Interaction between Arabidopsis Ca2+/H+ Exchangers CAX1 and CAX3*

In plants, high capacity tonoplast cation/H+ antiport is mediated in part by a family of CAX (cation exchanger) transporters. Functional association between CAX1 and CAX3 has previously been inferred; however, the nature of this interaction has not been established. Here we analyze the formation of “hetero-CAX” complexes and their transport properties. Co-expressing both CAX1 and CAX3 mediated lithium and salt tolerance in yeast, and these phenotypes could not be recapitulated by expression of deregulated versions of either transporter. Coincident expression of Arabidopsis CAX1 and CAX3 occurs during particular stress responses, flowering, and seedling growth. Analysis of cax1, cax3, and cax1/3 seedlings demonstrated similar stress sensitivities. When plants expressed high levels of both CAXs, alterations in transport properties were evident that could not be recapitulated by high level expression of either transporter individually. In planta coimmunoprecipitation suggested that a protein-protein interaction occurred between CAX1 and CAX3. In vivo interaction between the CAX proteins was shown using a split ubiquitin yeast two-hybrid system and gel shift assays. These findings demonstrate cation exchange plasticity through hetero-CAX interactions.

Ca 2ϩ can accumulate to millimolar levels in the vacuole, whereas the concentrations are maintained in the micromolar range in the cytosol (3). This concentration gradient is established across the tonoplast in part by high capacity Ca 2ϩ /H ϩ exchange and via Ca 2ϩ pumping directly energized by ATP hydrolysis (4,5). Plant Ca 2ϩ /H ϩ exchangers were cloned by the ability of N-terminal truncated versions of the proteins to function in Saccharomyces cerevisiae mutants defective in vacuolar Ca 2ϩ transport (6 -8). The term CAX (cation exchanger) is now used to identify CAX1 and CAX2 as well as four other CAX transporters in the Arabidopsis genome (9). N-terminal truncations of these transporters are termed sCAXs. Interestingly, CAX3 (and sCAX3), which is most similar to CAX1 (77% identical at the amino acid level), is at best a weak vacuolar Ca 2ϩ transporter when expressed in yeast cells (10,11). Understanding of the biological roles these CAX transporters play in cell growth and in response to environmental stresses is only beginning to emerge.
Interplay between CAX1 and CAX3 has been suggested through genetic studies and yeast expression assays (11). In planta, deletions in CAX1 cause compensatory changes in gene expression among a battery of transporters, including heightened expression of CAX3 (12). Although cax1 and cax3 knock-out lines individually display subtle phenotypes, stunting and leaf chlorosis are readily apparent when both CAX1 and CAX3 are perturbed (11). In yeast suppression assays, when autoinhibited versions of CAX1 and CAX3 are co-expressed in yeast mutant cells, there is measurable Ca 2ϩ /H ϩ exchange activity that is not present when these regulated transporters are expressed individually. The challenge is to delineate the nature and specificity of this association and determine how the action of endomembrane Ca 2ϩ /H ϩ antiporters is integrated into various biological processes.
Frequently, transporters oligomerize to form pores. In some cases, the nature of the oligomer affects their function (13,14). Recent work with ammonium transporters from Arabidopsis suggests that allosteric interactions between isoforms may be essential for activity (15). In some cases, coupling between transporters may be a mechanism for increasing the dynamic range of transporter regulation and function.
Here we detail the interaction between the Arabidopsis Ca 2ϩ /H ϩ exchangers CAX1 and CAX3. We demonstrate an interaction between CAX1 and CAX3 in planta and define this interaction using both plant and yeast assays. We conclude that CAX1 and CAX3 can be combined to form functional transporters with distinct transport properties.

Histochemical Assay of CAX::GUS Gene Expression
Histochemical assays for CAX::GUS activity were performed according to the protocol described previously (11). These CAX::GUS seeds were germinated on one-half strength MS medium (18) and monitored daily. Samples were photographed using a Nikon (Tokyo, Japan) E600W microscope.

Construction of CAX Split Ubiquitin Plasmids
Cloning of CAX1 and CAX3, was conducted using in vivo recombinational cloning (23). Standard PCR conditions were used to amplify and clone CAX open reading frames into mbSUS vectors. The primers contained a B1 or B2 linker, and CAX1B1 (5Ј-ACAAGTTTGTACAAAAAAGCAGGCTCT-CCAACCACCATGGCGGGAATCGTGACAGAG-3Ј) and CAX1B2 (5Ј-TCCGCCACCACCAACCACTTTGTACA-AGAAAGCTGGGTAAGATGAGAAAACTCCTCCTCC-3Ј) were used for full-length CAX1 amplification. Similarly, a pair of primers, CAX3B1 (5Ј-ACAAGTTTGTACAAAAAAGCA-GGCTCTCCAACCACCATGGGAAGTATCGTGGAGCCA-3Ј) and CAX3B2 (5Ј-TCCGCCACCACCAACCACTTTGTA-CAAGAAAGCTGGGTAAGCTGAGAAAACTTCTCCTCC-3Ј), were used for full-length CAX3 amplification. For the construction of NubX and NubWT fusions, the split ubiquitin vectors NubX and NubWT were cleaved with EcoRI/SmaI and mixed with the purified PCR products for CAX1, CAX3, N-CAX1, and C-CAX3 to transform THY.AP5 cells, and transformants were selected on SC medium lacking tryptophan and uracil. For Cub fusions, the vector metYCgate was cleaved with PstI/HindIII and mixed with the same PCR products to transform yeast strain THY.AP4, and transformants were selected on SC medium lacking leucine. After the initial interaction assay, plasmids were extracted from the yeast strains and amplified in E. coli. The interactions were then verified by repeating the assays using the purified plasmids, whose inserts were sequenced for confirmation.
