Plant mitochondria contain the protein translocase subunits TatB and TatC

The twin-arginine translocation (Tat) pathway differs from most other protein translocating systems in that it transports fully folded proteins (Lee et al., 2006). It is named after its targeting signal, which contains a pair of adjacent arginine residues (the twin arginines). Identified in all domains of life, Tat pathways play essential roles in a number of different cellular processes including: bacterial pathogen virulence, cell separation, phosphate and iron metabolism, photosynthetic and respiratory metabolism (Berks, 2015; Palmer and Berks, 2012).

The minimal Tat pathway that is found mostly in archaea and gram-positive bacteria consists of two subunits; TatA [possessing one transmembrane helix (TMH)] and TatC (possessing six TMHs) (Barnett et al., 2008). In general, most Tat systems – including the best-studied; that of Escherichia coli – contain a second functionally distinct member of the TatA family called TatB (Sargent et al., 1999). Outside of prokaryotes, a well-established Tat pathway has also been identified and characterized inside chloroplasts (Robinson and Klosgen, 1994; Cline and Henry, 1996). It is essential for the assembly of photosystem II and the cytochrome b₆f complex in thylakoid membranes (Molik et al., 2001). Genes for Tat pathway components have also been identified in a number of mitochondrial genomes including jakobids, whose mitochondria encode for a minimal Tat system of TatA and TatC (Jacob et al., 2004). TatC-like genes are also found in the mitochondrial genomes of higher plants (Unseld et al., 1997). Despite the conservation of Tat pathways across phyla, it was thought to have been completely lost from all opisthokont lineages, which includes Fungi, Metazoa and relatives. To date only two known exceptions have been described, the choanoflagellate Monosiga brevicollis (Burger et al., 2003), which is interesting as it is thought to be one of the oldest relatives to animals, and a member of the homoscleromorph sponges Oscarellidae (Wang and Lavrov, 2007), both of which have a mitochondrially encoded TatC. However, because most mitochondrial genomes only encode a TatC-like protein, no function has ever been attributed to these proteins.

Here, we present data that demonstrate that in Arabidopsis thaliana the mitochondrial-encoded TatC is not a pseudogene. It is also demonstrated that Arabidopsis mitochondria contain a second member of a Tat pathway, a TatB-like protein that is an essential gene. This mitochondrial TatB-like protein has been long sought after and strengthens the evidence that, in contrast to mitochondria from other eukaryotic lineages, plant mitochondria contain a Tat pathway.

It has been known for more than 20 years that the Arabidopsis thaliana mitochondrial genome contains a gene encoding a TatC-like protein (also known as either orfX or Mttb) (Sunkel et al., 1994; Unseld et al., 1997). However, no functional data has ever been ascribed to it. One of the main reasons for this is that the mitochondrially encoded TatC (mtTatC) has no classical start codon in Arabidopsis and has been thought of as a pseudogene. There are, however, several pieces of evidence to suggest that mtTatC is a functional gene that encodes a protein. The mtTatC transcript in Arabidopsis is edited at 36 individual sites (Bentolila et al., 2013). There are also several studies demonstrating that the Arabidopsis mtTatC is an expressed gene (van der Merwe and Dubery, 2007; Gutierrez-Marcos et al., 2007; Sunkel et al., 1994; Wang et al., 2012). To gather further evidence that mtTatC is a functional gene we searched every plant mitochondrial genome in the NCBI organelle genomes database (http://www.ncbi.nlm.nih.gov/genome/organelle/) for the presence or absence of a TatC-like gene. In all we searched 124 plant mitochondrial genomes that are displayed phylogenetically in Fig. 1A. Out of 124 plant mitochondrial genomes 102 of them contain a TatC gene (Fig. 1A) of which 90 contain classical start codons (Fig. 1A). These 102 mitochondrial genomes contained all sequenced higher plant species (Fig. 1A). The only major difference was observed in the Chlorophyte lineages (green algae), which is split in two; one half containing mainly the class Chlorophyceae and having no TatC gene, and the other half containing mainly the class Trebouxiophyceae, which contain a TatC gene (Fig. 1A). What can be determined is that the majority of plant mitochondrial genomes contain a TatC gene with a classical start codon. We believe that this conservation of a mitochondrial TatC (hereafter mtTatC) must mean it has some function within plant mitochondria, otherwise, like almost all animal, fungi and Chlorophyceae, mitochondrial genomes it would have been lost. Therefore, we first aimed to prove that the Arabidopsis mtTatC gene is translated into a protein.

