Inhibitory Interaction of the Plasma Membrane Na+/Ca2+ Exchangers with the 14-3-3 Proteins*

The three Na+/Ca2+ exchanger isoforms, NCX1, NCX2, and NCX3, contain a large cytoplasmic loop that is responsible for the regulation of activity. We have used 347 residues of the loop of NCX2 as the bait in a yeast two-hybrid approach to identify proteins that could interact with the exchanger and regulate its activity. Screening of a human brain cDNA library identified the ϵ and ζ isoforms of the 14-3-3 protein family as interacting partners of the exchanger. The interaction was confirmed by immunoprecipitation and in vitro binding experiments. The effect of the interaction on the homeostasis of Ca2+ was investigated by co-expressing NCX2 and 14-3-3ϵ in HeLa cells together with the recombinant Ca2+ probe aequorin; the ability of cells expressing both NCX2 and 14-3-3ϵ to dispose of a Ca2+ transient induced by an InsP3-producing agonist was substantially decreased, indicating a reduction of NCX2 activity. The 14-3-3ϵ protein also inhibited the NCX1 and NCX3 isoforms. In vitro binding experiments revealed that all three NCX isoforms interacted with multiple 14-3-3 isoforms. 14-3-3 was bound by both phosphorylated and nonphosphorylated NCX, but the phosphorylated form had much higher binding affinity.

The maintenance of a low free Ca 2ϩ concentration is vital to the correct functioning of cells. To this aim, efficient systems have evolved to eject Ca 2ϩ or to transport it from the cytosol into subcellular organelles. Two major proteins perform this function in the plasma membrane: the Ca 2ϩ -ATPase (PMCA) 2 and the Na ϩ /Ca 2ϩ -exchanger (NCX) (1). The latter system belongs to a superfamily of proteins, which countertransport 3 Na ϩ ions for 1 Ca 2ϩ ion bidirectionally across the plasma membrane. Some cell types also contain a variant of the exchanger that, in addition to Na ϩ and Ca 2ϩ , also transports K ϩ (NCKX) (2). The driving force for the transport by NCX is the concentration gradient of Na ϩ and Ca 2ϩ and the membrane potential (3,4). Three NCX isoforms have been cloned as products of distinct genes: NCX1, NCX2, and NCX3 (5)(6)(7). They share biophysical and biochemical properties but exhibit differences in expression during development and in the various adult tissues. NCX1 is expressed at high levels in the heart and skeletal muscle but is also present in most other tissues. The expression of NCX2 is essentially restricted to brain and spinal cord, and that of NCX3 is essentially restricted to brain and skeletal muscles (8 -11).
The membrane topology of the exchangers has been studied particularly in NCX1, but it is very likely shared also by the two other isoforms. The predicted nine transmembrane segments (TMSs) can be divided into an N-terminal portion, composed of the first five TMSs, and a C-terminal portion composed of the last four TMSs. The two TMS portions are important for the binding and the transport of ions, and they are separated through a large intracellular loop of about 550 amino acids (12). Although this loop is not involved in Na ϩ and Ca 2ϩ translocation, it is responsible for the regulation of activity. Several factors are responsible for the regulation (3), among them the two transported ions, Na ϩ and Ca 2ϩ (13,14), the intracellular pH (15), metabolic components (e.g. ATP, phosphatidylinositol 4,5-bisphosphate (16), protein kinase A, and protein kinase C (17)), redox agents, hydroxyl radicals, H 2 O 2 , dithiothreitol, O 2Ϫ , Fe 3ϩ , Fe 2ϩ , Cu 2ϩ , and OH (18). So far, however, relatively little has become known about possible proteins that could interact with the exchanger and be involved in the regulation processes. One interesting early development in this context was the finding that the cardiac exchanger interacted with the cytoskeleton, specifically with ankyrin. The interaction was suggested as a possible mechanism responsible for the specialized localization of the exchanger to particular domains of the plasma membrane (19). More recently (20), the heart exchanger was found to interact with the C-terminal portion of calcineurin A␤. The interaction was inhibitory and was enhanced in cardiomyopathic hamster hearts. Clearly, the matter of possible NCX interactors is potentially interesting. It was thus decided to perform a yeast two-hybrid screening to identify other proteins that could interact with the exchanger, regulate its function, and, as a result, Ca 2ϩ signaling in the cell as well. It was decided to initiate the work with NCX2, by creating a "bait" construct made of a portion of the large cytoplasmic loop. A human brain cDNA library was screened, and the analysis of 2 ϫ 10 6 clones identified the 14-3-3⑀ and 14-3-3 proteins as possible interactors with NCX2; the interaction was further characterized using the 14-3-3⑀ protein. The effect of the interaction on the homeostasis of Ca 2ϩ in the living cells was examined by co-expressing the two proteins (NCX2 and 14-3-3⑀) in HeLa cells together with the recombinant Ca 2ϩ probe aequorin. The experiments showed that the ability of the cells that had expressed both NCX2 and 14-3-3⑀ to clear off a Ca 2ϩ transient induced by an InsP3-producing agonist was substantially decreased. This indicated a reduction of NCX2 activity by the 14-3-3⑀ protein. The 14-3-3⑀ protein also inhibited the NCX1 and NCX3 isoforms, and it was found that all three NCX isoforms interacted with multiple 14-3-3 isoforms. The 14-3-3 protein was bound by both the phosphorylated and nonphosphorylated forms of NCX, but the phosphorylated form had much higher binding affinity.

