Identification of Mammalian Mitochondrial Translational Initiation Factor 3 and Examination of Its Role in Initiation Complex Formation with Natural mRNAs*

Human mitochondrial translational initiation factor 3 (IF3mt) has been identified from the human expressed sequence tag data base. Using consensus sequences derived from conserved regions of the bacterial IF3, several partially sequenced cDNA clones were identified, and the complete sequence was assembled in silico from overlapping clones. IF3mt is 278 amino acid residues in length. MitoProt II predicts a 97% probability that this protein will be localized in mitochondria and further predicts that the mature protein will be 247 residues in length. The cDNA for the predicted mature form of IF3mt was cloned, and the protein was expressed inEscherichia coli in a His-tagged form. The mature form of IF3mt has short extensions on the N and C termini surrounding a region homologous to bacterial IF3. The region of IF3mt homologous to prokaryotic factors ranges between 21–26% identical to the bacterial proteins. Purified IF3mt promotes initiation complex formation on mitochondrial 55 S ribosomes in the presence of mitochondrial initiation factor 2 (IF2mt), [35S]fMet-tRNA, and either poly(A,U,G) or an in vitro transcript of the cytochrome oxidase subunit II gene as mRNA. IF3mtshifts the equilibrium between the 55 S mitochondrial ribosome and its subunits toward subunit dissociation. In addition, the ability ofE. coli initiation factor 1 to stimulate initiation complex formation on E. coli 70 S and mitochondrial 55 S ribosomes was investigated in the presence of IF2mt and IF3mt.

Mammalian mitochondria synthesize 13 polypeptides that are essential for oxidative phosphorylation. These 13 proteins are translated from nine monocistronic and two dicistronic mRNAs with overlapping reading frames (1,2). The proteinsynthesizing system of mammalian mitochondria has a number of interesting features not observed in prokaryotes or the cell cytoplasm (3). The mRNAs in this organelle have an almost complete lack of 5Ј-and 3Ј-untranslated nucleotides. The start codon is generally located within three nucleotides of the 5Ј end of the mRNA (1,4). Thus, mammalian mitochondrial ribosomes do not recognize the start codon using the Shine/Dalgarno interaction between the mRNA and the 16 S rRNA as observed in prokaryotes. Further, this system does not use a cap-binding and scanning mechanism such as observed in the eukaryotic cytoplasm.
Three translational initiation factors, IF1, IF2, and IF3, 1 are required for the initiation of protein synthesis in bacteria (5)(6)(7). Prior to the present report, the homolog of only one of these factors, IF2 mt , had been identified, cloned, and characterized in mammalian mitochondria (8 -12). Similar to its prokaryotic counterpart, IF2 mt promotes the binding of fMet-tRNA to the small subunit of mitochondrial ribosomes in response to synthetic polynucleotides such as poly(A,U,G).
The current report describes the identification and initial characterization of the mammalian mitochondrial factor equivalent to IF3. In prokaryotes IF3 has a number of roles in the initiation of protein synthesis. IF3 binds to the 30 S subunit and inhibits its association with the 50 S subunit, thus ensuring a supply of 30 S subunits for initiation (13,14). IF3 also promotes an adjustment of the position of the mRNA on the 30 S subunit facilitating codon-anticodon interactions between the AUG codon and fMet-tRNA in the P site (15)(16)(17)(18). IF3 acts to switch the decoding preference of the small ribosomal subunit from elongator tRNAs to the initiator tRNA in the P site, thus playing a proofreading role in initiation (19 -21). IF3 is a small protein of 180 amino acids that folds into two distinct domains separated by a long flexible linker. The C-terminal domain is thought to carry out most of the direct functions of this factor, whereas the N-terminal domain stabilizes the interaction of IF3 with the 30 S subunit (22).
