The evolutionarily conserved iron-sulfur protein INDH is required for complex I assembly and mitochondrial translation in Arabidopsis [corrected].

Respiratory complex I is composed of 48 protein subunits of both nuclear and mitochondrial genetic origin. This study provides a functional characterization of INDH, which is required for the assembly of complex I but is not part of it. Unlike other complex I mutants, INDH is also involved in male and female gametogenesis and plays a role in mitochondrial translation. The assembly of respiratory complexes is a multistep process, requiring coordinate expression of mitochondrial and nuclear genes and cofactor biosynthesis. We functionally characterized the iron-sulfur protein required for NADH dehydrogenase (INDH) in the model plant Arabidopsis thaliana. An indh knockout mutant lacked complex I but had low levels of a 650-kD assembly intermediate, similar to mutations in the homologous NUBPL (nucleotide binding protein-like) in Homo sapiens. However, heterozygous indh/+ mutants displayed unusual phenotypes during gametogenesis and resembled mutants in mitochondrial translation more than mutants in complex I. Gradually increased expression of INDH in indh knockout plants revealed a significant delay in reassembly of complex I, suggesting an indirect role for INDH in the assembly process. Depletion of INDH protein was associated with decreased 35S-Met labeling of translation products in isolated mitochondria, whereas the steady state levels of several mitochondrial transcripts were increased. Mitochondrially encoded proteins were differentially affected, with near normal levels of cytochrome c oxidase subunit2 and Nad7 but little Nad6 protein in the indh mutant. These data suggest that INDH has a primary role in mitochondrial translation that underlies its role in complex I assembly.


INTRODUCTION
Mitochondria are the powerhouses of the cell, generating ATP through oxidative phosphorylation (OXPHOS) mediated by a series of large membrane-bound protein complexes I to V. The majority of mitochondrial proteins are encoded in the nuclear genome and imported into the organelle, but ;10 to 50 proteins are encoded by the remnant mitochondrial genome in most eukaryotes. The dual-genetic origin adds an extra layer of complexity to the assembly of complexes I and III to V. Mitochondrial genes need to be faithfully decoded, requiring a complete expression system in the mitochondria. The subunits then need to be assembled in a coordinate manner with the imported, nuclear-encoded proteins and specific cofactors.
Respiratory complex I, NADH:ubiquinone oxidoreductase, oxidizes NADH, linking electron flow to proton pumping across the mitochondrial inner membrane (Brandt, 2006). The large ;1 MD protein complex has an L-shape, consisting of a hydrophilic arm, known as the peripheral arm, protruding in the mitochondrial matrix, as well as a hydrophobic arm embedded in the inner mitochondrial membrane. The peripheral arm carries out the oxidation of NADH and transfers the electrons to ubiquinone over a chain of seven iron-sulfur (Fe-S) clusters. The membrane arm pumps protons from the matrix to the intermembrane space using a suggested piston mechanism (Zickermann et al., 2009;Efremov and Sazanov, 2011). Of the $40 subunits, 7 are encoded by the mitochondrial genome in most eukaryotes and are part of the hydrophobic membrane arm, including NADH dehydrogenase 1 (ND1), ND2, ND3, ND4, ND4L, ND5, and ND6. In plants, an additional two subunits of the peripheral arm are also mitochondrially encoded, which are NADH dehydrogenase 7 and 9 (Nad7 and Nad9), homologs of the 49-kD and 30-kD subunits in bovine, respectively. The complex is assembled in a modular way with the help of at least 10 assembly factors Hoefs et al., 2012;Mimaki et al., 2012). Complex I in plants follows a similar assembly pattern, except that it builds on the assembly of the plant-specific carbonic anhydrase module as an early step Li et al., 2013). It is not yet known whether any of the mammalian assembly factors are conserved in plants.
Given its size and complex structure, it is perhaps unsurprising that defects in complex I function are commonly found. The majority of human OXPHOS disorders are a result of complex I deficiency, caused by mutations in one of the 44 structural subunits, in one of the assembly factors, or in mitochondrial gene expression (Tucker et al., 2011;Hoefs et al., 2012). Some cases of complex I deficiency in humans are caused by mutation of a specific mitochondrial tRNA (Calvo et al., 2010;Swalwell et al., 2011), whereas a growing number of nuclear mutants are being described that affect splicing or editing of the mitochondrial nad transcripts in plants (Colas de Francs-Small and Small, 2013).
An assembly factor called IND1 (for iron-sulfur protein required for NADH dehydrogenase) was recently identified in the yeast Yarrowia lipolytica (Bych et al., 2008b). IND1 has a similar phylogenetic distribution to complex I, although this association was not uncovered in previous bioinformatics and systems studies (Gabaldón et al., 2005;Pagliarini et al., 2008). The 30-kD protein belongs to a subfamily of P-loop NTPases involved in Fe-S protein biogenesis (Bych et al., 2008a;Lill, 2009). The Ind1 homologs are classified based on a predicted mitochondrial targeting peptide and are 30% identical (50% similar) in amino acid sequence (see Supplemental Figure 1 online). Deletion of IND1 in Y. led to an 80% decrease in complex I levels and activity, but no obvious effect on other respiratory complexes or other Fe-S proteins during vegetative growth. The Ind1 protein has a conserved CxxC motif in the C terminus and binds a labile Fe-S cluster. We showed that the cluster could be transferred to another single-subunit Fe-S protein in vitro, but whether it is transferred to Fe-S subunits of complex I in vivo has not yet been technically possible to demonstrate. The human homolog of IND1, Nucleotide Binding Protein 1-Like (NUBPL), is also specifically required for complex I assembly as shown using an RNAi approach in HeLa cells (Sheftel et al., 2009). Moreover, mutations in NUBPL have been linked to complex I deficiency in patients with mitochondrial disease (Calvo et al., 2010;Tenisch et al., 2012;Kevelam et al., 2013). Of the cases described thus far, one NUBPL allele is a null allele, whereas the other allele has both a missense mutation (G56R) and a branch-site mutation leading to a frameshift after Gly272 (Tucker et al., 2012;Kevelam et al., 2013). Reconstruction of the altered C terminus in the Y. lipolytica Ind1 protein showed that the mutant protein is instable (Wydro and Balk, 2013). The branch-site mutation is found in 1.2% (7 of 751, www.ensembl.org) of the European population, but how this affects individuals is currently unknown because little is yet known about IND1 and its homologs.
To learn more about the function of IND1, we studied mutant alleles of the homologous INDH gene (At4G19540) in the model plant Arabidopsis thaliana, in which the gene was first described but has thus far remained uncharacterized (Lezhneva et al., 2004;Bych et al., 2008a). Compared with previous investigations in Y. lipolytica and human cell lines, both vegetative cells, the plant model system allowed for a detailed analysis of mutant phenotypes during reproductive development. In addition, gradual "switch-on" of INDH expression in a knockout mutant helped to unravel primary from secondary phenotypes. Together, our results indicate that INDH has a primary role in mitochondrial translation, associated with gametogenesis defects in the heterozygote, that indirectly but specifically affects the assembly of complex I.

