A Novel Topology and Redox Regulation of the Rat Brain K+-dependent Na+/Ca2+ Exchanger, NCKX2*

In this study we have examined the roles of endogenous cysteine residues in the rat brain K+-dependent Na+/Ca2+ exchanger protein, NCKX2, by site-directed mutagenesis. We found that mutation of Cys-614 or Cys-666 to Ala inhibited expression of the exchanger protein in HEK-293 cells, but not in an in vitro translation system. We speculated that Cys-614 and Cys-666 might form an extracellular disulfide bond that stabilized protein structure. Such an arrangement would place the C terminus of the exchanger outside the cell, contrary to the original topological model. This hypothesis was tested by adding a hemagglutinin A epitope to the C terminus of the protein. The hemagglutinin A epitope could be recognized with a specific antibody without permeabilization of the cell membrane, supporting an extracellular location for the C terminus. Additionally, the exchanger molecule could be labeled with biotin maleimide only following extracellular application of β-mercaptoethanol. Surprisingly, mutation of Cys-395, located in the large intracellular loop, to Ala, prevented reduction-dependent labeling of the protein. The activity of wild-type exchanger, but not the Cys-395 → Ala mutant, was stimulated after application of β-mercaptoethanol. Co-immunoprecipitation experiments demonstrated self-association between wild-type and FLAG-tagged exchanger proteins that could not be inhibited by Cys-395 → Ala mutation. These results suggest that NCKX2 associates as a dimer, an interaction that does not require, but may be stabilized by, a disulfide linkage through Cys-395. This linkage, perhaps by limiting protein mobility along the dimer interface, reduces the transport activity of NCKX2.

are a crucial component of the Ca 2ϩ efflux process and have been extensively investigated in a wide range of tissues, particularly in the heart and brain (2,3). Various functional and molecular studies have revealed the existence of two families of Na 2ϩ /Ca 2ϩ exchanger proteins that share sequence similarity in two intramolecular homologous domains known as ␣-repeats (4). One family, Na ϩ /Ca 2ϩ exchangers (NCX), 1 are thought to catalyze the electrogenic exchange of either 3 or 4 Na ϩ for 1 Ca 2ϩ (2,5,6). The NCX family is made up of at least three distinct gene products: NCX1 (7), NCX2 (8), and NCX3 (9). NCX1 is expressed at high levels in cardiac muscle, brain, and kidney and is also present to a lesser extent in many other tissues (10,11). NCX2 and NCX3, in contrast, are expressed primarily in only two tissues: brain and skeletal muscle (8,9). All three exchangers share an overall amino acid identity of ϳ70% that rises to more than 80% within the predicted transmembrane segments (TMS) (9). The second family, K ϩ -dependent Na 2ϩ /Ca 2ϩ exchangers (NCKX), are believed to catalyze the electrogenic countertransport of 4 Na ϩ for 1 Ca 2ϩ and 1 K ϩ (12)(13)(14). NCKX exchangers differ from NCX proteins in their absolute requirement for K ϩ , lower Ca 2ϩ transport rates, and primary amino acid sequence outside the ␣-repeats (2). NCKX1 was initially cloned from bovine rod photoreceptors and was believed to play a central and unique role in the mammalian phototransduction pathway because its ionic stoichiometry represented an adaptation to the unusual ionic environment of the vertebrate eye (15,16). However, evidence from functional measurements revealed some Na ϩ /Ca 2ϩ exchange processes that were dependent on K ϩ in tissues other than eye, for instance brain synaptic plasma membrane (17) and platelet (18). This result led to the search for other putative NCKX family members. Consequently, NCKX2 was first cloned from rat brain (19) and then from chick and human cone photoreceptors (20), and NCKX3 was recently cloned from brain and skeletal muscle (21). Expansion of the NCKX family suggests a wider role for K ϩ -dependent Na ϩ /Ca 2ϩ exchange in maintaining cellular Ca 2ϩ homeostasis than previously anticipated. The tissue-specific expression patterns of these known NCKX members may reflect the different Ca 2ϩ handling properties of different tissues or cells.
Cysteine accessibility studies have suggested that the initial hydropathy-based topological model of NCX1 needed to be re-* This work was supported by grants from the Canadian Institutes of Health Research and the Alberta Heritage Foundation for Medical Research. 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. vised so that mature NCX1 is now thought to contain nine TMSs with two re-entrant loops (22)(23)(24). The current topological model of NCKX, based solely on hydropathy analysis, is reminiscent of the original NCX model before modification. Recently, examination of the hydropathy profile for NCKX3 gave rise to a new topological model in which the C-terminal hydrophobic domain contains only five TMSs, thus placing the C terminus of the exchanger protein outside the cell (21), in conflict with the initially proposed NCKX model in which the C-terminal half contains six TMSs and an intracellular C terminus. Indeed, experimental determination of the topology of the Escherichia coli inner membrane protein YrbG, a putative bacterial Na ϩ /Ca 2ϩ exchanger, suggested the C-terminal half has five TMSs and placed the C terminus extracellularly (25).
