Arabidopsis thaliana plants (Columbia ecotype, Col-0) were either grown in soil at 22°C with 70% humidity and 150 μmol⋅m^-2⋅s^-1 light with a 16-h light/8-h dark photoperiod (long day, LD) or in growth chambers under LD photoperiod at 22°C or 30°C (water cultures or agar plates). For growth in water cultures, ca. 200 sterilized seeds were added to 80 ml of growth medium (1/2 Murashige and Skoog medium (MS), 3% sucrose, 2 mM MES, 1.5 g/L Gamborg B5 vitamins (1000×), pH 5.7) and incubated on an orbital shaker at 80–100 rpm for 2 weeks. For growth on agar plates, sterile seeds were plated on 1/2 MS with 1.5% sucrose and 1% bactoagar, supplemented with either 300 mM mannitol or 0.5 μM paraquat. The T-DNA insertion mutant seeds (SALK_093517C, SALK_094274, SALK_077448, SALK_012114C, SALK_088054C, SAIL_135_G10, WiscDsLox507A03, GABI_893A04, SALK_080280, SALK_080262, SALK_080264, SALK_080272) were obtained from the Nottingham Arabidopsis Stock Centre (NASC). Homozygous mutants were selected by PCR on genomic DNA template using primers described in Supplementary Table S2. The position of the T-DNA insertion of the A. thaliana homozygous knockout plants was verified by sequencing. A lack of gene expression in the selected homozygous plants was determined by RT-PCR using primers listed in Supplementary Table S2. Amplification of the transcript for ACT2 gene (At3g18780) was used as a control. Growth progression was analyzed according to Boyes et al. (2001).

For cloning of plant or yeast proteases, the full-length coding sequences were amplified by the Phusion Polymerase (Thermo Scientific) with primers listed in Supplementary Table S2 using as a template cDNA synthesized from the total RNA of wild-type plants or genomic DNA from the wild-type yeast cells, respectively. The PCR products were cloned into the Gateway vector pENTR-D-TOPO (Thermo Scientific) to generate the entry clones. For the transient expression in protoplasts, the entry clones were recombined into p2GWF7 (Joubès et al., 2004) to yield C-terminal GFP-tagged peptidases. For the stable expression of C-terminal GFP-tagged peptidases, the entry clones were recombined into pGWB551 (Nakagawa et al., 2007). For the stable expression of FLAG-tagged AtOMA1, the entry clone was recombined into pGWB511 (Nakagawa et al., 2007). For the stable expression of FLAG-tagged AtATP23, a sequence coding for a triple FLAG tag was added to the 5′ end of the AtATP23 coding sequence by PCR. The fusion PCR product was cloned into the Gateway vector pENTR-D-TOPO (Thermo Scientific) and the obtained clone was recombined into pGWB514 (Nakagawa et al., 2007). For the yeast complementation analysis, the entry clones were recombined into pVV209 (Van Mullem et al., 2003) to yield C-terminal HA-tagged peptidases. All constructs were verified by DNA sequencing and are listed in Supplementary Table S4.

The protoplast isolation and transfection was performed according to Yoo et al. (2007). The photographs of protoplasts were taken with scanning confocal microscope (Olympus and Zeiss) equipped with a 40X objective NA = 0.95, using the excitation wavelength 488 nm for GFP and 561 nm for MitoTracker CMXRos (Life Technologies). Images of protoplasts were processed and analyzed with ImageJ software¹.

The transgenic Arabidopsis atp23flag^OE, oct1flag^OE, oma1flag^OE and imp1agfp^OE plants were generated via vacuum infiltration (Clough and Bent, 1998) using Agrobacterium tumefaciens strain LBA4404. The selection of transformants was performed on 1/2 MS medium supplemented with 1.5% sucrose and 30 μg/ml hygromycin.

