Divalent metal ions in plant mitochondria and their role in interactions with proteins and oxidative stress induced damage to respiratory function.

Understanding the metal ion content of plant mitochondria and metal ion interactions with the proteome are vital for insights into both normal respiratory function and the process of protein damage during oxidative stress. We have analyzed the metal content of isolated Arabidopsis mitochondria revealing a 26:8:6:1 molar ratio for Fe:Zn:Cu:Mn and trace amounts of Co and Mo. We show selective changes occur in mitochondrial Cu and Fe content following in vivo and in vitro oxidative stresses. Immobilized-metal affinity chromatography (IMAC) charged with Cu 2+ , Zn 2+ and Co 2+ was used to identify over one hundred mitochondrial proteins with metal-binding properties. There were strong correlations between the sets of IMAC-interacting proteins, proteins predicted to contain metal-binding motifs and protein sets known to be oxidized or degraded during abiotic stress. Mitochondrial respiratory chain pathways and matrix enzymes varied widely in their susceptibility to metal-induced loss of function, showing the selectivity of the process. A detailed study of oxidized residues and predicted metal-interaction sites in the TCA cycle enzyme, aconitase, identified selective oxidation of residues in the active site and show an approach for broader screening of functionally significant oxidation events in the mitochondria proteome. mitochondrion during oxidative stress. The interactions between metal ions and proteins were also investigated using immobilized-metal affinity chromatography (IMAC). Functional assays were used to determine the targets and consequences of metal ion interaction in the mitochondrion and to explore the linkages to the redox nature of the metal and the loss of mitochondrial functions. Finally, a detailed study of the oxidized peptides of aconitase was undertaken to probe the linkage between metal


Introduction
Transition metal ions are essential in a myriad of biochemical functions by being incorporated into or associate with proteins to elicit functions in living cells. In plant mitochondria, key functions of metal cofactors include metabolism, electron transport, ATP synthesis and the detoxification of reactive oxygen species. For example, Cu and Fe ions facilitate the transfer of electrons in the electron transport chain (Bligny and Douce, 1977;Pascal and Douce, 1993), proteins of the TCA cycle utilize metal ion cofactors to catalyse primary metabolic reactions (Miernyk and Randall, 1987;Jordanov et al., 1992), Mn and Fe are required for antioxidant defense enzymes (Alscher et al., 2002), and Zn is required for the protein import apparatus in both carrier protein transport to the inner membrane (Lister et al., 2002) and in presequence degradation (Moberg et al., 2003). Co is known to substitute for other metal ions in the activation of NAD-malic enzyme and succinyl-CoA ligase from plant mitochondrial extracts (Palmer and Wedding, 1966;Macrae, 1971), but it is not known whether there is an in vivo requirement for trace amounts of Co for plant respiratory metabolism.
Metal ions, however, can also be highly toxic to cells and cell organelle functions. The redox inactive heavy metal Cd exhibits strong affinity for O, N and S atoms (Nieboer and Richardson, 1980) and can inhibit enzyme activity by direct blocking protein function or displacement of natural metal centres. There are numerous reports of heavy metals depleting cellular glutathione pools leading to diminished antioxidant protection in the cell and resulting in ROS accumulation (Schutzendubel and Polle, 2002). Cd has been reported to both directly and indirectly inhibit plant mitochondrial function (Kesseler and Brand, 1994;Smiri et al., 2009) as have Co complexes (Guzhova et al., 1979). Redox active metal catalysts such as Cu and Fe cations can also be cytotoxic as they react with ROS via the Haber-Weiss reaction or Fenton-type reactions to produce the hydroxyl anion ( • OH) (Stohs and Bagchi, 1995). Inhibitory effects of exogenously added Cu and Fe on plant respiratory function have been reported (Kampfenkel et al., 1995;Padua et al., 1996;Padua et al., 1999). Therefore, the presence of free metal cations, both redox active or inactive, in mitochondria may significantly contribute to the initiation and perpetuation of oxidative stress. One of the best described mechanisms for metal-linked damage is metalcatalysed oxidation (MCO) of proteins which involves the oxidation of susceptible amino acids such as arginine, lysine, proline and histidine among a plethora of other poorly characterized consequences (Stadtman, 1990). It has been proposed that MCO of proteins can be a highly specific event where proteins are more susceptible to damage if they bind metal ions and when the site of protein oxidation can be defined on the protein surface that binds to the metal ions (Stadtman, 1990). One of the major consequences of MCO is the irreversible formation of reactive carbonyls on amino acid sidechains (Stadtman, 1990). Such carbonyls are known to accumulate in the wheat mitochondrial proteome during environmental stress, even more so than in other ROS-producing subcellular organelles of plants ( Bartoli et al., 2004). The selectivity of protein susceptibility to MCO was also demonstrated in rice where distinct subpopulations of the mitochondrial matrix proteome were carbonylated following Cu 2+ and H 2 O 2 treatment (Kristensen et al., 2004). The targeted damage of select sets of plant mitochondrial proteins has also been observed in other studies, but without clear linkage to the role of metal ions. For example, altered protein abundances has been observed in Arabidopsis (Sweetlove et al., 2002) and pea (Taylor et al., 2005) mitochondria after the initiation of oxidative or environmental stress. Additionally, inhibition of respiratory metabolism by the lipid peroxidation byproduct 4-hydroxy-2-nonenal has been shown to operate through modification of a specific subset of proteins (Taylor et al., 2002;Winger et al., 2005;Winger et al., 2007). However, the mechanisms of targeted oxidative modification, the role of metals, and the consequences for mitochondrial metabolic function are not very well understood. Furthermore, whether or not selectivity of protein damage in mitochondria is based on relative metal ion affinity and if the sites of damage can be predicted by the sites of metal ion binding is not known.
In the current study, we investigated metal homeostasis in the Arabidopsis mitochondrion during oxidative stress. The interactions between metal ions and proteins were also investigated using immobilized-metal affinity chromatography (IMAC). Functional assays were used to determine the targets and consequences of metal ion interaction in the mitochondrion and to explore the linkages to the redox nature of the metal and the loss of mitochondrial functions. Finally, a detailed study of