Cardiac Na ϩ /Ca 2ϩ exchanger activity was observed to be enhanced dramatically after treatment with a combination of reducing and oxidizing reagents (26). Thiol-disulfide interchange was proposed to be the molecular mechanism underlying redox modification of exchange activity, although the precise amino acid(s) involved have not yet been identified (27). To date, experimental evidence for dynamic regulation of NCKXtype exchangers is quite limited. In this study, we have used site-directed mutagenesis to investigate the role native cysteine residues play in NCKK2 exchanger protein stability and in transport activity. A preliminary report describing some of these results was published previously in abstract form (28).

EXPERIMENTAL PROCEDURES
All molecular procedures were performed according to standard protocols (29,30) or the directions of reagent manufacturers, unless noted otherwise. Common chemical reagents were obtained from Fisher, Sigma, or BDH and were of analytical grade or better, unless indicated otherwise. 3-(N-Maleimidylpropionyl)biocytin (biotin maleimide, or MPB) was from Sigma or Molecular Probes. 4-Acetamido-4Ј-maleimidylstilbene-2,2Ј-disulfonic acid (AMS) was from Molecular Probes.
Construction of NCKX2 Mutants-The construction of the wild-type and the FLAG-tagged full-length rat brain NCKX2 cDNA was described previously (19). Site-directed mutagenesis was performed with the polymerase chain reaction (PCR) overlap extension method using the Expand High Fidelity PCR system from Roche Molecular Biochemicals. Briefly, a pair of complementary primers in which cysteine-coding nucleotides were changed to those for alanine were synthesized. PCR fragments were generated using these mutagenic primers and external primers that flanked convenient unique restriction endonuclease sites in a two-step process. The purified fragment was digested and subcloned into the correspondingly digested full-length exchanger clone in pBluescript II SK (Ϫ) (Stratagene). The cDNA clone plasmid was then digested with KpnI and BamHI, and the ϳ2.5-kilobase pair fragment containing the full-length NCKX2 clone was subcloned into the mammalian expression vector pcDNA3.1ϩ (Invitrogen). We made point mutations for each of the last four native cysteine residues of NCKX2, Cys-395 3 Ala, Cys-614 3 Ala, Cys-633 3 Ala, and Cys-666 3 Ala. We also generated combined cysteine to alanine mutations named according to the linear order of cysteines in NCKX2: C1-4 (Cys-16 3 Ala, Cys-24 3 Ala, Cys-154 3 Ala, and Cys-224 3 Ala), C1-5 (Cys- 16 3 Ala, Cys-24 3 Ala, Cys-154 3 Ala, Cys-224 3 Ala, and Cys-395 3 Ala), and C1-5,7 (Cys-16 3 Ala, Cys-24 3 Ala, Cys-154 3 Ala, Cys-224 3 Ala, Cys-395 3 Ala, and Cys-633 3 Ala). An additional cysteine residue was reintroduced back into the C1-5,7 construct to substitute Ser-105 in the N terminus (named C1-5,7ϩSer-105 3 Cys), or one residue at a time at selected sites between the putative loops in the C-terminal half. An HA epitope was inserted at the C terminus of FLAG-tagged NCKX2 as a 9-amino acid peptide extension (YPYDVPDYA) by a similar PCR overlap mutagenesis procedure as described above, and the HA-tagged construct was designated as FLAG-NCKX2-HA671. All constructs were confirmed by sequencing to ensure that no polymerase errors were introduced into the amplified segments.
Antibody Preparation-Affinity-purified rabbit antibody PA1-926 (Affinity Bioreagents, Inc.) was generated against a synthetic peptide corresponding to amino acid residues 90 -102 (DLNDKIRDYTPQP) of the rat brain NCKX2. Polyclonal antibody F was prepared at the Southern Alberta Cancer Research Centre Hybridoma Facility by immunizing rabbits with a glutathione S-transferase fusion protein containing amino acids 384 -463 of the rat brain NCKX2.
Expression in HEK-293 Cells-Transient expression of Qiagen-purified cDNA constructs in HEK-293 cells was performed using a standard calcium-phosphate precipitation protocol with BES buffer essentially as described previously (19). Two days following transfection, postnuclear extracts were prepared by solubilizing transfected cells for 20 min in ice-cold lysis buffer (1% Triton X-100, 0.5% deoxycholate, 0.14 M NaCl, 10 mM EDTA, 25 mM Tris-Cl, pH 7.4, 100 units/ml aproptinin, 0.1 mM phenylmethylsulfonyl fluoride) followed by centrifugation for 30 min at 16,000 ϫ g. Protein concentration was determined by Bradford assay (reagent from Bio-Rad) using bovine ␥-globulin as a standard. Immunoblotting was performed as described previously (19,31) using PA1-926 antibody or M2 anti-FLAG monoclonal antibody (Sigma) and detected using Pierce SuperSignal Plus ECL reagents.
