Arabidopsis mTERF15 Is Required for Mitochondrial nad2 Intron 3 Splicing and Functional Complex I Activity

Mitochondria play a pivotal role in most eukaryotic cells, as they are responsible for the generation of energy and diverse metabolic intermediates for many cellular events. During endosymbiosis, approximately 99% of the genes encoded by the mitochondrial genome were transferred into the host nucleus, and mitochondria import more than 1000 nuclear-encoded proteins from the cytosol to maintain structural integrity and fundamental functions, including DNA replication, mRNA transcription and RNA metabolism of dozens of mitochondrial genes. In metazoans, a family of nuclear-encoded proteins called the mitochondrial transcription termination factors (mTERFs) regulates mitochondrial transcription, including transcriptional termination and initiation, via their DNA-binding activities, and the dysfunction of individual mTERF members causes severe developmental defects. Arabidopsis thaliana and Oryza sativa contain 35 and 48 mTERFs, respectively, but the biological functions of only a few of these proteins have been explored. Here, we investigated the biological role and molecular mechanism of Arabidopsis mTERF15 in plant organelle metabolism using molecular genetics, cytological and biochemical approaches. The null homozygous T-DNA mutant of mTERF15, mterf15, was found to result in substantial retardation of both vegetative and reproductive development, which was fully complemented by the wild-type genomic sequence. Surprisingly, mitochondria-localized mTERF15 lacks obvious DNA-binding activity but processes mitochondrial nad2 intron 3 splicing through its RNA-binding ability. Impairment of this splicing event not only disrupted mitochondrial structure but also abolished the activity of mitochondrial respiratory chain complex I. These effects are in agreement with the severe phenotype of the mterf15 homozygous mutant. Our study suggests that Arabidopsis mTERF15 functions as a splicing factor for nad2 intron 3 splicing in mitochondria, which is essential for normal plant growth and development.


Introduction
Mitochondria, which originated through the endosymbiosis of a-proteobacteria into ancestral host cells, are the cellular powerhouses and play vital roles in diverse eukaryotic cell processes through the production of ATP and various metabolic intermediates [1,2]. Recent studies also suggest that dysfunctional mitochondria are involved in many neurodegenerative diseases such as aging and cognitive decline in a wide range of metazoans, including humans [3]. Maintaining the structural and metabolic integrity of this semi-autonomous organelle is essential for the normal function of eukaryotic cells. Nevertheless, over the course of symbiotic evolution, the majority of mitochondrial genes migrated into the nuclear genome of the original host, leaving an incomplete set of essential genes in the mitochondrial genomes of most organisms, including plants [4][5][6]. Complicated and dynamic communication and coordination between the nucleus and mitochondria greatly impact many fundamental cellular processes in, and even the lives of, most eukaryotes [7]. Indeed, based on the complete sequence of the Arabidopsis mitochondrial genome, it has been reported that 57 mitochondrial genes encode the subunits of multiprotein complexes that are required for the respiratory chain, heme and cytochrome assembly, and mitochondrial ribosomes [4]. Additionally, plant mitochondria are more complex than those found in other kingdoms and exhibit unique RNA metabolism, including RNA transcription, splicing, editing, degradation, and translation [8][9][10]. The proteins involved in these processes are predominantly encoded by the nuclear genome and are imported into mitochondria after protein synthesis. For example, recent studies suggested that one protein family, called the mitochondrial transcription termination factors (mTERFs), plays important roles in regulating the organellar transcription machinery.
The mTERF proteins were first identified two decades ago as regulators of transcription termination in human mitochondria [11]. Phylogenetic analyses of mTERF homologs in metazoans and plants revealed the presence of 4 subfamilies, mTERF1 to mTERF4 [12]. These proteins share a common 30-amino-acid repeat module called the mTERF motif [13]. The proteins within this family possess diverse numbers and arrangements of these motifs, yet the folding patterns of these proteins are similar. Moreover, crystal structure studies of mTERF1, mTERF3 and mTERF4 suggest that the helical structure of the mTERF motifs may be essential for their nucleic acid-binding activities [14,15]. Human mTERF1 binds specific sites located at the 39-end of the 16S rRNA and tRNA Ler(UUR) genes to terminate mitochondrial transcription [16]. Additionally, mTERF1 binds to the mitochondrial transcription initiation site to create a DNA loop that allows for the recycling of the transcriptional machinery [16]. This simultaneous link between mitochondrial transcriptional initiation and termination sites may explain the high rate of mitochondrial rRNA biogenesis. mTERF2 binds to mitochondrial DNA (mtDNA) in a non-specific manner and associates with nucleoids. mTERF2 loss-of-function mice exhibit myopathies, memory deficits, and impaired respiratory function due to a reduction in the number of mitochondrial transcripts [17,18]. mTERF3 is essential, and mTERF3-knockout mice die during the early stages of embryogenesis; mTERF3 non-specifically interacts with mtDNA promoter regions and mediates the repression of mitochondrial transcription initiation [19]. Mouse NSUN4 (NOL1/NOP2/Sun domain family, member 4) is a methyltransferase involved in the methylation of ribosomes. Mouse mTERF4 first binds 16S and 12S rRNAs, then forms a stoichiometric complex with NSUN4, which is essential for proper mitochondrial ribosomal assembly and translation [14,20,21].