Yeast strains imp1Δ (IMP1::kanMX4), imp2Δ (IMP2::kanMX4), oct1Δ (oct1::kanMX4), icp55Δ (ICP55::kanMX4), and oma1Δ (OMA1::kanMX4) were isogenic to BY4742 and obtained from Euroscarf ². The CW3 strain (atp23Δ::HIS3MX6), isogenic to W303-1B, was kindly provided by prof. Thomas Langer (University of Cologne). The strain coa2Δ was generated by PCR-based gene replacement method using natMX4 cassette amplified from plasmid pAG25 (Goldstein and McCusker, 1999). The disruption of the COA2 gene in oma1Δ was generated using the HIS3MX6 cassette amplified from plasmid pFA6a-His3MX6 (Longtine et al., 1998). PCR primers used to amplify the cassette and verify the disruption are listed in Supplementary Table S3. The strains were transformed using the lithium acetate method according to Gietz et al. (1995). For the functional complementation assay yeast cells were grown overnight in a liquid complete synthetic drop-out medium (0.7% yeast nitrogen base without amino acids, 2% glucose) lacking appropriate auxotrophic markers. Serial dilutions of the same number of cells were plated on YPD (1% yeast extract, 2% peptone, 2% glucose) or YPG (1% yeast extract, 2% peptone, 2% glycerol) plates and incubated at 30°C or 37°C for 3 days.

For SDS-PAGE and BN-PAGE experiments, mitochondria were isolated from Arabidopsis water cultures according to Day et al. (1985). Typically, 1 mg of mitochondrial protein was obtained from each preparation. Isolation of yeast mitochondria was performed according to Daum et al. (1982). Isolation of chloroplasts was performed using Chloroplast Isolation Kit (Sigma Aldrich). Protein concentration was determined using the DC Protein Assay (Bio-Rad).

Hundred microgram of mitochondrial proteins from the oct1flag^OE, imp1agfp^OE, atp23flag^OE lines were suspended in 200 μl of 0.1 M Na₂CO₃, pH 11, and the suspension was sonicated three times for 10 s on ice, followed by centrifugation at 100000 ×g for 60 min. The resulting pellet (membrane fraction) and supernatant (soluble fraction) were directly dissolved in the electrophoresis sample buffer. Mitochondria from oma1flag^OE were suspended in 200 μl of 0.1 M Na₂CO₃, pH 10.7, and directly centrifuged. Afterward, the obtained pellet and supernatant were dissolved in the electrophoresis sample buffer with the addition of 8 M urea.

One-dimensional blue-native gel electrophoresis (BN-PAGE) and catalytic staining of mitochondrial respiratory chain complexes were performed as described in Kolodziejczak et al. (2007). Bands were quantified using ImageJ software¹. Two-dimensional BN-PAGE was carried out as described in Piechota et al. (2010).

Total protein extracts (30 μg) or mitochondrial proteins (30 μg) were resolved on SDS-PAGE according to Laemmli (1970) using 12% gel (unless otherwise indicated) and then transferred onto nitrocellulose or PVDF (Bio-Rad) membrane. Blots were probed using the primary antibodies listed in Supplementary Table S5. Anti-rabbit or anti-mouse secondary antibodies conjugated to horseradish peroxidase were used, and the signal was visualized with the enhanced chemiluminescent ECL reagent (Advansta) and the GBox imager (Syngene).

Total RNA was extracted from leaves of 4-week-old plants using GeneMATRIX Universal RNA Purification Kit (EURx). Reverse transcription analysis and the quantification of mRNA were performed by quantitative RT-PCR as previously described (Kwasniak et al., 2013). The wild-type plants served as a calibrator, and the ACT2 gene (At3g18780) was used as a reference. The data were analyzed using the LightCycler software version 4.0 (Roche Diagnostics). All primers used for qRT-PCR are listed in Supplementary Table S6.