In Vitro Translation-In vitro translation of wild-type or mutated NCKX2 was performed essentially as described previously (21). In brief, cDNA constructs in the pcDNA3.1(ϩ) vector were transcribed and translated in vitro using the TNT-T7 system (Promega) together with [ 35 S]-methionine (Amersham Biosciences), in the presence of 0.1% Triton X-100. Following an incubation of 90 min at 30°C, the products were resolved on an SDS-polyacrylamide gel, dried, and detected by autoradiography using Biomax MR film (Eastman Kodak Co.).
Indirect Immunofluorescence-Location of the HA epitope was determined using immunofluorescence essentially as described previously (19) with some modifications. In brief, HEK-293 cells transfected with the FLAG-NCKX2-HA671 construct were grown on glass coverslips that had been precoated with 1 mg/ml poly-D-lysine (Sigma). HA-tagged intracellular Ca 2ϩ release channel ryanodine receptor 3 (RyR3) construct (generous gift from Dr. W. Chen) was used as a control to demonstrate the integrity of the plasma membrane barrier. Transfected HEK-293 cells were rinsed in PBSCM (PBS supplemented with 0.1 mM CaCl 2 and 1 mM MgCl 2 , pH 7.4) and then incubated with the rabbit anti-HA polyclonal antibody (1:500) in PBSCM containing 0.2% gelatin for 1 h at room temperature. The cells were then fixed in 4% paraformaldehyde in PBSCM and blocked with 0.2% gelatin/PBSCM for 30 min. A rhodamine-conjugated anti-rabbit second antibody (1:500) was employed in 0.2% gelatin/PBSCM for 30 min. After extensive washing with PBSCM, the cells were then treated with M2 anti-FLAG monoclonal antibody (1:500) followed by a FITC-conjugated anti-mouse second antibody. In permeabilization experiments, the cells were first fixed with 4% paraformaldehyde and then permeabilized with 0.1% Triton X-100 before consecutive application of first and second antibodies as described above. Immunofluorescence microscopy was performed using standard epifluorescence optics on a Zeiss Axioscop II through a Fluar 63ϫ objective. Images were captured using a Spot digital camera and processed with Photoshop.
Cysteine-selective Labeling of NKCX2 Exchangers-Biotin maleimide labeling of the NCKX2 wild-type and mutated proteins followed methods described previously (32,33) with modification. In brief, 2 days after transfection, HEK-293 cells on 100-mm dishes were washed three times with 10 ml of PBSCM and then incubated with 3 ml of 2% ␤-mercaptoethanol (␤-ME) in PBSCM or 2% ethanol in PBSCM for 15 min at room temperature. Cells were washed three times with 10 ml PBSCM and treated with 2 ml of PBSCM containing 100 M biotin maleimide (20 mM stock in Me 2 SO) for 15 min. After washing with 10 ml of PBSCM twice, the reaction was quenched by application of 3 ml of 2% ␤-ME in PBSCM for 5 min followed by washing twice with 10 ml of PBSCM. In some experiments, 2 ml of 100 M AMS in PBSCM (20 mM stock in Me 2 SO) was applied before addition of biotin maleimide to block extracellular cysteine labeling. Postnuclear cell lysis and protein concentration measurements were carried out as described above.
All immunoprecipitation (IP) experiments were performed at 4°C. 1 mg of protein extract was used, and the volume of the supernatant was adjusted to 1 ml with IP buffer (lysis buffer supplemented with 0.1 mg/ml ovalbumin, 1 mM benzamidine, 2 g/l leupeptin, and 2 g/l pepstatin A). After centrifugation at 14,000 rpm for 5 min, the supernatant was precleared with protein A-Sepharose beads (Sigma) and transferred to a new tube. The supernatant was then mixed with 10 g of anti-FLAG monoclonal antibody by rotating for 2 h and followed by 100 l of 20% protein A beads for 30 min. The beads were washed consecutively by centrifuging at 3,000 rpm for 2 min, once each with IP buffer plus 0.5 M NaCl, IP buffer plus 0.1% SDS, and wash buffer (0.1% Triton X-100, 25 mM Tris-Cl, and 1 mM EDTA). The sample was then transferred to a fresh tube, and bounded proteins were recovered by adding 40 l of 4ϫ SDS sample buffer containing 8% ␤-ME and heating to 50°C for 5 min.