For detection of the CAX1-CAX3 complex in plants, the detergent-solubilized plant microsomal fractions were prepared as described above from 3-week-old 35S::HA-CAX3 transgenic seedlings with or without 15 mM LiCl for 8 h. As previously detailed, anti-CAX1 antibody was used to immunoprecipitate the CAX1-CAX3 protein complex. Extensively washed precipitates with CIP and PBST buffer were dissolved in protein loading buffer without DTT and resolved in native gel for PAGE. Similar to the yeast gel shift assay, after transfer to PVDF membrane, anti-HA and anti-CAX1 antibodies were used to detect protein complex formation.

Yeast Growth Conditions
Suppression Assay-Saccharomyces cerevisiae strain K667 or WX1 was transformed with yeast expression constructs as described previously (21). Transformed cells were tested for their capability to suppress the K667 Ca 2ϩ -sensitive phenotype (6) or WX1 NaCl-sensitive phenotype (20). Growth assays using K667 cells were conducted using YPD medium supplemented with or without CaCl 2 (50, 100, 150, or 200 mM), LiCl (30, 50, 60, or 100 mM), or NaCl (400, 500, or 600 mM). Lithium and CaCl 2 plates were made in YPD agar medium. The LiCl stock solution was filter-sterilized and added to the autoclaved YPD agar medium to final concentrations of 30, 50, 70, or 100 mM. K667 cells expressing various constructs were grown in SC selection medium overnight at 30°C. After a series of 5-fold dilutions with water, 5-l cell suspensions were spotted on the indicated media, and the cells were grown for 3 days. A growth assay with the WX1 strain was conducted on AP medium (28) supplemented with or without NaCl (20 -60 mM), and drop tests were conducted as described above. For these assays, cells were grown at 30°C for 4 days.
Inductively Coupled Plasma Mass Spectroscopy Analysis-Yeast culture conditions and sample processing were slightly modified from previous studies (29,30). For growth in Li ϩ -and Ca 2ϩ -supplemented media, K667 cells expressing the various constructs were inoculated in 5 ml of YPD plus 1 ⁄ 100 volume of 100ϫ mineral supplement stock (29), supplemented with 10 mM CaCl 2 or 500 M LiCl (final concentrations). For Na ϩ supplementation experiments, WX1 cells expressing the various constructs were inoculated in 5 ml of YPD supplemented with 100 mM NaCl (final concentration). The cultures were grown at 30°C to stationary phase, and 2.5 ml of each culture was collected by vacuum filtration using isopore membrane filters (1.2-m pore size) (Fisher). Cells were washed three times with 1 ml of 1 M EDTA disodium salt solution, pH 8.0, by vacuum filtration, followed by three washes with 1 ml of distilled, deionized H 2 O. The filters were dried in a 70°C oven for 48 h before inductively coupled plasma mass spectroscopy analysis was performed, as pervious described (31). Data from five repeats were calculated with a formula, (ion cax Ϫ ion vector )/ion vector ϫ 100, and expressed as means Ϯ S.E. (n ϭ 5).

Gel Mobility Shift Assay
Microsomal fractions from yeast cells expressing CAX variants were prepared using the glass bead method (32). Briefly, glass bead buffer (25 mM Hepes-KOH, pH 7.5, 10% sucrose, 3 mM EGTA, 2 mM DTT, 1 mM PMSF, 10 mM benzamidine, and 5 g/ml leupeptin) was added to the cells with an equal volume of glass beads (Sigma), and the mixture was vortexed for 3 ϫ 1 min at maximum speed with intervals of 10 min on ice. Finally, Triton X-100 was added to the lysate to a final concentration of 0.5%. The lysate was centrifuged at 5,000 ϫ g for 5 min, and the supernatant was saved. Equal amounts of protein (about 15 g) from samples were resolved on native 12% PAGE with SDS-PAGE running buffer, followed by transferring membranes and immunoblotting as described above. Proteins were detected with HA, c-Myc, and GFP antibodies (Covance, Berkeley, CA).

Structure and Topology Prediction for CAX1 and CAX3
Transmembrane domains and topology of CAX1 and CAX3 were predicted using the ARAMEMNON data base (available on the World Wide Web) and the transmembrane hidden Markov model (TMHMM version 2.0) program.

Split Ubiquitin Assay
The split ubiquitin system used in this assay was a matingbased split ubiqutin system developed previously (23,33). Approximately 40 clones of each THY.AP5 and THY.AP4 transformation with various Nub and Cub constructs were mixed and incubated in appropriate selective SC with and without G418. Cultures of the lag phase were used for mating, as described previously (13). After mating THY.AP4 and THY.AP5 strains on YPD for 8 h of incubation at 28°C, diploid cells were selected by replica plating on SC without tryptophan, leucine, and uracil for 2-3 days. Diploid cells were used to test on SC medium supplemented with 150 M methionine for 3 days. After sequencing confirmation, these Nub and Cub constructs containing CAX1 and CAX3 constructs were used for retransformation, and the interaction assay was repeated using the confirmed Nub and Cub constructs.
For the filter assay, diploid cells were grown for 3 days on sterilized filter sets placed on SC medium supplemented with His, adenine, and 150 M methionine. ␤-Galactosidase activity assays were conducted according to the standard X-gal filter assay protocol described previously (23). Color changes in the positive interactions could be detected within 3 h.

Preparation of Membrane Vesicles and Calcium Transport Measurements
Plants were grown hydroponically and membrane vesicles for yeast and plant Ca 2ϩ uptake were performed as described previously (12,21,34). Three-week-old plants (mainly roots) were collected after 16 h of treatment with or without 100 mM CaCl 2 (34). For the measurement of ⌬pH-dependent Ca 2ϩ uptake, vacuole-enriched membrane vesicles were incubated in buffer containing 0.3 M sorbitol, 5 mM Tris-MES (pH 7.6), 25 mM KCl, 0.1 mM sodium azide, and 0.2 mM sodium orthovana-date. The vesicles were added to 1 mM MgSO 4 and 1 mM ATP to reach steady state pH gradient for 5 min at 25°C before the addition of 45 Ca 2ϩ (6 mCi/ml; American Radiolabeled Chemicals, St. Louis, MO). The final concentration of Ca 2ϩ in the reaction mixture was 10 M and 1 mM. At the indicated times, aliquots of the reaction mix were removed and filtered through premoistened 0.45-m filters (Millipore, Bedford, MA), followed by washing with 1 ml of ice-cold wash buffer (0.3 M sorbitol, 5 mM Tris-MES, pH 7.6, 25 mM KCl, and 0.1 mM CaCl 2 ), the filters were air-dried, and radioactivity was determined by liquid scintillation counting. For metal competition experiments, ⌬pH-dependent 10 M 45 Ca 2ϩ uptake was measured at a 10 min time point in the presence of 100 M or 1 mM nonradioactive metals.

