Molecular definition of the ascorbate-glutathione cycle in Arabidopsis mitochondria reveals dual targeting of antioxidant defenses in plants.

Key components of the ascorbate-glutathione cycle in Arabidopsis cell organelles are encoded by single organellar targeted isoforms that are dual localized in the chloroplast stroma and the mitochondrion. We demonstrate the presence of the ascorbate-glutathione cycle in purified Arabidopsis mitochondria using enzymatic activity, proteomic and in vitro and in vivo subcellular targeting data that identify the gene products responsible. In vitro experiments using a dual import assay assessing mitochondrial and chloroplast imports simultaneously show dual targeting of ascorbate peroxidase, monodehydroascorbate reductase, and glutathione reductase gene products to mitochondria and chloroplasts, while a putative dehydroascorbate reductase protein is only imported into mitochondria. In vivo subcellular localization using green fluorescent protein fusion proteins show clear targeting of all gene products to mitochondria. Transcript levels show these genes are induced by oxidative chemical stresses targeted to chloroplasts and/or mitochondria and are elevated during photosynthetic operation in the light. Together these data present a model of an integrated ascorbate-glutathione antioxidant defense common to plastids and mitochondria that is linked at the level of the genome in Arabidopsis.

The ascorbate-glutathione cycle is catalyzed by a set of four enzymes, ascorbate peroxidase (APX), 1 monodehydroascorbate reductase (MDHAR), glutathione-dependent dehydroascorbate reductase (DHAR), and glutathione reductase (GR). This cycle operates in chloroplasts of plants in order to remove the large amounts of H 2 O 2 generated during photosynthetic operations in the light through thylakoid electron transport chain compo-nents with electrochemical potentials capable of direct reduction of O 2 (1). The highest rates of O 2 reduction occur on the reducing side of photosystem I (1). The high specific activity of these enzymes in plastids and extensive work on its operation clearly indicates that this is the major H 2 O 2 metabolizing pathway in these photosynthetic organelles (2). The primary peroxidation of ascorbate (Asc) by APX yields the monodehydroascorbate radical (MDHA) that is either directly reduced back to Asc by MDHAR (3) or undergoes non-enzymatic disproportionation to ascorbate and dehydroascorbate (DHA). Recovery of the DHA produced occurs via the glutathione-dependent reaction catalyzed by DHAR, and the oxidized glutathione dimers are re-reduced by the NADPH-dependent GR (4).
The chloroplast Asc-glutathione cycle is housed primarily in the stroma (see review in Ref. 5). A stromal APX and a thylakoid-bound APX have been identified and purified from several plant species. An extended C-terminal sequence on the thylakoid APX facilitates binding to the membrane and makes this isoform ϳ5 kDa larger than the stromal APX (5). Sequence analysis in a variety of plants clearly delineates the two chloroplast APX classes from the cytosolic APX isoforms (6). The MDHA formed in the lumen disproportionates to DHA and penetrates the thylakoid membrane into the stroma. MDHA produced by both APX isoforms is reduced by stromal MDHAR. This enzyme is not present in the lumen, and chloroplast isoforms are differentiated from cyotosolic isoforms by the presence of N-terminal targeting sequences (5,7). DHAR and GR activities convert the DHA translocated from the lumen and the DHA generated in the stroma. Detailed enzymatic investigation and establishment of the stromal location of these enzymes have been reported (5,8,9).
A series of reports have also measured the activity of some of the enzymes of the Asc-glutathione cycle in mitochondrial preparations from plants (9 -14). Evidence for a complete, chloroplast-like, Asc-glutathione cycle in plant mitochondria has also been presented. In green leaves, a mitochondrial cycle has been proposed to cope with photosynthetic and environmental stress-induced oxidative stress (15). In the microaerobic environment of N 2 -fixing legume root nodules, a mitochondrial Asc-glutathione cycle is proposed to remove mitochondrial-derived radicals to protect sensitive sites including the heme of leghemeoglobin (16,17). The latency of mitochondrial APX activities is generally low suggesting a location outside of the inner mitochondrial membrane (16,17). APX is predominantly membrane-localized in plant mitochondria (12,15,17). The best collective evidence for MDHAR, DHAR, and GR presence is from Pisum sativum leaves (9,14,15) and Phaseolus valgaris nodules (17). These biochemical data indicate much of the MDHAR, DHAR, and GR to be in the mitochondrial matrix (9,14,15). However the actual proteins and the associated genes * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. that are responsible for these activities in plants have not been systematically investigated. As a result, it has been unclear to what degree these biochemical data represent true mitochondrial enzymes and to what degree they may be compromised by the presence of plastidic, peroxisomal, or cytosolic contaminations.
