The impact of oxidative stress on Arabidopsis mitochondria


* For correspondence (fax +61 8 9830 1148; e-mail


Treatment of Arabidopsis cell culture for 16 h with H2O2, menadione or antimycin A induced an oxidative stress decreasing growth rate and increasing DCF fluorescence and lipid peroxidation products. Treated cells remained viable and maintained significant respiratory rates. Mitochondrial integrity was maintained, but accumulation of alternative oxidase and decreased abundance of lipoic acid-containing components during several of the treatments indicated oxidative stress. Analysis of the treatments was undertaken by IEF/SDS-PAGE, comparison of protein spot abundances and tandem mass spectrometry. A set of 25 protein spots increased >3-fold in H2O2/menadione treatments, a subset of these increased in antimycin A-treated samples. A set of 10 protein spots decreased significantly during stress treatments. A specific set of mitochondrial proteins were degraded by stress treatments. These damaged components included subunits of ATP synthase, complex I, succinyl CoA ligase, aconitase, and pyruvate and 2-oxoglutarate dehydrogenase complexes. Nine increased proteins represented products of different genes not found in control mitochondria. One is directly involved in antioxidant defense, a mitochondrial thioredoxin-dependent peroxidase, while another, a thioredoxin reductase-dependent protein disulphide isomerase, is required for protein disulfide redox homeostasis. Several others are generally considered to be extramitochondrial but are clearly present in a highly purified mitochondrial fraction used in this study and are known to play roles in stress response. Using H2O2 as a model stress, further work revealed that this treatment induced a protease activity in isolated mitochondria, putatively responsible for the degradation of oxidatively damaged mitochondrial proteins and that O2 consumption by mitochondria was significantly decreased by H2O2 treatment.


2′,7′-dichlorofluorescin diacetate


fluorescein diacetate




tricarboxylic acid


superoxide dismutase


reactive oxygen species


protein disulphide isomerase.


Aerobic organisms exploit the redox chemistry of oxygen to efficiently derive energy from oxidation of substrates. However, due to the tendency of molecular oxygen to gain single electrons and form reactive oxygen species (ROS), this energy production comes at a price. ROS, and the hydroxyl radical in particular, are highly reactive and can cause rapid and deleterious oxidation of biomolecules such as proteins, lipid and DNA. Plants are especially susceptible to oxidative damage, since the production of ROS is increased during stresses imposed by the environment such as extremes of temperature, suboptimal water availability and pollution (Smirnoff, 1998). Plants have, therefore, evolved a dynamic network of antioxidant defenses that serve to reduce the accumulation of ROS, to detoxify ROS, and to detoxify and repair oxidized molecules. The best characterized antioxidant defense network in plants is that of the chloroplast where the photosynthetic electron transport chain can generate large quantities of ROS (either singlet oxygen due to incomplete energy dissipation or superoxide by leakage of electrons at photosystem I). A Cu/Zn superoxide dismutase (SOD) reduces superoxide to hydrogen peroxide and a range of enzymes reduce hydrogen peroxide to water, including ascorbate peroxidase and 2-cys peroxiredoxin (Noctor and Foyer, 1998). Electrons for the latter reactions are supplied by NADPH via ascorbate, glutathione or thioredoxin.

The mitochondrial electron transport chain can also produce significant quantities of ROS, primarily due to the presence of the ubisemiquinone radical which can transfer a single electron to oxygen and produce superoxide (Halliwell and Gutteridge, 1989). The half life of ubisemiquinone increases if the electron transport chain is over-reduced and thus mechanisms that increase or maintain the flow of electrons out of the ubiquinone (Q) pool may reduce the production of ROS. Thus, the activity of the alternative oxidase (AOX) has been shown to reduce mitochondrial ROS accumulation (Maxwell et al., 1999), and a similar role has been proposed for the mitochondrial uncoupling proteins (Kowaltowski et al., 1998). However, despite this knowledge that plant mitochondria can produce ROS and contain mechanisms that may limit ROS accumulation, little is known about the antioxidant system of plant mitochondria nor the consequences for metabolism in this organelle if oxidative stress occurs (Millar et al., 2001a; Moller, 2001). A recent microarray experiment looking at the impact of antimycin A-induced oxidative stress on gene expression in Arabidopsis, identified cytochrome c and Aox as genes induced by this treatment, but no change was observed in the expression of nuclear genes encoding TCA cycle, other electron transport or ATP synthase components (Yu et al., 2001). Plant mitochondria do contain an Mn-type SOD, but it is not known how the H2O2 generated by this enzyme is metabolized (or dispersed). A mitochondrial-specific catalase isoform has been proposed in maize (Scandalios et al., 1980), however, no actual genes encoding such an enzyme have been documented. Nevertheless, it can be assumed that mechanisms do exist to prevent the accumulation of hydrogen peroxide, since it is known to be a potent inhibitor of aconitase (Verniquet et al., 1991) and would thus lead to disruption of the tricarboxylic acid cycle. In addition, in the presence of reduced transition metal ions such as Fe2+, the Fenton reaction can convert hydrogen peroxide to the hydroxyl radical. This radical can initiate oxidations that could have a major impact on the integrity and function of organelle lipids and proteins. For example, lipid peroxidation both impairs membrane function and also generates cytotoxic by-products such as 4-hydroxy nonenal that specifically inhibits decarboxylating dehydrogenase enzymes in the mitochondrial matrix (Millar and Leaver, 2000).

In this paper, we present the first systematic survey of the impact of oxidative stress on the protein composition and function of plant mitochondria. Using 2D gel electrophoresis in combination with MS/MS we were able to identify new, inducible components of the mitochondrial antioxidant system. In addition, we have shown that oxidative stress has a significant effect on the mitochondrial proteome leading to the degradation of a number of key proteins. Using a combination of this proteomic approach and functional studies of isolated mitochondria, it was possible to highlight major functional limitations that oxidative stress imposes on mitochondria. Plant mitochondria have been frequently overlooked both as a source of ROS and as a site of oxidative damage. Here, we have demonstrated that mitochondrial function is negatively affected by oxidative stress in a manner that will have implications across the cell, and that plant mitochondria utilize a number of antioxidant enzymes to scavenge and detoxify ROS to ameliorate this oxidative damage.


Induction of oxidative stress in Arabidopsis cell cultures.