The plant mitochondrial metallome and its changes during oxidative stress
The metal content of mitochondria isolated from Arabidopsis cell culture was analyzed by inductively-coupled plasma mass spectrometry (ICP-MS). This showed Fe, Cu, Zn and Mn to be the predominant species of transition metals in Arabidopsis mitochondria, with trace levels of Co and Mo also detected (Table IA). Fractionating mitochondria into soluble and membrane components and further isolation of an integral membrane portion, revealed the integral membrane proteome contained 3fold more Cu and Fe than the soluble proteome, on a protein basis (Table IB). This is likely to be due to the abundance of Cu-and Fe-containing electron transport chain components and their enrichment in the integral membrane fraction. Mn was evenly distributed between the soluble and the integral membrane compartment (Table IB).
The redox cycling metals, Cu and Fe, accounted for approximately 75% of the mitochondrial metallome (Table IA). As labile metals are a likely cause of oxidative damage to proteins, ICP-MS was then used to quantify changes in metal composition of mitochondria isolated from cells under oxidative stress. The chemical treatments of cells and timing of analyses were selected based on previous studies of plant mitochondrial oxidative damage in which the products of lipid peroxidation were measured to peak at 8 hours post-treatment (Winger et al., 2005).
Antimycin A is an inhibitor of Complex III that ultimately leads to superoxide production within mitochondria (Maxwell et al., 1999), while menadione treatment of cells is reported to induce a broad cellular superoxide production from membranes (Hollensworth et al., 2000). Treatment of cells with either chemical was compared to a methanol as well as an untreated control to ensure the carrier solvent was not responsible for the changes observed. Both chemicals induced a 1.5-2-fold accumulation of Fe in mitochondria on a protein basis (Table IIA). This accumulation of Fe in mitochondria isolated from stress treated cells could not be traced to either the soluble protein fraction or the integral membrane protein fraction from these mitochondrial extracts (Table IIA), suggesting it was probably in the peripheral membrane fraction that was stripped from the membranes by carbonate extraction prior to ICP-MS analysis. Menadione treatment also elicited a slight reduction in mitochondrial integral membrane Cu content (Table IIA), suggesting damage to membrane-embedded cupro-proteins. The accumulation of Fe following both 7 menadione and antimycin A treatments may suggest a common mechanism induced by superoxide, regardless of the cellular source.
Damage to metalloproteins from oxidative treatments of cells was most apparent following H 2 O 2 treatment. In isolated mitochondria, H 2 O 2 induced a 40% decrease in Cu content (Table IIA). This loss of Cu could be traced to both the soluble and integral membrane protein fractions (Table IIA), which implies damage to mitochondrial soluble cupro-proteins and also to cupro-components of the membrane bound electron transport chain. A 40-50% reduction of Fe and Mn content was also observed in the soluble protein fraction of mitochondria from H 2 O 2 treated cells (Table IIA). This suggested that matrix metalloproteins such as Fe-S containing aconitase and Mn-containing SOD are sensitive to damage by H 2 O 2 . The observed metal losses are consistent with the fact that hydrogen peroxide is known to be able to damage proteins by fragmentation, destroying metal coordination sites (Hunt et al., 1988) and by evidence that plant aconitase is easily inactivated by H 2 O 2 (Verniquet et al., 1991).
To provide further evidence of direct oxidative damage to metalloproteins causing loss of metal ions, mitochondria isolated from untreated cells were directly treated with H 2 O 2 in vitro and the metal content of whole mitochondria and across different mitochondrial compartments was measured by ICP-MS (Table IIB). In total mitochondria samples, a significant reduction in Fe content was observed, and this loss was localized to the soluble fraction and not the integral membrane protein fraction. This apparently release of Fe from the soluble compartment could not be detected in low molecular mass fractions obtained by size exclusion chromatography of mitochondrial extracts but this is probably due to the detection limitations of the ICP-MS (data not shown).