In mammalian mitochondria, glutathione peroxidases and glutaredoxins are found to function in the direct depletion of H 2 O 2 (18). A mitochondria matrix-localized thioredoxin-dependent peroxidase is also proposed to be responsible for H 2 O 2 depletion in mammals (19). In yeast both an intermembrane space localized cytochrome c peroxidase and a matrix thioredoxin-dependent peroxidase have been implicated in H 2 O 2 removal (20). We have already identified a putative thioredoxin-dependent peroxidase in Arabidopsis mitochondria, but its role in H 2 O 2 depletion is unknown (21).
If an Asc-glutathione cycle exists in plant mitochondria two possibilities can be entertained. First, a discrete set of genes encoding mitochondrial-specific components of this pathway could be responsible. Suggestions of mitochondrial-specific isoforms have been made in biochemical studies (12,(15)(16)(17). Second, a series of organelle-targeted genes could exist that are authentically multitargeted to chloroplasts and mitochondria. Evidence has been presented for the dual targeting of pea GR to both mitochondria and chloroplasts (9), and very recently the products of differential transcripts from a single Arabidopsis MDHAR gene have been proposed to be targeted to mitochondria and chloroplasts (7).
Purification of Plant Mitochondria and Chloroplasts-Mitochondria were isolated from 7-day-old soybean cotyledons and chloroplasts from 10-day-old pea leaf tissue (23)(24)(25). Arabidopsis mitochondria were purified from dark grown cell culture according to Millar et al. (26). Subfractionation of mitochondria to matrix, inner membrane, and intermembrane space/outer membrane fractions was performed according to Sweetlove et al. (27). The mitochondrial protein concentrations were determined using the Coomassie Plus Protein Assay Reagent (Pierce) and the chloroplast chlorophyll concentrations by spectrophotometic measurements.
Assay of Asc-glutathione Cycle Enzymes-APX was measured as Asc oxidation by H 2 O 2 at 290 nm (e ϭ 2.8 mM Ϫ1 cm Ϫ1 ) in the presence of 0.2 mM Asc and 0.2 mM H 2 O 2 (15). GR was measured as NADPH oxidation at 340 nm in the presence of 0.2 mM NADPH and 0.5 mM oxidized glutathione (14). MDHAR was measured as NADH oxidation at 340 nm of MDHA generated using the Asc/Asc oxidase system using 0.2 mM NADH, 1 mM Asc, and 1 unit of Asc oxidase (Cucurbita enzyme, EC 1.10.3.3, Sigma A0157). DHAR was measured as Asc formation at 265 nm (e ϭ 14 mM Ϫ1 cm Ϫ1 ) in argon-flushed solution containing 0.2 mM reduced glutathione and 1 mM DHA (28). For latency measurements, spectrophotometric assays were performed in a solution of 0.3 M mannitol and 25 mM TES (pH 7.5) with or without addition of 0.05% (w/v) Triton X-100. As both KCN and SHAM can inhibit some of these enzymes, alternative respiratory inhibitors (50 M nPG and 5 M myxothiazol) were incorporated to prevent electron transport chain NAD(P)H oxidation interfering with spectrophotometric assays of enzymes in the absence of Triton X-100. For assays in subfractions from mitochondria samples, assays were performed in a solution of 0.05% (w/v) Triton and 25 mM TES (pH 7.5).
Synthesis and in Vitro Import of Precursor Proteins-Precursor proteins for soybean alternative oxidase (AOX) (29), pea small subunit of ribulose 1,5-bisphosphate carboxylase/oxygenase (SSU) (30), Arabidopsis Tim23 (22), APX, MDHAR, GR, DHAR, and GDXN were synthesized using the TNT® in vitro coupled transcription/translation reticulocyte lysate system (Promega, Madison, WI) in the presence of [ 35 S]methionine. Mutants were created using the QuikChange TM Site-directed Mutagenesis kit (Stratagene, La Jolla, CA) and confirmed by sequencing. Import experiments using the dual import system were carried out as previously described (31). Following import into soybean mitochondria, IMS depletion was carried out by resuspending the mitochondria in a hypotonic osmoticum solution (32). Imported products were resolved by SDS-PAGE and subsequently exposed to a BAS TR2040 phosphorimaging plate (Fuji, Tokyo).
Proteomic Analysis of Mitochondrial Proteins by Mass Spectrometry-Q-TOF MS/MS was performed on an Applied Biosystems Q-STAR Pulsar (Q-TOF MS) using an IonSpray source. Proteins to be analyzed were cut from the two-dimensional PAGE gel, dried at 50°C in a dry block heater and stored at Ϫ70°C. For the sequencing analysis, the proteins were digested with trypsin, injected in 50% acetonitrile, and doubly charged peptides fragmented by N 2 collision and analyzed by MS/MS were selected (27). Mass spectra and collision MS/MS data were analyzed with BioAnalyst software (Applied Biosystems, Sydney) and Mascot (Matrix Science, London) to identify matching Arabidopsis protein entries.