EXPERIMENTAL PROCEDURES
DNA Constructs-The portion of the large cytoplasmic loop of NCX2 corresponding to residues 283-631 was amplified by PCR using the forward primer 5Ј-CGGGATCCGCACATTC-GTGGGC-3Ј and the reverse primer 5Ј-CGGTCGACTCTCC-ATCCCCTTG-3Ј and subcloned into the BamHI-SalI sites of the vector pGilda to create a bait for the two-hybrid screening.
Cell Cultures and Transfection-Human neuroblastoma SH-SY5Y and HeLa cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 2 mM glutamine, penicillin (60 g/l), and streptomycin (120 g/l) in 75-cm 2 Falcon flasks at 37°C. The cells were transfected using a calcium phosphate method. For the aequorin experiments the cells were plated onto 13-mm glass coverslips the day before transfection. 3 g of plasmid DNA (or 1.5:1.5 g in the case of co-transfection) were used for every well. 36 h after transfection, the cells were used. Cerebellar granule cells were dissociated from the cerebellum of 6 -7-day-old Wistar rats and plated on poly-L-lysinetreated plates in Dulbecco's modified Eagle's medium (Hepes modification; Sigma) supplemented with 10% fetal calf serum, 100 g/ml gentamicin, 7 M p-aminobenzoic acid, 100 g/ml pyruvate, and 100 microunits/ml insulin at a density of 2.5 ϫ 10 6 cells/cm 2 , in the presence of 5.3 mM KCl. After 24 h, 10 M cytosine arabinofuranoside was added to inhibit mitotic cell growth. The cultures were maintained at 37°C in a water-saturated 8% CO 2 , 92% air atmosphere for 4 days.
Yeast Two-hybrid Screening-The LexA-based two-hybrid screening was performed according the standard protocol (BD Biosciences) using a human brain cDNA library (Invitrogen). Yeast EGY48 cells first transformed with the bait vector (NCX2-pGilda) and a p8op-LacZ reporter gene plasmid (pSH18-34) and then co-transformed with the library of cDNA plasmids were plated on 24 ϫ 24-cm Petri dishes containing synthetic dropout (SD) agar without histidine, uracil, and tryptophan in the presence of glucose. The plates were incubated at 30°C until colonies appeared. 2 ϫ 10 6 library transformants were harvested from the plates and screened for two-hybrid interactions. To this aim, the library transformants were plated on a selective induction medium without histidine, uracil, tryptophan, and leucine in the presence of galactose and raffinose as carbon sources (SD/Gal/Raf/ϪHis/ϪUra/ϪTrp/ϪLeu) and incubated at 30°C until positive colonies appeared. The assay for ␤-galactosidase activity then followed on a selective medium supplemented with 5-bromo-4-chloro-3-indolyl-␤-Dgalactopyranoside (X-gal). Genomic DNA from positive yeast colonies was extracted, transferred to E. coli using a standard transformation protocol, and sequenced. To ensure that the two-hybrid interactions were specific, the cDNA from positive clones was retransformed into yeast EGY48 together with the bait construct (NCX2-pGilda) and the reporter gene plasmid pSH18-34.
Co-immunoprecipitation-Granule cells were cultured for 4 days at a density of 2.5 ϫ 10 6 cells/well in a 6-well plate at 37°C. The cells were rinsed with PBS, and crude membrane proteins were prepared from the lysed cells. The cells were centrifuged and solubilized in 10 mM Tris-HCl, pH 8.0, 1 mM EDTA, 0.5% SDS. NET buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 0.25% gelatin, 0.1% Nonidet P-40, 1 mM EDTA) was added to dilute the SDS to a final concentration of 0.2%. Triton X-100 and sodium deoxycholate were added to final concentrations of 0.3 and 0.5%, respectively. The mixture was incubated for 30 min at 4°C, and after centrifugation at 13,000 rpm the supernatant was incubated with the primary antibody (3 l of anti-NCX2) (10) at 4°C on a rocking plate for 1-2 h. To recover the immunoprecipitates, 50 l of protein A/G Plus-Sepharose (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) were added to the mixture, which was incubated overnight at 4°C under gentle rocking. The protein A/G Plus-Sepharose-primary antibody complex was recovered by centrifugation for 1 min at 13,000 rpm and washed four times with 20 volumes of PBS. The material bound to protein A/G Plus-Sepharose was released by the SDS-PAGE sample buffer. The coimmunoprecipitates were analyzed by SDS-PAGE.