Inhibition of respiratory function by labile metal cations
Transition metal ions generally do not exist as free cations in cells but are sequestered by organic acids or protein ligands to minimise damage resulting from redox cycling with reactive oxygen species (Rauser, 1999). However, in the current study, ICP-MS has provided evidence that metal homeostasis can be disrupted during oxidative stress. During this time of increased metal flux into and out of the mitochondria, it is likely that the metal ions exist in a labile state and hence bind to 8 and could have toxic effects on other proteins. Historical studies have noted that exogenous addition of metal ions affect respiratory complexes and consequently mitochondrial energy production (Skulachev et al., 1967;Kesseler and Brand, 1994;Padua et al., 1996;Kuznetsova et al., 2005). Hence, the respiratory capacity of mitochondria was measured following addition of Cu 2+ , a redox active ion, or Zn 2+ , a redox inert metal ion, to characterise the differential effects of free metal ions on mitochondrial respiratory function.
A dose response procedure was undertaken using treatments for 5 mins with metal ions at 10, 50 and 100 μM concentrations compared to untreated control samples. This gave metal ion:protein ratios of 83, 415 and 830 nmol metal ions per mg of protein, respectively. At 10 μM metal ions, no detrimental effects on respiration could be observed. However, at 50 μM and 100 μM, metal ions inhibit mitochondrial respiratory rate (Table III). The electron transport chain (ETC) and tricarboxylic acid (TCA) cycle substrates used were succinate, NADH, and glutamate+malate. Each of these substrates initiates mitochondrial respiration via a different pathway: succinatedependent respiration measures the activity of the succinate dehydrogenase complex and the subsequent ETC; NADH-dependent respiration measured external NADH dehydrogenase entry to the ETC, bypassing Complex I; whilst the organic acids glutamate+malate are substrates for the TCA cycle which ultimately produces NADH in the matrix that enters the ETC at Complex I and/or the internal alternative NADH dehydrogenases.
Succinate-dependent respiration was susceptible to Cu 2+ and Zn 2+ treatment resulting in at least a 40% decline in respiration at both 50 μM and 100 μM of each metal ion (Table IIIA). In contrast, external NADH-dependent respiration was diminished differentially by Cu 2+ and Zn 2+ . The toxicity of Cu 2+ toward NADHdependent respiration was more pronounced as almost 80% inhibition of respiration was observed at both 50 μM and 100 μM Cu 2+ concentrations. On the other hand, Zn 2+ could only induce 40% inhibition of respiration under the same metal ion concentrations examined. Cu 2+ and Zn 2+ did not significantly affect glutamate/malatedependent respiration.
To determine if the cytochrome or alternative pathway were differentially affected by metal ion treatment, inhibitors were used to analyse each one independently. Cytochrome pathway inhibition by metal ions could be observed at 50 and 100 μM concentrations but did not occur in a dose-dependent manner as the extent of the inhibition appeared to plateau (Table IIIB). A 40-60% inhibition of respiration via the cytochrome pathway with either succinate or NADH as a substrate was observed following addition of either metal ion. Zn 2+ but not Cu 2+ was found to significantly inhibit the glutamate/malate-driven cytochrome pathway respiration. The alternative oxidase pathway was inhibited sharply by Cu 2+ at 50 μM and this metal ion could almost completely abolish respiration at 100 μM during both succinate and NADH driven alternative oxidase-dependent respiration. Zn 2+ was also able to inhibit alternative pathway respiration but to a lesser extent (Table IIIC).

