Mitochondrial CLPP2 Assists Coordination and Homeostasis of Respiratory Complexes1[OPEN]

Jakob Petereit,a Owen Duncan,a Monika W. Murcha,a Ricarda Fenske,a Emilia Cincu,a Jonathan Cahn,a Adriana Pru zinská,a Aneta Ivanova,a Laxmikanth Kollipara,b Stefanie Wortelkamp,b Albert Sickmann,b,c,d Jiwon Lee,e Ryan Lister,a,f A. Harvey Millar,a,2,3 and Shaobai Huanga,3 ARC Centre of Excellence in Plant Energy Biology, School of Molecular Sciences, The University of Western Australia, Washington 6009, Australia Leibniz-Institut für Analytische Wissenschaften-ISAS-e.V., 44139 Dortmund, Germany Department of Chemistry, College of Physical Sciences, University of Aberdeen, Aberdeen AB24 3FX, Scotland, United Kingdom Medizinische Fakultät, Medizinische Proteom-Center, Ruhr-Universität Bochum, D–44801 Bochum, Germany Centre for advanced Microscopy, The Australian National University, Acton, Australian Capital Territory 2601, Australia The Harry Perkins Institute of Medical Research, Perth, Washington 6009, Australia

Mitochondria contain numerous protein complexes that catalyze important functions, including oxidative phosphorylation leading to ATP synthesis, respiratory carbon metabolism, consumption of oxygen, and the production of reactive oxygen species (ROS; Millar et al., 2011;Huang et al., 2016). Subunits for some of these protein complexes are encoded across mitochondrial and nuclear genomes. This requires separate gene expression, protein synthesis, and protein import into mitochondria followed by a coordinated complex assembly. The regulation at a transcriptional level is limited, owing to the general lack of specific mitochondrial gene transcript activation (Costanzo and Fox, 1990;Giegé et al., 2000;Kühn et al., 2009).
Some of the coordination of nuclear-and mitochondrialencoded protein complex assembly in yeast occurs at the translation level and is unidirectional (Couvillion et al., 2016), with cytosolic events influencing mitochondrial mRNA levels. In Arabidopsis (Arabidopsis thaliana) cell culture under carbon starvation and refeeding conditions, a transient dysregulation in the stoichiometry of complex V subunits ATP1 and ATP2, encoded by mitochondrial and nuclear genomes, respectively, has been identified (Giegé et al., 2005). ATP1 was increased at the transcript level but not at the level of ATP1 protein abundance, while ATP2 increased at the protein level, but the ATP2 transcript was unchanged. Giegé et al. (2005) proposed that posttranslational processes in assembly were responsible for rebalancing this process after refeeding of cells, but they did not identify factors responsible for it. It has previously been suggested that unassembled protein complex subunit degradation by mitochondrial proteolysis is the basis for such regulation (Sarria et al., 1998), but the specific protease(s) involved have not been identified.
Organellar protease networks exist in mitochondria and plastids to control intraorganellar protein degradation (van Wijk, 2015). ATP-dependent endopeptidases of the AAA1 superfamily of ATPases associated with various cellular activities, such as long-filament phenotype-1 (LON1), caseinolytic protease (CLPP), and filamentous temperature-sensitive H (FTSH), are the dominant components of these networks that are conserved across eukaryotes. In plants, mitochondrial LON1 acts as a chaperone, by stabilizing newly synthesized/imported proteins to aid proper folding, and as a protease, by degrading mitochondrial protein aggregates (Li et al., 2017). Loss of mitochondrial FTSHs leads to oxidative stress, and FTSHs are shown to be components of the defense against accumulation of carbonylated proteins in plant mitochondria (Smakowska et al., 2016). To date, there have not been any reports of the effect of loss of CLPP on plant mitochondria.
The CLPP protease has an active Ser-His-Asp catalytic triad and is widely present in bacteria, fungi, mammals, and plants (Yu and Houry, 2007;Bhandari et al., 2018). The x-ray structure of CLPP has been solved from several different organisms with similar features: the protease subunit is comprised of two heptameric rings forming a cylinder-like structure that encloses a large chamber containing the protease active sites (Yu and Houry, 2007;Bhandari et al., 2018). Some bacteria, such as the cyanobacterium Synechococcus elongatus, contain CLPR, which is an inactive version of CLPP lacking the catalytic triad (Schelin et al., 2002). In bacteria, CLPP forms complexes with the AAA1 chaperones CLPX and CLPA and acts as a selective filter for specific target substrates (Gerth et al., 2008).
Plants have a diversified and complex CLP family, with six active CLPPs (CLPP1-CLPP6) and four inactive CLPR paralogs (CLPR1-CLPR4) found in Arabidopsis (Sjögren et al., 2006;van Wijk, 2015). In addition, Arabidopsis have CLP/HSP100 AAA1 chaperones with seven class I CLPs (CLPB1-CLPB4, CLPC1 and CLPC2, and CLPD) in chloroplasts and three class II CLPs (CLPX1-CLPX3) in mitochondria (Peltier et al., 2004). Arabidopsis plastids have five CLPP proteases (CLPP1, encoded by the plastid genome, and CLPP3-CLPP6) and four inactive CLPRs, which together form the tetradecameric and asymmetric ;350-kD CLP protease core (van Wijk, 2015). Three CLP AAA1 chaperones (CLPC1, CLPC2, and CLPD) and the adaptor CLPS1 deliver protein substrates to the protease complex for degradation (van Wijk, 2015). Plastid clpp and clpr mutants in Arabidopsis display severe phenotypes, including disorder of embryogenesis, seedling development, and chloroplast biogenesis. Complete loss of CLPP5 and CLPP4 is embryo lethal, and the clpp3 null mutant can only germinate and develop seedlings under heterotrophic conditions . CLPP1 is known to be essential for leaf development in tobacco (Nicotiana tabacum; Shikanai et al., 2001). Knockdown of a number of subunits of the CLP complex in tobacco impacted the CLP complex composition and chloroplastic protein homeostasis (Moreno et al., 2017(Moreno et al., , 2018. Knockout lines of CLPR2 or CLPR4 cause embryogenesis delay and developmental arrest at the cotyledon stage . Upregulation of chaperone proteins and downregulation of photosynthesis coupled with upregulation of the ATP import pathway occur in clpr2, clpr4, and clpr3 mutants . Dysfunction of CLPP6 in rice causes developmental stage-dependent virescent yellow leaf with decreased chlorophyll accumulation and impaired photosynthesis (Dong et al., 2013). The proteolytic P-ring (CLPP) and the structurally similar R-ring (CLPR) form the core CLPPR complex. Therefore, the plant CLPPR protease complex appears to play an essential role in plastid and chloroplast biogenesis and development.
Arabidopsis mitochondria contain a homotetradecameric CLPP2 core protease encoded by a single nuclear gene and three CLPX chaperones (CLPX1-CLPX3; Peltier et al., 2004). The CLPP2 core has been detected by Blue Native PAGE (BN-PAGE) with a molecular mass of ;320 kD, consistent with a predicted tetradecamer structure (Peltier et al., 2004;Senkler et al., 2017), but no transfer DNA (T-DNA) lines for CLPP2 exist. In this study, we developed two knockout lines of CLPP2 in Arabidopsis by mutations using CRISPR-Cas9 technology, providing a tool for functional analysis of mitochondrial CLPP2. Our assessments using a range of omic technologies and specific follow-up experiments have revealed that whereas CLPP2 is not essential for plant growth, it does contribute to the regulation of nuclear and mitochondrial J.P. performed most of the experiments and data analysis; O.D. and R.F. assisted with and performed some of the mass spectrometry and data analysis; J.C., E.C., and R.L. assisted in the development of CRISPR-Cas9 knockout lines; A.P. assisted in RNA-sequencing analysis; L.K., S.W., and A.S. assisted with the ChaFRADIC, liquid chromatography-mass spectrometry, and data analysis; M.M. and A.I. assisted with in vitro imports; J.L. performed the transmission electron microscopy analysis; S.H., A.H.M., and J.P. designed experiments and analyzed and discussed results; S.H., A.H.M., and J.P. wrote the article; and all authors commented on the results and the article. [OPEN] Articles can be viewed without a subscription. www.plantphysiol.org/cgi/doi/10.1104/pp.20.00136 protein complex assembly and maintenance. The manner of CLPP2 function is consistent with previously proposed posttranslational processes that have been implicated in this control.
The first mutant, clpp2-1, showed two separate genetic events that caused the CLPP2 knockout: first, a single adenine insertion at the 25-bp position introduced a frame shift and disrupted the restriction site of NlaIV at 24 bp (Fig. 1B); and second, introduction of the adenine at CRISPR site 1 created a 39 end identical to the 59 end at CRISPR site 3. It is likely that this caused a second error from the nonhomologous end-joining (NHEJ) DNA repair mechanism or the alternative NHEJ DNA repair mechanism, microhomology-mediated end joining, known for insertions, deletions, and inversions (McVey and Lee, 2008), by inserting the complete fragment between CRISPR site 1 and CRISPR site 3 in a reverse-complemented orientation. Consequentially, five downstream stop codons (TGA or TAA) were generated and the second NlaIV restriction site was shifted to the position at 95 bp (Fig. 1B). The second mutant, clpp2-2, had a single NHEJ error at CRISPR site 2 with an adenine insertion at the position of 124 bp, resulting in a disruption of the NruI restriction site and the introduction of three downstream stop codons (TAA or TGA; Fig. 1B).
We confirmed the disruption of CLPP2 in the mutants in the genetic sequence, as well as its consequence on mRNA and protein levels. As the two mutations resulted in disruption of restriction sites within exon 1, genomic DNA from mutants and the wild type was amplified using primers F1 and R1 (Fig. 1A) and digested with NlaIV for clpp2-1 and NruI for clpp2-2. The NlaIV-digested wild-type PCR product resulted in three fragments, a long, 165-bp fragment between the first and second NlaIV restriction sites, an intermediate, 105-bp fragment from restriction site 2 to the 39 end of the PCR product, and a short 24-bp fragment (out of gel range) from the 59 end to restriction site 1 (Fig. 1C). For clpp2-1, NIaIV restriction site 1 was disrupted and restriction site 2 moved in a 59-end direction. Accordingly, the restriction digest pattern had two fragments in total, a 199-bp fragment from the new restriction site 2 to the 39 end of the PCR product and a 92-bp fragment from the 59 end to the new restriction site 2 (Fig. 1C). The NruI-digested wild-type exon 1 PCR product resulted in two fragments 171 and 120 bp in length, whereas clpp2-2 had a disrupted NruI restriction site and only showed the full 291-bp PCR product (Fig. 1C).
Based on RNA-sequencing (RNAseq) analysis, the transcript levels of CLPP2 were significantly reduced in clpp2-1 and clpp2-2 to ;15% of the wild-type level (Fig. 1D). The expression levels of representative housekeeping genes such as general TUBULIN6 (AT5G12250) and mitochondrial ATVDAC1 (AT3G01280) were unaffected in both mutants (Fig. 1D). Whereas CRISPR-Cas9 systems should not actively alter transcript levels, the introduced frameshifts forming several stop codons in the sequence of CLPP2, .50 bp away from the exonexon junction (Fig. 1B), created a high-confidence position for exon-exon junction nonsense-mediated mRNA decay, which may contribute to the observed low transcript abundance (Lloyd, 2018).
To detect the protein abundance of CLPP2, we isolated mitochondria and conducted multiple reaction monitoring (MRM)-based targeted proteomic analysis and included the theoretical peptide of the inverted CLPP2-1 section. The peptide MRM transitions are presented in Supplemental Table S1. CLPP2 protein abundance in mitochondria from both mutants was undetectable, whereas all selected peptides of CLPP2 were detected in the wild type (Fig. 1E). We also did not detect any evidence of theoretical CLPP2-1 peptides. Two representative mitochondrial housekeeping proteins (mitochondrial heat shock protein 70 [mtHSP70; AT4G37910] and a voltage-dependent ion channel [ATVDAC2-AT5G67500]) showed no significant change in protein abundance in either mutant (Fig. 1E). The virtual absence of CLPP2 proteins in both mutants could be explained by the introduction of stop codons at the first exon ( Fig. 1B) that fully disrupted the translation of any functional CLPP2 protein.

