Ascorbate Efflux as a New Strategy for Iron Reduction and Transport in Plants*

Background: Iron long distance transport in plants is underdocumented. Results: Iron is delivered to embryos as ferric complexes with citrate/malate. An ascorbate-mediated reduction step is further required to acquire iron. Conclusion: Ascorbate plays a key role for the chemical reduction and transport of Fe2+. Significance: The identification of iron ligands and the transport process is crucial to further understand how iron is distributed within the plant. Iron (Fe) is essential for virtually all living organisms. The identification of the chemical forms of iron (the speciation) circulating in and between cells is crucial to further understand the mechanisms of iron delivery to its final targets. Here we analyzed how iron is transported to the seeds by the chemical identification of iron complexes that are delivered to embryos, followed by the biochemical characterization of the transport of these complexes by the embryo, using the pea (Pisum sativum) as a model species. We have found that iron circulates as ferric complexes with citrate and malate (Fe(III)3Cit2Mal2, Fe(III)3Cit3Mal1, Fe(III)Cit2). Because dicotyledonous plants only transport ferrous iron, we checked whether embryos were capable of reducing iron of these complexes. Indeed, embryos did express a constitutively high ferric reduction activity. Surprisingly, iron(III) reduction is not catalyzed by the expected membrane-bound ferric reductase. Instead, embryos efflux high amounts of ascorbate that chemically reduce iron(III) from citrate-malate complexes. In vitro transport experiments on isolated embryos using radiolabeled 55Fe demonstrated that this ascorbate-mediated reduction is an obligatory step for the uptake of iron(II). Moreover, the ascorbate efflux activity was also measured in Arabidopsis embryos, suggesting that this new iron transport system may be generic to dicotyledonous plants. Finally, in embryos of the ascorbate-deficient mutants vtc2-4, vtc5-1, and vtc5-2, the reducing activity and the iron concentration were reduced significantly. Taken together, our results identified a new iron transport mechanism in plants that could play a major role to control iron loading in seeds.

Iron is an essential micronutrient for plants. It is used as an enzymatic cofactor in a wide range of metabolic processes, such as photosynthetic and respiratory electron transfer reactions, reduction of nitrate and sulfate, synthesis of fatty acids, and branched amino acids. Iron was selected on the basis of the redox capacity of the Fe 2ϩ /Fe 3ϩ couple that can exchange one electron. This chemical property is also responsible for the potential toxicity of iron in excess conditions because ferrous ions can promote, in the presence of oxygen, the generation of reactive oxygen species and oxidative stress. Plants, as sessile organisms, must maintain a strict iron balance to cope with deficiency or excess conditions. At the whole plant level, this is achieved by the fine regulation of root absorption, allocation to the aerial parts, storage, and remobilization.
Although very abundant in soils, iron is often poorly available to plants because the main iron form, iron(III), has a very limited solubility and is prone to precipitate in alkaline and neutral conditions. Therefore, plants have developed strategies to efficiently acquire iron. In roots of dicotyledonous species, such as the model plant Arabidopsis thaliana, the uptake of iron requires the acidification of the rhizosphere by the proton pumping activity of the ATPase AHA2 (1) to release Fe 3ϩ bound to humic substances and mineral phases, the reduction of ferric ions by the ferric chelate reductase encoded by FRO2 (2), 4 and the subsequent uptake of Fe 2ϩ by the metal transporter IRT1 (3)(4)(5)(6). * This work was supported by the Centre National de la Recherche Scientifique, by l'Institut National de la Recherche Agronomique, and by Agence Nationale pour la Recherche Grants 23643 (CIDS) and 07-3-18-8-87 (DISTRIMET). 1 Both authors contributed equally to this work. 2 Supported by a Ph.D. fellowship from the Plant Biology Department of the Institut National de la Recherche Agronomique. 3 To whom correspondence should be addressed: Laboratoire  When in root cells, iron atoms must reach the vascular system and enter the xylem vessels to be transported to the aerial parts. Xylem loading is controlled, at least in part, by the FRD3 protein that belongs to the multidrug and toxic compound extrusion family of transporters (7). FRD3 is a citrate effluxer expressed in the root pericycle, whose activity is required to solubilize iron in the xylem sap (8), where citrate was identified as the main iron ligand (9), hence promoting iron movement into the xylem. In leaves, unloading of iron from xylem vessels and distribution in the surrounding cells also requires FRD3 (10). Likewise, the oligopeptide transporter OPT3 (11) contributes to iron movement in leaves, but its substrate, whether it is an iron ligand or an iron complex, is still unknown (11). To date, the number of organic molecules shown to form complexes with iron in vivo is extremely limited. Beside citrate, most of the information concerns nicotianamine (NA). This aminopropyl polymer, enzymatically synthesized from S-adenosylmethionine, has a high affinity for iron, copper, and zinc. Several studies have shown that the chemical properties and the binding capacity of NA make it an ideal ligand for iron in neutral, cytosolic conditions, whereas in acidic, apoplastic conditions, citrate will prevail (12,13). In planta, NA is involved in the rootto-shoot transport of copper in tomato (14) and nickel and zinc in the hyperaccumulator species Thlaspi caerulescens (15,16). For nickel, the formation of a complex between nicotianamine and nickel in xylem exudates has been demonstrated unambiguously by the coupling of chromatography and mass spectrometry (15,18).