Selective interaction of mitochondrial proteins with immobilised metal cations
The interaction of metal ions with mitochondrial proteins that might be responsible for these toxic effects were investigated in vitro using immobilised metal affinity chromatography (IMAC) to trap metal-binding proteins. Initial studies used Ca 2+ , Co 2+ , Cu 2+ , Fe 3+ , Mg 2+ , Mn 2+ or Zn 2+ to charge the IMAC resin ( Figure 1A).
Arabidopsis mitochondrial proteins solubilised in 0.1% (v/v) Triton X-100 were introduced into IMAC columns charged with metal ions of interest. Proteins that were metal-binding were retained on the IMAC resin whilst unbound proteins were removed in wash steps. The bound proteins were eluted from the IMAC resin by stripping the metal ions with EDTA. The eluent was concentrated, desalted and analysed by SDS-PAGE and then compared to total mitochondrial protein extracts to determine the specificity of protein enrichment. This showed that only Co 2+ , Cu 2+ , Zn 2+ and Fe 3+ retained subsets of mitochondrial proteins ( Figure 1A). These metal ions were then used to explore the differential interactions of divalent metal ions with proteins and how these interactions potentially modulate protein function. To further eliminate apparent non-specific binding of proteins to IMAC, fractionation of the proteins into weakly and strongly interacting sets by electrostatic or competitive displacement, with NH 4 Cl or imidazole respectively, was conducted. The enriched bands were excised for protein identification by mass spectrometry (MS). For MS protein identification of SDS-PAGE bands, multiple proteins are often identified and only the major proteins identified from each band were reported (Tables S1-S4).
Cu 2+ -interaction proteins -Initial time course studies of proteins binding to Cu 2+ -IMAC revealed that proteins bound rapidly and 1 min incubations were sufficient to permit binding (data not shown). However, the protein profile of bound proteins was very complex and showed poor selectivity compared to whole mitochondrial samples ( Figure 1A). Various concentrations of imidazole were investigated but resulted in poor fractionation of proteins based on strength of binding (results not shown).
Ammonium chloride step gradients allowed electrostatic displacement of the Cu 2+bound proteins and improved protein fractionation providing distinct protein profiles based on the strength of protein affinity to Cu 2+ -IMAC ( Figure 1B). Thirty-five proteins where identified in bands from fractions between 0.1 M and 0.6 M NH 4 Cl and were designated weakly associating proteins (Supp Figure 1 and Table S1) whilst 48 strong Cu 2+ -IMAC interacting proteins were identified from 0.8 M to 1 M NH 4 Cl fractions and also from the EDTA stripped IMAC resin (Supp Figure 1 and Table S2).
Some of the proteins identified were also found by the Kung et al (2006) (Table S1,S2).
Interestingly, the list of Cu 2+ binding proteins identified here is remarkably similar to the reported list of rice matrix proteins that are carbonylated after Cu 2+ -induced oxidation of rice mitochondrial extracts (Kristensen et al., 2004).
Co 2+ -interaction proteins -Initially, unfractionated (UF) proteins which were bound to Co 2+ -IMAC were analysed by SDS-PAGE and revealed an enrichment of some protein bands when compared to the total mitochondrial protein sample ( Figure 1A,C).
The 10 bands that were enriched in the unfractionated sample were excised and identified ( Figure S2 and Table S3). The separation of weak and strong Co 2+interacting proteins was attempted using both competitive and electrostatic displacement of proteins. Imidazole fractionation of Co 2+ -IMAC bound proteins showed most eluted at concentrations of 10-20 mM. Five extra protein bands were enriched in the 20 mM imidazole fraction and excised for protein identification ( Figure 1C, Figure S2 and Table S3). Fractionation via electrostatic displacement was also conducted. The majority of Co 2+ -binding proteins could be displaced from the resin using 0.1 M NH 4 Cl showing the binding of proteins to Co 2+ -IMAC is significantly weaker than that of Cu 2+ -IMAC binding proteins (data not shown).
Using Co 2+ -IMAC, 45 proteins involved in detoxification, DNA synthesis, protein fate, protein synthesis, signal transduction and unknown functions were identified.
However, energy production and metabolism proteins were by far the best represented functional category (Table S3). Co 2+ -IMAC was able to purify proteins that were found using both Cu 2+ and Zn 2+ but the binding of proteins to Co 2+ appears more similar to Zn as the proteins NADH dehydrogenase subunit 9, cytochrome c oxidase subunit 5b, dihydrolipoamide dehydrogenase, malic enzyme, 2-oxoglutarate dehydrogenase and nucleoside diphosphate kinase were found in common between the Co 2+ and Zn 2+ sets but were not in the Cu 2+ set. Co 2+ -IMAC was also able to purify 19 proteins that could not be purified by the other divalent cations ( Figure 1E).
Examples of proteins exclusively purified by Co 2+ -IMAC include DAG proteins, subunits of Complex I, methylcrotonyl CoA carboxylase alpha subunit, arginase, and 4 other metabolic enzymes. Co 2+ -IMAC was also shown to be an effective tool in enriching low abundant proteins as 9 proteins previously uncharacterised by MS to be in mitochondria were identified (Table S3).
Zn 2+ -interaction proteins -Unfractionated (UF) proteins which were bound and eluted from Zn 2+ -IMAC were analysed by SDS-PAGE which revealed enrichment of specific proteins bands when compared to the total mitochondrial protein sample ( Figure 1A). The 11 bands that were enriched in the unfractionated fraction were excised for identification ( Figure S3 and Table S4). The separation of weak and strong Zn 2+ -interacting proteins was also attempted using both competitive and electrostatic displacement of proteins. As with Cu 2+ -and Co 2+ -IMAC, imidazole was not able to efficiently fractionate proteins bound to Zn 2+ -IMAC ( Figure 1D). Despite poor fractionation of all the IMAC protein sets with imidazole, relative protein binding strength could be determined based on the concentration of imidazole needed to elute a large portion of the proteins. Proteins bound to Zn 2+ -IMAC more tightly than Co 2+ -IMAC as the majority of proteins were displaced at 50 mM imidazole in Zn 2+ -IMAC compared to 10-20 mM imidazole in Co 2+ -IMAC ( Figure 1C and 1D).
Electrostatic displacement of Zn 2+ -interacting proteins using ammonium chloride also verified stronger protein binding to Zn 2+ compared to Co 2+ as proteins tended to eluted at 0.7 M NH 4 Cl compared to 0.1 M NH 4 Cl in Co 2+ -IMAC (data not shown).
Eleven additional protein bands observed to have eluted from Zn 2+ -IMAC in the 50 mM imidazole fraction were also excised and identified ( Figure S3 and Table S4).
Zn 2+ -IMAC appeared to have no success in the enrichment of the well-characterised Zn-dependent proteins such as TIMs (Lister et al., 2002) or the metalloprotease PreP (Moberg et al., 2003). Instead, the 74 proteins identified are likely to associate with Zn through a variety of mechanisms. Some examples of these include proteins that contain Zn fingers (cytochrome c oxidase 5b) (Kubo et al., 2006), use Zn as a cofactor (nucleoside diphosphate kinase, mitochondrial processing peptidase beta) (Parks and Agarwal, 1973;Luciano and Geli, 1996), are inhibited by Zn (methylcrotonyl-CoA carboxylase, 2-oxoglutarate dehydrogenase) (Diez et al., 1994;Brown et al., 2000), or interact with divalent cations (malic enzyme, glutamate dehydrogenase, succinate dehydrogenase Fe-sulphur subunits) (Massarini and Cazzulo, 1975;Kindt et al., 1980). Proteins with no known association with Zn ions such as the ATP synthase subunits were also found. Interestingly, Zn 2+ -IMAC was able to purify 11 subunits of the ATP synthase complex which was greater than any of the other metal-IMAC resins used.
Fe 3+ -interaction proteins -Enrichment of proteins with Fe 3+ -IMAC was very variable.
The major protein enriched in Figure 1A was identified as malate dehydrogenase (At1g53240), but this pattern and specificity can be changed and greatly dependent on the pH and ionic strength of the media used. Our data indicated we were largely looking at a pseudocation exchange effect which was in agreement with the findings by (Zachariou and Hearn, 1996) (see Figure S4 for more details).

Metal-binding motifs in mitochondrial IMAC-interacting protein sets -Metals often
bind proteins at specific coordination sites involving cysteine, histidine and methionine residues (Harding, 2004). Hence an analysis of the sequences of the proteins in our IMAC subsets was performed to determine if putative metal-binding motifs were more common that expected by random chance in these proteins. For simplicity, the metal-binding motif parameters used were the same as those used by Kung et al (2006) to validate their metal-binding proteins. Kung et al (2006) found that the top 6 statistically enriched motifs in their protein subsets could accounted for nearly 90% of the proteins identified. Kung et al (2006) was also able to show Cubinding to synthetic peptides carrying a range of these putative Cu-binding motifs but did not characterise the biological significance of such motifs. In our data, 10, 45, and 21 Cys-His-Met motifs were found to be significantly greater in frequency in the Cu, Co and Zn subsets respectively when compared to the entire Arabidopsis proteome or the known Arabidopsis mitochondrial proteome (Tables S5a-i). Figure 2 shows the top 6 enriched motifs for each of the divalent metal cation bind sets. In total 72% of proteins from our Cu-binding subset contained one or more of the top 6 Cu-binding motifs compared to 96% and 89% for Co and Zn respectively ( Figure S5), which is comparable to the findings of Kung et al (2006) using this same method. One of the top 6 motifs, 'H-(X) 5 -H', was common to all three datasets and also to Kung et al (2006), while 'HM' was shared between our Co 2+ and Zn 2+ datasets. His residues were significantly more enriched in the Co-binding protein subset, suggesting that His motifs may complex Co more readily than Cu or Zn.