Arabidopsis INDH Is a Mitochondrial Protein of Low Abundance
The INDH protein sequence has a high probability for mitochondrial targeting according to prediction programs (TargetP, Predotar, iPSORT) and was recently found by mass spectrometry in purified mitochondria (Klodmann et al., 2011). To confirm its cellular localization experimentally, antibodies were raised against recombinant INDH and used to detect their target in total cell extract and purified organelles from wild-type seedlings. A specific immunolabeling signal of 30 kD was observed in the mitochondrial fraction, but not in chloroplasts ( Figure 1A). No signal was detected in an equal amount of total protein extract, but this might be a result of the low abundance of INDH, which was estimated to be less than 0.1% of total mitochondrial protein, based on comparison with purified INDH. (A) Immunolocalization of endogenous INDH. Equal amounts of protein (30 mg) from total leaf extract, gradient-purified chloroplasts (Cp), or mitochondria (Mit) were separated by SDS-PAGE, blotted, and labeled with antibodies against the indicated proteins. NBP35 served as a cytosolic marker protein, PetC (Rieske) as a chloroplast marker, and porin as a mitochondrial marker. Known amounts of purified His-tagged INDH were included to assess the abundance of endogenous INDH. (B) Transient expression of INDH fused to GFP under the control of the CaMV 35S promoter in epidermal leaf cells (green). Autofluorescence of chloroplasts is red. Bar = 5 µm. (C) INDH is a soluble matrix protein. Purified mitochondria were subjected to three cycles of freeze-thawing and separated into supernatant (S10) and pellet (P10) by centrifugation at 10,000g for 10 min. Equal amounts (10 mg) of protein were separated by SDS-PAGE, blotted, and labeled with antibodies against INDH, ISU1 (a soluble matrix protein), Nad9 (a subunit of respiratory complex I), and porin (an integral membrane protein). (D) The first 21 amino acids of INDH are important for mitochondrial targeting. The full-length coding sequence of INDH, or INDH lacking the first 21 codons, was cloned into the pRS416 vector and expressed in the yeast S. cerevisiae. Cells were fractionated into total cell extract (Tot), mitochondria (Mit), and postmitochondrial supernatant (PMS) by differential centrifugation. Equal amounts of protein (10 mg) were subjected to immunoblotting with antibodies against INDH, and against the S. cerevisiae proteins Leu1 and Nfs1 as cytosolic and mitochondrial markers, respectively.
To consolidate the fractionation and immunolabeling results, the cDNA of INDH was cloned in frame with green fluorescent protein (GFP) under the control of the cauliflower mosaic virus (CaMV) 35S promoter and transiently expressed in Arabidopsis epidermal cells. Fluorescence was detected in a punctate pattern surrounding the chloroplasts, characteristic of mitochondria ( Figure 1B).
To establish in which mitochondrial subcompartment INDH resides, mitochondria were separated in a soluble matrix fraction (S10) and a membrane pellet fraction (P10) and subjected to immunoblot analysis. INDH was enriched in the soluble fraction, similar to the Fe-S scaffold protein Iron-Sulfur cluster U (ISU1) Figure 1C). By contrast, the Nad9 subunit of complex I and the integral membrane protein porin were, as expected, enriched in the membrane fraction.
Based on the electrophoretic mobility of INDH ( Figure 1A) and the calculated molecular mass of full-length protein, the Nterminal targeting sequence is predicted to be relatively short, with a cleavage site after Tyr22 (TargetP). Indeed, removal of the first 21 codons severely compromised import of INDH into mitochondria in the yeast Saccharomyces cerevisiae ( Figure 1D). Whereas expression of full-length INDH targeted the protein exclusively to the mitochondrial fraction (Mit), ;50% of the truncated protein was found in the cytosolic fraction (postmitochondrial supernatant).
In summary, these data show that INDH is localized in the mitochondrial matrix, to which it is directed by a canonical N-terminaltargeting peptide.