To this end we raised an antibody against the peptide VREEGWTSGMRESGIEKKNKSSPPPRTW, which corresponds to amino acids 253 to 281 of the Arabidopsis mtTatC. The collected whole serum was then affinity purified against the peptide to obtain a purified antibody. This antibody detected a band of ∼27 kDa in isolated Arabidopsis mitochondria that was not recognized by the pre-immune serum (Fig. 1B). To confirm the specificity of the antibody we pre-incubated it with the peptide, which abrogated binding of the 27 kDa protein (Fig. 1B). Although the protein size of 27 kDa is slightly lower than the predicted size of 33 kDa, it is a well-known phenomenon that TatC and other hydrophobic proteins run at different molecular masses than predicted (Jakob et al., 2009; Mori et al., 2001). From these results it can be concluded that the antibody is specific for the Arabidopsis mtTatC and that the mtTatC gene is functional.

As all known Tat pathways contain at least TatC and TatA but the majority contain TatA, TatB and TatC subunits, for plant mitochondria to have a functional Tat pathway other subunits are required. Thus, we sought to identify any possible TatA- or TatB-encoding genes in the Arabidopsis genome that might be targeted to mitochondria. We first excluded the chloroplast-targeted cpTatA and cpTatB as being dual-targeted to mitochondria, by using antibodies to the Pisum sativum (pea) chloroplast TatA (Mori et al., 1999) and TatB (Mori et al., 2001) proteins (hereafter cpTatA and cpTatB) on isolated chloroplastic and mitochondrial fractions from pea (Fig. 1C). As expected, cpTatA and cpTatB were only detected in the chloroplastic fractions (Fig. 1C). BLAST (Altschul et al., 1990) searches of the Arabidopsis genome using either E. coli TatA or TatB sequences, the Arabidopsis chloroplastic targeted TatA or TatB sequences, or the mitochondrial-encoded sequences of TatA from jakobids were also performed. The only TatA- or TatB-like sequences recovered by this approach were the known chloroplast Tat subunits.

Therefore, we tried another approach and used the program Phyre2 in the back phyre mode (Kelley et al., 2015). This utilizes the known structures of proteins to search genomes for similar proteins. So we used the known structures of the E. coli TatA (PDB: 2NM7) (Zhang et al., 2014b) and TatB (2MI2) (Zhang et al., 2014c) (hereafter EcTatA and EcTatB) to search the Arabidopsis nuclear genome (Table S1). Whereas both the known cpTatA and cpTatB proteins were identified with high confidence scores (Phyre2 confidence scores are a representation of the probability that the match between the sequence and the template is a true homology) a third unknown protein was also identified with similar confidence scores with the locus At5g43680 (Table S1). In fact, At5g43680 came back as the highest-ranked protein when using the structure for EcTatB as the search reference.

Using the protein sequence of At5g43680 we performed BLAST searches of higher plant genomes. In all cases a protein similar to At5g43680 could be identified. To confirm that what we had identified was a TatB or TatA subunit, phylogenetic analysis was utilized using TatA and TatB subunits from a variety of bacterial and jakobid species along with the chloroplastic TatA and TatB subunits from plant species (Fig. 2). We observed that the At5g43680 and related proteins grouped most closely with the TatB proteins from bacteria, indicating that At5g43680 is a possible TatB-like protein (Fig. 2). Although some of the bootstrap values are low we believe the tree is accurate. We also tested Bayesian homology using MrBayes (Ronquist et al., 2012) and an almost identical tree was obtained (data not shown). Additionally, when BLAST searches excluding plant species were carried out with At5g43680 the only hits were bacterial TatB proteins (data not shown). Therefore, we believe that At5g43680 is a TatB-like protein. This is also supported by basic sequence comparisons as At5g43680 shows a 14% identity and 29% similarity to EcTatB in comparison with 10% identity and 19% similarity to EcTatA. Furthermore, a sequence alignment between EcTatB and At5g43680 showed that it contains the conserved glutamate residue at amino acid position 8 at the start of the transmembrane helix (TMH) and also the invariant glycine between the TMH and the second alpha helix (APH) (Fig. 3A). One major difference, however, is the length; At5g43680 is 61 amino acids longer that EcTatB, with the difference found predominantly in the C-terminus.