Loss of CLPP2 Has an Impact on Transcript Levels of Mitochondrial Genes But Little Effect on Expression of Nuclear Genes
To evaluate global transcript abundance changes upon the loss of mitochondrial CLPP2, we used highthroughput RNAseq to find and quantify differentially expressed genes (DEGs). We analyzed the transcriptome of hydroponically grown seedlings of clpp2-1, clpp2-2, and the wild type, with three biological replicates for each genotype. Based on a filter of a log2 fold change (log2FC) exceeding 60.4 and adjusted P-values of #0.05, we identify 62 DEGs using the R package Deseq2 (Love et al., 2014), which showed a consistent pattern in both mutants (Fig. 2). Only 0.1% of nuclear gene transcripts had significant changes in abundance in clpp2-1 and clpp2-2. This was considerably lower than the 33% of mitochondrial transcripts that showed significant changes in abundance in both mutants (Fig. 2), indicating that disruption of CLPP2 has a specific impact on the abundance of transcripts from mitochondrial genes.
In clpp2-1 and clpp2-2, 41 of the 54 upregulated DEGs from nuclear and mitochondrial genomes encode mitochondrial-localized proteins, whereas three of eight downregulated DEGs other than CLPP2 encode a mitochondrial-localized protein (Table 1). Two genes that encode nonmitochondrial targeted proteins and have unknown function stand out, showing very high and consistent log2FC values. AT4G36850 encodes a conserved PQ-Motif protein (Pfam [PF04193], involved in lysosome targeting) and AT3G56730 encodes a putative endonuclease with a conserved LabA-like PIN domain of limkain b1 (cd10910), which is a common Figure 1. Development of two independent CLPP2 mutants using CRISPR-Cas9. A, Layout of the CLPP2 gene structure, with enlargement of exon 1 showing CRISPR-Cas9 target sites, primer positions, and restriction sites used for restriction digest-based confirmation of the mutations. UTR, Untranslated region. B, Mapping of CRISPR-Cas9 gene disruption for wild-type (WT) CLPP2 and two individual knockouts, clpp2-1 and clpp2-2. For the Clpp2-1 map, the wild type shows the exon 1 sequence map with the location of two NlaIV restriction sites. Clpp2-1 has an adenine insertion (A) at CRISPR site 1 disrupting (X) one NlaIV restriction site and an insertion of the reverse complemented wild-type sequence between CRISPR site 1 and CRISPR site 3, introducing five stop codons (TAA, TGA) resulting in a change of the second NlaIV restriction site. For the Clpp2-2 map, the wild type shows the exon 1 sequence map with the location of one NruI restriction site. Clpp2-2 has an A insertion (CRISPR site 2) resulting in the disruption of the NruI restriction site and insertion of three stop codons. C, Confirmation of CRISPR-Cas9 gene disruption based on PCR fragment restriction digest amplified by forward (F1) and reverse (R1) primers. D, Confirmation of CRISPR-Cas9 gene disruption at the transcript level based on RNAseq analysis. Normalized counts of the target gene, CLPP2, and the two control genes, Tubulin6 and ATVDAC1, are presented. Asterisks indicate significant difference (**P # 0.01; n 5 3). E, Confirmation of CRISPR-Cas9 gene disruption at the protein level based on a targeted proteomic approach. Fold changes of the target protein CLPP2 and the two control proteins, mtHSP70 and ATVDAC2, are presented. Asterisks indicate significant difference (***P # 0.0001; n 5 3 peptides with two replicates for CLPP2, 3 peptides with two replicates for mtHSP70, and 2 peptides with two replicates for ATVDAC2).
Of the 41 upregulated mitochondrial-localized DEGs, 39 originate from the mitochondrial genome (Table 1) and encode proteins of the oxidative phosphorylation (OXPHOS) system, mitochondrial biogenesis apparatus, and various hypothetical proteins, transposons, and retrotransposons. Notably, the expression level of mitochondrial DEGs encoding complex V ATP synthase subunits (such as ATP1, ATP6 [ATMG00410, ATMG01170], ATP8, and ATP9) was higher in both CLPP2 mutants compared to the wild type (Table 1). Some mitochondrial genes encoding complex I subunits (NAD1, NAD3, NAD5, and NAD7) and complex IV subunits (COX1-COX3) were also observed to have high expression in both mutants (Table 1). Expression of mitochondrial genes encoding RNA or DNA polymerase (ATMG00490 and ATMG00810) and intron maturase (ATMG00520) were upregulated in both mutants ( Table 1).