The proteins were separated on a 9% SDS-PAGE gel and transferred to nitrocellulose membranes. Biotin-labeled proteins were analyzed by incubating the membranes with PBS plus 0.1% Tween 20 containing 0.1 g/ml horseradish peroxidase-conjugated streptavidin for 1 h after blocking with 2% bovine serum albumin for 1 h. After washing, the membranes were then developed using SuperSignal Plus ECL reagents (Pierce). To assess the level of protein present in each lane, the membranes were stripped with 0.2 M NaOH for 15 min and reprobed with rabbit anti-NCKX2 polyclonal antibody F, followed by application of horseradish peroxidase-conjugated anti-rabbit IgG antibody. The membranes were developed using ECL reagents.
Calcium Imaging and Data Analysis-Calcium transport into transfected HEK-293 cells was measured by fura-2 fluorescent ratio digital imaging essentially as described previously (19,21) with modification. In brief, 2 days after transfection, HEK-293 cells grown on poly-Dlysine-precoated coverslips were loaded with 5 M fura-2 AM (Molecular Probes) and mounted in a perfusion chamber on a microscope stage. The ratio of fura-2 fluorescence was captured with excitation at 340 or 380 nm using the ImageMaster system (Photon Technology International). Several perfusion solutions were used: solution I (145 mM NaCl, 10 mM D-glucose, 0.1 mM CaCl 2 , 10 mM HEPES-trimethylamine, pH 7.4), solution II (in which the NaCl of solution I was replaced with 140 mM LiCl and 5 mM KCl), and solution III (in which the NaCl of solution I was substituted with 140 mM NaCl and 5 mM KCl). For testing activity of mutants, cells were initially perfused with solution I for 5 min, followed by alternating changes to solution II for 2 min. For investigating redoxdependent regulation of NCKX2, cells were first incubated with solution III for 17 min without collecting ratio imaging data. Upon changing to perfusion solution I for 2 min, fura-2 fluorescence measurements were started. Perfusion was changed successively to solution II for 2 min and solution I for 2 min. Then, the cells were incubated with either 2% ␤-ME or 2% ethanol in solution III for 15 min, followed by perfusion with solution III for 2 min. Finally, the cells were subjected consecutively to perfusion solutions I, II, and I for 2 min each.
Imaging data were analyzed as described previously (19,21) using the ImageMaster program and Excel (Microsoft). For redox experiments, all the ratio data were normalized to the height of the first peak ratio. The height of peaks (with base line subtracted) following treatment was compared with the control peak height before treatment. Data were then tested for statistical significance using one-way analysis of variance with Newman-Keuls multiple comparison.
Co-immunoprecipitation-The FLAG epitope-tagged NCKX2 (the FLAG epitope is amino acids DYKDDDDK) was created by altering the extracellular sequence found at amino acids 90 -97 (DLNDKIRD) in the rat brain NCKX2 as described previously (19). Consequently FLAGtagged NCKX2 is not recognized by the PA1-926 antibody (see Fig. 5), which is directed against amino acids 90 -102. Two days after cotransfection of the appropriate wild-type and FLAG-tagged constructs, HEK-293 cells were solubilized and NCKX2 was immunoprecipitated using M2 anti-FLAG monoclonal antibody as described above, except the only detergent in the lysis/IP buffer was 1% CHAPS. Protein A beads were then washed three times with IP buffer containing 0.3% CHAPS. Samples were eluted from protein A beads and analyzed by SDS-PAGE and immunoblotting using PA1-926 antibody and reprobed with anti-FLAG antibody, as describe above.
Native disulfide bonds are believed to play an important role in developing proper protein conformational folding (34). Thus, we speculated that Cys-614 and Cys-666 might form a structurally and functionally important cystine disulfide bond. Disulfide bonds are usually found extracellularly, as the cytoplasm is a reducing environment. This speculation regarding Cys-614 and Cys-666 would place the C terminus of NCKX2 outside the cell, in conflict with the original topological model for NCKX2 based on hydropathy analysis, where the C terminus as well as Cys-614 and Cys-666 were located intracellularly (19), as illustrated in Fig. 1A.
Location of the C Terminus of NCKX2-We previously used the epitope insertion approach to prove that the N terminus of mature NCKX2 was on the outside of the plasma membrane (19). To explore the location of the C terminus, an HA epitope was added at the C terminus of an NCKX2 construct that was also tagged with a FLAG epitope in the N-terminal extracellular loop (see Fig. 2C). The double-tagged construct named FLAG-NCKX2-HA671 was shown to be functional, as tested by calcium imaging (data not shown) and thus preserved the structural integrity of the NCKX2 protein. Cells transfected with vector alone revealed no remarkable fluorescent staining with FLAG and HA antibodies after permeabilization ( Fig. 2A,  Negative Control). To ensure the integrity of the plasma membrane barrier under our conditions, an HA-tagged endoplasmic reticulum-localized intracellular Ca 2ϩ release channel, the type3 ryanodine receptor, was used as a control (35). Indeed, the HA-tagged ryanodine receptor could be detected by HA antibody only after membrane permeabilization with 0.1% Triton X-100 ( Fig. 2A, HA-tagged RyR3). Controls using FLAGtagged NCKX2 (without the HA tag) and FLAG-NCKX2-HA671, omitting one or other of the primary antibodies, confirmed the specificity of the signal observed (Fig. 2B). Immunofluorescent analysis of cells expressing FLAG-NCKX2-HA671 using both anti-FLAG and anti-HA antibodies and two conjugated second antibodies (Fig. 2B) showed that both the FLAG tag at the N terminus and the HA tag added to the C terminus of NCKX2 could be recognized without permeabilization of the cell membrane, indicating the C terminus of NCKX2 is located extracellularly. This finding supports a new putative NCKX-type topology model (Fig. 2C), based on the hydropathy analysis of NCKX3, in which the C terminus is extracellular (21).

