Ј and 3Ј ends of the DNA, respectively. The amplified product was subcloned into pGEX-KG and was transformed into E. coli BL21(DE3) cells. The identity of the construct was verified by DNA sequencing. Bacterial growth, protein induction, and isolation of the fusion protein were performed as described for the GST-syntail construct. The GST-annexin XI construct was further purified by applica-tion to an FPLC Poros Q anion exchange column (PerSeptive Biosystems, Cambridge, MA) equilibrated with 10 mM Tris, pH 7.4, at 4°C and elution with a linear gradient of 0 -1 M NaCl in 10 mM Tris. GST-annexin XI N-terminal domain fusion protein eluted between 0.15 and 0.2 M NaCl. SDS-PAGE of the column fractions followed by Coomassie Blue staining of the gel revealed a major protein band at approximately 40 kDa and several minor, lower molecular weight degradation bands. Binding assays of recombinant sorcin to the annexin XI construct were performed as described above for measuring the Ca 2ϩ dependence of the interaction between sorcin and the synexin N-terminal domain.
Chromaffin Granule Binding and Aggregation Assays-Chromaffin granules were purified from bovine adrenal medullary tissue by differential centrifugation in 300 mM sucrose as described previously (18). Chromaffin granule aggregation was assayed by measuring the change in turbidity (absorbance at 540 nm) of 1-ml granule suspensions in aggregation buffer (40 mM Hepes, pH 7.0, 30 mM KCl, 240 mM sucrose) at room temperature (2). Absorbance measurements were made using a Beckman DU70 microprocessor-controlled spectrophotometer, which allowed for absorbance data collection as a function of time for four assays simultaneously. Ca 2ϩ -EGTA buffer was added to give a final concentration of 2.5 mM EGTA, and the desired Ca 2ϩ concentration was verified with a Ca 2ϩ electrode (Radiometer). The initial absorbance was measured for 1 min, and then aggregation was induced by the addition of protein. Absorbance was monitored for an additional 10 min. Recombinant protein binding to granules was analyzed by pelleting the granules by centrifugation for 5 min at 14,000 ϫ g. The pellets were resuspended in Laemmli sample buffer and run on 10% SDS-polyacrylamide gels. The supernatants were desalted by gel filtration, lyophilized, resuspended in Laemmli sample buffer, and also run on 10% SDSpolyacrylamide gels. Binding was quantitated by densitometry of the Coomassie Blue-stained gels using the Collage Image Information Program (Image Dynamics Corp.). Total protein values were calculated as the sum of the densitometric values of the protein bands on the pellet and supernatant gels. For phase microscopy, chromaffin granule suspensions were mixed with both protein and Ca 2ϩ (800 M) and were allowed to aggregate at room temperature for 30 min before samples were imaged with a Nikon Optiphot microscope.
Analytical Methods-Protein concentrations were determined using the method of Bradford (19) using bovine serum albumin as standard. One-dimensional gel electrophoresis was performed as described by Laemmli (20) using 10% gels. The following standards were used for molecular mass determination: phosphorylase b, 97 kDa; serum albumin, 67 kDa; ovalbumin, 45 kDa; carbonic anhydrase, 31 kDa; trypsinogen, 21 kDa; ribonuclease, 14 kDa. Gels were stained with Coomassie Blue. Western immunoblots were performed as described by Burnette (21), using horseradish peroxidase-coupled secondary antibodies and a chemiluminescent substrate system for detection (Pierce).

RESULTS
Ca 2ϩ -dependent Binding of a 22-kDa Protein to the Synexin N-terminal Domain-To determine whether cytosolic proteins of the bovine adrenal medulla were able to bind to the synexin N-terminal domain in a Ca 2ϩ -dependent manner, affinity chromatography with the GST-synexin N-terminal domain fusion protein bound to CNBr-activated Sepharose 4B was performed. The postmicrosomal supernatant from bovine adrenal medullary tissue with an added 2 mM CaCl 2 was applied to the GST-synexin N-terminal domain fusion protein column. A distinct protein peak eluted from the column when Ca 2ϩ -dependent binding proteins were released with buffer containing 2 mM EGTA (Fig. 1A). The protein found in the peak fraction that migrates at 22 kDa on an SDS-polyacrylamide gel appeared to bind the synexin N-terminal domain specifically because it was not observed eluting from the control column to which unfused GST was bound (Fig. 1, B and D, compare lanes 35 and 30,  respectively). The other proteins found as components of the peak fraction associate with both the GST-synexin N-terminal domain fusion protein column and the GST control column. Several of these proteins represent protein kinase C and members of the annexin family that have been shown previously (22) to associate with the matrix, possibly through an interaction with phospholipids present in the postmicrosomal supernatant that bind nonspecifically to the column.