Semiquantitative Reverse Transcription and Gene Expression Analysis
For testing CAX1 and CAX3 expression, Col-0 seeds were analyzed at different times during germination. For LiCl induction, germinating seeds on day 4 were treated with 15 mM LiCl, and seedling RNA was extracted 8, 16, and 24 h post-treatment. For CAX1 expression in cax1-1, Col-0, and 35S:: HA-CAX3 transgenic plants, roots of 3-week-old plants grown hydroponically in B5 medium were harvested for RNA extraction. Total RNA was extracted with the RNeasy plant minikit (Qiagen, Valencia, CA). First strand cDNA was synthesized with Superscript II RNase H-reverse transcriptase kit with oligo(dT) 12-18 primer (Invitrogen) according to the manufacturer's instructions. PCR was performed with the following program: 94°C for 2 min to denature DNA and then from 94°C for 30 s, 60°C for 30 s, to 72°C for 30 s for 22, 25, or 30 cycles, followed by 72°C for a 10-min extension. ACTIN1 expression was used as internal control. For CAX1, forward primer was 5Ј-GCG TTA ATT AAG GCG CGC CTT CCG GCC ATA CCT GCC GCC A-3Ј, and reverse primer was 5Ј-GCG GGA TCC ATT TAA ATC ACC GCG GTT TCT TGC TCC AC-3Ј. For CAX3, forward primer was 5Ј-GCG TTA ATT AAG GCG CGC CCA TGG GAG AAC AGC ACA CAA C-3Ј, and reverse primer was 5Ј-GCG GGA TCC ATT TAA ATC CGC GAC AGA GTG AGC GAC AT-3Ј. For ACTIN1, forward primer was 5Ј-GTGCTCGACTCTGGAGATGGTGTG-3Ј, and reverse primer was 5Ј-CGGCGATTCCAGGGAACATTGTGG-3Ј.

Phenotypes When Co-expressing CAX1 and CAX3 in Yeast-
We demonstrated previously that co-expression of CAX1 and CAX3 can suppress yeast vacuolar Ca 2ϩ transport defects, whereas expression of either transporter individually failed to do so (11). In fact, in yeast assays, we have never been able to measure Ca 2ϩ /H ϩ transport mediated by CAX1 or CAX3. Expression of the deregulated version of CAX1, sCAX1, confers high levels of Ca 2ϩ /H ϩ transport; however, we have not yet isolated a deregulated version of CAX3 that functions as a Ca 2ϩ /H ϩ transporter. In the present study, we were interested in the potential change in substrate specificity that might occur during the putative CAX1 and CAX3 association. As shown in Fig. 1, co-expression of CAX1 and CAX3 could produce Li ϩ tolerance, a phenotype that was not observed when deregulated versions of the transporters were expressed individually (Fig.  1A). Data suggest that both proteins localize to the yeast vacuole (supplemental Fig. 1), and we have previously shown equivalent expression levels of these transporters in yeast (10,21). We also consistently obtained this phenotype regardless of the promoters and vectors used to express either CAX1 or CAX3 (data not shown; supplemental Table 1). Furthermore, we were unable to recapitulate these phenotypes when using other Ca 2ϩ /H ϩ transporter pairs (data not shown). For example, CAX1 or CAX3 co-expressed with the tonoplast-localized Arabidopsis CAX2 or CAX4 (8) did not confer any measurable changes in yeast ion tolerance (data not shown). These phenotypes imply that CAX1 and CAX3 may interact to form oligomers that confer distinct transport properties.
We also assessed the impact of co-expressing both CAX1 and CAX3 transporters on yeast metal accumulation (Fig. 1). This experiment allows us to determine if the metal accumulation in the cells was consistent with the tolerance phenotypes. We expressed sCAX1 ϩ vector, CAX1 ϩ vector, CAX3 ϩ vector, or CAX1 ϩ CAX3 and the vector control in the K667 (vcx1 pmc1 cnb1) mutant yeast strain, which is deficient in vacuolar Ca 2ϩ transporters (19) and grew the cells in liquid medium supplemented with LiCl to levels sufficient to facilitate detection in cell extracts by inductively coupled plasma mass spectroscopy. Increased Ca 2ϩ , Na ϩ , and Li ϩ accumulation was consistently observed in CAX1 ϩ CAX3-expressing yeast cells compared with the cells expressing these transporters individually (Fig.  1A). Although CAX3-expressing cells demonstrated some increased Li ϩ content, the yeast cells expressing CAX3 were not tolerant to LiCl-containing media (Fig. 1A).
We then measured Ca 2ϩ uptake in the vector control-, sCAX1-, and CAX1 ϩ CAX3-expressing cells (Fig. 1B). The ⌬pH-dependent 10 M 45 Ca 2ϩ uptake was measured at several time points into yeast vacuole-enriched membrane vesicles isolated from strains expressing vector, sCAX1, and CAX1 ϩ CAX3. We were able to measure uptake with vesicle from sCAX1-and CAX1 ϩ CAX3-expressing cells (Fig. 1B). However, we were unable to measure Ca 2ϩ uptake from CAX1-or CAX3-expressing cells, presumably because these transporters are autoinhibited (11). The ability to measure vacuolar Ca 2ϩ transport in the co-expressing cells suggests that both transporters are still predominantly localized on the tonoplast.