Loss of CLPP2 Alters Mitochondrial Protein Homeostasis
To identify differentially expressed proteins (DEPs) in isolated mitochondria from hydroponically grown Arabidopsis seedlings, we conducted a quantitative proteomic approach. We detected 797 proteins from mitochondrial membrane fractions in mutants and the wild type (see Supplemental Table S2). Among them, we identified 22 DEPs with increased abundance and five DEPs with decreased abundance. These showed a consistent pattern in clpp2-1 and clpp2-2, with a log2FC exceeding 60.4 and P-values #0.05 (Fig. 3). The 27 DEPs are are grouped based on functional categories in  Table 2). Three of the overaccumulated complex I subunits belong to the N-module of the matrix arm, which is assembled before it is attached to the membrane arm of complex I (Ligas et al., 2019). We observed accumulation of proteins related to protein synthesis, such as three components of the mito-ribosome (AT4G3090, AT1G61870, and AT5G64670) and the putative RNA helicase RH9 (AT3G22130), as well as mitochondrial proteases such as MPPa-1,a-2; MPPb, CLPX-1, and PREP1 (Table 2). MPPs are mitochondrial peptidases that are embedded in complex III and are responsible for mitochondrial presequence cleavage after protein import Zhang et al., 2001). CLPX-1 is the chaperone subunit of the CLPXP complex (van Wijk, 2015) and PREP1 degrades mitochondrial presequences after their cleavage from proteins (Kmiec and Glaser, 2012;Kmiec et al., 2014). A change to the abundance of these proteases in the absence of CLPP2 could influence the degradative landscape within the mitochondria of the CLPP2 mutants. In addition to analysis of the membrane mitochondrial proteome, an analysis of matched samples of the soluble mitochondrial proteome produced similar results (including changes in the abundance of the 24-kD subunit, RH9, CLPX1, mtHSP70-1, and PPR proteins), albeit with different ratios of abundance for a range of proteins (including MPP, ATP2, and ATP3; Supplemental  Table S2).
To investigate a potential modification in presequence processing or partial degradation of mature Figure 2. Changes in global transcript abundances in CLPP2 mutants compared with the wild type (WT). The volcano plot illustrates overlapping significant changes of total transcript abundances of the two knockout mutants compared to the wild type. The negative Log10 transformed adjusted P-values are plotted against the log2FC of the transcript abundances measured by RNAseq. Dashed lines separate gene transcripts with significant changes in abundance using the log2FC cutoff of 60.4 and adjusted P-value cutoff of 0.05. Depleted gene transcripts are shown in red and accumulated gene transcripts in green. Transcripts with adjusted P-values ,1 3 10 215 are indicated by arrows. Transposable elements and hypothetical proteins are indicated by open circles. The inset table displays a summary of the total numbers of detected nuclear and mitochondrial genes and the numbers in each category with significant changes in abundance for both mutants relative to the wild type. ORF, Open reading frame. Table 1. Functional category of genes with significant changes in transcript abundance exceeding the cutoff in clpp2-1 and clpp2-2 Genes were classified into the following groups: mitochondrial-encoded (mito-encoded), nuclear-encoded (nuc-encoded), plastid encoded, and knockout (nuclear encoded CLPP2 protease knocked out in the mutants; KO). The log2FC for both mutants compared to the wild type and the corresponding adjusted P-values (P.adj) are shown, as are gene symbols (with their complex) and Arabidopsis gene identifiers (AGI), short gene descriptions, and subcellular localization of the proteins based on the SUBA4 database. CI, Complex I; CIV, complex IV; CV, complex V; Cyt, cytosol; ER, endoplasmic reticulum; EX, extracellular; M, mitochondrion; N, nucleus; P, plastid; PM, plasma membrane; V, vacuole. proteins, we conducted an approach combining protein labeling with isobaric tags for relative and absolute quantification (Ross et al., 2004) and charge-based fractional diagonal chromatography (ChaFRADIC; Venne et al., 2013Venne et al., , 2015 to analyze the N-terminal mature sequence of mitochondrial proteins. Sequence logo analysis indicated that there was no difference in presequence cleavage, as there were very conserved -2R and -3R cutting sites detected in mitochondrial proteins in both mutants and wild type (Supplemental Fig. S3, A and B). There was also no apparent difference in the abundance of any presequence peptides between mutants and wild type (Supplemental Fig. S3C). In addition, no significant difference was observed in cleavage sites in the middle or C terminus of the protein sequence between the two mutants and the wild type (Supplemental Fig. S3, C and D). Thus, loss of CLPP2 had no measurable impact on mitochondrial protein maturation and left no partially degraded parts of mitochondrial proteins that we could detect.

