AtPHT4;4 is a chloroplast-localized ascorbate transporter in Arabidopsis

Ascorbate is an antioxidant and coenzyme for various metabolic reactions in vivo. In plant chloroplasts, high ascorbate levels are required to overcome photoinhibition caused by strong light. However, ascorbate is synthesized in the mitochondria and the molecular mechanisms underlying ascorbate transport into chloroplasts are unknown. Here we show that AtPHT4;4, a member of the phosphate transporter 4 family of Arabidopsis thaliana, functions as an ascorbate transporter. In vitro analysis shows that proteoliposomes containing the purified AtPHT4;4 protein exhibit membrane potential- and Cl−-dependent ascorbate uptake. The AtPHT4;4 protein is abundantly expressed in the chloroplast envelope membrane. Knockout of AtPHT4;4 results in decreased levels of the reduced form of ascorbate in the leaves and the heat dissipation process of excessive energy during photosynthesis is compromised. Taken together, these observations indicate that the AtPHT4;4 protein is an ascorbate transporter at the chloroplast envelope membrane, which may be required for tolerance to strong light stress.

A scorbate (vitamin C) is an antioxidant and coenzyme for a number of metabolic reactions in living organisms 1,2 . Primates, including humans, have a defect in the enzyme responsible for ascorbate synthesis, L-gulono-1,4-lactone oxidase, and must therefore acquire ascorbate via the diet to maintain homeostasis. In plants, however, ascorbate is synthesized in the mitochondria in response to external stresses, distributed throughout the cells, and confers stress tolerance [2][3][4] . In particular, chloroplasts contain high concentrations of ascorbate (10-50 mM) 4,5 . When light strikes photochemical II (PSII) in the thylakoid membrane, water is disassembled into oxygen, electrons and protons. The protons then flow to photochemical I through the quinone molecule and cytochrome b6f, resulting in the synthesis of NADPH and ATP for carbohydrate synthesis from carbon dioxide. Excessive light energy and active oxygen species may damage the chloroplasts under conditions of light stress, leading to inhibition of growth (photoinhibition) 3,6-8 . Chloroplasts use ascorbate in these metabolic processes to eliminate active oxygen produced by electron transmission of PSII for the synthesis of NADPH in the stroma and as a coenzyme of violaxanthin de-epoxidase (VDE), which is involved in the release of photoenergy by heat dissipation in the xanthophyll cycle 4,6-8 . However, the mechanism by which ascorbate, which is synthesized in the mitochondria, passes through the envelope and thylakoid membranes of the chloroplast is poorly understood 9 . Although biochemical analyses indicated that the envelope membrane possesses a transporter that interacts preferentially with the reduced rather than the oxidized form of ascorbate (dehydroascorbate) as a transport substrate 9,10 , it is yet to be identified.
The SLC17 transporter family of Arabidopsis was originally reported as the Na þ or H þ /phosphate co-transporter (PHT4) family consisting of six genes 11 . Although the PHT4 family is widely distributed in plants, including rice, poplar, Physcomitrella patens subsp. californica, and so on, as well as Arabidopsis, the physiological relevance of this family is unknown. Gene expression-profiling studies indicated that AtPHT4;1, AtPHT4;4 and AtPHT4;5 genes are strongly expressed in the leaves, AtPHT4;3 and AtPHT4;6 genes are expressed in both roots and leaves, and the AtPHT4;2 gene is abundantly expressed in the roots 11,12 . Among these genes, only AtPHT4;1 and AtPHT4;4 showed B10-fold increases in expression on light exposure 12 . On the other hand, as the levels of expression of all AtPHT4s changed little even under conditions of phosphorus deficiency, they were assumed to have functions in addition to their roles as phosphate transporters 11 .