Oxidative Stress Treatments of Arabidopsis-Prior to RNA isolation, 17-day-old Arabidopsis plants were each treated with one of four different chemicals: 0.1 mM Solicam® DF Herbicide (here referred to as norflurazon) (Novartis Crop Protection Australasia Ltd., NSW), 1 mM menadione (Sigma, Castle Hill), 662.5 mg/liter Tryquat 200 (a commercially available herbicide combination of 437.5 mg/L paraquat and 225 mg/liter diquat, here referred to as paraquat) (Crop Care Australasia Pty Ltd., Pinkenba, Australia) and 25 M antimycin A. Chemicals (or water for control plants) were painted onto Arabidopsis using a paintbrush and leaf tissue was harvested 1 h and 6 h following treatment. Two independent RNA isolations were carried out on harvested leaves using the RNeasy Plant Mini kit (Qiagen, Clifton Hill).

Transient Expression of GFP Constructs in Vivo-
The expression cassette containing the cauliflower mosaic virus (CaMV) 35 S promoter, Arabidopsis ER chitinase signal peptide, m-gfp5 and the nopaline synthase terminator was amplified by PCR using plasmid pBIN m-gfp5-ER as a template and primers CaMV35S 5Ј F (5Ј-CGCAAGCTTCATGGA-GTCAAAGATTC-3Ј) and Nos term 3Ј-R (5Ј-CCGATCTAGTAACATAG-3Ј). The PCR product was cloned into a pGEM®-T cloning vector (Promega, Melbourne), generating construct pGEMGFP. The sequence corresponding to the AOX presequence, full-length TOC64, APX, MDH-AR, GR, and GDXN was amplified with appropriate restriction sites by PCR using plasmid DNA as templates and primers listed as follows: The ER signal peptide sequence was removed from the pGEMGFP construct by restriction digestion, PCR products were cloned upstream and in-frame of the m-gfp5 gene and verified by sequencing.
Suspension cells developed from soybean (Glycine max cv. Stevens) leaf tissue were grown in 250 ml flasks containing Gamborg media agitated at 130 rpm under approx. 20 Em Ϫ2 s Ϫ1 light at 28°C. The cell stocks were maintained in an exponential growth phase by subculturing (1:6 dilution) at 7-day intervals. For transient expression analysis, 7-day-old soybean suspension cells were aseptically coated onto filter paper (Whatman No. 1 55 mm diameter, Maidstone, England), which was then placed onto solid media containing soybean culture media supplemented with 200 mM mannitol, and incubated for 2 h at 28°C in the dark. Prior to transformation, 10 g of the pGEMGFP construct was precipitated onto 5 mg of 1 m (average diameter) gold particles (Chempur, Karlsruhe, Germany) using 2.5 mM CaCl 2 (Sigma, Castle Hill) and 100 mM Spermidine (Sigma, Castle Hill). Bombardment of soybean suspension cells using 0.5 mg of particles was carried out under vacuum using helium pressure of 1400 kPa. The cells were then incubated at 28°C in the dark overnight before detection of GFP using fluorescence microscopy (Axioskop2, Zeiss, Germany).

Asc-glutathione Cycle Enzyme Activities in Arabidopsis
Mitochondria: Latency and Subcompartmentalization-Measurement of APX, GR, DHAR, and MDHAR activities revealed that each could be detected in Arabidopsis mitochondria (Table I). The two-Percoll gradient method for the high purity separation of mitochondria from plastid and peroxisomes, yields mitochondria only contaminated 1.5% by plastids, and essentially free from peroxisomal and cytosolic compartments, based on marker enzymes (26). Maximal activities for each enzyme of the Asc-glutathione cycle measured in the presence of 0.05% Triton X-100 (Table 1A) were 2-5-fold lower than those reported in mitochondrial preparations from other plant species (15,17), but were clearly measurable above background levels of auto-oxidation. The latency of these activities suggested that GR and MDHAR activities were largely inaccessible in intact mitochondria, indicating a matrix location. In contrast more than half the APX and DHAR activities could be assayed by provision of substrates to intact mitochondria, indicating a significant portion of activity is probably on the intermembrane space side of the mitochondrial inner membrane (Table 1A). Under the same conditions, cytochrome c oxidase latency was in excess of 95%, proving a high degree of membrane integrity in our mitochondrial preparations. To further assess compartmentation of these activities, fractions enriched in matrix proteins (MA), intermembrane space and outer membrane vesicle proteins (IMS ϩ OM) and inner membrane proteins (IM) were prepared by osmotic stock and centrifugation. Contamination between compartments prepared by this protocol are Ͻ5% for the MA and IM samples, and about 25% MA and Ͻ5% IM in the IMS ϩ OM fraction (27,32,33). The IMS ϩ OM fraction represented 10%, the matrix sample 40% and the membrane fraction 50% of total mitochondrial protein. Based on these proportions, the distribution of activity between the fractions is presented as a percentage of total activity in mitochondria (Table 1B). By comparing the combined activities rate (Table  1B) with the ϩTriton solubilization data (Table 1A), the percentage activity recovery (70 -110%) during this subfractionation process was determined to be consistent. Nearly 90% of the APX and MDHAR activities were present in the inner membrane fraction, with about 10% in the matrix fraction. The low latency of APX (Table 1A) coupled to the subfractionation data for this enzyme (Table 1B) suggests that much of this activity is likely to be on the IMS side of the inner membrane. In contrast, the high latency of MDHAR (Table 1A) coupled to subfractionation data (Table 1B) suggests a predominantly matrix side localization. The GR and DHAR activities predominate in the matrix, but are also found in the other fractions, suggesting they are partitioned between compartments and can be partially membrane associated. These data are consistent with the view that the majority of the APX faces the outside of the inner membrane, while an Asc-glutathione generating system (MDHAR, DHAR, GR) operates in the matrix and to a lesser extent in the IMS to regenerate Asc.