Respiration of isolated mitochondria of 2-week-old wild-type and oma1-1 seedlings grown under LD, 22°C, was recorded using a Clark-type oxygen electrode (Hansatech). The measurements were performed at 25°C in an incubation buffer containing 0.3 M mannitol, 10 mM TES-KOH (pH 7.5), 3 mM MgSO₄, 10 mM NaCl, 5 mM KH₂PO₄ and 0.1% BSA. Succinate (SA) at 5 mM and NADH at 1 mM were the respiratory substrates. 0.5 mM ATP and 2 mM SHAM were added to the incubation medium to ensure full activity of SDH and to block AOX-mediated oxygen consumption, respectively. The ADP-stimulated respiration (state 3) was measured in the presence of 0.2 mM ADP. To estimate the production of ATP by isolated mitochondria, oligomycin A (2 μg/ml) was added to inhibit ATP synthesis.

Statistical analyses were conducted using Microsoft Excel. Error bars represent the standard deviation (SD) of at least three independent experiments. Statistical significance was assessed using Student’s t-test and considered significant at p ≤ 0.05.

In a previous report, through sequence comparisons, we have shown that the Arabidopsis genome contains homologs of all known yeast ATP-independent peptidases, with the exception of yeast thiol-aminopeptidase (Lap3) (Kwasniak et al., 2012). Besides the presence of already known plant ATP-independent peptidases, such as MPP, RBL12, and PREP, an in silico analysis revealed the presence of a homolog each of yeast Oct1 (At5g51540), Icp55 (At1g09300), Atp23 (At3g03420), Oma1 (At5g51740), and Prd1 (At5g65620), and six homologs of Imp (At1g53530, At3g08980, At1g23465, At1g29960, At1g06870, and At2g31140) in the A. thaliana genome. Recently, the predicted homologs of yeast Prd1 (AtOOP), Icp55 (AtICP55), and Oct1 (AtOCT1) were identified experimentally in Arabidopsis mitochondria (Kmiec et al., 2013; Carrie et al., 2015; Huang et al., 2015). In the present study, we investigated the mitochondrial localization of the remaining expected ATP-independent proteases, namely Atp23 (AtATP23), Oma1 (AtOMA1), and selected Imp homologs (AtIMP1a, AtIMP1b, and AtIMP2). Additionally, in order to explore the possible targeting of AtOCT1 into chloroplasts (Kleffmann et al., 2004), we decided to determine the subcellular targeting of AtOCT1, even though its mitochondrial localization was recently proven by in vitro import into mitochondria isolated from Arabidopsis (Carrie et al., 2015). We excluded three Imp homologs (At1g29960, At1g06870, and At3g08980) from this study, considering their non-mitochondrial localization based on prediction studies and proteomic data (Feiz et al., 2006; Ni et al., 2010; Hooper et al., 2014). Taking into account the presence of the diagnostic motif RX5P in Imp1 and NX5S in Imp2 (Burri et al., 2005), we were able to distinguish the plant homologs of yeast Imp1 (AtIMP1a and AtIMP1b) and Imp2 (AtIMP2) (Supplementary Figure S1). The nomenclature of the plant proteases analyzed in this study with reference to yeast enzymes is given in Table 1.

To verify the predicted mitochondrial localization, two approaches were used. In the first approach, constructs containing a green fluorescent protein (GFP) fused to the C-terminus of the full-length coding sequences of the proteases were used for transient expression in Arabidopsis protoplasts. Colocalization of the GFP signal with a mitochondria-specific fluorescent probe MitoTracker CMXRos was observed for AtOMA1-GFP, AtOCT1-GFP, and AtIMP2-GFP, suggesting the mitochondrial localization of these fusion proteins (Figure 1A). However, using transient expression, we were unable to detect the GFP signal from AtIMP1a-GFP and AtIMP1b-GFP fusion proteins, whereas the AtATP23-GFP fusion protein tended to aggregate into clumps. In the second approach, we decided to generate Arabidopsis transgenic lines constitutively expressing C-terminal FLAG-tagged full-length variants of AtOMA1, AtOCT1, and AtATP23 proteases and C-terminal GFP-tagged full-length variants of AtIMP1a, AtIMP1b, and AtIMP2. We were successful in generating all the lines, with the exception of lines expressing AtIMP1b-GFP and AtIMP2-GFP. The expression of fusion proteins was monitored by immunodetection (Figure 1B). In the case of AtIMP1a-GFP fusion protein, both fluorescence microscopy (Figure 1A) and immunoblotting (Figure 1B) were used. Detection of the tags exclusively in mitochondria, but not in chloroplasts, fully confirmed the results of the transient expression assay for AtOMA1 and AtOCT1, and clearly revealed the mitochondrial localization of the tagged variants of AtATP23 and AtIMP1a (Figure 1B). Table 1 summarizes the results obtained from the localization experiment.