A series of studies performed in our laboratory as well as those reported by other groups indicated that the mammalian SLC17 transporter family consists of nine members, which were shown to be membrane potential (Dc)-and Cl À -dependent organic anion transporters: SLC17A1-2 act as urate exporters at the apical membranes of renal proximal tubules, SLC17A4 acts as a urate exporter at the apical membranes of intestinal ducts, SLC17A5 acts as a vesicular excitatory amino-acid transporter in synaptic vesicles, SLC17A6-8 act as vesicular glutamate transporters in synaptic vesicles, and SLC17A9 acts as a vesicular nucleotide transporter in synaptic vesicles and secretory granules [13][14][15] . The substrate specificity of each transporter is achieved by slight differences in amino-acid residues around the active centre: SLC17A1-2 and 4 transport urate, SLC17A5 transports aspartate and glutamate, SLC17A6-8 transport glutamate and SLC17A9 transports nucleotides [13][14][15] . On the basis of the above findings, we hypothesized that members of the AtPHT4 family also function as Dc-dependent organic anion transporters, and that at least one of these proteins transports ascorbate anions.
The results of the present study indicate that AtPHT4;4 encodes an ascorbate transporter expressed at the envelope membranes of chloroplasts. In addition, both the levels of the reduced form of ascorbate in the leaves and the process of heat dissipation of excessive energy during photosynthesis are decreased in Arabidopsis thaliana pht4;4 (atpht4;4) gene knockout mutants.

Results
Identification of an ascorbate transporter. The PHT4 family can be classified into four groups according to amino-acid sequence homology (Fig. 1a). To identify the ascorbate transporter from the PHT4 family, we selected one gene from each subgroup of the Arabidopsis PHT4 family (subgroup 1: AtPHT4;3, subgroup 2: AtPHT4;5, subgroup 3: AtPHT4;6, and subgroup 4: AtPHT4;4), and their cDNAs were cloned into Escherichia coli expression vectors with a His-tag and soluble a-helix protein (b) coupled to both ends 16 . Each transporter was overexpressed in E. coli, solubilized and purified using Ni-NTA affinity column chromatography. The purified proteins were then electrophoresed and stained with Coomassie Brilliant Blue ( Fig. 1b left), and their immunological properties were confirmed by immunoblotting with anti-6 Â His antibodies (Fig. 1b right). The final fractions contained the major protein bands of the expected apparent molecular masses and immunological properties (Fig. 1b). These purified proteins were incorporated into proteoliposomes. By analogy to mammalian SLC17 family transporters, we investigated whether the transporters possess Na þ -dependent transport activity for inorganic phosphate (P i ). The Na þ /P i transport activity was detected in proteoliposomes containing all of these transporters, supporting the suggestion that all of the purified recombinant transporters were active in nature (Fig. 1c). Using the same batch of proteoliposomes, we employed Dc (positive-inside) by addition of valinomycin in the presence of K þ . The proteoliposomes established an inside-positive Dc of B90 mV through K þ diffusion, as reported previously 17 . Under these conditions, only proteoliposomes containing purified AtPHT4;4 exhibited significant ascorbate uptake activity, while those containing AtPHT4;3 or AtPHT4;6 did not (Fig. 1d). Proteoliposomes containing purified AtPHT4;5 exhibited slightly Dc-dependent ascorbate uptake activity.
Characterization of AtPHT4;4-mediated ascorbate uptake. We further characterized the ascorbate uptake by proteoliposomes containing purified AtPHT4;4. Valinomycin-induced Dc was maximal at 1-2 min after addition of valinomycin ( Supplementary  Fig. 1). In parallel with the degree of Dc formed, proteoliposomes showed maximal ascorbate uptake at 2 min, which decreased gradually thereafter (Fig. 2a). Liposomes lacking AtPHT4;4 showed only background uptake level. The Dc-dependent ascorbate uptake exhibited Michaelis-Menten-type kinetics with Km and Vmax of 1.2 mM and 520 nmol min À 1 mg À 1 , respectively (Fig. 2b). Bioenergetics analysis under conditions of defined Dc, DpH and/or DpNa þ indicated that Dc primarily triggered ascorbate uptake, while DpH and DpNa þ did not (Fig. 2c). Imposing DpH (outside-acidic) had a slight effect. Ascorbate uptake showed an absolute requirement for Cl À similar to mammalian SLC17 family transporters [13][14][15] , and no ascorbate uptake was detected in the absence of Cl À . Ascorbate uptake showed marked activation with 2-4 mM Cl À and reached a plateau at 10 mM Cl À (Fig. 2d). Both Evans blue and 4,4 0diisothiocyano-2,2 0 -stilbenedisulphonic acid, which are typical inhibitors of mammalian SLC17 family transporters, inhibited ascorbate uptake (Fig. 2e) [13][14][15] . Experiments were performed to examine the effects of a spectrum of possible cis-inhibitors, and the results indicated that Dc-dependent L-ascorbate uptake was insensitive to dehydroascorbate(oxidized L-ascorbate), Disoascorbate (a stereoisomer of L-ascorbate), P i , glutamate, ATP, p-aminohippuric acid (PAH, a typical substrate of mammalian organic anion transporter) and tetraethylammonium (a typical substrate of mammalian organic cation transporters; Fig. 2f).