To get a better understanding of the possible function of At5g43680 we built a model of its 3D structure. This was carried out using Phyre2 in the intensive mode, which returned a model fitting the identified structure of EcTatB (Fig. 3B) (Zhang et al., 2014c; Kelley et al., 2015). The model is based on the first 104 amino acids of EcTatB and the first 125 amino acids of At5g43680 as the C-termini of both proteins are thought to be unstructured. EcTatB contains four alpha helices that adopt an L-shape conformation (Fig. 3B) (Zhang et al., 2014c). The model we obtained for At5g43680 has an almost identical structure containing four alpha helices in an L-shape conformation, further supporting the phylogenetic analysis that it is a TatB-like protein. To determine the quality of our model we calculated an overall RMSD of 1.866 and a Q-score of 0.501 when compared with EcTatB using Chimera (Pettersen et al., 2004). This indicates that our model for At5g43880 is very close to the structure of EcTatB as evidenced by the superimposed structures in Fig. 3B. All these data indicate that At5g43680 is most likely a TatB-like protein.

As yet little or no data is available about the localization of At5g43680. We only know that several peptides of this protein have been identified in whole protein fractions in the pep2pro proteomics database (Baerenfaller et al., 2011), indicating that it is an expressed protein. To determine the subcellular localization of At5g43680 we first performed in vivo GFP targeting analysis. Firstly, we cloned the full-length coding sequence for At5g43680 in frame with a C-terminally located GFP tag. Tobacco leaves were infiltrated with agrobacterium carrying the At5g43680–GFP construct. Analysis of the GFP expression in protoplasts isolated from the transformed tobacco leaves displayed a GFP signal that overlapped with mitochondria as visualized using mitotracker (Fig. 3C). As At5g43680 colocalized with mitochondria we decided to rename it as AtmtTatB.

To test the AtmtTatB–GFP localization results, in vitro import assays into isolated Arabidopsis mitochondria were performed using radiolabeled precursor proteins. When AtmtTatB was incubated with isolated mitochondria, a protease-resistant product was observed after incubation with proteinase K of the same size as the precursor protein (Fig. 4A, lanes 1–3), indicating that AtmtTatB is imported into mitochondria but does not contain a cleavable pre-sequence. Pre-treatment of the mitochondria with the ionophore valinomycin prior to import abolished the appearance of the protease-resistant band of AtmtTatB (Fig. 4A, lanes 4 and 5). Valinomycin dissipates the mitochondrial membrane potential that is required for the import of proteins into or across the inner membrane, indicating that AtmtTatB is imported into the matrix or inner membrane of plant mitochondria. When the outer membrane was removed prior to the addition of proteinase K, AtmtTatB remained protease-resistant, indicating that if it is located in the inner mitochondrial membrane, the majority is facing towards the matrix (Fig. 4A, lanes 6–9). Finally, the addition of 1% Triton X-100 prior to protease treatment demonstrated that when total mitochondria were ruptured AtmtTatB is accessible to the added protease (Fig. 4A, lane 10). The Glycine max Alternative oxidase 1a (AOX) and Arabidopsis Translocase of the inner membrane protein of 23 kDa (Tim23) proteins were used as controls in the import assays as both are associated with the mitochondrial inner membrane. AOX was imported in a membrane-potential-dependent manner and processed to a 32 kDa mature protein that was protease resistant even when the outer membrane was ruptured (Fig. 4A, lanes 1–9). Tim23 was also imported in a membrane-potential-dependent manner and was shown to be protease resistant in intact mitochondria but produced its characteristic smaller membrane-protected fragment after outer membrane rupture and protease treatment (Fig. 4A). Both proteins were also fully digested by protease K when the mitochondria were ruptured by 1% Triton X-100 treatment (Fig. 4A, lane 10). These experiments demonstrated that AtmtTatB is targeted to and imported into mitochondria and is most likely located either within the mitochondrial inner membrane or mitochondrial matrix.