Loss of CLPP2 Disrupts the Coordination of Mitochondrial-and Nuclear-Encoded Subunits of Respiratory and Ribosomal Complexes
To directly compare between mitochondrial transcript and protein abundance changes for individual subunits located in the same protein complex, we compiled the changes in transcript level and protein abundance for all subunits of the mitochondrial OXPHOS system and the mitochondrial ribosome (Fig. 4). Complex I, the largest protein complex of the mitochondrial OXPHOS system, is encoded by nine mitochondrial-encoded genes and 47 nuclear-encoded genes (Ligas et al., 2019;Meyer et al., 2019). Our data showed a consistent transcript upregulation for five mitochondrial-encoded subunits (NAD1, NAD3, NAD5, and NAD7) in both mutants, but we did not find NAD7 protein accumulation in either mutant (Fig. 4A). By contrast, complex I subunits (24 kD, 51 kD, 75 kD, and B14), encoded by four nuclear genes, showed high protein abundance with no change in transcript levels (Fig. 4A). A similar pattern was observed for complex V: of six mitochondrial transcripts, five (ATP1, ATP6 [ATMG0041 and ATMG01170], ATP7, and ATP8) had higher abundance in both mutants, but ATP1 and ATP8 did not show any accumulation at the protein level (Fig. 4E). There was no consistent transcript upregulation of nuclear genes encoding complex V subunits in either mutant, but there was high accumulation of the ATP2 protein, encoded by the nuclear genome, in both mutants (Fig. 4E).
Complex III contains only one mitochondrialencoded subunit Meyer et al., 2019), and its expression appeared to be unaffected transcriptionally in both mutants. Changes in protein abundance of MPPa-1, MPPa-2, and MPPb were observed, but without any changes in transcript level for these nuclear-encoded genes (Fig. 4C). Complex IV has three mitochondrial subunits, with all showing up-regulation in transcript abundance in both mutants compared to the wild type, and 13 nuclearencoded subunits of complex IV showed no change in transcript or protein abundance (Fig. 4D). Complex II of the OXPHOS system contains only nuclear-encoded subunits and appeared to not be affected by a disruption of the mitochondrial CLPP2 protease at the transcript or protein level (Fig. 4B). We did not find consistent transcriptional response in either CLPP2 mutant for mitochondrial-encoded subunits of the mito-ribosome (Fig. 4F); however, two nuclear-encoded ribosomal large subunits (AT4G30930 and AT5G64670) Figure 3. Changes in protein abundances in the membrane fraction of isolated mitochondria in CLPP2 mutants compared with the wild type (WT). The volcano plot illustrates overlapping significant changes of label-free quantification protein intensities of the two knockout mutants compared to the wild type, using a log2FC cutoff of 60.4 and an adjusted P-value cutoff of 0.05 (dashed blue lines). Depleted proteins are shown in red and accumulated proteins in green. Proteins with adjusted P-values ,1 3 10 215 are displayed as arrows. and one nuclear-encoded ribosomal PPR336 (AT1G61870) increased in protein abundance in both mutants (Fig. 4F).

Loss of CLPP2 Impacts Protein Homeostasis and Incorporation of New Subunits into Complex I
Changes in the transcript level and protein abundance of specific complex I subunits led us to assess the impact on respiratory chain complex I in both mutants as an exemplar of a complex affected in CLPP2 mutants.
We first analyzed complex I activity in both mutants compared to the wild type as the rate of de-amino NADH-dependent FeCN reduction but did not find any difference in enzymatic activity (Fig. 5A). To determine the abundance of assembled complex I, we separated mitochondrial complexes using BN-PAGE and stained gels to visualize proteins with Coomassie (Fig. 5B). We did not observe any accumulation of complex I or supercomplex I1III 2 in either mutant compared with the wild type (Fig. 5B). Complex I in gel activity staining using NADH and nitroblue tetrazolium Table 2. Functional category of proteins with significant changes in abundance in clpp2-1 and clpp2-2 Proteins were classified as OXPHOS, mitochondrial biogenesis (mtBIO), Protein homeostasis, knockout (nuclear encoded CLPP2 protease knocked out in the mutants; KO), and other (ungrouped genes). The log2FC of protein abundance in both mutants compared to the wild type, and the corresponding adjusted P-values (P.adj) are shown, as are the gene symbol, Arabidopsis gene identifier (AGI), a short description, and the subcellular protein localization based on the SUBA4 database.Cyt, Cytosol; ER, endoplasmic reticulum; EX, extracellular; M, Mitochondrion; N, nucleus; P, plastid; PM, plasma membrane; V, vacuole.  The heatmaps show the combined changes in transcript abundance (from RNAseq data; Fig. 2) and protein abundance (from quantitative proteomics data; Fig. 3) for the RNA and proteins indicated. The subunits with accession number and abbreviations from individual complexes are listed at the left. In heatmaps, the color code represents the log2FC of the individual mutants compared to the wild type, with green for accumulation and red for depletion in the mutants. Gray indicates undetected proteins. Asterisks indicated significant difference (*P # 0.05, **P # 0.01, and ***P # 0.001; n 5 3 for transcripts and 4 for proteins).
(NBT) as substrate (Yan et al., 2007) showed that supercomplex I1III 2 (band a) and complex I (band b) had similar in-gel activity in both mutants and wild type (Fig. 5B), consistent with their protein abundance and total complex I enzymatic activity. However, we found a residual in-gel activity stain in a lower-molecular-mass complex (band c) in both mutants but not in the wild type (Fig. 5B). This indicated the existence of a potential subcomplex of complex I containing a functional FMN cofactor to convert NADH to NAD 1 and to transfer the free electron onto NBT. To further confirm this hypothesis, we cut the gel band c region from both mutants and wild type and subjected them to mass spectrometry (MS) analysis. We detected 51-and 24-kD subunits of complex I in both mutants but not in the wild type (Fig. 5C), pointing to the existence of a substantial amount of a soluble unassembled N-module subcomplex. We then measured the protein abundance of the three accumulated N-module subunits (24, 51, and 75 kD) by targeted MRM MS in purified mitochondria after fractionation into membrane and soluble fractions by centrifugation. In the membrane fraction, a significant 1.5-to 2-fold accumulation of the 24-and 75-kD subunits in clpp2-1 and clpp2-2 was observed when compared to the wild type (Fig. 5D). However, in the soluble fraction, accumulation was 4-to 6.5-fold that of the wild type and significant for all three subunits (Fig. 5D). This indicated that the accumulation of specific complex I N-module subunits appeared to be an accumulation of an unassembled subcomplex even though the steady-state intact complex I abundance was not affected in either mutant. We conducted an in vitro protein import assay of a radio-labeled 24-kD subunit to understand the nature of the accumulation of the upregulated complex I subunits in clpp2. After cloning and expression, the 35 S Met-labeled Figure 5. Changes in protein abundance for complex I N-module subunits and complex I enzymatic activity and in vitro import for complex I subunits in CLPP2 mutants. A, Complex I enzyme activity in isolated mitochondria from mutants and the wild type (WT) using de-amino NADH as substrate and FeCN as donor. Measurements were made by photospectrometer assay (n 5 5). B, BN-PAGE of isolated mitochondria of the wild type, clpp2-1, and clpp2-2 stained with Coomassie blue (left) and complex I activity stain (NADH and NBT; right). Major OXPHOS components are marked on the gel for size comparison. Lowercase letters at right indicate bands of interest: supercomplex I1III 2 (a), complex I (b), and a potential complex I assembly intermediate (c). C, MS evidence for N-module proteins (24 and 51 kD) present in band c in the BN-PAGE. D, Protein abundance of three complex I matrix arm N-module subunits (24, 51, and 75 kD) in membrane and soluble fractions from isolated mitochondria. Data are shown as log2FC in peptide abundance compared to the wild-type median. Asterisks indicated significant difference (**P # 0.01; n 5 4). E, In vitro import of the radio-labeled 24-kD subunit of complex I into isolated mitochondria of the wild type and clpp2-1 at 5, 10, and 15 min incubation. F, In vitro import of radio-labeled complex I 51-and 23-kD subunits into isolated mitochondria of the wild type and clpp2-1 at 2 h incubation time. The blue-native gels were imaged for radioactive intensity. Loading control gels stained with Coomassie blue are presented in Supplemental Figure S4.
24-kD subunit was incubated with freshly isolated mitochondria. We did not observe any apparent difference in the rate of import between the wild type and clpp2-1 (Fig. 5E). We then tested the incorporation of the radio-labeled 24-kD subunit into complex I using BN-PAGE analysis, but we failed to detect any obvious bands that would suggest incorporation within a protein complex. (Supplemental Fig.  S4). Therefore, we tested another complex I N-module component, the 51-kD subunit (Fig. 5F). After 2 h of import, the radio-labeled, 51-kD subunit was observed to incorporate into complex I in the wild type and clpp2-1, with a stronger signal observed in the wild type (Fig. 5F). We then selected the 23-kD Q-module subunit that is known to interact with the N-module (Ligas et al., 2019). Weaker labeling of complex I and supercomplex I1III 2 with the 23-kD subunit was observed in clpp2-1 compared with the wild type (Fig. 5F). To determine whether the labeling pattern was different with other complex I subunits, we selected the P-module CAL1 subunit and the membrane arm component, B14.7 subunit (Ligas et al., 2019;Supplemental Fig. S4). After 2 h of import of those subunits, we observed stronger labeling of complex I and supercomplex I1III 2 bands in the wild type compared to clpp2-1 (Supplemental Fig. S4). We conclude that for all four subunits tested, labeled subunits can be incorporated into complex I and supercomplex I1III 2 and that this incorporation appeared weaker in mitochondria isolated from clpp2-1 compared to wild-type mitochondria.