Expression and localization of AtPHT4;4 in leaves. Quantitative PCR was performed to examine the level of AtPHT4;4 gene expression. Consistent with previous observations 12 , the AtPHT4;4 gene was expressed at higher levels in the leaves than the roots ( Supplementary Fig. 2a), and its level of expression increased on light exposure ( Supplementary Fig. 2b).
We prepared a specific polyclonal antibody against AtPHT4;4 to examine its expression and localization. In a parallel experiment shown in Fig. 1b, the polyclonal antibody detected the AtPHT4;4 protein but not AtPHT4;3, AtPHT4;5 or AtPHT4;6 ( Fig. 3a left), while pre-absorbed anti-AtPHT4;4 antibody did not bind to the AtPHT4;4 protein ( Fig. 3a right) indicating the immunological specificity of the antibody for AtPHT4;4. On indirect immunofluorescence microscopy with the antibody, AtPHT4;4 immunoreactivity was detected in chloroplasts of the palisade tissue rather than spongy tissue from the leaves of Arabidopsis (Fig. 3b). Examination at higher magnification indicated that the AtPHT4;4 signal was present outside chlorophyll ( Fig. 3c upper). The pattern of AtPHT4;4 localization was very similar to that of TIC40, which is an envelope membrane marker, but not light-harvesting chlorophyll protein 2 (LHC2), which is a thylakoid membrane marker ( Fig. 3c middle and lower, respectively).
AtPHT4;4 gene knockout decreases reduced ascorbate in leaves. Two lines (ET4970; pht4;4-1 and GT5039; pht4;4-2) of Ds transposon-tagged atpht4;4 mutants were obtained from the Cold Spring Harbor Laboratory. Disruption of the AtPHT4;4 gene in both mutant lines was confirmed using RT-PCR (Fig. 4a). The loss of the AtPHT4;4 protein in chloroplasts of both mutant lines was confirmed by immunoblotting and immunofluorescence microscopy (Fig. 4b,c). On microscopic observation, no morphological differences were detected between wild-type control and mutant lines (Fig. 4c). The atpht4;4 mutant lines had a normal appearance compared with wild-type controls (that is, maximum rosette radius: 22.9±1.4, 20.8±1.6 and 24.0±0.7, 24.1 ± 1.5 mm for control-1, pht4;4-1 and control-2, pht4;4-2, respectively; Fig. 4d). Although the mutants were exposed to high light (300 mmol photons m À 2 s À 1 ) following initial growth under conditions of low light (100 mmol photons m À 2 s À 1 ), there were no significant changes in appearance compared with wild-type controls. Moreover, we measured the levels of ascorbate in the leaves of wild-type controls and atpht4:4 mutant lines before and after exposure to high light stress. The levels of the reduced form of ascorbate in wild-type control leaves were increased by high light, whereas those of the reduced form of ascorbate in the atpht4:4 mutant leaves under conditions of high light were decreased by B35% compared with wild-type controls (Fig. 4e). On the other hand, no significant changes were observed in oxidized ascorbate levels between wild-type control and mutant leaves. Total ascorbate level was slightly decreased in the mutant leaves (Fig. 4e). Total ascorbate in the fraction containing chloroplasts from mutants was reduced to B70% of that in wild-type controls ( Supplementary Fig. 3    with 540 mmol photons m À 2 s À 1 , both mutant lines showed decreases in nonphotochemical quenching (NPQ and qN), corresponding to the dissipation of excess absorbed light energy as heat, but not Fv/Fm (the maximum quantum efficiency of PSII), Fv 0 /Fm 0 (the efficiency of open reaction centre in light), (Fm 0 -Ft)/ FmF (the quantum yield of electron transfer at PSII) or photochemical quenching (the redox state of the primary quinone acceptor of PSII; Fig. 5a). The NPQ induction curves were compared between the two atpht4;4 mutant lines (Fig. 5b,c).