To further test the mitochondrial location of AtmtTatB we raised an antibody against the full-length protein. Immunoblotting against Arabidopsis mitochondria using the purified AtmtTatB antibody detected a band of 30 kDa that was absent when the pre-immune serum was used (Fig. 4B). Incubation of the antibody with the antigen prior to western blotting abrogated binding of the antibody to this 30 kDa protein, indicating that the antibody is specific (Fig. 4B). Using the antibody against mitochondria and ruptured mitochondria treated with proteinase K indicated that AtmtTatB is resistant to protease treatment even when the outer membrane is ruptured (Fig. 4C), supporting the import results showing that AtmtTatB is located either in the inner mitochondrial membrane or the matrix. To verify that AtmtTatB is located in the membrane, mitochondrial membrane fractions were extracted with carbonate. Like the membrane control COXII, AtmtTatB was located in the pellet membrane fraction, as opposed to the soluble fraction that contained subunit H of the glycine decarboxylase (Fig. 4D). Putting these results together we could demonstrate that AtmtTatB is located within the inner mitochondrial membrane, most likely with its C-terminus facing the matrix. A similar orientation of the EcTatB protein within the cytoplasmic membrane has been described previously (Koch et al., 2012).

As AtmtTatB is nuclear-encoded, it is possible to test T-DNA insertion lines for its physiological role to be inferred. We identified two potential insertion lines in CSHL_GT11254 and SALK_003481, and characterized them. The T-DNA insertion for CSHL_GT11254 was found to be directly downstream of the ATG start codon and the T-DNA insertion for SALK_003481 was found to be located in exon five (Fig. 5). We attempted to identify homozygous plants using PCR; however, for both lines we never obtained homozygous plants, only heterozygous and wild-type plants were found. When we examined the siliques of self-fertilized heterozygous plants we observed that one quarter of all embryos were aborted (displayed by the white embryos in Fig. 5). This corroborates the observation that offspring of heterozygous plants produced heterozygous to wild-type ratios of 2:1 (Fig. 5). To confirm these results, we PCR-screened the progeny of self-fertilized heterozygous plants. After screening 140 plants from the line CSHL_GT11254 we obtained 95 heterozygous plants and 45 wild-type plants, which gives a ratio of 2.1:1. After screening 157 plants from the line SALK_003481 we obtained 106 heterozygous plants and 51 wild-type plants, which also gives a ratio of 2.1:1. These results are almost identical to the predicted results of a heterozygous-to-wild-type ratio of 2:1 as predicted by Mendelian genetics for a self-fertilized embryo lethal gene. We conclude that AtmtTatB is an essential gene that causes the homozygous null mutant embryos to abort.

In other organisms the TatB and TatC subunits are normally found in a stable complex termed the TatBC complex (Cline and Mori, 2001; Bolhuis et al., 2001). To determine if the AtmtTatB (henceforth mtTatB) and mtTatC proteins are in a stable complex of similar molecular weights we used two-dimensional Blue Native (BN)-SDS-PAGE followed by immunoblotting. Using Tom40 from the TOM complex (300 kDa), COXII from complex IV (240 kDa) and Qcr7 from complex III (500 and 1500 kDa) as controls, we could determine that mtTatB is located in one complex with the molecular weight of 1500 kDa and mtTatC was located in complexes of molecular weights of 1500 kDa and 150 kDa (Fig. 6). As the super-complex of complexes I and III also runs at 1500 kDa it could be interpreted that mtTatB and mtTatC are part of these complexes. However, extensive work on identifying the subunits of plant mitochondrial complexes I and III has never identified either mtTatB or mtTatC (Meyer et al., 2008; Peters et al., 2013; Klodmann and Braun, 2011; Klodmann et al., 2010), which means it is unlikely they are part of those complexes. The fact that both mtTatB and mtTatC are both found in a high molecular weight complex of the same size is a good indication that they form a stable complex together in a similar manner to that of other organisms.