The genome of Arabidopsis contains 35 identified mTERFs, which vary in the number and arrangement of their mTERF motifs [22]. Several recently characterized Arabidopsis mTERF proteins displayed diverse and complicated functions during mitochondrial and plastid transcription, such as SINGLET OXYGEN-LINKED DEATH ACTIVATOR10 (SOLDAT10) [23], BELAYA SMERT/RUGOSA2 (BSM/RUG2) [22,24], MTERF DEFECTIVE IN ARABIDOPSIS 1 (MDA1) [25] and SUPPRESSOR of hot1-4 1 (SHOT1) [26]. SOLDAT10, the first Arabidopsis mTERF to be characterized, is targeted to plastids and regulates plastid homeostasis. Arabidopsis soldat10 mutants suppressed the 1 O 2 -mediated cell-death and growth-retardation phenotypes of the conditional fluorescent (flu) mutant, in which excess levels of the reactive oxygen species (ROS) 1 O 2 were generated within chloroplasts after the dark-to-light shift [23]. BSM/RUG2 is essential for embryogenesis and postembryonic development in Arabidopsis and displays DNA-binding activity towards plastid DNA. Additionally, BSM/RUG2 localizes to both plastids and mitochondria, and the perturbation of BSM/RUG2 affects the processing and steady-state levels of plastid transcripts [22] and decreases the expression of mitochondrial transcripts [24]. Arabidopsis MDA1 is a plastid-localized protein involved in abiotic stress tolerance. MDA1 mutation leads to significant defects in plant growth and chloroplast development but enhanced tolerance to salt and osmotic stress, perhaps via reduced sensitivity to abscisic acid [25]. Mutating SHOT1 suppresses the heathypersensitive phenotype caused by the loss of heat shock protein 101 in Arabidopsis [26]. However, SHOT1 is a mitochondrialresident mTERF protein, and the thermotolerance phenotype of shot1 may be caused by reduced ROS-mediated oxidative damage during heat stress. Additionally, SHOT1 plays an important role in regulating the expression of mitochondrial-encoded genes and mitochondrial genome copy number. Although it is known that many mTERF mutations cause defective development and stress responses, studies of the precise molecular functions of these mTERF proteins are limited in Arabidopsis organelles.
In Arabidopsis, several protein families that participate in organellar RNA splicing, including those containing the chloroplast RNA splicing and ribosome maturation (CRM) domain [27][28][29], the pentatricopeptide repeat (PPR) [30][31][32][33][34][35], the plant organellar RNA recognition (PORR) domain [36,37] and the ACCUMULATION OF PHOTOSYSTEM ONE (APO) motif [38]. These domains are recognized as RNA binding domains and interact with specific elements within introns. Although several splicing factors have been identified in plastids and mitochondria, the detailed mechanism of RNA splicing in plant organelles is still being investigated. Recently, maize mTERF4 was identified as a splicing factor associated with several introns and splicing factors in chloroplasts [39].
In the present study, we investigated another Arabidopsis mTERF protein, mTERF15, and identified its role in post-transcriptional modification of mitochondrial RNA processing. Mutating mTERF15 significantly disturbed normal plant vegetative and reproductive growth, which may be the result of defective mitochondrial development and/or function. Interestingly, mTERF15 is a mitochondria-localized protein that binds to RNA but not to double-stranded DNA (dsDNA). Moreover, an analysis of mitochondrial intron splicing in mterf15 mutants revealed defective RNA splicing of nad2 intron 3, which led to a deficiency in complex I activity. The possible molecular function of mTERF15 in regulating Arabidopsis mitochondrial RNA metabolism is discussed.