When illuminated at 540 mmol photons m À 2 s À 1 (Fig. 5b) or 230 mmol photons m À 2 s À 1 (Fig. 5c), the wild-type controls showed rapid establishment of NPQ within 2 min. In the atpht4;4 mutants, however, NPQ was always B20% lower than that in wild-type controls, although the level was similar to that in the wild-type controls in the dark period. NPQ is a process by which xanthophylls, accessory pigments of LHC2, convert violaxanthin at a higher light-condensation rate to antheraxanthin and then zeaxanthin at a lower rate in order by VDE releasing excessive light energy by heat dissipation [6][7][8] . As ascorbate functions as a coenzyme of VDE, the xanthophylls and other pigments of wild-type and atpht4;4 mutants were assayed by high-performance liquid chromatography (HPLC; Fig. 6a).
The levels of zeaxanthin and antheraxanthin in the mutants were decreased by high-light treatment for 2 min, whereas the level of violaxanthin increased slightly (Fig. 6b, Supplementary Fig. 4). This is the first study to detect changes of xanthophyll cycle activities in vivo with short-term illumination for 2 min. There were no changes in the levels of lutein (constitutional isomer of zeaxanthin) or neoxanthin (one of the products of violaxanthin), neither of which is involved in NPQ (Fig. 6b, Supplementary   Fig. 4). The above findings indicated that the levels of products of VDE using ascorbate as a coenzyme were decreased in the atpht4;4 mutants. With regard to pigments other than xanthophylls, the levels of b-carotene were decreased by B30% with high-light exposure in comparison with wild-type controls (Fig. 6b, Supplementary Fig. 4).  and may be required for photoinhibition tolerance. This is the first report regarding the identification of an ascorbate transporter in plants.

Discussion
AtPHT4;4 exhibited Na þ /P i co-transport activity, similar to the mammalian SLC17 family transporters. As chloroplasts do not possess a Na þ gradient as a driving force, this activity is considered to be an evolutionarily conserved function of the ancestor protein. In addition, AtPHT4;4 also shows Cl Àdependent ascorbate-transport activity making use of Dc as the driving force (Fig. 2) [13][14][15] . Although plants have evolved differently and have different transport substrates from mammals for the SLC17 transporter family, the fundamental transport mechanisms, such as Cl À and Dc dependence, have been preserved (Supplementary Fig. 5). The results of aminoacid sequence comparison indicated conservation of Arg230 in the fourth transmembrane domain of AtPHT4;4 in almost all SLC17 family transporters in mammals and Arabidopsis ( Supplementary Figs 6 and 7). Our previous biochemical analyses of mammalian SLC17 family transporters by sitedirected mutagenesis and homology modelling with bacterial glycerol 3-phosphate transporter indicated that this arginine residue is essential for Dc-dependent organic anion transport activity and that the second and fourth transmembrane domains are important for binding to the transport substrate ( Supplementary Fig. 7) [13][14][15] . The small differences in aminoacid residue around this domain are thought to determine the transport substrate specificity within the SLC17 transporter family. In addition, the amino-acid residues around this domain in AtPHT4;4 are also thought to be important in transport of the reduced form of L-ascorbate.  NATURE COMMUNICATIONS | DOI: 10.1038/ncomms6928 ARTICLE AtPHT4;4 is expressed more abundantly in the envelope membranes of chloroplasts of the palisade tissue, which is exposed to stronger photostress than the spongy tissue (Fig. 3b,c), suggesting that AtPHT4;4 may play a role in the transport of ascorbate into the chloroplast. The quantity of ascorbate transported to the stroma of the chloroplast is assumed to be controlled by Dc as the driving force of AtPHT4;4, that is, the concentration gradient of ascorbate is dependent on the Dc gradient. Although the chloroplast contains abundant ascorbate, its concentration gradient in the chloroplast and cytosol does not vary greatly because Dc at the envelope of the chloroplast is small under physiological conditions 18 . The substrate specificity of ascorbate transport at the envelope membrane of the chloroplast was almost identical to that of AtPHT4;4 (refs 9,10). Moreover, the level of the reduced form of ascorbate in the leaves in Arabidopsis mutants with a defect in the AtPHT4;4 gene was B35% lower than that in wild-type controls, almost corresponding to the content of ascorbate in the chloroplast (25-30% of the whole, Fig. 4e) 9 . The ascorbate content was also decreased in the fraction containing the chloroplasts of mutant leaves ( Supplementary Fig. 3). On the basis of these findings, we concluded that AtPHT4;4 transports ascorbate from the cytosol into the chloroplast.