Immunofluorescence Analysis-HeLa cells were transfected with NCX1 or co-transfected with NCX1 and 14-3-3⑀ expression vectors. The cells were fixed with 3.7% formaldehyde in PBS for 20 min. After permeabilization of membranes with a 5-min incubation with 0.1% Triton X-100, the cells were washed with 1% gelatin (type B from bovine skin; Sigma) in PBS and immunostained with a primary antibody against NCX1 at a 1:100 dilution in PBS and the secondary antibody Alexa Fluor 594 (Molecular Probes). The images were acquired using an Olympus IX80 microscope (Olympus Optical Co., Ltd., Japan) with a 40ϫ Plan Neofluar objective and a Photometrics cooled CCD camera with a 35-mm shutter. The intensity of the fluorescence in the plasma membrane was quantified with a homemade program, in which the region of interest was specifically concentrated on the plasma membrane.
GST Pulldown Assay-Fusions of NCX genes and GST were produced in E. coli (BL21) cells. Protein expression was induced by adding 1 mM isopropyl 1-thio-␤-D-galactopyranoside to the growing culture and the cells were incubated at 30°C for 1 h. The cells were centrifuged at 5000 rpm for 30 min, resuspended in 10 ml of lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, mix of protease inhibitors (Roche Applied Sciences), 1 mM phenylmethylsulfonyl fluoride and disrupted using a French pressure cell. The cells were then recentrifuged at 12,000 rpm for 20 min, and the GST-NCX recombinant proteins were purified by incubating with glutathione-Sepharose 4B at 4°C for 2 h. The supernatant was removed and the glutathione-Sepharose 4B pellet was washed three times with lysis buffer. The clear HeLa cell extract was prepared by lysing cells in 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.25% deoxycholic acid, 1% Nonidet P-40, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, mixture of protease inhibitors for 30 min in ice. The lysate was cleared by centrifugation at 13,000 rpm for 30 min. Different concentrations of cell lysate were added to GST-NCX beads and mixed by gentle rotation at 30°C for 1 h. The beads were recovered by centrifugation (2500 rpm for 5 min at 4°C), washed three times with lysis buffer, and resuspended in SDS-PAGE loading buffer prior to the immunoblot analysis.
In Vitro Phosphorylation and Dephosphorylation-GST-NCX fusions were expressed in BL21 cells and purified on glutathione-Sepharose 4B as described previously. The phosphorylation of GST-NCX was performed in the HeLa cell lysates (150 -200 g of protein) supplemented with 25 Ci of [␥-33 P]ATP and 15 mM MgSO 4 , to support phosphorylation in the presence of 10 nM of the phosphatase inhibitor calyculin A (Calbiochem) for 1 h at 37°C. The GST-NCX beads were then centrifuged at 2500 rpm for 5 min and washed three times with lysis buffer. Dephosphorylation was performed in the HeLa cell lysate supplemented with 25 Ci of [␥-33 P]ATP in the presence of 10 mM EDTA for 1 h at 37°C or after incubation of GST-NCX with HeLa cell lysate prepared in protein phosphatase lysis buffer (10 mM Tris-HCl, pH 7.0, 0.1 mM EDTA, 5 mM dithiothreitol, 0.5% Triton X-100) supplemented with 1 mM MnCl 2 in the presence of protein phosphatase-1 (10 units; New England Biolabs Inc., Ipswich, MA) or shrimp alkaline phosphatase (10 units; Fermentas, Ontario, Canada). The beads were washed three times with lysis buffer. The reactions were stopped by adding loading buffer, and the proteins were run on SDS-PAGE. The gel was then dried on a Slab Gel Dryer GD2000 (Hoefer; Amersham Biosciences) for 1 h, and the signal was analyzed on a STORM 820 PhosphorImager (Amersham Biosciences).