Plants Endure Loss of Mitochondrial CLPP2 under Normal and Stress Conditions
We evaluated the growth of clpp2-1 and clpp2-2 compared to wild-type plants under a range of growth conditions to find any phenotypical consequences caused by loss of mitochondrial CLPP2. Arabidopsis seedlings grown for 2 weeks under long-day control conditions showed no phenotypic differences between the wild type, clpp2-1, and clpp2-2 (Fig. 6A). In mouse (Mus musculus) myoblasts, disruption of mitochondrial CLPP is known to lead to an altered mitochondrial morphology (Deepa et al., 2016). Plant roots are a mitochondria-rich tissue for energy production without chloroplasts. We employed transmission electron microscopy to investigate the morphology of CLPP2deficient Arabidopsis mitochondria in root tissues and found no visible differences between the wild type, clpp2-1, and clpp2-2 (Fig. 6B). We grew Arabidopsis seedlings on vertical Murashige and Skoog media agar plates to investigate root elongation under control and various stress conditions. Under control conditions (Fig. 6C) and various stress conditions (salt, osmotic stress, heat, and carbon starvation; Supplemental Fig.  S5), we did not find significant changes in root elongation between the wild type and either clpp2-1 or clpp2-2. We also treated plants with inhibitors of complex I (rotenone), complex III (antimycin A), and the mitochondrial biogenesis inhibitor (doxycycline), but again no difference in growth was found (Supplemental Fig. S5). So far, the only stress condition that affected clpp2-1 and clpp2-2 differently from the wild type was nitrogen starvation, which, after 14 d (Fig. 6D), showed significant reduction in root growth to ;60% of the wildtype root length. Therefore, Arabidopsis seedlings can compensate for the loss of CLPP2 under most growth conditions, but under nitrogen starvation scenarios it Figure 6. Phenotypes associated with loss of mitochondrial CLPP2 under different conditions. A, Arabidopsis seedlings of the wild type (WT), clpp2-1, and clpp2-2 grown on soil for 15 d under long-day conditions. B, Representative transmission electron microscope images of dissected roots of 2-week-old hydroponically grown Arabidopsis seedlings of the wild type, clpp2-1, and clpp2-2. Mitochondria are indicated by red arrows. C and D, Representative plate images (top) and root-length measurements (bottom) of 2-week-old Arabidopsis seedlings grown on vertical plates under control long-day conditions under normal nitrogen supply (C) or without nitrogen (D). Dots represent individual root measurements. Asterisks indicated significant difference (**P # 0.01).
appears that CLPXP may help to acclimate to stress conditions efficiently.