Plant material and growth conditions
The mTERF15 T-DNA insertion mutant (SALK_134099) was obtained from the Arabidopsis Biological Resource Center (Ohio State University, USA). Seeds from Arabidopsis thaliana ecotype Columbia-0 and mterf15/+ plants were surface-sterilized by chloride gas after stratification for 3 d at 4uC and then either germinated on half-strength Murashige and Skoog medium (Duchefa Biocheme, The Netherlands) containing 1% sucrose and 0.7% phytoagar (Duchefa) or grown in soil inside a growth chamber on a 16-h light/8-h dark cycle at 22uC.

Complementation of the mterf15 mutant
Approximately 3.3 kb of the genomic sequence corresponding to locus At1g74120, including a 2-kb putative promoter and the 1.3-kb coding region, was amplified using Phusion polymerase (Finnzymes, http://www.finnzymes.com) and then ligated into a modified pPZP221 vector containing a C-terminal GFP sequence followed by the Nopaline synthase (NOS) terminator. The primers used for cloning are in Table S1. After verification of the DNA sequence, the construct was introduced into Agrobacterium tumefaciens strain GV3101. Arabidopsis plants heterozygous for mTERF15 were transfected with Agrobacterium by the vacuum infiltration method [40]. Seeds corresponding to the T 1 generation were collected and grown on solid half-strength Murashige and Skoog medium (Duchefa) containing 1% sucrose, 0.7% phytoagar (Duchefa) and the antibiotic G418 at 100 mg/ml. Leaves from T 1 plants were harvested for genotyping. Plants with an mterf15 homozygous background harboring mTEFR15p::mTERF15-GFP were collected for further studies.
Ultrastructural sample preparation and tetramethylrhodamine methyl ester perchlorate (TMRM) staining For transmission electron microscopy, samples were frozen in a high-pressure freezer (Leica EMPACT2, 2000-2050 bar). Freeze substitution involved the use of anhydrous acetone (containing 1% OsO 4 and 0.1% UA) with a Leica EM AFS2. The samples were kept at 285uC for 3 d, at 260uC for 1 d, at 220uC for 1 d, and at 0uC for 1 d and then brought to room temperature. The samples were infiltrated and embedded in Spurr resin. Ultrathin sections (70-90 nm) were cut using a Reichert Ultracut S or Leica EM UC6 (Leica, Vienna, Austria) and collected on 100 mesh copper grids. After staining with 5% uranyl acetate in 50% methanol for 10 min and 0.4% lead citrate for 4 min, sections were observed under a Philips CM 100 Transmission Electron Microscope at 80 KV; images were captured using a Gatan Orius CCD camera.
TMRM staining (Molecular Probes, Invitrogen) was performed to examine the mitochondrial membrane potential in protoplasts. Isolated protoplasts were stained with 20 nM TMRM and immediately observed under a laser scanning confocal microscope (Zeiss, LSM510, Carl Zeiss, http://www.zeiss.com). TMRM fluorescence and chlorophyll autofluorescence were induced by excitation with a 543-nm HeNe laser and a 488-nm Argon laser. The emission signals from TMRM and from chlorophyll were recorded with using 565-to 615-and 650-nm band-pass filters, respectively. The fluorescence intensity and area were quantified using LSM510 software (Carl Zeiss, http://www.zeiss.com).

Northwestern blot analysis
Northwestern blot analysis was performed to test the RNAbinding ability of mTERF15 as described previously [41]. Briefly, mTERF15 cDNA lacking the 34 N-terminal amino acids was ligated into the pET52b vector to produce the mTERF15-strep recombinant protein. The recombinant proteins were expressed in E. coli strain BL21 and incubated with StrepTactin-Sepharose resins (GE Healthcare) according to the manufacturer's instructions. The resins were washed extensively with incubation buffer before elution of the recombinant proteins. The mTERF15-strep recombinant protein and the MBP-strep control protein were transferred from 12% SDS-polyacrylamide gels onto a PVDF membrane and then renatured overnight in renaturation buffer (0.1 M Tris-HCl, pH 7.5 and 0.1% (v/v) NP-40) at 4uC. After 4 washes with renaturation buffer (15 min), blots were blocked with blocking buffer [(10 mM Tris-HCl, pH 7.5, 5 mM Mg acetate, 2 mM DTT, 5% (w/v) BSA and 0.01% (v/v) Triton X-100)] at room temperature for 5 min. Full length of nad2 intron 3 was in vitro synthesized using SP6 polymerase (Promega). Total RNA and in vitro synthesized nad2 intron3 were labeled with c 32 P ATP using T4 polynucleotide kinase (Fermentas). Hybridization was performed at 4uC in the presence of 5 mg labeled RNA. Blots were washed 4 times (5 min each) in a washing buffer (10 mM Tris-HCl, pH 7.5, 5 mM Mg acetate, 2 mM DTT) to remove unbound RNA. The radioisotope signals were detected via autoradiography on X-ray film (Kodak).