The effect of oxidative stress on the mitochondrial proteome was investigated by treating 7-day-old Arabidopsis cell cultures with three different toxins: H2O2, menadione (a redox active quinone that generates intracellular superoxide), and antimycin A (an inhibitor of complex III of the mitochondrial respiratory chain). An initial screen revealed that the minimum concentrations of these compounds that induced significant and reproducible changes in the mitochondrial proteome were: H2O2, 88 mm; menadione, 400 μM; antimycin A, 25 μM (data not shown). Incubation of cells with these compounds for 16 h decreased cell growth rate, but did not lead to appreciable loss of cell viability as determined by FDA staining in combination with fluorescence microscopy. Further, the cells were still capable of sustained rates of respiration (Figure 1) and no DNA degradation or laddering was evident (data not shown) when DNA was analyzed using 2% (w/v) agarose-TAE gels according to McCabe et al. (1997). All three treatments led to increased DCF fluorescence in the first 30 min after application and all resulted in significant increases in the extent of lipid peroxidation, as determined by MDA content at 16 h (Figure 1).

Figure 1.

Induction of oxidative stress in Arabidopsis cell cultures.

Cell culture aliquots treated with 400 µm menadione, 88 mm H2O2 or 25 µM antimycin A were immediately assayed for 2′,7′-dichlorofluorescein (DCF) fluorescence and cell viability compared to controls. After 16 h of treatment, cell viability was re-analyzed and the rate of O2 consumption and dry weight determined. Cells frozen after 16 h were later assayed for MDA equivalents. Averages of at least three experiments (±SEM) are shown, all data shown are significantly different to controls for top three panels (P < 0.05).

It is notable that much lower concentrations of H2O2 and menadione than those used here, have been reported to induce programmed cell death in other plant cell cultures (Desikan et al., 1998; Houot et al., 2001; Tiwari et al., 2002). This discrepancy may be due to differences in cell density and culture vigor between the cell cultures used by different laboratories. In addition, Arabidopsis cell suspension cultures have been shown to be particularly efficient scavengers of ROS, leading to an extremely short half life for such molecules (Desikan et al., 1998). Although the concentration of extracellular H2O2 used here does not represent a physiological concentration, the action of highly efficient apoplastic peroxidases means that the concentration of intracellular H2O2 was probably much lower than that added to the culture medium (Amor et al., 2000). The amount perceived by the mitochondrion may well be lower still due to the action of cytosolic peroxidases and catalases.

Analysis of mitochondrial structural integrity and known oxidative stress responsive components

After treatment of cells as described above, mitochondria were isolated using a double-Percoll gradient method that has been demonstrated to yield highly purified mitochondria (Millar et al., 2001b). Each mitochondrial isolation was from five pooled 120 ml cell cultures. Each treatment was replicated at least twice using independently cultured batches of cells. Immunodectection of mitochondrial proteins (loaded on constant protein basis), revealed no major changes in the abundance of the outer membrane channel, VDAC, or the major matrix soluble protein HSP60, suggesting the outer membrane was present and matrix protein was not depleted. This suggests no change in the general structural integrity of the mitochondrial samples occurred from the different stress treatments (Figure 2). Alternative oxidase (AOX) abundance is commonly reported to rise in response to oxidative stress, most notably in the presence of cytochrome pathway inhibitors such as antimycin A (Vanlerberghe and McIntosh, 1997). AOX was slightly induced in antimycin A-treated cells but not in the other treatments, perhaps owing to the high endogenous amount of this protein present in control cells (Figure 2). Lipoic acid is a vital co-factor for TCA cycle function and a known site for oxidative stress-induced lipid peroxidation damage (Millar and Leaver, 2000). The abundance of lipoic acid co-factors on PDC and OGDC acyltransferase catalytic subunits was decreased by stress treatments (Figure 2).

Figure 2.

Immunodetection of marker protein abundance in mitochondrial samples from menadione, H2O2- and antimycin A-treated cells.

Mitochondrial proteins were separated by SDS-PAGE, transferred to nitrocellulose and probed with antibodies raised to the voltage-dependent anion channel (VDAC), mitochondrial heat shock protein 60 (HSP60), protein-bound lipoic acid (LA) and the mitochondrial alternative oxidase (AOX).

Analysis of the mitochondrial proteome following oxidative stress

Mitochondrial proteins were further analyzed by separation using 2D-gel electrophoresis, staining with colloidal Coomassie G-250 and imaging by flatbed scanning (Figure 3). Protein spots were detected in these images, quantified and matched across gels using Z3 software (Compugen, Tel Aviv, Israel). Default settings were used within the Z3 software with the exception of altering the minimum spot area to 300. Spot matching was done automatically and refined where necessary by the manual addition of ‘registration anchors’. Spot quantities (after normalization and background subtraction) were compared across the gels. A spot was considered to be significantly induced if its spot value relative to the matching spot in the control gel was >3 and significantly decreased if <0.3. This three-fold threshold represents changes in spot quantity that are significantly greater than the variation in spot quantity between samples. In pairwise comparisons of 2D gels of three independent samples of control mitochondria, 99% of the spots varied by less than this three-fold threshold (data not shown). The 1% of spots that did vary by >3-fold were a consequence of localized staining artefacts, inconsistent resolution of spots at the edges of the gel and incorrect spot detection (particularly inconsistent splitting of larger spots). Such spots were removed from our analysis by manual inspection of all spots that altered in abundance by >3-fold. Spots that were significantly altered in abundance are highlighted in Figure 3(a) which shows the mitochondrial proteome after H2O2 treatment. The same changes were apparent to varying degrees in H2O2 and menadione treatments, while only a subset of these changes were observed in antimycin A-treated cells.

Figure 3.

Changes in protein abundance in IEF-SDS-PAGE separated mitochondrial samples from menadione, H2O2- and antimycin A-treated cells. Proteins separated by isoelectric focussing on 3–10 non-linear immobilized pH gradients were subsequently separated by SDS-PAGE and arrays from different treatments compared by Z3 analysis software.

(a) An H2O2-treated sample overlaid with circles representing proteins increased in abundance (red and blue) or decreased in abundance (green) compared to a control and analyzed in Table 2.

