Loss of Lon1 in Arabidopsis changes the mitochondrial proteome leading to altered metabolite profiles and growth retardation without an accumulation of oxidative damage.

Lon1 is an ATP-dependent protease and chaperone located in the mitochondrial matrix in plants. Knockout in Arabidopsis (Arabidopsis thaliana) leads to a significant growth rate deficit in both roots and shoots and lowered activity of specific mitochondrial enzymes associated with respiratory metabolism. Analysis of the mitochondrial proteomes of two lon1 mutant alleles (lon1-1 and lon1-2) with different severities of phenotypes shows a common accumulation of several stress marker chaperones and lowered abundance of Complexes I, IV, and V of OXPHOS. Certain enzymes of the tricarboxylic acid (TCA) cycle are modified or accumulated, and TCA cycle bypasses were repressed rather than induced. While whole tissue respiratory rates were unaltered in roots and shoots, TCA cycle intermediate organic acids were depleted in leaf extracts in the day in lon1-1 and in both lon mutants at night. No significant evidence of broad steady-state oxidative damage to isolated mitochondrial samples could be found, but peptides from several specific proteins were more oxidized and selected functions were more debilitated in lon1-1. Collectively, the evidence suggests that loss of Lon1 significantly modifies respiratory function and plant performance by small but broad alterations in the mitochondrial proteome gained by subtly changing steady-state protein assembly, stability, and damage of a range of components that debilitate an anaplerotic role for mitochondria in cellular carbon metabolism.

Functional proteins are assembled from polypeptides that have undergone many steps of folding and posttranslational modification. As such, many pitfalls exist where anomalous polypeptide species or configurations can form and affect biological functions. These aberrant proteins can result from errors during translation, misfolding of the nascent polypeptide, damage to the native proteins from oxidation events, or even incorrect stoichiometry of protein complex subunits (Adam, 2000;Schaller, 2004). In the cytosol, these aberrant molecules are largely removed via the 26S proteasome pathway, which involves ubiquitination of the protein substrate and subsequent degradation into peptides and eventually amino acids that can then be recycled (Hershko and Ciechanover, 1998). Oxidized cytosolic proteins are primarily degraded by the 20S core particle of the 26S proteasome in a ubiquitin-independent manner (Davies, 2001). However, these major proteasome pathways are not directly available to turnover proteins in intraorganellar environments. Therefore, mitochondria, chloroplasts, and peroxisomes possess alternate pathways governed by networks of chaperone and protease functions for the maintenance of protein homeostasis.
The AAA + protease class (ATPases associated with diverse cellular activities) of proteases (Neuwald et al., 1999;Ogura and Wilkinson, 2001) are largely responsible for this organellar protein homeostasis in plants (Janska, 2005). The ATP-dependent proteases (including the Clp, FtsH, and Lon subclasses) are multimeric ring structures (Lupas et al., 1997;Schmidt et al., 1999) that selectively bind a substrate, unfold, and translocate the substrate via ATP hydrolysis into the proteolytic chamber formed by the ring structure and then finally cleave the substrate into large peptides (Wickner et al., 1999;Chandu and Nandi, 2004). In addition to this proteolytic function, ATP-dependent proteases in organelles also serve as molecular chaperones and thereby have a dual function (Neuwald et al., 1999;Iyer et al., 2004;Ngo and Davies, 2007). These dual functions allow ATP-dependent proteases to participate in subcellular targeting, enzyme activation, protein maturation, complex formation, and protein degradation.
In mitochondria, Clps, FtsHs, and Lons have been shown to participate both in the maintenance of mitochondrial integrity by selective proteolysis as well as protein assembly (Käser and Langer, 2000). Mitochondria, as the site of cellular energy production via respiration and ATP synthesis, and specifically the mitochondrial electron transport chain (ETC), are inherent sources of free radicals and oxidative stress (Marcillat et al., 1988;Zhang et al., 1990;Cadenas and Davies, 2000). Consequently, mitochondrial proteins are regularly exposed to oxidants. Oxidatively modified proteins can become nonfunctional and serve as degradation targets. Lon protease has been implicated in degradation of oxidatively modified proteins, such as oxidized aconitase (Bota and Davies, 2002), oxidized pyruvate dehydrogenase, acetohydroxyacid reductoisomerase, and HSP60 (Bayot et al., 2010), and has also been indirectly implicated in the assembly of ETC respiratory complexes (Major et al., 2006).
Lon protease was first identified in Escherichia coli mutants that formed long, undivided filaments (Donch and Greenberg, 1968). Since its discovery, Lon has been found ubiquitously and is highly conserved across kingdoms. To date, most organisms, both eukaryotic and prokaryotic, have been found to possess only a single form of Lon, which is located in mitochondria. However, plants have evolved four separate isoforms of Lon protease located in mitochondria, chloroplasts, and peroxisomes (Janska, 2005). The lon1 gene in plants was first identified in maize (Zea mays; Barakat et al., 1998) and shortly after ORF239, a protein associated with cytoplasmic male sterility in common bean (Phaseolus vulgaris), was identified as a target of Lon1 in plant mitochondria (Sarria et al., 1998). More recently, two Lon mutants, lon1-1 and lon1-2, were found to contribute to a postgerminative growth retardation in Arabidopsis (Arabidopsis thaliana), but to different extents (Rigas et al., 2009a). lon1-1 has the more severe phenotype and bears a point mutation in Lon1, which has been postulated to lead to an aberrant protease or chaperone, while lon1-2 is a T-DNA knockout and has a milder, but similar, growth phenotype. While transcriptional rates for mitochondrial components were equal to those of the wild type, the mutants were found to have lower respiratory capacity of succinate and cytochrome c, suggesting some impairment of Complexes II and IV. Additionally, the activities of key citric acid cycle enzymes were lower in lon1 mutants when compared with the wild type (Rigas et al., 2009a). However, to date, it has been unclear how these activity changes were linked to Lon1 function, why lon1-1 and lon1-2 differ, and how the whole-plant phenotypes were linked to respiratory function.
In this study, we have sought to examine the mitochondrial proteomes of the two known Arabidopsis Lon1 mutants in depth to determine if the induced impairments in mitochondrial enzyme function are due to differences in enzyme abundance/modification, if protein oxidation is a key factor in lon1-1 and lon1-2 phenotypes, and how the impact on mitochondrial impairment is linked to plant growth. We also investigated the distinctions between the two mutants in terms of Lon1 abundance to understand the reason for their different phenotypes. We identified a range of differences in protein abundance relative to wild-type Arabidopsis and specific changes in the organic acid pools and mitochondrial stress responses in the mutants. While we found no evidence of widespread oxidation-related damage in either mutant, we did find evidence for more oxidative modification of specific proteins in lon1-1 and elevation of some antioxidant defenses. We propose that both the chaperone function and the proteolytic function of Lon1 underlie a significant number of the changes observed in Arabidopsis in both mutants.