No factor equivalent to IF1 has been observed in the mitochondria from any system nor can an EST for this protein be identified in the human EST data bases. A gene for IF1 is, however, apparent in many chloroplast genomes. This small protein (less than 90 residues) binds to the 30 S subunit around helix 44 in the region that will become the A site (23). By binding to this site, IF1 is postulated to prevent accidental initiation from the A site and to promote the correct positioning of fMet-tRNA in the P site (24,25). In the current report, IF3 mt has been identified and characterized, and the effects of bacterial IF1 on the function of IF2 mt and IF3 mt have been investigated.

Preparation of Ribosomes and Initiation
Factors-Bovine mitochondria and 55 S ribosomes were prepared as described previously (26). Escherichia coli ribosomes were prepared as described (27,28), and tight couples were collected from a sucrose gradient in the presence of 5 mM Mg 2ϩ (29). Bovine IF2 mt , yeast [ 35 S]fMet-tRNA, and E. coli initiation factors were prepared as described (12,28). E. coli IF2 was also prepared from an expression construct providing a mixture of the ␣ and * This work was supported by National Institutes of Health Grant GM32734. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF410851 and AAL04150.
‡ To whom correspondence should be addressed.  (3). The location of the proteolytic cleavage observed in a portion of the factor during expression in E. coli is indicated by the arrow (1). The asterisks indicate residues implicated in the binding of bacterial IF3 to 30 S subunits. C, alignment of the amino acid sequence of human IF3 mt with its homologs from Bos taurus (Bovine), Mus musculus (Mouse), F. rubripes (Fugu), and D. melanogaster (Drosophila). The alignment was done with the CLUSTALW program in Biology Workbench, and the results are displayed in BOXSHADE. The full sequence of the F. rubripes IF3 mt is not shown for convenience. D, alignment of human IF3 mt with the putative IF3 mt from S. pombe.
␤ forms of IF2 (kindly provided by Angela Coursey, University of North Carolina). The genes for E. coli IF3 and IF1 (kindly provided by Drs. Roberto Spurio and Claudio Gualerzi, University of Camerino, Italy, and Dr. Rebecca Alexander, Wake Forest University, respectively) were also amplified by PCR and cloned into pET-21(c). The constructs carrying E. coli IF3 and IF1 were transformed into an E. coli BL21(DE3) strain that also carried the plasmid pArgU218 (kindly provided by Dr. Yamada, Mitsubishi Chemical Corp., Yokohama, Japan). The Histagged forms of the E. coli initiation factors were purified on Ni-NTA resins as described below.
Cyberprobing for Mitochondrial Translational Initiation Factor 3-EST and genomic data base searches for human IF3 mt were performed using BLAST (National Center for Biotechnology Information) and the sequences of various prokaryotic IF3s as virtual probes (30). Sequence analysis was done using the GCG DNA analysis software package (Wisconsin Package, version 10, Genetics Computer Group, Madison WI), Vector NTI (Informax Inc.), and Biology WorkBench 3.2. The results were displayed using BOXSHADE (written by K. Hofmann and M. Baron). Prediction of the cleavage sites for the mitochondrial signal sequence was carried out using PSort and MitoProt II (31,32). Protein secondary and tertiary structures were predicted using Internet-based software, PHDsec and SWISS-Model, respectively (33,34).
Cloning of Human IF3 mt into an Expression Vector-A full-length cDNA clone in vector pT7T3D-Pac carrying the human mitochondrial IF3 mt cDNA was obtained from the American Culture Type Collection (number 526483). The region predicted to be present in the mature form of human IF3 mt (residues 32-278) was amplified by PCR using the full-length cDNA as a template. The portion of the IF3 mt cDNA predicted to correspond to the mature protein was cloned between the NdeI and XhoI sites of pET-21(c) using the forward primer 5Ј-CGCGGATC-CAATTCATATGGCTGCTTTTTCT-3Ј and the reverse primer 5Ј-CGCGGATCCGCTCGAGCTGATGCAGAACAT-3Ј. This vector provides a His tag at the C terminus. The construct carrying the human IF3 mt was transformed into E. coli BL21(DE3) carrying the plasmid pArgU218 (Dr. Yamada, Mitsubishi Chemical Corp., Yokohama, Japan), which provides the gene for the isoacceptor of tRNA Arg recognizing the AGA and AGG codons.
