Heterodimerization of Arabidopsis calcium/proton exchangers contributes to regulation of guard cell dynamics and plant defense responses

The cation exchanger family members CAX1 and CAX3 form homomeric and heteromeric complexes that are likely to modulate calcium transport during signaling events.

Both CAX1 and CAX3 proteins have been ascribed a functional role based on in planta expression analysis, ectopic expression, and mutant analysis in plants, and by heterologous expression in yeast (Shigaki and Hirschi, 2000;Manohar et al., 2011b). Until now, CAX1 and CAX3 expression has been shown to overlap in reproductive tissues at the organ level, but to localize differentially within the vegetative organs. CAX3 expression has been shown to localize primarily to root tips, whereas CAX1 expression is predominantly localized to leaf tissues. CAX1 regulates elemental accumulation across specific leaf cell types and subcellular compartments (Hirschi et al., 1996;Catalá et al., 2003;Conn et al., 2011b), whereas root growth in cax3-1 plants is lower than that of wild-type or cax1-1 plants under saline conditions, a phenotype that has been attributed to the greater inhibitory effects of Na + (and Li + ) on CAX3 Ca 2+ transport compared with CAX1 (Cheng et al., 2003;Manohar et al., 2011b).
CAX expression in leaves appears to be variable and subject to unresolved regulatory mechanisms (Conn and Gilliham, 2010;Gilliham et al., 2011). For example, in cax1-1 plants, expression of CAX3 and CAX4 increase along with a vacuolar Ca 2+ -ATPase ACA4 (Cheng et al., 2003). It has been hypothesized that enhanced CAX3 expression complements for the loss of CAX1, as cax1-1 plants do not show the major physiological perturbations of cax1-1/cax3-1 plants; however, it has not been shown whether CAX3 directly replaces CAX1 in the mesophyll (Cheng et al., 2003;Conn et al., 2011b). Such a complex system of cross-talk among genes is proposed to account for the subtleties in phenotypes that are often associated with loss-of-function genetic studies for transport proteins (Connorton et al., 2012). In order to elucidate these compensatory changes, in the present study we designed experiments to analyze the potential interactions between CAX transporters.
Despite the existence of phenotypic differences between cax knockout plants, there are also notable similarities. For instance, germination in both cax1-1 and cax3-1 knockouts is abscisic acid (ABA) and sugar sensitive, and ethylene inhibits seedling growth of both mutants, which suggests some shared functionality of these proteins (Zhao et al., 2008). Previous functional assays of CAX1 and CAX3 proteins in yeast, although insightful, have limitations because they used engineered CAX variants that lack N-terminal autoregulatory domains (Pittman and Hirschi, 2016). The functional significance of CAX1 and CAX3 interactions suggested by genetic and yeast assays is yet to be fully understood (Zhao et al., 2009b). Such interactions may occur in the guard cell, as expression of both CAX1 and CAX3 has been detected in guard cell protoplasts (Leonhardt et al., 2004;Cho et al., 2012). Moreover, genetic analysis has implied that a putative CAX1-CAX3 complex may influence guard cell closure and apoplastic pH (Cho et al., 2012).
Here, we provide evidence that heteromeric CAX complexes have physiological roles in plants. A split luciferase assay demonstrated that CAX1 and CAX3 form homo-and heterodimers in plant tissue, and, using yeast-based functional assays, we have shown for the first time that full-length CAX1-CAX3 has distinct transport characteristics compared with homomeric truncated-deregulated CAX proteins. The functional role for a CAX1-CAX3 complex in the plant was probed with stomatal assays, and both CAX1 and CAX3 appear to be required for correct functioning of the stomata. Our results highlight that expression of CAX3 and, to a lesser extent, CAX1 is induced in the mesophyll during defense responses, and that both proteins are required in the guard cell for the control of gas exchange. The interactions between CAX1 and CAX3 identified here suggest the possibility of regulatory plasticity in tonoplast Ca 2+ transport during signaling events.

RNA extraction
Total RNA was extracted from shoot tissue or mesophyll protoplasts of 5-6-week-old Col-0 plants treated as indicated in the respective figure legends, using TRIzol reagent (Invitrogen) and the DNasetreated by Turbo DNA-free kit (Ambion). Reverse transcription was used to synthesize cDNA from 2 µg RNA from each sample using SuperScript ® III Reverse Transcriptase (Invitrogen) with Oligo(dT) 20 as previously described by Conn et al. (2011b).
Gene cloning and plasmid construction PCR was used to amplify DNA fragments from Arabidopsis cDNA to clone the CAX1 and CAX3 coding sequences without start or stop codons (primers listed in Supplementary Table S2). Then, the DNA fragments were cloned into the pCR8/GW/TOPO TA cloning vector and transformed into TOP10 chemically competent Escherichia coli (Invitrogen). The genes of interest in pCR8/ GW/TOPO vectors were recombined into serial pDuEx-Bait/Prey expression vectors for a split luciferase interaction assay (Nakagawa et al., 2007), and a subsequent Cre-Lox recombinase reaction was performed to produce dual gene expression vectors for simultaneous expression of NLuc-CAX1 and CAX3-CLuc (or CLucN-CAX3-CLuc) (Creator TM DNA Cloning Kit, Clontech). CAX1 and CAX3 promoters (2 kb region upstream of the gene start codon ATG) as described by Cheng et al. (2005) were amplified from A. thaliana (Col-0) genomic DNA. Primers incorporated EcoRI/HindIII restriction sites (Supplementary Table S2) with the amplicon subcloned into pNO::Luc vectors using T4 DNA Ligase (New England Biolabs). The CAX1 and CAX3 artificial miRNA (amiRNA) was designed to achieve CAX-specific transcript reduction. CAX1 and CAX3 amiRNA sequences were designed using Web MicroRNA Designer v2 (Schwab et al., 2006) and cloned into the pCR8/GW/ TOPO vector. Subsequently, CAX1 and CAX3 amiRNA was recombined into a 2× CaMV 35S overexpression vector (pTOOL2) and used for miRNA expression (Plett et al., 2010).