The INDH Gene Is Important for Early Vegetative Growth
To study the function of INDH, we searched the Arabidopsis stock centers for insertion mutants. Only one insertion line was found: GK_956A05 in the Gabi-Kat collection. In this line, the T-DNA is inserted in the seventh exon (Figure 2A), 20 bp upstream of the codons for the conserved CxxC motif. T3 seed from a heterozygous plant was germinated on basic salt medium and 10-d-old seedlings were analyzed by PCR for the presence of the T-DNA insertion and wild-type allele. Heterozygous and wild-type seedlings segregated in a ;1:1 ratio and were similar in appearance, but no homozygous indh mutants were found (Figures 2B, left panels, and 2C; see Supplemental Tables 1 and 2A online). The segregation ratio was confirmed by scoring seedlings for resistance to sulfadiazine, the selection marker contained within the T-DNA, which was 142:129 (resistant: sensitive). The ratio is clearly different from the Mendelian 2:1 ratio expected for an essential gene.
Upon closer inspection, we noticed that many seeds failed to germinate or that seedlings were arrested in early growth. Therefore, T3 seeds were planted on medium with 1% (w/v) Suc. More than 95% of the seed germinated, 19% of which (175 of 923) was severely delayed in vegetative development ( Figure 2C). Small seedlings were analyzed by PCR and found to be homozygous for indh ( Figure 2B, right panels) and lacking INDH protein (see below). PCR analysis further showed that of 20 normal-sized seedlings, 10 were heterozygous and 10 were wild type. Therefore, on Suc medium, the indh knockout allele segregated 0.7:1:1 (indh:heterozygous:wild type). The indh seedlings were transferred to soil and grown under long-day conditions. They bolted after ;6 weeks and produced seeds after 9 weeks, compared with 4 and 6 weeks, respectively, in the wild type. The indh inflorescences were highly branched and reached only half the height of wild-type plants ( Figure 2D). We also noticed a significant decrease in seed set, both in the heterozygous and homozygous indh plants, of ;50% and 30% compared with the wild type, respectively ( Figure 2E; see Supplemental Table 1 online). There was no bias for seed set with respect to distance from the stigma, arguing against a defect in pollen tube growth.

Heterozygous indh/+ Plants Have Sporophytic Defects in Female and Male Gametophyte Development
To further investigate the unusual segregation ratio of heterozygous indh/+, we examined gametophyte development using microscopy. Seeds that failed to develop were small and white, resembling unfertilized ovules ( Figure 3A). Inspection of immature siliques showed two sizes of ovules: the larger one contained a developing embryo and the smaller ovules arrested at the stage of a mature embryo sac (Figures 3B to 3D).
Pollen development proceeded normally in indh/+ until after meiosis ( Figure 3E). At stage 11 (Sanders et al., 1999), the cytoplasm came away from the pollen wall in ;40% of the cells, followed by collapse of the pollen grain. The tapetum tissue in indh/+ appeared like that of the wild type and underwent its usual developmental program of cell death (Balk and Leaver, 2001).
The 0.7:1:1 segregation ratio and the observed ovule and pollen abortion prompted us to investigate the inheritance of the indh allele from either the male or female heterozygous parent. Reciprocal crosses were performed between indh/+ and wild type, and seeds germinated on selective medium. The sulfadiazineresistance marker present in the T-DNA (indh allele) was efficiently transmitted through both the female and male gametophyte with 96% and 89% efficiency, respectively ( Figure 3G). To verify that the sulfadiazine resistance was linked to the transmitted T-DNA, 14 resistant seedlings derived from the crosses were picked randomly and genotyped by PCR. In all of those, the T-DNA insertion in INDH was indeed detected. The slightly lower transmission efficiencies are in agreement with the observed ;19% homozygous indh segregating from a heterozygous plant on Suc medium (against 25% expected), but it does not explain the 1:1 ratio of heterozygous to wild-type offspring.
To confirm that the collapse of an individual pollen grain was independent of the presence of the indh allele, the recessive quartet1-2 mutation (Francis et al., 2006) was crossed into indh/+ for tetrad analysis. The number of collapsed pollen varied from zero to four in the tetrads ( Figure 3F), although there was a bias for 2:2 segregation (see Supplemental Figure 2 online). These results indicate that the observed pollen death is caused either by a dominant negative effect of INDH disruption in the heterozygote, or by haploinsufficiency in combination with non-Mendelian (cytoplasmic) segregation during gametogenesis. In either case, the collapse of the pollen is a result of a sporophytic defect.