Expression and Purification of IF3 mt -Induction of IF3 mt with 50 M isopropylthiogalactoside was carried out for 5 h at 37°C after the cell density had reached 0.5 at OD 600 . The cells were harvested by low speed centrifugation, and IF3 mt was purified through a Ni-NTA column as described (35). Protein concentrations were determined by the Micro-Bradford method (Bio-Rad). Because of the presence of the 19-kDa form of IF3 mt found in the Ni-NTA column preparations, a second step of purification was carried out using high performance liquid chromatography. In this procedure, the partially purified IF3 mt preparation prepared from 2 liters of cell culture (2 mg/ml, 6 mg) was dialyzed against 100-fold excess of Buffer A (20 mM HEPES-KOH, pH 7.6, 10 mM Mg 2 Cl, 6 mM ␤-mercaptoethanol, 225 mM KCl, and 10% glycerol) for 1.5 h. The dialyzed sample was applied at a flow rate of 0.5 ml/min to a TSKgel SP-5PW column (7.5 ϫ 75 mm, TosoHas Inc., Japan) that had been equilibrated in Buffer A except that the KCl was adjusted to 240 mM. The column was washed until the absorbance at 280 nm returned to base line. The column was then developed with a linear gradient (50 ml) from 0.24 to 0.30 M KCl in Buffer A at a flow rate of 0.5 ml/min. The fractions (0.5 ml) were collected. The fractions containing IF3 mt and its major degradation product were pooled separately and fast-frozen in a dry ice isopropyl alcohol bath and stored at Ϫ70°C. The N-terminal sequences of the expressed forms of IF3 mt were determined using a Perkin Elmer/ABI Procise model 492 protein/peptide sequencer.
Preparation of mRNA Transcripts-The previously described clone carrying the bovine cytochrome oxidase subunit II (CoII) gene (36) was modified to provide a sequence of 30 A residues at the 3Ј end. The transcript prepared from this vector mimics mitochondrial mRNAs produced in vivo, which generally have poly(A) stretches up to 70 residues added following transcription and processing (37). In vitro transcripts were prepared as described previously (36).
Initiation Complex Formation Assays-The activity of IF3 mt in promoting initiation complex formation with E. coli and mitochondrial ribosomes was assayed using conditions basically described previously (8,9). Reaction mixtures (100 l) contained 0.5 OD 260 units of E. coli 70 S or 0.075-0.15 OD 260 units of mitochondrial 55 S tight couples, 12.5 g of poly(A,U,G), or 10 pmol of the CoII mRNA, 3.8 pmol of [ 35 S]fMet-tRNA, and the indicated amounts of various initiation factors. All of the initiation complex formation assays were incubated at 37°C for 15 min and analyzed as described (8).
Dissociation of Mitochondrial 55 S Ribosomes by IF3 mt -The reaction mixtures (100 l) were prepared containing 25 mM Tris-HCl, pH 7.6, 2 mM Mg 2ϩ , 100 mM KCl, 0.5 mM EDTA, 1 mM dithiothreitol, 5% glycerol, 0.2 OD 260 units of 55 S ribosomes, and variable amounts of IF3 mt (0 -1.72 g). The reactions were incubated for 15 min at 37°C. The Mg 2ϩ concentration was then adjusted to 7 mM by the addition of 2.5 l of 0.1 M MgCl 2 , and the samples were analyzed for 28, 39, and 55 S particles on 10 -30% (w/v) linear sucrose gradients prepared in the buffer described above containing 7 mM Mg 2ϩ and analyzed as described previously (38).