RNA isolation and qRT-PCR analysis
Total RNA was isolated using the Qiagen RNeasy Mini Kit (QIAGEN) and then treated with TURBO DNA-free DNase (Ambion) to remove contaminating genomic DNA. Reverse transcription of RNA was performed using the M-MLV transcriptase system (Invitrogen) and random hexamer primers.
Gene expression analysis via qRT-PCR was carried out using Power SYBR Green Supermix (Applied Biosystems) running on an ABI Prism 7000 sequence detection system (Applied Biosystems). Primer sets were designed as described previously [31] (Table S2) to monitor the expression of genes involved in the splicing of junctions within mitochondria.

Northern blot analysis
A total amount of 5 mg RNA was denatured at 65uC for 10 min and then separated on 2% formaldehyde-agarose gels in MOPS buffer and transferred to a nylon membrane (PerkinElmer). The nad2 probes were labeled using the Rediprime II DNA labeling system (GE Healthcare). Probe primers are listed in Table S1. The prehybridization, hybridization and washing steps were performed following the manufacturer's instructions.

Crude mitochondria isolation and complex I Activity assay
Crude mitochondria were prepared as described previously [36,42]. Approximately 200 mg fresh tissue was extracted in 2 ml extraction buffer (75 mM MOPS-KOH, pH 7.6, 0.6 M sucrose, 4 mM EDTA, 0.2% polyvinylpyrrolidone 40, 8 mM cysteine, 0.2% bovine serum albumin) at 4uC. The lysate was centrifuged first at 13006g for 5 min (twice), after which the supernatant was centrifuged at 22,0006g for 20 min at 4uC. The resulting pellet (which usually contains most of the thylakoid and mitochondrial membranes) was resuspended in buffer (10 mM MOPS-KOH, pH 7.2, 0.3 M sucrose) and stored at 280uC.
Clear native electrophoresis was performed as described previously [43]. Crude mitochondria were washed with distilled water, resuspended with buffer (50 mM NaCl, 50 mM imidazole/ HCl, 2 mM 6-aminohexanoic acid, 1 mM EDTA, pH 7.0) and solubilized by adding DDM (10%) at a DDM-to-protein ratio of 2.5 (g/g). After centrifugation at 100,0006g for 15 min, 10% glycerol and 0.02% Ponceau S were added to the supernatant for 4-16% native PAGE with anode buffer (25 mM imidazole/HCl, pH 7.0) and cathode buffer (50 mM Tricine, 7.5 mM imidazole, pH 7.0, 0.01% DDM and 0.05% DOC). Clear native gels were washed 3 times with distilled water for 5 min and incubated with 1 mM nitroblue tetrazolium and 0.14 mM NADH in 0.1 M Tris, pH 7.4. The reaction was stopped with 40% methanol and 10% acetic acid after a dark blue signal appeared on the gel.

Disruption of the Arabidopsis mTERF15 gene results in significant growth and developmental retardation
A screen of Arabidopsis T-DNA insertional mutants from the SALK collection allowed us to identify several mutants that were defective in post-embryonic development and/or seed germination. We found one T-DNA mutant showing abnormal phenotype in progenies segregated from SALK_134099 heterozygotes. The progeny of SALK_134099 heterozygotes showed growth retardation at a ratio of 3:1 (normal to defective) (x 2 , p = 0.924; Table 1). Most abnormal progenies derived from SALK_134099 heterozygous plants were unable to germinate on solid medium (17.02%), although a few grew slowly (8.21%, Table 1). The SALK_134099 homozygous mutant containing a T-DNA insertion in the coding region of Arabidopsis mTERF15 (At1g74120; Figure 1A) and showing significant growth and development retardation was named mterf15 ( Figure 1B). The full-length mTERF15 transcript was not detected in the homozygous mterf15 plant ( Figure 1C), suggesting that all of the observed phenotypes resulted from the null mutation of the mTERF15 allele (SALK_134099). Developmental defects were widespread in homozygous mterf15 plants. Compared with wild-type seedlings, mterf15 seedlings were small in stature, had small cotyledons and displayed reduced growth of primary roots ( Figure 1D). The rosette leaves of mterf15 were smaller, wrinkled and twisted whereas the wild type plant showed larger and smooth round-shaped leaves. Moreover, the extended petiole region of mterf15 was shorter than the wild type plant. Compared with 1.5-month-old wild-type plants that were ready for blossoming, 3-month-old adult mterf15 plants were dwarfed and just about ready for flowering ( Figure 1E). The flowers of mterf15 plants were abnormal, with misplaced petals, extremely defective anthers and a limited amount of pollen ( Figure 1F, upper panel). Unlike wild-type siliques, mature mterf15 siliques were much shorter ( Figure 1G), and no viable seeds were found ( Figure 1F, lower panel).