Isolation and Characterization of Fully and Partially Complemented indh Mutant Lines
To confirm that the observed phenotypes were attributable to disruption of the INDH gene, we complemented the GK_956A05 line with cDNA of wild-type INDH (cINDH). The distance between the 59 untranslated region of INDH and the 39 untranslated region of the upstream gene (AT4G19530) is only 100 bp, leaving uncertainty about the length of the INDH promoter, which could either be short or overlapping with the upstream gene. Therefore, cINDH was cloned behind the commonly used CaMV 35S promoter (35S:cINDH) in the pBIN vector. In addition, a second construct was made for expression of cINDH using the promoter of the UBIQUITIN11 gene (UQ11:cINDH) in pART27, because the 35S promoter is known to be poorly expressed in pollen.
Heterozygous indh/+ plants were transformed, and the T1 seeds planted on growth medium with antibiotics to select independent transformants that also carried the indh allele (sulfadiazine resistance). Five of 17 plants were homozygous for indh and contained the cINDH transgene, whereas the other 12 were heterozygous for indh, as confirmed by PCR ( Figure 4A). The expression of INDH protein was assessed by protein blot analysis. The transgenic INDH protein levels were generally much higher than in the wild type, and were easily detectable in total cell extract ( Figure 4B). Vegetative growth in the complemented mutants was like that of the wild type, without any sign of deleterious effects of INDH overexpression ( Figure 2C; see Supplemental Figure 3 online). Pollen viability was fully restored by expression of cINDH from either promoter. However, while seed set was similar to that of the wild type in the UQ11:cINDH lines, it was only 70% in the 35S:cINDH-complemented indh lines, compared with 57% in heterozygous indh/+ plants grown in parallel ( Figures 4C and 4D).
In summary, expression of the wild-type sequence of INDH complements both the vegetative and reproductive phenotypes observed in the indh mutant, demonstrating the importance of INDH for plant growth and development.

INDH Is Required for Complex I Assembly
To investigate whether INDH plays a role in complex I assembly in plants, the levels and integrity of respiratory complexes were analyzed by blue-native polyacrylamide gel electrophoresis (BN-PAGE) and activity staining. Mitochondria were isolated from the wild-type, indh mutant, and complemented seedlings, and protein complexes were solubilized with dodecyl maltoside and separated by BN-PAGE. Gels were stained with Coomassie blue, or incubated with NADH and nitro blue tetrazolium to visualize the NADH dehydrogenase activity of complex I, which runs at the top of the gel because of its large ;1 MD size. Neither Coomassie blue-stained protein nor activity staining was detected at the expected position of complex I in the indh  Figure 5B). By contrast, the assembly and activity of complex I was restored in the line overexpressing wild-type INDH cDNA in the indh knockout line (indh + cINDH). While complex V was intact in the indh mutant, a decrease in the levels of complex III could sometimes be observed ( Figure 5C). The levels of the Fe-S binding protein aconitase and the Fe-S scaffold ISU1 were normal in indh seedlings ( Figure 5B). The results indicate that, like in Y. lipolytica and in humans, depletion of INDH specifically affects complex I assembly and there is little or no effect on other respiratory complexes or mitochondrial Fe-S proteins.
Following recent progress in our knowledge of assembly intermediates of complex I Mimaki et al., 2012), we investigated the assembly of complex I in the indh mutant. Blots of BN-PAGE-separated membrane complexes were probed with antibodies against Nad6 or gamma carbonic anhydrases (CAs). Nad6 is a conserved integral membrane subunit, whereas CAs form a subdomain of the membrane arm of complex I that is unique to plants (Perales et al., 2004). The labeling patterns with either Nad6 or CAs antibodies showed a weak signal at 650 kD in the indh mutant that is absent from wild type. The relatively low intensity of the 650-kD signal compared with fully assembled complex I in the wild type, suggests that only a small amount of the assembly intermediate accumulates. Minor amounts of 450-kD and 400-kD intermediates were revealed upon longer exposure ( Figure 5C). Based on analysis of Arabidopsis mutants in structural complex I subunits, the 650-, 450-, and 400-kD intermediates are thought to correspond to assembly modules of the membrane arm . We also investigated the presence of assembly intermediates and complex I subunits in a weaker allele of indh, named B10 (see below). The 650-kD assembly intermediate is also detected in this line ( Figure 5D). Immunolabeling of the same mitochondrial preparations separated by SDS-PAGE showed that the levels of different complex I subunits varied ( Figure 5E). The level of mitochondrially encoded Nad6 was strongly diminished and Nad7 was slightly increased, whereas Nad9 was decreased to ;50% in the mutant compared with the wild type. The nuclearencoded NADH Dehydrogenase Ubiquinone Fe-S Protein 4 (NDUFS4) subunit was also strongly decreased. Although it is not possible to obtain a full overview of the protein levels of all individual subunits because of the limited availability of antibodies, the results indicate that some complex I subunits are present but not assembled (Nad7, Nad9), whereas others may be degraded or downregulated (Nad6, NDUFS4).