(b–d) Enlargement of gel segments from control and representative gels of each of the three treatments showing particular examples of (b) decreases in known mitochondrial proteins (green) (c) increases in breakdown products of known mitochondrial proteins (blue) and (d) increases in novel mitochondrial proteins (red).

The spots circled in Figure 3(a) were excised from the gel, digested with trypsin and identified using tandem mass spectroscopy (Table 1). The thresholds shown in Table 1 are based upon spot quantities expressed relative to the matching spot in the control gels. Relative spot quantities are mean values of two replicates and values are shown for each of the three treatments. Protein identification was based upon collision-induced mass spectra for at least three tryptic peptides from each protein. Masses of collision-induced ion fragments were searched against the equivalent theoretical masses derived from the NCBInr protein database using Mascot ( Searches were scored using the MOWSE algorithm and given a probability value. In all cases, at least three peptides matched as the best hit to the protein entry noted. However, in some cases the best hit for individual peptides had a MOWSE score that gave a non-significant P-value (P > 0.05). The number of matching peptides with a significant MOWSE score (P < 0.05) is shown in Table 1. In several cases, highly similar gene copies for particular proteins (with 95–99% amino acid identity) could not be distinguished by the MS/MS data. These cases are noted in the legend of Table 1. We found nine proteins that were significantly increased in abundance in the H2O2 and menadione treatments, but not in the antimycin A treatment. In addition, 10 proteins were significantly decreased in abundance in at least two of the three treatments. We also identified 16 proteins that were increased in abundance in all treatments that were breakdown products of larger proteins. Examples of proteins from each of these categories (increased in abundance, decreased in abundance and breakdown products) are shown in enlarged gel sections for each treatment (Figure 3b,c,d, respectively).

Table 1.  Identification of 2D-separated protein spots from Arabidopsis mitochondria decreased or increased in abundance following treatment with menadione, H2O2 and antimycin A-treated cells. MS/MS spectra derived from trypsinated peptides of proteins were matched at MASCOT against a translated NCBI database
SpotNo. MP
(P < 0.05)
AGI accessionTPPRPutative IDH2O2Menanti A
  • Table headings: No. MP, number of peptides matching to predicted protein sequence (P < 0.05); GEL MM/pI, observed molecular mass (MM) and pI of the sample from the gel in Figure 3(a); MATCH MM/pI predicted molecular mass (MM) and pI of matched sequence; TP, predicted localization of sequence by TargetP, M1–5 = mitochondrial, C1–5 = chloroplast, S1–5 = secretory pathway, O1–5 = other (1 = high probability, 5 = low probability); PR-predicted localization of sequence by Predotar, mitochondrial (M), plastidic (P) or other (N). Z3 analysis of gels from different treatments identified fold changes compared to control, a cut off of three-fold induction (>3) or three-fold decrease (<0.3) was used.

  • a

    Spots 4 and 5 match At1g13440 and At3g04120 equally. Due to differences in MM/pI on gels the two gene copies were assigned as indicated.

  • b

    The ATP synthase beta subunit copies At5g08660, At5g08670 and At5g08680 could not be differentiated by MS/MS spectra obtained (spots 13,14 and 37).

  • c

    Spots 19 and 20 could be succinyl CoA ligase alpha subunit At5g08300 or At5g23250.

  • d Identification of spots 36 and 37 was based on peptide mass fingerprinting analysis (PMF) previously published in Millar et al. (2001b).

Increased in abundance
 1392000/4.066485/4.48At5g60640S1Nprotein disulphide isomerase>3>3
 7318000/5.715319/5.94At5g59880O2PActin depolymerising factor 3>3>2
 9215000/6.811696/6.41At2g21640O5NUnknown protein FtsX like>3
 10322000/6.454971/6.23ATP1O4NATP synthase alpha subunit>3>3>3
 11329000/5.054971/6.23ATP1O4NATP synthase alpha subunit>3>3>3
 12331000/5.354971/6.23ATP1O4NATP synthase alpha subunit>3>3>3
 13214000/7.059671/6.18At5g08670bM4MATP synthase beta subunit>3>3>3
 14115000/6.959671/6.18At5g08670bM4MATP synthase beta subunit>3>3
 15216000/8.627597/6.27At2g21870M1MATP synthase 27 kDa subunit (ATP7)>3>3>2
 16227000/7.568863/5.38At3g52200M1MPyruvate dehydrogenase E2 subunit> 2>3>3
 18421000/5.136992/8.54At1g53240M4MNAD malate dehydrogenase>3>3>3
 19113000/6.536152/8.55At5g08300cC5MSuccinyl CoA ligase alpha subunit>3>3>3
 20225000/9.236152/8.55At5g08300cC5MSuccinyl CoA ligase alpha subunit>3>3>3
 21234000/4.945346/6.30At2g20420M2PSuccinyl CoA ligase beta subunit>3>3>2
 24313000/6.225735/9.28At4g11010C4MNucleoside diphosphate kinase III> 2>2>3
 25327000/7.529211/7.84At5g15090O3NVoltage-dependent anion channel (porin)>3>3
Decreased in abundance
 26392000/6.498093/5.79At2g05710O2NAconitate hydratase<0.3
 27380000/6.181130/6.24At5g37510M3MNADH-ubiquinone oxidoreductase 75 kDa subunit<0.3<0.3<0.3
 28355000/5.765885/8.83At2g14170M4MMethylmalonate semialdehyde dehydrogenase<0.3<0.3<0.3
 30234000/6.735187/8.07At3g22200O2NGABA aminotransferase<0.3<0.3
 31148000/6.850134/9.19At5g55070M4M2-oxoglutarate dehydrogenase E2 subunit<0.3<0.3<0.3
 33126000/6.226490/6.87At4g25100M2NSuperoxide dismutase (Fe)<0.3<0.3
 34118000/7.925718/9.28At4g11010C4MNucleotide diphosphate kinase 3<0.3>0.3<0.3
 35329000/6.228388/8.03At4g02580M2MNADH dehydrogenase 24 kDa subunit<0.3>0.3<0.3
 36PMFd53000/6.454971/6.23ATP1O4NATP synthase alpha subunit<0.5
 37PMFd54000/5.359671/6.18At5g08670bM4MATP synthase beta subunit<0.5<0.5<0.5