In A. thaliana, several genetic approaches have allowed to pinpoint the role of NA in the transport of iron. The impairment of NA production, through the inactivation of all four NA synthase-encoding genes, provokes the expected phenotype of interveinal chlorosis described for the NA-free tomato mutant chloronerva and, more interestingly, a reduced accumulation of iron in seeds (19). Such an impact on seeds has also been reported in mutants of the yellow stripe 1-like (YSL) family of NA-metal transporters. The mutation in AtYSL1 results in decreased iron and NA accumulation in seeds (20), most likely caused by a reduced remobilization from old leaves, as shown for the Arabidopsis ysl1ysl3 double mutant (20,21). Taken together, these results indicate that disturbing the long distance transport of iron has a major impact on the process of seed loading (reviewed in Ref. 22). Seeds represent the most important sink for plants and the end point of long distance circulation of nutrients (23,24). Virtually nothing is known about the transport pathway that leads to iron loading in the seeds and on the chemical forms (i.e. the speciation) of iron during this process. The speciation of a metal ion in a specific biological compartment is crucial information to further understand how the metal is kept soluble, transported across membranes, and delivered to its final target.
In this work, the pea (Pisum sativum) was chosen as a biochemical model to study the mechanisms of iron acquisition by the embryo. Grain legumes are particularly suited to study the processes of transport in seeds. One specific feature of these seeds is that the bulk flow of nutrients delivered by the seed coat accumulates in large amounts, forming a liquid endosperm also termed embryo sac liquid (ESL). This particular liquid is of high biological importance because it serves directly to feed the embryo (25). We first analyzed the speciation of iron in isolated ESL using chromatography coupled to mass spectrometry and synchrotron radiation x-ray absorption spectroscopy. This information was crucial to further characterize the transport machinery that is expressed at the embryo surface to acquire iron from maternal tissues.

Liquid Endosperm Sampling
Developing pods were dissected with a surgical blade. Two holes were pierced into each seed using a glass capillary tube. The liquid endosperm was then extracted quickly with another glass capillary using a peristaltic pump and blown gently at the bottom of an Eppendorf tube kept on ice. Pumping was carried out carefully to avoid bubbling. Samples were then maintained frozen in liquid nitrogen and kept at Ϫ80°C until further analysis.

Iron Speciation by X-ray Absorption Spectroscopy
Iron K-edge x-ray absorption near edge structure (XANES) measurements were performed on LUCIA Beamline at Soleil Synchrotron, Gif sur Yvette, France, operating at 2.75 GeV (26). The x-ray beam was monochromatized using a Si(111) twocrystal monochromator, and spectra were collected in fluorescence mode using a Bruker Silicon Drift diode and a nonfocused beam. All measurements were done in a vacuum and in cryo conditions with a He cryostat providing a temperature of 110 K on the sample holder. A few drops of glycerol were added to liquid references to minimize the formation of ice crystals. No glycerol was added to the ESL because it already contains high concentrations of sugars. Energy was calibrated with metallic iron. The XANES spectra were corrected from background absorption by subtracting a linear function before the edge and normalizing using a linear or to-degree polynome. A fingerprint approach was then used to simulate the experimental spectra by a linear combination of iron reference compound spectra. The quality of the fits was estimated by the normalized sum-squares residuals NSS ϭ ⌺ (Xanes experimental Ϫ Xanes fit) 2 / ⌺ (Xanes experimental) 2 ϫ 100 in the 7090-7380 eV range. The iron valence was also estimated by the examination of the preedge structure (27).
Iron reference compounds were prepared for comparison with ESL samples. Aqueous 50 mM iron(II) and iron(III) was prepared by dissolving FeCl 2 and FeCl 3 in deionized water at pH 2. Iron(II) preparation was performed in an argon atmosphere just before analysis to prevent iron oxidation. 25 mM iron(III)-citrate, iron(II)-citrate, and iron(III)-malate were prepared with metal/ligand (M/L) ϭ 1/10 at pH 5.6 and 5.2, respectively. On the basis of the Visual MINTEQ software for chemical speciation (28), under these conditions, more than 99% of iron is found as organic complexes. Iron(III)-histidine was synthesized at pH 7.2 with a 1/10 ratio and a total iron concentration of 3.1 mM.

