Yeast AEP3p Is an Accessory Factor in Initiation of Mitochondrial Translation*

Initiation of protein synthesis in mitochondria and chloroplasts normally uses a formylated initiator methionyl-tRNA (fMet-tRNAfMet). However, mitochondrial protein synthesis in Saccharomyces cerevisiae can initiate with nonformylated Met-tRNAfMet, as demonstrated in yeast mutants in which the nuclear gene encoding mitochondrial methionyl-tRNA formyltransferase (FMT1) has been deleted. The role of formylation of the initiator tRNA is not known, but in vitro formylation increases binding of Met-tRNAfMet to translation initiation factor 2 (IF2). We hypothesize the existence of an accessory factor that assists mitochondrial IF2 (mIF2) in utilizing unformylated Met-tRNAfMet. This accessory factor might be unnecessary when formylated Met-tRNAfMet is present but becomes essential when only the unformylated species are available. Using a synthetic petite genetic screen in yeast, we identified a mutation in the AEP3 gene that caused a synthetic respiratory-defective phenotype together with Δfmt1. The same aep3 mutation also caused a synthetic respiratory defect in cells lacking formylated Met-tRNAfMet due to loss of the MIS1 gene that encodes the mitochondrial C1-tetrahydrofolate synthase. The AEP3 gene encodes a peripheral mitochondrial inner membrane protein that stabilizes mitochondrially encoded ATP6/8 mRNA. Here we show that the AEP3 protein (Aep3p) physically interacts with yeast mIF2 both in vitro and in vivo and promotes the binding of unformylated initiator tRNA to yeast mIF2. We propose that Aep3p functions as an accessory initiation factor in mitochondrial protein synthesis.

Initiation factor 2 (IF2) 3 plays a critical role in the initiation of translation. IF2 binds the initiator tRNA, then positions it over the initiation codon of the mRNA on the ribosome in a GTPdependent process (1). IF2 is conserved in bacteria and mitochondria (2) and is assumed to have a native structure similar to the cytoplasmic initiation factor eIF5B found in Archaea and eukaryotes (3). Bacterial and mitochondrial IF2s share similar domain structures and biochemical properties (4 -9).
In eubacteria, the initiator Met-tRNA f Met is formylated by the enzyme methionyl-tRNA formyltransferase. Formylated methionyl-tRNA (fMet-tRNA f Met ), along with Met-tRNA formyltransferase activity, is also observed in mitochondria of eukaryotes such as Saccharomyces cerevisiae, Neurospora crassa, and mammals (10 -13). These observations along with the identification of formylmethionine at the N terminus of several mitochondrially synthesized proteins (14) led to the widely accepted dogma that mitochondrial protein synthesis initiation requires fMet-tRNA f Met in a process involving mitochondrial IF2. However, it is now recognized that, at least in S. cerevisiae, mitochondrial protein synthesis can initiate with non-formylated Met-tRNA f Met , as demonstrated in yeast mutants in which the nuclear gene for mitochondrial Met-tRNA f Met transformylase (FMT1) has been deleted (15,16). Mitochondrial IF2 (mIF2) is the only initiation factor thus far identified for the yeast mitochondrial system. Three proteins (IF1, IF2, and IF3) are essential for initiation in eubacteria (17), and there must certainly be additional factors involved in initiation of mitochondrial protein synthesis. Koc and Spremulli (18) have characterized a mammalian mitochondrial IF3, but no IF3 homolog has been found in the S. cerevisiae genome. Mammalian mIF3 promotes initiation complex formation on mitochondrial ribosomes and promotes ribosome dissociation (18), like its bacterial IF3 counterpart. A possible yeast homolog was suggested by BLAST analysis, but this gene (YGL159w) is not required for mitochondrial respiration (19) and is, thus, not likely to encode the yeast mitochondrial IF3. IF1 homologs are present in all three domains of life (bacteria, Archaea, eucaryotes) (20), but no IF1 homolog has been detected in any mitochondrial translation system. Instead, biochemical and molecular modeling studies suggest that mammalian mIF2 may perform the functions of both IF2 and IF1 (21).
Mammalian mIF2 is reported to be 20 -50-fold more active with formylated versus non-formylated Met-tRNA f Met in vitro (22). Yeast mIF2 (ymIF2) is also more active with the formylated initiator tRNA in vitro but has an enhanced ability to bind unformylated initiator Met-tRNA relative to the bovine protein (8,9). Yet both of these mIF2s support translation initiation in vivo with unformylated Met-tRNA f Met (16). Based on these observations, we hypothesize the existence of an accessory factor that assists mIF2 in utilizing unformylated Met-tRNA f Met . This accessory factor might be unnecessary when formylated Met-tRNA f Met is present but becomes essential when only the unformylated species is available.