To analyze the properties of the putative CAX1 ϩ CAX3 transporter, competition experiments were performed (Fig.  1C). The pH-dependent 10 M 45 Ca 2ϩ uptake was measured at a single 10 min time point into yeast membrane vesicles isolated from strains expressing vector, sCAX1, and CAX1 ϩ CAX3. This approach allowed us to determine the effect of co-expressing the CAXs in terms of cation selectivity in comparison with sCAX1. Ca 2ϩ uptake determined in the absence of excess nonradioactive metal (control) was compared with Ca 2ϩ uptake determined in the presence of two concentrations (10ϫ and 100ϫ) of various nonradioactive metals (Fig. 1C). Inhibition of Ca 2ϩ uptake by nonradioactive Ca 2ϩ was used as a control, and as expected, Ca 2ϩ uptake was inhibited in both sCAX1 and CAX1 ϩ CAX3 transporters by excess Ca 2ϩ . Nonradioactive Ca 2ϩ , particularly 10ϫ concentrations, did not completely inhibit Ca 2ϩ uptake, further highlighting the low Ca 2ϩ affinity of the transporters. Ca 2ϩ uptake by sCAX1-expressing cells was inhibited by Cd 2ϩ , whereas the CAX1 ϩ CAX3-mediated Ca 2ϩ transport was not significantly inhibited by Cd 2ϩ . In contrast, CAX1 ϩ CAX3-expressing cells displayed more Ca 2ϩ uptake inhibition by Li ϩ (Fig. 1C). Both sCAX1-and CAX1 ϩ CAX3expressing cells displayed similar levels of Ca 2ϩ uptake inhibited by excess Na ϩ (Fig. 1C).
To obtain a clearer understanding of the potential function of CAX1 ϩ CAX3 in NaCl tolerance, we repeated these experiments in yeast strains lacking only the prevacuolar localized Na ϩ /H ϩ transporter NHX1p (35). In this strain, only yeast cells co-expressing CAX1 ϩ CAX3 displayed sodium tolerance (Fig.  1D). This phenotype was obtained regardless of the promoters and vector sets used to co-express either CAX1 or CAX3 (data not shown; supplemental Table 1). When this experiment was performed using CAX1 or CAX3 co-expressed with the tonoplast-localized Arabidopsis CAX2 or CAX4, the cells did not display a sodium-tolerant phenotype (data not shown). Furthermore, using whole cell metal accumulation measurements, these same cells grown in high levels of NaCl accumulated slightly more Na ϩ and Li ϩ than controls or cells expressing either transporter individually (Fig. 1D).
Functional Interactions between CAX1 and CAX3 in Planta-Our previous findings suggested that CAX1 and CAX3 functionally associate in Arabidopsis cells (11). In mature plants, CAX1 is expressed predominantly in green tissues, whereas CAX3 is expressed in roots (11). However, there are several instances when the two transporters are co-expressed temporally and spatially. We have shown that CAX3 expression is generally low in shoot/leaf tissue but not completely absent. In fact, CAX3 is up-regulated by abscisic acid in guard cells where CAX1 is highly expressed (36). Additionally, CAX1 and CAX3 are co-expressed in floral tissue and many other tissues during  Table 1). The empty vectors were expressed as controls. In the top panel, saturated liquid cultures of K667 containing various plasmids were diluted and then spotted onto selection medium or the yeast extract peptone dextrose medium supplemented with 50 mM LiCl. Photographs were taken after 2-3 days of growth at 30°C. Data from the metal content are from five repeats and were calculated with a formula, (ion cax Ϫ ion vector )/ion vector ϫ 100, and expressed as means Ϯ S.E. (n ϭ 5). B, Ca 2ϩ uptake mediated by sCAX1 and CAX1 ϩ CAX3. Time course of 45 Ca 2ϩ uptake into vacuolar vesicles prepared from the yeast strain K667 co-expressing sCAX1 with vector, CAX1 ϩ CAX3, and vector controls. Solid circle, pH-dependent 45 Ca 2ϩ uptake; empty square, uptake in the presence of the protonophore gramicidin. The Ca 2ϩ ionophore, A23187 (5 M), was added at 12 min. The data represent means of three replications, and the bars indicate S.E. C, Ca 2ϩ uptake by sCAX1 or CAX1 ϩ CAX3 into yeast endomembrane vesicles in the presence of other metals. Uncoupler-sensitive (⌬pH-dependent) uptake of 10 M 45 Ca 2ϩ , estimated as the difference between uptake with and without 5 M gramicidin, was measured in the absence (control) or presence of 10ϫ or 100ϫ nonradioactive CaCl 2 , CdCl 2 , LiCl, or NaCl after 10 min. Ca 2ϩ uptake values are shown following subtraction of the gramicidin background values and expressed as percentages of the control in the absence of any excess nonradiolabeled metals. The data represent the means of at least four repeats from three independent membrane preparations and are expressed as means Ϯ S.E. (n ϭ 3). D, the top panel shows that co-expression of CAX1 ϩ CAX3 in nhx1 mutant (WX1) suppressed the NaCl-sensitive phenotype. In the lower panel, lithium and sodium content of WX1 cells expressing CAX1 ϩ vector and CAX3 ϩ vector and co-expressing CAX1 ϩ CAX3 in YPD medium supplemented with 30 and 50 mM NaCl is shown. The plasmids were expressed as described above, but the yeast strains were grown in AP medium supplemented with NaCl. Photographs were taken after 3-4 days of growth at 30°C. Data from five repeats were calculated with a formula, (ion cax Ϫ ion vector )/ion vector ϫ 100, and expressed as means Ϯ S.E. (n ϭ 5). senescence (11) (see the Arabidopsis Membrane Protein Library on the World Wide Web). As shown in Fig. 2A, analysis of CAX promoter::GUS (␤-glucuronidase) reporters and RT-PCR demonstrated that both transporters were expressed in young seedlings during germination. This coincident expression during germination could signify that the CAX1-CAX3 complex has an important role in development. We thus investigated germination rates among control, cax1, cax3, and cax1/ cax3 lines on lithium-and calcium-containing media (Fig. 3A). In all stress conditions, germination rates from wild-type Col-0 were significantly different from cax1 or cax3 single or double mutant lines (p Ͻ 0.05). On 50 mM CaCl 2 -containing media, germination rates for cax1 or cax3 single mutants were significantly different from cax1/cax3 double mutants (p Ͻ 0.05), as determined by Student's t test (Fig. 3A). In contrast, on lithiumcontaining media, the cax1/3 double mutant had similar germination rates compared with both cax1 and cax3 lines. Using RT-PCR, we demonstrated that during calcium treatments, CAX1 expression significantly increased, whereas CAX3 was slightly induced (10, 37) (Fig. 3B), and exposure to lithium increased both CAX1 and CAX3 expression levels (Fig. 3C).