In yeast and mammalian mitochondria the protein Bcs1 is responsible for the insertion of the Rieske Fe/S (Rip) protein into complex III (Wagener et al., 2011), which has also been hypothesized as a substrate for a potential mitochondrial Tat pathway (Hinsley et al., 2001; Pett and Lavrov, 2013). Thus, we sorted a selection of organisms based on their mitochondrial evolution in the same manner as before by using the protein sequences of COXI and COB. Next, we overlaid the structural arrangement of that organism’s Bcs1 and then also included whether or not that organism contains a mitochondrial TatC and a mitochondrial TatB-like protein (Fig. 7). We observed that the majority of organisms that contain a complete Bcs1-like protein encompassing both the Bcs1 and AAA ATPase domains do not contain genes encoding either mitochondrial TatC or mitochondrial TatB-like. The opposite can then also be observed; the majority of organisms with genes that do encode a mitochondrial TatC and mitochondrial TatB-like do not possess a protein with a Bcs1 domain, but only the AAA ATPase domain. This is in agreement with what was previously reported (Pett and Lavrov, 2013), the only two exceptions being Naegleria gruberi and Chlamydomonas reinhardtii. This observation indicates that organisms that have evolved a complete Bcs1 protein, i.e. the AAA ATPase and Bcs1 domains, have been able to lose their mitochondrial Tat pathway as it is no longer required for Rip insertion. Conversely, organisms that do not contain a complete Bcs1 protein have retained their mitochondrial Tat pathway, potentially required for Rip insertion. A similar theory was also proposed by Pett and Lavrov (2013).

It has previously been demonstrated that some mitochondrial Rip proteins contain Tat-like targeting signals immediately preceding the transmembrane domain (Hinsley et al., 2001; Pett and Lavrov, 2013). To test if the Tat-like targeting signal is functional we analyzed the import and assembly of a mutated Rip protein. We mutated the potential Arabidopsis Rip Tat signal of KR to QQ; glutamines were chosen because substitutions with glutamines had previously been shown to completely abolish Tat pathway export in bacteria (Kreutzenbeck et al., 2007). Using in vitro import assays into isolated Arabidopsis mitochondria it was observed that Rip-KR imports and assembles into complex III and the super-complex of I and III very efficiently, as seen by the increase in radioactive signal over time (Fig. 8A). However, in contrast, Rip-QQ displays a far weaker signal in the complexes, indicating that it is not assembled efficiently in Arabidopsis mitochondria (Fig. 8A). Interestingly, we noticed that in the Rip-QQ samples, the signal peaked at 60 mins and subsequently decreased in the further time points, whereas Rip-KR continued to increase (Fig. 8A). This suggests that exchanging the Rip KR Tat signal to QQ disrupts the stable assembly into complex III. In order to test if both the Rip-KR and Rip-QQ proteins are imported and stable within Arabidopsis mitochondria, we repeated the experiment but used SDS-PAGE instead of BN-PAGE. Both proteins were imported and processed to the mature form in a time-dependent manner with similar efficiencies (Fig. 8B). Importantly, we also checked that both Rip-KR and Rip-QQ were inserted into the inner membrane. This was confirmed by extracting the membranes using 0.1 M Na₂CO₃ (Fig. S1). Both Rip-KR and Rip-QQ were located within the membrane fractions (Fig. S1). From these results we conclude that the Rip-QQ was imported and stable within Arabidopsis mitochondria but cannot be efficiently assembled into complex III, indicating that the insertion of the Rip protein into complex III requires a functional Tat pathway.