Lon-1 Loss Decreases Plant Biomass But Not Respiratory Capacity of Plant Tissues
Lon1 protease mutants have been reported to show a retarded growth of both shoots and roots when compared with wild-type plants (Rigas et al., 2009a). To confirm this, plants were sown in soil and grown under controlled conditions. Shoots of the mutants had significantly lower biomass (Fig. 1A), approximately 17% and 32% of the wild type for lon1-1 and lon1-2, respectively, on a dry weight basis. Likewise, the roots of mutant seedlings grown on plates were significantly shorter (Fig.  1B), approximately 27% of wild-type seedlings.
Lon1 is localized to mitochondria in plants and its loss reported to lead to lower maximal activity of respiratory enzymes (Sarria et al., 1998;Rigas et al., 2009a). To test if these differences alter respiratory activity of shoots and roots, the oxygen consumption of shoot and root tissues isolated from hydroponically grown plants was assessed (Fig. 1,C and D). However, no significant differences in tissue respiration were found in either of the mutants.
The lon1-1 mutation, produced through ethyl methanesulfonate mutagenesis, results in a premature stop codon due to a point mutation at residue 807 in the 18th of 19 exons, while the lon1-2 mutant is a T-DNA insertion line, where the T-DNA resides also in the 18th exon, beginning at residue 730. It was previously postulated that the difference in severity of phenotypes between lon1-1 and lon1-2 could be due to partial translation of the Lon1 polypeptide in lon1-1 mutants (Rigas et al., 2009b). However, no information on the relative abundance of Lon1 in the mutants was available. Using multiple reaction monitoring (MRM), to a wide selection of Lon1 peptides, we were able to detect and quantify eight peptides derived from across the sequence of Lon1. The average abundance of Lon1 peptides in both lon1-1 and lon1-2 mutants is ,10% that of the wild type in mitochondria-enriched fractions, and the two mutants do not have a significant difference to each other in the abundance of these Lon1 peptides (Fig.  1E).
lon1 Mutation Alters the Abundance of a Diverse Set of Proteins in Mitochondria Previously, mitochondria from lon1 mutants were found to have decreased respiratory capacity of some pathways of ETC as well as decreased activity of key tricarboxylic acid (TCA) cycle enzymes (Rigas et al., 2009a). To determine if differences in protein abundance were responsible for both the retarded postgerminative growth phenotype as well as the decreased mitochondrial activity reported, mitochondria were isolated from 3-week-old hydroponically grown shoots of wild-type, lon1-1, and lon1-2 plants. We examined the differences between the mitochondrial proteomes using both two-dimensional (2D) fluorescence difference gel electrophoresis (DIGE) and isobaric tags for relative and absolute quantitation (iTRAQ). For both types of analyses, each lon1 mutant was compared with the wild type. While both mutants showed similar staining patterns on 2D gels when compared with the wild type, the DIGE analysis focused on alterations in the abundance of specific protein spots (Fig. 2). While some abundant proteins altered in abundance between replicates, most of the protein spots that reproducibly changed in abundance across replicates were low abundance protein spots (Fig. 2). These protein spots with fold changes in abundance of greater than 61 (P , 0.05) from both experiments were isolated, digested with trypsin, and identified by matrix-assisted laser-desorption ionization (MALDI) time-of-flight (TOF) tandem mass spectrometry (MS/MS). In total, 30 and 35 of the isolated protein spots could be identified from the lon1-1 and lon1-2 comparison experiments, respectively (Supplemental Table S2). iTRAQ labeling of peptides from trypsin digestions were performed in experiments where each mutant was compared separately to the wild type and labeled peptides were identified by liquid chromatography (LC)-MS/MS. Based on the iTRAQ data, 50 proteins with significant differences Figure 1. Growth habit, biomass, tissue respiratory capacity, and Lon1 abundance in lon1 mutants of Arabidopsis. A, Phenotype and dry weight of 3-week-old aerial tissues (n = 10; mean 6 SE). B, Root length of 15-d-old seedlings (n = 15; mean 6 SE). C, Shoot respiration rates (n = 13; mean 6 SE). D, Root respiration rates (n = 13; mean 6 SE). E, Average peptide abundance in MRM transitions (n = 6; mean 6 SE). Significant differences compared with the wild type are indicated: *P # 0.05 and #P # 0.01. FW, Fresh weight; WT, the wild type.
in abundance were found in lon1-1 and 24 with significant differences in abundance were found in lon1-2, relative to the wild type (Supplemental Table S3). DIGE average ratios and iTRAQ reporter ratios were then converted to fold change to allow for direct comparison. Overall, the results from the different analysis methods were in good agreement regarding the specific proteins or the enzyme complex that changed in abundance as well as in the fold changes compared with the wild-type levels of those same proteins. In total, 84 unique proteins were found to change across genotypes and techniques (Table I).
In both lon1-1 and lon1-2 mutants, subunits from several components of the ETC were decreased in abundance. Subunits from Complexes I, III, and IV and the ATP synthase complex were affected. In the case of Complex I (NADH-ubiquinone oxidoreductase), a total of 11 different subunits (seven in lon1-1 and six in lon1-2) were identified as quantitatively lower in abundance in the mutants than in wild-type plants.
Most of these subunits are thought to be situated either in the matrix arm or in the plant-specific matrix domain of Complex I (Meyer et al., 2011). Nine different subunits of the ATP synthase complex (from both the F 1 and F o subcomplexes) showed changing abundances in the mutants when compared with the wild type. In general, most of the changing subunits were decreased in abundance in the mutants. Isoforms of all nine identified ATP synthase subunits were lower in lon1-1, while seven of the identified subunits are lower in lon1-2. Two subunits of Complex III (cytochrome bc 1 ) were determined to be of lower abundance in lon1-2. Cytochrome-c oxidase (Complex IV) core subunits are hydrophobic and very underrepresented in proteome data sets. However, we did find evidence of less of the Complex IV substrate cytochrome c, the peripheral subunit Cox6b was also lower in lon1-1 and lon1-2, and two isoforms of Cox5b were found to be slightly more abundant in lon1-1 (ETC Associated; Table I).
In addition to the changes in protein abundance in the ETC, another major trend to emerge was the number of enzymes in the TCA cycle whose abundances increased in the mutants (Table I). Between both DIGE and iTRAQ data, lon1-1 had six different TCA cycle enzyme subunits that were increased in abundance, including mitochondrial lipoamide dehydrogenase (At1g48030), dihydrolipoamide S-acetyltransferase (At3g13930), malate dehydrogenase (At1g53240), isocitrate dehydrogenase (At3g09810), succinyl-CoA ligase (At2g20420), and 2-oxoglutarate dehydrogenase (At3g55410). The lon1-2 mutant also had enzymes increased in abundance, including two isoforms of isocitrate dehydrogenase (At4g36260 and At5g03290) and two isoforms of 2-oxoglutarate dehydrogenase (At5g65750 and At3g55410). Interestingly, several enzymes that were identified in both DIGE and iTRAQ methods and can be thought of as metabolically adjacent to the TCA cycle, such as Glu dehydrogenase (At5g18170 and At5g07440), formate Arabidopsis mutants. Samples of mitochondria isolated from shoots of hydroponically grown lon1 mutants are compared with Col-0. A, lon1-1 (labeled with Cy3) compared with Col-0 (labeled with Cy5); yellow protein spots represent proteins of equal abundance between genotypes, while protein spots more abundant in lon1-1 mutant are green and protein spots more abundant in Col-0 are red. B, lon1-2 (labeled with Cy5) compared with Col-0 (labeled with Cy3); protein spots more abundant in lon1-2 mutant are red, and protein spots more abundant in Col-0 are green. Arrows indicate proteins unambiguously identified by MS; the numbers correlate with Table I. MW, Molecular weight.  Table S2) and iTRAQ (Supplemental Table S3) were identified by MS/MS. Accession number (Arabidopsis Genome Initiative), protein name, and protein abundance differences are shown. When the latter are pink, they are significantly less abundant in the mutant (P , 0.05), and when they are green, they were significantly more abundant in the mutant (P , 0.05). "X" indicates identification but without the ability to define a ratio in the iTRAQ experiment. "-" indicates an absence of a protein spot matching to this protein.
A number of heat shock proteins and other proteins involved in the stress response of plants were found in greater abundance in both mutants compared with the wild type. Most notably, in both mutants, HSP60 (At2g33210 and At3g23990) and HSP70 (At4g37910 and At5g09590) were increased in abundance according to DIGE and iTRAQ experiments. Other mutant-induced stress-related enzymes were monodehydroascorbate reductase (At1g63940) in lon1-2 and prohibitins (1, 3, and 4; At4g28510, At5g40770, and At3g27280) and peroxiredoxin (At5g06290) in lon1-1.