Small Gene Families Exist for Each Enzyme of the Ascglutathione Cycle in Arabidopsis-To determine which Arabidopsis genes encoded the proteins responsible for these enzymatic activities (Table I), the Arabidopsis-predicted protein set from v3.0 of The Institute for Genomic Research (TIGR) genome annotation was searched by description to identify genes encoding the enzymes of the Asc-glutathione cycle (Table II). These protein sequences were then used in BLAST analysis of the protein data set to identify unannotated protein sequences with significant similarity to suggest they also represented members of these gene families. This analysis revealed there TABLE I Asc-glutathione enzyme activities in Arabidopsis mitochondria APX, MDHAR, DHAR, GR, and the mitochondrial marker cytochrome c oxidase (COX) were assayed according to the procedures outlined under "Experimental Procedures." A, the latency of each enzyme (% of activity dependent on solubilization) was calculated from measurements before and after solubilization with 0.05% (w/v) Triton X-100. B, the subcompartmentation of these activities was measured in matrix (MA), inner membrane (IM) and intermembrane space plus outer membrane (IMS ϩ OM) fractions and is expressed as % of the combined activities from all fractions. This combined sum of fraction activities is also noted in nmol min Ϫ1 mg Ϫ1 protein and is compared as % recovery with the ϩTriton rates in A. All data are from independent mitochondrial preparations, mean Ϯ S.D. (n ϭ 3) are likely 9 APX genes, 5 MDHAR genes, 5 DHAR genes, and 2 GR genes in Arabidopsis. Further, there is an extensive set of glutaredoxins (GDXN) that are likely to catalyze DHAR activities in Arabidopsis (full data not shown in Table II). DHAR activities purified from mammalian tissues are found to be GDXN (34, 35), the mammalian mitochondrial GDXN has DHAR activity (36) and DHAR activity of plant GDXNs is widely reported (37). All of these genes, except the APX At1g43220, were represented in EST databases at the time of this analysis, suggesting they are all expressed genes (Table  II).
Predicted Organelle Targeting of Asc-glutathione Cycle Gene Products-A simple comparison of apparent molecular masses of each predicted gene product within a family suggests that different classes of each protein are represented and that some have N-or C-terminal extensions that may represent targeting information for localization within the cell (Table II). Analysis of the predicted targeting of each gene product for APX, MDHAR, DAHR, and GR by three targeting prediction programs is presented in Table II. Based on the predictions by each program we have made a consensus prediction for each product as organellar targeted, cytosolic, or unclear (Ϫ). From this analysis we predict that only 6 of the 21 genes encode proteins that are likely to be organellar-targeted. In the case of MDHAR, GR, and the classical DHAR family, only one gene product of each class has a clearly predicted organelle-targeting presequence (Table II). The three organellar APXs have already been assigned to chloroplast locations in the lumen (At4g09010), thylakoid membrane (At1g77490) and the stroma (At4g08390) (6). The organellar GR and MDHAR are also assigned as the main chloroplast isoforms in Arabidopsis (7,38,39). A GDXN with strong mitochondrial targeting prediction is also included in Table II.
Proteomic Evidence for Presence of Proteins in Mitochondria-We have also been undertaking a comprehensive program of Arabidopsis mitochondrial proteome analysis (21,33) using two-dimensional gel electrophoresis and direct analysis of whole mitochondrial trypsin digestions. During this analysis peptides have been identified for several Asc-glutathione cycle enzymes that are present in purified mitochondria isolated from Arabidopsis cell culture (Table I). A set of 3 peptides match to APX At4g08390, a series of 4 peptides match to the MDHAR At1g63940, one peptide from the GR At3g54660 has been identified, one peptide from the DHAR At1g19570 and a set of 3 peptides matched to a GDXN At3g15660 (Table III). Matching of these peptides by ProID software against the full Arabidopsis protein set proved that these identifications were highly significant with most scores of greater than 90 for matches of MS/MS spectra and delta masses of Ͻ0.1 Da for matches of the parent peptides masses to predicted peptide masses of the matched proteins. Thus with the exception of DHAR At1g19570, the proteins detected by proteomic analysis are all predicted to be organelle-located.