To gain additional insights into the function of the putative CAX1-CAX3 complex, we sought to compare and contrast the transport properties from lines expressing CAX1, CAX3, and CAX1 ϩ CAX3. Given that CAX3 expression levels are lower than CAX1, we ectopically expressed a HA-tagged CAX3 fusion driven by the CaMV 35S promoter in Col-0. After we verified expression of HA-CAX3 in the F3 generation (supplemental Fig. 1) and the induction of CAX1 by CaCl 2 (Fig. 3B), we prepared vacuole-enriched membrane vesicles from 35S::HA-CAX3 lines, vector controls, and cax1-1 lines with and without CaCl 2 treatment. Vector controls without CaCl 2 express low levels of both CAX3 and CAX1, and we have previously demonstrated that we were able to measure any Ca 2ϩ /H ϩ antiport in vacuole-enriched membrane vesicles from these lines (12,37). Control lines treated with CaCl 2 predominantly express CAX1, and we have shown that we can measure Ca 2ϩ /H ϩ

. Phenotypes of CAX mutants. A, germination of CAX mutants in
LiCl and CaCl 2 stress conditions. Stratified cax1-1, cax3-1, cax1/cax3, and wild type Col-0 seeds were sown on one-half strength MS medium supplemented with or without LiCl or CaCl 2 at the indicated concentrations. The germination rate at day 4 was calculated as means Ϯ S.E. At least three independent experiments totaling ϳ150 seeds were tested. B, CAX1 expression in Col-0, cax1-1, and 35S::HA-CAX3 lines, as determined by RT-PCR analysis, when plants were treated with or without CaCl 2 prior to membrane preparations. The intensities of the CAX1 PCR bands are given as a ratio to ACTIN1 expression. Experiments were repeated at least twice with similar results from each replicate. C, time course of CAX1 and CAX3 expression, as determined by RT-PCR analysis, in seedlings treated with LiCl from 0 to 24 h after transfer. The intensities of CAX1 and CAX3 PCR bands are given as a ratio to ACTIN1 expression. Experiments were repeated at least twice, with similar results from each replicate. exchange in these vesicles (Fig. 4A) (12,37). With cax1-1 lines, we could assess the influence of CaCl 2 treatment without CAX1 present. In the 35S::HA-CAX3 lines, we could obtain vesicles predominantly expressing HA-CAX3 (without CaCl 2 treatment) and vesicles expressing CAX1 ϩ HA-CAX3 (treatment with CaCl 2 ). Without CaCl 2 treatment, Col-0, cax3-1, and cax1-1 lines displayed similar low levels of Ca 2ϩ /H ϩ transport activity, and the 35S::HA-CAX3 lines showed very modest activity (Fig. 4A) (data not shown). CaCl 2 treatment significantly increased vacuolar Ca 2ϩ /H ϩ transport activity from both Col-0 and 35S::HA-CAX3 lines (Fig. 4A). We postulate that by comparing the activity in these conditions, we can compare and contrast the activity of CAX1 with that of the putative CAX1-CAX3 complex.
To analyze the transport properties in these vacuole-enriched membranes, we performed competition studies similar to those done previously in yeast (Fig. 1C). As stated earlier, in the 35S::HA-CAX3 lines without CaCl 2 , we were unable to measure substantial Ca 2ϩ /H ϩ transport (Fig. 4A); thus, the competition measurements were not done under these conditions. Only during CaCl 2 treatment, when we propose that CAX1 is active, were we able to measure uptake and perform competition measurements. In tonoplast-enriched vesicles from CaCl 2 -treated Col-0 lines (CAX1 activity high), Cd 2ϩ appeared to inhibit Ca 2ϩ transport; however, the inhibition was not as robust in CaCl 2 -treated 35S::HA-CAX3 lines (putative CAX1-CAX3 complex formation; Fig. 4C). In addition, lines expressing both CAX1 and CAX3 appeared to be more strongly inhibited by Li ϩ and Na ϩ than CAX1-expressing lines (CaCl 2 -treated Col-0). At 100ϫ concentrations of Li ϩ , the HA-CAX3 CaCl 2 -treated lines (where we postulate CAX3 ϩ CAX1 interactions) had only 30% of the Ca 2ϩ /H ϩ transport activity, compared with the CAX1-expressing lines (CaCl 2 -treated Col-0), which were 55% as active under these conditions (Fig. 4, B and C).
Co-immunoprecipitation of CAX1 and CAX3 in Planta-To analyze the interaction between the transporters in plants, we performed co-immunoprecipitation using proteins from microsomal preparations of transgenic plant leaves and HAand CAX1-specific antibodies (12) (Fig. 5). These microsomal fractions contain vacuoles, prevacuolar compartments, and other light vesicles; however, these membrane preps should allow preliminary observations regarding co-localization of CAXs. As controls, the V-ATPase antibody and other tonoplast antibodies were used to precipitate these protein samples (supplemental Fig. 2). All immunoprecipitates were resolved by SDS-PAGE and detected by immunoblotting with either HA, CAX1, or V-ATPase antibodies. As shown in Fig. 5, detection with CAX1 antibody demonstrated that both CAX1 and HA antibodies clearly precipitated CAX1 proteins of the appropriate size (ϳ50 kDa) from controls and HA-CAX3-expressing plants but failed to precipitate the same protein from cax1-1 lines, suggesting high efficiency and specificity of immunoprecipitation and binding between CAX1 and CAX3. Detection with the HA antibody showed HA-CAX3 proteins immunoprecipitated with both CAX1 and HA antibodies from HA-CAX3 transgenic plants but not from controls and cax1-1 lines, further suggesting that CAX1 and CAX3 physically interact (Fig.  5B). As expected, the V-ATPase antibody failed to precipitate either CAX1 or HA-CAX3, although it did precipitate V-ATPase proteins from all plants (Fig. 5C). Similar negative interactions with the CAX transporters were found using antibodies against the Arabidopsis H ϩ -pyrophosphatase (AVP1), TPC1 (two-pore channel 1), and TIP (tonoplast intrinsic protein) (supplemental Fig. 2). Thus, a nonspecific interaction due to incomplete solubilization between CAX1 and CAX3 is unlikely, since neither HA-CAX3 nor CAX1 interacted with several other tonoplast proteins.