Identification of proteins involved in oxidative and stress responses

Of the nine induced proteins that were present in mitochondrial proteomes from both the H2O2- and the menadione-treated cells, two represented novel putative mitochondrial antioxidant proteins. One, At3g06050 is listed in the TAIR database as ‘unknown protein’ and has previously been observed in the Arabidopsis mitochondrial proteome (Kruft et al., 2001). The amino acid sequence of this protein is homologous to a number of bacterial peroxiredoxins and it seems likely that it participates with the mitochondrial thioredoxin system to reduce hydrogen peroxide to water (Laloi et al., 2001). The other protein is a disulphide isomerase (At5g60640), a protein family that is classically located in the ER and is involved in protein folding and thiol homeostatis (Freedman et al., 1994). Other proteins induced in mitochondria after oxidative stress (after H2O2 and menadione, but not antimycin A treatments) included calreticulin, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and a glutathione S-transferase (GST1). These proteins are generally considered to be non-mitochondrial and none contain known mitochondrial targeting sequences, but all three are known to be induced by stress conditions in other organisms (see Discussion). Interestingly, the mitochondrial Mn-superoxide dismutase, identified on 2D gels in our previous study (Millar et al., 2001b), did not change in abundance during any stress treatments of cells (data not shown).

Oxidative stress is associated with significant decreases in abundance of a number of key mitochondrial proteins

A number of mitochondrial proteins were significantly decreased in abundance following oxidative stress (Table 1). The effect was most pronounced after treatment with H2O2, with several proteins reduced below the limit of detection. Although the effect of menadione was less pronounced, the majority of proteins highlighted in Table 1 were reduced below the threshold of 0.3 (spot value relative to the control). We found that several proteins of the tricarboxylic acid (TCA) cycle were significantly reduced, including aconitase, the E2 subunit of the 2-oxoglutarate dehydrogenase complex, fumarase and succinyl CoA ligase. In addition, two subunits of complex I were significantly reduced in abundance, as was the β subunit of the ATP synthase complex. These decreases in protein abundances are likely to impose significant flux restrictions on the tricarboxylic acid cycle and electron transport, thereby limiting the synthesis of ATP. A variety of other proteins representing a range of metabolic activities were also decreased in abundance. These included methyl malonate semialdehyde dehydrogenase (MMSD), an aldehyde dehydrogenase involved in valine and pyrimidine metabolism, a GABA transaminase which participates in the GABA shunt around the TCA cycle (Busch and Fromm, 1999), and nucleoside diphosphate kinase III which is located in the mitochondrial intermembrane space and functions to equilibrate nucleoside triphosphate pools (Sweetlove et al., 2001).

A specific set of proteins is degraded following all three oxidative stresses

A number of proteins were increased in abundance in all three treatments that on the basis of their mobility in the second dimension of gel electrophoresis, were of a molecular mass considerably less than the predicted molecular mass of the mature gene product. These proteins presumably represent fragments of proteins generated as a consequence of protein degradation. The identity of these degradation products is consistent in a number of cases with a decrease in abundance of the respective parent proteins (Table 1). For example, we found degradation products of both the α and β subunits of the ATP synthase complex, aconitase, succinyl CoA ligase and nucleoside diphosphate kinase all of which showed a decrease in abundance of the respective parent protein. The absence of breakdown products derived from some of the proteins shown to decrease in Table 1 presumably reflect their more complete degradation to small peptides or the low abundance of intermediate breakdown products which did not allow their detection in this analysis. The fact that most of the proteins in control mitochondria did not change during the stress treatments (data not shown), suggests that a specific set of proteins are degraded, presumably those that incur oxidative damage.

Induction of a protease activity in mitochondria by oxidative stress

In order to investigate whether a specific protease might be responsible for protein degradation of putatively damaged components, we assayed protease activity (using resorufin-labeled casein as a substrate) in mitochondria isolated from control cells and from those treated with 88 mm H2O2 for 16 h (Figure 4). The incubation medium contained 5 mm CaCl2 and 5 mm MgCl2 as well as 0.1% (v/v) Triton X-100 to ensure lysis of mitochondria. Under these conditions, there was no significant difference in protease activity in mitochondria from control and H2O2-treated cells (t-test, P < 0.05). However, the addition of 5 mm ATP increased mitochondrial protease activity in the H2O2-treated cells to approximately two-fold higher than in the control cells. EDTA abolished all protease activity by chelating Mg2+ and thereby preventing formation of Mg-ATP required for ATP-dependent protease activity. The complete inhibition of protease activity by EDTA suggests that the residual ATP-independent protease activity is due to the activity of a metallo-protease. The induced ATP-dependent protease activity was decreased by inhibitors of serine proteases (AEBSF) but not aspartate proteases (pepstatin) or cysteine proteases (E-64). No protease activity could be detected using this assay in IEF sample buffer-containing urea, thiourea and CHAPS, ruling out the possibility that enhanced proteolysis occurred during IEF separation of mitochondrial proteins samples derived from stress treatments (data not shown).

Figure 4.

Protease activity in mitochondria isolated from control cells, and cells treated with H2O2. Isolated mitochondria from control cells (open bars), or cells treated with 88 mm H2O2 (filled bars), were ruptured with Triton-X100 and incubated with resorufin-labeled casein at 25°C for 16 h. Proteolytic degradation of casein released resorufin which was detected by spectrophotometry. ATP and protease inhibitors (in the presence of ATP) were added as indicated.

The impact of oxidative stress on whole chain respiratory function of mitochondria

The degradation and decrease in abundance of a number of protein subunits of complexes of the TCA cycle and the electron transport chain suggests that oxidative stress may significantly restrict the respiratory capacity of mitochondria. This was investigated by isolating mitochondria from control cells and cells that had been treated with 88 mm H2O2 for 16 h and comparing respiratory oxygen consumption (Table 2).

Table 2.  O2 consumption by Arabidopsis mitochondria isolated from control cells and cells treated with H2O2
SubstrateMitochondria from control cells
(nmol O2 min−1 mg−1 protein)
Mitochondria from H2O2-treated cells
(nmol O2 min−1 mg−1 protein)
  1. Respiratory assays were performed according to materials and methods. Means ± sem (n = 3) for O2 consumption rates are presented, outer membrane integrity is the percentage of mitochondria in the sample with intact outer membranes.