It should be noted that AtPHT4;1 shows a high degree of identity with AtPHT4;4 (identity of 70% at the amino-acid sequence level). Previous studies indicated that both AtPHT4;1 and AtPHT4;4 gene expression are induced by light 12 . AtPHT4;4 is present in the envelope membrane of the chloroplast, as shown in Fig. 3c, while AtPHT4;1 is localized at the thylakoid membrane of the chloroplast 19,20 . Thus, AtPHT4;1 is likely to also function as an ascorbate transporter at the thylakoid membrane. AtPHT4;4 transports ascorbate into the stroma of the chloroplast, and AtPHT4;1 transports ascorbate from the stroma to the thylakoids to control the dynamic state of ascorbate in the chloroplasts (Fig. 7). Further studies of the transport function and physiological role of PHT4;1 are currently in progress in our laboratory.
AtPHT4;4 gene defect results in a decrease of ascorbate content in the chloroplasts. Although the AtPHT4;4 gene defect was not associated with any visible phenotype, we found marked variations in the xanthophyll cycle during photosynthesis in the mutants. As VDE requires ascorbate as a coenzyme, this variation may have been due to the AtPHT4;4 gene defect 6,7 . To date, two types of mutant lines with decreased NPQ in Arabidopsis have been reported, that is, ascorbate synthetic enzyme mutant lines (vtc) and VDE mutant lines (npq1) 4,21 . With regard to vtc mutant lines, four genes (vtc1-4) have been reported as variants that reduce the ascorbate levels in young leaves to B30-50% of those in wild-type controls. Although NPQ in measurement of chlorophyll fluorescence was decreased in vtc mutant lines, there were no changes in other parameters, as observed in atpht4;4 mutant lines. Moreover, it was also reported that npq1 mutant lines showed decreased NPQ and zeaxanthin levels in the xanthophyll cycle as observed in atpht4;4 mutant lines. Interestingly, there were no marked changes in growth in vtc or npq1 mutant lines, and their b-carotene levels were decreased under conditions of high-light exposure, as observed in atpht4;4 Z e a x a n t h i n A n t h e r a x a n t h i n V i o l a x a n t h i n L u t e i n N e o x a n t h i n β -C a r o t e n e ** * *  Figure 7 | Schematic model of ascorbate transport in chloroplasts. Upon photostress, PHT4;4 gene expression is enhanced, and the PHT4;4 protein at the envelope membranes takes up ascorbate from mitochondria, which is transferred into the thylakoid through an as yet unknown transporter. PHT4;1 is a candidate ascorbate transporter at the thylakoid membrane. mutant lines 22,23 . The similarities in variations among the mutants of ascorbate synthetic enzyme, ascorbate transporter and VDE strongly support our conclusion.