Iron Complexes by Chemical Analysis
Instrumentation-Chromatographic separations were performed using a model 1100 HPLC pump (Agilent, Wilmington, DE) as the delivery system. The HILIC column was a TSK gel amide 80 (250 ϫ 1 mm inner diameter) (Tosoh Bioscience, Germany). All connections were made of fused silica/PEEK tubing (0.050 mm inner diameter). The ICP MS instrument was Agilent 7500cs (Agilent Technologies, Japan) equipped with a collision cell with hydrogen as collision gas. The interface between HPLC and the Agilent 7500cs consisted of a glass Cinnabar cyclonic spray chamber (Glass Expansion, Australia), a Micromist U-series nebulizer (Glass Expansion), a 1-mm inner diameter quartz torch (Agilent Technologies), and a set of platinum cones (Agilent Technologies). The ESI MS instrument was an LTQ Orbitrap Velos mass spectrometer (Thermo Scientific, San Jose, CA).
LC MS Experimental Conditions-The mobile phase was 5 mmol/liter ammonium acetate buffer (pH adjusted to 5.5 by addition of acetic acid) (A) and acetonitrile (B). Analytical reagent-grade chemicals were used for buffer preparation and were purchased from Sigma-Aldrich (Saint Quentin, Fallavier, France). Water (18.2 M⍀/cm) was obtained with a Milli-Q system (Millipore, Bedford, MA). The optimized HILIC gradient started with 10% of A and was increasing to 50% of A during 25 min. The samples (diluted in acetonitrile 1/2, v/v, and 7-l injection) were eluted at a flow rate of 50 l/min. Iron-containing species were detected by direct online coupling of the HILIC column to the ICP MS instrument (following signals for 54 Fe and 56 Fe isotopes) and to the ESI MS instrument. The settings and parameters of ICP MS and ESI MS instruments were optimized daily for highest intensities and lowest interferences using a built-in software procedure (ICP) and a standard mixture recommended by the constructor (ESI). ES MS spectra were acquired with a maximum injection time of 400 ms in the range of 230 -1000 mass units. The electrospray voltage was set at 3 kV in positive ion mode and Ϫ3 kV in negative ion mode.
The ions were fragmented by higher-energy collisional dissociation at various energy levels.
Ferric Reduction Assays-Ferric chelate reductase activity was estimated as the quantity of iron-bathophenanthroline disulfonate (BPDS) complexes formed in the assay solution calculated from the optical density measured at 535 nm with a Hitachi U-2800 spectrophotometer. A 535 was measured after incubation in the dark at 22-25°C with 250 rpm shaking in an assay solution that contained 5 mM MES buffer (pH 5.5), 300 M (BPDS), and 100 M iron(III)-EDTA unless stated otherwise. Pea embryos were dissected with a surgical blade as fast as possible and kept in 5 mM MES buffer (pH 5.5) for 30 min, representing the time to dissect and weigh around 15 embryos. Embryos were then incubated in the assay solution.
To measure root ferric chelate reductase activity, whole root systems from individual pea plants were excised, rinsed three times with demineralized water, and incubated in 50 ml of assay solution following the same protocol. The comparison of embryo ferric chelate reductase activities was repeated five times independently, and each experiment consisted of the incubation of five embryos in 5 ml of assay solution with the usual protocol. Final results are the mean Ϯ S.D. of the five values obtained. Arabidopsis root ferric chelate reductase activity was measured from individual root systems of iron-deficient seedlings in 1 ml of assay solution.
Ferric Reduction Activity Measurement of Exudates-Exudates consisted of 5 ml Milli-Q water in which five pea embryos were incubated, representing a total fresh weight of 0.7-1 g. For the efflux kinetics, embryos were incubated 10, 20, and 30 min or immersed and removed immediately for the 0-min time point. 100 M iron(III)-EDTA and 300 M BPDS were then added to exudates and allowed to react for 1 h in the dark prior to A 535 determination. For the comparison of exudate activity with embryo activity, exudates were collected in Milli-Q water after 30 min of incubation and allowed to react with 100 M iron(III)-EDTA and 300 M BPDS for 1 h in the dark. In the case of ascorbate oxidase (AOX) treatment, 1.5 unit of AOX/ml of exudate was added 2 min before iron(III)-EDTA and BPDS addition. In this experiment, embryo ferrireduction activity was quantified following 30

Iron Speciation in The Embryo Sac Liquid by XANES-
The aim of the XANES spectroscopy measurements was to examine the redox status of iron and the potential iron complexes in the ESL under conditions minimizing the possible changes of iron species during sample preparation and analyses. For that, the ESL sample was frozen immediately after collection and analyzed without any chemical additive and in the frozen state to limit speciation change and beam radiation damage.