NADH350 ± 7157 ± 2360 ± 2 (17)3 ± 1211 ± 2193 ± 3742 ± 14 (20)2 ± 1
Succinate362 ± 7165 ± 2367 ± 19 (19)3 ± 1198 ± 9128 ± 1348 ± 4 (24)3 ± 1
NADH + succinate473 ± 13224 ± 1089 ± 8 (19)2 ± 2236 ± 10119 ± 1045 ± 2 (19)3 ± 2
Malate + pyruvate167 ± 1263 ± 1127 ± 5 (16)1 ± 182 ± 540 ± 618 ± 4 (22)1 ± 1
Cytochrome c263 ± 21   264 ± 34   
OM integrity (%)98 ± 1   98 ± 1   

Outer membrane integrity of both control and oxidatively stressed mitochondria was greater than 98% and respiration was coupled to ATP formation. Using either NADH, succinate or malate and pyruvate as respiratory substrates, it was found that the respiratory rates in the presence of ADP were significantly lower from mitochondria from H2O2-treated cells than those from control cells. The respiratory rate with exogenous cytochrome c as a substrate was not significantly different (t-test, P > 0.05), indicating that the inhibition by stress occurred before the cytochrome c oxidase step of the respiratory chain. Although the rate of KCN-resistant respiration was reduced in mitochondria from H2O2-treated cells, when expressed as a percentage of the respiratory rate prior to the addition of KCN, there was no change. This KCN-resistant respiration was almost completely inhibited by the alternative oxidase inhibitor, n-propyl gallate. These results indicate that oxidative stress reduces the respiratory capacity primarily by inhibiting electron flow through the respiratory dehydrogenases into the Q pool and by reducing the activity of the TCA cycle rather than affecting the activity of respiratory oxidases.


In order to investigate the effects of sub-lethal oxidative stress on mitochondrial function, we have identified concentrations of H2O2, menadione or antimycin A that when added to Arabidopsis cell suspension cultures caused increased fluorescence of DCF and an increase in lipid peroxidation but did not kill the cells. In comparison to control cells, cell viability did not significantly decrease and cells continued to respire at high rates, but growth rate was slowed. Following the treatment of Arabidopsis cells with sub-lethal doses of oxidative stress it was possible to observe significant changes in the mitochondrial proteome. Protein identifications revealed these changes were caused either by the increase in abundance of mitochondrial-targeted proteins that have an antioxidant or stress-response function, or the selective association of other cell proteins with mitochondria which have potential roles in stress response or the degradation of specific mitochondrial proteins, putatively those that were oxidatively damaged.

Identification of novel antioxidant defense proteins in plant mitochondria

Two proteins both of which function in the thioredoxin-based redox pathway and which may form part of an antioxidant defense/post-oxidation recovery system in plant mitochondria were enhanced by stress treatments. The first of these proteins belongs to the peroxiredoxin family of thioredoxin-dependent peroxidases. This protein carries a mitochondrial targeting presequence and has previously been observed in the Arabidopsis mitochondrial proteome (Kruft et al., 2001). However, this protein is annotated by homology only and its function remains unproven. Peroxiredoxins form a ubiquitous group of peroxidases found in bacteria (Tartaglia et al., 1990), yeast (Chae et al., 1994), animals (Kim et al., 1988) and higher plants (Baier and Dietz, 1996, 1997) and can be classified according to the number of conserved cysteine residues; hence 2-cys peroxiredoxin and 1-cys peroxiredoxin. Alignment of the At3g06050 amino acid sequence against other Arabidopsis peroxiredoxins reveals that it is a member of the 1-cys peroxiredoxin group (data not shown). While the function of 2-cys peroxiredoxins as enzymes that accept electrons from thioredoxin to reduce alkyl hydroperoxides and hydrogen peroxide is well established (Chae et al., 1994), the function of the 1-cys peroxiredoxins is less clear. Indeed, the electron donor of this group remains a source of controversy. In plants, 1-cys peroxiredoxin has been suggested to play a role in the maintenance of seed dormancy, although overexpression of a rice 1-cys peroxiredoxin did not confirm this (Lee et al., 2000). Recently, it has been shown that some members of the 1-cys group are functionally related to 2-cys peroxiredoxins in that they act as thioredoxin-dependent peroxidases (Pedrajas et al., 2000; Verdoucq et al., 1999). These include a yeast 1-cys peroxiredoxin in the mitochondrion and the Arabidopsis AtTPX2. Comparison of the amino sequence of At3g06050 with all Arabidopsis peroxiredoxins, reveals that it shares the greatest homology with AtTPX2. In addition, a search for conserved domains (pFam) reveals that At3g06050 contains a conserved alkyl hydroperoxide reductase/thiol-specific antioxidant domain present in all peroxiredoxins. Thus, it is highly likely that At3g06050 encodes for a thioredoxin-dependent peroxidase (peroxiredoxin) and as is the case in yeast, the Arabidopsis mitochondrial peroxiredoxin is a 1-cys peroxiredoxin. The identification of this peroxiredoxin completes the mitochondrial thioredoxin system in plants and provides a mechanism by which plant mitochondria can metabolize H2O2 (Laloi et al., 2001). We did not see inductions of the other proteins of the thioredoxin system (thioredoxin and thioredoxin reductase). However, thioredoxin may run off the bottom of our 2D gels and the thioredoxin reductase has been shown to be resistant to identification by mass spectrometry (Rabilloud et al., 2001).

The second protein induced belongs to the protein disulphide isomerase (PDI) family. These proteins are best known for their role as chaperones in protein-folding reactions in the endoplasmic reticulum (Freedman et al., 1994). However, mitochondrial isoforms have been documented in mammals (Rigobello et al., 2000, 2001) and a chloroplast isoform has been identified in Chlamydomonas (Trebitsh et al., 2001). These proteins have ER retention signals at their carboxyl terminus, but purification and compartmentation studies in the case of the mitochondrial isoforms and in vitro import assays in the case of the chloroplast isoform showed that they are still specifically imported to these organelles. PDIs contain two thioredoxin like domains which cycle between dithiol and disulphide oxidation states allowing the formation, reduction or rearrangement of disulphides in protein substrates (Gilbert, 1997). PDI is reduced by thioredoxin reductase, and may act in mitochondria in several ways: to remove abberant disulphides introduced by oxidative conditions; to enhance protein folding of newly synthesized proteins that replace damaged counterparts; or directly act to reduce disulphides required to activate proteins involved in antioxidant defense (Rigobello et al., 2000, 2001). One such protein is AOX, which exists in a reduced active form and disulphide bridged inactive form (Vanlerberghe and McIntosh, 1997). To date, no mechanism for AOX redox regulation has been identified, but it is possible that the mitochondrial PDI identified here may be involved. Notably, of the seven disulphide isomerase genes identified in Arabidopsis genome sequencing, that identified here (At5g60640) is a minor isomer based on representation in EST databases (TIGR, The Institute for Genomic Research) with only 8 ESTs, while the dominant ER form (At1g21750) is represented by 55 ESTs.