To search for other factors potentially involved in the initiation of mitochondrial protein synthesis, we developed a genetic screen using a strain (⌬fmt1) lacking formylated initiator tRNA (15). Mitochondrial protein synthesis is required for respiratory function in mitochondria, and mutation of genes encoding mitochondrial translation components invariably leads to a respiratory-deficient (petite) phenotype (23). Thus, the screen was designed to identify mutants that cause a respiratory defect only in combination with the ⌬fmt1 mutation (i.e. synthetic petites). One of the genes identified in this screen (AEP3) encodes a peripheral mitochondrial inner membrane protein reported to stabilize the bicistronic transcript of the mitochondrially encoded ATP6 and ATP8 genes (24). Here we show that the AEP3 protein (Aep3p) physically interacts with ymIF2 and promotes the binding of unformylated initiator tRNA to ymIF2. S. cerevisiae strains and plasmids are listed in Table 1. Yeast cells were grown in YPD (1% yeast extract, 2% Bacto-peptone, and 2% glucose), YPEG (1% yeast extract, 2% Bacto-peptone, 2% ethanol, and 3% glycerol), or YMD (0.7% yeast nitrogen base without amino acids, 2% glucose) as described previously (16). Plates used for eviction of URA3 plasmids contained 0.1% 5-FOA (Biovectra; Charlottetown, PEI, Canada). Yeast genomic DNA was isolated using the Qiagen DNeasy Blood and Tissue kit. Haploid ⌬aep3 cells (DLY2) were generated from EUROSCARF strain Y22823 (Table 1) by sporulation and tetrad dissection as described previously (25). DLY2 was maintained with a plasmid (pRS416-AEP3, Table 1) encoding the wild-type AEP3 gene, as aep3 deletion mutants are respiratory-deficient (petite) and exhibit mitochondrial DNA instability (24). Other AEP3 constructs could then be transformed into DLY2 on pRS415 and studied after eviction of the wild-type pRS416-AEP3 plasmid by growth on 5-FOA.

EXPERIMENTAL PROCEDURES
Full-length ymIF2 as well as the C2-subdomain of ymIF2 were prepared as described previously (8,9). Yeast cytoplasmic C 1 -THF synthase was purified as described (26).
Polyclonal antibodies against purified recombinant Histagged ymIF2 were produced in rabbits by the MD Anderson Cancer Center (Bastrop, TX). The IgG-fraction of the immune serum was purified by Protein A-Sepharose (Sigma) chromatography. This IgG fraction was used in the experiments described below.
Genetic Screen for Mutations Creating a Synthetic Petite Phenotype in the Presence of ⌬fmt1-The FMT1 gene, including 360 bp of upstream sequence, was PCR-amplified from DAY4 yeast genomic DNA (using oligonucleotides FMT1f2 (5Ј-CAGTGGATCCACACCCAATTGCGAGCC-TAA-3Ј) and FMT1r2 (Ј-GTAGATCGATATGTAGAGCC-GGGTTACAGG-3Ј), cleaved with BamHI and ClaI, and ligated into the centromeric shuttle vector, pRS416. The Y03039 (⌬fmt1) strain was transformed with the resulting pRS416-FMT1 construct, selecting for Uraϩ transformants. The resulting strain was grown in minimal media to late log phase and exposed to UV mutagenesis in a Stratalinker UV cross-linker for 6 s, resulting in ϳ20% cell survival. Mutagenized cells were diluted and plated for single colonies on YPEG (rich complete, nonfermentable) medium and incubated at 30°C for 6 -7 days. Colonies were replica-plated from the YPEG plates onto minimal glucose media (YMD) containing 1 g/liter 5-FOA and amino acids necessary for Y03039 growth. After the forced eviction of the pRS416-FMT1 plasmid, colonies from the YMD/5-FOA plates were replica-plated once again onto YPEG medium. Approximately 10,000 colonies were screened (200 plates with ϳ50 colonies/plate), and mutants able to grow on the first YPEG plate (with the wildtype FMT1 gene on a plasmid) but unable to grow on the second YPEG plate (after plasmid eviction) were isolated. The synthetic petite phenotype was verified by re-streaking the cells on the same series of plates. Putative synthetic petite mutants were crossed with a ⌬fmt1 strain of opposite mating type, and resulting diploids were spread onto YPEG/5-FOA plates, testing for the presence of a recessive nuclear mutation responsible for the synthetic petite phenotype. Eight synthetic petite strains were isolated from the screen; the synthetic petite strain 175 (s.p. 175) was used for the studies reported herein.
Cloning Yeast AEP3 for Protein Expression in E. coli-The wild-type AEP3 ORF (YPL005W) was amplified from genomic DAY4 DNA using primers AP3-pMAL-For (5Ј-CCGGGATC-CATGAATACATTAAGG-3Ј; the BamHI site is underlined, and the start codon is in bold) and AP3-pMAL-Rev (5Ј-CCG-AAGCTTTCAAACCTCCCCAAC-3Ј; the HindIII site is underlined, and the stop codon is in bold). PCR was carried out using KOD Hot Start DNA polymerase. The resulting PCR fragment was digested with BamHI and HindIII, gel-purified, inserted behind the maltose-binding protein (MBP) ORF of pMALc2H 10 T (a gift from Dr. John Tesmer, University of Michigan) (29), and transformed into E. coli JM109 to generate MBP-Aep3p. The construct was verified by sequencing at the DNA Core Facilities of the Institute for Cellular and Molecular Biology at the University of Texas at Austin. Recombinant plasmid was transformed into E. coli Rosetta(DE3) cells (Novagen) to express the MBP and decahistidine tag fused to the fulllength Aep3 protein (MBP-Aep3). Heterologous protein was expressed by growth in Luria broth and induction with 0.1 mM isopropyl-␤-D-thiogalactopyranoside at 15°C overnight. The cells were harvested, washed, and stored at Ϫ80°C.