Dual Targeting of APX, MDHAR, and GR Precursor Proteins to Chloroplasts and Mitochondria-Using the bioinformatic and proteomic analysis we decided that the most likely candidates for mitochondrial-targeted proteins were APX At4g08390, MDHAR At1g63940, and GR At3g54660. The situation with DHAR was not clear, as proteomic data pointed to either At1g19570 or the GDXN At3g15660, while targeting predictions pointed to At5g16710. To further resolve this situation all three proteins were used for import studies, further, the only other DHAR significantly expressed, At1g75270, was also included for completeness.
To ensure that import, rather than simply association of proteins with organelles was occurring, a dual import system established in our laboratory (31) was utilized for import of in vitro synthesized [ 35 S]methionine-labeled precursors of the test proteins and two control proteins (Fig. 1A). Assessing the import properties of APX At4g08390, MDHAR At1g63940, and GR At3g54660 using this system indicated that they were all imported and cleaved to a mature product in both chloroplasts and mitochondria. The mature products were protected from digestion by addition of thermolysin. In the case of APX there was an additional methionine residue 25 residues from the initiator methionine that resulted in an additional band upon translation. In order to determine which protein was dual targeted we removed the first and second methionines independently. Removal of the first methionine (APX⌬M1) resulted in a single precursor protein that was still dual targeted (Fig.   TABLE II Asc-glutathione cycle enzymes in Arabidopsis thaliana Predicted targeting to mitochondria (M), chloroplasts (C), or neither organelle (-) by Predotar, TargetP, or MitoProtII. EST numbers are derived from TIGR. The predicted molecular mass for each sequence is shown and asterisks are used to match these data with the experimental identifications of the mitochondrial isoforms in Table III and Figs. 1 and 2. An * indicates a positive identification by this approach. An X in the import column indicates that it was not imported into mitochondria using in vitro assays. 1A). In contrast changing methionine 2 to an isoleucine residue resulted in targeting only to mitochondria (see Supplementary  Materials). Mitochondrial import of APX, MDHAR, and GR proteins were not completely valinomycin sensitive, and re-mained protease protected (Fig. 1A). This suggests that a proportion of these proteins were not crossing the inner membrane. Such an observation is consistent with the earlier reports of partitioning of these enzyme activities between the TABLE III Mass spectrometry-based identification of Asc-glutathione cycle enzymes in Arabidopsis mitochondria Mitochondrial protein was digested by trypsin and subjected to MS/MS analysis. Peptides from five Asc-glutathione cycle components were identified, and the mass differences between observed and predicted peptides (delta mass), the ProID match score for the MS/MS spectra (score), and the peptide molecular mass (MM) are shown.  1. Targeting of precursor proteins for Asc-glutathione cycle enzymes to mitochondria and chloroplasts. A, import of precursor proteins into dual organelle import system. Lane 1, precursor protein alone. Lane 2, precursor protein incubated with pea leaf chloroplasts and soybean cotyledon mitochondria and chloroplasts re-purified. Lane 3, as lane 2 but with addition of thermolysin. Lane 4, precursor protein alone. Lane 5 precursor protein incubated with pea leaf chloroplasts and soybean cotyledon mitochondria and mitochondria re-purified. Lane 6, as lane 5 but with addition of thermolysin after import reaction. Lane 7, as lane 5 with the addition of valinomycin prior to import. Lane 8, as lane 7 but with addition of thermolysin after import reaction. B, import of precursor proteins into IMS-depleted mitochondria. Lane 1, precursor protein alone. Lane 2, precursor protein incubated with soybean cotyledon mitochondria. Lane 3, as lane 2 but with addition of proteinase K after import reaction. Lane 4, as lane 2 but with IMS-depleted mitochondria in place of mitochondria. Lane 5, as lane 4 but with proteinase K added. p and m refer to the precursor and mature forms of the protein, respectively. C, import of precursor proteins for Asc-glutathione cycle enzymes into Arabidopsis mitochondria. Lane 1, precursor protein alone. Lane 2, precursor protein incubated with Arabidopsis mitochondria. Lane 3, as lane 2 but with addition of proteinase K after import reaction. Lane 4, as lane 2 but with addition of valinomycin prior to import. Lane 5, as lane 4 but with proteinase K after the import reaction. IMS and MA fractions for MDHAR and GR, and the low latency of APX activity that suggested an outer face of the inner membrane orientation (Table I). To determine intraorganellar localization, imports into mitochondria were repeated and subsequent IMS-depletion (rupturing the outer membrane and removing the IMS) allowed assessment of the proportion of each imported protein that was accessible to protease in the IMS (Fig. 1B). AOX and TIM23 were used as controls to monitor the activity of proteases. AOX is an interfacial membrane protein that is not exposed to the IMS (40), and thus should remain fully protected when the outer membrane is ruptured. TIM23 is an inner membrane space protein with the N and C termini in the IMS. Upon rupture of the outer membrane and protease treatment the IMS domains are digested but the membrane portions are protected (22). This results in a shift in apparent molecular mass on SDS-PAGE and confirms that the added