Next, we fractionated mitochondria isolated from Arabidopsis lines constitutively expressing the tagged fusion proteases in order to examine if the enzymes being investigated were soluble or membrane localized. Similar to their yeast counterparts, AtATP23 behaved like a loosely associated with the inner membrane Cyt c, whereas AtOMA1 and AtIMP1a were detected in the membrane protein fraction (Figure 1C). Surprisingly, AtOCT1 was also found along with the membrane proteins. This result is in contradiction with the matrix localization of the AtOCT1 homolog in yeast mitochondria (Figure 1C) (Isaya et al., 1994).

To check whether the function of Arabidopsis mitochondrial ATP-independent proteases is similar to their yeast counterparts, a functional complementation assay was conducted in yeast. We tested the Arabidopsis proteases identified in this study (AtIMP1a, AtIMP2, AtATP23, and AtOMA1) (Figures 2A,B) along with AtOCT1 and AtICP55, the ATP-independent proteases described earlier by Carrie et al. (2015) and Huang et al. (2015) (Figure 2C). The phenotypes of yeast mutants lacking the respective protease growing on non-fermentable carbon sources were described in the following publications: imp1Δ and imp2Δ (Burri et al., 2005), oct1Δ (Isaya et al., 1994), atp23Δ (Osman et al., 2007; Zeng et al., 2007), oma1Δcoa2Δ (Khalimonchuk et al., 2012), and icp55Δ (Vögtle et al., 2009). After transformation, the presence of HA-tagged Arabidopsis proteases in yeast mutant mitochondria was confirmed by immunoblotting (Supplementary Figure S2). However, the Arabidopsis ATP23 protein could not be stably expressed until it was fused with a mitochondrial targeting sequence derived from the yeast cytochrome b2 protein (Cyb2) (Supplementary Figure S2B).

When yeast knockout mutants lacking the IMP1 or IMP2 gene were transformed with constructs carrying their plant counterparts (AtIMP1a or AtIMP2), the defect in respiratory growth could not be restored in yeast cells grown in a medium containing glycerol as the carbon source (YPG medium) (Figure 2A). Similarly, in the functional complementation experiments, we found that the AtATP23 protein did not restore respiratory function of the atp23Δ yeast mutant (Figure 2A) and was also unable to process yeast Atp6 (Figure 2B). Furthermore, the plant homolog of Oct1 could not restore respiratory function in the oct1Δ yeast mutant (Figure 2C), suggesting a functional divergence between the plant and yeast proteins, which has been postulated previously as well (Carrie et al., 2015).

In contrast to the above results, AtOMA1 and AtICP55 were able to substitute for their yeast homologs in the functional complementation assays. It has been reported that a depletion of Oma1 results in progressive impairment of the respiratory function in aging oma1Δ mutants (Bohovych et al., 2015). However, the same results were not obtained in repeat experiments. Therefore, we decided to use another functional complementation system, the yeast mutant lacking both Oma1 and Coa2 (the assembly factor for Cox1 biogenesis required during cytochrome c oxidase maturation). It has been demonstrated that Oma1 protease mediates degradation of newly synthesized, unassembled Cox1 in cells lacking the Coa2 assembly factor (Khalimonchuk et al., 2012). Thus, the lack of Oma1 in coa2Δ leads to stabilization of the unassembled Cox1 and restores respiratory function. In other words, while the coa2Δ mutant is not able to grow on glycerol, the double mutant coa2Δoma1Δ displays growth on media containing non-fermentable carbon sources. As can be inferred from Figure 2A, the plant homolog of Oma1 in coa2Δoma1Δ was able to partially replace the degradation function of the yeast protease. It has also been shown that the AtICP55 protease performs a function similar to its yeast counterpart, which is critical for the stability of mitochondrial proteins (Carrie et al., 2015; Huang et al., 2015). Our complementation studies are in agreement with this previous finding since AtICP55 was able to restore respiration in impaired yeast cells lacking ICP55 (Figure 2C).