To further verify that these defects were indeed caused by the disruption of the mTERF15 allele, mterf15 heterozygotes were complemented with the mTERF15 native promoter-driven mTERF15 coding sequence tagged with GFP (mTERF15p::m-TERF15-GFP). The T 2 progeny of homozygotes for the mTERF15 mutation which harbors at least one copy of the transgene mTERF15p::mTERF15-GFP showed similar phenotypes as 7-d-old wild-type seedlings and adult plants (Figure 1 and Figure S1). All of the above results show that functional mTERF15 is essential for normal plant growth and development.
mTERF15, a plant-specific and mitochondria-localized protein, is important for mitochondrial biogenesis and membrane potential The Arabidopsis mTERF15 locus contains a single exon and appears to encode a 51-kDa protein. Architecture analysis with SMART (http://smart.embl-heidelberg.de) revealed that mTERF15 contains 5 typical mTERF motifs ( Figure S2A). A search of the NCBI database (http://www.ncbi.nlm.nih.gov) revealed mTERF15 homologs present in Vitis vinifera, Populus trichocarpa, Brachypodium distachyon, Sorghum bicolor and Zea mays. Comparing protein similarities and identities between flowering plant mTERF15 orthologs and metazoan mTERF1, Arabidopsis mTERF15 is evolutionally closer to plant mTERF15 orthologs than metazoan mTERF1. This result suggests that mTERF15 is a plant-specific mTERF protein (Table S3). The phylogenetic tree branched into 2 groups, monocots and dicots ( Figure S2B). Amino acid sequence alignments of all 6 homologs revealed stronger conservation of the C terminus compared with the N terminus. All of these homologs appeared to contain 5 mTERF motifs with a conserved proline residue at position 8, except for the first motif ( Figure S2A).
In silico analysis of the BIO-array (http://bar.utoronto.ca/ efp_arabidopsis/cgi-bin/efpWeb.cgi) and Genevestigator (https:// www.genevestigator.com/gv/plant.jsp) microarray databases suggested various expression levels of mTERF15 in different tissues, with the highest and lowest expression levels in mature seeds and pollen, respectively. However, in the Genevestigator microarray database, mTERF15 showed the highest expression in sperm cells and the lowest in leaf phloem protoplasts. Quantitative real-time PCR spatiotemporal analysis in Arabidopsis revealed mTERF15 expression in all organs, with high levels of accumulation in seedlings, flowers and siliques ( Figure S3).
Transient protoplast assays and stable transgenic studies suggested that mTERF15 is a mitochondrial protein [22]. To confirm the subcellular localization of mTERF15, we observed the root hairs of 7-d-old mTERF15p::mTERF15-GFP transgenic seedlings via confocal microscopy after staining with the mitochondrial indicator MitoTracker orange. The co-localization of the mTERF15-GFP and MitoTracker signals demonstrated that mTERF15 is a mitochondrial protein ( Figure S4).
We further investigated the morphology of chloroplasts and mitochondria from the leaves of 1-month-old wild-type and 2month-old mterf15 plants by transmission electron microscopy. The sizes and features of mterf15 chloroplasts were similar to those of wild-type chloroplasts (Figure 2A, D). Although the sizes of the mitochondria did not differ between mterf15 and wild-type plants, the inner membrane systems were aberrant in mterf15 mitochondria; wild-type mitochondria ( Figure 2B-2C) showed normal cristae structures with clear inner spaces, whereas mterf15 mitochondria ( Figure 2E-2F) lacked any obvious cristae structure. Therefore, the integrity of the mitochondrial inner membrane system may have been compromised, potentially resulting in decreased mitochondrial membrane potential.