Identification of p22 as Sorcin-The 22-kDa protein band was excised from an SDS-polyacrylamide gel with subsequent electroelution of the protein. Following digestion of the protein with a lysyl peptidase, two of the resulting peptides were sequenced by Edman degradation. The sequences of these peptides (PFNLETCRLMVSMLDRDM and ITFDDYIACCVK) exactly match residues 69 -86 and 154 -165 of both hamster (23) and human (24) sorcin.
Verification of the Interaction between Sorcin and the Synexin N Terminus-Since the postmicrosomal supernatants used in the column experiments may contain small amounts of endogenous lipid, it was necessary to determine whether sorcin interacted directly with the synexin N-terminal domain or whether the association of the two proteins was lipid-dependent. Therefore, an overlay procedure was used in which the proteins from the peak fractions of the GST-synexin N-terminal domain fusion protein column were first separated by electrophoresis on an SDS-polyacrylamide gel and were then transferred to nitrocellulose. The immobilized proteins were then incubated in the presence of Ca 2ϩ with either unfused GST or the GST-synexin N-terminal domain fusion protein. When binding was detected with anti-GST antiserum, the interaction between sorcin and the synexin N-terminal domain was confirmed (Fig. 2, GST-syntail). Sorcin was not able to bind to the unfused GST control (Fig. 2, GST).
Sorcin Binding to Full-length Synexin and Cytosolic Proteins of the Adrenal Medulla-To demonstrate that recombinant sorcin was able to bind full-length synexin, a sorcin overlay was performed of human synexin that had been expressed in yeast without a fusion tag. Immobilized synexin was incubated with 2 g/ml recombinant human sorcin in the presence of Ca 2ϩ and sorcin binding was detected with polyclonal sorcin antisera (Fig. 3A). Sorcin bound full-length synexin (Fig. 3A, lane 2). Additionally, sorcin overlays of total adrenal medullary cytosol demonstrated that the only protein to which sorcin bound was a 47-kDa protein, which is the expected molecular mass of synexin (Fig. 3B, lane 2). No binding of sorcin to proteins found in the total membrane fraction was observed (data not shown). Control Western blots demonstrated that background binding of the sorcin antiserum to synexin was insignificant for both experiments (Fig. 3, A and B, lane 1). Not only do these results demonstrate that sorcin binds full-length synexin, they also FIG. 1. Protein concentration and SDS gels of affinity column fractions. A, protein concentration of fractions obtained from a GST synexin N-terminal domain fusion protein column. The postmicrosomal supernatant was applied to the column prior to treatment with buffers containing Ca 2ϩ or EGTA as indicated at the top of the graph. Ca 2ϩcontaining buffer was applied at fraction 1. The maximum of protein concentration observed at the far left of the graph represents cytosolic proteins not bound to the column. Column volume was approximately 15 ml, and the fraction size was 2.5 ml. B, 1 ml of each of the indicated fractions from the GST-synexin N-terminal domain fusion protein column was desalted, lyophilized, and run per lane on an SDS-polyacrylamide gel. The gel was then stained with Coomassie Blue. C, protein concentration of fractions obtained from a GST control column that was run as described above in A. D, 2.5 ml of each of the indicated fractions from the GST control column were treated as described in B.

FIG. 2. Confirmation of a direct interaction between the synexin N-terminal domain and sorcin by overlay assay.
Four micrograms of the proteins in the peak fraction of the GST-synexin Nterminal domain fusion protein column that contained sorcin were separated on an SDS-polyacrylamide gel. The proteins were then transferred to nitrocellulose, and the blot was incubated with either 5 g/ml GST protein only (GST) or 5 g/ml GST-synexin N-terminal domain fusion protein (GST-syntail). Bound GST-protein was detected by sequentially incubating with rabbit anti-GST antiserum (Sigma), a goat anti-rabbit HRP-conjugated antibody, and a chemiluminescent substrate system (Pierce). This experiment was performed with 2 mM Ca 2ϩ in all solutions.
suggest that synexin is the only soluble protein to which sorcin binds.