Redox-modulated Accessibility of Cysteine Residues to Biotin Maleimide and Effects of Cys-395 3 Ala Mutation-
The accessibility of the cysteine residues in wild-type and functional cysteine mutants was examined by covalently modifying the protein expressed in transfected HEK-293 cells with the membrane-permeant cysteine-selective reagent MPB. Cysteine res-

FIG. 2. Immunofluorescence studies of the C-terminal HA-tagged NCKX2 and a new putative topological model for NCKX2. A, HEK-293 cells
were transfected with either vector alone (Negative Control) or with HA-tagged intracellular Ca 2ϩ channel ryanodine receptor 3 (RyR3) and then subjected to immunofluorescent analysis using either monoclonal anti-FLAG antibody and FITC-conjugated anti-mouse antibody, or polyclonal anti-HA antibody and rhodamine (Rhod)-conjugated anti-rabbit antibody. UP, cells unpermeabilized before antibody addition; P, cells permeabilized with 0.1% Triton before antibody addition. B, unpermeabilized HEK-293 cells transfected with FLAG-NCKX2-HA671 were double-stained using monoclonal anti-FLAG antibody followed by FITC-conjugated anti-mouse antibody, and polyclonal anti-HA antibody followed by rhodamine-conjugated anti-rabbit antibody. Control experiments used FLAG-NCKX2 (without the HA tag) doublestained using both sets of antibodies, and FLAG-NCKX2-HA671 stained with only one or other of the primary antibodies. The images shown are representative of 12 microscopic fields analyzed from four different experiments. C, a new putative topological model for NCKX2 is proposed. In this model the C terminus is extracellular and putative transmembrane segment M6 is placed inside the cytoplasm to form part of the large intracellular loop. Cys-614 and Cys-666 may form a cystine disulfide bond. Sites for the FLAG and HA epitopes are shown, as well as the site of glycosylation (CHO). Other details are similar to those of Fig. 1A. idues in wild-type NCKX2 were not reactive when treated with extracellular application of 100 M MPB for 15 min (Fig. 3A). To certify the effectiveness of MPB labeling and the capability of the membrane-impermeant sulfhydryl-selective compound, AMS, to block extracellular cysteine labeling, the C1-5,7ϩSer-105 3 Cys construct was generated and shown to be functional by calcium imaging (data not shown). Cys-105 is located between the FLAG epitope (amino acids 90 -97) and the glycosylation site (Asn-112) (36). Therefore, Cys-105 is located extracellularly and should be accessible to both cysteine-selective reagents. Indeed, Cys-105 could be detected by extracellular application of MPB without any pretreatment of transfected HEK-293 cells, and this labeling was eliminated by preincubation with AMS (Fig. 3A). We speculated that all endogenous cysteine residues in NCKX2 must either undergo posttranslational modifications or be embedded within the protein so that MPB was not able to reach them (37).
Because the two remaining cysteine residues in the C1-5,7 mutant were unreactive with the MPB, the cognate cDNA was suitable as a background construct for further topological studies using the cysteine accessibility method. Mutants with a single cysteine reintroduced into the C1-5,7 background one at a time in the putative C-terminal loops (Ser-471, Ala-503, His-531, Lys-567, and Leu-569), as well as the C1-5 mutant in which the endogenous Cys-633 remained intact, were expressed in HEK-293 cells and analyzed by immunoblot and Ca 2ϩ imaging. C1-5 and the cysteine mutants at Ser-471, Ala-503, and His-531 were all expressed and functional. Lys-567 and Leu-569 cysteine mutants were expressed but were not functional. However, none of these expressed proteins could be labeling with the MPB reagent, and they were thus uninformative for our topological studies.