Reassembly of Complex I Is Significantly Delayed upon Expression of INDH Protein
One of the 35S:cINDH-complemented indh homozygous lines, which we named B10, was delayed in early development, but growth was more vigorous than in indh after the initial 2 weeks, which was particularly noticeable in the absence of Suc (see Supplemental Figure 3 online). Immunoblot analysis on isolated mitochondria showed that INDH protein levels in the B10 line were below the detection level in germinating seedlings (stage 0.7, according to Boyes et al. [2001]). However, INDH levels increased gradually during vegetative development, reaching ;250% in mature rosette leaves (stage 3.90) (Figures 6 and 7). The low expression of cINDH in young seedlings correlated with poor resistance to hygromycin, suggesting that the T-DNA was inserted in a heterochromatin region that suppressed expression at early development. Backcrossing and genetic analysis of the B10 line showed that the observed phenotypes were linked to partial complementation of indh and were not attributable to disruption of another gene (see Supplemental Table 2B online).
We noticed that in the B10 line, the level of INDH protein correlated poorly with fully assembled complex I. For example, in the mitochondria preparation used for Figures 5D and 5E, INDH levels were 50% of the wild type, but no fully assembled complex I was detectable by immunoblot analysis. To quantify the relationship between INDH protein and complex I levels, data from different mitochondrial preparations were arranged by growth stage of the seedlings. This reduced variation between B10 seed batches, but also paired B10 and the wild type of the same growth stage. The level of fully assembled complex I was assessed from scans of formazan-stained BN-PAGE gels, and INDH expression was quantified from immunoblotting data, in both cases using ImageJ software. We found that reassembly of complex I lagged significantly behind de-repression of INDH ( Figure 6). Complex I levels in the B10 line reached 100% only at growth stage 3.90, or mature rosettes, which is 2 weeks after growth stage 1.04 when INDH expression reached >100% of wild-type levels. The dynamics of complex I assembly in Arabidopsis was recently described for the first time (Li et al., 2013). Reassembly of complex I was apparent 20 min after reintroduction of the CA2 subunit in isolated mitochondria of a ca2 mutant. Therefore, our data suggest that INDH may have an indirect role in complex I assembly.

INDH Is Required for Mitochondrial Translation
The indirect correlation between INDH and complex I levels suggested that INDH is involved in mitochondrial gene expression. This idea was also hinted at by the unusual combination of phenotypes of indh/+ and indh mutants during the reproductive phase, which are more similar to those reported for mutants in mitochondrial translation than for complex I mutants (Table 1; see Discussion). To investigate whether INDH has a function in translation and how this would correlate with complex I assembly, we isolated mitochondria from different growth stages of the B10 line, followed by 35 S-Met labeling in organello. The earliest growth stage at which we managed to obtain mitochondria were germinating seeds from which the root was just emerging (growth stage 0.5). The total protein pattern of the organellar fraction was similar in B10 and the wild type ( Figure  7A, left panel). However, very little radiolabel was incorporated in the INDH-depleted mitochondria compared with the wild type, indicating that the rate of translation is strongly decreased. We also isolated mitochondria at growth stages 0.7 and 1.02 with 20% and 58% of wild-type INDH protein levels, respectively ( Figure 7B). Both the translation rate and protein pattern changed markedly during seedling germination and greening, concomitant with increased mitochondrial biogenesis and Mitochondria were purified from wild-type (WT) and B10 mutant plants at growth stage 0.7 (hypocotyl and cotyledon emergence), stage 1.02 (green seedlings with two leaves >1 mm), stage 1.04 (four leaves >1 mm), and stage 3.90 (fully grown rosettes). The amount of fully assembled complex I was quantified from BN-PAGE gels, as well as INDH from immunostaining signals using ImageJ software. Representative immunostaining results are shown. Values are as a percentage of wild type at the same growth stage, 6 SD (n = 2). Antibodies against pyruvate dehydrogenase (PDH) E1a were used as a loading control. (A) 35 S-Met labeling of purified mitochondria from wild-type (WT) and INDH-depleted seedlings (B10) at different growth stages. Stage 0.5 is root emergence; see Figure 6 for the other stages. The seedlings were grown under sterile conditions to avoid problems with bacterial contamination in the translation assay. After exposure to reveal the radioactivity signal, total protein was stained with Coomassie blue. The arrowheads point to noticeable differences in band intensities between WT and B10. The assignment of proteins on the right is based on other reports and molecular mass, in parentheses. A question mark indicates that the assignment is tentative. Atp, subunit of ATP synthase (complex V); Cob, cytochrome b; Cox, subunit of cytochrome c oxidase. changes in metabolic functions (Law et al., 2012). Although the rate of translation was restored correlating with expression of INDH, minor differences in de novo synthesis of specific proteins could be seen in B10 mitochondria with 20% INDH (indicated by white arrowheads in Figure 7A; see also the protein profile plot in Supplemental Figure 5 online).
Unfortunately, the mitochondrial in organello labeling pattern is poorly characterized in plants compared with humans. The most prominent upper and lower signals are thought to be Atp1 and Atp9, two highly abundant mitochondrial proteins of the ATP synthase complex (Giegé et al., 2005); however, the identity of other translation products is uncertain. To investigate the steady state levels of mitochondrially encoded proteins, in particular at stage 0.5, mitochondrial fractions were labeled with antibodies against cytochrome c oxidase subunit2 (Cox2), Nad6, and Nad9 ( Figure 7C). Cox2 levels were 50% less in the B10 mitochondria, while Nad6 and Nad9 levels fell below the detection limit. A prominent lower-mass protein of ;18 kD with affinity to the Nad6 antibodies was specifically detected in B10 mitochondria. This may be a breakdown product of Nad6. Note that at stage 1.02 of development when INDH expression was ;50%, Cox2 protein levels were restored to normal and Nad9 levels were 50% compared with the wild type ( Figure 5E). Nad6 protein levels were still severely depleted at this stage.
Last, we investigated whether the decrease in specific mitochondrial proteins was a result of lower transcript levels rather than a posttranscriptional defect. Total RNA was isolated from wild-type and B10 seedlings at different developmental stages and probed for selected mitochondrial and nuclear transcripts using RNA gel blotting ( Figure 7D). mRNA levels of cox2, nad2, nad6, and nad7 and the rRNA rrn26 were normal or increased; however, nad9 and the nuclear-encoded NDUFS4 transcripts were decreased in the B10 line.
These data suggest that INDH has a role in the translation of mitochondrial proteins, with a differential effect on specific polypeptides, which could cause the delay in complex I assembly upon de-repression of INDH.