Semi-quantitative PCR
Semi-quantitative PCR (semi-qPCR) for cell-specific CAX1 and CAX3 expression analysis was performed on cDNA separately synthesized from RNA of 5-8-week-old Arabidopsis mesophyll and epidermal cells. RNA preparation was performed as described previously using single cell sampling (SiCSA) (Conn et al., 2011b). Transcripts amplified were Actin2 (At3g18780; normalization control for both epidermis and mesophyll), CAX1 (At2g38170), and CAX3 (At3g51860); primers used are listed in Supplementary Table  S2. Amplification for this analysis was performed using Phire Hot Start Taq DNA Polymerase (Finnzymes) with the following cycling conditions: first round: 98 °C for 1 min, then 25 cycles of 98 °C for 10 s, 50 °C for 10 s, and 72 °C for 30 s. Primers were then removed using Nucleospin Extract II (Macherey-Nagel), and 1 μl of the eluate was used as the template for the second round: 98 °C for 1 min, then 25 cycles of 98 °C for 10 s, 55 °C for 10 s, and 72 °C for 10 s.
Semi-qPCR for CAX1 and CAX3 expression analysis was performed on cDNA samples reverse-transcribed from mesophyll protoplast RNA isolated from 5-6 week-old Arabidopsis leaves, following the protoplast isolation method detailed below. Transcripts amplified were Actin2, CAX1, CAX3, and GC1 (At1g22690; a marker gene for guard cells) with primers listed in Supplementary Table S2. Amplification of Actin2, CAX1, CAX3, and GC1 transcripts was performed using Phire Hot Start DNA Polymerase (Finnzymes) with the following cycling conditions: 98 °C for 1 min; 35 cycles of 98 °C for 5 s, 56 °C for 5 s, and 72 °C for 15 s; and then 72 °C for 1 min.
In situ PCR Leaves of 5-6-week-old Arabidopsis plants grown in short-day conditions were infiltrated with either ultrapure H 2 O or 2 µM flg22 in ultrapure H 2 O and left for 12 h. Then, leaves were detached, cut into 5 mm strips, and fixed in ice-cold formalin-acetic-alcohol solution [63% (v/v) ethanol, 5% (v/v) acetic acid, 2% (v/v) formalin] and washed in 1× PBS before being embedded in 5% agarose. Embedded leaf tissue was cross-sectioned using a VT 1200 S Vibrating Microtome (Leica) into 70 μm sections and transferred into a PCR tube. Then, the in situ PCR protocol of Athman et al. (2014) was followed using gene-specific qPCR primers as listed in Supplementary  Table S2, with the following cycling conditions: 98 °C for 1 min; 35 cycles of 98 °C for 5 s, 56 °C for 5 s, and 72 °C for 15 s; and then 72 °C for 1 min.

Real-time quantitative PCR
Real-time quantitative PCR (RT-qPCR) was performed on 0.2 µl cDNA using an iCycler Thermal cycler equipped with an iQ multicolor optical assembly module (Bio-Rad) and using KAPA SYBR ® FAST qPCR Kits (KAPA Biosystem), according to the following program: 95°C for 3 min; 40 cycles of 95 °C for 20 s, 55 °C for 20 s, and 72 °C for 20 s; with melt curve analysis from 52 °C to 92 °C in 0.5 °C increments. Primers for RT-qPCR analysis are listed in Supplementary Table S2. RT-qPCR result analysis followed the method described by Schmittgen and Livak (2008) using 2 −ΔCт to calculate gene expression level normalized to Actin2 (At3g18780) as an internal control.

Protoplast isolation
Isolation and transformation of protoplasts was carried out according to a previously described method (Yoo et al., 2007). Briefly, mesophyll protoplasts were isolated from leaf strips of 5-6-weekold A. thaliana by a 3-hour digestion in an enzyme solution containing 1.5% cellulase R10 and 0.4% macerozyme R10 (Yakult Pharmaceutical). Protoplasts were transformed via the polyethylene glycol 1450-mediated introduction of plasmid DNA in buffer solution. Modifications to this method included the use of one cell incubation medium, W2 [4 mM 2-(N-morpholino)ethanesulfonic acid (MES), 0.4 M mannitol, 15 mM KCl, 10 mM CaCl 2 , and 5 mM MgCl 2 , adjusted to pH 5.7 with KOH], as a replacement for both WI and W5 solutions. Protoplasts were incubated at room temperature for 0 or 24 h before harvesting for RNA extraction as indicated in the legend of Fig. 1.