To investigate the impact of mitochondrial ATP-independent proteases on plant growth and development, we selected homozygous lines with a T-DNA insertion in the gene of interest. PCR analysis was performed in order to select two homozygous lines each for AtATP23, AtOMA1, and AtICP55, and one line each for AtIMP1a and AtOCT1 (Supplementary Figure S3A). Localization of T-DNA insertion sites in the respective genes is shown schematically in Supplementary Figure S4. We observed that the insertion lines for AtICP55 and AtOCT1 identified by us were different from the ones selected by Carrie et al. (2015) and Huang et al. (2015). We could not identify T-DNA insertions in any of the AtIMP2 lines examined (SALK_080262, SALK_080280, SALK_080264, and SALK_080272) (Supplementary Figure S3B). RNA analysis revealed that the full-length transcript was not synthesized in four out of the eight selected homozygous lines, including imp1a-1 (SALK_094274), oma1-1 (SALK_088054C), oct1-1 (SALK_077448), and icp55-1 (GABI_893A04) (Supplementary Figure S3C). All these null mutants carried a T-DNA insertion in the coding region (Supplementary Figure S4) and were used for further analysis.

A growth stage-based progression analysis was performed for imp1a-1, oma1-1, oct1-1, and icp55-1 growing under optimal conditions, following the method of Boyes et al. (2001). No significant developmental alterations were observed in any of the examined knockout lines (Supplementary Figures S5A,B). Analysis of seedling morphology revealed that only oma1-1 displays a slight reduction in root length (Figure 3A, Supplementary Figure S5C). The Arabidopsis eFP Browser indicates that the expression of AtOMA1 is induced by heat stress (Winter et al., 2007). Therefore, we examined the growth of oma1-1 plants as well as other null mutants under moderate heat stress (LD, 30°C) (Figure 3A). Growth of imp1a-1, oct1-1, and icp55-1 was indistinguishable from wild-type plants. However, oma1-1 revealed strong phenotypic differences under this condition: an overall decrease in seedling size with an 80% reduction in root length compared with the wild-type plants (Figure 3A). We also found that plants lacking AtOMA1 exhibited increased sensitivity to osmotic stress (imposed by high concentrations of mannitol) and oxidative stress (induced by paraquat) (Figure 3B). To confirm that the observed morphological alterations in oma1-1 were caused due to the lack of AtOMA1 protein, we performed a complementation assay by transforming the oma1-1 line with the full-length cDNA of At5g51740 (Supplementary Figure S3D). Analysis of seedling morphology under optimal as well as stress conditions indicated recovery of the wild-type phenotype in complemented lines (Figure 3C), thus, confirming that the phenotype of oma1-1 mutant is caused solely by the lack of AtOMA1.

We investigated the impact of the examined proteases on the functionality of the OXPHOS system by using a combination of blue-native polyacrylamide gel electrophoresis (BN-PAGE) and in-gel histochemical staining to detect the enzymatic activities of complexes I, IV, and V. This analysis revealed that the loss of AtIMP1a, AtOCT1, or AtICP55 did not cause significant changes in activity and protein level of the investigated complexes in plants grown under optimal conditions (Supplementary Figure S6).