RESULTS AND DISCUSSION
Identification of the cDNA for IF3 mt and Analysis of the Coding Region-Although the mammalian mitochondrial ribosome has a low percentage of rRNA and a high protein content compared with bacterial ribosomes, portions of the rRNA where IF3 is thought to bind are present (39). Further, the ribosomal proteins with which IF3 interacts (S7, S11, and S18) have homologs in the 28 S subunit. Hence, it was logical to postulate that mammalian mitochondria contain a homolog of bacterial IF3. Probing the human ESTs with the amino acid sequence of E. coli or most other IF3 species fails to provide any convincing evidence for a mammalian mitochondrial homolog of IF3. However, extensive data base searches with the sequences of IF3 from the Mycoplasma and the IF3 homology domain of Euglena gracilis chloroplast IF3 provide a hit in both the human and mouse EST data bases. The sequence detected  (Fig. 1A). MitoProt II gives this protein a 97% probability to be localized in mitochondria and predicts that the mature protein will be 247 residues in length. The mature form of IF3 mt is predicted to have an N-terminal extension of about 30 residues (Fig. 1, A  and B) that can form a coiled region followed by an ␣-helix. An N-terminal extension of about 150 residues has been noted on E. gracilis chloroplast IF3 (IF3 chl ) (40). IF3 mt also has a Cterminal extension just over 30 residues long. Overall, it is quite hydrophilic and highly charged having nine acidic and five basic residues. The C-terminal extension, like the N-terminal extension, is predicted to have significant helical content. A 63-residue acidic C-terminal extension on E. gracilis IF3 chl has been shown to reduce the activity on the chloroplast factor in initiation complex formation and may serve as a potential regulatory region (41).
Alignment of IF3 mt with prokaryotic and chloroplast IF3 indicates that the mitochondrial factor has diverged considerably from other IF3s (Table I). Overall, it has only 20.8% identity to E. coli IF3, which explains the failure of data base searches with the sequence of E. coli IF3 to locate the corresponding mitochondrial factor. IF3 mt is 25.9% and 23.7% identical to Bacillus stearothermophilus IF3 and to E. gracilis chloroplast IF3, respectively (40). Alignment of the sequence of IF3 mt with the bacterial factors indicates that regions of identity are rather scattered throughout the structure (Fig. 1B). Residues that are responsible for the binding of IF3 to the small subunit are thought to be located primarily in the C-terminal domain. Crystallography experiments place the C-terminal domain of IF3 on the solvent side of the platform on the 30 S subunit (42), whereas cryo electron microscopy and footprinting suggest that it is located on the interface side (39,43). Important regions of IF3 include residues 99 -116, 127-137, 145-155, and 168 (E. coli numbering) as indicated by NMR experiments, mutagenesis, and structural studies (39,42,44). A number of the residues in these regions are conserved or are conservative replacements in the C-terminal domain of human IF3 mt .
One of the major roles of prokaryotic IF3 is the discrimination of the initiation codon (AUG or occasionally GUG or UUG) from other codons. This property can be observed in the isolated C-terminal domain of bacterial IF3 (22) but is strongly affected by conserved residues in the linker region (17,45,46). Interestingly, these highly conserved residues, Tyr 70 , Gly 71 , and Tyr 75 , in prokaryotic IF3s are not conserved in human IF3 mt . In mammalian mitochondria, both AUG and AUA (normally an isoleucine codon) serve as initiation codons. Consequently, the proofreading properties of human IF3 mt could be quite different from those of the bacterial factors.
Analysis of the mouse and bovine EST data bases indicates the presence of mammalian homologs of human IF3 mt that are 65-70% identical to the human factor (Table I and Fig. 1C). In addition, BLAST searches indicate the presence of IF3 mt in Fugu rubripes (puffer fish) and in Drosophila melanogaster (Table I and Fig. 1C). No homolog can be detected in Caenorhabditis elegans. It is quite reasonable to assume that this organism will have a corresponding factor. However, IF3 mt does not appear to be highly conserved throughout the animal kingdom, and it may be difficult to detect using BLAST searches.