Split luciferase protein-protein interaction and native promoter luciferase report assay
Direct protein-protein interactions between CAX1 and CAX3 were probed using a split luciferase complementation assay (Fujikawa and Kato, 2007). Mesophyll protoplasts were transformed with an equal amount of expression pDuEx-Bait and/or pDuEx-Prey plasmids and incubated for at least 16 hours. Native promoter reporter assays were performed using an equal amount of pNO::Luc expression plasmid fused to either the 35S, CAX1, or CAX3 promoter transformed into mesophyll protoplasts; pNO::Luc containing the promoter-free LUC gene was used as a negative control. Luciferase activity in the protein-protein interaction and promoter activity assay was detected using the ViviRen Live Cell Substrate (Promega) in a Polarstar Optima plate-reading spectrophotometer with luminescence detection capabilities (BMG Labtech). Following overnight incubation of transfected protoplasts, the ViviRen substrate was dissolved in DMSO and added to 500 µl protoplasts (2 × 10 5 cells ml −1 ) to 60 µM, mixed briefly, and aliquots of 100 µl (equivalent to 4 × 10 4 cells) were dispensed into white 96-well plates, in triplicate. Luminescence was measured immediately. Peak luminescence was observed at 300 s after substrate addition (gain=4095), and data from this time point were used for further analysis.
Ca 2+ tolerance and uptake assay in Saccharomyces cerevisiae Saccharomyces cerevisiae strain K667 (vcx1::hisG cnb1::LEU2 pmc1::TRP1 ade2-1 can1-100 his3-11,15 leu2-3,112 trp1-1 ura3-1)  was transformed with sCAX1 or CAX1 and CAX3 using (SC-Ura) transformation (Sherman et al., 1986). The Ca 2+ growth tolerance assay of S. cerevisiae was performed as previously described by Manohar et al. (2011b). Briefly, the assay was carried out via growing yeast expressing genes of interest at 30 °C for 3 days on solid YPD medium and supplemented with the appropriate amount of CaCl 2 . A vacuolar-enriched membrane fraction was prepared from yeast, following the method described by Manohar et al. (2011b). Yeast cells were collected by centrifugation at 4000× g for 5 min until the density reached an OD 600 of ~1.5.
The collected cell pellet was washed in spheroplast buffer (100 mM potassium phosphate buffer, 1.2 M sorbitol, pH 7.0) and resuspended in the same buffer plus 10 mM dithiothreitol (DTT) and 1% dextrose. Membrane vesicles of yeast cells were isolated using 1.5 units of zymolyase and incubated at 30 °C for up to 2 h. Timedependent 45 Ca 2+ /H + transport into these endomembrane vesicles was measured as described previously by Pittman et al. (2005).

Western blotting analysis
Western blotting analysis was performed as previously described by Manohar et al. (2011b). A monoclonal antibody to human influenza hemagglutinin (HA) (Berkeley Antibody Co., Richmond, CA, USA) was used at a 1:1000 dilution.

Gas exchange and photosynthesis measurements
Gas exchange and photosynthesis rate of whole rosettes were measured in 5-to 8-week-old Arabidopsis plants with treatments as indicated in the corresponding figure legends, using a LI-6400 infrared gas exchange analyzer (LiCOR) equipped with an Arabidopsis whole-plant chamber. Individual plants were exposed to light intensity of ~350 μmol m −2 s −1 at least 30 minutes prior to the start of measurement. The rosette was allowed to acclimatize inside the Arabidopsis whole-plant chamber for at least 5 minutes before gas exchange data were recorded, with reference CO 2 concentration set at 500 μmol mol −1 , flow rate at 500 μmol s −1 , light intensity at 350 μmol m −2 s −1 , and relative humidity at 56%. Leaf area of the whole plant was calculated using MATLAB on the basis of the image of whole plants, as described by Conn et al. (2013).

Agrobacterium-mediated transformation
Agrobacterium-mediated Arabidopsis seedling transformation followed the Fast Agrobacterium-mediated Seedling Transformation (FAST) method described by Li et al. (2009). Briefly, Arabidopsis seedlings were grown on 0.25× Murashige and Skoog medium for 5-6 weeks before being transferred into fresh 0.25× Murashige and Skoog liquid medium in a Petri dish containing an additional 100 μM acetosyringone and 0.005% (v/v) Silwet L-77, and co-cultivation with Agrobacterium tumefaciens cells at OD 600 =0.5 for 2 days. Assays were then performed on these seedlings.
Data and statistical analysis Statistical tests are described in the figure legends. All graphing and statistical analysis were performed in GraphPad Prism v6 and 7.