To further investigate whether the Arabidopsis mitochondrial Rip protein is inserted by a Tat pathway, we analyzed the assembly in the presence of nigericin; an inonophore that dissipates the ΔpH of the membrane potential but does not effect the electrical component of the membrane potential (Yuan and Cline, 1994). Import and assembly of mitochondrial proteins has previously been demonstrated to rely heavily on the electrical component of the membrane potential and not the ΔpH (Martin et al., 1991; Pfanner and Neupert, 1985). Therefore, we first incubated wild-type Rip-KR with mitochondria under normal import conditions for a pulse time of 7.5 min. After the pulse period, the mitochondria were re-isolated and washed to remove unimported protein. The reaction was then split into two, one containing normal import buffer and the other containing 2 µM nigericin in addition to normal import buffer. Assembly into complex III was analyzed by BN-PAGE (Fig. 8C). Only the reactions not containing nigericin showed a clear assembly of Rip-KR into complex III and the super-complex of I and III as indicated by the increase in radioactive signal over time (Fig. 8C). The fact that Arabidopsis Rip requires the ΔpH is another indication that it is inserted into complex III by the Tat pathway. Incidentally, the Tat pathway was originally called the ΔpH pathway as many of its substrates in both bacteria and chloroplasts are inhibited by nigericin in a similar manner as was seen for the Arabidopsis chloroplast Rip protein (Molik et al., 2001).

Since the first plant mitochondrial genome sequences were obtained it has been hypothesized that plant mitochondria contained a Tat pathway. However, only one subunit gene, TatC, was ever identified and was mostly thought of as a pseudogene, particularly in Arabidopsis as the mtTatC gene did not contain a classical start codon. But even with a functional TatC gene, plant mitochondria were still missing the other subunits required to complete a functional Tat pathway.

The identification of a plant mitochondrial TatB-like protein opens up a new whole field of research in plant mitochondrial biology. Interestingly, mtTatB lacks a cleavable transit peptide but is still correctly targeted to the mitochondria. Most likely this would mean that mtTatB is imported by the carrier import pathway. The carrier import pathway is specialized in the import of inner membrane proteins, the majority of which lack cleavable presequences (Sirrenberg et al., 1996). Targeting signals of the carrier import pathway are thought to be mediated by the transmembrane domains of substrate proteins (Endres et al., 1999). Therefore, the mitochondrial targeting signal for mtTatB is most likely its N-terminal transmembrane domain. Several questions, however, still require answering such as: where is mtTatA, does the mitochondria Tat pathway require a TatA subunit, why are mtTatB proteins much longer than other TatB proteins, what are the substrates for the mitochondrial Tat pathway and why have plant mitochondria retained a Tat pathway whereas other eukaryotes have replaced it?

To answer the first question, we could not as yet identify any protein within Arabidopsis that is an obvious target for being a mtTatA protein. This might be related to the second question of why mtTatB is much longer than other TatB proteins. It might be possible that mtTatB has gained some sort of dual functionality and performs both the roles of TatA and TatB. This, though, will require extensive biochemical testing to confirm. So far, in our hands attempts to complement E. coli Tat mutants with the mitochondrial subunits have not worked. Additionally, attempts to overexpress full-length mtTatC in E. coli have also failed, therefore limiting any in vitro reconstitution assays. As for the substrates of the mitochondrial Tat pathway, this study has identified that the Rip protein from complex III as a good candidate. Rip is an interesting candidate as, in the mitochondria of yeast and humans, it undergoes a unique import and assembly process. Firstly, Rip is synthesized on cytosolic ribosomes and imported in a post-translational manner through the outer and inner membrane by the general import pathway, utilizing the translocation complexes TOM and TIM17:23 (Hartl et al., 1986). Secondly, after reaching the matrix, Rip has its targeting signal removed in two steps. Thirdly, Rip has its iron sulfur cluster inserted and the C-terminus is fully folded (Kispal et al., 1997, 1999). This is all thought to happen in the matrix and at this stage Rip is found as a soluble intermediate (Hartl et al., 1986). Fourthly, Rip is chaperoned by the protein Mzm1 to Bcs1 in the inner membrane where Rip is then inserted into the membrane and the fully folded C-terminus is passed back through the inner membrane to the intermembrane space (Cui et al., 2012; Wagener et al., 2011). Fifthly, Rip is the last protein inserted into the so-called pre-Bc1 complex, forming the complete and functional complex (Wagener et al., 2011).