To assess mitochondrial membrane integrity, we treated protoplasts from leaves of the wild-type, complementation line and mterf15 plants with tetramethylrhodamine methyl ester (TMRM), a lipophilic, cationic, red-orange fluorescent dye that accumulates within mitochondria in living cells. The intensity of fluorescence is associated with the mitochondrial membrane potential [44,45]. We observed large and strong fluorescent TMRM signals in protoplasts from the wild-type and complementation lines, which suggests that mitochondria were active in these lines ( Figure 2G, I); however, TMRM signals were slightly weaker in mterf15 protoplasts ( Figure 2H). The mean fluorescence intensity for mterf15 protoplasts was reduced by ,80% compared with wild-type protoplasts ( Figure 2J). These results suggest that mTERF15 is crucial for maintaining mitochondrial biogenesis and membrane integrity. mTERF15 is an RNA-binding protein that is involved in mitochondrial nad2 intron 3 splicing and normal complex I activity Because most eukaryotic mTERFs function as nucleic acidbinding proteins, we performed in vitro binding studies with dsDNA-cellulose and northwestern blot analyses to determine the dsDNA-and RNA-binding abilities of mTERF15. After incubation with commercial dsDNA-cellulose resin, most of the mTERF15 was present in the unbound fraction ( Figure S5A), which implies a lack of dsDNA-binding activity for mTERF15. Additionally, southwestern blot analysis was employed to validate the DNA-binding ability of mTERF15. As shown in Figure S5B, there is no DNA-binding signal for the mTERF15 protein in southwestern blots. The above results imply that mTERF15 may not be a DNA-binding protein. Therefore, we investigated the RNA-binding ability of mTERF15 by northwestern blot analysis (Figure 3). Strep-tagged mTERF15 lacking the first N-terminal 34 amino acids, which contain a putative targeting signal (MW = 48 kDa), was separated by SDS-PAGE, transferred to a PVDF membrane and probed with a specific antibody against strep-tagged or radiolabeled total RNA prepared from 7-d-old Arabidopsis seedlings. The presence of the mTERF15-strep and MBP-strep fusion protein were confirmed by immunoblot analysis using an antibody against the strep-tag, and 48-kDa bands corresponding to mTERF15-strep were detected in a dose-dependent manner in the RNA-binding assay but not in the control. These results suggest that Arabidopsis mTERF15 functions as an RNA-binding protein.
Previously, Babiychuk et al. (2011) showed that one mTERF protein, BSM, is required for embryogenesis and plays a role in chloroplast intron splicing. Therefore, we examined 23 intronsplicing events in mitochondria by quantitative RT-PCR (qRT-PCR). In mitochondria, 9 genes require intron splicing for the complete maturation of their respective transcripts, and 4 of these contain a single intron: rps3, cox2, ccmFc and rpl2. The others (nad1, nad2, nad4, nad5 and nad7) contain multiple introns and are required for the normal function of mitochondrial complex I. Using the specific primer sets designed by Longevialle et al. (2007), we examined individual splicing events in mitochondria obtained from wild-type, mterf15 and complemented plants. All splicing events showed similar results in both the wild-type and complemented plants; however, the RNA splicing of nad2 intron 3 (nad2-3) was significantly reduced in mterf15 plants ( Figure 4A).
In Arabidopsis, the mature nad2 mRNA is formed by merging exons from 2 different transcripts, nad2a and nad2b, through one trans-splicing and 3 cis-splicing events. The genomic fragment of nad2a carries exons 1 and 2 of nad2 separated by nad2 intron 1. The other 3 exons of nad2, exons 3 through 5, are from nad2b transcripts ( Figure 4B). To validate the results of the qRT-PCR analysis, we used RT-PCR to examine the accumulation of unspliced nad2b transcripts in mterf15 plants; the amplification of nad2 exon 5 by primers 3F and 3R was used for normalization (lower panel in Figure 4B). RT-PCR with primers 1F and 2R revealed no mature transcripts from the splicing of nad2 intron 3 in mterf15 plants ( Figure 4B). Additionally, the transcript levels were greater in mterf15 plants than in wild-type plants upon amplification of the exon-intron junction with the primer sets 1F&1R and 2F&2R. This nad2 intron 3 splicing defect was further confirmed by northern blot analysis. The nad2b transcripts revealed by probes for nad2 exon 3 and 5 showed no mature nad2 mRNAs in mterf15 plants (m in Figure 4C) and was accompanied by the overaccumulation of un-spliced transcripts (u in Figure 4C). A similar accumulation of un-spliced transcripts in mterf15 was confirmed using probes specific to nad2 intron 3 (u in Figure 4C). This splicing defect was fully rescued in complement-ed plants. Another cis-splicing transcript, nad1 exon 3, was used as a positive control, with comparable results to those observed in qRT-PCR: the splicing was normal, but the accumulation of nad1-3 was greater in mterf15 plants than in wild-type and complemented plants ( Figure 4C).