protease is active in digesting IMS located proteins. Upon rupture of the outer membrane these controls show that while AOX is fully protected, TIM23 is fully converted to a lower molecular mass (Fig. 1B). Upon rupture and salt washing of mitochondria imported APX protein was completely lost even without protease treatment indicating that it was removed in this treatment (Fig. 1B, lane 4). We have previously reported that this washing procedure also removed cytochrome c from the inner membrane (27,32). Imported MDHAR and GR mature proteins were partially resistant to proteolysis in mitoplasts further supporting a direct partitioning of these products between the IMS and the matrix spaces during import. These Arabidopsis proteins could also be imported into mitochondria from Arabidopsis cell culture. In each case the same pattern of import was observed into Arabidopsis mitochondria (Fig. 1C) as had been seen in soybean mitochondria (Fig. 1A).
The DHAR At1g75270 and At1g19570 were not cleaved to a mature product after incubation with either organelle and were resistant to all the proteases tested in the presence and absence of organelles, thus protease protection of these proteins could not be assessed to determine if import was occurring (data not shown). The DHAR At5g16710, predicted to be the main organellar targeted form in Arabidopsis (Table II), did not appear to be imported into either chloroplasts or mitochondria in vitro (data not shown). The GDXN At3g15660 was clearly imported and cleaved in mitochondria in a valinomycin-sensitive manner, but was not imported into chloroplasts (Fig. 1A).
Additionally, we have also conducted in vivo subcellular localization experiments using full-length APX⌬ M1, MDHAR, GR, and GDXN fused to green fluorescent protein (GFP). Constructs were introduced by biolistic transformation into nongreen plant cell cultures and fluorescence observed (Fig. 2). Non-green cultures were used as green plant material yields high chloroplast autofluorescence and chloroplast-targeted proteins were likely to obscure subtle co-targeting to mitochondria. We used the soybean AOX presequence fused to GFP as a mitochondrial control and Arabidopsis TOC64 fused to GFP as a plastid control. The AOX control produced the expected pattern for mitochondria, reticulate, punctate spots less than 1 m in size. The TOC64 control produced a plastid pattern of oval and rounded shapes of ϳ5 m in diameter. Fusion of the full-length protein for APX⌬ M1, MDHAR, GR, and GDXN to GFP all produced patterns that indicated mitochondrial targeting. However only GR displayed a pattern that also indicated plastid targeting. Thus under the conditions used we could definitely confirm a mitochondrial location for all four proteins using GFP as a marker for in vivo subcellular localization.
Transcript Abundances and Gene Regulation-Comparison of transcript abundance in leaves and roots revealed that all five Asc-glutathione cycle genes were expressed well in both tissue types (Fig. 3). In contrast, nuclear-encoded ribosomal transcripts for the plastidial rps1 dominated in leaves, while transcripts for the nuclear-encoded mitochondrial rps13 were more evenly distributed. These data provide further evidence that this set of organellar-targeting gene products are unlikely to be destined for photosynthetic roles in leaf tissue alone but rather are being synthesized in both root and shoot tissue, consistent with a dual location in chloroplasts and mitochondria.
Further assessment of gene expression was undertaken during oxidative stress treatments (Fig. 4). A wide variety of stress conditions and photosynthetic operation are known to induce transcripts of the cytosolic and chloroplast enzymes of the Asc-glutathione cycle in plants (5,(41)(42)(43). When we recorded transcript abundances in light grown and dark-adapted leaves, a 2-6-fold increase in the light in 4 of 5 transcripts was revealed. Notably, GDXN, the only transcript not considered to be multitargeted, is the only transcript that does not respond to light/dark changes. Transcript changes induced by specific chemical stresses in lighted leaves were then assessed. Norflurazon and paraquat are both capable of inducing chloroplast oxidative stress, by inhibiting carotenoid synthesis and by acting as an artificial electron acceptor from photosystem I, respectively. Menadione leads to oxygen activation and lipid peroxidation generally in membranes of the cell acting as a quinone-like analogue in one-electron donation reactions. Antimycin A is a specific inhibitor of complex III in the inner mitochondrial membrane and induces superoxide formation from a highly reduced ubiquinone pool in vitro and in vivo. The significant transcript fold changes induced by treatments (Fig.  4, asterisks) reveal that both plastidial and mitochondrial targeted stresses induce the MDHAR, GR and DHAR/GDXN transcripts. APX was not readily inducible by chemical stresses. Overall norflurazon and menadione yielded a later response peaking Ͼ6 h, while antimycin A and paraquat yielded more rapid changes in transcript abundance. However, even within a stress, some transcript responses were rapid (ϳ1 h) and had abated to near control levels by 6 h (such as MDHAR and GR with paraquat) while other transcripts were still increasing at 6 h (notably DHAR with paraquat). This complexity of response magnitude and timing hints at further levels of control on this antioxidant cycle, but its dual response to functional stresses emanating from the two organelles reveals a coordination of response consistent with transcriptional regulation of this gene set.