Purification of MBP-Aep3-Cells from a 2-liter culture of E. coli expressing MBP-Aep3 were lysed by sonication in 50 ml of 1ϫ binding buffer (5 mM imidazole, 20 mM Tris-HCl, pH 7.9, 500 mM NaCl) to which one tablet of protease inhibitor (Roche Applied Science) was added. Cells were lysed by sonication on ice (15 s with 15-s intervals for 6 cycles). The supernatant was collected after centrifugation at 17,000 rpm for 30 min and applied to a 5-ml His⅐Bind column that had been charged with Ni 2ϩ and equilibrated in 1ϫ binding buffer. After washing with 50 ml of binding buffer, loosely bound protein was eluted with wash buffer containing 60 mM imidazole, 500 mM NaCl in Tris buffer. Tightly bound protein was eluted by 250 mM imidazole in the same Tris/NaCl buffer. The pooled protein fractions were dialyzed against 500 ml of 20 mM Tris-HCl, pH 7.5, for 4 h and stored at Ϫ80°C in small aliquots. Protein concentration was determined by the Bradford assay (30), and purity was analyzed by SDS-PAGE (31). The maltose-binding protein alone was expressed from pMALc2H 10 T and purified as described above to serve as a control in the binding assays.
In Vitro Filter Binding Assay-The filter binding assay described by Sundari et al. (32) was used to study tRNA binding to yeast ymIF2 and MBP-Aep3. Different amounts of each protein were incubated for 20 min on ice in 50 l of total volume containing 50 mM Tris-HCl, pH 7.0, 2 mM dithiothreitol, and 10 g of bovine serum albumin with 4 pmol of 35 S-labeled initiator tRNA (fMet-tRNA f Met or Met-tRNA f Met ). [ 14 C]Lys-tRNA Lys and MBP were used in control reactions. Binding reactions were stopped with 2 ml of ice-cold solution containing 50 mM Tris-HCl, pH 7.0, and 2 mM dithiothreitol and then filtered through nitrocellulose filters followed by washing 5 times with the same buffer. Filters were dried, and radioactivity on filters was determined by liquid scintillation counting. Each assay was performed in duplicate.
MBP Pulldown Assay-MBP-Aep3 and ymIF2 (concentrations indicated in the legends to Figs. 3, 4, and 6) were incubated at room temperature with gentle rotation for 2 h in 200-l final volume containing 20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA, 2 mM dithiothreitol. Thirty l of a slurry containing amylose affinity beads (New England Biolabs; diluted 50% with phosphate-buffered saline) were added and incubated for an additional 1 h while rotating at room temperature. The beads were spun down and removed from the supernatant (unbound fraction) and then washed six times using the same buffer. Bound proteins were eluted from the washed beads by boiling for 5 min in SDS-PAGE sample buffer and analyzed by SDS-PAGE on 10% gels. MBP alone and yeast cytoplasmic C 1 -THF synthase were used as control proteins. Immunoblotting was performed using antibodies against ymIF2 or yeast cytoplasmic C 1 -THF synthase (33). The immunoblots were visualized with the enhanced chemiluminescence detection system (GE Healthcare).
In Vivo Co-immunoprecipitation-An epitope-tagged version of AEP3 with the hemagglutinin (HA) epitope fused to its C terminus was constructed by PCR amplification of the AEP3 gene using forward primer A3-FOR (5Ј-AGCCTCGAGCAT-AGCAGATTATACTA-3Ј; the XhoI site is underlined) and reverse primer AP3-HA-Rev (5Ј-GGCGGATCCTCAAGC-GTAGTCTGGGACGTCGTATGGGTAAACCTCCCCAAC-TATTCTCCTCTTTAC-3Ј; the BamHI site is underlined, and the stop codon is in bold; HA epitope tag sequence italicized). The resulting PCR product was digested with XhoI and BamHI, gel-purified, and ligated into pRS415. The resulting construct (pRS415-AEP3-HA) was sequence-verified.
Yeast cells containing HA-tagged AEP3 were grown in YMD, harvested, and washed twice with ice-cold water. Cells were resuspended in 500 l of co-immunoprecipitation buffer (34) containing 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 10% glycerol, 1 mM dithiothreitol, 10 mM MgCl 2 , and one complete-mini EDTA-free protease inhibitor mixture tablet. Cells were disrupted with glass beads using a FastPrep FP120 cell disrupter (Qbiogene). A cleared soluble lysate was obtained by centrifugation at 16,000 rpm for 30 min. This was saved, and the glass beads were washed in 500 l of same buffer and centrifuged again. The two supernatants were combined (1 ml total) and then incubated with 10 l of anti-HA antibodies (1:100 dilution; Covance, Trenton, NJ) at 4°C for 4 h with gentle rotation.