ETC Components and Activities Are Lower in lon1 Mutants
As our discovery-based quantitative proteomics approach identified changes in the abundance of several components of the mitochondrial ETC, we sought to confirm these proteomic changes in a number of ways. First, replicate mitochondrial protein extracts isolated from hydroponically grown shoots of mutant and wild-type plants were separated by SDS-PAGE and transferred to nitrocellulose membranes. Membranes were then probed with antibodies against a number of mitochondrial ETC subunits for Complex I (NDUFA1, CA2, and NDUFS4) and ATP synthase (a and b). In all cases except ATP synthase a-subunit, the relative protein amounts as determined by densitometry were decreased in the mutants when compared back to the wild type (Fig. 3, A and B). Porin, used as a loading control, remained unchanged between genotypes.
Given that there was no difference in overall tissue respiration rate of either roots or shoots ( Fig. 1, C and D) but that many components of the mitochondrial respiratory complexes were decreased in abundance (Table I), we also examined the respiratory capacity of ETC complexes in isolated mitochondria. The lon1 mutants were previously shown to have significantly decreased oxygen consumption in Complexes II and IV (Rigas et al., 2009a). As our proteomic evidence suggested significant alterations in Complexes I and IV, we examined Complexes I, II, and IV for altered Table I. (Continued from the previous page.) Figure 3. Mitochondrial respiratory complex activity, abundance, and subunit abundance of lon1 Arabidopsis mutants. A, Representative images of immunodetection of mitochondrial ETC subunits in SDS-PAGE separations of purified mitochondria. B, Relative protein amount of immunodetected mitochondrial ETC subunits (n = 3; mean 6 SE). C, Oxygen consumption of purified mitochondria (n $ 6; mean 6 SE). D, SDH activity of purified mitochondria (n $ 3; mean 6 SE). E, Relative abundance of mitochondrial ETC complexes determined from Coomassie-stained blue native-PAGE separations of isolated mitochondrial proteins (n = 3; mean 6 SE). Significant differences compared with the wild type (WT) indicated: *P # 0.05 and #P # 0.01.
activities. The oxygen consumption of Complex I was measured by using the selective substrate deamino-NADH. Mitochondria of both lon1 mutants showed lower deamino-NADH-dependent oxygen consumption rates than the wild type (Fig. 3C). The oxygen consumption rate of Complex IV examined in isolated mitochondria was also significantly lower in both lon1 mutants when compared with the wild type (Fig. 3C). Because of the large effect that decreased Complex IV activity had in the mitochondria of mutant plants, and as that large effect would have a consequent effect on all other oxygen consumption measurements, we chose to examine the activity of Complex II by monitoring the reduction of 2,6-dichlorophenol-indophenol (DCPIP) by direct succinate dehydrogenase (SDH) activity assay. Complex II activity in isolated mitochondria was significantly decreased in both mutants when compared with the wild type, but to a lesser extent than Complexes I and IV (Fig. 3D).
As our results show impairment of mitochondrial respiratory complexes in the mutants as well as decreased abundances in a number of ETC subunits, we examined the overall integrity of mitochondrial respiratory complexes using blue native-PAGE. In general, the abundance of the individual complexes did not appear to change significantly between the wild type and lon1-2 for the complexes examined. When compared with lon1-1, the average band intensities were often lower, but only the band corresponding to Complex I was statistically significantly lower in abundance compared with the wild type (Fig. 3E).

The TCA Cycle Composition and Its Function Is Affected in lon1 Mutants
To confirm changes to the TCA cycle and related enzymes observed by DIGE and iTRAQ, we performed western analysis on SDS-PAGE-separated mitochondrial proteins using antibodies against formate dehydrogenase and lipoamide dehydrogenase (E3). While the abundance of formate dehydrogenase decreased in the mutants according to DIGE and iTRAQ, the abundance by western analysis did not significantly change (Fig. 4, A and B). Analysis of the position of the spots increasing in the DIGE analysis ( Fig. 2) revealed many were breakdown products that decreased in abundance in the lon1 mutants. The abundance of lipoamide dehydrogenase (E3) on western blots trended up in the mutants but was not significantly changed across the biological replicates (P , 0.05). Porin was used as a loading control and remained unchanged from mutant to the wild type. We also tested for changes in aconitase, which is a known substrate for Lon1 in yeast (Bota and Davies, 2002;Ngo and Davies, 2007). Analysis of aconitase changes in DIGE gels is complicated by its appearance as a long chain of spots close to other major enzymes at a high molecular mass where there is little size separation. Aconitase significantly increased in abundance in the mutants (Fig. 4, A and B) even though it was not identified as a protein changing in abundance in the iTRAQ datasets.
Upon further analysis of DIGE gel spots from the TCA cycle enzymes, it was observed that some of the significantly different DIGE-identified TCA cycle enzymes were based on protein spots that arose from acidic shifts in pI. An acidic shift can be due to oxidative modification of a protein (Rabilloud et al., 2002). Malate dehydrogenase showed the greatest difference in pI shift from an expected pI of the mature protein of approximately 7.0 to an observed pI of approximately 5.5 for the spot 13 found to change in lon1-1 (Fig. 2). This indicates that some of the changes observed in the TCA cycle may be posttranslational changes in only a portion of the protein, rather than evidence for up-or down-regulation of the protein abundance as a whole.
An altered abundance, regulation, or oxidative state of TCA cycle enzymes could have an effect on metabolic flux in the affected plants. Given this, we examined the metabolomes of both lon1 mutants against the wild type by total metabolite pool analysis of hydroponically grown shoot material isolated at both midday (.4 h light) and midnight (.4 h dark). The total pools of a number of organic acids were found to be lower in lon1-1 than in the wild type plants in the light and lower in both mutants than the wild type in dark collected tissues. The TCA cycle intermediates fumarate, citrate, succinate, and malate were most decreased in abundance, notably in lon1-1 (Fig. 4C). g-Aminobutyrate (GABA), closely associated with the TCA cycle through its involvement in the GABA shunt, was also found to be lower in both mutants than in the wild type in both light and dark conditions (Fig. 4C).