To investigate whether Cys-614 and Cys-666 form a disulfide bond, transfected HEK-293 cells were treated with the reducing agent ␤-ME, or ethanol as a control for ␤-ME treatment, for 15 min before incubation with the MPB reagent. The results demonstrated that endogenous cysteine residue(s) in wild-type NKCX2 and C1-4 became accessible for MPB labeling only following extracellular application of ␤-ME (Fig. 3B). Interestingly, C1-5,7 could not be labeled even after ␤-ME treatment (Fig. 3B). Thus we concluded that Cys-614 and Cys-666 were not accessible for labeling by the hydrophilic MPB even after treatment with ␤-ME, and so might be concealed within the hydrophobic portion of the membrane. Because the NCKX2 construct, C1-4, was labeled with MPB, but C1-5,7 was not, we reasoned that Cys-395, Cys-633, or both, might be involved in the ␤-ME-dependent labeling of the NCKX2 exchanger. By using the C1-5 and the Cys-395 3 Ala (C5) mutants we demonstrated that mutation of Cys-395 to Ala, the only cysteine residue located in the large intracellular loop, completely abolished the ␤-ME-dependent labeling of NCKX2 by MPB (Fig.  3B). We also found that after ␤-ME treatment of transfected HEK-293 cells, application of AMS could not block the subsequent MPB labeling of wild-type NCKX2 (Fig. 3C), confirming that Cys-395, the single cysteine underlying reduction-dependent labeling of NCKX2, was located inside the cell as predicted by hydropathy analysis (19).
Redox Regulation of Wild-type NCKX2 Activity and Involvement of Cys-395-The observation that MPB labeling of NCKX2 was ␤-ME-dependent led to the hypothesis that reducing agent might also play a role in regulating exchanger activity, as reported for cardiac NCX1 (26,27). Therefore, cells were subjected to two sequential switches to Na ϩ -free solution II (140 mM Li ϩ , 5 mM K ϩ ) to measure NCKX2 activity, separated by a 15-min incubation with either 2% ␤-ME or ethanol in physiological solution III (140 mM Na ϩ , 5 mM K ϩ ) followed by a 2-min wash with solution III. The magnitude of the difference between peak and base line for each perfusion-induced fluctuation in fura-2 ratio was compared before and after the application of reagents. Fig. 4A shows the effects of 2% ␤-ME or ethanol on the increase in fura-2 ratio for transfected HEK-293 cells. Cells transfected with vector alone (traces a and b) had no significant change in fura-2 ratio. After a 15-min application of 2% ␤-ME, the fura-2 ratio peak of wild-type NCKX2 (trace c) was enhanced to an average of 113.2 Ϯ 3.2%, and the peak of the Cys-395 3 Ala mutant (trace e) was decreased to 86.33 Ϯ 7.3%. In comparison, the control treatment with 2% ethanol for 15 min caused the fura-2 ratio peak after treatment in cells expressing wild-type (trace d) and Cys-395 3 Ala (trace f) to decrease to 81.4 Ϯ 8.4 and 71.6 Ϯ 19.8%, respectively. Similar small inhibitory effects of ethanol have been observed for a number of neurotransmitter receptors and ion channels, such as voltage-gated Ca 2ϩ channels and the glutamate receptor (38). One-way analysis of variance revealed that Ca 2ϩ transport activity of wild-type NCKX2 treated with ␤-ME differed significantly from that of the other three groups (p Ͻ 0.01). No significant difference was observed among results for the Cys-395 3 Ala mutant treated with ␤-ME, the wild-type treated with ethanol, and the Cys-395 3 Ala mutant treated with ethanol (p Ͼ 0.05). 3 Cys (C1-5,7ϩC) were treated with the MPB reagent with (ϩ) or without (Ϫ) preincubation of the AMS reagent. Postnuclear extracts from the treated cells were immunoprecipitated with monoclonal anti-FLAG antibody, subjected to SDS-PAGE, and transferred to a nitrocellulose membrane. The immunoblots were probed first with horseradish peroxidase-streptavidin (top), then stripped and reprobed with polyclonal anti-NCKX2 antibody F (bottom). B, HEK-293 cells transfected with vector alone (Neg), FLAG-NCKX2, FLAG-NCKX2 C1-4, C1-5, C1-5,7, or C5 mutants were treated first with either ␤-ME or EtOH before application of biotin maleimide. Subsequent experimental details are similar to those for panel A. C, HEK-293 cells transfected with vector alone (Neg), FLAG-NCKX2, and FLAG-NCKX2-C1-5,7-Ser-105 3 Cys (C1-5,7ϩC) were all treated first with 2% ␤-ME, and then subjected to MPB reagent application with (ϩ) or without (Ϫ) pretreatment with AMS reagent. Subsequent experimental details similar to those for panel A. These data are representative of at least three similar experiments.