XANES spectra of iron references show that the edge of iron(II) compounds is shifted to lower values compared with iron(III) compounds, as expected (Fig. 1A). Iron complexes involv-ing malate and citrate, where the metal is coordinated by oxygen atoms from carboxyl groups, show similar spectral features but can be distinguished from iron-histidine.
The ESL spectrum edge mainly exhibits edge features of iron(III) compounds (Fig. 1A). However, the spectrum did not match one iron species only, and results of linear combination fitting showed that a mixture of 79% iron(III)-citrate and 21% iron(II) aq gave a satisfactory fit (NSS ϭ 0.0015, Fig. 1B). A correct match is also obtained for a combination of iron(III)-citrate and iron(II)-citrate (data not shown, NSS ϭ 0.0020), suggesting the presence of a minor iron(II) species, that was not clearly identified. Fits with iron(III)-malate instead of iron(III)-citrate gave a poorer agreement (NSS ϭ 0.0022), and the improvement of the simulation, when considered as a third component, was not significant (NSS ϭ 0.0014). We can thus infer that the major part of iron is present as iron(III) in the ESL, probably as organic acids, whereas a minor iron(II) species can occur but remains unidentified. Although it is difficult to clearly rule out a possible photoreduction under beam exposure (29), this effect was checked by successive collection of spectra on the samples. No reduction was observed, neither on the ESL nor on the iron(III) model compounds, and, thus, we can infer that the occurrence of iron(II) is not an artifact.
Iron Speciation in the Embryo Sac Liquid-The analysis of iron speciation in the embryo sac liquid was performed by alternatively coupling hydrophilic liquid chromatography to two different mass spectrometry instruments, an ICP MS to determine the number of significant iron-containing species within this sample and quantify iron and an ESI MS to identify the corresponding species (retention time matching) using the same chromatographic conditions. HILIC chromatography has been established recently as an adequate technique to preserve and separate metal complexes originating from raw samples, allowing their identification and quantification by mass spectrometry (18). It was applied here to the liquid endosperm samples. Prior to analyses, iron recovery was tested during sample preparation (mixing with acetonitrile followed by centrifugation) and during the chromatographic run. An overall 85 Ϯ 8% of iron species were conserved after the two consecutive procedures, which meant that significant amounts of the original iron species were conserved by this method. After chromatographic coupling with ICP MS, two major iron species as well as a minor iron species were detected ( Fig. 2A). By HILIC ESI MS in positive ion mode, these iron-containing compounds were identified, respectively, as compound 1 at m/z 811.84, compound 2 at m/z 869.85, and compound 3 at m/z 437.97 (Fig. 2B). MS 2 fragmentation of compounds 1 and 2 allowed to determine that organic molecules containing exclusively carbon, hydrogen, and oxygen with carboxylic acid groups were surrounding a nucleus constituted of three ions of iron, which was consistent with the observed isotopic profile (Fig. 2C). It was deduced that these two metal complexes were constituted of a mixture of citrate (Cit) and malate (Mal) binding three iron(III) ions. Each of these two complexes then contains four organic acid molecules, two citrate molecules, and two malate molecules in compound 1 (Fe(III) 3 Cit 2 Mal 2 ) and three citrate molecules and one malate molecule for compound 2 (Fe(III) 3 Cit 3 Mal 1 ). Compound 3 was found to be a complex between two citrate molecules and one iron(III) ion (Fe(III)Cit 2 ). According to this, the chemical formula of each complex could be determined, and experimental and theoretical masses could be compared: Ϫ0.5 ppm. The analysis of a standard mixture of iron(III)-citrate-malate (0.2:1:5 ratio, buffered at pH 5.5 with ammonium acetate) definitively confirmed the formation of these iron-citrate-malate complexes through the appearance of the abovementioned m/z (compounds 1 and 2) in MS spectra. The mixed complex structures is unclear, but is probably similar to the iron(III) 3 Cit 3 complex structure as postulated by Fukushima et al. (17). That is, a Fe 3 core is expected to occur with iron ions bound to each other through ␣-hydroxy groups of organic acids, and carboxylic groups are surrounding iron ions. The absence of an Fe 3 Cit 3 complex in the analyzed sample seems to indicate that the mixed iron(III)-citrate-malate would be an even more stable species.