Lon-1 Loss Enhances Stress Responses But Not Oxidative Damage Accumulation in Mitochondria
Western analysis using antibodies raised against HSP60, HSP70, AOX (Fig. 5A), and lipoic acid (Fig. 5B) were performed on SDS-PAGE separated mitochondrial proteins to confirm results from quantitative proteomic analyses and further examine the extent of oxidative modification in the plants. Both DIGE and iTRAQ showed elevated abundances of HSP60 and HSP70; however, while the mean levels of these proteins increased in western analysis, these changes were not statistically significant (P , 0.05). But it should be noted that multiple HSP60 and HSP70 isoforms exist in Arabidopsis mitochondria and several could be being detected by the antibodies. The oxidative stress-induced marker protein AOX was lower in abundance in the mutants, although only significantly in lon1-1. We have previously shown that lipoic acid moieties on several mitochondrial proteins are targets for damage by oxidative stress from chemical treatments and environmental conditions (Taylor et al., 2002;Winger et al., 2007). Upon examination of the 17-and 78-kD bands of the antilipoic acid immunoblot that correspond to Gly decarboxylase H-protein (GDC-H) and dihydrolipoamide acetyltransferase (PDC-E2), respectively, no difference in the abundance of lipoate moieties could be determined between the mutants and the wild type (Fig. 5B).
Because Lon1 is believed to selectively degrade oxidatively damaged proteins (Bota and Davies, 2002;Bayot et al., 2010) we wanted to examine if the small growth phenotype of the lon1 mutant plants might be caused in part by an accumulation of oxidatively damaged proteins. To assess this, we examined the extent of protein carbonylation, a common measure of accumulated oxidative damage to plant proteins (Johansson et al., 2004;Gibala et al., 2009). Densitometry of the total SDS-PAGE-separated protein lanes revealed no overall differences in the amount of total carbonylation between wild-type and lon1 mutant samples (Fig. 5C).
A detailed examination of oxidatively modified peptides within the iTRAQ experiments was performed as outlined in "Materials and Methods." This showed no significant increase in oxidation in the mutant plants when all mitochondrial peptides were considered together. In lon1-1 plants, of the total quantified oxidatively modified peptides of mitochondrial proteins, only 86 (40%) were more abundant in lon1-1 compared with 127 (60%) that were more abundant in the wild type. Similarly, in the lon1-2 plants, 46 (43%) of the total quantified oxidatively modified peptides more abundant in lon1-2 compared with 61 (57%) in the wild type. The oxidized peptides were broadly distributed across 36 proteins in lon1-1 and 21 proteins in lon1-2. However, it was notable in several cases that the ratios were focused in one direction (Supplemental Table S4). All 14 of the reported spectra matched to oxidized malate dehydrogenase peptides were more abundant in the mutants than in the wild type (12 of them in lon1-1), four of the five oxidized peptides of HSP60 were in lon1-1 mutants, and six of the nine oxidized peptides in aconitase were more abundant in lon1-1. In most other cases, oxidized peptides were evenly distributed or biased toward being more abundant in wild-type plants (Supplemental Table  S4). Specific analysis of the peptides for aconitase did not reveal evidence for accumulation of the oxidationmodified peptides observed during chemical induced oxidative inactivation of the enzyme (Tan et al., 2010). Oxidation of HSP60 in lon mutants is consistent with reports in yeast (Bayot et al., 2010). Taken together, these data suggest that the loss of Lon1 protease does not contribute to an overall increase in oxidative stress in the mitochondria of Arabidopsis plants but may contribute to elevation of oxidation status for some specific proteins, most notably in lon1-1.

DISCUSSION
Difference between lon1-1 and lon1-2 in Arabidopsis Previous reports have suggested that differences in the nature of the mutation in lon1-1 and lon1-2 might explain the apparent difference in the severity of the phenotypes of these plants (Rigas et al., 2009b). Our own phenotyping confirms these differences between the mutants (Fig. 1, A and B). However, the predicted presence of a mutated protein in lon1-1 that might be responsible for phenotypes of the plants had not been confirmed. Resolving this issue was important if these mutants were to be used to understand the role of Lon1 in plants.
MRM is a sensitive technique and can determine the relative amounts of specific peptides within a complex protein mixture in plants (Wienkoop and Weckwerth, 2006), and we previously used this approach to observe changes in mitochondrial carrier proteins in plants . Using specific Lon1 peptides as targets, we have shown that mitochondrial extracts of both the lon1-1 and lon1-2 mutants have ,10% of the relative amount of Lon1 protein than do mitochondria from wild-type plants (Fig. 1E). In total, eight different peptides, all unique to Lon1 and dispersed throughout the sequence, were analyzed and their average relative response reported. While lon1-1, but not lon1-2, plants synthesize a transcript of the Lon1 gene (Rigas et al., 2009a), it is clear from our MRM data that the transcript from the lon1-1 plants is not processed into polypeptides and accumulated more than the background signals observed in lon1-2. While it is clear the primary defect in both mutants is the loss of the Lon1 protein, we can't exclude a very Figure 5. Abundance of stress-related proteins and analysis of oxidation of Arabidopsis lon1 mutants. A, Relative protein amount of immunodetected stress-related proteins (n = 3; mean 6 SE) and representative images of immunodetection of stress-related proteins in SDS-PAGE separations of purified mitochondria. B, Relative amount of lipoate moieties in PDC-E2 and GDC-H protein of purified mitochondria (n = 3; mean 6 SE) and representative images of SDS-PAGE separated mitochondria immunodetected with antibodies to lipoic acid. C, Relative amount of carbonylation in purified mitochondria (n = 3; mean 6 SE) and representative lanes of SDS-PAGE separated mitochondria immunodetected with carbonyl antibodies. Significant differences compared with the wild type (WT) indicated: *P # 0.05. low abundance mutant Lon1 product in lon1-1 with a detrimental impact.
The clearest molecular evidence we have separating lon1-1 and lon1-2 is the significantly lower organic acid pools in lon1-1 during the day (Fig. 4C), the clearly lower Complex I assembly in lon1-1 (Fig. 3), and the greater number of oxidized peptides in lon1-1 samples (Supplemental Table S4). On general protein abundance measures (Table I) and many activity measurements ( Fig. 3; Rigas et al., 2009a), the two mutants could not be easily distinguished.
The Importance of Lon1 to the Respiratory Apparatus and Viability The absence of Lon protease in Arabidopsis plants results in a small growth phenotype in both roots and shoots (Rigas et al., 2009a;Fig. 1, A and B), but as the mutants are able to successfully develop to maturity and generate viable seed, this protease is clearly not essential for plant survival. Aberrant growth phenotypes of Lon mutants have also previously been observed in yeast (Suzuki et al., 1994;Van Dyck et al., 1994) and bacteria (Donch and Greenberg, 1968). The nonessential nature of Lon1 has been reported in many organisms (Tsilibaris et al., 2006), indicating the role of this protease is at least partly functionally redundant within the protease network in the cell and probably also within the protease network of mitochondria themselves.
Loss of Lon1 in yeast (Suzuki et al., 1994;Van Dyck et al., 1994) and in human lung fibroblasts (Bota et al., 2005;Ngo and Davies, 2007) is reported to contribute to significant overall respiratory defects in these organisms. Despite the effect that the loss of the Lon1 protein has on overall growth in Arabidopsis, no effect on the respiratory activity of root or shoot tissues was observed (Fig. 1, C and D). However, it is clear that maximal respiratory capacity is affected when substrate saturated assays are conducted on isolated mitochondria of Lon1 mutants in Arabidopsis (Rigas et al., 2009a;Figs. 3 and 4). Our quantitative proteomic analysis provides evidence that these defects do operate, at least partially, at the level of protein abundance differences. The subtlety of the differences noted is consistent with the small degrees of change in the overall proteome noted in similar comparisons of mutants in the yeast homolog PIM1 (Major et al., 2006;Bayot et al., 2010). We show that several subunits of mitochondrial ETC and OXPHOS are significantly decreased in mitochondria of lon1 mutants when compared with wild-type mitochondria. Most notable is the decrease in the abundance of protein subunits in Complexes I, III, IV, and V of OXPHOS (Table I), which is likely to directly impact the oxygen consumption capacity of mitochondrial samples. Confirmation of these protein abundance changes with western blots and activity assays is consistent with the published activity results (Rigas et al., 2009a) for Complex IV but also shows significant decrease in Complex I at the protein, activity, and assembly levels. Absence of the Lon1 homolog in yeast has been shown to affect the expression and abundance of several elements of the cytochrome c oxidase complex (Complex IV) and cytochrome b/c1 complex (Complex III), including cytochrome a, cytochrome a3, cytochrome b (Van Dyck et al., 1994), COX1 and COB (van Dyck et al., 1998), cytochrome b5 (Major et al., 2006), QCR2, RIP1, and COX4 (Bayot et al., 2010). COX expression is also impaired in the human Hslon mutants (Hori et al., 2002;Fukuda et al., 2007). ATP synthase subunits (Complex V) are affected in yeast lon mutants, including ATP1, ATP2, and ATP7 (Suzuki et al., 1994;Major et al., 2006;Bayot et al., 2010). To our knowledge, this is the first evidence of an impact of Lon1 loss on Complex I of the respiratory chain.