The IF3 mt species detected in animals generally have N-and C-terminal extensions that surround a central section that has homology to the bacterial IF3s. The N-terminal extension on puffer fish IF3 mt is considerably longer than that observed on  the mammalian factors (Fig. 1C). D. melanogaster IF3 mt has a very short N-terminal extension. The mammalian and puffer fish IF3 mt s all have C-terminal extensions of around 30 residues compared with the bacterial IF3s. However, D. melanogaster IF3 mt again lacks a significant extension at the C terminus. The linker regions of the mitochondrial factors are charged as observed for the prokaryotic proteins.
Two potential homologs of IF3 mt can be found in Arabidopsis thaliana. One of these genes probably encodes the chloroplast factor, whereas the other encodes the mitochondrial factor. These two forms (NP-179984 and NP-174696) differ considerably in length (312 and 574 residues, respectively). Alignments of these two species with E. gracilis chloroplast IF3 suggests that the shorter form is more likely to encode the chloroplast factor based on the percentage of identity. However, the shorter form also has a higher percentage of identity to human IF3 mt than the longer form, making it difficult to assign these two species clearly to one or the other compartments.
Cyberprobing of the recently completed genome of Schizosaccharomyces pombe allows the tentative identification of IF3 mt in this organism (47). The protein encoded by this gene is 20.9% identical to human IF3 mt and 25% identical to E. coli IF3 (Table I and Fig. 1D). If the predicted import signal is cleaved from S. pombe IF3 mt , no N-terminal extension would be present. IF3 mt from S. pombe does not have any observable C-terminal extension. Searching the genome of Saccharomyces cerevisiae with the sequence of human IF3 mt fails to indicate the presence of a yeast homolog. However, a possible candidate can be detected that is 23.9% identical to the S. pombe factor over the IF3 region. The S. cerevisiae protein is considerably longer than traditional IF3s, and its classification as a mitochondrial IF3 remains to be clarified.
Development of Three-dimensional Models for the N and C Domains of IF3 mt -The coordinates for the crystal structures of N and C domains of B. stearothermophilus IF3 were used to model the N and C domains of human IF3 mt using Swiss-Model. The N-terminal domain of IF3 has a globular ␣/␤ topology consisting of a single ␣-helix packed against a fourstranded ␤-sheet ( Fig. 2A). This domain leads into the connecting linker, which is helical in the crystal structure but is quite flexible in solution (48). As indicated in Fig. 2B, the N-terminal domain of IF3 mt is predicted to fold into a highly similar structure. The linker region is not shown in this model.
The C-terminal domain of bacterial IF3 also consists of an ␣/␤ fold with two helices packed against a four-stranded sheet (Fig. 2C). The C-terminal domain of IF3 mt could not be fully modeled because of low sequence conservation and the unclear alignment of portions of the molecules (Fig. 2, C and D). However, the first ␣-helix and the first two strands of the ␤-sheet can be modeled to resemble the prokaryotic factors quite well. It is likely that the remainder of the C-terminal domain will have a similar overall fold to that observed in prokaryotic IF3 despite the low sequence conservation.
The linker region separating the N and C domains of IF3 is a rigid helix in the crystal structure of B. stearothermophilus IF3 but is more flexible in the NMR structure of E. coli IF3 (48 -52). Structural studies suggest that the linker must be flexible to allow IF3 to interact with distant regions of the small  4. Effects of IF3 mt on the dissociation of mitochondrial ribosomal subunits. A, mitochondrial 55 S ribosomes (0.2 OD 260 ) were incubated as described under "Materials and Methods" at 2 mM Mg 2ϩ in the absence of added IF3 mt , then the Mg 2ϩ concentration was subsequently raised to 7 mM, and the mixture was analyzed for monosomes and subunits by sucrose density gradient centrifugation as described previously (38). B, mitochondrial ribosomes (0.2 OD 260 , ϳ6 pmol) were incubated at 2 mM Mg 2ϩ in the presence of 1.72 g of IF3 mt . The Mg 2ϩ concentration was subsequently raised to 7 mM, and the mixture analyzed for monosomes and subunits by sucrose density gradient centrifugation.
subunit (39,42,43). Secondary structure predictions on the linker region of IF3 mt indicate that it could form a helical conformation, particularly as it exits the N-terminal domain. However, the linker in IF3 mt contains two proline residues near its junction with the C-terminal domain. Proline residues are not observed in the linker regions of prokaryotic IF3s. These residues would be expected to reduce the flexibility of the linker and may help confer a specific orientation between the N-and C-terminal domains in the mitochondrial factor. Both proline residues are conserved in the mammalian factors, whereas the second is also seen in puffer fish IF3 mt .