Ca 2ϩ Measurement with Recombinant Aequorin-Transfected cytAEQ was reconstituted by incubating HeLa or SH-SY5Y cells for 3 h with a 5 M concentration of the aequorin prosthetic group coelenterazine wild type (Molecular Probes) in Dulbecco's modified Eagle's medium supplemented with 1% fetal calf serum at 37°C in a 5% CO 2 atmosphere. Where indicated, the cells were loaded with 20 M 5-(and 6-)-carboxyeosin diacetate (succinimidyl ester; Molecular Probes) for 20 min in a Krebs-Ringer solution (135 mM NaCl, 5 mM KCl, 0.4 mM KH 2 PO 4 , 1 mM MgSO 4 , 20 mM Hepes, pH 7.4, at 37°C) (Krebs-Ringer; KRB) supplemented with 1 mM CaCl 2 and 0.1% glucose. The cells were then superfused with carboxyeosin-free solution for a further 10 min to allow carboxyeosin de-esterification and to wash off the residual compound in the bath solution. 13-mm round glass coverslips with the transfected cells were placed in a perfused, thermostated (37°C) chamber placed in close proximity to a low noise photomultiplier, with a built-in amplifier discriminator. The experiments were performed in the KRB physiological solution supplemented with 0.1% glucose and 1 mM CaCl 2 . The cytoplasmic Ca 2ϩ concentrations were measured after the addition of the InsP3-generating agonists ATP (100 M) in the KRB in the case of HeLa cells and 100 nM bradykinin (Sigma) in KRB in the case of SH-SY5Y cells. The experiments were terminated by lysing the cells with 100 M digitonin (Sigma) in a hypotonic Ca 2ϩ -rich solution (10 mM CaCl 2 in H 2 O) to discharge the remaining aequorin pool. The light signal from the discriminator was collected by a Thorn-EMI photon counting board and stored in an IBM-compatible computer for further analysis. The aequorin luminescence data were calibrated off-line into [Ca 2ϩ ] values, using a computer algorithm based on the Ca 2ϩ response curve of wild type aequorin as described in Ref. 21.

NCX2 Interacts with the ⑀ and Isoforms of the 14-3-3 Protein in a Yeast
Two-hybrid System-The portion of the large cytoplasmic loop of NCX2 corresponding to amino acids 283-631 was used as a bait in a yeast two-hybrid approach (Fig. 1A). The bait was cloned in the LexA-DNA binding domain-containing two-hybrid vector pGilda, and 2 ϫ 10 6 clones of a human brain cDNA library were screened and further tested for the activation of both lexAop-LEU2 and lexAop-lacZ reporter genes. From the 105 colonies that grew on a selective medium lacking histidine, uracil, tryptophan, and leucine, 12 were positive in a ␤-galactosidase selection assay. After sequencing the cDNA isolated from the positive clones, two isoforms of the 14-3-3 protein family (14-3-3⑀ and 14-3-3) were identified. The specificity of the interaction between the 14-3-3 proteins and the NCX2 bait construct were confirmed in the yeast two-hybrid assay (Fig. 1B). Yeast EGY48 were co-transformed with the expression vectors for NCX2 and protein 14-3-3, and the transformants were grown on a selective medium without leucine and further checked for ␤-galactosidase activity. No interaction was detected between the identified proteins and a bait construct only encoding the LexA-binding domain. For further experiments it was decided to focus on the 14-3-3⑀ isoform.
Interaction of the 14-3-3⑀ Protein with NCX2 in Rat Cerebellar Granule Cells-To investigate whether the interaction between NCX2 and 14-3-3⑀ occurred in a cell type where both proteins are endogenously expressed, co-immunoprecipitation experiments were performed on granule cells isolated from the cerebellum of 7-day-old rats, where the expression level of NCX2 is high. The cells were grown in the presence of physiological concentrations of KCl (5.3 mM) to prevent NCX2 degradation (10). The 14-3-3⑀ protein was found to be associated with the NCX2 protein of the granules (Fig. 2). No immunoreactivity was detected among proteins that bind nonspecifically to the control antibody, indicating that the association of 14-3-3⑀ with NCX2 was specific.
Monitoring of the Cytosolic Ca 2ϩ Concentration in HeLa Cells Overexpressing NCX2 and the 14-3-3⑀ Protein-The functional consequences of the interaction between 14-3-3⑀ and NCX2 were studied by measuring the transport activity of NCX2. The overexpression of the exchanger increases the cell FIGURE 1. Interaction between NCX2 and 14-3-3 proteins in a yeast two-hybrid system. A, membrane topology and domain structure of NCX with the sequence of the portion of the large cytoplasmic loop (amino acids 283-631) of NCX2 used to create the "bait" construct. The topology scheme refers to NCX1 but can be assumed to be valid for NCX2 and NCX3 as well. B, specific interaction between NCX2 and the 14-3-3⑀ and -proteins in a two-hybrid screening. Yeast EGY48 cells were co-transformed with the "bait" construct, the 14-3-3 expression plasmid, and the reporter vector pSH18-34. Transformants were grown in the selective medium lacking histidine, uracil, tryptophan, and leucine (left) and checked for ␤-galactosidase activity in the same medium in the presence of X-gal (right) (for additional details, see "Experimental Procedures"). ability to extrude Ca 2ϩ and, consequently, leads to the reduction of Ca 2ϩ concentration in the cytoplasm. The changes of Ca 2ϩ concentrations were thus measured in the cytoplasm of HeLa cells transfected with the recombinant Ca 2ϩ -sensitive photoprotein, aequorin (21), targeted to the cytoplasm (cytAEQ); co-transfected with cytAEQ and NCX2; or co-transfected with cytAEQ, NCX2, and 14-3-3⑀ expression vectors. The extrusion of Ca 2ϩ was studied following the induction of a cytosolic Ca 2ϩ transient obtained by applying to the cells 100 M ATP, a purinergic receptor agonist that is coupled to the generation of InsP3. The rapid rise in cytosolic Ca 2ϩ induced by the stimulation was followed by a gradual decline of the trace to a lower plateau. The decline comprises two components: the Ca 2ϩ efflux from the cytoplasm mediated by the endogenous PMCA and sarcoplasmic reticulum Ca 2ϩ -ATPase pumps and by NCX (when cells were transfected with NCX) and the Ca 2ϩ influx into the cell via plasma membrane (capacitative) Ca 2ϩ channels activated by the decrease of the Ca 2ϩ concentration in the endoplasmic reticulum. The changes in the cytosolic Ca 2ϩ transients induced by the overexpression of the proteins are shown in Fig. 3. The average peak value of 2.49 Ϯ 0.24 M (n ϭ 39) for the Ca 2ϩ transient in control cells (only transfected with cytAEQ) was reduced to 1.77 Ϯ 0.17 M (n ϭ 25) in the cells expressing NCX2. However, in the cells expressing NCX2 and the 14-3-3⑀ protein, the peak of the Ca 2ϩ transient was significantly higher (peak amplitude 2.31 Ϯ 0.12 M; n ϭ 25) than in those expressing only NCX2 and was followed by a slower declining phase. The data suggest an inhibitory effect of the 14-3-3⑀ protein on the function of NCX2.