Traces of an iron(II)-NA complex (compound 4, C 12 H 18 N 3 O 6 Fe Ϫ , m/z exp 356.05507, m/z theo 356.05506, ␦m ϭ 0.0 ppm) were also detected in negative ion mode (Fig. 2B), even when oxic (because of sampling and chromatography) and acidic (pH 5.5) conditions were not favorable to the presence of this form compared with iron(III)-organic acid complexes. Even when iron(III) was the preponderant iron form in the sample, iron(III)-NA could not be detected, which seemed to indicate that NA would have a relative higher affinity for iron(II) than for iron(III) in the presence of citrate and malate, suggesting a differentiated speciation depending on the iron redox status.
Pea Embryos Are Capable of Ferric Reduction through Ascorbate Efflux-Strategy I plants are not able to perform direct transport of ferric iron, but, instead, rely on a ferric reduction step prior to ferrous ion uptake. Therefore, we wondered whether ferric chelate reduction was also part of the iron uptake machinery of the embryo. Embryonic ferric chelate reductase activity was assayed by incubating isolated immature pea embryos with the Fe 2ϩ chelator BPDS in the presence or the absence of Fe 3ϩ . With Fe 3ϩ in the assay, the embryos were capable of generating Fe 2ϩ , visualized by the formation of a pink iron(II)-BPDS 3 complex (Fig. 3A). In contrast, no coloration was detected when Fe 3ϩ was omitted in the incubation medium (Fig. 3A, ϪFe 3ϩ ϩ BPDS), thereby ruling out the possibility that production of an iron(II)-BPDS 3 complex is caused by the secretion of Fe 2ϩ by the embryo. Therefore, we concluded that pea embryos are capable of reducing exogenous Fe 3ϩ .
To check whether the ferric reduction activity was catalyzed by a FRO-like transmembrane reductase or mediated by the efflux of reductive compounds, we compared the ferric reduction activity measured with embryos in the medium to the activity of embryo exudates (i.e. by removing the embryos before measuring iron reduction). Surprisingly, the reduction activity measured after removing the embryos was almost as high as the one obtained with embryos in the medium (Fig. 3B). This result indicates that, contrary to what has been described in roots, most of the reduction activity of pea embryos is due to the efflux of potent reductants in the medium.
Because the most obvious candidates reducing molecules in plants are ascorbic acid and glutathione, we looked whether these compounds were present in embryo exudates by ESI MS. As expected, ascorbate and glutathione could be detected by ESI MS analysis of embryo exudates (Fig. 3, C and D). To discriminate between these two molecules, we tested their capac- ity to reduce ferric iron in vitro under the conditions used with embryos (100 M Fe 3ϩ , 300 M BPDS). Under these conditions, we could show that ascorbate, but not glutathione, was necessary and sufficient to spontaneously reduce iron in vitro (Fig.  3E). To confirm that the iron-reducing molecule effluxed by embryos was ascorbate, we tested the effect of AOX on the reducing activity of isolated embryos over 50 min of incubation. The time course analysis of iron reduction, linear for 50 min, was strongly inhibited by AOX in a dose-dependent manner (Fig. 3F), confirming that ascorbic acid secreted by the embryos was responsible for the ferric reduction activity measured with isolated embryos.
We then checked whether this reduction activity was regulated by the iron status, as described previously for the root reductase. Pea plants were thus grown in hydroponic culture under standard (50 M iron) or deficiency conditions (no iron added). As expected, the root reductase activity was strongly induced under iron deficiency conditions (Fig. 4A). However, the reduction activity of embryos isolated from ϩiron and Ϫiron plants was not significantly different (Fig. 4B). We also used the iron signaling mutant dgl that constitutively expresses the iron acquisition machinery in roots (30). Consistent with previous reports, the root reductase activity was constitutively high, regardless of the iron status (Fig. 4A), although the reduction activity of embryos was not significantly different, neither between ϩiron and Ϫiron nor between the wild-type (DGV) and the mutant embryos (B). In conclusion, contrary to the root system, the ferric reduction activity expressed at the embryo surface is not regulated by the iron status.
Pea Embryos Take up Ferrous Iron-To assess the importance of ascorbate-mediated ferric reduction for embryo iron uptake, we quantified the accumulation of the radioactive isotope 55 Fe, provided as iron(III)-EDTA, into isolated embryos. To evaluate Fe 2ϩ and Fe 3ϩ uptake rates by the embryo, an uptake assay was set up by adding either AOX, to prevent the formation of Fe 2ϩ because of ascorbate-mediated reduction, or BPDS, which is expected to scavenge newly formed Fe 2ϩ . The rationale for this experiment was that if an ascorbate-mediated reduction step is required, then both BPDS and AOX should inhibit the uptake.