DISCUSSION
CLPP has been well studied in bacteria and some eukaryotes, such as human (Homo sapiens), mice, and fungi, and in particular in the plant chloroplast, where it plays an essential role in chloroplast biogenesis and development (Yu and Houry, 2007;van Wijk, 2015;Bhandari et al., 2018). While it has long been known that a CLPP exists in plant mitochondria (Peltier et al., 2004), there was no information on its function due to lack of genetic resources for its disruption and analysis. In this study, we successfully developed two mutant lines using CRISPR-Cas9 to knock out the Arabidopsis mitochondrial CLPP2 gene. These mutants were designed to provide direct insight into CLPP2 function both in mitochondrial protein homeostasis and in coordinated regulation of assembly of plant respiratory complex subunits, which exhibit mixed mitochondrial and nuclear genomic location.
To our surprise, the knockout of mitochondrial CLPP2 did not cause any phenotypic variation in either CRISPR-Cas9 mutant when compared with the wild type (Fig. 6). By contrast, the plastid-encoded CLPP1 in tobacco plays an essential role in shoot development (Shikanai et al., 2001). In Arabidopsis, CLPP4 and CLPP5 null mutants were embryo lethal, and null mutants of CLPP3 were seedling lethal in soil. Partial downregulation of CLPP4 and CLPP6 by antisense RNA techniques reduced plant growth and development and resulted in pale green plants (Sjögren et al., 2006;Zheng et al., 2006;Olinares et al., 2011). Unlike its chloroplast counterpart, the plant mitochondrial CLPP appears to not be essential for plant growth and development. As bacteria lack a proteasome protein degradation system, CLPP and LON proteases contribute to ;80% of their cellular protein degradation (Goldberg et al., 1994). Bacterial CLPP plays an important role in the degradation of proteins involved in nutrient starvation, stationary phase adaptation, heat-stress response, cell-cycle progression, biofilm formation, cell motility, and metabolism (Frees et al., 2003;Gerth et al., 2008). The plant mitochondrial LON protease plays a key role in regulation of mitochondrial protein homeostasis, and loss of mitochondrial LON1 causes a severe reduction in plant growth and development (Rigas et al., 2009;Solheim et al., 2012;Li et al., 2017). By contrast, there is no information on the function of Lon protease in chloroplastic protein degradation, even though one isoform, LON4, is targeted to the chloroplast (Ostersetzer et al., 2007). The contrasting roles of CLPP and LON proteases in chloroplasts and plant mitochondria might indicate a shift from Lon protease to CLP protease for the respective regulation of critical aspects of chloroplastic and mitochondrial protein homeostasis.
Dysfunction of mitochondrial CLPP causes significant phenotypes in other eukaryotes. Loss of mitochondrial CLPP in humans is linked to infertility and sensorineural hearing loss (Gispert et al., 2013;Szczepanowska et al., 2016). Mitochondrial CLPP in humans is required for protein homeostasis (Szczepanowska et al., 2016), which is involved in the degradation and regulation of several enzymes of the electron transport chain and other cellular metabolic pathways such as mitochondrial translation/tRNA modification, mitochondrial transcription, protein folding, and proteolysis (Fischer et al., 2013;Cole et al., 2015;Szczepanowska et al., 2016). Loss of mitochondrial CLPP in mice leads to general infertility in both males and females, reduction in physical growth, and sensorineural deafness (Gispert et al., 2013;Szczepanowska et al., 2016). Remarkably, deletion of mitochondrial CLPP leads to improved health and increased life span in the filamentous fungus Podospora anserina (Fischer et al., 2013). In this study, loss of plant mitochondrial CLPP did not result in any apparent defects in plant growth and development. Therefore, mitochondrial CLPP has diverse impacts on growth and development based on phenotypic analysis among different eukaryotic organisms, potentially due to redundant roles in the mitochondrial protease network.
In this study, we observed that loss of CLPP2 affected the expression of only ;0.1% of nuclear genes ( Fig. 2A) but nearly 33% of mitochondrial genes ( Fig. 2A), indicating an important impact of this protease on the landscape of mitochondrial gene expression patterns. Mitochondrial ATP-dependent proteases, such as LON and CLPP, are proposed to serve roles in mitochondrial DNA functions including packaging and stability, replication, transcription, and translation (Matsushima and Kaguni, 2012). CLPP may play an indirect role in the regulation of mitochondrial DNA replication, transcription, and translation (Matsushima and Kaguni, 2012). We observed that loss of CLPP2 caused high expression levels of genes encoding potential RNA/DNA polymerases (ATM00490 and ATMG00810), a putative intron maturase (ATMG00520), and a nucleic acid-binding protein (ATMG00560; Fig. 2). In addition, the abundance of mitochondrial RNA helicases RH9 (AT3G22310; Fig. 3) and RH53 (AT3G22330; Supplemental Fig. S2) increased in both mutants relative to the wild type. Similarly, the protein abundance of chloroplast AtRH3 (AT5G26742), which is closely related to mitochondrial RH9 and RH53 (Asakura et al., 2012), was strongly (.5-fold) accumulated in clpr2-1 and clpr4-1 (Rudella et al., 2006;Kim et al., 2009;Zybailov et al., 2009). Chloroplast RH3 DEAD Box RNA Helicases in maize (Zea mays) and Arabidopsis function in splicing of specific group II introns and affect chloroplast ribosome biogenesis (Asakura et al., 2012). So far, there is no detailed information on the global impact of loss of chloroplast CLPP on expression of chloroplastic genes, but such studies are likely to be hard to interpret given the severity of the plastid phenotypes of these mutants.
The effect of CLPP2 loss on mitochondrial accumulation of nuclear-encoded proteins for complexes with mixed subunits derived from both mitochondrial and nuclear genomes can be easily seen for complex I, complex V, and the mito-ribosome (Fig. 4, A, E, and F). This is reminiscent of previous evidence that a posttranslational process is responsible for balancing the availability of nuclear subunits for complex assembly (Sarria et al., 1998;Giegé et al., 2005). We have shown through a detailed examination of complex I that this effect is primarily due to an excess of subunit availability and not to a block on import, preprotein processing, or accumulation of assembled and functional complexes, indicating that CLPP2 prevents such protein accumulations. It is known that CLPP targets subunits within protein complexes in a range of other organisms, providing a housekeeping mechanism for mixed-origin complexes (Flynn et al., 2003;Feng et al., 2013;Fischer et al., 2013;Szczepanowska et al., 2016). Indeed, Arabidopsis orthologs of CLPP substrates identified by trapping approaches in other organisms (including the 24-and 75-kD subunits of complex I) were identified as DEPs in this study (Supplemental Table S3).
Whereas Giegé et al. (2005) previously identified the phenomenon of nonstoichiometry of mixed-origin subunits in a transition from starved to refed Arabidopsis callus culture cells, we found here that carbon starvation did not produce a phenotype in clpp2 plants (Supplemental Fig. S5), but transient response to nitrogen starvation did (Fig. 6). This implies that nonstoichiometric accumulation of unused subunits is not in itself egregious for plant mitochondrial function ( Fig. 5; Supplemental Fig. S5). The reasons for the specific need for mixed subunit complexes to turn over in this way are unknown. However, there is additional evidence for such maintenance of complex I. The nuclear-encoded matrix arm subunits of complex I, including the 75-kD, 51-kD, 39-kD, 18-kD, and B17.2 subunits, exhibited relatively higher turnover rates compared to other subunits of complex I in Arabidopsis (Li et al., 2013). Blue native activity assays and targeted proteomic analysis indicated that the N-module subcomplex was highly accumulated in the matrix fraction in both mutants, indicating increased stability of this separated part of complex I in clpp2 plants ( Fig. 5;  Supplemental Fig. S6). In CLPP-deficient mice, reduced protein levels in complex I subunits have also been observed (Deepa et al., 2016;Szczepanowska et al., 2016). Recently, it was discovered that turnover of the N-module within BN-PAGE-purified intact mitochondrial complexes was faster in proliferating than in differentiated mammalian cell cultures (Szczepanowska et al., 2020). It was proposed that ClpXP protease plays a key role in complex I maintenance, acting through surveillance and replacement of expended N-modules (Szczepanowska et al., 2020). It was shown that to regenerate stalled complex I, ClpXP can recognize, disassemble, and rapidly degrade impaired N-module proteins, allowing an effective replacement of N-module components in preexisting complex I (Szczepanowska et al., 2020). In plants, it is still not clear how dysfunction of CLPP2 impairs the incorporation of complex I subunits into complex I and supercomplex I1III 2 based on in vitro import analyses ( Fig. 5F; Supplemental Fig. S4); however, this new work in mammalian systems provides a hypothesis to be tested. Future work on in vivo protein turnover using stable isotope progressive labeling (Li et al., 2017) would be required to dissect the impact of CLPP2 on the assembly versus the repair of protein complexes in plants.