To determine if LiCl treatment altered CAX1-CAX3 interactions, we carried out gel shift assays with 35S::HA-CAX3 plant protein extracts with and without LiCl treatment. As shown in Fig. 5D, after treatment with 15 mM LiCl for 8 h, 35S::HA-CAX3 plants accumulated much more putative CAX1-CAX3 complex resolved by a native gel shift assay than control lines. However, even in control lines, the complex could be detected in SDS-polyacrylamide gels after immunoblotting (Fig. 5D).
Yeast Split Ubiquitin Assay of CAX1 and CAX3 Interaction-To test whether CAX1 and CAX3 can physically interact in yeast, an optimized split ubiquitin system was used (13). This system allows detection of interactions between membrane proteins in vivo. Three-week-old Col-0, cax1-1, and 35S::HA-CAX3 seedlings were grown hydroponically with or without CaCl 2 for 16 h prior to membrane extraction. Whole seedlings (mainly roots) were then collected for vacuolar enriched membrane preparations. ⌬pH-dependent uptake of 10 M 45 Ca 2ϩ , estimated as the difference between uptake with and without 5 M carbonyl cyanide p-(trifluoromethoxy)phenylhydrazone (protonophore), was measured in the absence (control) or presence of 10ϫ or 100ϫ nonradioactive CaCl 2 , CdCl 2 , LiCl, or NaCl after 10 min. Ca 2ϩ uptake values are shown following subtraction of the carbonyl cyanide p-(trifluoromethoxy)phenylhydrazone background values and expressed as percentages of the control in the absence of any excess nonradiolabeled metals. Data represent means Ϯ S.E. from at least four replicates from three independent membrane preparations. A, the absolute Ca 2ϩ uptake from membrane preparations after the subtraction of the carbonyl cyanide p-(trifluoromethoxy)phenylhydrazone values. B, membrane transport analysis from Col-0 plants treated with CaCl 2 (expressing predominantly CAX1). Uncoupler-sensitive ⌬pH-dependent uptake of 10 M 45 Ca 2ϩ uptake in the presence of 10ϫ or 100ϫ (100 M or 1 mM) nonradioactive CaCl 2 , CdCl 2 , LiCl, or NaCl after 10 min. All results are means Ϯ S.E. of experiments from three independent membrane preparations. C, membrane transport analysis from 35S::HA-CAX3 plants treated with CaCl 2 (high CAX3 and CAX1 expression). Competition experiments were done as previously described.
Protein fusions of ubiquitin were constructed (Fig. 6, A  and B), and the interaction of CAX1 and CAX3 was monitored by the release of the artificial transcription factor PLV, activating lexA-driven reporter genes in the nucleus. As a positive control, we demonstrated the interaction between AtKAT1 subunits (13, 23) (Fig. 6). As negative controls, CAX1 and CAX3 did not interact with the membrane protein AtKAT1. Using this system, interactions were observed between CAX1 and CAX3 (Fig. 6). The interaction of CAX1 (Cub or Nub) and CAX3 (Cub or Nub) was also detected by measuring ␤-galactosidase activity of LacZ (Fig.  6D). These results provided independent evidence that CAXs were capable of assembling in a heterodimeric protein complex. FIGURE 5. In planta coimmunoprecipitation of CAX1 and CAX3. Microsomal proteins were extracted from wild type, cax1-1, and 35S::HA-CAX3 transgenic plants (from two independent lines). Solubilized proteins were immunoprecipitated (IP) with monoclonal HA antibody (A), polyclonal CAX1 antibody (B), and monoclonal V-ATPase antibody (C), respectively. After pulling down with protein A-agarose beads and extensive washing, proteins were resolved by SDS-PAGE and detected by using immunoblotting (IB) with either the HA antibody, CAX1 antibody, or V-ATPase antibody. Experiments were repeated at least three times, with a typical experiment shown. D, LiCl induction of CAX1-CAX3 complex. The membrane vesicles were prepared from 3-week-old 35S::HA-CAX3 seedlings treated with LiCl or, water (control) for 8 h. The dissolved proteins were immunoprecipitated with anti-CAX1 antibody, and the precipitates were resolved in native PAGE for immunoblotting with anti-CAX1 antibody and anti-HA antibody, respectively. Experiments were repeated at least three times with similar results. FIGURE 6. Protein-protein interactions. Schematic representation of the topology of full-length CAX1 and CAX3 and truncated proteins with or without a ubiquitin (N-terminal part (Nub); C-terminal part (Cub)) fusion. Topology was predicted using transmembrane domain hidden Markov model (TMHMM version 2.0). Rectangles represent the vacuolar membrane. In, cytosol side; Out, vacuolar lumen side. A, CAX1 full-length protein (CAX1) without or with Nub and Cub fusions at the N terminus (Nub-CAX1 or Cub-CAX1) or at the C terminus (CAX1-Nub or CAX1-Cub). B, CAX3 full-length protein (CAX3) without or with Nub fusions at the N terminus (Nub-CAX3). C, interaction growth assay. CAX1 and CAX3 interact in a mating-based split ubiquitin system. Interaction between CAX1 and CAX3 was determined by growth assay on synthetic minimal medium lacking all six amino acids but supplemented with 150 M methionine for 3 days. The empty Cub vector was used as a negative control, and the potassium channel, KAT1, from Arabidopsis thaliana was used as a positive control. CAX1-Cub, KAT1-Cub, or empty Cub vector (IV) was used as bait vector for interaction in combination with empty Nub vector, KAT1, Nub-CAX1, and Nub-CAX3. All experiments were repeated at least three times with similar results obtained in each replicate. D, X-gal filter assay. Yeast filters set on synthetic minimal medium lacking adenine and histidine but supplemented with 150 M methionine were grown for 3 days and then incubated for 30 min in the X-gal filter assay. All experiments were repeated at least three times with similar results obtained in each replicate.