Untreated embryos accumulated 105 (Ϯ27) nmol/g of fresh weight/h when incubated at 30°C and 43 (Ϯ11) nmol/g FW/h at 4°C. In the presence of AOX and BPDS, the iron accumulation rates, not significantly different from the 4°C control, were, respectively, 62 (Ϯ8) and 22 (Ϯ4) nmol/g FW/h (Fig. 5). When expressed as net influx, by subtracting the iron accumulation at 4°C, it appeared clearly that both BPDS and AOX significantly inhibited the uptake of iron by isolated embryos (Fig. 5, inset). On the basis of these results, it was concluded that ascorbate-mediated ferric reduction is an absolute requirement for iron uptake by pea embryos, thus uncovering a new role for ascorbate in plants.
Ascorbate Efflux Is Not Limited to the Pea-These new findings on pea embryos prompted us to check whether this ascorbate efflux activity was a general mechanism, found in other species such as the model plant A. thaliana. Embryos were isolated (bent cotyledon stage) from developing siliques of Arabidopsis (Fig. 6A, inset), and iron reduction activity was measured as described previously for peas (Fig. 3). Arabidopsis embryos were indeed capable of reducing iron, and this activity was reduced significantly by the addition of AOX in the incubation medium (Fig. 6A). This ferric reduction activity appeared to be fully attributable to ascorbate efflux because the rates obtained with embryos or with exudates (i.e. after removal of embryos from the medium) were not significantly different. The addition of AOX to embryo exudates completely abolished iron reduction, confirming the identity of ascorbic acid as the main reductant effluxed by Arabidopsis embryos (Fig. 6A). As a control for AOX specificity on the ascorbate-mediated reduction, we checked the effect of AOX on the well established root ferric reductase activity mediated by the membrane protein FRO2. We thus assayed the reductase activity of Arabidopsis roots isolated from iron-sufficient and iron-deficient plants (Fig. 6B). As shown previously, iron-deficient roots exhibited higher fer-  . Ascorbate-mediated ferric reduction is an obligatory step for iron influx. 55 Fe influx was measured in isolated pea embryos. Embryos were incubated in uptake buffer containing 100 M iron(III)-EDTA, and, when stated, 4 mM BPDS or 1.5 AOX units/ml (ϩAOX) was added. Incubation at 4°C with the complete uptake medium was also realized to measure passive adsorption of iron at the embryo surface. Inset, net influx of 55 Fe deduced by calculating the difference between iron accumulation at 30°C (complete, ϩAOX, ϩBPDS) and the 4°C control incubation. Each measurement was performed on four to five embryos of 20 -50 mg, and experiments were repeated three times independently. Results are the mean Ϯ S.E. (n ϭ 3). Significant differences (indicated by the different letters) were determined with a nonparametric Kruskal-Wallis test (p Ͻ 0.05).
ric chelate reductase activity than iron-replete roots (Fig. 6B). Interestingly, this activity was not significantly inhibited by AOX, indicating that ferric reduction in roots is biochemically different from that in embryos.
Finally, we sought to make a genetic link between iron reduction and accumulation with ascorbate metabolism. We thus analyzed the reduction activity and iron concentration in seeds and embryos on several vitamin C-defective (VTC) mutants, affected in ascorbate synthesis (31). We first isolated a new T-DNA insertion line of the VTC2 gene, vtc2-4 (Fig. 7A) and showed, by RT-PCR, that it was a knockout allele (Fig. 7B). This mutation led to a 70% decrease in ascorbate accumulation in leaves (Fig. 7C), which is in agreement with previous results on other vtc2 alleles (31). Embryos were then isolated from vtc2-4, vtc5-1, and vtc5-2 mutant seeds to measure ferric reduction activity.
Compared with the wild type, the iron reduction activity of the vtc mutant embryos was reduced significantly (Fig. 7D). Furthermore, the iron accumulation in mature embryos of vtc2-4, vtc5-1, and vtc5-2 was also significantly lower than in the wild type, as shown by Perls staining that reveals iron by a blue coloration (Fig. 8A). Wild-type mature embryos contained detectable amounts of iron in the veins, whereas, in vtc5 embryos, the staining was very faint, and vtc2-4 embryos did not contain detectable amounts of iron (Fig. 8A). The total iron content in whole dry seeds was also significantly lower in vtc mutants compared with the wild type (Fig. 8B). Collectively, these data demonstrate that embryos acquire iron through a new strategy of iron reduction and uptake that is dependent on, and mediated by, ascorbate synthesis and efflux.