Interaction between Recombinant GST-NCX Proteins and 14-3-3⑀-It was then decided to examine whether 14-3-3⑀ interacted in vitro with NCX2 and the other two isoforms, NCX1 and NCX3, by using pulldown assays. To this aim, the portions of the large cytoplasmic loop of NCX1 and NCX3, corresponding to that of NCX2, which had been used to create the bait, and that of NCX2 were fused in frame with a GST protein using a pGEX4T1 expression vector. The GST-NCX1, GST-NCX2, or GST-NCX3 constructs were expressed in BL21 cells as described under "Experimental Procedures." Fig. 4 shows that 14-3-3⑀ was bound by all three NCX isoforms. The binding of NCX2-GST to 14-3-3⑀ was compared with that of NCX1-GST and NCX3-GST. Equivalent amounts of 14-3-3⑀ protein were bound by all three NCX isoforms. No interaction was detected between 14-3-3⑀ and glutathione-Sepharose beads containing GST alone.   4. 14-3-3⑀ interacts with NCX1, NCX2, and NCX3 in vitro. GST-NCX fusion proteins were expressed in BL21 cells for 1 h at 30°C. The bacteria were then lysed and the supernatant was incubated with glutathione-Sepharose 4B for 2 h at 4°C. The HeLa cells lysate (1-2 mg/ml) was incubated with the Sepharose beads containing GST-NCX1, GST-NCX2, or GST-NCX3 fusions (0.5 mg/ml) or with GST alone. Proteins from the cell lysate (Lysate) and retained by the Sepharose beads were separated by SDS-PAGE and immunoblotted with an anti-14-3-3⑀-specific antibody (ab).
Effect of 14-3-3⑀ Overexpression on NCX1 and NCX3 Transport Activity in HeLa Cells-It was then examined whether the 14-3-3⑀ protein had the same effect on the activity of NCX1 and NCX3 as it had on that of NCX2. HeLa cells were transfected with recombinant cytAEQ or co-transfected with cytAEQ and NCX1 (or NCX3) or with cytAEQ, NCX1 (or NCX3), and 14-3-3⑀ expression vectors. Ca 2ϩ transients were induced by stimulation with the InsP3-generating agonist ATP, as described above.
The changes in the Ca 2ϩ transients are shown in Fig. 5. As expected from the results on NCX2, the 14-3-3⑀ protein also inhibited the activity of NCX1 and NCX3. The peak of the Ca 2ϩ transient, [Ca 2ϩ ] i , in the cells co-transfected with NCX1 (or NCX3) and 14-3-3⑀ was significantly higher than in those transfected only with the exchangers. In the case of NCX1, the average peak value of the transient was increased from 1.24 Ϯ 0.

The Overexpression of 14-3-3⑀ Does Not Affect the Level of Expression or the Membrane Targeting of Recombinant NCX-
Experiments were performed to rule out the possibility that the inhibition of NCX activity by recombinant 14-3-3⑀ was due to a decreased expression of NCX or to a change in its targeting to the plasma membrane. Fig. 6A shows a Western blot of membrane fractions of HeLa cells transfected with NCX1 or cotransfected with both NCX1 and 14-3-3⑀; the level of expression of NCX was the same in the two cell batches. Fig. 6B shows the immunofluorescence of recombinant NCX1 in HeLa cells transfected with the exchanger alone or co-transfected with the exchanger and 14-3-3⑀. The fluorescence in the plasma mem-brane was evaluated using the program described under "Experimental Procedures." The histograms on the right of Fig.  6B show the equivalent levels of NCX1 in the plasma membrane of a large number of transfected and co-transfected cells. Fig. 6C shows that the recombinant exchanger and 14-3-3⑀ colocalize at the plasma membrane.