Here we postulate a mechanism for the reduction in the level of b-carotene in these mutant lines from the viewpoint of antioxidant action as follows. Reactive oxygen species are produced during photosynthesis within the chloroplasts of plants and inhibit photosynthesis. Ascorbate has an antioxidant action for detoxification; active oxygen is converted to H 2 O 2 in the stroma by superoxide dismutase and the H 2 O 2 is then converted to H 2 O by ascorbate peroxidase 3,4 . On the other hand, b-carotene is an accessory pigment of LHC2, and it has been reported that b-carotene plays an important role in photosynthesis as an antioxidant 24,25 . b-Carotene nonenzymatically converts active oxygen species produced by photosynthesis to oxygen inside the thylakoid membrane for detoxification. Thus, in atpht4;4 mutant lines, b-carotene is inferred to function as an antioxidant in place of ascorbate to protect against photoinhibition. In addition, it has been suggested that C3 plants, which only have the Calvin-Benson cycle in the carbon fixation reaction, consume oxygen and release carbon dioxide through photorespiration to prevent oxidation by high concentrations of oxygen under conditions of high-light exposure 26 . In atpht4;4 mutant lines, as the levels of accessory pigments in the xanthophyll cycle are decreased, the level of light energy absorbed by plants under conditions of highlight exposure is assumed to be decreased as is the level of oxygen. Thus, excessive oxygen is not produced easily, resulting in a mitigation of oxidative stress. We speculate that these actions serve as an alternative mechanism of antioxidant action when the level of the reduced form of ascorbate inside the chloroplasts is decreased, which is a factor not observed in variations other than NPQ in atpht4;4 mutant lines.
In summary, we identified the ascorbate transporter in the chloroplast envelope membrane in Arabidopsis, and demonstrated that it plays a role in the xanthophyll cycle during photosynthesis. Heritable transporter genetic modification technology may provide a means of developing photoinhibitiontolerant plants. Expression and purification of AtPHT4. The expression and purification of AtPHT4 were carried out as described previously 16 . E. coli C43 (DE3) cells were transformed with expression vectors and grown in TB medium containing 20 mg ml À 1 kanamycin sulphate at 37°C. E. coli cells were grown until A 600 reached 0.6-0.8, and then isopropyl-b-D-thiogalactopyranoside was added to a final concentration of 1 mM and culture was continued for 16 h at 18°C. The cells were then harvested by centrifugation and suspended in a buffer containing 20 mM Tris-HCl (pH 8.0), 300 mM sucrose and 1 mM phenylmethylsulphonyl fluoride. The cell suspension was then disrupted by sonication with a TOMY UD200 tip sonifier (OUTPUT4), and centrifuged at 5,856 Â g at 4°C for 10 min to remove large inclusion bodies and cell debris. The resultant supernatant was carefully obtained and centrifuged again at 150,000 Â g for 1 h at 4°C. The pellet was suspended in buffer containing 70 mM Tris-HCl (pH 8.0), 100 mM NaCl, 10 mM KCl, 15% glycerol and 1 mM phenylmethylsulphonyl fluoride, and the protein concentration was adjusted to 10 mg ml À 1 . Then, the membranes were treated with 1.5% Fos-choline 14 (Affymetrix) and centrifuged at 150,000 Â g at 4°C for 1 h. The supernatant containing AtPHT4 was obtained, diluted twofold with buffer containing 70 mM Tris-HCl (pH 8.0), 100 mM NaCl, 10 mM KCl, 15% glycerol and 1 mM phenylmethylsulphonyl fluoride, and then applied to a column containing 1 ml of nickel-NTA Superflow resin (Qiagen) equilibrated with buffer containing 70 mM Tris-HCl (pH 8.0), 100 mM NaCl, 10 mM KCl and 15% glycerol. After incubation for 1 h at 4°C, the column was washed with washing buffer containing 70 mM Tris-HCl (pH 8.0), 5 mM imidazole, 100 mM NaCl, 10 mM KCl, 20% glycerol and 0.1% n-decyl-b-D-thiomaltopyranoside (Affymetrix). The AtPHT4 protein was eluted with buffer containing 20 mM Tris-HCl (pH 8.0), 300 mM imidazole, 100 mM NaCl, 10 mM KCl, 20% glycerol and 0.1% n-decyl-b-D-thiomaltopyranoside, and then stored at À 80°C, at which it was stable without loss of activity for at least several months.