Determination of the Ca 2ϩ Requirement for Binding-To determine the Ca 2ϩ requirement for the in vitro interaction between sorcin and the synexin N-terminal domain, binding of recombinant sorcin to GSH-Sepharose immobilized GST-synexin N-terminal domain fusion protein was assayed at various Ca 2ϩ concentrations (Fig. 4). The proteins were mixed for 20 min on ice in Ca 2ϩ buffer, and then the beads were rapidly washed and loaded onto an SDS-polyacrylamide gel. Binding of sorcin to the synexin N terminus is observed over the range of pCa 2ϩ ϭ 4.7-3.1 (20 -800 M Ca 2ϩ ) with no binding detected at pCa 2ϩ ϭ 5.4 (4 M Ca 2ϩ ). In control experiments using unfused GST, no binding was observed at any pCa 2ϩ value (data not shown). Therefore, these data confirm the Ca 2ϩ requirement for the association of these proteins and suggest that low micromolar levels of Ca 2ϩ are required for binding.
Mapping of the Sorcin Binding Site on Synexin-Eight synexin deletions were constructed and expressed as GST fusion proteins. Two sets of deletions are represented, those where C-terminal portions and those where N-terminal portions of the first 145 amino acids that comprise the synexin N-terminal domain are deleted (Table I). Sorcin overlays were performed to assay the ability of sorcin to bind to each of the deletion constructs. Controls performed without any incubation with recombinant sorcin showed insignificant background binding of the sorcin antiserum (data not shown). Sorcin is able to bind the GST construct composed of amino acids 1-31 of synexin (Fig. 5B). In addition, deletion of amino acids 1-8 of synexin abolishes sorcin binding (Table I). Therefore, the results of these experiments suggest that the sorcin binding site on synexin is located at the very N terminus of the protein.

Ca 2ϩ -dependent Binding of Sorcin to the N-terminal Domain
of Annexin XI-Human synexin and annexin XI share complete sequence identity within their first eight N-terminal amino acids (MSYPGYPP). Additionally, the extreme N-terminal regions of both of these annexins contain a repeat of the sequence GYPP, although the spacing between the repeats varies between the two proteins (Fig. 10). Since deletion mapping studies suggested this region of synexin was necessary for sorcin binding, a GST-annexin XI N-terminal domain fusion protein (amino acids 1-121) was used to assay sorcin binding to annexin XI.
Binding of recombinant sorcin to the GST-annexin XI Nterminal domain fusion protein attached to GSH-Sepharose was assayed at several Ca 2ϩ concentrations (Fig. 6). Sorcin binds the annexin XI N-terminal domain in a Ca 2ϩ -dependent manner. It appears that sorcin may bind annexin XI with lower affinity than synexin, since smaller amounts of sorcin are bound than seen with the synexin N-terminal domain (compare with Fig. 4). In control experiments using unfused GST, no binding was observed at any pCa 2ϩ value (data not shown). These data suggest that sorcin may recognize the GYPP motif Ϫ NT-(80-145) Ϫ a Sorcin overlays (2 g/ml) of the GST-synexin N-terminal domain deletion constructs were used to determine sorcin binding. All assays were performed in the presence of Ca 2ϩ . Sorcin binding was detected using a polyclonal anti-sorcin antiserum followed by a HRP-conjugated secondary and chemiluminescent detection.
found in both synexin and annexin XI as its binding target sequence.