CAX3 expression can be induced in mesophyll cells by cax1 knockout, wounding, or pathogen stress
To directly test whether it is possible for CAX3 to replace and compensate for the loss of CAX1 expression in cax1-1 plants, we examined the expression profile of CAX1 and CAX3 in mesophyll cells of A. thaliana ecotype Col-0 and cax1-1 using laser capture microdissection (LMD) qRT-PCR (Fig. 1A). We confirmed that CAX1 was expressed at high levels in the mesophyll cells of wild-type (Col-0) plants, whereas CAX3 was not detected. In contrast, CAX3 expression was significantly induced in cax1-1 mesophyll, when CAX1 was absent, as predicted by Conn et al. (2011b). This suggests that it is possible for CAX3 to be expressed in the mesophyll under some conditions. However, the lack of CAX3 expression detected in Col-0 mesophyll cells (Fig. 1A), contrasts with previous observations made from protoplasts, where both CAX1 and CAX3 have been detected (Leonhardt et al., 2004;Cho et al., 2012). We further explored this disparity.
Immediately following the isolation of mesophyll protoplasts, CAX3 expression was barely detected, but the expression was significantly increased 24 h after protoplast isolation. Interestingly, CAX1 was highly abundant at both stages, with reduced expression after 24 h (Fig. 1B). A guardcell-specific marker (GC1) (Yang et al., 2008) was close to the detection limits at both time points (Fig. 1B). These results indicate that there was no or minimal guard cell contamination in our mesophyll preparation, and that CAX3, but not CAX1, showed inducible expression within mesophyll cells during the protoplasting procedure.
To assess whether these changes were due to either the stability of the mRNA or an increase in CAX promoter activity, we generated native CAX promoter::luciferase reporter constructs for transient expression in mesophyll protoplasts. To make this construct, we cloned the CAX1 and CAX3 promoter fragments, identified in Cheng et al. (2005) as reporting native expression patterns. We then inserted a promoter upstream of a luciferase protein derived from Renilla reniformis (sea pansy), and transfected the constructs into Arabidopsis mesophyll protoplasts. At 24 h after transfection, equivalent to the time point used in Fig. 1B, we could detect the expression of luciferase in the protoplasts driven by either the CAX1 or the CAX3 promoter (Fig. 1C), or via three sets of a truncated promoter for each CAX (F1, F2, and F3) (see Supplementary  Fig. S1). Interestingly, the CAX3 promoter drove 3-to 15-fold stronger luciferase activity compared with the respective CAX1 promoter (Fig. 1C). This suggests that gene transcription of CAX3 increases during protoplasting.
We then investigated whether we could induce CAX3 expression in the mesophyll of Col-0 plants under any conditions. Using SiCSA and semi-qRT-PCR, under our standard conditions (i.e. growth in BNS), CAX1 transcript was detected in both adaxial epidermal and palisade mesophyll cells of Col-0 leaves, whereas CAX3 was detected only from cax1-1 plant mesophyll (Fig. 1D). CAX1 transcript was abundant in RNA extracted from whole leaves of Col-0 plants, consistent with its expression in epidermis and mesophyll, while the presence of CAX3 transcript was below the level of our assay's detection limits under our standard growth conditions (see Supplementary Fig. S2). In an attempt to overload the leaf with apoplastic Ca 2+ , to mimic the situation in cax1/cax3 plants (Conn et al., 2011b), we increased the concentration of Ca 2+ in the root growth solution from 2 to 11 mM Ca 2+ (HCS). Although both transcripts were induced by HCS ( Fig. 1D; Supplementary Fig. S2), we did not observe a change in the cell-type localization of CAX1 and CAX3 expression under high Ca 2+ conditions (Fig. 1D). Furthermore, we compared the leaf vacuolar and apoplastic Ca 2+ content of Col-0, cax1-1, and cax3-1, and no significant differences were observed between the genotypes (Supplementary Fig. S3).
To determine whether the changes in CAX transcript abundance translated into increases in CAX protein levels, we used immunochemistry to measure the abundance of translational pCAX(1 or 3):CAX::HA fusions. In Col-0, the relative amount of CAX1 and CAX3 protein reflected the differences in transcript abundance, with CAX1 ~24-fold as abundant as CAX3 (Fig. 2). In cax1-1 plants transformed with these constructs, CAX1 protein abundance was comparable to wild-type, whereas CAX3 protein abundance was increased ~17-fold compared with wild-type (Fig. 2). These results strongly indicate that the changes in transcript abundance for CAX1 and CAX3 in whole leaves and in protoplasts reflect changes in protein abundance. This finding encouraged us to conduct a broader analysis to determine whether there were other stimuli that alter CAX expression.
A survey of existing Arabidopsis expression data (eFP BAR; http://bbc.botany.utoronto.ca) identified several biotic and abiotic stresses that might regulate CAX1 and CAX3 expression. CAX3 abundance was increased in leaves/shoots by osmotic stress, salt stress, and infection with Pseudomonas syringae or Botrytis cinerea (Table 1). To further examine the effect of P. syringae on CAX expression, we infiltrated 5-week-old Col-0 leaves with 1 μM flg22 (as a proxy for Pseudomonas infection) or water (as a control) 12 h before performing qPCR. Consistent with the result in the eFP BAR database, CAX3 expression was increased in flg22-infiltrated leaves, whereas CAX1 expression was unchanged (Fig. 3A). In addition, using in situ PCR, CAX3 was detected in the mesophyll of flg22-infiltrated leaves but not in the leaves of waterinfiltrated controls (Fig. 3B). Protoplasting induces wound responses (Ecker and Davis, 1987) and flg22 is a pathogen mimetic; both induce CAX3 expression, which implies that CAX3 has a role in plant defense responses. Using the BAR Expression Angler, we identified 236 genes strongly co-regulated with CAX3 (r 2 >0.75) in Arabidopsis leaves in response to P. syringae, B. cinerea, and their corresponding elicitors, whereas there were only 16 genes co-regulated in response to abiotic stresses (Austin et al., 2016). Furthermore, microarray analysis of cax1/cax3 plants shows alteration in expression of many pathogen-related genes (see Supplementary Table  S3). These findings suggest that co-expression of CAX1 and CAX3 within the mesophyll occurs during defense responses and as such there is potential for a CAX1-CAX3 complex to have a physiological role.