Fifty l of Protein A-Sepharose beads (Invitrogen) was then added, and the incubation was continued at 4°C overnight. Immunocomplexes were collected by centrifugation at 1500 rpm for 2 min, washed 5 times with 500 l of co-immunoprecipitation buffer, and eluted from the beads with 50 l of SDS-PAGE sample buffer. Five l of each immunoprecipitate were analyzed by SDS-PAGE on a 10% gel followed by immunoblotting with the ymIF2 polyclonal antibody (1:10,000 dilution).
Labeling of Mitochondrial Translation Products-Yeast cells were inoculated in 10 ml of synthetic complete medium (0.7% yeast nitrogen base and standard concentrations of adenine, uracil, and amino acids) (35) containing 2% (w/v) galactose as the carbon source. Leucine was omitted from the synthetic completeϪgalactose medium for growth of the DLY2 strains transformed with LEU2 plasmids (pRS415-AEP3 and pRS415-aep3Y305N), and uracil was omitted for growth of strain s.p. 175 harboring the pRS416-FMT1 plasmid. In some experiments Uraϩ plasmids (pRS416-FMT1 or pRS416-AEP3) were evicted from AEP3 mutant strains by streaking on YMD/5-FOA plates before metabolic labeling. Cells were grown to mid-log phase, harvested, and washed with minimal galactose medium (0.7% yeast nitrogen base, 2% galactose) without amino acids. In vivo labeling of mitochondrial translation products using [ 35 S]methionine in the presence of cytoplasmic protein synthesis inhibitor cycloheximide was performed as described previously (36). Translation products were separated by SDS-PAGE using gels (15% acrylamide, 0.1% bisacrylamide) prepared as described (36). Typically, 25% of the final resuspended protein extract was loaded on the gel for SDS-PAGE analysis.
Yeast strains used in these metabolic labeling experiments were also tested for respiratory deficiency using a tetrazolium overlay assay (15,37). Cells grown in synthetic com-pleteϪgalactose medium were plated on rich glucose me-dium, and colonies were covered with agar containing 0.1% 2,3,5triphenyltetrazolium chloride to determine the number of respiratory-defective (petite; white) and respiratory-competent (red) colonies.

Isolation of Mutants That Are
Synthetically Petite with ⌬fmt1-Deletion of the FMT1 gene has no effect on respiratory growth (15). Because the only phenotype of a ⌬fmt1 mutant is the lack of formylated initiator tRNA (15), we reasoned that genes identified in this screen would likely encode proteins specifically involved in the initiation process of mitochondrial translation. The genetic screen was designed to generate and identify mutations that, in combination with the fmt1 deletion, lead to respiratory deficiency. Approximately 10,000 colonies were screened, and eight recessive synthetic petite mutants were isolated. One mutant strain, s.p. 175, was transformed with a yeast genomic library, resulting in four colonies (of 6000 screened) that harbored plasmids complementing the synthetic petite phenotype. These plasmids were rescued and their inserts sequenced, identifying a region of chromosome XVI spanning nucleotides 547857-557733. This region contains only three complete ORFs: AEP3, LSP1, and ULA1 (Saccharomyces Genome Database). Because AEP3 was previously characterized as a mitochondrial protein (24), the AEP3 open reading frame was subcloned into the LEU2 pRS415 vector and transformed into s.p. 175 yeast harboring the FMT1 gene on pRS416 (URA3 vector). The resulting Uraϩ Leuϩ transformants were re-streaked onto YPEG/5-FOA plates to select cells that evicted the pRS416-FMT1 vector and to test for growth on a non-fermentable carbon source (YPEG; YP ϩ ethanol/glycerol). The growth of AEP3-expressing cells on YPEG/5-FOA demonstrates complementation of the s.p. 175 synthetic petite mutation (Fig. 1A). The AEP3 locus was PCR-amplified from s.p. 175 genomic DNA. Sequence analysis revealed a single T-to-A point mutation that results in a tyrosine-to-asparagine substitution at amino acid 305 (Y305N). This point mutation alone (in a wild-type FMT1 background) does not cause respiratory deficiency, whereas the null aep3 mutant is unable to grow on the non-fermentable carbon source YPEG (Fig. 1B). The aep3-Y305N point mutation does, however, cause a slight growth deficit on YPEG, but not on YPD, which supports growth by fermentation (Fig. 1B).