Purification of Overexpressed IF3 mt -The portion of the cDNA for human IF3 mt corresponding to the region predicted to be present in the mature form of the protein (amino acids 32-278) was cloned into an expression vector providing a His tag. When the mature form of IF3 mt was expressed, two major bands of protein were observed on SDS-PAGE following purification on Ni-NTA resins (Fig. 3A, lane 2). The highest molecular mass form migrated at 29 kDa, the size expected for the mature form of this factor. A second shorter form of IF3 mt migrated at 19 kDa (Fig. 3A). Both of these bands cross-reacted with the antibody prepared against E. gracilis IF3 chl on Westerns (data not shown). These two forms of IF3 mt were purified by high performance liquid chromatography (Fig. 3). N-terminal analysis of the 29-kDa form showed that it begins with the sequence TAP, indicating that it was expressed from the start site predicted for the mature form of the protein following removal of the initiating Met. N-terminal sequencing of the 19-kDa species gave the sequence GNMHRAN, indicating that it arose from the proteolytic degradation of IF3 mt at amino acid 97 (arrows in Figs. 1B and 2B), which is located in the helical segment in the middle of the N-terminal domain of IF3 mt .
Effect of IF3 mt on the Equilibrium between the 55 S Ribosome and Its Subunits-Bacterial IF3 acts as a ribosome dissociation factor. The ability of IF3 mt to affect the equilibrium between the mitochondrial 55 S ribosome and its 28 and 39 S subunits was examined in a two-step assay. In the first step, IF3 mt was incubated with mitochondrial ribosomes at 2 mM Mg 2ϩ . At this concentration of Mg 2ϩ , a significant fraction of the 55 S ribosomes dissociates into subunits giving IF3 mt access to 28 S subunits (38). In the second stage, the Mg 2ϩ concentration was raised to 7 mM, promoting the reassociation of the subunits. A blank representing the amount of label retained on the filter in the absence of IF3 (0.1 pmol) has been subtracted from each value. B, stimulation of fMet-tRNA binding to 70 S ribosomes with the indicated amount of either E. coli IF3 or IF3 mt and in the presence of a saturating amount of E. coli IF2 and E. coli IF1. A blank representing the amount of label retained on the filter in the absence of IF2 (about 0.3 pmol) has been subtracted from each value. C, fMet-tRNA binding to 70 S ribosomes in the presence of either E. coli IF3 or IF3 mt with a saturating amount of IF2 mt . A blank representing the amount of label retained on the filter in the absence of IF3 (about 0.9 pmol) has been subtracted from each value. For unknown reasons, background values in the presence of IF2 mt are higher than those observed in the presence of the other initiation factors. The E. coli IF3 and IF3 mt are shown with (E) and (s), respectively.
Binding of IF3 mt to the 28 S subunits in the first step would be expected to result in an increased amount of ribosomal subunits following the increase in the Mg 2ϩ concentration. The distribution of ribosomal particles was monitored by sucrose density gradient centrifugation. The significant fraction of the mitochondrial ribosomes were present as 55 S particles following the increase in the Mg 2ϩ concentration to 7 mM (Fig. 4A). However, in the presence of IF3 mt , a substantial increase in the presence of 28 and 39 S subunits was observed (Fig. 4B). This observation demonstrates that IF3 mt acts as a subunit antiassociation factor in mammalian mitochondria as it does in bacteria.