CAX3 in guard cells affects stomatal responses
To determine whether the putative CAX1-CAX3 interaction may have functional roles in plants, we surveyed the cellular expression of CAX1 and CAX3 in leaves. CAX1 expression was detected in mesophyll cells, vascular bundles, adaxial epidermal cells, and abaxial epidermal peels (Fig. 4). The only tissue assayed in which CAX3 expression was detected was abaxial epidermal peels; these peels will contain viable stomatal guard cells, indicating that CAX1 and CAX3 are coexpressed in guard cells.
To investigate whether co-expression is required in planta for normal guard cell function, we examined the apertures of stomatal pores in Col-0, cax1-1, cax3-1, and cax1/cax3 plants. Apertures of cax1-1, cax3-1, and cax1/cax3 stomata from isolated epidermis were smaller than those of Col-0 (Fig. 5A). Moreover, none of the mutants had reduced apertures in the presence of supplemental extracellular [Ca 2+ ], unlike Col-0 (Fig. 5B). Coupled with the fact that EGTA treatment opens cax1/cax3 stomata (Conn et al., 2011b), this result indicates that stomata in these epidermal peels were already partially closed in a Ca 2+ -dependent manner in both the single and double cax mutants. We previously demonstrated that while cax1/cax3 exhibits reduced leaf gas exchange as a result of higher apoplastic calcium, which causes reduced stomatal aperture (Conn et al., 2011b), apoplastic Ca 2+ concentration was not different from wild type in cax1-1 or cax3-1 plants under steady-state conditions (see Supplementary Fig. 3). This is likely due to CAX1 protein, or CAX3 in the case of cax1-1 plants, effectively buffering apoplastic Ca 2+ in the mesophyll of the single-knockout plants.
We therefore explored whether the Ca 2+ -sensitive stomatal phenotype of cax1-1 and cax3-1 plants could be recreated in leaves of intact plants or whether it was an artifact of the epidermal peel system. In order to achieve this, we first optimized the growth of cax1/cax3 plants to ensure that the guard cell phenotype observed in intact cax1/cax3 plants was not a consequence of the dwarf stature or delayed development of these plants when they are grown in standard BNS (2 mM Ca 2+ ) growth solution (Conn et al., 2011b). Previously, growing cax1/cax3 plants in low Ca 2+ solution (LCS, 50 μM Ca 2+ ) mitigated growth inhibition and increased stomatal conductance compared with growth in BNS (Conn et al., 2011b) (see Supplementary Fig. 4). However, at this low level of supplied Ca 2+ , plants developed necrotic lesions after 7 days. To overcome Ca 2+ deficiency and optimize growth, we germinated and grew cax1/cax3 plants in an optimized solution with low but sufficient calcium to support growth comparable to that of Col-0 (SLCS, 300 μM Ca 2+ ) (Fig. 6). After 5 weeks we transferred plants from SLCS to high calcium solution (HCS, 11 mM Ca 2+ ) for 1 week. While Col-0 plants did not appear to be adversely affected by this treatment and continued to develop normally, the growth of the cax1/cax3 rosette was greatly inhibited (Fig. 6A). In addition, the photosynthesis and transpiration rate of cax1/cax3 plants (per mm rosette surface  (Li et al., 2009). The following native promoter/gene/reporter fusions were constructed: pCAX1::AtCAX1::YFP::2xHA (pSIM1; GenBank accession number: HM750245.1) and pCAX3::AtCAX3::YFP::2xHA (pSIM3; GenBank accession number: HM750246.1). pSIM1 and pSIM3 proteins were predicted to have molecular weights of 81.9 and 81.6 kDa, respectively (ExPASy Compute pI/Mw tool). Arabidopsis Col-0 seedlings were co-cultivated with Agrobacterium carrying (1) pSIM3, (2) pSIM1; the cax1-1 line was transformed with (3) pSIM3 and (4) pSIM1; and (5) d35S::AtCAX3 served as a negative control. (A) Western blot of total protein extracted from the shoots of eight transformed seedlings, transferred to nitrocellulose membrane, and probed with anti-HA, HRP-conjugated primary antisera (Cell Signalling Technology, CS2999), demonstrates that CAX3 compensates at the protein level for the loss of CAX1 in the A. thaliana cax1-1 T-DNA insertion line. (B) Coomassie-stained 10% SDS-PAGE gel as a loading control for the Western blot. (C) Quantification of band intensities across three biological replicates using QuantityOne software (Bio-Rad Laboratories), normalized to pSIM3-transfected Col-0 (lane 1). area) was lower than that of Col-0 (Fig. 6B, C). Plants grown in SLCS for 6 weeks were used to compare the effects of rootfed Ca 2+ on gas exchange as a proxy for stomatal aperture. Both photosynthesis and rosette conductance of the single cax knockout and wild-type lines were not significantly different in SLCS, but were decreased in the cax1/cax3 line treated with HCS for 18 hours ( Supplementary Fig. 5); this difference was sustained 7 days after treatment (data not shown). We found a decrease in mean rosette conductance in cax1-1 and cax3-1 plants, although this was less significant than in cax1/cax3. When the length of exposure to HCS was reduced from 18 to 2 hours, we found that the cax1-1, but not cax3-1, plants had significantly reduced photosynthetic and rosette conductance rates compared with Col-0 ( Supplementary Fig. 6). This suggests that the transporters that buffer increases in apoplastic [Ca 2+ ] around mesophyll cells in cax1-1 plants, which include CAX3, are less effective than in wild-type plants containing CAX1, but only in the short term.