Reconstitution. Aliquots of 20 mg of purified AtPHT4 were mixed with 500 mg of liposomes and frozen at À 80°C for at least 10 min. The mixture was diluted 60fold with reconstitution buffer containing 20 mM MOPS-Tris (pH 7.0), 0.15 M sodium acetate and 5 mM magnesium acetate. The buffer composition was changed as necessary. Reconstituted proteoliposomes were pelleted by centrifugation at 200,000 Â g for 1 h at 4°C, and then suspended in 0.2 ml of reconstitution buffer. Asolectin liposomes were prepared as described previously 17 . Soybean lecithin (10 mg ml À l ; Sigma Type IIS) was suspended in the buffer containing 20 mM MOPS-Tris (pH 7.0) and 1 mM dithiothreitol. The mixture was sonicated until clear in a bath-type sonicator, and stored at À 80°C until use.
Quantitative PCR. Total RNA was prepared from the leaves and roots of 4-to 5-week-old Columbia wild-type plants using an RNeasy Plant Mini Kit (Qiagen). cDNA was generated from total RNA with a PrimeScript RT reagent Kit (Takara) using 1 mg of total RNA as the template. Quantitative PCR was carried out with specific forward and reverse primers at 0.4 mM and 5 units ml À 1 of SYBR Premix Ex Taq II (Takara). Reactions were performed for 35 cycles of denaturation at 95°C for 15 s and annealing/extension at 60°C for 30 s. The primer set used for detection of AtPHT4;4 was as follows: 5 0 -TCGGGTCTCTACTCTAATCATCAAG-3 0 and 5 0 -AACACATCATCCCATGAACCTCG-3 0 . The level of AtPHT4;4 expression was evaluated relative to that of the housekeeping gene (AtActin2) by relative standard curve method using the StepOne Software v2.2.2 (Life Technologies). The primer set used for detection of AtActin2 was as follows: 5 0 -CCATCCAAGCTGTTCTCT CCTTG-3 0 and 5 0 -GGTAATCAGTAAGGTCACGTCCAG-3 0 .
Immunohistochemistry. Immunostaining of AtPHT4;4 in leaves of 4-to 5-weekold plants was performed as described previously 27 . Leaves were fixed in 4% (w/v) paraformaldehyde and 60 mM sucrose buffered with 50 mM cacodylic acid (pH 7.4) for 2 h at room temperature with occasional degassing. After three washes with 60 mM sucrose and 50 mM cacodylic acid (pH 7.4), the fixed samples were embedded in 5% agar and cut into sections 80 mm thick with a microslicer (ZERO 1; Dosaka EM). Sections were placed on microscope slides, incubated with PBS (10 mM, pH 7.4, 138 mM NaCl, 2.7 mM KCl) containing 0.1% (w/v) pectolyase Y-23 (Seishin) at 30°C for 2 h and then reincubated in PBS containing 0.3% (v/v) Triton X-100 at 30°C for 2 h, washed three times with PBS and blocked with 5% (w/v) bovine serum albumin in PBS. Slides were incubated in a chamber at 37°C with purified rabbit anti-AtPHT4;4 polyclonal antibody (1:500 dilution in PBS). After three washes in PBS and blocking with 5% (w/v) bovine serum albumin in PBS, the slides were exposed to secondary antibody (1:2,000 dilution in PBS, Alexa Fluor 488 goat anti-rabbit IgG) for 2 h at room temperature, washed five times in PBS and mounted with 50% (v/v) glycerol in PBS. Samples were examined with a laser-scanning confocal microscope (LSM510; Carl Zeiss).
Preparation of chloroplasts for immunoblotting. Leaves of 4-to 5-week-old plants were homogenized with a biomasher III (Nippi) in buffer containing 20 mM MOPS-Tris (pH 7.0), 330 mM sorbitol, 0.2 mM MgCl 2 , 10 mg ml À 1 pepstatin A and 10 mg ml À 1 leupeptin. The extract was centrifuged at 500 Â g at 4°C for 1 min to remove cell debris. The resultant supernatant was carefully obtained and centrifuged again at 3,000 Â g for 5 min at 4°C. The pellet (chloroplast fraction) was suspended with the same buffer. The fraction (50 mg) was separated by 10% SDS-PAGE and analysed by immunoblotting with rabbit polyclonal antibodies against AtPHT4;4 (1:1,000 dilution) and AtLhcb1 (1:5,000 dilution) 17 . Full-size blot images are shown in Supplementary Fig. 8b.