The Protease versus Chaperone Function of Lon1
Transcripts of both mitochondrial-and nuclearencoded subunits of Complexes III and IV remained unchanged in lon1 mutants (Rigas et al., 2009a), and yeast PIM1 mutants showed no difference in transcript levels of COX1 and COB (Van Dyck and Langer, 1999). This indicates that Lon1 acts directly or indirectly to alter the abundance of selected proteins posttranscriptionally (and most likely following translation). It has been suggested that at least some of the respiratory enzymes that are altered in abundance in lon1 mutants are direct targets of Lon (Van Dyck and Langer, 1999;Hori et al., 2002;Major et al., 2006;Fukuda et al., 2007). If this is the case in Arabidopsis and Lon1 is acting predominately via proteolysis, we might reasonably expect protein targets to have higher abundances in mutants lacking Lon1. In our data, a range of respiratory chain subunits were decreased in abundance and mitochondria had concurrently lowered activities, while a number of TCA cycle enzymes had higher abundance (Fig. 3) while mitochondria had concurrently lowered activities (Rigas et al., 2009a;Fig. 3). In the case of mitochondrial proteases, of which many seem to possess both proteolytic and chaperone activities (Janska, 2005), these differences could be the result of several scenarios. First, it could be suggested that ETC subunits experience high turnover rates when some downstream binding partner is accumulating due to the lack of a proteolytic event. Alternatively, ETC subunits could be targets of Lon1's chaperone function and the loss of Lon1 could then result in lower abundance as ETC partners cannot be assembled properly into respiratory complexes, which would be consistent with a number of our data sets (Fig.  6). In the case of TCA cycle enzymes, increased abundance with lowered activity is consistent with a more direct role of Lon1 in degradation of damaged or misfolded protein subunits (Fig. 6). Our quantitative proteomic analysis of the mitochondria from lon1 mutant Arabidopsis plants showed increases in specific chaperones of the HSP60, HSP70, and prohibitin classes. These could be viewed as compensatory changes due to loss of the Lon1 chaperone activity or to a general stress response in the mitochondrial matrix and membrane proteome due to the inhibition of proteolysis. In yeast, prohibitins and HSP60 also increase in lon mutants, the former were shown to also be binding partners of Lon by TAP-tagging, while the latter was oxidized and accumulated in mutants (Bayot et al., 2010).
The other plant mitochondrial proteases that have been studied in detail with protease and chaperone functions are from the FtsH class (Kolodziejczak et al., 2007;Gibala et al., 2009). Those authors have shown that Arabidopsis deficient in FtsH4 also suffers from decreased activity of respiratory complexes, specifically Complexes I and V. In FtsH mutants, the decreased abundance of subunits of Complex V and increased HSP70 and prohibitins has also been confirmed. Transcripts of ATP6 were unchanged in FtsH4 mutants, while the protein decreased in abundance (Gibala et al., 2009). Overall, the response in Lon1 mutants clearly shares some similarities with the mitochondrial FtsH mutants.