It seemed likely that the genetic interaction between FMT1 and AEP3 was due to the lack of formylated mitochondrial Met-tRNA f Met in the fmt1 mutant rather than some direct (physical) interaction between the proteins encoded by these two genes. If this is indeed the case, the Aep3p-Y305N mutant should be synthetically petite with any mutation that eliminates formy-  175 (⌬fmt1 aep3-Y305N), harboring full-length FMT1 on a URA3 plasmid, was transformed with either wild-type pRS415-AEP3 (top) or empty plasmid (pRS415, bottom) as a negative control, streaked onto a 5-FOA/YPEG plate, and incubated at 30°C for 5 days. B (top), strain DLY2 (⌬aep3 FMT1 ϩ ), harboring wild-type AEP3 on a URA3 plasmid (pRS416), was transformed to leucine prototrophy with either pRS415-aep3-Y305N, wild-type pRS415-AEP3, or empty plasmid (pRS415) as a negative control. The pRS416-AEP3 plasmid was evicted by growth on 5-FOA, and then the transformants were streaked onto a YPEG plate and incubated at 30°C for 5 days. Bottom, 10-fold serial dilutions of DLY2 harboring either the wild-type or mutant aep3 plasmid were spotted on YPD or YPEG plates and incubated at 30°C for 3 or 6 days, respectively. C, the original synthetic petite 175 strain (⌬fmt1 aep3-Y305N leu2 MIS1/pRS416-FMT1) and a Leu ϩ integrant of this strain were streaked onto YPEG plates and incubated at 30°C for 5 days to test for respiratory growth. MIS1, wild-type MIS1 locus, ⌬mis1::LEU2, disrupted mis1 locus. See "Experimental Procedures" for details of MIS1 gene disruption. DECEMBER 4, 2009 • VOLUME 284 • NUMBER 49 JOURNAL OF BIOLOGICAL CHEMISTRY 34119 lated mitochondrial Met-tRNA f Met . The MIS1 gene encodes the mitochondrial isozyme of trifunctional C 1 -THF synthase, which is responsible for synthesis of mitochondrial 10formyltetrahydrofolate (38). 10-Formyltetrahydrofolate is the formyl donor for the Met-tRNA formyltransferase (FMT1 gene product) (39). Deletion of the MIS1 gene eliminates the substrate for the formylation reaction, and mis1 mutants contain only unformylated Met-tRNA f Met in their mitochondria (15). Thus, we tested whether the Aep3p-Y305N mutant exhibits a synthetic interaction with a mis1 deletion. The s.p. 175/ pRS416-FMT1 strain (leu2 MIS1 ϩ aep3-Y305N FMT1 ϩ ) was transformed with a PCR-amplified LEU2 cassette with MIS1flanking regions on each end of the LEU2 gene to direct recombination at the MIS1 locus. Genomic DNA was isolated from Leuϩ transformants, and successful mis1 gene disruption (mis1::LEU2) was verified by PCR (data not shown). Cells disrupted at the MIS1 locus but still harboring the wild-type FMT1 gene from the pRS416-FMT1 plasmid were unable to grow on YPEG plates (Fig. 1C), demonstrating a genetic interaction between MIS1 and AEP3. This result confirms a role for Aep3p specifically in the initiation of mitochondrial protein synthesis in the absence of formylated initiator tRNA.

AEP3p Functions in Mitochondrial Translation Initiation
The genetic interaction between AEP3 and FMT1 (or MIS1) suggested that Aep3p might physically interact with factors that are involved in the initiation stage of mitochondrial protein synthesis. Mitochondrial IF2 is responsible for binding and delivering the initiator tRNA to the small ribosomal subunit, and the yeast mIF2 shows a 3-4-fold binding preference for formylated versus unformylated Met-tRNA f Met (8). We speculated that Aep3p might interact with ymIF2 to stabilize initiator tRNA binding. Alternatively, Aep3p might interact directly with the initiator tRNA. To distinguish these two possibilities, yeast Aep3p was purified for in vitro binding assays.
Expression of Aep3p as a Fusion Protein-Attempts to express in E. coli the yeast Aep3p fused to an N-terminal polyhistidine tag produced only insoluble protein (data not shown). Given the tight membrane association of Aep3p in yeast mitochondria (24), we constructed a MBP fusion in an attempt to increase its solubility (40). Therefore, the full-length AEP3 ORF (606 amino acids) was cloned into the vector pMALc2H 10 T to express N-terminal MBP and a decahistidine tag fused to Aep3p (MBP-Aep3). E. coli cells expressed this fusion protein at high levels in soluble form. MBP-Aep3 was purified by nickel affinity chromatography, resulting in nearly homogeneous preparation of fusion protein that migrated with an apparent mass of 110 kDa, consistent with the combined size of Aep3p plus MBP (Fig. 2). The MBP tag provides a useful handle in the pulldown assays described below. The Aep3p-Y305N mutant protein was also expressed as an MBP fusion, but the protein degraded rapidly (data not shown) and could not be used for in vitro assays.
Aep3p Does Not Exhibit Significant tRNA Binding-A filter binding assay was used to test whether Aep3p binds initiator tRNA. MBP-Aep3 exhibited no significant binding activity with either formylated or unformylated Met-tRNA f Met nor with Lys-tRNA Lys (Fig. 3). Yeast mIF2 showed significant binding activity with unformylated Met-tRNA f Met (positive control). MBP alone showed no binding (negative control).
Aep3p Interacts with ymIF2 in Vitro and in Vivo-We next used a MBP pulldown assay to test whether Aep3p physically interacts with ymIF2. Purified MBP-Aep3p and ymIF2 were incubated for 2 h at room temperature. Purified MBP alone was used as a control. Complexes were isolated with amylose affinity beads and analyzed by SDS-PAGE and immunoblotting with antibodies against ymIF2. Fig. 4 shows that ymIF2 (ϳ70 kDa) is pulled down with MBP-Aep3p (lane 2) but not with MBP alone (lane 4). MBP-Aep3 also efficiently pulls down a 130-amino acid C-terminal fragment of ymIF2 (ϳ16 kDa) (lane 6). This C-terminal segment contains the C2 subdomain, responsible for initiator tRNA binding (9). To rule out a nonspecific interaction due to Aep3p being a "sticky" protein (a common prob-  lem with membrane-associated proteins), we tested MBP-Aep3 with purified yeast cytoplasmic C 1 -THF synthase (33). An immunoblot of the pulldown fraction with antibodies against C 1 -THF synthase were negative (Fig. 4, lane 8), suggesting the interaction between Aep3p and ymIF2 is specific.