Measurement of ascorbate. Ascorbate levels were measured as described previously 28,29 . The same batch of chlorophyll was immediately frozen in liquid nitrogen. The frozen leaves were ground to a fine powder and extracted with 400 ml of 0.2 N HCl for 30 min. The extract was neutralized with 0.2 N NaOH and 0.2 M NaH 2 PO 4 (pH 5.6), centrifuged at 16,000 Â g for 10 min at 4°C and the supernatants were pooled. Total ascorbate was determined by spectrophotometry to measure ultraviolet absorption by the reduced form of ascorbate at 265 nm (A 265 ). In this assay, ascorbate oxidase was used to oxidize all ascorbate, and the amount of ascorbate was determined from the difference in A 265 before and after addition of the enzyme. To obtain values for oxidized and total ascorbate, samples were reduced by addition of 5 mM dithiothreitol. The values were corrected relative to chlorophyll a.
Extraction of ascorbate from chloroplast fraction. Plants were germinated and grown on MS medium containing 1% (wt/vol) sucrose and 0.8% agar in a growth chamber at 22°C under a 16-h light/8-h dark cycle. Leaves of 2-week-old plants were sliced in buffer containing 20 mM MOPS-Tris (pH 7.0), 330 mM sorbitol, 0.2 mM MgCl 2 and 0.2% BSA. The extract was passed through cell strainer with 40 mm nylon mesh (Falcon), and centrifuged at 500 Â g at 4°C for 1 min to remove cell debris. The resultant supernatant was carefully obtained and centrifuged at 3,000 Â g for 5 min at 4°C. The pellet (chloroplast fraction) was extracted with 200 ml of 0.2 N HCl for 10 min. The extract was neutralized with 2 N NaOH and 0.2 M NaH 2 PO 4 (pH 5.6), centrifuged at 16,000 Â g for 10 min at 4°C and the supernatants were pooled.
Measurement of chlorophyll fluorescence. Standard modulated chlorophyll fluorescence measurements were performed with 15-min dark-adapted plant leaves using a miniaturized pulse-amplitude-modulated photosynthesis yield analyser (PAM 101/103; Walz) as described previously 30 . Leaves of 4-to 5-week-old plants were subjected to a saturating light pulse and then illuminated (230 mmol photons m À 2 s À 1 or 540 mmol photons m À 2 s À 1 ) for 10 min followed by 4 min of darkness. NPQ was calculated as (Fm À Fm 0 )/Fm 0 , where Fm 0 and Fm are the maximum PS II fluorescence in the light-adapted state and the dark-adapted state, respectively.
Measurement of pigments. Measurement of pigments was carried out as described previously 8,28 . The same batch of chlorophyll was immediately frozen in liquid nitrogen. The frozen leaves were ground to a fine powder and extracted with 150 ml of 80% (v/v) acetone by vortexing for 1 min. The extract was centrifuged at 16,000 Â g for 10 min at 4°C and the supernatant was saved. Another 150 ml of 80% (v/v) acetone was added to the pellet and mixed thoroughly. The extract was centrifuged again, and the supernatants were pooled. Aliquots of 50 ml of the supernatant were subjected to HPLC and separated on a Spherisorb S5 ODS 4.6 Â 250 mm cartridge column (Waters) at 25°C. Pigments were eluted with a linear gradient from 100% (v/v) solvent A (acetonitrile:methanol:0.1 M Tris-HCl, pH 8.0; 84:2:14 [v/v]) to 100% (v/v) solvent B (methanol:ethyl acetate, 68:32 [v/v]) for 15 min, followed by 3 min of solvent B. The solvent flow rate was 1.2 ml min À 1 . Pigments were detected by A 445 with a reference at 550 nm by a diode array detector. The values were corrected relative to chlorophyll a.
Data analysis. All numerical values are shown as the means ± s.e.m.; n ¼ 3-12, unless otherwise specified. Statistical significance was determined by Student's t-test. Significance was defined as *Po0.05 or **Po0.01.