Lon1 Loss Alters TCA Cycle and Cellular Carbon Metabolism But Not Protein Oxidation
Our analysis of the mitochondrial proteomes of the two lon1 mutants showed increases in the abundance of several enzymes involved in the TCA cycle. Using western blots of SDS-PAGE separated mitochondrial proteins, we also found accumulation of aconitase. Aconitase activity was significantly lower in lon1 mutants (Rigas et al., 2009a). Aconitase has long been thought to be a direct target/substrate of Lon1, and it has been proposed that Lon1 is responsible for degrading oxidized aconitase polypeptides (Bota and Davies, 2002). It would follow that if oxidized aconitase is a Lon1 substrate, then the absence of Lon1 protein might result in accumulation of nonfunctional aconitase. When we examined the level of TCA cycle intermediates in lightand dark-collected leaf tissues, a number of these organic acids were decreased in relative abundance in both lon1 mutants. Most notably, these were intermediates in the second half of the TCA cycle, namely, succinate, fumarate, and malate (Fig. 6). It was previously shown that the activities of several TCA cycle enzymes in the first half of the TCA cycle were decreased in Arabidopsis lon1 mutants, notably citrate synthase, aconitase, isocitrate dehydrogenase, and 2-oxoglutarate dehydrogenase (Rigas et al., 2009a). Accumulation of TCA cycle enzymes and their subsequent modifications observed on 2D gels (Fig. 2) might explain increased protein abundance despite specific activities of the enzymes being lower (Rigas et al., 2009a). The lowered abundance of GDH and 4-aminobutyrate transaminase (Table I) and the low abundance of GABA (Fig. 4C) suggest that the GABA shunt may be decreased in activity, thus not providing a bypass of the lowered 2-OGDC activity (Rigas et al., 2009a) or an alternative source of succinate for TCA cycle operation (Michaeli et al., 2011). However, direct assay of GABA shunt flux in vivo is required to confirm this.
As Lon1 has long been believed to be involved in the removal of oxidatively damaged proteins from the mitochondrial matrix (Bota and Davies, 2002;Ngo and Davies, 2007;Bayot et al., 2010;Bender et al., 2010) and given the common aberrant growth phenotypes of lon1 mutants across kingdoms, it has generally been thought that the loss of Lon1 would contribute to an increase in overall protein oxidation in plant mitochondria. We Figure 6. Diagram of the impact of Lon1 loss on mitochondrial metabolism. Red metabolites indicate organic acid pools decreasing in mutants (Fig. 4C). Red or green enzyme names indicate lower or higher protein abundance (Table I;  . Red arrows indicate evidence of loss of enzyme activity (Figs. 2-4;Rigas et al., 2009a).
found peroxiredoxin and MDHAR increasing in abundance in iTRAQ or DIGE in lon1 mutants (Table I), which could be analogous to the Mn-superoxide dismutase accumulation observed in yeast lon mutants (Bayot et al., 2010). But subsequent examination of the carbonylation patterns in isolated mitochondria from lon1 mutant plants, the quantified abundance of lipoic acid moieties on mitochondrial proteins, the quantified oxidation of peptides in the proteomes of lon1 mutants, and the expression of the classic oxidative stress marker of mitochondria (AOX), all showed no evidence of heightened oxidative stress. While we have some evidence of enhanced oxidation of a small number of proteins (Supplemental Table S3), our data as a whole do not currently support the hypothesis of an overall increase in the steady-state oxidation of the plant mitochondrial matrix environment associated with Lon1 loss. This is in contrast to the data presented for the knockout of FtsH4 that did claim enhanced steadystate carbonylation of the Arabidopsis mitochondrial proteome as a whole (Gibala et al., 2009).
To extend this analysis and make it more comparative at the protein level, we aligned data on protein degradation in oxidative stress of mitochondria in Arabidopsis (Sweetlove et al., 2002) with our data on changes in lon1 mutants (Table I) and included comparisons to quantitative proteomics of the yeast PIM knockout (Major et al., 2006) and related analysis on protein degradation in oxidative stress of mitochondria in yeast (Bender et al., 2010). This shows that while there are some proteins identified in all these sets, there is only a moderate correlation between the direction of the responses in yeast and plant mitochondria during oxidative stress, but some overlap between the proteins that change in abundance in lon mutants between Arabidopsis and yeast (Supplemental Table S5). The similarities include changes in prohibitin abundance in lon mutants in species; changes in HSP60, ATP1, and ATP2 in both lon and pim mutants and oxidative stress in Arabidopsis; changes to lipoamide dehydrogenase, succinyl CoA, aconitase, and aldehyde dehydrogenase abundance in both Arabidopsis and yeast; and increase in peroxiredoxin in both lon and pim mutants along with changes in its abundance in both species during oxidative stress (Supplemental Table S5). To directly determine if these changes in Arabidopsis proteins are evidence that they are direct substrates of Lon1, or an indirect consequence of Lon loss, a series of experiments would be required. This could include import of specific proteins into isolated mitochondria and assessing their degradation rates in the wild type and mutant, in vivo studies of protein turnover using radio-or stable isotopes, plus in vitro direct assays of proteolysis of pure natively folded putative targets using purified Lon1. To date, this evidence is not available for any one protein in mitochondria, but combinations of evidence for changes in lon mutants, evidence of oxidation, and some targeted assays of degradation (Supplemental Table S5) are available. However, the clarification of the Arabidopsis proteins that change in abundance in lon1 mutants provides the necessary foundation to define the subset that is Lon targets in the future.

CONCLUSION
Through combining data on the protein abundances, protein complex assembly and activities, oxidative modifications, and organic acid pool sizes (Figs. 1-5) with phenotypic and enzymatic data (Rigas et al., 2009a), we can now consider in depth the impact of Lon1 loss on mitochondrial function and thus attempt to unravel the role of this chaperone and protease. The lack of anaplerotic functioning of the TCA cycle leading to depletion of organic acid pools could be an important determinant of the slow growth rates of the mutants (Fig. 6), despite whole-tissue oxygen consumption rates in Lon1 mutants being close to normal. Enhanced oxidative stress appears to be unlikely to be able to explain the mutant phenotypes. Distinct differences of lon1 to the molecular phenotypes of ftsh in Arabidopsis (Gibala et al., 2009) and lack of evidence for oxidative impairment suggest a different role in the plant mitochondrial protease network for Lon1. Overall, the broadness of the impact (Fig. 6) suggests that Lon1 has a wide variety of targets and likely plays roles as both a chaperone for membrane complexes and a protease for soluble enzymes to optimize the role of plant mitochondria in cellular organic acid metabolism.
Biomass measurements were made on soil-grown plant material. Wetted (deionized water) seeds were stratified for 2 d at 4°C in the dark and planted on a soil mix (3:1:1; compost: vermiculite:perlite) and grown under long-day conditions (as above). Rosettes were removed from 3-week old plants and dried for at least 48 h at 65°C before dry weight determination. For root length measurements, 48-h stratified (4°C in the dark) seeds were germinated on plates (onehalf-strength Murashige and Skoog, 1% [w/v] Suc, and 0.8% [w/v] agar) in long-day conditions. After 5 d, seedlings were transferred to new agar plates (1% [w/v] agar) and stood upright to promote straight root elongation. After a further 7 d, root length was measured from photos using ImageJ software version 1.44 (Abramoff et al., 2004).

Measurements of ETC Function
Respiration measurements of hydroponically grown root and shoot tissues were carried out in the dark at 25°C in a Clark-type oxygen electrode (Hansatech Instruments) containing 1 mL of respiration buffer (5 mM KH 2 PO 4 , 10 mM TES, 10 mM NaCl, 2 mM MgSO 4 , and 0.1% [w/v] BSA, pH 7.2). Respiration measurement of isolated mitochondria was performed as above with the addition of 0.3 M Suc to the respiration buffer. Different substrates, cofactors, and inhibitors were added to the reaction chamber to assess the oxygen consumption of Complexes I and IV.
For Complex I-specific NADH-dehydrogenase activity, 1 mM deamino-NADH was used as substrate. The reaction was inhibited with the addition of 5 mM rotenone. Complex IV (cytochrome c oxidase [COX]) activity was measured by the addition of reduced cytochrome c (10 mM ascorbate and 2.5 mM cytochrome c). The reaction was facilitated by the addition of Triton X-100 (0.04% [v/v]) and inhibited with 1 mM KCN. The direct activity of Complex II (SDH) was measured spectrophotometrically by monitoring the reduction of DCPIP  with modifications. Briefly, 20 mg of mitochondrial proteins were added to 1 mL reaction solution (10 mM K 2 HPO 4 , pH 7.4, 100 mM EDTA, 0.1% BSA [w/v], 10 mM KCN, and 10 mM sodium succinate) in a 1-mL cuvette. The absorbance was offset after with the addition of 1.6 mM phenazine methosulfate prior to the addition of 120 mM DCPIP and the absorbance monitored at 600 nm at 25°C in a UV-1800 spectrophotometer (Shimadzu).