Recruitment of Sorcin to Chromaffin Granule Membranes in the Presence of Ca 2ϩ by Synexin-Previous studies of sorcin demonstrated that sorcin pelleted with crude heart membrane fractions in the presence of Ca 2ϩ (25). To determine whether synexin is able to recruit sorcin to phospholipid membranes in a Ca 2ϩ -dependent manner, the ability of both sorcin and synexin to bind chromaffin granules in the presence of Ca 2ϩ was assayed. Equal molar amounts of recombinant human sorcin and synexin were added to chromaffin granules at various Ca 2ϩ concentrations. Association of sorcin with granule-bound synexin was determined by densitometry of SDS-polyacrylamide gels of the granule pellets and corresponding supernatants (Fig. 7). In control assays of sorcin alone, approximately 5.9 Ϯ 1.0% (n ϭ 8) of sorcin is associated with the chromaffin granule pellet in the presence of EGTA (Fig. 7A, far right lane, *E). However, approximately 25.6 Ϯ 3.2% (n ϭ 8) of sorcin is associated with the chromaffin granule pellet at pCa 2ϩ ϭ 3.1 (800 M Ca 2ϩ ) (Fig. 7A, far right lane, *3.1). Despite this Ca 2ϩ -dependent pelleting of sorcin, which may represent either membrane binding or pelleting of the protein itself, synexin is able to recruit 90.7 Ϯ 1.6% (n ϭ 8) of the total sorcin to the chromaffin granule membrane surface in the presence of Ca 2ϩ . Sorcin remains bound to synexin and is found associated with the pellet down to approximately pCa 2ϩ ϭ 4.7 (20 M Ca 2ϩ ), below which synexin is unable to bind to the granule membrane and both proteins are found in the supernatant. In control incubations with synexin alone, the amount of synexin In control experiments using unfused GST, no binding was observed at any pCa 2ϩ value. Note that the annexin XI construct has two faint degradation bands that run below the full-length protein band, but above the sorcin band.

FIG. 7.
Synexin recruits sorcin to the chromaffin granule membrane. Eight micrograms recombinant human synexin and 4 g recombinant human sorcin were added to chromaffin granules at various Ca 2ϩ concentrations, and the mixtures were incubated at room temperature for 10 min. Then the granules and any associated proteins were pelleted, the supernatants removed, the pellets resuspended in Laemmli sample buffer, and these pellets and supernatants run on separate SDS-polyacrylamide gels. A, results of the densitometry of the sorcin; B, synexin bands from the SDS-polyacrylamide gels; shaded boxes represent the percent of total protein associated with the pellet, and open boxes represent the percent of total protein found in the supernatant at the indicated pCa 2ϩ value. Total protein values were calculated as the sum of the densitometric values of the protein bands on the pellet and supernatant gels. Binding assays controlling for sorcin binding to granule membranes in the presence of EGTA or at pCa 2ϩ 3.1, respectively, are represented by the far right-hand lanes of A, labeled *E and *3.1.
bound and the Ca 2ϩ dependence of synexin binding to the granules were identical to those seen in the presence of sorcin. Therefore, sorcin does not interfere with or enhance binding of synexin to the granule membrane.
Inhibition of Synexin-mediated Chromaffin Granule Aggregation by Sorcin-Since synexin is able to bring two membranes together in a Ca 2ϩ -dependent manner as demonstrated by its ability to aggregate chromaffin granules (2), we wanted to determine whether the recruitment of sorcin by synexin to chromaffin granule membranes alters synexin-mediated aggregation. In particular, we wanted to determine whether sorcin is able to alter the Ca 2ϩ requirement of synexin-mediated aggregation, since p11 binding to annexin II decreases the Ca 2ϩ requirement for annexin II-mediated aggregation (26). Sorcin alone was found to have no chromaffin granule aggregating ability. However, addition of equal molar sorcin and synexin to the aggregation assay inhibits the aggregating ability of synexin by approximately 40% at pCa 2ϩ 3.1 (800 M Ca 2ϩ ) when measured as the percentage of the initial absorbance at 540 nm. This inhibition is saturable with increasing molar concentrations of sorcin with a maximal inhibition of approximately 70% of the initial absorbance at pCa 2ϩ 3.1 observed at a 2 to 1 molar excess of sorcin (Fig. 8). Densitometry of the synexin and sorcin bands on gels of the pellets and supernatants from these experiments yielded an estimate of the stoichiometry of the interaction of 1 mol of synexin to 2-3 mol of sorcin. The inhibition of synexin-mediated granule aggregation by sorcin was also directly visualized by phase microscopy. Although granules were aggregated by synexin at pCa 2ϩ 3.1 (Fig. 9A), this aggregation was visibly inhibited in the presence of sorcin (Fig.  9B). Large aggregates were observed upon the addition of synexin in the presence of Ca 2ϩ , but there were few large aggregates and many more smaller aggregates and single granules observed when sorcin was present. Although these data show that sorcin is able to inhibit synexin-mediated aggregation, sorcin was found to have no effect on the Ca 2ϩ requirement for aggregation.