DISCUSSION
The Case for Molecular Definition of the Mitochondrial Cycle-We have used a variety of molecular approaches to identify the proteins involved in the Asc-glutathione cycle in Arabidopsis mitochondria. The defined set of proteins involved has dual activities in the intermembrane space and matrix. Further we show that many of these proteins are dual-targeted to chloroplasts and mitochondria, thus not representing a specific mitochondrial cycle from a molecular point of view. Rather the mitochondrial antioxidant defense cycle is linked to the chloroplast antioxidant defense cycle at the level of the nuclear genome.
In virtually all cases of authentically dual targeted proteins between mitochondria and chloroplasts, the chloroplast target is the stroma (44). Of the three organellar targeted APXs predicted in Table II, it was the stromal APX (At4g08390) that was found to be dual targeted to mitochondria and chloroplasts (Table III, Figs. 1 and 2). Orthologs of APX At4g08390 in tea, spinach and tobacco also have confirmed chloroplast stroma locations (45)(46)(47). A stromal location for GR and MDHAR activities, providing the necessary NAD(P)H for these reactions, is also well known in chloroplasts (2,5). So overall it is the chloroplast stromal antioxidant pathway that appears to be dual-targeted to mitochondria in Arabidopsis.
The in vivo subcellular localization (Fig. 2) only shows a mitochondrial localization for APX and MDHAR whereas from reports in the literature and the in vitro import assays ( Fig. 1) a plastid localization is also expected. This difference could simply be due to a technical limitation of the in vivo approach. A report of a sigma factor from maize that is located both in mitochondria and plastids indicated that GFP fusions could be detected in mitochondria or plastids but not both together in the same cell (48). Although no mechanistic reason is apparent from this the exclusive mitochondrial localization observed for APX and MDHAR in our hands may represent a similar phenomenon. Another possibility is that the plastid import path-way responsible for the import of these proteins is not very active in the cells used. Different import pathways have been described for plastid proteins (49), and developmental regulation is also observed (50). Alternatively, these targeting signals may not be able to support the import of GFP into plastids. Artificial passenger proteins have previously been reported to affect protein import into mitochondria and plastids (51). For plastid import it has been shown that regions in the mature protein can affect binding kinetics or stability (52). Thus the addition of the GFP moiety to these proteins may affect this interaction for plastid import but not for mitochondrial import. A previous report on Arabidopsis MDHAR also reported dual targeting but by means of multiple transcription initiation sites (7). However the data presented in this report is based on EST evidence for the different initiation sites and the authors do not conclusively demonstrate the presence of two independent transcripts in full-length cDNAs from Arabidopsis plants. Our own data on the same gene show that MDHAR proteins produced from both the initiator and the second methionine at position 8 can be imported into both mitochondria and plastids ( Fig. 1

and Supplementary Materials).
Role of the Cycle in Mitochondria and Functionality Cross Subcompartments-The probable role of this antioxidant cycle in plant mitochondria is complicated. All eukaryotic mitochondria appear to directly deal with superoxide generated by their electron transport chains through the use of superoxide dismutase (SOD), mostly notably the Mn-SOD type in the mitochondrial matrix. Superoxide needs to be dealt with directly at the site of action as its charge prevents easy diffusion across membranes and its rapid reaction retards the veracity of spatially separated detoxification systems. The product of the SOD reaction, H 2 O 2 , can readily diffuse from the site of production and has a substantially longer half-life in cells than superoxide. The potential for the cytosol, peroxisomes or even chloroplasts to facilitate scavenging of plant mitochondrial synthesized H 2 O 2 is therefore plausible. In yeast and mammalian systems, mitochondria are clearly the major source of active oxygen species (AOS), and it makes sense to deal with these radicals at the point of production. In yeast and mammals, cytochrome c peroxidase, thioredoxin peroxidase, and glutathione peroxidase play the dominant role (18 -20,53). In plants, chloroplasts and peroxisomes predominate in terms of both AOS formation and breakdown rates (2). The argument that plant mitochondria need their own defenses against H 2 O 2 has been presented in terms of the damage to other components by mitochondrial H 2 O 2 in non-photosynthetic organs such as root nodules (16,17), direct damage to mitochondrial enzymes by H 2 O 2 (21), prevention of lipid peroxidation-induced damage (54) and the requirements of mitochondria and chloroplasts for H 2 O 2 antioxidant defenses in response to environmental stresses such as rising NaCl concentrations (55). The proportion of self-formed H 2 O 2 and extra-mitochondrial H 2 O 2 that mitochondria need to deal with in vivo in mitochondria is entirely unknown. Thus in plants, a mitochondrial Asc-glutathione antioxidant defense against H 2 O 2 could be as much about protection from AOS generated by other organelles as about defense from self.