DISCUSSION
In this work, we studied the mechanism of iron entry into the seed through two approaches: an iron speciation analysis at the interphase between the seed coat and embryo in the pea, associated with a biochemical dissection of iron transport on isolated embryos of the pea and Arabidopsis. We established that iron is delivered by the maternal tissues as ferric complexes with citrate and malate, that the embryos are capable of reducing the ferric ions of these complexes through the efflux of ascorbate, and that this reducing activity is an obligatory step for iron transport as iron(II). The analysis of iron speciation is by far the least documented aspect of iron homeostasis, both in animals and plants. This is principally due to the difficulties of obtaining biologically reliable samples (i.e. without mixing cellular compartments and creating artifact complexes) and the low concentration of iron in living tissues. To our knowledge, the iron speciation analysis reported in this study is one of the few unambiguous reports on the identification of iron complexes in vivo. Ferric-citrate complexes have long been proposed to occur in apoplastic compartments, but only recently a has a speciation approach led to the molecular identification of a tri-iron(III) tricitrate complex as the major iron-circulating form in the xylem of tomato plants (9). The ESL of pea seeds, although highly concentrated in nutrients (sugars, amino acids, etc.) from symplastic origin (32), is characterized by an acidic pH (5.4) that is similar to the pH of an apoplastic compartment. Thus, our findings reinforce previous in silico (12) and in vitro studies (13) that demonstrated that, under mildly acidic conditions, ferric citrate would be the most stable and abundant type of complex, despite the presence of other ligands like nicotianamine. Moreover, our data identified malate as a new player in FIGURE 6. The ascorbate efflux mechanism is conserved in A. thaliana embryos. A, embryos at the bent torpedo stage were dissected and incubated to measure the ferric reduction activity, either with embryos in the medium or after removing the embryos to measure the activity on the exudates. When stated, AOX was added to the incubation medium. Results are the mean Ϯ S.D. of three independent replicates. Each condition in an individual experiment was realized with 30 -50 embryos. Different letters indicate significant differences among means (p Ͻ 0.05 by Tukey's test). B, the root ferric chelate reductase activity is not inhibited by AOX. Ten-day-old seedlings grown in 50 M iron(III)-EDTA (ϩFe) were transferred in a medium without iron and supplemented with 30 M FerroZine (-Fe) for 3 days and then assayed for ferric reductase activity. When stated, AOX was added to the incubation medium. Values are the mean of five plants, and different letters indicate significant differences among means (p Ͻ 0.05 by Tukey's test). iron speciation in plants. Although one of the metabolic responses to iron deficiency is the accumulation of organic acids, including malate (34), a direct role of malic acid has never been clearly established. Instead, malate has been involved in the hyperaccumulation of nickel in T. caerulescens leaf extract (18) and Alyssum serpyllifolium with the formation of complexes with nickel in the xylem sap (35). Malate was also found to be involved in the storage of zinc in the leaves of the hyperaccumulator Arabidopsis halleri (36). Malate is also a key molecule for the tolerance to aluminium because it is excreted from roots by aluminium-activated malate transporters to chelate aluminium in the rhizosphere (37)(38)(39). Aluminium-activated malate transporter family members, therefore, represent putative candidate transporters whose function could be to load malate in the embryo sac liquid. Belonging to the same multidrug and toxic compound extrusion family of transporters as aluminium-activated malate transporters, FRD3 is a citrate efflux transporter that has been proposed recently to play a critical role in the movement of iron between tissues that are not symplastically connected. A detailed analysis of the expression pattern of FRD3 and a phenotypical analysis of frd3 knockout mutants have demonstrated the importance of FRD3-mediated citrate efflux in iron unloading in the leaf (from the xylem to adjacent cells) and in the anther (from the tapetum to pollen grains) (10). Interestingly, the seed coat and the embryo protodermis are also disconnected completely, and FRD3 is highly expressed in both tissues in A. thaliana (10). Thus, citrate efflux by FRD3 could be important as well to load the ESL with citrate and, subsequently, induce the formation of ironcitrate complexes. More generally, our results strengthen the importance of citrate in the transport of iron between cells and organs. Citrate may, in fact, be a ubiquitous and important intercellular iron ligand. Indeed, in mammals citrate is the main ligand of the non-transferrin-bound iron species (40,41), which appear to be the most abundant chemical form of iron circulating in the extracellular medium of the brain (42,43), used as a source of iron for brain cells (44). Most of the iron present in the liquid endosperm was found to be Fe 3ϩ bound to a mixture of citrate and malate. It is the first observation in vivo of such mixed complexes. Compounds 1 and 2 correspond to two iron complexes that most likely participate in binding iron(III) to form a soluble, stable (contrary to the iron(III)Cit 2 complex, for example), but low-reactive species. Indeed, iron(III) alone would precipitate, leading to iron deficiency, whereas its presence in a weakly complexed form would probably lead to increased stress for the plant because of expected frequent random exchanges of the Fe 3ϩ ion with other nonspecific binding sites in surrounding molecules. Therefore, these mixed complexes with organic acids probably participate in making iron(III) available in the liquid endosperm for transport/reaction, such as reduction by ascorbate in the vicinity of the embryo, without inducing significant metal or oxidative stress.