Relationship between metal ion interaction and modulation of protein function
A range of major TCA cycle enzymes, that were identified to interact in IMAC here and have been reported to be carbonylated in the literature, were selected for activity measurements. The aim was to determine if there was a clear relationship between observed metal ion interactions observed using IMAC, reports of protein oxidation, and modulation of enzyme activity.
Aconitase was found to interact with Cu 2+ (Table S2) and is reported to be oxidatively modified by Cu-catalysed mechanisms (Kristensen et al., 2004).
Aconitase was inhibited 25% by H 2 O 2 treatment (Table IV) and this is in agreement with previous studies in plants and mammals (Verniquet et al., 1991;Brazzolotto et al., 1999). Cu 2+ was able to diminish aconitase activity by 80% and the primary mechanism of this inhibition appeared to be independent of reducing or oxidising agents (Table IV). Fe 3+ and Zn 2+ did not affect aconitase activity directly or by MCO.
Interestingly, the presence of Fe 3+ or Zn 2+ in the H 2 O 2 treatment was infact able to protect aconitase activity (Table IV) (Brazzolotto et al., 1999).
The activity of pyruvate dehydrogenase complex and α-ketoglutarate dehydrogenase complex showed similar patterns of inhibition by metal ions. Both activities were significantly affected by Cu 2+ and Zn 2+ by direct mechanisms and were found to be resistant to MCO under the conditions examined (Table IV). The observed toxicity of Cu 2+ towards pyruvate and α-ketoglutarate dehydrogenase complex activities agrees with in vivo and in vitro studies in mammalian mitochondria (Sheline and Choi, 2004). The toxicity of Zn 2+ towards α-ketoglutarate dehydrogenase complex activity could at least in part explain the inhibition of glutamate/malate driven cytochrome oxidase-dependent respiration by Zn 2+ (Table   III). Subunits of pyruvate dehydrogenase complex was found by IMAC to interact with Co 2+ and Cu 2+ and subunits of α-ketoglutarate dehydrogenase complex to interact with Co 2+ and Zn 2+ , indicating mechanisms for the direct effect of metal ions on these enzymes. Fumarase was not found as a metal-interacting protein in our IMAC studies but has been reported to be carbonyl-modified in rice mitochondria by Cu 2+ -catalysed oxidation (Kristensen et al., 2004). In our hands the activity of this enzyme was not affected by MCO (Table IV). However it was the only enzyme investigated that showed sensitivity to ferric ions which caused a 30% decrease in activity (Table IV).
The TCA cycle enzymes isocitrate dehydrogenase and malate dehydrogenase were also found to be putatively oxidised in rice mitochondria (Kristensen et al., 2004 Hence, while a number of TCA cycle enzymes that were selected for analysis showed a functional impact of incubation with metal ions, there was a complex relationship with the metal interaction data (Table IV). As MCO of proteins may not necessarily occur at the active site, protein carbonylation may not be noticeably detrimental in protein function assays. The specificity of carbonylation was further investigated to assess the susceptibility of putative metal-binding sites to oxidation using aconitase as a case study.
Linking metal binding sites and oxidized residues to protein structure of aconitase.
Aconitase interacts with Cu 2+ -IMAC ( Figure 1B, Figure S1 and Table S2), contains all but one of the top 6 putative Cu 2+ binding motifs, has reduced activity on Cu 2+ /H 2 O 2 exposure (Table 4) and is reported to be oxidatively modified by Cucatalysed mechanisms (Kristensen et al., 2004). The 3D structure of the Arabidopsis mitochondrial aconitases (At2g05710 and At4g26970) can also be predicted by using the crystal structure of the human cytosolic aconitase, PDB accession 2b3y (Dupuy et al., 2006), which shares ~60% amino acid identity to the Arabidopsis mitochondrial aconitase proteins.
The location of the putative Cu-binding motifs H-X 5 -H, M-X 3 -H, C-X 7 -H, H-X 1 -C, M-X 7 -H and H-X 2 -M were mapped onto the aconitase structure using a BLAST alignment of the protein sequences ( Figure S6). Based on the 2b3y crystal structure, the 2 coordinating His residues of the H-X 5 -H motif are 18.93Å and are not likely to form a metal-coordinating pocket. The M-X 3 -H motif and the conserved C-X 7 -H, and H-X 1 -C motifs are clustered together on the protein surface indicating that this region is likely to attract Cu ions, however the distances between the respective coordinating residues are 6.51 Å, 11.58 Å, and 5.25 Å for M-X 3 -H, C-X 7 -H, and H-X 1 -C respectively, which is relatively large as the criteria for metal-coordination normally requires the electron donor to be within ~0.75 Å of the metal ion (Harding, 2004). H-X 2 -M was the only motif found within the core of the protein, near the aconitase catalytic and substrate recognition site. The predicted distance between the coordinating residues of this motif is 5.75 Å. Whether this site is responsible for Cu inactivation of aconitase activity remains to be investigated. However, it is unlikely the motif is responsible for aconitase binding to the Cu 2+ -IMAC resin as the site is buried within the protein core and is likely to have restricted access to the immobilised Cu 2+ .
To test whether sites near putative metal-interacting motifs had an enhanced susceptibility to MCO, the soluble protein fraction from Arabidopsis mitochondria was treated with 100 μM Cu 2+ and H 2 O 2 to elicit MCO. The band corresponding to aconitase on SDS-PAGE was excised trypsin digested and peptides analysed by mass spectrometry. The variable modifications of oxidation of the amino acids C, D, F, K, M, N, P, R, and Y were used to determine if MCO elicited any detectable changes between mock and treated extracts. In triplicate experiments there were 11 tryptic peptides from aconitase consistently found in both mock and Cu 2+ treated samples, in addition three peptides from aconitase were consistently found only in the mock samples and one multiply oxidised peptide was found consistently and only in the Cu 2+ treated samples ( Figure 3A). This latter peptide is in a highly conserved region of aconitase and includes an Asp and His residue of the active site ( Figure 3B, C). A range of other oxidised peptides of aconitase from the oxidised samples were found in single experiments but were not able to be repeatedly observed. These apparently random events typically mapped to the surface of the 3D structure, but are generally not in close proximity to the metal-binding motifs (Table S6, Figure S6).