Until recently, it was thought that a similar pathway also existed in plant mitochondria. However, as stated before, the closest related protein to human or yeast Bcs1 in plant mitochondria is located in the outer membrane and plays no role in the assembly of the bc1 complex (Zhang et al., 2014a). As for Rip assembly in plant mitochondria, it is most probable that the first four of the five steps outlined above are exactly the same. Arabidopsis Rip is a nuclear-encoded gene, with its protein product synthesized in the cytosol and post-translationally imported into mitochondria. Similar to yeast, AtRip contains a cleavable pre-sequence and one transmembrane domain, and therefore most likely also uses the TOM and TIM17:23 complexes for translocation through the outer and inner membranes. The plant mitochondrial matrix also contains a homolog of Mzm1 and all the required proteins for iron sulfur biogenesis (Balk and Pilon, 2011). Thus, plant mitochondrial Rip must probably also have a soluble intermediate in the matrix while its iron sulfur cluster is assembled. Therefore, the only divergent part in Rip assembly in plant mitochondria is the use of a Tat pathway for membrane insertion and passing the fully folded C-terminus back into the intermembrane space. This is fully consistent with our observations that mutation of the AtRip Tat targeting signal and also that AtRip requires a ΔpH for assembly. Rip proteins of chloroplasts and bacteria have also been demonstrated previously to require a Tat pathway for proper insertion (Molik et al., 2001; Aldridge et al., 2008; Bachmann et al., 2006; De Buck et al., 2007). These observations are strengthened by the correlation of mitochondrial Tat proteins with the absence of a complete Bcs1 protein in a variety of organisms.

It could be argued that AtRip could also be inserted back into the inner membrane by way of the Oxa pathway; however, this seems unlikely as the Oxa pathway has never been demonstrated to translocate fully folded proteins. This argument can also be countered by the fact that yeast and humans do not use Oxa for Rip assembly, which would negate the need for Bcs1. It might also be argued that AtRip could be laterally inserted from the TIM17:23 complex into the inner membrane. In this case, AtRip would not have a soluble intermediate and the C-terminus would never reach the matrix. However, all the iron sulfur cluster biogenesis machinery and the Mzm1 chaperone specific for Rip are located within the mitochondrial matrix. Therefore, if the C-terminal domain of AtRip never gets to the matrix there is no known mechanism to insert its iron sulfur cluster. Thus, after excluding other possibilities and looking at the data presented in this paper it seems most likely that AtRip uses a Tat pathway for assembly into complex III.

The data presented here demonstrate that plant mitochondria contain the translocation subunits TatB and TatC. As the TatB subunit is an essential gene it can be extrapolated that this potential Tat pathway is required for plant mitochondrial biogenesis. However, further work is required to determine if this potential mitochondrial Tat pathway requires a TatA subunit or somehow functions with only TatB and TatC subunits. We have also sought to demonstrate that the AtRip protein is a substrate of this potential mitochondrial Tat pathway. AtRip requires the ΔpH and a Tat-like targeting signal for proper assembly into complex III. Functional analysis of viable mtTatB mutants (e.g. RNAi or antisense knockdowns) will prove invaluable in identifying more possible substrates and also the functional role of the potential plant mitochondrial Tat pathway.

As to why plant mitochondria have retained a Tat pathway in contrast to other eukaryotes is a difficult question to answer. The most obvious explanation is that Bcs1 evolved later in eukaryotes, after the split of plants from other eukaryotes. There is some evidence to support this with the base animal Monosiga brevicollis containing both mitochondrial TatC and TatB-like proteins and also lacking a complete Bcs1 protein, indicating that the mitochondrial Tat pathway was lost rather late in opisthokont evolution. It is also interesting to note that a selection of related green algae also seem to lack both a complete Bcs1 and mitochondrial Tat subunits (e.g. Chlamydomonas reinhardtii). How these organisms assemble their mitochondrial Rip is another interesting question for future investigation.

The Agrobacterium tumefaciens strain AGL1 was transformed with the full coding sequence of AtmtTatB (At5g43680) fused to GFP in the vector pK7FWG2 (Karimi et al., 2002) and used to infiltrate 4–6 week old Nicotiana benthamiana leaves as described previously (Schweiger et al., 2012). Protoplasts were prepared as outlined in Koop et al. (1996), except cell walls were digested for 90 min at 40 rpm in 1% cellulase R10 and 0.3% macerase R10 after vacuum infiltration. Mitotracker (Life Technologies) was added to the protoplast suspensions to a final concentration of 500 nM. Fluorescence was observed with a confocal laser scanning microscope at 20°C (Leica TCS SP5).