Because mTERF15 functions as an RNA-binding protein and is involved in nad2 intron 3 splicing, mTERF15 may directly interact with nad2 intron 3. First, we confirmed the interaction between mTERF15 and full-length nad2 intron 3 (,2.6 kb) by northwestern blot analysis (Figure 3, lowest panel). The interaction between mTERF15 and nad2 intron 3 was stronger than the interaction with total RNA. We first examined the interaction between mTERF15 and full-length nad2 intron 3 (,2.6 kb) by northwestern blot analysis (Figure 3, lowest panel). The interaction between mTERF15 and nad2 intron 3 was stronger than the interaction with total RNA. Then RNA-EMSA (RNA-Electro- phoresis Mobility Shift Assay) was used to confirm the RNAbinding activity of mTERF15 ( Figure S6). Additionally, two unrelated mitochondrial intron sequences, ccmF C and rpl2, were used as the control probes in RNA-EMSA. Since the full length of nad2 intron 3 is too large to be used as the probe for RNA-EMSA, we divided it into five fragments (approximately 500 bp each), named Fragment 1-5 ( Figure S6A). Fragment 1, part of domain 1, is the largest and substantially varies in sequence and secondary structure among characterized group II introns. Fragment 5 contains phylogenetically conserved and typical RNA structure found in all mitochondrial introns [51]. As shown in Figure S6B, only Fragment 1 could be recognized by mTERF15 in a dosage- dependent manner. Neither Fragments 2-5 nor ccmFc and rpl2 could be recognized by mTERF15. Above results suggest that mTERF15 is an RNA-binding protein and is involved in the proper splicing of nad2 intron 3.
In Arabidopsis, the nad2 gene encodes a subunit of mitochondrial complex I (an NADH:ubiquinone oxidoreductase), which is required to translocate protons from the mitochondrial matrix into the intermembrane space. Complex I activity was therefore investigated in mterf15, in which the splicing of nad2 intron 3 is defective. Isolated crude mitochondria were subjected to complex separation on clear native gels to visualize protein complexes by silver staining (Figure 5A) or to complex I activity assays ( Figure 5B). The amount and activity of complex I were greatly reduced in mterf15, whereas the wild-type and complementation lines showed normal activity. Therefore, intron-splicing defects in nad2 intron 3 led to reduced complex I formation and activity. Disturbing mitochondrial complex I impairs energy generation and leads to the extreme retardation of diverse aspects of growth and development in mterf15 plants (Figure 1).
The steady-state levels of several mitochondrial proteins, including Nad9, Cox2, Cox3, AtpA, alternative oxidase (AOX) and the channel-forming protein Porin, were examined by immunoblot analysis of crude mitochondrial extracts from wildtype and mterf15 plants ( Figure 6). The steady-state levels of 3 proteins, Cox2, Cox3 and AtpA, were greater in mterf15 plants than in wild-type plants. The level of alternative oxidase (AOX) was significantly increased in mterf15 plants, which is consistent with a defect in NADH oxidation [46]. The level of another complex I subunit, Nad9, was decreased in mterf15 plants, which may be due to an imbalance in the stoichiometry of multi-subunit complexes [47].

Discussion
Our genetic, molecular and biochemical results indicate that mTERF15 is a unique, mitochondria-localized protein ( Figures S2  and S4) with RNA-binding activity (Fig. 3). This protein is critical for post-transcriptional modification in the RNA splicing of mitochondrial nad2 intron 3 ( Figure 4). Mutating mTERF15 impaired the normal activity of mitochondrial respiratory chain complex I (Figure 5), thereby resulting in abnormal mitochondrial development ( Figure 2) and the widespread retardation of plant growth and development (Figure 1). mTERFs constitute a broad family of eukaryotic proteins that are essential for the initiation and termination of organellar transcription and the translation and replication of the organellar transcription machinery [12]; however, mounting evidence suggests an emerging role for these proteins in RNA splicing in plants [39]. The human genome contains only 4 genes encoding mTERF paralogs, whereas the Arabidopsis genome contains at least 35 genes for mTERF proteins with a diverse arrangement and number of mTERF motifs. Using transient assays and stable transgenic plants expressing GFP-tagged mTERF proteins, Babiychuk et al. (2011) identified 11 and 17 mTERF proteins that localize to chloroplasts and mitochondria, respectively. However, only a few mTERFs have been well studied and are essential for vegetative growth [24][25][26] and embryogenesis [22,23]. Whether Arabidopsis mTERF proteins share similar conserved molecular functions with their mammalian counterparts or have evolved additional regulatory mechanisms is unclear. In contrast to animal cells, plant cells harbor 2 types of nucleoidcontaining organelles -chloroplasts and mitochondria. Moreover, plant mitochondrial genomes are much larger and more complex, requiring intron splicing for proper gene expression [48]. The biological functions of mTERF proteins may be complicated in plant cells because recent co-expression analyses of the 35 Arabidopsis mTERF members indicated the association of mTERF proteins with DNA and RNA metabolism [49]. Recently,  . Relative accumulation of mitochondrial proteins in mterf15 mutants. Crude mitochondria were isolated from the wildtype, mterf15 and complementation lines. Following protein quantification based on the Bradford assay, equal amounts of protein were subjected to immunoblot analysis using various antibodies recognizing Nad9 [55], Cox2, Cox3, AtpA [56], and AOX [57], as well as Porin. doi:10.1371/journal.pone.0112360.g006 it was found that Zm-mTERF4, an ortholog of Arabidopsis BSM/ RUG2, is required for the splicing of several RNAs necessary for plastid translation in maize [39]. Therefore, an understanding of the mTERF family may provide new insights into the plantspecific functions of mTERF proteins in the transcriptional and post-transcriptional regulation of organellar nucleoids, as reported in this study.