CAX1 and CAX3 interact as homo-and heterodimers in planta and facilitate Ca 2+ transport when co-expressed in yeast
Despite previous reports identifying the capacity of CAX1 and CAX3 proteins to interact in yeast and in plants under the control of ubiquitous promoters (Zhao et al., 2009a(Zhao et al., , 2009b, the nature of their interactions remains largely unexplored. We sought to address this latter point by expressing fulllength CAX1 and CAX3 using a split luciferase reporter construct in mesophyll protoplasts (Fig. 7). We found that both CAX1 and CAX3 could homodimerize and heterodimerize, and these interactions were abolished by co-transfecting with an artificial miRNA designed against one of the genes (see Supplementary Fig. 7). Furthermore, by switching the split luciferase between the N-and C-terminus, we found that the interaction was mediated by the N-termini of both proteins (head-to-head fashion) ( Fig. 7; Supplementary Fig. 7).  In situ PCR of CAX3 expression in wild-type leaf cross sections. 5-6-week-old Arabidopsis leaves were infiltrated with either H 2 O or 1 µM flg22 for 12 h before fixation. CAX3 and 18S rRNA transcripts were amplified with primers as listed in Supplementary Table S1 before staining; 18S rRNA was used as a positive control to show the presence of cDNA in all cell types; a no reverse transcription (RT) control was included to show lack of genomic DNA contamination. Scale bars=100 µm.
To gain insights into the functional relevance of the interaction between CAX1 and CAX3, we utilized yeast expression assays. Previously, it has been demonstrated that co-expression of CAX1 and CAX3 can suppress yeast vacuolar Ca 2+ transport defects, whereas expression of either transporter individually fails to do so (Cheng et al., 2005;Manohar et al., 2011b). The negative regulatory domains within CAX1 and CAX3 prohibit the functional expression of either transporter when individually expressed in yeast cells (Cheng et al., 2005;Manohar et al., 2011b). To avoid potential artifacts arising from the plasmid-dependent CAX overexpression approaches used previously, we modified yeast to integrate both CAX transporters into the genome to ensure stable expression levels. Only strains harboring both constructs conferred Ca 2+ tolerance to yeast mutants defective in vacuolar Ca 2+ transport (Fig. 8). Immunoblot analysis demonstrated that both CAX1 and CAX3 proteins accumulated to comparable levels in yeast cells (see Supplementary  Fig. 8).
We measured transport properties by measuring 45 Ca 2+ uptake activity in membrane vesicles isolated from yeast cells expressing integrated CAX1-CAX3 and the plasmid-based deregulated sCAX1, an artificially truncated form of CAX1 with the autoinhibitory domain removed. In this system, the pH gradient across yeast vacuolar membrane vesicles was generated by activation of the vacuolar H + -ATPase. The vesicles of sCAX1-and CAX1-CAX3-expressing cells took up 45 Ca 2+ from the medium in a pH-and time-dependent manner for up to 12 min (Fig. 9A). The accumulated 45 Ca 2+ was released after the addition of the Ca 2+ ionophore A23187. The addition of gramicidin, a protonophore that dissipates the pH gradient, eliminated membrane vesicle Ca 2+ uptake activity. Membrane vesicles of yeast cells expressing an empty vector had negligible activity (data not shown). Interestingly, CAX1-CAX3-expressing yeast cells demonstrated transport activity that differed from that of the deregulated sCAX1-expressing cells (Fig. 9A). Moreover, Michaelis-Menten kinetic analysis of the data showed that CAX1-CAX3-expressing cells displayed a K m of 21.64 µM for Ca 2+ , while sCAX1-expressing cells demonstrated a K m of 13.10 µM (Fig. 9B).
To analyze and compare the substrate specificity of the putative CAX1-CAX3 transporters, competition experiments were performed. This approach allowed us to determine the effect of co-expressing CAX proteins in terms of cation selectivity in comparison to the deregulated sCAX1. Initially, we measured Ca 2+ uptake in sCAX1-and CAX1-CAX3-expressing cells. The pH-dependent 10 µM 45 Ca 2+ uptake into yeast microsomal vesicles isolated from strains Fig. 5. Stomatal aperture and calcium responsiveness of Col-0, cax1-1, cax3-1, and cax1/cax3 isolated epidermal strips. (A) Stomatal pore aperture, expressed as width/length ratio, in Col-0, cax1-1, cax3-1, and cax1/cax3, n=175, 173, 173, 176, respectively. (B) Change in stomatal pore apertures in adaxial epidermal peels of Col-0, cax1-1, cax3-1, and cax1/cax3 measured with an additional 1 mM extracellular Ca 2+ (n=173, 183, 146, 166), plotted as the change in aperture between each genotype with or without the addition of 1 mM Ca 2+ . Asterisks indicate significant differences within a genotype with Ca 2+ treatment: (A) two-way ANOVA with Tukey's post-hoc test, **P=0.0054, *P<0.05; (B) Sidak's multiple comparisons test, ****P<0.0001. expressing empty vector (data not shown), sCAX1, and CAX1-CAX3 (Fig. 9C) was measured at a single 10 min time point. Ca 2+ uptake determined in the absence of excess nonradioactive metal (control) was compared with Ca 2+ uptake determined in the presence of two concentrations (10× and 100× excess) of various non-radioactive metals (Zhao et al., 2008). Inhibition of Ca 2+ uptake by non-radioactive Ca 2+ was used as an internal control; as expected, excess Ca 2+ inhibited the typical rosette size of Col-0 and cax1/cax3 grown for 6 weeks in sufficient but low calcium solution (SLCS) or grown for 5 weeks in SLCS followed by 1 week in high calcium solution (HCS). Transpiration (B) and photosynthesis (C) rates were determined using a LI-6400 infrared gas exchange analyzer (LiCOR) equipped with an Arabidopsis whole-plant chamber, set up according to Conn et al. (2011b). Data are presented as mean±SEM, n=20. ****P<0.0001 (two-way ANOVA). Fig. 7. Split luciferase protein-protein interaction assay to determine CAX1 and CAX3 interactions in mesophyll protoplasts (Fujikawa and Kato, 2007). Full-length CAX1 and CAX3 lacking a stop codon were recombined into split luciferase vectors with either the N-terminal (Nluc) or C-terminal (Cluc) half of luciferase fused to the N-or C-terminus of CAX1 or CAX3. Full-length Renilla luciferase was used as a positive control.
Ca 2+ uptake in both sCAX1-and CAX1-CAX3-expressing yeast. Non-radioactive Ca 2+ , particularly the 10× concentration, did not completely inhibit Ca 2+ uptake, further highlighting the low Ca 2+ affinity of the transporters. Ca 2+ uptake by sCAX1-expressing cells was strongly inhibited by a 10× concentration of Cd 2+ , whereas CAX1-CAX3-mediated Ca 2+ transport was only moderately inhibited. Interestingly, microsomes from CAX1-CAX3-expressing cells, compared with sCAX1-expressing cells, displayed less Ca 2+ uptake inhibition by Li + and Na + . These data demonstrate that the CAX1-CAX3 complex has altered Ca 2+ affinity and transport capacity compared with the deregulated sCAX1. These observations imply that the Ca 2+ dynamics may be different in plant cells containing the CAX1-CAX3 complex compared with cells containing only CAX1 or CAX3.