Activity of IF3 mt in Initiation Complex Formation and Ribosome Specificity-The ability of IF3 mt to promote initiation complex formation on mitochondrial ribosomes was examined by testing its effect on fMet-tRNA binding to ribosomes in the presence of IF2 mt and poly(A,U,G). As indicated in Fig. 5A, the presence of IF3 mt increased the amount of fMet-tRNA binding observed. This result is expected based on the ability of IF3 mt to increase the availability of 28 S subunits required for the activity of IF2 mt . IF3 mt did not stimulate binding of fMet-tRNA to 28 S subunits, suggesting that its effect in the initiation  7. Effects of E. coli IF1 on fMet-tRNA binding to 55 and 70 S ribosomes. A, initiation complex formation was monitored on mitochondrial 55 S ribosomes. The reaction mixtures were prepared as described under "Materials and Methods" and contained the indicated amount of IF3 mt , a saturating amount of IF2 mt , and either 0.14 g of over-expressed and purified E. coli IF1 (q) or a compensating amount of buffer (Ⅺ). A blank representing the amount of label retained on the filter in the absence of IF3 (0.15 pmol) has been subtracted from each value. B, fMet-tRNA binding to E. coli 70 S ribosomes was examined in the presence of a limiting amount of E. coli IF2 in the presence (q) or absence (Ⅺ) of 0.7 g of E. coli IF1 and in the presence of a saturating amount (50 pmol) of E. coli IF3. A blank representing the amount of label retained on the filter in the absence of IF2 (0.25 pmol) has been subtracted from each value. C, fMet-tRNA binding to E. coli 70 S ribosomes was examined in the presence of a limiting amount of IF2 mt in the presence (q) or absence (Ⅺ) of 0.7 g of E. coli IF1 and in the presence of a saturating amount (50 pmol) E. coli IF3. A blank representing the amount of label retained on the filter in the absence of IF2 mt (0.19 pmol) has been subtracted from each value. D, initiation complex formation on 70 S ribosomes in the presence of a saturating amount of IF2 mt (2.2 pmol) and the indicated amounts of IF3 mt and in the presence (q) and absence (Ⅺ) of 0.7 g of native E. coli IF1. A blank representing the amount of label retained on the filter in the absence of IF3 mt but in the presence of IF2 mt (0.7 pmol) has been subtracted from each value. assay arises primarily from its ability to promote the dissociation of ribosomes.
Previously, it has been shown that E. coli IF3 promotes the dissociation of mitochondrial ribosomes into 28 and 39 S subunits (38,53). This observation suggests that it might be active in promoting initiation complex formation on mitochondrial ribosomes. As indicated in Fig. 5A, E. coli IF3 also promotes initiation complex formation on mitochondrial 55 S ribosomes as expected from its activity as a ribosome dissociation factor. In contrast to the activity on E. coli IF3 on mitochondrial ribosomes, E. coli IF2 is not active on 55 S ribosomes (9).
The activity of IF3 mt on E. coli 70 S ribosomes was tested by examining its ability to promote the binding of fMet-tRNA to these ribosomes in the presence of E. coli IF2 (Fig. 5B). Somewhat surprisingly, IF3 mt showed no activity on the bacterial ribosomes under these conditions. This observation is in contrast to the activity of IF2 mt which is quite active on 70 S ribosomes (8). Two possible explanations can be put forward to explain this observation. First, it is possible that IF3 mt cannot bind to bacterial ribosomes. Second, IF3 mt might bind to bacterial small subunits but not permit the binding of E. coli IF2, which would be required for fMet-tRNA binding. To distinguish between these two possibilities, fMet-tRNA binding assays to 70 S ribosomes were carried out using bovine IF2 mt , which is active on bacterial ribosomes. As indicated in Fig. 5C, under these conditions, IF3 mt is quite active in promoting fMet-tRNA binding to E. coli ribosomes. This observation indicates that IF3 mt can bind bacterial ribosomes but that its presence is incompatible with the activity of E. coli IF2. Examination of the structures of E. coli and mitochondrial IF2 indicates that the mitochondrial factor is significantly shorter than E. coli IF2. Indeed, IF2 mt lacks both domains I and II in the six-domain model of E. coli IF2 (12,54). Domain II has been implicated in the binding of E. coli IF2 to 30 S subunits (55). The absence of this domain in IF2 mt may allow both IF3 mt and IF2 mt to bind to the small subunit at the same time.