The identification of iron(II)-nicotianamine in the ESL represents the definitive proof of the in vivo role of nicotianamine in the movement of iron, although, in our case, the proportion of this complex is several orders of magnitude lower than the iron bound to organic acids. Nevertheless, the presence of iron-NA in the ESL could be interpreted as the signature of the activity of YSL transporters. The fact that iron translocation to the seed is controlled, in part, by YSL transporter activity has been shown in Arabidopsis. The closely related members AtYSL1 and AtYSL3 have been involved in the loading of iron, zinc, and copper to the seeds (20,21). In particular, the mutation in AtYSL1, which is expressed in the chalazal endosperm of the seed, provokes a reduced accumulation of iron and NA in mature seeds (20). Thus, the presence of iron-NA complexes, the bona fide substrate of YSL transporters, in the ESL leads to the assumption that iron is delivered from the seed coat to the ESL by YSL proteins.
Up to now, there has been no direct functional link between the metabolism of ascorbate and iron homeostasis. Instead, several reports had described the induction of ascorbate accumulation in response to iron deficiency and proposed that it could represent a mechanism to cope with potential reactive oxygen species production (45,46). The most unexpected finding of this study was the discovery of ascorbate efflux as the mechanism to reduce and acquire iron from the ferric complexes. Although other organic molecules, such as phenolics, can be excreted by roots of strategy I plants in response to iron deficiency, these molecules are not capable of reducing iron(III) in the medium (47). Furthermore, the iron reduction activity has, so far, only been attributed to membrane proteins of the FRO family, such as the plasma membrane proteins FRO2, FRO4, and FRO5 (2, 48) and the chloroplastic protein FRO7 (49). In contrast, this report describes a new strategy of iron acquisition on the basis of a reduction activity that is independent of the FRO proteins but, instead, relies on the property of ascorbate as a powerful reductant. Unlike the root ferric reductase activity, the ascorbate-mediated reduction was not induced by iron deficiency or in the dgl mutant that constitutively expresses the iron deficiency responses, irrespective of the iron nutritional status, strengthening the idea that this new mechanism is molecularly distinct from the FRO-based iron transport and not regulated by the iron deficiency signaling pathway. Furthermore, we have shown that the ascorbate efflux activity is also expressed in A. thaliana embryos. To further confirm the link between ascorbate metabolism and iron acquisition, we used mutants affected in ascorbate synthesis. The VTC2 and VTC5 genes encode two isoforms of GDP-L-galactose phosphorylase, one of the last enzymes of the ascorbate biosynthetic pathway (31,50). In terms of expression, VTC2 is predominant, and disruption of this gene provokes a 70% decrease in ascorbate accumulation in leaves, whereas mutations in VTC5 barely affect ascorbate accumulation (31). Both genes are expressed in A. thaliana embryos, and we show here that both mutations greatly impact the ascorbate efflux, iron accumulation in seeds and mature embryos. These data confirm the biochemical characterization performed in pea embryos and establish a direct and genetic link between ascorbate metabolism and iron transport in the plant.
Interestingly, a comparable iron uptake mechanism has been proposed in mammals. Erythroleukemia K562 cells and astrocytes are capable of reducing iron(III) using ascorbate efflux, generating iron(II) and dehydroascorbate that is reabsorbed to regenerate an ascorbate pool in the cytosol (51,52). Transmembrane cycling of ascorbate coupled to ferrous transport by the divalent transporter DMT1 appears as one possible route for iron accumulation in these cell types. Yet, in contrast to the constitutively high ascorbate efflux activity measured in embryos, ascorbate efflux was only triggered when cells were supplied with elevated dehydroascorbate levels. The molecular identity of the efflux transport system remains unknown in both plants and mammals, but the biochemical features of ascorbate efflux in mammalian K562 cells were compatible with the activity of an anion channel (51). Ascorbate efflux by pea embryos was insensitive to anion channel inhibitors (data not shown) but required a pH gradient, suggesting that it could be mediated by an H ϩ /ascorbate antiporter. The importance of the proton motive force has already been reported for the transport of nutrients in pea embryos on the basis of the specific expression in the outer cell layer of the cotyledons of genes encoding H ϩ /sucrose and H ϩ /amino acid cotransporters (33,53). Future work will be aimed at molecularly identifying this ascorbate efflux system.