Discussion
We found that both CAX1 and CAX3 are expressed in stomatal guard cells, unlike most other leaf tissues, under standard conditions. This corroborates previous studies that have detected CAX1 and CAX3 RNA in isolated guard cell protoplast preparations (Leonhardt et al., 2004;Yang et al., 2008). These studies also found CAX3 to be expressed in the mesophyll. However, when we extracted RNA from non-protoplasted leaf tissue, using LMD or SiCSA, we could not detect CAX3 transcript in mesophyll cells (Fig. 1A). Only when protoplast incubation was extended to 24 h could CAX3 expression be detected at moderate levels in the mesophyll (Fig. 1B). The cell wall damage that occurs during protoplasting has been suggested to mimic physiological perturbations that occur in the cell wall in response to pathogen infections (Ecker and Davis, 1987). Our finding that flg22, a bacterial elicitor peptide, stimulates CAX3 expression in the mesophyll cells supports the conclusion that CAX3 expression is induced by wounding and pathogens, but is otherwise normally absent from the mesophyll (Table 1; Fig. 3). As both CAX1 and CAX3 are present in mesophyll cells in response to bacterial elicitors, and a functional CAX1-CAX3 complex can form, it is likely that the transport properties of this complex could participate in the physiological response to pathogen attack.
It is interesting to note that cax1/cax3 plants, which have an altered capacity for Ca 2+ secretion into mesophyll cells, had an increased apoplastic Ca 2+ concentration, and an altered transcript profile enriched in pathogen-responsive genes (see Supplementary Table S3). For instance, PR1 and PR2, in addition to many cell-wall-related genes, were upregulated in the cax1/cax3 rosette; expression could be reduced to wild-type levels by transferring plants to a low Ca 2+ condition ( Fig. 6A; Supplementary Fig. S9) (Conn et al., 2011b). Altered Ca 2+ compartmentation into the mesophyll and apoplast may also mimic some plant responses to pathogen infection. For example, heterologous expression of Arabidopsis sCAX1 in tomato results in upregulated expression of two pathogenrelated proteins, PR P2 precursor and PR leaf protein 4-like, homologs of PR1 and PR4 from A. thaliana (~8-and ~23fold, respectively) (De Freitas et al., 2011). Similarly, four PR genes were induced in cax1/cax3 plants, including PR1 (17fold) and PR5 (11-fold) (Conn et al., 2011b). The increased membrane leakage and blossom end rot symptoms in tomato fruits were considered to be due to the impact of enhanced vacuolar Ca 2+ storage on Ca 2+ -signaling-related proteins and disturbed apoplastic [Ca 2+ ], as well as cell wall modification (De Freitas et al., 2011). The upregulation of PR genes in sCAX1-expressing tomato and Arabidopsis cax1/cax3 lines suggests that a modification in CAX-mediated Ca 2+ transport in cells may also occur during pathogen responses of plants (Hocking et al., 2016). Variation in the intracellular Ca 2+ concentration in a plant cell is known to be a critical step for early defense signaling pathways (Lecourieux et al., 2006).
The observation that CAX1 and CAX3 are expressed in guard cells (Fig. 4) suggests a role for the CAX1-CAX3 complex in stomatal function. A smaller mean stomatal aperture was found in epidermal strips of cax1-1, cax3-1, and cax1/ cax3 plants relative to Col-0 (Fig. 5A). These findings are consistent with a previous study Cho et al. (2012), which found significantly reduced steady-state stomatal apertures in epidermal strips in cax1-1, cax3-1, and cax1/cax3 plants  Ca 2+ uptake into vacuolar vesicles prepared from the yeast strain K667 expressing sCAX1 or CAX1/CAX3. Results are shown in the absence and presence of a protonophore (gramicidin; G) The Ca 2+ ionophore A23187 (5 µM) was added at 12 min and uptake was measured at 22 min. Data are representative of three independent experiments. (B) Michaelis-Menten kinetic analysis of the initial rate of metal/H + exchange. A preset steady-state pH gradient was generated in vacuolar-enriched vesicles from yeast cells expressing sCAX1 and CAX1-CAX3 by activation of the V-ATPase. Initial rates of H +dependent Ca 2+ uptake were calculated over a range of Ca 2+ concentrations from 0 to 50 µM. Data are representative of three independent experiments. (C) Inhibition of Ca 2+ uptake by sCAX1 or CAX1-CAX3 into yeast vacuolar-enriched vesicles in the presence of other metal ions. Uncoupler-sensitive (∆pH-dependent) uptake of 10 μM 45 Ca 2+ was measured in the absence (control with 100% activity, shown by the dotted line) or presence of 10× or 100× non-radioactive CaCl 2 , LiCl, NaCl, NiSO 4 , MnCl 2 , CdCl 2 , ZnCl 2 , FeCl 3 , or CoCl 2 after 10 min. Data are averages of at least three replications from two independent membrane preparations, and are presented as mean±SEM.