Quantitative Mitochondrial Proteomics
For 2D DIGE-SDS-PAGE and MALDI-MS/MS analysis, mitochondrial proteins from wild-type (Col-0) and mutant (lon1-1 and lon1-2) plants were acetone precipitated overnight at 220°C and pelleted at 20,000g for 20 min at 4°C. Pellets were resuspended in lysis buffer (8 M urea, 40 mM Tris, and 4% [w/v] CHAPS) and centrifuged at 20,000g for 10 min to remove unsolubilized material. Biological replicate samples (50 mg) were labeled on ice, in the dark for 30 min with the addition of 400 nmol of CyDye (Cy3 or Cy5) according to the manufacturer's instructions (GE Healthcare). A reference standard was made by combining equal amounts of all samples and labeled with Cy2. Two independent experiments were performed comparing the wild type to either of the lon1 mutants. Pooled samples were loaded on a 24-cm, pH 3 to 10 nonlinear Immobiline DryStrip (GE Healthcare). Strips were rehydrated and first-dimension isoelectric focusing (IEF) was performed in an IPGPhor (GE Healthcare). Following IEF, strips were successively equilibrated for 15 min each in equilibration buffer (6 M urea, 50 mM Tris, pH 8.8, 2% [w/v] SDS, 26% [v/v] glycerol, and bromophenol blue) plus 65 mM dithiothreitol and then 135 mM iodoacetamide. Strips were briefly rinsed in 1.5 M Tris (pH 8.8) with 1% SDS prior to layering on top of a 12% [w/v] acrylamide gel. Following separation, gels were imaged with a Typhoon scanner (GE Healthcare) and the images compared using DeCyder software version 6.5 (GE Healthcare).
A 1-mg protein sample made in the same sample ratios as the Cy2 standard sample was used for a preparative gel for each comparison and run in the manner described above. Gels were stained with colloidal Coomassie, and spots of interest identified through DIGE analysis were excised. Spots were digested overnight at 37°C with trypsin (Invitrogen). Extracted peptides were dried in a vacuum centrifuge and resuspended in 5 mL of 5% acetonitrile with 0.1% trifluoroacetic acid (TFA). Peptides were analyzed with an UltraFlex III MALDI-TOF/TOF mass spectrometer (Bruker Daltonics). Samples were mixed 1:1 with spotting matrix (90% ACN and 10% saturated a-cyano-4hydroxycinnamic acid in TA90 [90% ACN and 0.01% TFA]) on an MTP 384 MALDI target plate. Spots were allowed to dry completely before a 10-s wash with 10 mL cold washing buffer (10 mM NH 4 H 2 PO 4 and 0.1% TFA). Spots were analyzed at 50% to 85% laser intensity with up to 1200 shots for mass spectrometry (MS) analysis per spot. Ions between 700 and 4,000 mass-to-charge ratio were selected for MS/MS experiments using 3% additional laser power. MS/MS data were analyzed using Biotools (Bruker Daltonics) and an in-house Arabidopsis database comprising ATH1.pep (release 9) from The Arabidopsis Information Resource and the Arabidopsis mitochondrial and plastid protein sets (33,621 sequences; 13,487,170 residues), using the Mascot search engine version 2.3.02 and error tolerances of 61.2 D for MS and 6 0.6 D for MS/MS; "max missed cleavages" set to 1; variable modifications of oxidation (M) and carbamidomethyl (C).
For iTRAQ labeling of whole mitochondrial protein samples and LC-MS/ MS analysis, total mitochondrial protein samples (100 mg) were precipitated with six volumes of cold acetone. Following recovery, the proteins were resolved in dissolution buffer, denatured, reduced, and had Cys residues blocked according to the manufacturer's instructions (AB Sciex). Samples were then digested overnight (37°C) with 10 mg trypsin (Invitrogen). Peptides from triplicate samples were then labeled with three different iTRAQ isobaric tags, after which samples were pooled. Excess label and interfering salts were removed by diluting the pooled sample four times in a buffer (10 mM KH 2 PO 4 in 25% [v/v] ACN, pH 3.0) and subjecting it to strong cation exchange chromatography along an OPTI-LYNX cartridge (Optimize Technologies). The eluted fraction was dried in a vacuum centrifuge and resuspended in 5% (v/v) ACN and 0.1% (v/v) formic acid.
iTRAQ-labeled samples were analyzed on an Agilent 6510 quadrupole-TOF mass spectrometer using an HPLC Chip Cube source driven by an Agilent 1100 nano/capillary HPLC. Chips housed a 160-nL Zorbax 300SB-C 18 5-mm enrichment column and a Zorbax 300SB-C 18 5-mm, 150-mm separation column. Peptides were loaded onto the enrichment column at 4 mL/min in 5% ACN and 0.1% (v/v) formic acid and then separated and eluted along a 60min gradient (10%-45% ACN with 0.1% formic acid) at 300 nL/min. The Q-TOF was run in positive mode and MS scans collected from mass-to-charge ratio 250 to 1,400 at four spectra/s. Precursor ions were selected for auto MS/ MS when an absolute threshold of 200 and a relative threshold of 0.01 were met, with maximum three precursors per cycle and active exclusion set to two spectra and released after 3 min. Charge state selection of precursors was set to 2 + and 3 + . The resulting MS/MS spectra were searched against an in-house Arabidopsis database comprising ATH1.pep (release 9) from The Arabidopsis Information Resource, using Mascot version 2.3.02 (Matrix Science) with the following settings: MS and MS/MS error tolerances of 100 ppm and 0.7 D, respectively, maximum missed cleavages set to 1, methylthio (C), iTRAQ8plex (N-term), and iTRAQ8plex (K) as fixed modifications and carbimidomethyl (C), oxidation (M), and iTRAQ8plex (Y) as variable modifications, peptide charge as .2 + , and the instrument set as electrospray ionization-Q-TOF. The resulting searches were exported and all peptides identified (P , 0.05) were extracted to create an exclusion list for subsequent LC-MS/MS runs. Successive runs were combined using mzDataCombinator v.1.0.6 (West Australian Centre of Excellence in Computational Systems Biology) and submitted for database searching as outlined above. A total of 10 runs were performed, successively combined, and database searched. Two independent experiments were conducted comparing each of the lon1 mutants to the wild type. Quantitation of peptides and proteins was performed on the final concatenated file using the Mascot settings outlined above. Individual peptide ratios were reported for matched peptides and were combined using a weighted average approach for the protein ratio. Outlier removal was set to none. Ratios for proteins with normal distributions are reported as geometric means with SD. For non-normally distributed proteins, geometric means of the ratios were calculated manually using the individual peptide ratios and significance determined by t test (P # 0.05) using Analyze-It version 2.22.