In flowering plants, recombinogenic events such as the creation of intron-split genes greatly affect the dynamic nature of the mitochondrial genome [50]. The Arabidopsis mitochondrial genome contains 23 group II introns, and most are found dispersed in nad genes. The exons, including the flanking introns of these genes dispersed among the mtDNA, are transcribed and are the mRNAs generated by the splicing machinery. Until now, knowledge of the splicing machinery found in Arabidopsis mitochondria had been limited by the difficulty of organellar genome manipulation. However, a growing number of studies have revealed the involvement of a nuclear-encoded splicing factor in the excision of these introns [51]. Several proteins are involved in nad2 intron splicing. For example, 2 proteins, ABA overlysensitive 5 (ABO5) and RCC1/UVR8/GEF-like 3 (RUG3), were identified as splicing factors regulating the cis-splicing of nad2 intron 3 [35,52]. ABO5, which encodes a PPR protein, was isolated from a mutational screen for ABA sensitivity and is required for cis-splicing of nad2 intron 3 in mitochondria [35]. Another protein, RUG3, which encodes an RCC1/UVR8-like protein, is responsible for the efficient splicing of nad2 introns 2 and 3 in mitochondria [52]. The mTERF15 protein, identified in this study, is required for nad2 intron 3 RNA splicing, as demonstrated by RT-PCR and by northern blotting showing the accumulation of nad2 intron 3 in mterf15 plants ( Figure 4B-C). Moreover, RNA-binding assays also demonstrated the direct interaction between mTERF15 and nad2 intron 3. Our results suggest that mTERF15 is a new splicing factor in mitochondria. The next steps are to identify the specific elements of nad2 intron 3 that are required for the binding of mTERF15 and to explore the relationships and/or potential interactions between ABO5, RUG3 and mTERF15 in nad2 intron 3 splicing.
The nad2 intron 3 splicing defect led to the loss of mitochondrial complex I activity in mterf15. Similar results were observed with the RUG3 mutant [52]. Both mutants showed reduced root growth, small stature, and growth retardation. Another complex I mutant, ndufs4 (NADH dehydrogenase [ubiquinone] fragment S subunit 4), also exhibited a similar phenotype [53]. Although approximately one-third of mitochondrial ATP production is associated with complex I, the presence of alternative NADH dehydrogenases allows the electron transport chain to bypass complex I when complex I efficiency is reduced or when plants encounter stress [54]. Interestingly, abo5, rug3 and ndufs4 plants produced viable seeds and propagated to the next generation, whereas mterf15 showed a more severe phenotype. Homozygous mterf15 plants showed a similar phenotype as these mutants in terms of complex I deficiency; however, seeds from homozygous mterf15 plants could not germinate, even in solid medium over three months. It is possible that mTERF15 in maternal tissues of heterozygotic mterf15 embryo may partially contribute to the developing homozygous mterf15 seeds, which is completely absent in the developing seeds of homozygous mterf15 plants. This suggests that the germination defect of mterf15 is not only due to embryo lethality but also caused by aberrant reproductive organs, especially maternal tissues. The defective maternal tissues may reduce the flow of nutrient into the endosperm and eventually cause seed lethality [58]. Therefore, mTERF15 may have additional roles in mitochondria, as observed in other eukaryotes, during embryogenesis and seed development.

Supporting Information
Table S1 Primer sequences used in mTERF125 cloning, gene expression, and probe preparation for northern blotting and RT-PCR for the nad2 exon-intron junction. (DOCX)