Since annexin I is also able to aggregate chromaffin granules in a Ca 2ϩ -dependent manner, it was used as a control for the annexin VII-mediated aggregation experiments. Sorcin was unable to inhibit annexin I from aggregating granules in the presence of Ca 2ϩ (data not shown). Additionally, examination of the granule pellets from the aggregation assays revealed that annexin I did not recruit sorcin to the granule membrane (data not shown).

DISCUSSION
The results presented here demonstrate that the EF-handcontaining, Ca 2ϩ -binding protein sorcin binds to the N-terminal domain of synexin in a Ca 2ϩ -dependent manner. Sorcin stands for soluble resistance-related Ca 2ϩ -binding protein. It was first identified as a protein overexpressed in conjunction with the membrane-bound drug transporter, P-glycoprotein, in multidrug-resistant cell lines selected for with natural product chemotherapeutic drugs such as colchicine or the vinca alkaloids (23). Amplification of the gene for sorcin in these cell lines, The suspensions were allowed to aggregate at room temperature for 30 min before samples were taken for imaging by phase microscopy. Note large aggregates upon the addition of synexin, but only small aggregates and single granules in the presence of synexin and sorcin.
FIG. 8. Aggregation assays at varying stoichiometries of sorcin and synexin show that the inhibition of aggregation is saturable. Aggregation data are represented as the percent of initial absorbance at 540 nm, 10 min after initiation of aggregation by addition of protein versus pCa 2ϩ value. The Ca 2ϩ concentration of each buffer was verified with a Ca 2ϩ -selective electrode. The curves on the graph represent the following molar ratios of synexin to sorcin; ϫ ϭ 1:0, f ϭ 1:0.5, E ϭ 1:1, å ϭ 1:2, * ϭ 1:3. Maximal inhibition is observed at a 1:2 molar ratio of synexin to sorcin. however, may be due to linkage with P-glycoprotein-encoding genes (27).
Direct binding of Ca 2ϩ to sorcin has been demonstrated by 45 Ca 2ϩ overlays (28,29), and affinity measurements have been made by studying intrinsic fluorescence changes of recombinant sorcin that occur at 0.1-1 M Ca 2ϩ (25). The sequence of this protein predicts that it has four putative Ca 2ϩ -binding domains, two with strong homology to calmodulin EF-hand motifs (23). Additionally, the N-terminal domain of sorcin is rich in glycine, proline, and tyrosine residues and is homologous to the corresponding domain of the calpain light chain. The tissue distribution of sorcin and synexin have been independently demonstrated by Western (30) and Northern blot (32) analysis to be similar. Both proteins are expressed in liver, heart, brain, lung, spleen, kidney, and, as found in our study, the adrenal medulla.
The binding sites of S100C on annexin I (10), p11 on annexin II (11), and calcyclin on annexin XI (31) have been determined to be within the N-terminal domain of the annexin protein. Therefore, the sorcin-synexin interaction represents the fourth example of an annexin interacting with an EF-hand, Ca 2ϩbinding protein via the N-terminal domain of the annexin. Additionally, like the annexin I-S100C and annexin XI-calcyclin interactions, the synexin-sorcin interaction is Ca 2ϩ -dependent. The observed Ca 2ϩ dependence of the synexin-sorcin interaction suggests that changes in local calcium concentration may regulate the interaction under physiological conditions. In fact, Ca 2ϩ may play a role in controlling the intracellular localization of the complex, since the Ca 2ϩ requirements for synexin-mediated membrane binding mirror the Ca 2ϩ requirements for sorcin binding to synexin.