Mobilization of known stress responsive proteins to the mitochondrial fraction

Three proteins were found in the mitochondrial 2D gel profile of H2O2 and menadione treatments but are not generally considered to be mitochondrial proteins based on targeting prediction studies and the available literature. All three proteins – GST1, GAPDH and calreticulin – have clear links to stress response from studies in other eukaryotes.

A number of GSTs have been shown to detoxify lipid peroxidation products (Gronwald and Plaisance, 1998; Hayes and Pulford, 1995) some acting within the mitochondria in mammals (Gardner and Gallagher, 2001). Expression of GST1 transcripts in Arabidopsis is induced by oxidative-stress (Kliebenstein et al., 1999; Reuber et al., 1998) and elevated mRNA levels of this GST are widely used as early markers for biotic and abiotic stress in this model plant (Grant et al., 2000; Rate and Greenberg, 2001). There is an obvious role for a mitochondrial GST for detoxification of lipid peroxidation products in this organelle. Very recently, the import of the major cytosolic GSTA4-4 from mouse into mitochondria following oxidative stress has been reported and a role for this enzyme in removal of lipid peroxidation products has been established (Raza et al., 2002).

The glycolytic enzyme glyceraldehyde-3-phosphate dehydrogenase, GAPDH, has a number of well characterized secondary roles in mammals such as a cellular protein kinase, facilitator of membrane fusion, an mRNA binding protein and following oxidative damage as a DNA repair enzyme with uracil DNA glycosylase activity (Meyer-Siegler et al., 1991). GAPDH has recently been shown to be translocated from the cytosol to the nucleus following oxidative stress where it is thought to act in one of its secondary functions (Dastoor and Dreyer, 2001). Our data might suggest a similar translocation to or association with mitochondria, perhaps also to allow a DNA repair or signalling role.

Finally, calreticulin has been implicated in mammalian systems in the Ca2+ homeostasis of mitochondria through Ca2+ loading of these organelles from junctions with the endoplasmic reticulum specifically rich in this Ca2+ binding protein (Simpson et al., 1997). The select association of mitochondria with these ER junctions during oxidative stress may facilitate changes in Ca2+ levels within mitochondria under these conditions. Changes in the Ca2+ homeostasis of mitochondria are critical during stress response in mammalian systems and can trigger the permeability transition pore and thereby engage PCD pathways (Gunther and Pfeiffer, 1990; Pfeiffer et al., 2001). The select association of mitochondria with these ER-junctions during oxidative stress may point to the dynamics of changing Ca2+ levels under these conditions within mitochondria.

Given the high abundance of these three proteins in cellular compartments other than mitochondria, it is possible that they simply adhered to the mitochondria during isolation and, therefore, are contaminants. However, other cytosolic markers are absent from our preparations of mitochondria (Millar et al., 2001b) and no such association was seen with the control mitochondria. Nonetheless, functional association with mitochondria will require further experimentation.

The differential effect of three oxidative stress treatments

It was very apparent from the 2D proteomics studies and Western blots of stress protein induction that H2O2 and menadione revealed a very similar response while antimycin A resulted in an overlapping but distinct set of responses. In this context, it is important to consider the differences between these stress treatments. H2O2 and menadione treatments were both non-specific oxidative stresses that have the potential to act throughout the cell, they do not directly have a specific impact on a single site in metabolism or biosynthesis. Protection from these forms of damage likely requires removal of oxidants and repair of damage, that is, antioxidant and post-oxidant defense. So called pre-oxidative defense mechanisms such as AOX (Millar et al., 2001a; Moller, 2001), which lower the redox poise of reactive proteins, for example, would be of little use and are notably not induced in response to these chemical stresses. Antimycin A, by contrast, directly targets mitochondrial electron transport resulting in elevated ubiquinol levels and ROS formation in the inner mitochondrial membrane. Blocking the cytochrome pathway also has the effect of greatly decreasing ATP formation thus debilitating energy requiring biosynthesis and uncoupling glycolytic rate from adenylate control. Protection from this form of oxidative stress could be gained in part by pre-oxidative defenses such as AOX which decreases Q pool reduction levels and re-initiates some level of ATP synthesis to drive biosynthesis. At the proteome level, antimycin A only lead to breakdown products of existing mitochondrial proteins and not to the induction of novel components including peroxiredoxin, protein disulplhide isomerase, calreticulin, GST1 and GAPDH. One simple explanation for this might be that lowered ATP availability severely restricted protein synthesis in antimycin A treatments. It is worth noting, that with the exception of AOX, few transcripts encoding mitochondrial proteins were found to be induced by antimycin A in tobacoo or Arabidopsis (Maxwell et al., 2002; Yu et al., 2001). Thus, while the intramitochondrial stress of antimycin A is triggering a response that induces AOX and potentially other protein synthesis events directly connected to the re-initiation of electron transport chain function, the more general stress of H2O2 and menadione trigger a more general response to oxidative stress across the whole cell which includes defenses located in mitochondria.