Effect of 14-3-3⑀ on the NCX Activity in HeLa Cells Loaded with Carboxyeosin-Previous work in the laboratory (22) had shown that PMCA4 was also inhibited by the 14-3-3⑀ protein.
PMCA4 is the plasma membrane Ca 2ϩ -ATPase isoform present endogenously in HeLa cells (23). Therefore, it was necessary to eliminate the possibility that the overall inhibition of Ca 2ϩ extrusion in HeLa cells co-transfected with NCX and the 14-3-3⑀ protein was partially (or even totally) due to the inhibition of the endogenous PMCA4. The effect of 14-3-3 on the Ca 2ϩ extrusion in HeLa cells expressing NCX was thus examined in the presence of the cell penetrant version of the inhibitor of PMCA ATPases, 5-(and 6-)-carboxyeosin diacetate, succinimidyl ester (carboxyeosin). Control HeLa cells transfected with the cytAEQ expression vector, treated with carboxyeosin, and exposed to ATP showed a markedly slowed decay of the Ca 2ϩ peak and a modestly higher [Ca 2ϩ ] i level at the peak of the transient with respect to control cells. This can be logically attributed to the inhibition of the endogenous plasma membrane Ca 2ϩ -ATPase. On average, carboxyeosin peak increased the half-time of [Ca 2ϩ ] i decay by 100%, from 64 Ϯ 8 to 125 Ϯ 16 s, n ϭ 10 (Fig. 8A). The height of the peak, by contrast, was markedly higher when cells were co-transfected with the NCXs and 14-3-3 with respect to those transfected only with the NCXs. This was in line with the accepted low affinity of NCX for Ca 2ϩ ; it was expected that its efficiency of ejection would be higher when Ca 2ϩ in the cell was at the peak level. For this reason, the effect of 14-3-3 on the height of the peak must be considered as a more significant parameter in the evaluation of the inhibitory effect of 14-3-3 on NCX. HeLa cells were then co-transfected with cytAEQ expression vec- tor and the plasmids encoding NCX1 (or NCX2/NCX3) or with cytAEQ expression vector, the plasmids encoding NCX1 (or NCX2/ NCX3), and 14-3-3⑀. Prior to the Ca 2ϩ measurements, the cells were loaded with 20 M carboxyeosin for 20 min. Fig. 8B shows that even in the presence of carboxyeosin, which eliminated PMCA4 activity, the expression of 14-3-3⑀ increased the height of the peaks of [Ca 2ϩ ] i in cells overexpressing the exchangers (in the case of NCX1, from 1.00 Ϯ 0. It was decided to test whether recombinant 14-3-3⑀ also inhibited the activity of endogenous NCX in cells that express it. SH-SY5Y were used for the experiments. Fig. 9 shows that the expression of 14-3-3⑀ indeed inhibited the NCX activity of SH-SY5Y cells. The figure shows a representative experiment. The average height of the peak of the Ca 2ϩ transient in the cytoplasm was 2.88 Ϯ 0.06 M (n ϭ 7) in the controls and 3.3 Ϯ 0.07 M (n ϭ 7) in the case of cells transfected with 14-3-3⑀. Control experiments were performed in the presence of carboxyeosin to rule out the possible contribution of the inhibition of the PMCA pumps to the effect of 14-3-3⑀. Only a minor prolongation of the decay phase of the trace after the peak of the Ca 2ϩ transient was observed, in keeping with the accepted minor contribution of PMCA pumps with respect to NCXs to the overall Ca 2ϩ extrusion activity in excitable cells (not shown). A, crude membrane proteins were obtained as described under "Experimental Procedures" from HeLa cells transfected with NCX1 or co-transfected with NCX1 and 14-3-3⑀ expression vectors and separated by SDS-PAGE. The NCX1 expression level was evaluated with an anti-NCX1 polyclonal antibody (ab). B, HeLa cells were transfected with NCX1 or co-transfected with NCX1 and GFP-tagged 14-3-3⑀ expression vectors. NCX1 was visualized by incubation with a primary rabbit anti-NCX1 polyclonal antibody and the Alexa Fluor 594-conjugated secondary antibody. The panel on the right shows a quantification of the NCX1 in the plasma membrane. The bars in the histograms indicate S.D. C, HeLa cells were co-transfected with NCX1 and GFP-tagged 14-3-3⑀ expression vectors; the image on the right shows the superposition of the two staining (red, Alexa Fluor 594-conjugated secondary antibody). FIGURE 7. GST-NCXs interact with multiple 14-3-3 isoforms. GST-NCX fusions were expressed in bacteria and purified on glutathione-Sepharose. Glutathione-Sepharose beads containing recombinant NCX1, NCX2, or NCX3 (1 mg/ml) proteins were incubated with HeLa cell extract (1-2 mg/ml) containing endogenous 14-3-3 proteins at 30°C for 1 h. The proteins bound to the beads were separated by SDS-PAGE and analyzed by immunoblotting using antibodies against the 14-3-3␤, -⑀, -, and -isoforms.