DISCUSSION
The evolutionary drive for both increased size of OXPHOS complexes and retention of selected genes in the mitochondrial genome comes with a great cost. Many extra proteins are required to orchestrate mitochondrial gene expression, to import nuclear-encoded proteins, and to assemble the numerous subunits and prosthetic groups in an orderly fashion. In addition, the complex assembly process is prone to errors, which can cause loss of fitness, including debilitating diseases. The Ind1 protein was previously identified as an assembly factor of complex I in the yeast Y. lipolytica (Bych et al., 2008b) and subsequently in human cells (Sheftel et al., 2009) and was linked to pathogenic complex I deficiencies (Calvo et al., 2010;Tenisch et al., 2012;Kevelam et al., 2013). Here, we show that INDH, the homolog of Ind1 in the model plant Arabidopsis, is also specifically required for complex I assembly. However, we found that its function(s) may be more compound than initially proposed.
INDH can be defined as an assembly factor of complex I in Arabidopsis in the sense that it is not associated with mature complex I ( Figure 1C) and mutation resulted in the specific loss of complex I ( Figure 5). In addition, slow vegetative growth is generally observed in Arabidopsis complex I mutants (see Supplemental Figure 3 online). Based on its homology to P-loop NTPases involved in Fe-S cluster assembly (Roy et al., 2003;Lezhneva et al., 2004;Netz et al., 2007;Schwenkert et al., 2010), and binding of a transferable Fe-S cluster, we previously proposed that this protein functions as a Fe-S scaffold specific for complex I (Bych et al., 2008b). Indeed, the assembly intermediates ND, not determined. a Segregation is given in the following order, homozygous mutant: heterozygous: wild type. b N16195 has a T-DNA insertion in exon 5 in gene AT5G37510 and was previously found in a screen for embryo-lethal mutants (www.seedgenes.org).
AT5G37510 encodes the 75-kD subunit of complex I, which binds three Fe-S clusters. A very low frequency (1 in 50) of homozygous mutants was obtained, which were viable and produced low numbers of seed. c RPL11 is not an essential ribosomal subunit, but is added for comparison.
of complex I found in Arabidopsis mitochondria depleted of INDH would be consistent with this idea (Figures 5C and 5D). The accumulation of a 650-kD subcomplex is similar to that in Arabidopsis mutants lacking the peripheral arm subunits NDUFS4 or 39-kD (NDUFA9) . The 650-kD subcomplex in the ndufs4 mutant was shown to lack the N-module that contains Fe-S cluster binding subunits. However, the relatively small amounts of 650-kD intermediate in the indh knockout mutant ( Figure 5A) suggest that there may be an additional defect in the assembly of the membrane arm, which does not contain Fe-S proteins. Interestingly, our in-depth characterization of Arabidopsis indh mutant alleles indicates that INDH has a function in mitochondrial translation. One explanation is that the lack of complex I, and consequently lower ATP levels, leads to impaired translation; however, we think this is not the case. First, the 35 S-Met incorporation studies were done using isolated organelles with added ATP and GTP. Second, at growth stage 1.02, complex I levels were only 30%, whereas in organello translation rates were restored to normal (Figures 6 and 7A). Third, recent studies in human cells showed that depletion of early complex I assembly factors resulted in increased proteolysis of the ND1 subunit, rather than a block in its translation (Zurita Rendón and Shoubridge, 2012). The long delay in complex I restoration upon de-repression of INDH ( Figure 6) is also indicative of a translation defect. In human cell lines, reassembly of complex I upon expression of a missing structural subunit took less than 24 h (Dieteren et al., 2012), whereas reassembly after transient inhibition of mitochondrial translation took 2 to 4 d (Ugalde et al., 2004).
Further support for the idea that INDH may play a role in mitochondrial translation comes from the unusual combination of phenotypes in heterozygous indh/+ plants. Aberrant segregation ratios, and abortion of both ovules and pollen but near-normal transmission of the mutant allele ( Figure 3) have also been observed in plants that are heterozygous for insertion alleles of genes encoding essential mitochondrial ribosomal subunits or amino acid tRNA synthetases (Table 1). By contrast, mutants that completely lack complex I (de Longevialle et al., 2007;Meyer et al., 2009) or lack the 75-kD Fe-S binding subunit (Table 1) do not show gametophytic defects in the heterozygotes, and segregate like essential genes. How could loss of one functional allele required for mitochondrial translation cause random abortion of gametophytes? First, reproductive development in plants is known to reveal even minor impairments in OXPHOS, such as in a class of mitochondrial mutants known as cytoplasmic male sterility mutants. Second, nuclear gene expression shuts down during meiosis; therefore, the early stages of gametogenesis rely on sufficient reserves in metabolites but also amino-acylated tRNAs. Apparently, the amount of amino acid tRNA in the mitochondria is only just enough to survive the shutdown period, since haploinsufficiency of a specific amino acid tRNA synthetase can lead to random death of female and male gametophytes after meiosis (Berg et al., 2005). Something similar might occur in the indh/+ mutant. Analysis of the indh allele in the quartet1-2 background showed that death of a particular pollen daughter cell is independent of meiotic segregation of the indh allele or positional cues in the pollen sack ( Figure 3F; see Supplemental Figure 6 online). Instead, it is likely that a daughter cell dies because of uneven segregation of a mitochondrial population with variable translation capacities. It would be interesting to investigate whether cell death is the consequence of low ATP levels or another aspect of mitochondrial function. We have analyzed ATP levels in illuminated rosette leaves, which was not significantly lower in soil-grown indh compared with the wild type despite severe growth retardation in indh (see Supplemental Figure 7 online). Moreover, in seedlings grown on a medium containing Suc, ATP levels were in fact higher in the indh mutant, indicating that mechanisms have been induced to compensate for lack of complex I.
The Arabidopsis tRNA synthetase mutants could perhaps be thought of as the equivalent to human mitochondrial tRNA (mt-tRNA) mutations, which have thus far not been found in plants. Interestingly, mutations in several human mt-tRNAs have been reported to cause isolated complex I deficiency, with no obvious effect on other respiratory complexes. These include the mt-tRNA for Pro, Trp, Lys, Leu (UUR), and Ser (AGY) (Shoffner et al., 1990;Arenas et al., 1999;Blakely et al., 2009;Da Pozzo et al., 2009;Calvo et al., 2010;Swalwell et al., 2011). Although little is known about mitochondrial translation in plants, a recent study (Kwasniak et al., 2013) described how silencing of RPS10, a nuclear-encoded subunit of the small ribosomal subunit in mitochondria, resulted in an altered pattern of in organellotranslated peptides. The protein level and activity of complex I was the most affected of the respiratory complexes. These data indicate that a generic defect in mitochondrial translation can have a differential effect on individual peptides, and that complex I is most vulnerable because of its large number of mitochondrially encoded subunits.
In conclusion, our results indicate that INDH plays an indirect role in complex I assembly, through the biogenesis of mitochondrially encoded proteins. At the moment, our data do not fully exclude the previously proposed role in inserting Fe-S clusters, but this is less likely given the time-dependent difference in INDH levels and fully assembled complex I ( Figure 6). Although the role of INDH in translation could be specific for Arabidopsis, it may be worth investigating for Y. lipolytica and human. It was previously noted that ind1D Yarrowia could not go through meiosis, which is perhaps a result of gametogenesis problems (Bych et al., 2008b); in addition, other respiratory complexes in NUBPL-depleted cells are also mildly affected (Calvo et al., 2010). Ultimately, any further knowledge on the precise role of IND1 and its homologs will help to elucidate why this conserved protein is important for complex I assembly and why certain aspects of mitochondrial biogenesis are critical for both male and female gametogenesis.