MRM to Measure Lon1
Total mitochondrial samples (50-250 mg) as prepared above were digested overnight at 37°C with 1 mg of trypsin per 10 mg of protein. Extracted peptides were dried in a vacuum centrifuge, resuspended in 5% ACN and 0.1% formic acid to a final concentration of 1 mg/mL, and cleaned up with a 0.22-mm macro spin column (Millipore).
An in-house database of all mitochondrial matrix peptides identified previously by HPLC-ChipCube Q-TOF analysis in our lab was queried for previously identified lon1 peptides. For selection, peptides needed to be unique to lon1 (At5g26860). A data file containing all Q-TOF-acquired peptide information for Arabidopsis mitochondrial matrix proteins was submitted to MASCOT for database searching according to the parameters mentioned above (without iTRAQ modifications). A lon1 peptide library was then constructed with Skyline version 1.1.0.2905 (Stergachis et al., 2011) using the developer defined settings for peptides by charge state to determine the optimum collision energies (CEs) for the candidate MRM transitions. Candidate peptides were limited to those between eight and 25 amino acid residues in length, charge states of 2 + and 3 + , y-series ions only, and no product ions ,300 D. CE optimization was then performed on wild-type samples using an Agilent 6430 TripleQuad LC-MS with an Agilent 1200 nano/cap HPLC and a ChipCube housing a 160-nL Zorbax 300SB-C 18 5-mm enrichment column and a Zorbax 300SB-C 18 5-mm, 150-mm separation column and was controlled by Agilent MassHunter Workstation Acquisition B.03.01 (B2065). Samples were loaded onto the enrichment column at 3 mL/min and eluted at 300 nL/min along a 15-min gradient (5%-45% ACN with 0.1% formic acid). All measurements were made in positive mode using dynamic MRM scan type with a retention time window of 2 min. Data were then reimported into Skyline where CE optimization data were manually verified and poorly responding peptides were culled from the list. A quantifier and at least one qualifier ion were assigned for each peptide transition that could be used for the relative quantitation of the lon1 protein in mitochondrial-enriched samples. Technical duplicates of triplicate biological samples of the wild type and both lon1 mutants were performed. MRM data were quantified using Agilent Mass-Hunter Quantitative Analysis (B.04.00/build 4.0.225.19) using default settings for quantifier/qualifier ratios. The mean peptide response for the eight peptides analyzed was calculated as ratio to the wild type with SE of the mean.

Gas Chromatography-MS Metabolite Analysis
Hydroponically grown root and shoot tissues were collected midway through the light and dark periods and excess water removed by vacuum filtration prior to flash-freezing in liquid N 2 . Total metabolites were extracted from 20 to 50 mg of frozen-fresh tissue in 500 mL of methanol buffer, containing ribitol as an internal standard, with a thermomixer (1,200 rpm; 75°C; 20 min). Samples were centrifuged at 13,200 rpm for 3 min and 100 mL of supernatant removed and dried in a vacuum centrifuge. Online derivatization was performed with an MPS2 XL-Twister autosampler (Gertel) where samples were resuspended in 20 mL of 20 mg/mL methoxyamine hydrochloride in pyridine and incubated at 37°C for 90 min with constant shaking at 750 rpm. A 30-mL aliquot of N-methyl-N-(trimethylsilyl) trifluoroacetamide + 1% Nmethyl-N-trimethylsilyltrifluoroacetamide + 1% trimethylchlorosilane was added and samples again were incubated at 37°C for 30 min with constant shaking at 750 rpm. Following a final 30-min incubation at room temperature, 1 mL was injected splitless into an Agilent 7890A gas chromatograph and 5975C inert XL MSD with triple-axis detector with a Varian Factor 4 column. Raw gas chromatography-MS spectra were processed with Agilent's Chemstation (E.02.00.493) where data were normalized first to the internal standard and second to the sample weight. Ten randomized biological samples were run per genotype, with each sample being extracted from a different plant. Dixon's Q-test was applied to determine outliers and results are presented as mean 6 SE of the mean.

Immunoblot Analysis and Detection of Carbonyl Groups
Mitochondrial proteins (5-25 mg) were separated by SDS-PAGE along 10% to 20% Criterion Tris-HCl polyacrylamide gels (Bio-Rad). For immunodetection, separated proteins were then transferred to Hybond-C nitrocellulose membranes using a Hoefer semidry apparatus and incubated at room temperature for 1 h with primary antibodies. Appropriate secondary antibodies conjugated to horseradish peroxidase were applied to washed membranes for 1 h prior to detection of the chemiluminescent signal using the ECL Advance western blotting detection kit (GE Healthcare) and an ImageQuant RT ECL imager (GE Healthcare). Densitometry of the replicate bands was performed using ImageJ. The list of antibodies used is given in Supplemental Table S1. Detection of carbonyl groups was done on replicate (n = 3) 10-mg samples of mitochondrial proteins using the OxyBlot protein oxidation detection kit (Millipore) according to the manufacturer's instructions. Densitometry was performed as above.

Production of HSP Antibodies
Monoclonal antibodies to HSP60 and HSP70 were obtained from a set of hybridoma lines raised against Arabidopsis proteins and produced by Ab-Solutions based in the Western Australian Institute of Medical Research (Perth, Australia). A/J mice were immunized with 50 mg of Arabidopsis mitochondrial matrix protein emulsified in Complete Freund's adjuvant intraperitoneally, followed by a boost in Incomplete Freund's adjuvant followed by an aqueous boost 4 and 8 weeks later. Mice were bled 2 weeks after the third injection, and sera were assayed by ELISA to determine antibody titers. The best responder was boosted 4 d before fusion. Spleen cells were fused with Sp2/O myeloma cells according to standard procedures (Goding, 1996). Antibody-containing supernatants were identified by ELISA. Hybridomas producing antibodies were selected for further study. Selected hybridoma lines were then analyzed for specific monoclonal antibody production by western analysis of 2D separated matrix proteins and identification of the corresponding gel spot by MALDI-MS/MS (as above). Hybridoma lines expressing antibody reacting with HSP70 and HSP60 were isolated by this screen.

Native Gel Analysis of Mitochondrial Respiratory Complexes
Mitochondrial samples (200 mg) were solubilized in 50 mg/mL digitonin in buffer (30 mM HEPES, pH 7.4, 150 mM potassium acetate, and 10% [w/v] glycerol) on ice for 20 min and centrifuged at 18,300g for 20 min at 4°C prior to gel loading. Protein complexes were separated along blue native acrylamide gels (4% acrylamide stacking gel; 4.5%-16% acrylaminde gradient separating gel) at 4°C and 15 mA/gel. Gels were stained with colloidal Coomassie and imaged with a Typhoon scanner (GE Healthcare). Abundance of protein complexes was assessed by densitometry using ImageJ software.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Table S2. Identification of proteins differentially abundant in the wild type and lon1-1 and lon1-2 in DIGE analysis.
Supplemental Table S3. Identification of proteins differentially abundant in the wild type and lon1-1 and lon1-2 in iTRAQ analysis.
Supplemental Table S4. Quantified oxidized peptides from the iTRAQ analysis between lon1 mutants and the wild type.
Supplemental Table S5. Comparison of proteins identified to change in abundance in lon mutants and those that change during oxidative stress in Arabidopsis and yeast.