The 14-3-3 Protein Interacts with Both Phosphorylated and
Nonphosphorylated NCX-14-3-3 proteins generally (albeit not invariably) bind to phosphorylated serine or threonine res-idues located in defined consensus motifs (24). The analysis of the primary structure of the large cytoplasmic loop of NCX failed to reveal sequences corresponding to the 14-3-3 binding motif. However, all of the NCX isoforms are known to be phosphorylated on some serine residues of the large cytoplasmic loop (17); these residues could conceivably still serve as binding sites for the 14-3-3 protein. Alternatively, 14-3-3 could bind NCX in a phosphorylation-independent manner (such cases have been described; see "Discussion"). Pulldown assays were thus performed with either the phosphorylated or dephosphorylated forms of the NCX protein. GST-NCX was phosphorylated in vitro in the HeLa cell lysates by supplementing these with [␥-33 P]ATP/Mg 2ϩ and the inhibitor of protein serine-threonine phosphatases, calyculin A. The dephosphorylated form of NCX was obtained by supplementing instead the HeLa cell lysates with 10 mM EDTA during the incubation. As shown in Fig. 10A, the incubation with Mg 2ϩ /ATP led to the phosphorylation of GST-NCX by endogenous kinases in the lysates, whereas the incubation in the presence of EDTA totally prevented phosphorylation. Controls were run in which the samples were treated with alkaline phosphatase or with protein phosphatase-1 to rule out the possibility of nonspecific effects of EDTA. The results repeated those obtained with EDTA (not shown). The samples were then analyzed for binding of 14-3-3 by Western blotting; 14-3-3 bound to GST-NCX after incubation in both lysate types; however, the amount of 14-3-3 that became bound under the phosphorylating conditions was significantly higher. The ability of NCX to bind 14-3-3 was then examined at different times of incubation with Mg 2ϩ /ATP (Fig. 10B) (i.e. at presumably different levels of phosphorylation of Ser/Thr targets in the lysates). The amount of 14-3-3 bound to phosphorylated NCX increased between 0 and 1 min, remained approximately constant at 5 min, and then decreased progressively at 30 and 60 min. The late decrease could be due to competition for 14-3-3 binding by other proteins in the lysates, which would become phosphorylated with different kinetics under these conditions. Phosphorylated and Nonphosphorylated NCX Bind 14-3-3 with Different Affinities-As mentioned, cases have been reported in which 14-3-3 may bind to ligands via a mechanism not involving phosphoserines or phosphothreonines (25,26). However, the phosphorylation of the 14-3-3 ligands generally increases the affinity of 14-3-3 binding. Therefore, it was exam-  ined whether the phosphorylated and nonphosphorylated forms of NCX had different affinities for the 14-3-3 protein.
Bacterially expressed GST-NCX was incubated with different concentrations of HeLa extracts, which contain endogenous 14-3-3 proteins, under phosphorylating or dephosphorylating conditions. Immunoblotting tests (Fig. 10C) showed that after 1 h of incubation, the phosphorylated form of NCX bound the 14-3-3 protein even in the presence of the smallest amount of the lysate tested and in the presence of the presumed competition with other proteins of the lysate. By contrast, nonphosphorylated NCX was only able to bind 14-3-3 in the presence of much higher amounts of lysate.

DISCUSSION
The experiments described here were performed to identify possible regulatory protein partners of the NCXs. They have shown that various isoforms of the 14-3-3 protein family interact with NCX2 (but also with the other two exchangers) in a yeast twohybrid system. The 14-3-3 protein family consists of at least seven isoforms (14-3-3␤, -␥, -⑀, -, -, -, and -), which are capable of homo-or heterodimerization, are highly conserved, and are abundantly expressed in a wide range of eukaryotes. They were logical candidates as potential regulators of NCX, since they play an important role in multiple signaling pathways and regulate the activity of numerous proteins, including some membrane proteins (27). The central groove of each subunit of the 14-3-3 dimer, which is the most highly conserved region both within and across species, is a ligand-binding site. Only small differences in ligand binding specificity thus exist among isoforms (28). The present work had shown that the large cytoplasmic loop of NCX, which contains ␤ repeat sequence, conserved in all three isoforms, is the binding site for the 14-3-3 proteins.