The only exception was oma1-1 mutant, which showed an evident reduction (∼50%) in the activity of complex V (even though the total protein level remained unchanged) as compared with the wild-type plants (Figures 4A,B). The functionality of complex V was also affected when the oma1-1 plants were grown at 30°C (Figure 4), resulting in a decrease in activity by 60% and a slight but statistically relevant decline in the total amount of protein (Figures 4A,B). Interestingly, protein level and activity of complex IV, both were slightly higher in the mutant (Figures 4A,B), particularly under optimal conditions. In contrast, reduction in the activity of complex I and supercomplex I + III₂, which was correlated with decreased protein level, was observed exclusively under moderate heat stress conditions in oma1-1. Furthermore, analysis using BN/SDS-PAGE followed by immunoblotting with antibodies directed against subunits of the examined complexes led to similar results on the abundance of supercomplex I + III₂ and complexes I, IV, and V in oma1-1 mitochondria under stress conditions (Supplementary Figure S7). It should be emphasized that protein amount and activity of the OXPHOS complexes were restored to the wild-type levels in the complemented line of oma1-1, thus, confirming that the observed defects resulted from AtOMA1 deficiency (Figure 4A).

There could have been many probable reasons for the alterations in the protein amount of complexes. To better understand this issue, we quantified the abundance of mRNAs encoding representative components of complexes I, IV, and V. The levels of these transcripts were generally unaffected by the loss of AtOMA1 under both optimal and moderate heat stress conditions, except for a slight increase in ATP2-3 under optimal conditions, coxII under elevated temperature, and CA2 under both conditions (Supplementary Figure S8). These findings clearly imply that the reduction in the abundance of OXPHOS complexes in the oma1-1 mutant was not due to changes in the transcript levels. Similarly, accumulation of complex IV in oma1-1 was not due to transcriptional induction of genes encoding its subunits, since a higher level of only one transcript (coxII) was observed, and that too, exclusively under elevated temperature conditions.

In contrast to BN-PAGE or BN/SDS-PAGE, detection of proteins using SDS-PAGE not only allows the visualization of complex-assembled subunits, but also of unassembled subunits. The western blot experiment presented in Figure 4C indicates that in mitochondria lacking AtOMA1, the abundance of NAD9 (complex I) as well as ATP1 and ATP2 (complex V) was unaltered under optimal growth conditions, whereas a decline in the level of these subunits was found under stress conditions. On the other hand, the protein amount of complex IV subunit consistently increased under both conditions. Thus, the SDS-PAGE (complex subunit) and BN-PAGE (whole complex) results indicated similarities in the abundance of OXPHOS complexes.

To confirm the aberrant functionality of OXPHOS in oma1-1, we measured the rate of oxygen uptake by mitochondria isolated from wild-type and oma1-1 plants grown under optimal conditions, with succinate and NADH as the respiratory substrates. We observed a significantly lower basal respiration rate (state 2) in oma1-1 in comparison with the wild-type mitochondria (Figure 5A). Furthermore, while the respiration rate increased in wild-type mitochondria on addition of ADP (state 3), the oma1-1 mitochondria were unable to respond to this treatment (Figure 5A). In addition, the ATP production rate (measured as the respiration rate after inhibition of ATP synthase by oligomycin A) was lowered in the oma1-1 mutant in comparison with the wild-type (Figure 5B). Thus, our data indicated that oma1-1 mitochondria showed impaired ATP production and were not capable of reaching full respiratory capacity, even though the electron transfer chain was stimulated to work at a maximum rate.

It is possible that the loss of a single protease is compensated by the functions of the other proteases, and therefore, no phenotypic changes are visible. This compensation may result from redundancy between mitochondrial proteases and/or induction of other proteases at the transcript and/or protein level. To check if this second possibility explains the observed lack of altered phenotype in almost all of the investigated mutants under optimal growth conditions, the transcript abundance of a bulk of known mitochondrially localized proteases was quantified using real-time PCR (Figure 6A). Surprisingly, no clear transcriptional response was observed from genes encoding different mitochondrial proteases in the mutants lacking a single protease, with the exception of PREP2, the targeting peptide degrading protease/oligopeptidase (Kmiec et al., 2014), which was slightly up-regulated in three out of the four mutants analyzed (Figure 6A). Next, the protein level of several mitochondrial proteases was evaluated by immunoblotting using available antibodies (Figure 6B). No significant changes were observed in any mutant line, including PREP, except for FTSH4 in oma1-1 and FTSH10 in imp1a-1, which showed an increase of approximately 50 and 30%, respectively (Figure 6B).