Plant Material and Growth
The Arabidopsis thaliana line GK_956A05 was obtained from the Gabi-Kat collection (Kleinboelting et al., 2012). Insertion of a single T-DNA in the INDH gene (AT4G19540) was confirmed by PCR, sequencing of the left border, and the segregation of sulfadiazine resistance. The heterozygous ova2/+, ova6/+, and emb1467/+ lines were obtained from the Nottingham Arabidopsis Stock Centre (stock IDs SALK_088424, SALK_045080, and N16195, respectively). The quartet1-2 mutant was provided by Ian R. Henderson (University of Cambridge). The ndufs4 and rug3-1 mutants were previously described (Meyer et al., 2009;Kühn et al., 2011). Primers used to confirm the genotype of the lines by standard PCR procedures are listed in Supplemental Table 3 online. Plants were grown on agar plates containing one-half-strength Murashige and Skoog salts, with 1% (w/v) Suc when indicated, and antibiotics when required for selection. Seedlings for mitochondria preparations were grown in hydroponic culture (Sweetlove et al., 2007). All plants were grown under a 16/8 h light/dark cycle with a photon flux density of 100-150 mmol m 2 s 21 at 20°C and 65% relative humidity. Because of the developmental delay in the B10 line, we compared plant material at the same growth stages as defined by Boyes et al. (2001).

cINDH Expression
A cDNA corresponding to full-length INDH (cINDH) was cloned in fusion with GFP in the pOL-GFP-S65C vector using the SpeI and SalI restriction sites (Peeters et al., 2000), and expressed transiently using particle bombardment of Arabidopsis leaves. For complementation of the mutant allele, cINDH was cloned behind the 35S CaMV promoter and translational enhancer in the pBIN vector, or behind the UBIQUITIN11 promoter (UQ11) in the pART27 vector (Gleave, 1992). The restriction sites NcoI and XbaI were used to clone cINDH behind the 35S promoter, and AscI and PacI to transfer 35S:cINDH to pBIN. The restriction sites XhoI and BamHI were used to clone cINDH behind the UQ11 promoter, and NotI to transfer UQ11: cINDH to pART27. The pBIN and pART27 vectors carried hygromycin B or kanamycin resistance, respectively, as a selectable marker. Plants were transformed using Agrobacterium tumefaciens-mediated insertion of the T-DNA. Primers for the cloning and for PCR detection are listed in Supplemental Table 3 online.