compared with wild-type. These mutants were also nonresponsive to increased Ca 2+ concentrations, indicating that each isoform is required in guard cells for a normal response to apoplastic Ca 2+ (Fig. 5B). We propose that the CAX1-CAX3 complex may function in modulating apoplastic Ca 2+ signaling and is required for maintenance of normal stomatal aperture in response to changes in external Ca 2+ concentration; this might occur through the sensing of external Ca 2+ signals through the CAS apoplastic sensor pathway or due to misregulation of cytosolic free Ca 2+ affecting intracellular signaling in the guard cell (Wang et al., 2014).
Interestingly, however, we found no statistical difference in gas exchange rate in the single cax1-1 or cax3-1 mutants under most conditions (see Supplementary Figs S5 and S6). Previously, we have demonstrated that mesophyllic CAX1 controls leaf apoplastic [Ca 2+ ] and that CAX3 could compensate for loss of CAX1 in cax1-1 lines (Cheng et al., 2005;Conn et al., 2011b). Here, we demonstrate that CAX3 expression was induced in the mesophyll cells of cax1-1 plants, substituting for CAX1 (Fig. 1). Under standard Ca 2+ conditions, ectopic expression of CAX3 in the mesophyll in cax1-1 plants prevents excessive Ca 2+ accumulation in the apoplast and allows the maintenance of growth rate and gas exchange in cax1-1 plants (Supplementary Figs S3-S5) (Cheng et al., 2003). However, after a 2 h pulse of high Ca 2+ to the roots of cax1-1 plants, the gas exchange rates were reduced compared with wild-type or cax3-1 plants ( Supplementary Fig. S6); this indicates that CAX3 cannot fully complement cax1-1. This reduction in gas exchange was not observed in cax1-1 plants when high Ca 2+ was supplied over 18 h or 7 days, suggesting that the plants can adapt to cope with this Ca 2+ load over a longer time period.
In this study, we demonstrate that integrated expression of both CAX1 and CAX3 can catalyze vacuolar Ca 2+ uptake and rescue the Ca 2+ -hypersensitive phenotype of yeast strains defective in Ca 2+ transport (Fig. 8A). Further analysis on yeast vesicles showed that the heteromeric CAX1-CAX3 complex has similar Ca 2+ transport properties to deregulated CAXs ( Fig. 9; Supplementary Fig. S6) (Cheng et al., 2005;Manohar et al., 2011b). These yeast assays support the hypothesis that CAX3 may act as an activator of the negatively regulated CAX1 in specific plant tissues (Conn et al., 2011b;Cho et al., 2012). The transport affinity of the CAX1-CAX3 complex in yeast assays was different from that of the deregulated transporters and this finding suggests that coupling between CAX transporters could be a mechanism for increasing the range of transporter functions.

Conclusions
Evidence from non-plant studies is beginning to provide confirmation that CAX proteins are able to modulate Ca 2+ signals (Guttery et al., 2013;Melchionda et al., 2016). These studies have the advantage that the CAX proteins are encoded by single genes, and therefore genetic dissection of CAX Ca 2+ signaling is not hampered by genetic redundancy. Our study highlights that the multigene CAX families found throughout the plant kingdom may allow the formation of complex functional heteromeric complexes. In yeast-based assays, CAX1-CAX3 displayed transport properties that could not be recreated by highlevel expression of either native transporter individually. We also investigated the significance of these interactions for a variety of plant physiological responses. We found that the CAX1-CAX3 complex can occur in leaf mesophyll in response to pathogen attack. Additionally, CAX3 and the CAX1-CAX3 complex may be important in guard cells for maintenance of normal calcium responses and signaling pathways. Further work is required to determine the full extent of the signaling pathways in which CAX1-CAX3 may play a role.