Association of NCKX2 Monomers Does Not
Require Cys-395-To determine whether Cys-395 might be involved in forming an NCKX2 dimer, we co-transfected wild-type and FLAG-tagged NCKX2, or their Cys-395 3 Ala mutants. FLAGtagged NCKX2 proteins from extracts of the transfected HEK-293 cells were immunoprecipitated with anti-FLAG antibody and co-association of wild-type NCKX2 was detected by immunoblot with PA1-926 anti-NCKX2 antibody (Fig. 5). The amino acid sequence of NCKX2 recognized by the PA1-926 antibody was replaced by the FLAG epitope in the FLAG-tagged NCKX2 so that the PA1-926 antibody would only detect wild-type NCKX2 protein as seen in the FLAG only lanes of Fig. 5. Controls also indicated that the anti-FLAG antibody did not immunoprecipitate wild-type NCKX2. The data presented in Fig. 5 reveal that wild-type NCKX2 was co-immunoprecipitated together with FLAG-tagged NCKX2, thus indicating that NCKX2 existed as an oligomeric complex when transfected in HEK-293 cells. However, the Cys-395 3 Ala mutation in either FLAG-tagged or wild-type NCKX2 did not prevent co-immuno-precipitation, and thus Cys-395 was not required for the NCKX2 oligomeric complex.

DISCUSSION
In this study, we have prepared cDNA constructs that express the plasma membrane NCKX2 exchanger protein with various single or combined mutations of cysteine residues, to investigate the role of native sulfhydryls in the expression and function of NCKX2. We demonstrated that, of eight endogenous cysteine residues, both Cys-614 and Cys-666 were critical for functional expression of NCKX2 in HEK-293 cells. In mammalian cells, the cytosol is a reducing environment, which prevents the formation of inter-or intrachain disulfide bonds between intracellularly exposed cysteine residues. Therefore, disulfide bonds of an integral membrane protein most likely exist either extracellularly, embedded internally in the protein structure, or within the lipid bilayer. Thus, we speculated that Cys-614 and Cys-666 might form a structurally and functionally important cystine disulfide bond, exposed on the extracellular side of the plasma membrane.
To test this hypothesis, we examined the location of the nearby C terminus of NCKX2 with carefully controlled immunofluorescence experiments. Our data confirmed an extracellular location of the C terminus. On the basis of these data, we propose a new topology model for NCKX2 (Fig. 2C) that is consistent with both the prediction for NCKX3 (21) and the data on the putative bacterial Na ϩ /Ca 2ϩ exchanger protein, YrbG (25). These findings give rise to the possibility that NCKX-type exchangers may have a different topology than NCX-type exchangers in which the C terminus is believed to be inside the cell (3).
NCKX-type exchangers and NCX-type exchangers share no significant similarity in their amino acid sequences outside the ␣-repeat regions. However, both new models for NCKX-and NCX-type exchangers place the ␣-repeat regions on the opposite face of the membrane (22). Thus, it is possible that NCX and NCKX exchangers have similar conserved structural elements formed by the ␣-repeat regions, surrounded by a different overall transmembrane structure. It remains an intriguing possibility that such structural differences between NCKXand NCX-type exchangers may underlie their distinctive ion stoichiometry. The accuracy of this new NCKX-type exchanger topology will need more supportive proof from further experimental studies.
Studies using cysteine-scanning mutagenesis of NCX1 have revealed a novel C-terminal structure that differs remarkably from the previous putative topological model based on hydropathy analysis (3). A cysteine-labeling experiment using biotin The height of the second peak as a percentage of the first peak before application of reagents, is plotted (with base line subtracted in both cases). The peak ratio for wild-type NCKX2 subjected to 2% ␤-mercaptoethanol incubation was significantly different from the peak ratio for wild-type NCKX2 incubated with 2% ethanol (p Ͻ 0.01). No significant difference was found between the peak ratios for the Cys-395 3 Ala mutant after incubation with either 2% ␤-mercaptoethanol or 2% ethanol (p Ͼ 0.05). WT, wild-type.
FIG. 5. Cys-395 3 Ala mutation did not inhibit co-immunoprecipitation of wild-type and FLAG-tagged NCKX2. The left immunoblots demonstrate expression of the indicated constructs in the postnuclear cell lysate from HEK-293 cells co-transfected with wild-type NCKX2 and vector (WT), FLAG-NCKX2 and vector (FLAG), wild-type NCKX2 and FLAG-NCKX2 (WTϩFLAG), FLAG-NCKX2-Cys-395 3 Ala and wild-type NCKX2 (FLAG-C5ϩWT), and wild-type NCKX2-Cys-395 3 Ala and FLAG-NCKX2 (WT-C5ϩFLAG), detected by anti-FLAG antibody (top) and reprobed with PA1-926 anti-NCKX2 antibody (bottom). The right immunoblots show the results of co-immunoprecipitation of wild-type and FLAG-NCKX2 from cell lysates using anti-FLAG antibody, and detected with anti-FLAG (top) and with PA1-926 anti-NCKX2 antibody (bottom), respectively. maleimide demonstrated that the endogenous sulfhydryls of NCKX2 could not be detected under normal conditions. Therefore, cysteine residues were reintroduced, one at a time, at sites in the putative C-terminal loops. None of these reintroduced cysteine residues reacted with biotin maleimide, even after ␤-ME treatment, suggesting they may be buried in the membrane and hence inaccessible for labeling. Furthermore, and discussed below, ␤-ME-dependent labeling of NCKX2 did not involve the endogenous Cys-614 and Cys-666 proposed to have an extracellular disposition. These results may indicate that the current model for threading of the C terminus protein of NCKX2 through the membrane needs substantial revision, as demonstrated by the topological studies of NCX1 (22,23).