To test whether Aep3p binds ymIF2 in vivo, we prepared an HA-tagged version of Aep3p in a low-copy number vector (pRS415-AEP3-HA). This construct was transformed into a heterozygous diploid aep3/AEP3 strain (Y22823; Table 1). The transformed strain was sporulated and dissected to obtain a haploid ⌬aep3 strain harboring pRS415-AEP3-HA. This strain grew normally on non-fermentable carbon sources (data not shown), confirming that the HA tag did not interfere with AEP3 function. Extracts prepared from these cells were immunoprecipitated with antibodies against HA as described under "Experimental Procedures." However, immunoblots of the immunoprecipitates using antibodies against ymIF2 were negative (data not shown). In fact, using these antibodies we are unable to detect ymIF2 in whole cell extracts of IFM1 ϩ yeast unless it is overexpressed from a plasmid (data not shown), suggesting that endogenous levels of ymIF2 are quite low. Therefore, we repeated the experiment using a strain co-expressing ymIF2 from a multicopy plasmid (pVT101U-ymIF2) and pRS415-AEP3-HA (low copy number). This strain (LOY1, Table 1) is disrupted at the chromosomal IFM1 locus (encoding ymIF2) and expresses ymIF2 only from the plasmid. As shown in Fig. 5, anti-HA antibodies co-immunoprecipitated ymIF2 from extracts of the yeast strain expressing both AEP3-HA and ymIF2 (lane 2). No signal for ymIF2 was seen in an HA coimmunoprecipitate from an extract prepared from a strain lacking ymIF2 (lane 3). The diffuse immunostaining below 50 kDa in lanes 2 and 3 is a nonspecific cross-reaction of the anti-ymIF2 antibodies with additional proteins pulled down with the HA-tagged Aep3p. This cross-reaction is not observed in lane 1, as much less total protein is present in the immunoprecipitate due to lack of HA-tagged Aep3p in this control extract. Nonetheless, the 70-kDa ymIF2 band is clearly enriched in the HA-Aep3p pulldown assay (lane 2). These data suggest that the Aep3p-ymIF2 interaction observed in vitro also occurs in vivo.
Aep3p Stimulates the Binding of Unformylated Initiator tRNA to ymIF2-Our genetic and binding data suggest that Aep3p might interact with ymIF2 to stabilize initiator tRNA binding. The filter binding assay was used to test whether Aep3p affected the initiator tRNA binding by ymIF2 (Fig. 6).   Aep3p-Y305N Causes a Mitochondrial Protein Synthesis Defect in Combination with ⌬fmt1-The Y305N point mutant of AEP3 was identified as a petite mutation when combined with deletion of FMT1. To try to understand the cause of the petite phenotype, mitochondrial protein synthesis was analyzed in various AEP3/FMT1 backgrounds. In an FMT1 ϩ background, aep3-Y305N supported robust mitochondrial protein synthesis, including Atp6p (Fig. 7, lane 2) (our gel system does not resolve Atp8p from Atp9p, so we could not determine whether Atp8p was synthesized in the aep3-Y305N FMT1 strain). Cox1 labeling appeared to be somewhat decreased compared with the AEP3 wild-type strain (Fig. 7, lane 1 versus 2). On the other hand, the aep3-Y305N point mutant causes a severe mitochondrial protein synthesis defect after FMT1 is evicted (Fig. 7, lane 3 versus 4). Labeling of all the mitochondrial products is depressed, similar to that observed in the aep3 deletion strain (Fig. 7, lane 5). We determined the percentage of petites in each population at the time of harvest for the protein synthesis assays using a tetrazolium overlay assay (15,37). The strains represented in lanes 3-5 of Fig. 7 (aep3-Y305N FMT1, aep3-Y305N ⌬fmt1, ⌬aep3 FMT1) were 25, 80, and Ͼ99% petite, respectively. These results are consistent with the respiratory phenotypes observed in Fig. 1 and suggest that the petite phenotype of the ⌬fmt1/aep3-Y305N genotype results from defective mitochondrial protein synthesis.