DISCUSSION
CAX1 is a tonoplast Ca 2ϩ /H ϩ exchanger (12,37,38) that appears to be autoinhibited in planta, requiring trans-acting factors for optimal activity (8,30,34). Here we identify a potentially new mechanism for CAX regulation through the formation of "hetero-CAX" complexes, which may alter CAX transport properties. We show that CAX1 and CAX3 can be present in the same cells of plants and that they can interact directly, and we provide some arguments why this interaction could be physiologically important.
Tissue Expression of CAX1 and CAX3-Although CAX1 and CAX3 differentially accumulate in vegetative tissue, there are several conditions where their expression patterns overlap. For example, CAX1 and CAX3 both accumulate in reproductive organs and germinating seeds and during senescence (11). Furthermore, CAX3 expression can be enhanced more than 4-fold in the leaf guard cells, where CAX1 is highly expressed (36). The data in guard cells are the most definite example of coincident expression of CAX1 and CAX3 in the same cells. Other microarray data suggest that during osmotic stress, UV light treatment, and wounding, CAX3 expression in the aerial portions of the plants can reach levels equivalent to CAX1 (see the Arabidopsis Membrane Protein Library on the World Wide Web). Here, we have used CAX1::GUS and CAX3::GUS lines as well as RT-PCR analysis to demonstrate overlap in expression patterns during germination (Fig. 2). In general, CAX1 expression is higher than CAX3 and does not display the dynamic fluctuations in gene expression displayed by CAX3. We speculate that CAX3 may sometimes act as a co-factor with CAX1 to modulate transport function in response to various hormonal or environmental stresses. Guard cells may be the ideal model to further explore this dynamic Ca 2ϩ /H ϩ transport, given that basal levels of CAX1 are high in this tissue, and during abscisic acid treatment, both CAX1 and CAX3 are highly expressed.
CAX1 and CAX3 Forms a Transporter with Distinct Functions-Our results indicate that the CAX1-CAX3 complex in the plant may possess distinct transport functions. Here we show that co-expression of CAX1 with CAX3 produced LiCl tolerance in yeast, a unique phenotype not associated with expression of any of the deregulated CAXs (Fig. 1). It is interesting to note that sCAX1 expression in yeast conferred increased sensitivity to LiCl and NaCl levels as opposed to the salt tolerance for the CAX1 ϩ CAX3-expressing cells (Fig. 1). The ability to confer LiCl tolerance in yeast without producing appreciable NaCl tolerance has been documented (39). Presumably, LiCl accumulation at the yeast vacuole (mediated by NHX1) confers only a degree of LiCl tolerance (39), and such accumulation/ tolerance can be improved by expression of the CAX1-CAX3 complex. We have used transport assays to further demonstrate the differences among CAX1, CAX3, and CAX1 ϩ CAX3 (Fig. 1). Only yeast cells expressing CAX1 ϩ CAX3 demonstrated Ca 2ϩ uptake rates that could be inhibited by Li ϩ , further suggesting that this hetero-CAX complex has different transport properties.
The inability to measure CAX3-mediated Ca 2ϩ /H ϩ exchange suggests that the Ca 2ϩ /H ϩ measurements presented here are not the product of the additive effect of the individual transporters. To date, we have been unable to measure the CAX3-mediated Ca 2ϩ /H ϩ exchanger in yeast, and CAX3 does not suppress yeast mutants defective in vacuolar Ca 2ϩ transport. In Arabidopsis, cax3-1 lines treated with CaCl 2 had Ca 2ϩ /H ϩ exchange activity similar to that of controls (11). Furthermore, 35S::HA-CAX3 lines did not exhibit high levels of Ca 2ϩ /H ϩ exchange. It should be noted that our in planta transport assays lack precision, given that other transporters are certainly active in these membrane preparation. Furthermore, our previous work has shown that mutants in CAX transporters cause altered activity of other transporters. Despite these limitations, together with our data regarding the physical interactions (see below), we interpret the changes in Ca 2ϩ /H ϩ transport in planta as being a consequence of the interaction between the transporters.
In our yeast assays, co-expression of both CAX1 and CAX3 (CAX1 ϩ CAX3) demonstrated more salt accumulation than either transporter expressed individually (Fig. 1, A and D). We have thus used whole yeast metal accumulation as a rapid inference of these CAX complex function(s). For example, where we measured vacuolar Ca 2ϩ uptake inhibited by Li ϩ , we were also able to measure increased accumulation of Li ϩ (Fig. 1A). Taken together, these observations further suggest that we are measuring phenotypes caused by an interaction of the transporters rather than an additive effect of the two individual transporters.
Additionally, we demonstrated that CAX1 ϩ CAX3 cells could suppress the NaCl sensitivity of nhx1 yeast strains (Fig.  1C), a phenotype that could not be recapitulated by expressing sCAX1, CAX1, or CAX3, as well as several other combinations of CAX transporters (data not shown).