Stimulation of Initiation Complex Formation in the Presence of Natural mRNAs by IF3 mt -It has not yet been possible to assemble an initiation complex using a "natural" mRNA in the mammalian mitochondrial system. To further test the effect of IF3 mt on fMet-tRNA binding in initiation, its activity was examined in the presence of an in vitro transcript of the cytochrome oxidase subunit II gene (CoII mRNA). As indicated in Fig. 6, IF3 mt stimulated fMet-tRNA binding with CoII mRNA on 55 S ribosomes. As mentioned earlier, mitochondrial mRNAs have an almost complete lack of 5Ј-and 3Ј-untranslated nucleotides. The translational start codon is generally located within three nucleotides of the 5Ј end of the mRNA (1, 4), a situation very similar to the rare leaderless mRNAs found in several prokaryotic systems. Recent studies performed with leaderless mRNAs in bacteria have suggested that IF3 antagonizes translation initiation on these mRNAs, at least with ribosomes containing ribosomal protein S1 and in the presence of a competing mRNA carrying a Shine/Dalgarno sequence (56 -59). Because mitochondrial ribosomes do not have a protein equivalent to S1 (60) and have no mRNAs with Shine/ Dalgarno sequences, different constraints may be operating in the mitochondrial system permitting initiation on mRNAs with essentially no 5Ј leader.
Requirement for Other Initiation Factor(s) in Mammalian Mitochondrial Initiation-Prokaryotic translational initiation requires IF1 in addition to IF2 and IF3. With the current work, two of these factors, IF2 mt and IF3 mt , have been identified in mammalian mitochondria. Extensive searches of the human and mouse EST data bases and of the genomes of S. cerevisiae, D. melanogaster, and C. elegans have failed to provide evidence for the presence of a factor equivalent to IF1 in mitochondria. This small protein presents a considerable challenge to identify in such searches because of its small size (about 70 amino acids) and low degree of sequence conservation. Biochemical tests have also failed to date to identify a factor equivalent to IF1 in mammalian mitochondrial extracts. However, such a protein would only be present in trace amounts making its detection a challenge.
To help assess the possible need for a factor equivalent to IF1 in the mitochondrial system, the effect of E. coli IF1 on initiation complex formation was examined using both E. coli and mitochondrial 55 S ribosomes as well as with E. coli and mitochondrial IF2 and IF3. As indicated in Fig. 7A, the presence of E. coli IF1 has essentially no effect on initiation complex formation on 55 S ribosomes in the presence of IF3 mt and IF2 mt . The observation suggests that the mitochondrial system may not require a factor directly equivalent to IF1.
Further insight into the question of an interaction between E. coli IF1 and the mitochondrial initiation factors was obtained by examining the effects of this factor on initiation complex formation on 70 S ribosomes. As a control, the previously reported stimulation of E. coli IF2 by IF1 was examined, and a substantial stimulation of fMet-tRNA binding was observed (Fig. 7B). In contrast to the stimulation of E. coli IF2 by IF1, no stimulation of the activity of IF2 mt on 70 S ribosomes was observed under identical conditions (Fig. 7C). IF2 mt alone actually stimulates initiation complex formation on 70 S ribosomes as effectively as E. coli IF2 in the presence of IF1.
The activity of E. coli IF3, like that of IF2, is stimulated by IF1 (data not shown). However, E. coli IF1 again fails to stimulate the activity of IF3 mt on 70 S ribosomes in the presence of saturating levels of IF2 mt (Fig. 7D). Taken together, these results suggest that IF2 mt and IF3 mt function efficiently in initiation complex formation in the absence of IF1. These result suggest that the conformational change caused by the binding of IF1 to 30 S subunits (23) can also be generated by the binding of these mitochondrial initiation factors to 30 S subunits.