CRISPR-Cas9-Guided Mutagenesis of clpp2
The guide RNAs were designed for each target at CLPP2 exon1 using CRISPRdirect (Naito et al., 2015) and were further analyzed with E-CRISPR (Heigwer et al., 2014). The cassettes containing two guide RNAs for each mutant were ordered as gBlocks and were inserted into the pHEE2E backbone, provided by Dr. Qi-Jun Chen , by restriction-ligation using BsaI (New England Biolabs; Supplemental Fig. S1). All plants were Arabidopsis (Arabidopsis thaliana) plants (ecotype Columbia-0) grown on soil as described below and were transformed by Agrobacterium-mediated DNA insertion using the floral-dipping procedure (Clough and Bent, 1998). Plants transformed with the CRISPR-Cas9 construct were selected by resistance to hygromycin (15 mg mL 21 ). Genomic DNA was extracted from leaf tissue according to a protocol adapted from Edwards et al. (1991). Primers F1 (ATGAGGGGTCTCGTTTCC G) and R1 (GCCGCCAGGGGAATTGAGA) were used for PCR genotyping. Homozygous T1 plants were determined by the presence of a single trace including the desired mutations after Sanger sequencing. T2 plants were selected for the absence of CRISPR-Cas9 T-DNA by PCR. T2 and T3 seedlings were used for subsequent experiments.

Plant Growth
Arabidopsis seeds (wild type [Columbia-0] and CRISPR-Cas9 knockout mutants clpp2-1 and clpp2-2) were sown on a perlite:vermiculite:compost , supplemented with 15 mg mL 21 hygromycin B, and placed in the dark at 4°C for 3 d. Then plates were transferred to long-day conditions as described above. For root length measurements, seeds were surface sterilized and placed on Murashige and Skoog Gamborg B5 plates as described above in a vertical orientation. For root stress treatments, Arabidopsis seedlings were grown on vertical Murashige and Skoog agar (pH 5.8, 1% [w/v] Suc) plates, with addition or removal of specific growth conditions: rotenone (addition of 5 mM rotenone), antimycin A (addition of 50 mM antimycin A), doxycycline (addition of 1 mg L 21 doxycycline), salt stress (addition of 100 mM NaCl), osmotic stress (addition of 175 mM mannitol), carbon starvation (no Suc supplied), and heat (seedlings on plates grown at 30°C day/night).
For mitochondria isolation and RNA extraction, Arabidopsis seeds were surface sterilized and dispensed into one-half strength Murashige and Skoog Gamborg B5 liquid media within enclosed, sterilized 100 mL polypropylene containers. The containers were seated on a rotating shaker in the long-day conditions as described above and seedlings were harvested 2 weeks after germination.

RNAseq Analysis
RNA was extracted from three biological replicates per genotype using the Spectrum Plant Total RNA Kit (Sigma-Aldrich) according to the instructions. The RNA quality and quantity were assessed using a nanodrop ND-1000 spectrophotometer (ThermoFisher). RNAseq library preparation and 50-bp single-end read sequencing on BGISEQ-500 was performed by the BGI Group. RNAseq data were received from BGI as fastq file format. Following quality control with FastQC (https://www.bioinformatics.babraham.ac.uk/ projects/fastqc/), on average, 42 million reads were obtained per sample, and these were mapped to Arabidopsis reference genome Ensembl v34 and quantified with Salmon (v0.14.0; Patro et al., 2017). Differential analysis was performed with the R package DESEQ2 (Love et al., 2014), and transcripts with a log2FC exceeding 60.4 for the mutant relative to the wild type and an adjusted P-value #0.05 were considered differentially expressed.
RNA sequencing data was deposited to the Gene Expression Omnibus ( Barrett et al., 2013) with the accession number GSE141942.

Mitochondria Isolation, Protein Purification, and Trypsin Digestion
The 2-week-old hydroponically grown Arabidopsis seedlings were used for mitochondria isolation . The isolated mitochondria were quantified using the Bradford (1976) method and the aliquots were stored at 2 80°C until further analysis.
For MRMs used to confirm CLPP2 loss at the protein level, 200 mg mitochondria was precipitated with 93 volumes ice-cold acetone for 24 h at 220°C and centrifuged at 20,000g for 20 min at 4°C. The pellets were collected for further analysis. For quantitative untargeted MS and MRM MS analysis of different mitochondrial fractions, 200 mg mitochondria were lysed using 33 freeze-thaw cycles (20 min at 220°C followed by 20 min at 4°C) and centrifuged at 20,000g for 20 min at 4°C. The supernatant (soluble fraction) and pellet (membrane fraction) were collected and precipitated with 93 volumes cold acetone for 24 h at 220°C. The pellets from soluble and membrane fractions were collected for further targeted MS analysis using MRM. For quantitative untargeted MS and MRM MS experiments, samples were alkylated and trypsin digested as follows: The above acetone-precipitated pellets were resuspended with 100 mL solution containing 50 mM ammonium bicarbonate and 10 mM dithioreithol (pH 8.0), and incubated at 60°C for 30 min. Samples were cooled to room temperature and alkylated with 100 mL 50 mM ammonium bicarbonate and 25 mM iodoacetamide for 30 min. Samples were trypsin digested by adding digestion solution (1:

Quantitative Untargeted MS
Samples with 1 mg protein were analyzed by liquid chromatography (LC)-MS on a Thermo orbitrap fusion mass spectrometer using data-dependent acquisition. Analysis consisted of direct injection onto a self-packed 150 mm 3 75 mm Reprosil-Pur 120 C18-AQ 1.9-mm column (Dr. Maisch). Water/acetonitrile gradients with 0.1% (v/v) formic acid were formed by an Ultimate U3000 nano pump running at 250 nL min 21 from 2% to 27% (v/v) acetonitrile over 30 min.
Thermo raw files were database searched and quantified using MaxQuant (v1.6.7.0; Cox and Mann, 2008) and analyzed using the R package DEP (Zhang et al., 2018). Based on PCA analysis, one wild-type sample and one clpp2-2 sample were flagged as outliers (showing high overall variance compared to all other samples) and removed.

Targeted MRM MS
MRM MS was conducted as described previously (James et al., 2019). In brief, 10 mg protein in 10 mL of each sample were loaded onto an AdvanceBio Peptide Map column (2.1 3 250 mm, 2.7 mm particle size; part number 651750-902, Agilent), using an Agilent 1290 Infinity II LC System. The column was heated to 60°C. Peptides were eluted over a 15-min gradient (0-1 min, 3% [v/v]  In total, 12 samples corresponding to four biological replicates of each genotype i.e. wild type, clpp2-1, and clpp2-1 were used and divided into two experiments. Briefly, isolated mitochondria pellets (; 100 mg of protein) of each genotype and their respective biological replicates were first solubilized with 10 mL of 10% (w/v) SDS containing complete mini protease inhibitor cocktail and samples were diluted to 100 mL with 50 mM NH 4 HCO 3 buffer (pH 7.8). Reduction and subsequent alkylation steps were carried out with 10 mM DTT and incubation at 56°C for 30 min; followed by alkylation of free thiol groups using 30 mM iodoacetamide and incubation at room temperature for 30 min in dark. Next, each sample was diluted 10-fold with ice-cold ethanol in 1:10 ratio, vortexed and stored at -40°C for 60 min followed by centrifugation in a precooled (4°C) centrifuge at 18,000g for 30 min. Next, the supernatant was discarded and to each protein pellet 100 mL of ice-cold acetone were added, briefly vortexed, and centrifuged as above. The supernatant was discarded and the protein pellets were dried under a laminar flow hood. Labeling of proteins using reagents for isobaric tags for relative and absolute quantitation, enrichment of the N-terminal peptides based on the ChaFRADIC method, LC-MS/ MS, and data analysis were performed as previously described (Venne et al., 2013(Venne et al., , 2015. MS proteomics data were deposited to the ProteomeXchange Consortium via the PRIDE (Perez-Riverol et al., 2019) partner repository with the dataset identifier PXD016263 (reviewer account details: Username, reviewer75079@ebi.ac.uk; Password, 2y7IvNFO).