SALK_003481 and CSHL_GT11254 T-DNA insertion lines were obtained from the Nottingham Arabidopsis Stock Centre (NASC) and Cold Spring Harbor Laboratory (CSHL), respectively (Alonso et al., 2003; Sundaresan et al., 1995). T-DNA insertions were genotyped by PCR and insertion sites were confirmed by sequencing.

Mitochondria for in vitro import and assays and western blot analysis were harvested from 14-day-old Arabidopsis thaliana seedlings grown in liquid culture as previously described (Lister et al., 2007). Typically, between 2 and 4 mg of mitochondrial protein was obtained from each preparation.

For in vitro import studies, the full coding sequence of AtmtTatB (AT5G43680) was cloned into the destination vector, pDest14 (Invitrogen). The clones for AOX (X68702), Tim23 (At1g72750) and Rip (At5g13430) have been described previously (Murcha et al., 2003; Whelan et al., 1995; Carrie et al., 2015). [³⁵S]Met-labeled precursor proteins were synthesized using Flexi Rabbit Reticulocyte Lysate (Promega) as previously outlined (Chang et al., 2014). In vitro mitochondrial imports were then performed using isolated mitochondria as previously described (Whelan et al., 1995; Lister et al., 2007). All in vitro imports were obtained using radiography and images were scanned using a Typhoon scanner (GE Healthcare).

Outer-membrane-ruptured mitochondria were prepared by osmotic shock as previously described (Murcha et al., 2005). Carbonate extraction of mitochondrial membrane proteins was carried out as previously described (Tan et al., 2012).

BN-PAGE was performed as in Eubel and Millar (2009) using 5% digitonin (Serva). Immunodetection of proteins was performed as described previously (Murcha et al., 2005). Unless specified, the equivalent of 50 µg of protein was used in each lane. For the production of the antibody against AtmtTatC, the peptide VREEGWTSGMRESGIEKKNKSSPPPRTW, which corresponds to amino acids 253 to 281 of the Arabidopsis TatC, was produced by Ganaxxon bioscience (Germany) and injected into two New Zealand white rabbits as per standard protocols in accordance with all relevant institutional and national animal welfare regulatory standards (Cooper and Paterson, 2008). For the antibody against AtmtTatB the full protein sequence of AtmtTatB was expressed in E. coli fused to a 6×His tag. The protein was purified using denatured IMAC followed by electro-elution. The final protein was concentrated to ∼2 mg/ml and then sent to Pineda Antikörper-service (Germany) for antibody production. Both antibodies were affinity purified using their respective antigens. The antibodies against COXII (A S04 052) and GDC-H (A S05 074) were purchased from Agrisera (Sweden). Antibodies against Tom40 and Qcr7 have been previously published (Kuhn et al., 2011; Carrie et al., 2009).

For the phylogenetic tree in Fig. 1 the protein sequences of Cytochrome c oxidase I (COXI) and Cytochrome b (COB) were concatenated prior to analysis. Protein sequences were obtained from the NCBI organelle genomes database. The protein sequences were first aligned using MEGA5 (Tamura et al., 2011) with the Muscle algorithm (Edgar, 2004). The evolutionary history was then inferred using Maximum Likelihood method based on the Whelan and Goldman model (Whelan and Goldman, 2001). The bootstrap consensus tree was inferred from 1000 replicates with the percentage of replicate trees in which the associated proteins clustered together in the bootstrap test shown next to branches. For the phylogenetic tree of the TatA and TatB sequences in Fig. 2 the process was exactly the same. For the phylogenetic tree in Fig. 6 the procedure and use of COX1 and COB sequences was exactly the same. For species and sequence information see Table S2 for all phylogenetic trees.

The Tom20 antibody was a kind gift from Dr Serena Schwenkert, Ludwig-Maximilians-Universität München, Planegg-Martinsried, Germany. The pea cpTatA and cpTatB antibodies were a kind gift from Prof. Kenneth C. Cline, University of Florida, Gainesville, USA.