The interaction between the 14-3-3 proteins and NCX in HeLa and SH-SY5Y cells limited the ability of the exchangers to dispose of a cytosolic Ca 2ϩ load induced by the stimulation of the cells with InsP3-generating agonists. Previous work (22) had shown that the 14-3-3 proteins had a similar inhibitory effect on at least one of the four isoforms of the other system for the extrusion of Ca 2ϩ from cells, the PMCA pump. The relative contributions of the PMCA pumps and of the NCXs to the total Ca 2ϩ ejection power vary from cell to cell, the exchanger being specially prominent in excitable cells like neurons. The model cells used in the experi- The probes were run on SDS-PAGE, and the gel was then dried and the NCX phosphorylation was analyzed on a STORM 820 PhosphorImager (Amersham Biosciences) or immunoblotted with an anti-14-3-3 monoclonal antibody (ab). The densities of the 14-3-3 bands were analyzed as described under "Experimental Procedures" and shown as histograms. B, NCX-GST was phosphorylated in the cell extract (20 g of protein) supplemented with Mg 2ϩ /ATP/calyculin, and the phosphorylation was stopped by adding 10 mM EDTA after 0, 1, 5, 30, and 60 min of incubation at 30°C. The samples were then analyzed for 14-3-3 binding. The histograms represent the density of the 14-3-3 bands on Western blot. C, GST-NCX was incubated with different concentrations of HeLa cell lysate under phosphorylating (Mg 2ϩ/ ATP) and dephosphorylating (10 mM EDTA) conditions for 1 h at 30°C. The proteins were then run on SDS-PAGE and analyzed by immunoblotting for the binding of 14-3-3 protein. Additional details are found under "Experimental Procedures." ments described here contained the isoform of the PMCA pump (PMCA4) that is inhibited by the 14-3-3 proteins. Thus, as expected, a portion of the decrease in total Ca 2ϩ extrusion ability of cells overexpressing the NCXs and the 14-3-3 protein was related to the inhibition of PMCA4. However, the experiments with the PMCA pump inhibitor carboxyeosin have shown that the exchangers were still inhibited by the 14-3-3 protein. In heart cells or in neurons, where the contribution of PMCAs to the total Ca 2ϩ extruding power is negligible, it is very likely that the inhibitory effect on the NCX will have significant consequences on the overall control of Ca 2ϩ homeostasis. Indeed, the experiments on SH-SY5Y cells transfected with 14-3-3⑀ have shown that the effect of 14-3-3⑀ was retained almost completely in the presence of carboxyeosin. In the majority of cases, target phosphorylation regulates the interaction with 14-3-3 proteins. Although the binding domain in target proteins generally contains phosphoserine or phosphothreonine residues within the consensus sequences RSX-pSXP or RXXXpSXP, a number of 14-3-3 target proteins do not contain sequences that conform precisely to these motifs. The existence of more than one "imperfect" binding motif significantly increases the affinity of the peptide that contains them for the 14-3-3 dimer. Thus, two "imperfect" sites may be sufficient to bind to the 14-3-3 dimer if they are in antiparallel orientation and able to achieve the correct spacing. In fact, a binding mode involving two imperfect sites may even be a more common mode of interaction between 14-3-3 proteins and their targets (28). As mentioned, the 14-3-3 proteins can also bind unphosphorylated targets, as was recently shown to be the case, for instance, for PMCA4 (22). These targets bind to 14-3-3 in the same location and can compete with phosphopeptides for binding (25,26). The canonical consensus 14-3-3 binding motif is not present in the sequence of the large cytoplasmic loop, but all three NCX isoforms are known to be phosphorylated (17) on a number of serine residues located within different motifs in the loop. For instance, putative protein kinase C phosphorylation sites inside the cytoplasmic loop of NCX1, which become phosphorylated after treatment with protein kinase C activators, are Ser-249, Ser-250, and Ser-357. Ser-357 is located inside the portion of the NCX1 used for the experiments described here and could be the serine that becomes phosphorylated to facilitate 14-3-3 binding. The other two NCX isoforms (NCX2 and NCX3) have been studied less extensively with respect to their possible phosphorylation sites. However, they are known to be phosphorylated as well. 14-3-3 interacted in vitro with both the phosphorylated and nonphosphorylated forms of NCX, but its affinity for the phosphorylated form was much higher. The number of proteins that may interact with the 14-3-3 proteins in the cell environment has been claimed to approach 200 (24), and a fierce competition must thus be assumed to prevail among cellular targets for the binding of the 14-3-3 protein. Thus, even if in principle nonphosphorylated binding sites on NCX could participate in the interaction with the 14-3-3, it is likely that the phosphorylation of NCX, which increases its affinity for the 14-3-3 protein, would augment its ability to interact with 14-3-3 in a competitive environment. Then, once bound to the phosphoserine-binding motif of the 14-3-3 protein through its phosphopeptide binding pocket, the ability of other 14-3-3 residues and/or the second half of the dimer to interact with nonphosphorylated sites on the target surface could become significantly increased. Furthermore, the assembly of the 14-3-3 dimer may interfere with these secondary, phosphorylation-independent interactions, since it could be influenced by the high affinity binding of the phosphopeptide. Finally, the finding that the same interactor regulates the activity of PMCA and NCX is interesting and suggests that cells may have developed a common strategy to modulate their Ca 2ϩ extrusion activity.