Furthermore, treatment of NCKX2 with reducing agent stimulated its activity, an effect that was abolished by mutation of Cys-395 to Ala. Redox signaling has been shown to participate in modulating the activity of several ion channels and transporters (39), such as the N-methyl-D-aspartate receptor NR1 subunit (40), the cystic fibrosis transmembrane conductance regulator (41), the ryanodine receptor 1 (42,43), and the G protein-coupled inwardly rectifying K ϩ channel (44).
The activity of cardiac Na ϩ /Ca 2ϩ exchanger was also shown to be markedly stimulated after incubation with a combination of both reducing and oxidizing agents (26). This observation was verified using cloned canine NCX1.1 expressed in Xenopus oocytes (27). Interconversion of thiol and disulfide groups was initially speculated to be the molecular mechanism involved in redox modulation of exchange activity (26). Analysis of redox stimulation using mutated NCX1.1 constructs, although ruling out the involvement of individual cysteine residues, did not identify which amino acids were responsible. Indeed, redox-dependent stimulation of wild-type NCX1.1 was suggested to be primarily an elimination of Na ϩ -dependent inactivation (27), a process that involves several regions of the intracellular loop (45,46).
Our data, on the other hand, clearly demonstrate the necessity of Cys-395 in the reduction-dependent modification and stimulation of NCKX2. The remaining question is what molecular component might have been associated with Cys-395 before the reducing reagent abolished the interaction, hence rendering Cys-395 available for MPB labeling? An intracellular cysteine may be subject to reducible modifications, such as S-nitrosylation (47), palmitoylation (48), and, if protected from the reducing environment of the cytoplasm, a disulfide bond. The level of nitric-oxide synthase in HEK-293 cells is too low to be detected (49), so it is very unlikely that Cys-395 is S-nitrosylated. The common protein palmitoylation motif requires the palmitoylated cysteine residue to be located either within a TMS or at the cytoplasmic side of the membrane near a TMS, for instance 12 amino acids away as found in the ␣ 2A -adrenergic receptor or ␤ 2 -adrenergic receptor (50,51). These considerations indicate that Cys-395, found in the middle of the large cytoplasmic loop, is unlikely to undergo palmitoylation. Thus, formation of a disulfide bond is left as the most likely explanation.
We have demonstrated that NCKX2 monomers associate in a homo-oligomeric complex, based on the observation of co-immunoprecipitation of wild-type and FLAG-tagged NCKX2 proteins. However, our data show that the Cys-395 3 Ala mutation in neither wild-type nor FLAG-tagged NCKX2 inhibited self-association of NCKX2 monomers. This implies that oligomerization of NCKX2 monomers is primarily based on noncovalent, possibly hydrophobic, interactions. A similar situation has been documented for the extracellular Ca 2ϩ -sensing receptor, where dimerization still occurs without covalent interactions through intermolecular disulfide bonds (52). Once formed, however, such oligomeric complexes are then often "locked" in place, giving rise to a structural constraint conferred by the formation of the covalent disulfide bond (52).
Self-association of ion transporters can play an important role in function, regulation, or possibly cellular location (53), as found in the oligomerization of serotonin transporters (54) and glutamate transporters (55). The stimulation of NCKX2 activity induced by ␤-ME may be mediated through disruption of the putative Cys-395 disulfide bond, but appears not to require elimination of NCKX2 oligomerization, because NCKX2 mutants lacking Cys-395 can still form oligomers. Recently, NCKX1 has been reported to exist as a homodimer in bovine rod photoreceptors (56) and an inhibitory protein domain was proposed to be present at the contact site of the dimerized NCKX1 exchanger (57). Thus, even though Cys-395 of NCKX2 is not required for oligomerization, this covalent linkage may constrain conformation changes required for NCKX2 transport function, in a manner analogous to NCKX1. Therefore, reducing this bond, although not destroying oligomerization, may relieve a structural constraint, allowing increased exchange activity. This speculation is consistent with the requirement that Cys-395 is not accessible to the glutathione-mediated reducing environment of the cytoplasm, but once reduced by the small ␤-ME molecule, becomes accessible to MPB labeling.
The redox-dependent modification of Cys-395 in NCKX2 and its subsequent functional consequence is a novel aspect of dynamic regulation of K ϩ -dependent Na ϩ /Ca 2ϩ exchangers, although the detailed molecular mechanisms underlying this modulation still need to be resolved. It seems likely that redox regulation of NCKX2 serves a potential protective mechanism by increasing efflux of Ca 2ϩ in response to hypoxic and ischemic conditions.