Microscopy
Immature ovules were imaged with differential interference contrast microscopy after tissue clearing with Hoyer's solution (Ruzin, 1999). Flowers were embedded with polyethylene glycol, sectioned, and stained with toluidine blue. Pollen was stained with Alexander's stain (Alexander, 1969). Stained sections and pollen were analyzed using conventional light microscopy. Green fluorescent protein and chlorophyll autofluorescence were visualized by confocal microscopy (Leica DM RXA).

Purification of Mitochondria and Other Cell Fractions
Plant mitochondria were purified from germinating seeds using differential centrifugation. When using hydroponic seedlings or rosette leaves as source material, the mitochondria were further purified using a Percoll and polyvinylpyrrolidine-40 gradient (Sweetlove et al., 2007). Yarrowia lipolytica mitochondria were purified by differential centrifugation (Daum et al., 1982). Chloroplasts were isolated from rosette leaves. Total cell extract was prepared from leaves by grinding tissue with 2 vol of buffer (50 mM Tris-HCl 8.0, 0.1% [w/v] sodium dodecyl maltoside, 1 mM ETDA, 1 mM phenylmethylsulfonyl fluoride, 1 mM dithiothreitol) and centrifuging for 10 min at 16,000g to remove debris. The protein concentration was determined using the Bradford method (Bio-Rad Protein Assay).

Protein Blot Analysis
Total cell extract or purified organelles were mixed with Laemmli buffer and separated on standard SDS-PAGE gels. Proteins were transferred to nitrocellulose (Schleicher & Schuell BA 83), and labeled with antibodies. Immunolabeling was detected using secondary horseradish peroxidaseconjugated antibodies and chemiluminescence.

Blue-Native PAGE and Activity Staining
Blue-native PAGE was performed as described in Wittig et al. (2006) for activity staining, and as described in Meyer et al. (2011) for immunolabeling. Gels were incubated with 0.2 mM NADH and 0.1% (w/v) nitroblue tetrazolium in 0.1 M Tris-HCl 7.4 for NADH dehydrogenase activities. For immunolabeling, gels were blotted onto polyvinylidene difluoride.

In Organello Translation
Purified mitochondria from plants grown under sterile conditions were labeled with 35 S-Met as previously described with minor modifications (Giegé et al., 2005). In brief, mitochondria (100 mg protein) were pelleted and resuspended in 0.1 mL 0.3 M mannitol, 50 mM HEPES-KOH pH 7.4, 60 mM KCl, 10 mM MgCl 2 , 5 mM KH 2 PO 4 , 0.1% (w/v) BSA, 10 mM Namalate, 10 mM Na-pyruvate, 2 mM GTP, 4 mM ADP, 4 mM ATP, and 2 mM dithiothreitol. Unlabeled amino acid mix without Met (2.5 mL 1 mM each, Promega) was added and the mixture was preincubated for 15 min at 23°C attached to a rotating wheel (15 rpm). After addition of 35 S-Met (30 mCi of EasyTag L-[ 35 S]-Methionine; PerkinElmer), the mitochondria were incubated for 1 h. Label incorporation was stopped by adding 0.5 mL 0.3 M mannitol, 50 mM HEPES-KOH pH 7.4, 10 mM Met. Mitochondria were pelleted by centrifugation and resuspended in 13 Laemmli buffer. The samples were separated on 15% SDS-PAGE gels, dried and exposed to a phosphor-imaging plate. Exposure ranged from 1 week (Percollpurified mitochondria from 10-d-old seedlings) to 12 weeks (mitochondria from germinating seeds). After exposure, the gel was rehydrated and stained with Coomassie Brilliant Blue R 250 (InstantBlue; Expedeon).

RNA Isolation and RNA Gel Blot Analysis
Total cellular RNA was isolated from 30-to 50-mg seedlings using an RNeasy Plant Mini (Qiagen) kit according to the manufacturer's instructions. For RNA gel blot analysis 2.5 to 5 mg of RNA was separated on 1.2% (w/v) agarose gels using formaldehyde as denaturing agent. Gel electrophoresis, blotting, and hybridization were done as previously described (Sambrook and Russel, 2001). Nucleic acids were blotted onto Hybond-N+ membrane (GE Healthcare) and hybridized with 32 P radiolabeled oligonucleotides (cox2, rrn23) or PCR-generated probes (nad2, nad6, nad7, nad9). Oligonucleotides used are listed in Supplemental Table 3 online.

Statistical Analysis
For genetic analysis, the chi-squared test for goodness of fit was applied. Numerical values are the average of independent experimental replicates, with n indicated in the figure legends. Error bars represent the standard deviation.

Supplemental Data
The following materials are available in the online version of this article.  Table 1. Segregation Analysis of the T-DNA Insertion in the INDH Gene.
Supplemental Table 3. Primers Used in This Study.