Discussion
Metals are important cofactors in many biological reactions but to date there has been little systematic analysis of the metal composition of subcellular organelles in plants. Our screen of the metallome of Arabidopsis cell culture mitochondria is to our knowledge the first multi-elemental profiling of a subcellular organelle in plants.
Fractionation of mitochondrial samples revealed the integral membrane fraction had a 6-fold greater Cu and Fe content than the soluble protein compartment on a protein basis, consistent with the redox transition metals involved in the electron transport chain. The Arabidopsis mitochondrial Fe and Mn content of 3.2 and 0.12 nmol per milligram of protein respectively, is relatively similar to that of yeast mitochondria which was found to be 5-10 and 0.16-0.36 nmol per milligram of protein, respectively (Luk and Culotta, 2001;Luk et al., 2003;Luk et al., 2005;Yang et al., 2006). The discovery of trace amounts of Co 2+ and Mo 2+ in Arabidopsis mitochondria was somewhat unexpected, however, reports of Co 2+ substituting for other metals in metal-dependent enzyme reactions in plant mitochondria (Palmer and Wedding, 1966;Macrae and Moorhouse, 1970;Macrae, 1971) and of a mitochondrial Mo 2+ carrier protein in Arabidopsis (Baxter et al., 2008) are consistent with these data.
Little is known regarding the subcellular perturbations in the metal content during oxidative stress despite the general acceptance that MCO is a common source of oxidative modification in biological macromolecules (Stohs and Bagchi, 1995).
Comparing the impact of oxidative stress on the metallome of the Arabidopsis mitochondria using the same chemical elicitors reported by Sweetlove et al (2002) and Winger et al (2005Winger et al ( , 2007 allowed the changes in respiratory activity, lipid peroxidation and protein degradation reported in these studies to be considered in light of metal-catalysed reactions investigated here. H 2 O 2 treatment resulted in a detectable loss of Cu from total mitochondria (Table II) (Tables S1-S4). Menadione treatment, like H 2 O 2 , was able to induce loss of Cu in the membrane fraction (Table II) (Table II) study of hamster fibroblasts in which menadione treatment increased the lability of both these redox active transition metals (Calderaro et al., 1993).
This current study is also in agreement with wider proteomic studies investigating the changes in protein abundance in Arabidopsis during abiotic stress.
Aconitase is known to release ferrous ions when oxidatively damaged (Verniquet et al., 1991;Brazzolotto et al., 1999) and has been reported to decrease in abundance in cadmium-stressed (Sarry et al., 2006), salt-stressed (Jiang et al., 2007 and salt and osmotic-stressed (Ndimba et al., 2005) plants. Another mitochondrial metalloprotein that have been shown to have altered abundance following oxidative stress is the Zncontaining mitochondrial processing peptidase (MPP) (At3g02090.1) which increased in abundance in salt and osmotic stress (Ndimba et al., 2005). The increase in abundance of MPP may contribute to the import of new proteins to replace degraded proteins. In contrast, while we have not seen any evidence of increases in Mn content in mitochondria, MnSOD has been observed to increase in abundance in salt-stressed (Jiang et al., 2007) and in salt and osmotic stressed Arabidopsis plants (Ndimba et al., 2005).
While the release of metal ions from metalloproteins and the import of new metal cofactors into the mitochondrion to synthesise or repair metalloproteins are likely to contribute to the changes in the mitochondrial metallome, they also raise the need to study the way in which metals interact with the proteome as a whole.
Examining this metal-protein interactome is extremely complicated and fraught with technical changes. Here we have used IMAC in an attempt to mimic protein binding to free metal ions in vivo. Under the native conditions explored, only exposed metalbinding surfaces are studied and there is no assessment of metal-coordinating sites that are buried in the protein core (Ueda et al., 2003). In addition, steric hindrance caused by the use of immobilised metal ions on a fixed ligand will limit access to metal ion binding sites in some proteins (Ueda et al., 2003). Additionally, the strength of metal coordination in a native metalloprotein may be too strong to allow for an exchange of metal ion cofactors with the IMAC resin. Like the IMAC study of Kung et al (2006), we have found that IMAC selects subsets of proteins that contain significant numbers of putative metal interaction motifs on their surfaces. In the metal-protein interaction studies of Cu ions and liver disease (She et al., 2003;Smith et al., 2004), nickel hypersensitivity in human B cells (Heiss et al., 2005), and Cu ion homeostasis in Arabidopsis roots (Kung et al., 2006) no attempt was made to localise the metal-binding motifs on homologous proteins. However, the surface localisation of the putative metal-binding motifs is supported by reports of the use of IMAC to assist in protein crystallisation through surface histidine residues (Frey et al., 1996), and reports of protein contaminants in metal affinity purification of His-tagged recombinant proteins due to natural surface metal-binding motifs (Cai et al., 2004;Bolanos-Garcia and Davies, 2006). While the functional significance of these putative surface metal-interacting sites remains undefined, we noted that there was a good correlation between these proteins and experimentally observed sets of proteins that Detailed analysis of the activity of major matrix-located enzymes following expose to oxidative conditions revealed very different responses (Table IV) indicating that the impact of oxidation on function is highly variable, dependent on different sets of conditions and different metal ions and thus specific studies will be required to assess the impact of chemical stresses on each protein (Table IV). In support of this, studies in castor bean peroxisomal proteins have also demonstrated a lack of correlation between the extent of protein oxidation and inhibition of protein function following metal-catalysed oxidation (Nguyen and Donaldson, 2005). The prospect of having to perform functional analysis on all oxidised proteins to infer effects is daunting. However, our study of aconitase revealed that MS analysis of the peptides from damaged proteins and layering onto 3D structures (Figure 3, Table S6) can provide important information. For example, by revealing putative active site damage consistent with the loss of function of this enzyme during oxidative conditions as well as a range of surface localised oxidation sites.
The general proposition that electron-donor dense regions on protein surfaces coordinate metal ions and, in the presence of ROS, the specific metal-coordinating residue(s) are oxidised (Stadtman, 1993), is supported by a series of specific studies.
For glutamine synthetase, it was observed that in a peptide containing a stretch of oxidation-prone amino acids, Met 268 -His 269 -Cys 270 -His 271 -Met 272 , only the metalcoordinating His 269 was oxidised (Farber and Levine, 1986). The theory of 'caged MCO' has been applied to the characterisation of Cu-binding residues of the cuproproteins Cu/ZnSOD and azurin (Bridgewater et al., 2006) and the known metalinteracting proteins angiotensin I and bacitracin (Bridgewater et al., 2006(Bridgewater et al., , 2006. The sites of oxidation which were found to be in the vicinity of the Cu-interacting site were determined by manual interpretation of the product ions derived from both MS/MS and MS/MS/MS analysis of peptide fragments (Bridgewater et al., 2006(Bridgewater et al., , 2006. Also, specific surface histidine residues of the cuproprotein ceruloplasmin have been shown to coordinate Cu and promote the oxidation of low density lipoprotein (Mukhopadhyay et al., 1997). Our analyses to date have not provided evidence for a link between metal binding sites and MCO in the case of the plant mitochondrial aconitase. However, studies of complex lysates have yet to be performed in any system to demonstrate MCO specificity on a large scale and thus determine if this concept is the exception or the rule in understanding metal-induced damage to proteins.
While it is generally considered that oxidative modification leads to protein dysfunction by affecting the structural integrity of the protein, promoting the formation of protein aggregates, and potentially damaging the active site (Starke-Reed and Oliver, 1989), protein oxidation may not be entirely detrimental to protein function. Hence, among the protein oxidation events in plant mitochondria may be important triggers for the stress response of the organelle. Protein oxidation can promote protein degradation and turnover (Rivett, 1985;Davies and Lin, 1988;Marcillat et al., 1988). It can also provide protective outcomes for the cell, for example, a bacterial transcription factor has been identified that senses and promotes an appropriate cellular response to an oxidative environment through MCO (Lee and Helmann, 2006). In addition, methionine residues on protein surfaces has been shown to act as antioxidants to protect the active site of enzymes (Levine et al., 1996)