Sequence Logo Analysis
Sequence logos of N-terminal peptides detected by ChaFRADIC MS were produced using the R package "ggseqlogo". Sequence logos are created from frequency data, whereas the ChaFRADIC MS results finally yield intensity data. The intensities were normalized and transformed into frequency data. RH9 (AT3g22310) and RH53 (AT3g22330) were identified as outliers (i.e. they contributed .503 sequences) and removed for this calculation. Transformation and normalization were performed using the following rules. The raw peptide intensity was normalized to the median of all peptide intensities. Transformation was achieved using a normalized intensity of #0.1 at a frequency of 1 followed by normalized intensity of #0.2 at a frequency of 2, and then increasing the frequency from 1 to 25 by 0.1 increments. Sequences were then multiplied based on their frequencies and the sequence logo was calculated.

Complex I Enzymatic Activity Assay
Complex I enzymatic assays were carried out as described previously (Huang et al., 2015). Deamino-NADH:Q reductase (complex I) specific activity was measured by the following reaction. The reaction solution contained 50 mM Tris-HCl [pH 7.2], 50 mM NaCl, 1 mM FeCN, and 0.2 mM deamino-NADH. The decrease in A 420 at 25°C was recorded after adding 10 mg mitochondrial protein to a 1-mL reaction solution. The extinction coefficient of FeCN at 420 nm is 1.03 mM 21 cm 21 .

BN-PAGE and Complex I In-Gel Activity Staining
For gel separation, 300 mg mitochondrial proteins were dissolved with 5 mg digitonin mg protein 21 and then separated with a 4.5% to 16% gradient BN-PAGE gel (Schertl and Braun, 2015). Part of the gel was stained with Coomassie blue. For the complex I in-gel activity assay, the gel was washed three times with distilled water for 5 min and incubated in the reaction medium (0.14 mM NADH, 1.22 mM NBT, and 0.1 M Tris-HCl [pH 7.4]). When the dark blue stain was strong enough, the reaction was stopped by transferring the gel to 40% (v/ v) methanol and 10% (v/v) acetic acid (Schertl and Braun, 2015).

In-Gel Spot Identification
Gel spots (;1 mm 3 ) of interest were excised from the blue-native gel, destained twice in appropriate volumes of 50% (v/v) methanol:50% (v/v) 50 mM ammonium bicarbonate and 0.1% (v/v) formic acid for 30 min at 1,000 rpm, twice dehydrated in appropriate volumes of 50% (v/v) acetonitrile:50% (v/v) 50 mM ammonium bicarbonate and 0.1% (v/v) formic acid for 1 min, and for 30 s in 100% (v/v) acetonitrile. Samples were then reduced, alkylated, dehydrated, digested, and cleaned up as described above in appropriate volumes and identified using an Agilent 6550 Q-TOF as described previously (Nelson et al., 2014).

Protein Import Assays
[ 35 S]-Met-labeled proteins were translated using the rabbit reticulocyte TNT in vitro transcription translation kit (Promega) with the full-length complementary DNA of the 24-kD subunit (AT4G02580) cloned into pDEST14 as previously described (Ivanova et al., 2019). In vitro protein import assays were carried out as described previously in 30 mg of freshly isolated mitochondria (Duncan et al., 2015). BN-PAGE import assays were carried out using [ 35 S]-Metlabeled proteins as previously described (Duncan et al., 2015). In brief, radiolabeled protein was incubated with 30 mg of freshly isolated mitochondria for 2 h, and the mitochondria were pelleted by centrifugation and processed for BN-PAGE analysis (Eubel et al., 2005), except that commercially available precast native-PAGE gels were used (Novex Tris-Gly, Thermofisher). Full-length complementary DNAs of the 51-kD (AT5G08530) and 23-kD (AT1G79010) subunits, CAL1 (At5g63510), and B14.7 (At2g42210) of complex I were cloned using Gateway technology into pDONR201 and subsequently recombined into pDEST14 for [ 35 S]-Met-labeled protein translation as described above.

Transmission Electron Microscopy
Root tissues of 2-week-old hydroponic seedlings were used, and the analyses were performed in Australian National University's Electron Microscope laboratory according to the method described previously (Hyman and Jarvis, 2011). The dissected root tissues were fixed overnight in 2.5% (v/v) glutaraldehyde:2% (v/v) paraformaldehyde in 0.1 M sodium phosphate buffer (pH 7.2) and then postfixed with 1% (w/v) OsO4:1.5% (w/v) potassium ferricyanide. The tissues were then dehydrated gradually in ethanol and propylene oxide, embedded in Spurr's resin, and cured for 24 h at 70°C. Ultrathin sections;70 nm thick were cut using a Leica UC7 ultramicrotome, collected onto copper mesh grids, and stained with 1% (w/v) uranyl acetate and Reynolds' lead citrate. The images were taken with a Hitachi7100 transmission electron microscope at 75 kV. Images were acquired using a Gatan Orius CCD camera.

Accession Numbers
Sequence data from this article can be found in the GenBank/EMBL data libraries under accession numbers AT5G23140, AT5Gg53350, AT5G49840, and AT1G33360.

Supplemental Data
The following supplemental materials are available.
Supplemental Figure S1. Cloning history of the plasmid used to create the CRISPR-Cas9-mediated stable CLPP2 knockout lines.
Supplemental Figure S2. Changes in protein abundances in the soluble fraction of isolated mitochondria from CLPP2 mutants compared with the wild type.
Supplemental Figure S3. Sequence logos of N-terminal peptides obtained from ChaFRADIC MS of purified mitochondrial proteins from the wild type, clpp2-1, and clpp2-2.
Supplemental Figure S4. BN-PAGE separations of in vitro imports of radio-labeled complex I subunits to determine the rate of label incorporation into complex I and supercomplex I1III 2 .
Supplemental Figure S5. Arabidopsis seedlings of the wild type, clpp2-1, and clpp2-2 grown on Murashige and Skoog agar plates under stress conditions or additional chemicals.
Supplemental Figure S6. Schematic position of N-module subunits in complex I and the unassembled N-module subcomplex in the mitochondrial matrix.
Supplemental Table S1. Peptide transition list used for MRM MS analysis.
Supplemental Table S2. Changes in protein abundance in membrane and soluble fractions of isolated mitochondria in CLPP2 mutants compared with the wild type.
Supplemental Table S3. List of subunits in mitochondrial complexes as substrates for CLPP targets through trapping approaches in other organisms.