The results of the overlay assays of the deletion constructs suggest that sorcin interacts specifically with the first 31 amino acids of synexin. Additionally, residues 1-8 are necessary to promote binding. It may be of importance that this extreme N-terminal region of synexin contains a three-fold sequence repeat (GYPP). This repeat unit (Fig. 10) is conserved between the human (32) and mouse (33) homologues. A glycine-, tyrosine-, and proline-rich region is also observed in Xenopus synexin (34) and the Dictyostelium homologue has 19 sequential GYPPQQ repeats (35,36). Sequence modeling of proline repeat units has produced results that suggest these sequences form novel secondary structures composed of polyproline ␤-turn helices (42). In fact, the secondary structure of the annexin I and II N-terminal domains may play an important role in S100 protein binding (45). The p11 binding site on annexin II (amino acids [1][2][3][4][5][6][7][8][9][10][11][12][13][14] has been shown to form an amphipathic ␣-helix with the hydrophobic amino acids Val-3, Ile-6, Leu-7, and Leu-10, which lie on one face of the helix, representing the major contact sites for p11 (44,46). Additionally, the S100C binding site on annexin I (amino acids 2-18) is predicted to form a similar helix with Met-2, Val-3, Phe-6, and Leu-7 providing potential contacts for S100C (10). These precedents suggest that the secondary structure of the synexin repeat unit may play a role in sorcin's recognition of its binding site on synexin.
Sorcin binding to the annexin XI N-terminal domain was unexpected due to the fact that calcyclin was previously identified as a Ca 2ϩ -dependent binding protein for annexin XI. However, it is important to note that the calcyclin binding site on annexin XI has been determined to be between amino acids 27-53 of the protein (31). Since human synexin and annexin XI share complete sequence identity within their first eight amino acids and annexin XI also contains two GYPP repeats, the binding data suggests that sorcin binds annexin XI early within its N-terminal domain. The GYPP repeat region of annexin XI ends at amino acid 15 and, therefore, the proposed sorcin binding site would not overlap with the known calcyclin binding site.
The studies with chromaffin granules demonstrated that sorcin is able to alter synexin-mediated aggregation, but not membrane binding. The control experiments with annexin I suggest that sorcin is not simply preventing the granules from making contact with each other. If this were the case, annexin I-mediated aggregation would be inhibited as well. Nor is sorcin decreasing the ability of synexin to aggregate granules simply by binding the membrane and preventing or excluding the annexin from interacting with the phospholipid membrane. In fact, the results of the densitometry from the chromaffin granule binding studies show that the phospholipid binding ability and Ca 2ϩ sensitivity of synexin appears unaltered in the presence of sorcin. Therefore, the protein-protein interaction between sorcin and synexin appears to play a role in this observed inhibition of aggregation. Since the inhibition was saturated at 2 mol of sorcin/mol of synexin, sorcin may bind to synexin as a dimer. Synexin has been shown to undergo selfassociation in vitro in the presence of Ca 2ϩ (43). The selfassociation has the same Ca 2ϩ dependence and cation specificity as the aggregation of chromaffin granules by synexin. Additionally, it has been proposed that synexin self-association may play a role in synexin-mediated granule aggregation. Therefore, the binding of a sorcin dimer to synexin may interfere with synexin self-association and may act to inhibit synexin-mediated granule aggregation.
This study of the in vitro biochemical properties of the synexin-sorcin interaction may provide important tools to enhance understanding of the physiological functions of both proteins. The Ca 2ϩ -dependent binding of sorcin to synexin may regulate synexin involvement in membrane trafficking, lipid organization, or ion channel formation. As demonstrated with the chromaffin granule binding studies, synexin may regulate sorcin by recruiting it to membranes. Recently, it has been demonstrated that sorcin associates with (30) and modulates (39) the cardiac ryanodine receptor when added to the cytoplasmic side of the receptor. In this role, sorcin, and possibly synexin by association, may play a role in Ca 2ϩ homeostasis in certain cell types. In the future, disruption of the synexin-sorcin interaction in FIG. 10. A conserved sequence motif is present in the synexin N terminus. Residues 1-31 of synexin contain a threefold GYPP repeat. This repeat unit is conserved between the human and mouse homologue. A glycine-, tyrosine-, and proline-rich region is also observed in Xenopus synexin and the Dictyostelium homologue has 19 sequential GYPPQQ repeats. Human annexin XI also contains two GYPP repeats at the N terminus of the protein, and the first 8 amino acids of this protein are identical to those found in human synexin. cells using peptides or expression constructs incorporating the sequence of this binding site may provide insight into the cell biological role of this complex.