DISCUSSION
We demonstrate here a functional interaction between Aep3p and mitochondrial translation initiation factor 2, revealed when only unformylated initiator tRNA is available. A point mutant of the AEP3 gene, encoding a protein with a tyrosine-to-asparagine substitution at position 305, was found to cause a synthetic respiratory defect in cells lacking formylated Met-tRNA f Met . Aep3p was found to physically interact with mitochondrial IF2, and this interaction stimulated the binding of unformylated Met-tRNA f Met to ymIF2. We propose that Aep3p is an accessory initiation factor that stabilizes the mIF2-Met-tRNA f Met interaction in the 37 S initiation complex. Aep3p was initially characterized as a mitochondrial inner membrane protein required for stabilization and/or processing

AEP3p Functions in Mitochondrial Translation Initiation
of the bicistronic transcript of the mitochondrially encoded ATP6/ATP8 genes (24). The ATP6 and ATP8 genes encode subunits of the mitochondrial ATP synthase. Deletion of the AEP3 gene causes respiratory incompetence (petite phenotype) and reduces the levels of cytochromes b and aϩa3 as well as the activities of NADH-cytochrome c reductase, cytochrome c oxidase, and oligomycin-sensitive ATPase. AEP3 deletion causes instability of the mitochondrial DNA, leading eventually to rho Ϫ /rho 0 cells. Indeed, one of the hallmarks of S. cerevisiae strains impaired in mitochondrial protein synthesis is extreme instability of the mitochondrial DNA, giving rise to rho Ϫ and rho 0 mutants at high frequency (41). In vivo labeling of mitochondrial translation products revealed that both subunits 6 (Atp6p) and 8 (Atp8p) were almost undetectable (24). Northern analysis showed that in the aep3 deletion strain, the long ATP8/6 transcript was reduced to 34% that of wild-type levels, and the short ATP8/6 transcript was undetectable. Finally, Aep3p possesses one or more degenerate pentatricopeptide repeat (PPR) motifs, which are implicated in RNA binding (42). Based on these data, Ellis et al. (24) propose that Aep3p interacts with the ATP6/8 mRNA at the matrix side of the mitochondrial inner membrane, stabilizing the transcript and facilitating insertion of newly translated Atp6p and Atp8p subunits into the membrane for assembly into the functional F o component of the ATP synthase. Our data suggest an additional, more general, role for Aep3p in mitochondrial translation; that is, as an accessory initiation factor that interacts with mIF2. This general translation initiation function is compatible with Aep3p inner membrane localization and specific involvement in ATP6/8 mRNA translation. Mitochondrial protein synthesis is known to occur on the matrix face of the inner membrane (43), where Aep3p is localized. Although Ellis et al. (24) noted the dramatic loss of ATP6/8 expression, translation of nearly all of the mitochondrially encoded proteins was depressed in their aep3 deletion strain. Our in vivo labeling results (Fig. 7) are qualitatively similar.
The aep3-Y305N point mutant, on the other hand, supports respiration ( Fig. 1) and mitochondrial protein synthesis (including Atp6p; Fig. 7) as long as formylated Met-tRNA f Met is available. However, when formylation of the initiator tRNA is prevented (either by fmt1 or mis1 mutation), the Y305N point mutant can no longer support robust mitochondrial protein synthesis and is respiratory-incompetent (petite). Translation of most of the mitochondrially encoded proteins is severely depressed under these conditions (Fig. 7). Thus, we propose that the petite phenotype of the aep3-Y305N ⌬fmt1 strain is caused by defective mitochondrial protein synthesis.
Physical interaction between Aep3p and mIF2, the initiation factor responsible for binding initiator tRNA, provides a bio-chemical mechanism underlying the genetic interaction. This interaction, which was also observed in vivo, somehow stabilizes the binding of unformylated Met-tRNA f Met to ymIF2 (Fig. 6). In contrast, Aep3p appears to have an inhibitory effect on the binding of formylated Met-tRNA f Met by ymIF2. The binary interaction between IF2 and initiator tRNA, even with formylated Met-tRNA f Met , is inherently weak and may not represent a physiologically relevant complex. The ternary complex of initiator tRNA, IF2, and 30 S subunit is much more stable than any of the binary complexes alone (44,45). Thus, Aep3p may have a different effect in vivo than in these in vitro binding experiments. Nonetheless, the stimulatory effect seen with unformylated Met-tRNA f Met indicates that the physical interaction between Aep3p and mIF2 has functional consequences.
These in vitro binding data fit nicely with previously published genetic and metabolic data (15,16). Both mammalian and yeast mIF2 show a distinct preference for binding formylated Met-tRNA f Met , yet both of these proteins support translation initiation in vivo with unformylated Met-tRNA f Met (16). We propose that Aep3p is a normal component of the mitochondrial translation initiation complex. When mitochondria contain formylated initiator tRNA, the major role of Aep3p is in facilitating ATP6/8 translation (24). However, under conditions where formylation of the initiator tRNA is prevented, e.g. when mitochondria are deficient of folates or one-carbon units, Aep3p has a critical role in stabilizing the otherwise weak binding of unformylated Met-tRNA f Met by mIF2. The mechanistic details of this stabilizing effect remain to be determined. Initiator tRNA can bind to the 30 S subunit in the absence of IF2 (45), and in fact, mutant yeast completely lacking mIF2 retain the ability to synthesize low levels of some mitochondrially encoded proteins (16). The structure of the 30 S initiation complex from Thermus thermophilus reveals binding interactions between the tRNA and the 30 S subunit, the tRNA and IF2, and IF2 and the 30 S subunit (45). It is the cooperativity of these interactions that provides for stable binding of the IF2-fMet-tRNA f Met subcomplex to the 30 S subunit. Assuming a similar structure exists for the mitochondrial small subunit (37 S) initiation complex, Aep3p may contribute further stability through its interactions with mIF2.