Only certain proteins in the mitochondrion are degraded in response to oxidative stress

A number of TCA cycle enzymes such as aconitase, pyruvate dehydrogenase complex (PDC) and 2-oxoglutarate dehydrogenase complex (OGDC) have been shown to be sensitive to oxidative damage either directly by H2O2 or indirectly through lipid peroxidation products (Millar and Leaver, 2000; Verniquet et al., 1991). Our immunoblots revealed that lipoic acid moieties are damaged primarily by menadione and antimycin A stress treatments that would produce superoxide capable of initiating lipid peroxidation (Figure 2). In 2D gel analysis, it was evident that a PDC lipoic acid-containing component was lost from the profiles and a breakdown product of the lipoic acid-containing OGDC subunit was detected during some oxidative stresses (Table 1). Further, two subunits of complex I were significantly lost during oxidative stress, notably these two subunits of complex I are known to contain motifs for binding Fe-S clusters (Rasmusson et al., 1998). Aconitase, another Fe-S-containing enzyme was specifically lost following H2O2 treatment, the severe sensitivity of this enzyme has been documented previously in the purified enzyme from potato mitochondria (Verniquet et al., 1991). Such losses in function will reduce NAD-substrate-dependent respiratory function as was evident during malate-dependent respiratory rate in Table 2. Similar declines in abundance were also apparent for less well-known mitochondrial proteins. The methyl malonate semialdehyde dehydrogenase (MMSD) was previously identified by the authors as a mitochondrial protein in plants (Millar et al., 2001b) and has a known mammalian mitochondrial ortholog (GenBank L32961.1). The GABA transaminase is the missing link in a GABA shunt proposed by the authors previously (Millar et al., 2001b) where both the glutamate dehydrogenase and succinic semialdehyde dehydrogenase were previously identified. The type III nucleoside diphosphate kinase was previously identified in Arabidopsis by the authors and others (Kruft et al., 2001; Sweetlove et al., 2001) and an ortholog has been shown to bind to an unknown protein during heat stress in pea (Escobar Galvis et al., 2001). This is the first evidence of changes in the stability/abundance of these proteins during stress conditions (Table 1).

The electron transport chain and ATP synthase were also affected by oxidative stress treatments. The breakdown in ATP synthase subunits, in particular, is evident in the breakdown products list of Table 1, but this probably reflects the great abundance of these subunits rather than any particular sensitivity of ATP synthase relative to other enzymes during oxidative stress. ATP synthase α subunit was decreased during one stress, while ATP synthase β subunit was significantly decreased in all three stress treatments. Clearly these changes will have a significantly affect on oxidative phosphorylation capacity. The results suggest that sensitive components of the TCA cycle, together with the Fe-S centers of complex I and ATP synthesis subunits, are likely to be the greatest casualties of oxidative stress, since specific respiratory oxidase activities were unaffected by H2O2 addition (Table 2).

Part of the recovery process following oxidative stress will clearly be the breakdown and removal of damaged proteins so that they can be replaced de novo. We assume that this is the case with the proteins discussed above and that their degradation was catalysed by the ATP-dependent protease induced by the stress treatment. This suggests that plant mitochondria possess a stress-responsive protease specifically for this purpose. Such a protease system has been identified previously in mammalian mitochondria and shown to specially degrade oxidatively denatured proteins (Marcillat et al., 1988).

New insights into mitochondrial response of oxidative stress

By investigating the effect of oxidative stress on mitochondrial function and protein composition in a relatively global manner in a model plant system we have highlighted new molecular mechanisms involved in mitochondrial defense and catalogued degradations of sensitive protein components. Further, we have provided intriguing data pointing to differences between stress treatments that have implications for the nature of the integration of stress responses between plant cell compartments.

Materials and methods

Maintenance of cell culture

A heterotrophic Arabidopsis thaliana cell culture, established from callus of ecotype Lansberg erecta stem explants, has been maintained for over 9 years by weekly subculture. Media used for this cell culture was Murashige and Skoog basal media supplemented with 3% (w/v) sucrose, 0.5 mg L−1 naphthaleneacetic acid and 0.05 mg L−1 kinetin (May and Leaver, 1993). The cell cultures were maintained in 250 ml conical flasks in the dark at 22°C in an orbital shaker (150 g). At 6–7 days, each flask (120 ml of cell culture) contained 8–10 g FW cells and growth was approximately in the middle of the log phase. Subculture of 20 ml of culture to 100 ml of fresh media initiated the cycle again.

Cell growth, oxidative stress and cell viability measurements

Cell growth rate was determined from cell dry weights obtained by vacuum filtration of 10 ml aliquots of cultured cells onto 0.8 μm nitrocellulose filters, heating at 60°C for 24 h followed by mass determination. Oxidative stress was detected using 2′,7′-dichlorofluorescin-diacetate (DCF-DA) or by measuring the accumulation of malondialdehyde equivalents. DCF-DA accumulates in cells, where the action of intracellular esterases cleave the diacetate group. The resulting non-fluorescent 2′,7′-dichlorofluorescin is converted to the fluorescent 2′,7′-dichlorofluorescein by oxidation. DCF-DA (0.2 µm) was added to cells diluted in 3 ml of fresh MS media in a stirred cuvette and fluorescence (excitation, 488 nm; emission 525 nm) intensity recorded at 5, 15 and 30 min. These data were used to calculate relative rates of change for comparison between samples and experiments. The thiobarbituric acid reactive substances (TBARS) assay of Hodges et al. (1999) was used to assess the amount of malondialdehyde (MDA) equivalents. Cell viability was determined as described by Swidzinski et al. (2002). Living cells, stained with the vital stain, fluorescein diacetate (F-DA) at a final concentration of 0.002% (w/v) and dead, non-fluorescent cells were counted using fluorescence microscopy. No distinction was made between necrosis and programmed cell death. A minimum of 500 cells were counted for each treatment which were performed in triplicate. Results are presented as means ± SEM.

Mitochondrial isolation

A total of 1.0–1.2 L of 7-day-old dark grown cell suspension culture was filtered through gauze to remove media and then the cells disrupted in a Waring blender by three successive 15 s bursts. Disruption of 60 g of cells was performed in 200 ml of grinding medium (0.45 m mannitol, 50 mm sodium pyrophosphate, 0.5% (w/v) bovine serum albumin, 0.5% (w/v) PVP-40, 2 mm EGTA, 20 mm cysteine pH 8.0). Filtered cell extract was separated by differential centrifugation and mitochondria purified from the resultant organelle pellets on two Percoll gradients according to Millar et al. (2001b).