The cycle we present in Arabidopsis has very similar subcompartmentation characteristics as those described in pea and bean mitochondria (15,17). APX is located in the IMS in plant mitochondria, which is analogous to the location of cytochrome peroxidase in yeast (20,53). In plants this localization makes use of the production of Asc in the IMS by galactonogamma-lactone dehydrogenase that is also attached to the inner membrane (56), while the yeast counterpart utilizes reduced cytochrome c from the electron transport chain on the IMS side of the inner membrane. However, the MDHA and DHA that results from APX in plant mitochondria must move to the matrix for reduction back to Asc. The largely matrix location of MDHAR, DHAR, and GR are presumably required for the provision of reductant in the form of NADH and NADPH to fuel these reactions from the tricarboxylic acid cycle and/or photorespiratory glycine decarboxylation. Transporters for this function are not known in plant mitochondria to date, however, a diverse range of possible MDHA and DHA transport routes have been proposed in different plant compartments (2,5). As the latency of both DHAR and MDHAR were not as high as cytochrome c oxidase in both our studies and those in pea and bean mitochondria (15,17), an unidentified route for mitochondrial inner membrane transport of these compounds seems likely.
Dual Targeting between Plastids and Mitochondria in Plants and the Implications for Oxidative Stress Defense-Almost twenty proteins have been reported to be dual targeted using an ambiguous targeting signal directing a mitochondrial and plastidic location (44,51), and several other genes encode dual located proteins by means of alternative splicing or translation initiation defining an organelle specific targeting signal on the same protein (57)(58)(59). Almost half of these proteins play roles in DNA or RNA metabolism such as repair, binding, or aminoacyl-tRNA synthetases. The remainder fall into various groups such as biosynthetic pathways, phosphoribosyl aminoimidazole synthase and phosphatidylglycerophosphate synthase I (60,61), protein modification functions such as methionine aminopeptidase (62), and anti-oxidant activities such as the first dual-targeted protein GR from pea (9) and now our evidence of further Asc-glutathione cycle enzymes. This list is likely to expand with the identification of genes involved in various DNA functions clustered on chromosome 3 in Arabidopsis, many suggested to be dual-targeted (63). All these are common housekeeping functions required in mitochondria and plastids.
Thus it appears that many diverse functions may be linked by a single gene in the genome, which helps coordinate the function and biogenesis of mitochondria and plastids without the need for additional pathways to facilitate intra-organellar communication (64,65). So if any independent regulation needs to occur it must be via a different mechanism at the level of protein import or post-translational modification. We (66) and others (67,68) have previously discussed the need for signaling pathways to coordinate antioxidant defense in different cellular organelles in plants. But it appears that at least in the context of the mitochondrial and the chloroplast Asc-glutathione cycle, this coordination occurs by dual targeting rather than subtle retrograde signaling. Indeed in our stress induction experiments it appears transcripts respond to both mitochondria and chloroplast-localized stresses redundantly (Fig. 4). This makes most sense if mitochondrial H 2 O 2 defense is significantly against stray chloroplast H 2 O 2 rather than against moderate internal mitochondrial production, or at least that chloroplast-dependent mitochondrial functions, such as photo- FIG. 4. Real-time PCR analysis of transcript abundance changes in response to five AOS producing treatments. Two independent preparations of total RNA (1 g) from Arabidopsis tissue was reverse-transcribed, and each cDNA was assayed in duplicate using real-time quantitative PCR. Tissue was harvested 1 and 6 h following treatment. Treatments were light versus dark adapted (light), norflurazon (norf), menadione (mena), paraquat (para), and antimycin A (AA). Fold differences (ratio) between treatments and control at 1 h (white) or 6 h (black) are shown. For light treatment the fold difference is (light)/(dark adapted) as outlined in methods. Asterisks indicate significant changes (Student's t test, p Ͻ 0.05). Transcripts for APX (At4g08390), MDHAR (At1g63940), GR (At3g54660), DHAR (At1g19570), and GDXN (At3g15660) were assessed. respiration, are a primary requirement for a mitochondrial Asc-glutathione cycle in plants.