Induction of oxidative stress in Arabidopsis and isolation of mitochondria
Seven-day old dark-grown heterotrophic Arabidopsis thaliana (cv. Landsberg erecta) cell suspension cultures (May and Leaver, 1993) were treated with either 10 mM H 2 O 2 , 400 μM menadione prepared in methanol, or 25 μM antimycin A prepared in methanol for 8 hours. The concentrations of antimycin A and menadione have previously been optimised (Winger et al., 2005). Equivalent volumes of either water or methanol were added as controls for respective treatments. Mitochondria were isolated from as described previously (Millar et al., 2001).

ICP-MS analysis of metal content
To determine sites of altered metal content, mitochondria were fractionated into soluble protein and integral membrane protein fraction. Intact isolated mitochondria were suspended in milliQ water before lysis by 3 freeze/thaw cycles.
Soluble proteins were collected in the supernatant following centrifugation at 20,000 x g. Peripherally attached proteins were depleted from the total membrane fraction using 0.1 M Na 2 CO 3 treatment to produce the integral membrane protein fraction (Fujiki et al., 1982). Re-distilled concentrated HNO 3 (kindly provided by Prof. John Watling, Centre for Forensic Science, UWA) was used to breakdown organic material by heating at 160 o C for 1 hr. The acid digest was diluted to <5% (v/v) HNO 3 , and passed through 0.22 μm filters (Millipore). Metal speciation and quantification was performed by Centre for Forensic Science, UWA using Perkin Elmer Elan 5000 ICP-MS calibrated against elements of interest.

Respiratory assays
Oxygen consumption by cell suspension cultures and isolated mitochondria was measured by a Clark-type oxygen electrode (Hansatech Instruments, UK).
Respiratory data were collected and analysed using OxyGraph Plus v1.01 software (Hansatech Instruments, UK). Liquid phase calibration was performed by adding excess sodium dithionite to 1 ml autoclaved water to remove all oxygen at 25 o C.
Mitochondrial respiration assays were conducted at 25 o C by adding 100-120 μg www.plantphysiol.org on August 15, 2017 -Published by Downloaded from Copyright © 2009 American Society of Plant Biologists. All rights reserved. mitochondrial protein to 1 ml respiration medium (0.3 M sucrose, 5 mM KH 2 PO 4 , 10 mM TES, 10 mM NaCl, 2 mM MgSO 4 , 0.1% (w/v) BSA, pH 7.2). Respiration via Complex II was initiated by adding succinate and ATP. Respiration initiated through the external NADH dehydrogenase pathway was measured using NADH, CaCl 2 , and rotenone. Respiration initiated through matrix NADH dehydrogenases was measured using malate, glutamate, CoA, TPP, and NAD + . In all respiratory assays, following the addition of substrates to initiate respiration, ADP was added to induce maximum state 3 respiratory rates. Respiration via the alternative and cytochrome oxidase pathway was determined by the addition of the inhibitors KCN and nPG, respectively.