Physiological Relevance of CAX Interactions-Here we have addressed the interplay among CAXs within plant cells using both phenotype analysis and transport studies. We demonstrated that cax1, cax3, and cax1/3 lines have similar germination defects when grown under high lithium conditions (Fig. 3). In addition, cax1, cax3, and cax1/3 lines display similar responses to sugar stress, ethylene, and abscisic acid during seed germination (17). These genetic data are consistent with a model where both transporters are required together for these responses. In agreement with these observations, transgenic lines that simultaneously express high levels of HA-CAX3 ϩ CAX1 had altered substrate specificity, as measured by cation selectivity comparisons (Fig. 4). The Ca 2ϩ /H ϩ transport measured from tonoplast-enriched vesicles from the HA-CAX3 ϩ CAX1-expressing lines demonstrated increased inhibition in the presence of excess nonradioactive Na ϩ and/or Li ϩ compared with lines expressing high levels of each transporter individually.
The intriguing question is how CAX1 and CAX3 interact and coordinately regulate transport. To understand the function of any specific transporter, much less the function of transporter complexes, is difficult. The Arabidopsis genome contains over 150 cation transporters, many with redundant functions (40). We speculate that the CAX1-CAX3 complex formation is regulated temporally by alterations in transporter abundance (Fig.  2) (11). In this study, both CAX1 and CAX3 displayed similar expression levels during seed germination. This may signify that CAX1-CAX3 complexes have an important role during germination. As a consequence, the cax1, cax3, and cax1/3 lines have similar germination phenotypes (Fig. 3) (17).
CAX1 and CAX3 also have discrete roles in plant growth and development. CAX1 and CAX3 can also form functional homooligomers. Previous studies with a mung bean CAX in yeast demonstrated homo-CAX formation (7). In planta, we speculate that in maturing roots, a homo-CAX3 complex is active. Only CAX3 is highly expressed in roots, and it appears to be responsible for 10-day-old plant responses to NaCl and low pH stress (17). Homo-CAX1 complexes may also be important for ion tolerance; however, 10-day-old cax1 seedlings display no difference in growth compared with wild-type grown on lithium-containing medium (12). Interestingly, ectopic expression of the soybean cation/H ϩ antiporter GmCAX1 in Arabidopsis does provide lithium tolerance (41), and 10-day-old cax3 lines display some alterations in lithium tolerance (17). These findings suggest that homo-Arabidopsis CAX transporters as well as the CAX1-CAX3 complex may have roles in particular stress tolerances.
CAX1 and CAX3 Interact-The CAX1-CAX3 interaction is demonstrated by the following findings. 1) In yeast assays, the co-expression of CAX1 and CAX3 conferred unique growth and transport properties not exhibited by the expression of either transporter individually ( Fig. 1) (11). 2) Split ubiquitin assays demonstrated physical interaction among different CAXs (Fig. 6). 3) In vivo interaction in plant cells was shown through coimmunoprecipitation of CAX1 with HA-CAX3 (Fig.  5). 4) Genetic analysis between the loss-of-function alleles of cax1 and cax3 is consistent with the gene products operating together during stress responses (Fig. 3A) (17). 5) Transport studies in vacuole-enriched membrane vesicles isolated from Arabidopsis lines expressing HA-CAX3 and CAX1 demonstrate transport properties distinct from membranes expressing high levels of either CAX1 or CAX3 (Fig. 4).
Here we have focused our plant CAX1-CAX3 analysis on lithium phenotypes. Our data suggest that the putative CAX1-CAX3 complex may be required for tolerance to particular stress conditions. Both CAX1 and CAX3 are highly expressed when seedlings are exposed to high levels of LiCl (Fig. 3B). Furthermore, the complex formation appears to be enhanced during LiCl treatment (Fig. 5D). These complexes may be subject to degradation, as indicated by our inability to resolve distinct bands during the isolation procedure after prolonged LiCl treatments ( Fig. 5D; data not shown). Both yeast and plant data support the idea that CAX1-CAX3 complexes can transport Li ϩ more efficiently than either transporter individually (Figs. 1  and 4). We postulate that the interaction among the CAXs alters substrate affinity though a conformational change. However, future work is required to directly measure Li ϩ transport kinetics.
The physical interaction between CAX1 and CAX3 in planta has been demonstrated using co-IP experiments using plant protein extracts derived from microsomes. We have utilized a number of tonoplast antibodies as negative controls (V-ATPase, AVP1, TPC1, and TIP) to document that this interaction is direct and specific. Although it remains a formal possibility that CAX1 and CAX3 could co-exist in the same membrane protein complex/domain, our co-IP data, in tandem with the membrane transport changes in CAX1 ϩ HA-CAX3-expressing plants, strongly favor a model where CAX1 directly interacts with CAX3. Future experiments in planta with split-GFP or a fluorescence resonance energy transfer system using isolated vacuoles can be used to look at the dynamics of this interaction in more detail.
CAX Regulation via Oligomerization-Oligomerization may represent a regulatory mechanism for many transporters. Both plant and animal sugar transporters form regulatory complexes (42,43). In lipid transport, the small CER5 ABC transporter may form a homo or heterodimer to facilitate fatty acid transport to the cuticle (44,45). Oligomerization of S. cerevisiae Nha1p is essential for its Na ϩ /H ϩ antiporter activity (46). Recently, a cytosolic trans-activation domain has been identified as being essential for ammonium uptake by AMT transporters (15,47). A molecular model of AtAMT1;2 provides a mechanism where the C terminus of one monomer directly contacts the neighboring subunit. These alterations in the C-terminal domain may provide conformation coupling between monomers to allow tight regulation of transport and sensing (15,48).
Like oligomerization of other transporters, CAX1 and CAX3 interactions could allow dynamic flexibility for membrane transport. In plant tonoplasts, CAX1, CAX3, and CAX1-CAX3 complexes may simultaneously exist, and during growth and adaptation the ratio of these complexes may be regulated. Perturbing both CAX1 and CAX3 causes significant morphological phenotypes (11); however, not all endomembrane Ca 2ϩ /H ϩ activity is abolished. These observations suggest that other CAX transporters may also be involved in this regulatory interplay.
Conclusions-We have demonstrated that hetero-CAX interactions can have distinct functions when expressed in yeast and plant cells. These complexes may provide plants with dynamic transport flexibility and help in adaptation to environmental stresses.