We do not know how the Y305N point mutation leads to defective function of Aep3p. The mutation may destabilize the protein, as we were unable to purify full-length recombinant mutant protein either alone or as a MBP fusion protein. Nonetheless, the Y305N mutant protein is apparently able to fulfill its ATP6/8 stabilizing role in vivo in an FMT1 ϩ background, as mitochondrial protein synthesis appears normal (Fig. 7), with the possible exception of a reduction in Cox1p. The Y305N mutation occurs in a region of high sequence identity with Aep3p homologs in other related Saccharomyces species (Fig. 8) (46). This region lies between two pairs of degenerate PPR motifs. Another mutation in this same region, Q320P, causes an acetate growth defect (47). Because acetate utilization requires mitochondrial respiration, this is also consistent with a role for FIGURE 8. Predicted domain structure of Aep3p. The 606-amino acid AEP3 ORF was analyzed for PPR motifs using TPRpred (Bioinformatics Toolkit). The four highest scoring PPR motifs are indicated along with the residue numbers at their boundaries. The two point mutations discussed in the "Discussion" (Y305N, Q320P) are located in the region of highest sequence identity with Aep3p homologs in other related Saccharomyces species (46). MTS, mitochondrial targeting sequence.
Aep3p in mitochondrial protein synthesis. Ellis et al. (24) suggested that this region "may be important for some other conserved function of Aep3p," and it is tempting to speculate that this region is involved in the interaction with ymIF2.
It is likely that other mitochondrial translation initiation factors remain to be discovered. Vial et al. (48) showed that the phenotype of an fmt1 disruption depends on the genetic background of the strain; growth on glycerol at a high temperature (37°C) in rich medium or growth on glycerol in minimal medium at 30°C was slower in some, but not all, ⌬fmt1 strains. This suggests that other gene products interact with the initiator tRNA in mitochondrial protein synthesis. In this context, Williams et al. (49) used a similar synthetic petite screen to search for genes that interact with RSM28, a dispensable component of the mitochondrial ribosomal small subunit. Their screen identified three interacting genes: FMT1, IFM1, and RMD9. The synthetic interaction between FMT1 and RSM28 reveals that initiator tRNA formylation is essential in the absence of Rsm28p. Thus, either the Y305N mutation in Aep3p (this work) or the loss of Rsm28p function (49) leads to a requirement for initiator tRNA formylation to support normal mitochondrial protein synthesis and respiratory growth. Genetic interaction with the genes encoding methionyl-tRNA formyltransferase and ymIF2 provides strong evidence that Rsm28p has a role in translation initiation. Williams et al. (49) speculate that the small subunit protein Rsm28p might provide IF1 and/or IF3-like functions in yeast mitochondrial translation initiation.
The other gene identified in their screen, RMD9, encodes an extrinsic membrane protein located on the matrix face of the mitochondrial inner membrane (50). RMD9 was independently isolated as a high-copy suppressor of a respiratory-deficient oxa1 mutant and was shown to control the processing and/or stability of several mitochondrial mRNAs (50). These data are consistent with the inner membrane localization of most of the mitochondrial gene expression machinery, including ribosomes (51,52), translational activators (53), RNA processing and stability factors (24,54,55), and even RNA polymerase (54,56). Nouet et al. (50) have proposed that Rmd9p delivers mRNAs to the inner membrane-localized ribosomes for translation and insertion of the nascent proteins into the membrane. The genetic interaction between RMD9 and RSM28 observed by Williams et al. (49) links Rmd9p function, like that of Aep3p, specifically with the initiation stage of translation.
Finally, it is worth noting that all four of the proteins in S. cerevisiae predicted to possess PPR motifs (Aep3p, Rmd9p, Pet309p, and Ccm1p) are mitochondrial proteins involved in the processing or stabilization of mitochondrial mRNAs (24,50,55,57). Both Aep3p and Pet309p were initially discovered as factors that control the stability of a single mRNA (ATP6/8 in the case of Aep3p; COX1 in the case of Pet309p) (24,58). However, the COX1 and ATP6/8 mRNAs are both derived from a single polycistronic COX1-ATP6/8 transcript that is processed to yield the two mRNAs (24), so it might be expected that Aep3p and Pet309p function together. Tavares-Carreón et al. (59) showed that the PPR motifs in Pet309p are necessary for Cox1p translation but not for stabilization of the COX1 mRNA. Rmd9p (50) and Ccm1p (57) appear to control the processing and stability of multiple mitochondrial mRNAs. The function(s) of the PPR motifs shared by these four proteins are not known in any detail, but PPR motifs have been implicated in RNA binding and post-transcriptional gene regulation in organelles, including transcript processing and stabilization, tRNA binding, and translation (42,60,61). The discovery that Aep3p (this work) and Rmd9p (49) are also functionally linked to mitochondrial translation initiation factors suggests that the coupling between mRNA stability and translation may be a general phenomenon mediated by RNA-binding proteins and occurs mechanistically at the initiation stage.