Enzyme assays
Assays were performed at 25 o C using a temperature controlled spectrophotometer (U-2810 spectrophotometer, Hitachi) and data were collected by UV Solutions software v2.1 (Hitachi). For inhibition assays, mitochondrial samples equivalent to 50 μg of protein were treated with 100 μM of metal ions for 5 min prior to the assay. The metal ions were added at an excess of 1.5 μmol per mg of protein to induce maximal toxicity. Equimolar ascorbate or H 2 O 2 was added to catalyse MCO.
The protein samples were then directly added to the reaction medium for enzyme assays. All assays were conducted according to published methods (Lee et al., 2008).

Statistical analyses
Unless stated otherwise, all data obtained from experiments were expressed as mean ± standard error about the mean (SEM). Statistical significances were evaluated by the two-tailed unpaired Student's t-test using Microsoft Office Excel 2003 or Kaleidagraph v3.6 (Synergy Software) where appropriate.

Immobilized metal affinity chromatography
HiTrap Chelating HP 1 ml column (Amersham Biosciences) consisting of prepacked iminodiacetic acid conjugated to agarose beads was used. All chromatography was conducted manually using syringes with the flow rate maintained at approximately 1 ml/min. The column was charged with 1 column volume (CV) of 0.1 M metal ion solution and excess metals were removed with 5 CV of milliQ water. The charged column was equilibrated with 10 CV of binding buffer (20 mM NaH 2 PO 4 , 0.5 M NaCl, 0.1% (v/v) Triton X-100) unless stated otherwise. Prior to sample loading, 2 mg of pelleted mitochondrial proteins were suspended in 500 μl of milliQ water and subjected to 3 freeze/thaw cycles. The lysate was passed through 0.22 μm filter (Millipore) and 500 μl of 2-fold concentrated binding buffer was added to the filtered lysate so that the proteins were suspended in 1 ml lysate. The lysate was injected into the column, allowed to incubate at RT for 1-5 min, and then washed with at least 10 CV of binding buffer to remove unbound proteins. Proteins were fractionated with varying concentrations of NH 4 Cl or imidazole in the binding buffer. All remaining proteins were removed from the column by stripping the metal ions with an eluent containing 50 mM EDTA and 50 mM NaCl. Due to the high salt content of the eluent fractions, the proteins were concentrated using 5 kDa molecular weight cut-off centrifugal filter units (Millipore) and the excess salt diluted following buffer exchange with milliQ water.

Metal-binding motif analysis
Metal-binding motif analysis used was based on analysis of similar work on Arabidopsis root Cu-binding proteins where histidine, methionine and cysteine residues in any combination and up to 12 amino acids apart were considered putative metal-binding motifs (Kung et al., 2006). The algorithm for motif screening was written in MySQL and 117 potential metal-binding motifs were analysed. The gene accession codes of the Arabidopsis proteome, a total of 26738 putative proteins, were extracted from TAIR. The list of accession codes were obtained from SUBA where parameters for mitochondrial proteins were "annotated in SwissProt" or "found in mitochondria by mass spectrometry" or "found in mitochondria by GFP" culminating in 742 proteins (Heazlewood et al., 2005). The standard normal distribution (Z score) statistical analysis was conducted on the frequency of occurrence of the IMAC motifs comparing to the entire Arabidopsis and the Arabidopsis mitochondrial proteomes.
Removal of IMAC-interacting proteins from the Arabidopsis and Arabidopsis mitochondrial proteome list was performed in order to conduct statistical analyses on 2 independent samples. The Cu ion / ascorbate / O 2 system is the preferred mechanism for metalcatalysed oxidation (MCO) as the major product of this MCO system is the formation of 2-oxohistidine (Uchida and Kawakishi, 1986). Dose response assays determined that 1 nmol metal ion per milligram of protein was optimum to induce maximal metalcatalysed protein oxidation with minimal protein degradation. Equimolar ascorbate was added to catalyse MCO. All MCO treatments were performed at RT with constant agitation for 5 min. The MCO reaction was stopped by the addition of 5 mM EDTA. The metal ions were then removed from the proteins by overnight dialysis at 4 o C against 50 mM HEPES-NaOH, pH 7.2.

In-gel protein digestion and peptide extraction
Gel plugs from protein bands of interest were excised and de-stained twice for 45 minutes with 50% (v/v) acetonitrile in 25 mM NH 4 HCO 3 . The plugs were dehydrated at 50°C on a dry block heater for 30 minutes and re-hydrated with 12.5 μg/ml trypsin in 25 mM NH 4 HCO 3 and incubated overnight at 37°C. Peptides were extracted by adding 15 µl acetonitrile with vigorous shaking for 15 minutes, removing liquid and secondly, adding 15 µl of 50% acetonitrile and 5% formic acid to the gel plugs followed by another 15 minutes of shaking. The second extraction step was repeated and all the samples were pooled after each extraction step and lyophilised.
Normalised data are expressed as % control ± SEM (n≥3, where * = p<0.05 and ** = p<0.01, comparing each sample to the total oxygen consumption in untreated mitochondria

Figure 1. IMAC affinity purification of proteins from Arabidopsis mitochondria.
A) selective purification of mitochondria protein subsets using different metals to charge IMAC resin and EDTA elution. B) Different strengths of Cu2+ affinity to IMAC using NH4Cl step gradient elution. C) Different strength of Co2+ and D) Zn2+ using imidazole gradient elution. E) Venn diagram of proteins identified from each metal binding set using mass spectrometry (See Tables S1-S4 for details).  (2b3y). Oxidised (red) and active site (green) residues are indicated on B and C.
Substrate recognition site (yellow) and FeS cluster ligation site (orange) residues are shown in B. In C, * Asp 125 is a conserved active site residue and was found to be oxidised.  Figures   Table S1. List of weak Cu 2+-IMAC binding proteins eluted at 0.1 M to 0.6 M NH 4 Cl.