Respiratory measurements

O2 consumption by whole cells and isolated mitochondria was measured in a Clark-type O2 electrode in 1 ml volume. For mitochondria, the reaction medium contained: 0.3 m mannitol, 10 mm TES-KOH pH 7.5, 5 mm KH2PO4, 10 mm NaCl, 2 mm MgSO4 and 0.1% (w/v) bovine serum albumin; for whole cells the supplemented MS media was used (as above). In mitochondrial experiments, pyruvate (5 mm), malate (0.5 mm), succinate (10 mm), NADH (1 mm), ADP (0.5 mm), KCN (0.5 mm) and nPG (0.05 mm) were added as indicated to modulate O2 consumption rates of mitochondria. ATP (0.5 mm) was added to ensure full activation of succinate dehydrogenase. Cytochrome c oxidase activity was measured as ascorbate (5 mm), cytochrome c (25 µm)-dependent O2 consumption in the presence of 0.05% (w/v) Triton X-100. Outer mitochondrial membrane integrity was assayed as the latency of cytochrome c oxidase activity (Neuberger, 1985).

Protease measurements

Protease activity was assayed by determining the rate of degradation of resorufin-labeled casein (Roche Diagnostics Ltd, Lewes, UK). Isolated mitochondria were incubated in a volume of 200 µl with 0.2 m Tris–HCl, 10 mm MgCl2, 10 mm CaCl2, 50 mm ATP, 1% (w/v) resorufin-labeled casein, pH 7.5 at 25°C for 16 h. The reaction was terminated, and undigested casein precipitated, by the addition of 480 µl of 5% (w/v) trichloroacetic acid. After centrifugation at 20 000 g for 5 min, 400 µl of the supernatant was added to 600 µl of 0.5 m Tris–HCl, pH 8.8 and the absorbance at 574 nm measured. Protease inhibitors (5 mm EDTA, 1 mm AEBSF (4-(2-aminoethyl)-benzenesulfonyl fluoride hydrochloride), 1 μm pepstatin and 28 μm E-64 (N-[N-(l-3-trans-carboxirane-2-carbonyl)-l-leucyl]-agmatine) were added as indicated. All protease inhibitors were supplied by Roche Diagnostics.

One-dimensional electrophoresis and immunodetection

One dimensional SDS-PAGE was performed according to standard protocols using 12% (w/v) polyacrylamide 0.1% (w/v) SDS gels. For immunodectection, separated proteins were transferred onto a nitrocellulose membrane and incubated with primary antibodies raised to VDAC, HSP60, alternative oxidase or lipoic acid. Antibodies were from the following sources: alternative oxidase, VDAC and HSP60 (Dr Tom Elthon, University of Nebraska, Lincoln) and lipoic acid (Dr K.M. Humphries and Prof P.I. Szweda, Case Western Reserve University, Cleveland). A chemiluminescence detection system linked to horseradish peroxidase was used as a secondary antibody, and quantitative light emission was recorded using a Luminescent Image Analyser (LAS 100, Fuji, Tokyo).

Two-dimensional gel electrophoresis

For IEF-SDS/PAGE, mitochondrial protein samples (800 µg) were acetone-extracted by addition of acetone to a final concentration of 80% (v/v) at −80°C, samples stored at −20°C for 4 h and then centrifuged at 20 000 g for 15 min The pellets were re-suspended in an IEF sample buffer consisting of: 6 m urea, 2 m thiourea, 2% (w/v) CHAPS, 2% (v/v) ampholytes (pH 3–10), 2 mm tributylphosphine and 0.001% (w/v) bromophenol blue. Aliquots of 330 µl were used to re-swell dried 180 mm, pH 3–10 non-linear IPG strips (Immobiline DryStrips, APBiotech, Sydney) overnight and then IEF was performed for 19.5 h reaching a total of 49 KVh at 20°C on a flat-bed electrophoresis unit (Multiphor II, APBiotech, Sydney). IPG strips were then transferred to an equilibration buffer consisting of 50 mm Tris–HCl (pH 6.8), 4 m urea, 2% (w/v) SDS, 0.001% (w/v) Bromophenol blue and 100 mmβ-mecaptoethanol and incubated for 20 min at room temperature with rocking. The equilibrated strips were then slotted into central single wells of 4% acrylamide stacking gels above 0.1 cm × 18.5 cm × 20 cm, 12% (w/v) acrylamide, 0.1% (w/v) SDS-polyacrylamide gels. Strips were overlaid with 0.5% (w/v) agarose in SDS-PAGE running buffer. Gel electrophoresis was performed at 25 mA per gel with circulating cooling (4°C) and completed in 5 h. Proteins were visualized by colloidal Coomassie (G250) staining. Gels were placed in a solution of 17% (w/v) ammonium sulfate, 34% (v/v) methanol, 3% (v/v) phosphoric acid, 0.1% (w/v) Coomassie brilliant blue G250 for 16 h and de-stained in 0.5% (v/v) phosphoric acid for 24 h. The sensitivity and linearity of this staining procedure have been extensively documented over the protein concentration range used in this study (Berggren et al., 2000; Consoli and Damerval, 2001; Neuhoff et al., 1988). Stained gels were scanned using a APBiotech (Sydney) ImageScanner in transparency mode. The optics of this scanner has been optimized and linearized above 3.4 OD for 2D gel analysis. MW and pI standards from APBiotech (Sydney) were used to confirm fixed pH gradient positioning on first dimension separations and to identify apparent molecular masses on the second dimension separation.

Quadrupole time-of-flight mass spectrometry (Q-TOF MS)

Q-TOF MS/MS was performed on an Applied Biosystems Q-STAR Pulsar (Q-TOF MS) using an IonSpray source. Protein spots to be analyzed were cut from the 2D PAGE gel, dried at 50°C in a dry block heater and stored at −70°C. For the sequencing analysis, the proteins were digested with trypsin according to Sweetlove et al., (2001), injected into the electrospray source in 50% (v/v) methanol/0.1% (v/v) formic acid, and selected doubly charged peptides, identified in MS-TOF mode, fragmented by N2 collision and analyzed by MS/MS. Mass spectra and collision MS/MS data were analyzed with Analyst QS and BioAnalyst software (Applied Biosystems, Sydney).

Bioinformatic data analysis

Sub-cellular targeting of predicted protein sequences were performed with TargetP ( as directed by Emanuelsson et al. (2000) and by Predotar ( as directed on this website. Molecular mass and pI were determined using the ProtParam program on an ExPASy website (


LJS was supported by a BBSRC David Phillips Fellowship and AHM by an Australian Research Council QEII Fellowship. Funds from the ARC to AHM and DAD and from the BBSRC to CJL are gratefully acknowledged.